Milan van Lange Emotional Imprints of War

**Digital Humanities Research** | Volume 6

#### **Editorial**

Digital Humanities is an evolving, cross cutting field within the humanities employing computer based methods. Research in this field, therefore, is an interdisciplinary endeavor that often involves researchers from the humanities as well as from computer science. This collaboration influences the methods applied as well as the theories underlying and informing research within those different fields. These implications need to be addressed according to the traditions of different humanities' disciplines. Therefore, the edition addresses all humanities disciplines in which digital methods are employed. **Digital Humanities Research** furthers publications from all those disciplines addressing the methodological and theoretical implications of the application of digital research in the humanities. The series is edited by Silke Schwandt, Anne Baillot, Andreas Fickers, Tobias Hodel and Peter Stadler.

**Milan van Lange**, born in 1992, works as a researcher at the NIOD Institute for War, Holocaust and Genocide Studies in Amsterdam. He studied history at the Radboud Universiteit in Nijmegen and did his doctorate at the NIOD and Utrecht University (2016-2021). His research interests are at the intersection of contemporary and social history, war studies, digital humanities, digitisation of historical archives, and innovation in research methodology, including the integration of computational methods, text mining, data analysis, and statistics in historical research.

Milan van Lange **Emotional Imprints of War**

A Computer-assisted Analysis of Emotions in Dutch Parliamentary Debates, 1945-1989

Funded by the NIOD Institute for War, Holocaust and Genocide Studies.

This work is based on the author's PhD thesis defended at Utrecht University in 2021.

#### **Bibliographic information published by the Deutsche Nationalbibliothek**

The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http:// dnb.d-nb.de

This work is licensed under the Creative Commons Attribution 4.0 (BY) license, which means that the text may be remixed, transformed and built upon and be copied and redistributed in any medium or format even commercially, provided credit is given to the author. For details go to https://creativecommons.org/licenses/by/4.0/

Creative Commons license terms for re-use do not apply to any content (such as graphs, figures, photos, excerpts, etc.) not original to the Open Access publication and further permission may be required from the rights holder. The obligation to research and clear permission lies solely with the party re-using the material.

**First published in 2023 by Bielefeld University Press An Imprint of transcript Verlag © Milan van Lange**

Cover layout: Maria Arndt, Bielefeld Printed by Majuskel Medienproduktion GmbH, Wetzlar Print-ISBN 978-3-8376-6485-0 PDF-ISBN 978-3-8394-6485-4 https://doi.org/10.14361/9783839464854 ISSN of series: 2747-5476 eISSN of series: 2749-1986

Printed on permanent acid-free text paper.

## **Contents**




## **1. Introduction | On War, Emotions, and Computers in History**

#### **1.1 Hurting Feelings**

'[B]ecause of all the sensitivities, I had to choose my words very carefully', stated Gerdi Verbeet, chair of the Dutch national commemoration committee *Nationaal Comité 4 en 5 mei* in 2016. 'You don't want to say something that hurts people's feelings', she added.<sup>1</sup> More than seven decades after the end of the German occupation of the Netherlands, Verbeet raised this issue in an interview about her role in the national World War II commemoration. It is not surprising that people associate a major devastating historical event with emotions experienced in the past, or that its history or consequences continue to elicit emotional responses in the present. 'The War', as people still refer to it, is more than a moral cornerstone or a prominent frame of reference. The German occupation during World War II left a profound emotional imprint on Dutch society. As Verbeet demonstrated, it has also become something about which it seems unimaginable, or at least socially unacceptable, to think, speak, or write without either a (subtle) display of emotion, or at least an explicit mention of the supposed emotionality of others.<sup>2</sup> That, however, is not all.

Emotions are omnipresent in academic historical studies dealing with the aftermath of World War II in the Netherlands. Scholars use emotions in subordinate

<sup>1</sup> Gerdi Verbeet stated this in an interview in the leading national newspaper, NRC Handelsblad, when she was interviewed around the time of national Remembrance Day in 2016. See Danielle Pinedo, '"De oorlog houdt ons meer bezig dan ooit"', *NRC Handelsblad*, 30 April 2016, https://www.nrc.nl/nieuws/2016/04/30/de-oorlog-houdt-ons-meer-bezig-dan-ooit-16 13560-a1234765.

<sup>2</sup> This seems to be an implicit social code: Emotions should not be forgotten, and sensitivities are to be respected. This phenomenon became very clear during the Jenninger affair in Germany in 1988. Philip Jenninger, president of the West German Bundestag, caused a scandal when he, in the eyes of his critics, failed to express sentiments and emotions of remorse about the atrocities of the war in a public remembrance ceremony of the Kristallnacht. See Ernestine Schlant, *The Language of Silence: West German Literature and the Holocaust*, 1st ed. (New York: Routledge, 1999), 185–86.

clauses, as unsubstantiated remarks, or as explanatory factors in their historical claims.<sup>3</sup> Emotions gained a role as one of the most popular side notes in the historiography of the post-war dealing with the consequences of World War II in the Netherlands. Not only does the connection between war and emotion seem self-evident, the converse is also true. In thinking, speaking, and writing about societal developments in emotionality, references to the war and its lasting consequences are frequent.WorldWar II is a prominent theme in academic literature dealing with emotions or the emergence of a so-called 'emotion culture'in contemporary history.<sup>4</sup> However, studies that explicitly scrutinise their interrelatedness – between dealing with the consequences of the war and emotions – are an exception. This is remarkable, given the vast amount of literature addressing different aspects of the way in which the consequences of World War II have been dealt with.<sup>5</sup> As a result, when it

<sup>3</sup> For some examples of studies wherein emotions play this role, see Martin Bossenbroek, *De Meelstreep* (Amsterdam: Bakker, 2001); Jolande Withuis, *Erkenning. Van oorlogstrauma naar klaagcultuur* (Amsterdam: De Bezige Bij, 2002); Hinke Piersma, *Bevochten recht: politieke besluitvorming rond de wetten voor oorlogsslachtoffers* (Amsterdam: Boom, 2010); Harry Oosterhuis, 'Mental Health, Citizenship, and the Memory of World War II in the Netherlands (1945–85)', *History of Psychiatry* 25, no. 1 (1 March 2014): 20–34.

<sup>4</sup> See for example: Henri Beunders, *Publieke tranen. De drijfveren van de emotiecultuur* (Amsterdam: Contact, 2002), 125,205; Remieg Aerts, 'Emotie in de politiek. Over politieke stijlen in Nederland sinds 1848', in *Jaarboek Parlementaire Geschiedenis 2003: Emotie in de politiek*, ed. Carla van Baalen et al. (Nijmegen, Den Haag: Centrum voor Parlementaire Geschiedenis, SDU Uitgevers, 2003), 21.

<sup>5</sup> Withuis, *Erkenning*; Jolande Withuis and Annet Mooij, 'From Totalitarianism to Trauma − A Paradigm Change in the Netherlands', in *The Politics of War Trauma: The Aftermath of World War II in Eleven European Countries*, ed. Jolande Withuis and Annet Mooij (Amsterdam: Amsterdam University Press, 2010), 193–216; In addition, the following publications are, in this context, especially noteworthy and relevant: Peter Romijn, *Snel, streng en rechtvaardig. Politiek beleid inzake de bestraffing en reclassering van 'foute' Nederlanders, 1945–1955* (Houten: De Haan, 1989); Bossenbroek, *De Meelstreep*; Harald Fühner, *Nachspiel: die niederländische Politik und die Verfolgung von Kollaborateuren und NS-Verbrechern, 1945–1989*, Niederlande-Studien 35 (Münster: Waxmann, 2005); Regula Ludi, 'Who Is a Nazi Victim? Constructing Victimhood through Post-War Reparations in France, Germany, Switzerland', *UCLA: Center for European and Eurasian Studies*, 18 August 2005; Hinke Piersma, *De Drie van Breda: Duitse oorlogsmisdadigers in Nederlandse gevangenschap, 1945–1989*, 1st ed. (Amsterdam: Balans, 2005); Ismee Tames, *Besmette jeugd: de kinderen van NSB'ers na de oorlog* (Amsterdam: Balans, 2009); Piersma, *Bevochten recht*; Elly Touwen-Bouwsma, *Op zoek naar grenzen. Toepassing en uitvoering van de wetten voor oorlogsslachtoffers* (Amsterdam: Boom, 2010); Chris van der Heijden, *Dat nooit meer: de nasleep van de Tweede Wereldoorlog in Nederland*, 1st ed. (Amsterdam: Contact, 2011); Helen Grevers, *Van landverraders tot goede vaderlanders: de opsluiting van collaborateurs inNederland en België, 1944–1950*, Erfenissen van Collaboratie (Amsterdam: Uitgeverij Balans, 2013); Ismee Tames, *Doorn in het vlees:fouteNederlanders in de jaren vijftig en zestig* (Amsterdam: Balans, 2013); Ismee Tames, 'Ashamed About the Past: The Case of Nazi Collaborators and Their Families in Postwar Dutch Society', in *Reverberations of Nazi Violence in Germany and Beyond: Disturbing Pasts*, ed. Stepha-

comes to gaining a deeper understanding of the role of emotions in dealing with the consequences of the war, there is much still unknown.

The manner in which the consequences of the war in the Netherlands were dealt with in the post-war period is not a neglected issue. This investigation therefore builds on an elaborate existing historiography with new perspectives and methods. A central aim is to approach an already relatively well-investigated historical case from a new and different angle. My re-evaluation of this historiography employs two fundamental strategies: the placing of emotions as a central object of research, and the application of a computer-assisted methodology. Yet this approach goes beyond merely assessing whether emotions were present. The study also reflects on issues raised in the field known as the 'history of emotions', which asks whether the study of emotions can be of additional value in historical research, and if so, how. The central question that this book addresses, is therefore not only *whether,* but also *how*emotions played a role in the ways Dutch politics dealt with the consequences of the war in the Netherlands between 1945 and 1989. This leads in the first instance to issues that are of a more fundamental epistemological nature. Questions discussed in Chapter 2 are, for example: How to define emotions? How knowable are emotions at all – not only for the contemporary observer, but also for the historical researcher? How can we identify (changes in) the expression of emotions? And how to analyse emotions historically? These questions also relate to current claims on historical developments in emotionality and are addressed later in this chapter and in Chapter 5. First, however, this chapter proceeds with a reflection on common historical research practices and methodologies, and asks how we might approach historical emotions from a new perspective and why this is needed.

#### A Computer-assisted Methodology

From a methodological perspective, this investigation builds upon a computerassisted research methodology.<sup>6</sup> Such a prominent role for the computer distinguishes this study from most 'traditional' historical investigations. That the application of computational research methods in historical research is not (yet) very common does not mean that it is a recent development or a very novel one per

nie Bird et al. (London; New York: Bloomsbury Publishing, 2016), 47–64; Ismee Tames and Peter Romijn, 'Transnational Identities of Dutch Nazi-Collaborators and Their Struggle for Integration into the National Community', *Journal of Modern European History* 16, no. 2 (2018): 247–68.

<sup>6</sup> The term 'computer-assisted' is used here to describe a methodology that consists of computational techniques, but does not discard close reading of the sources. I consider the overall research methodology as computer-assisted, as there is still an important role for traditional qualitative methods within the field of historical research.

se.<sup>7</sup> The use of computers in historical research has been stimulated by a relatively recent increase in the availability of large-scale digitised historical sources, open source software, language technology, and other resources.<sup>8</sup> Such developments have also fuelled the rapid rise of the broader field now known as the 'digital humanities'.<sup>9</sup> Inspired by experimentation, exploration, tools, and reflection on the potential of using a computer in the digital humanities, this investigation aims for a practical, analytical approach that also critically evaluates their methodological and epistemological implications.<sup>10</sup> This is achieved not only by reflecting on, but also by the application of digitised historical sources, computational techniques, and resources in the practice of historical research.

The computer-assisted methodology of this study connects to 'the historian's macroscope' – a term coined by scholars Shawn Graham, Ian Milligan, and Scott Weingart.<sup>11</sup> 'The macroscope' is used here as a metaphor that brings together common historiographic practices and the application of a computational method known as 'emotion mining' in a systematic, quantitative 'distant-reading' analysis

8 See for example: Ludovic Rheault et al., 'Measuring Emotion in Parliamentary Debates with Automated Textual Analysis', *PLoS ONE* 11, no. 12 (22 December 2016); Luke Blaxill and Kaspar Beelen, 'A Feminized Language of Democracy? The Representation of Women at Westminster since 1945', *Twentieth Century British History* 27, no. 3 (1 September 2016): 412–49; Melvin Wevers, 'Consuming America. A Data-Driven Analysis of the United States as a Reference Culture in Dutch Public Discourse on Consumer Goods, 1890–1990' (PhD thesis, Utrecht, Utrecht University, 2017); Luke Blaxill, *The War of Words: The Language of British Elections, 1880–1914* (Suffolk and Rochester: Boydell & Brewer, 2020); Melvin Wevers and Marijn Koolen, 'Digital Begriffsgeschichte: Tracing Semantic Change Using Word Embeddings', *Historical Methods: A Journal of Quantitative and Interdisciplinary History* 53, no. 4 (13 May 2020): 226–43; For an overview of the approaches and methods used in digital historical research, see C. Annemieke Romein et al., 'State of the Field: Digital History', *History* 105, no. 365 (14 May 2020): 291–312.

9 Inger Leemans, 'A Smell of Higher Honey: E-Humanities Perspectives', *BMGN − Low Countries Historical Review* 128, no. 4 (2013): 150–51; Johanna Drucker et al., *Introduction to Digital Humanities: Concepts, Methods, and Tutorials for Students and Instructors* (Los Angeles: UCLA, 2014); Nan Z. Da, 'The Digital Humanities Debacle', *The Chronicle of Higher Education*, 27 March 2019; Silke Schwandt, ed., *Digital Methods in the Humanities − Challenges, Ideas, Perspectives*, 1st ed., vol. 1, Digital Humanities Research (Bielefeld: Bielefeld University Press, Transcript, 2020).

<sup>7</sup> For a brief history of using computers in historical research (in the Netherlands), see Gerben Zaagsma, 'On Digital History', *BMGN − Low Countries Historical Review* 128, no. 4 (16 December 2013): 3–29; Adam Crymble, *Technology and the Historian: Transformations in the Digital Age*, 1st ed., Topics in the Digital Humanities (Urbana: University of Illinois Press, 2021), 17–45.

<sup>10</sup> Zaagsma, 'On Digital History', 5.

<sup>11</sup> Shawn Graham, Ian Milligan, and Scott B. Weingart, *Exploring Big Historical Data: The Historian's Macroscope*, 1st ed. (London: Imperial College Press, 2016); The term 'macroscope' was borrowed from Katy Börner. See Crymble, *Technology and the Historian*, 32.

of digitised historical sources.<sup>12</sup> This includes combining a 'baseline' – a more distant, long-term perspective – with zooming in on particular events or incidents in their specific context. The fresh perspectives enabled by the use of computational methods in historical research, as well as the particularities and limitations of those methods, is addressed in part 1.4 of this chapter and in Chapter 4. The results of combining traditional and computational approaches to historical sources are discussed using multiple case studies in Chapters 5, 6, 7, 8, and 9. For now, this chapter proceeds with an introduction to the historiographical framework of this study.

#### **1.2 Historiography (of a Side Note)**

Recent historiographies of the consequences of World War II have certainly considered the role of emotions. Whether mentioned explicitly or not, emotions are present in studies on the effects of war and wartime experiences on different communities, both victims and perpetrators (or their descendants). Various studies, for example in the 2016 edited volume by Stephanie Bird et al., show how emotions played – and continue to play – a role in dealing with the lasting legacies of the war, as well as the ways in which such legacies continue to be disrupted and instrumentalised because of the lasting presence of emotions.<sup>13</sup> Scholars who explicitly address emotions often focus on the role of relatively specific emotions (e.g., guilt or shame) and their interaction with the responses to, and understanding of, questions related to the aftermath or legacies of the war (e.g., culpability and taking responsibility for war-related crimes, violence, and wrongdoings).<sup>14</sup> Such studies build on the assumption that a strong connection exists between (the role of) emotions and the wartime past, and offer insights in understanding how different individuals or groups respond emotionally to particular events or situations. Their findings also

<sup>12</sup> Frédéric Clavert, 'Lecture des sources historiennes à l'ère numérique', *L'histoire contemporaine à l'ère numérique* (blog), accessed 27 May 2021, https://histnum.hypotheses.org/1061; cited in Zaagsma, 'On Digital History', 24.

<sup>13</sup> Withuis and Mooij, 'From Totalitarianism to Trauma'; Tames,*Doorn in het vlees*; Mary Fulbrook, 'Troubling Issues: Guilt and Shame among Persecutors and Persecuted', in *Reverberations of Nazi Violence in Germany and Beyond: Disturbing Pasts*, ed. Stephanie Bird et al. (London; New York: Bloomsbury Publishing, 2016), 15–32; Tames, 'Ashamed About the Past: The Case of Nazi Collaborators and Their Families in Postwar Dutch Society'; Ulrike Weckel, 'Shamed by Nazi Crimes: The First Step towards Germans' Re-Education or a Catalyst for Their Wish to Forget?', in *Reverberations of Nazi Violence in Germany and Beyond: Disturbing Pasts*, ed. Stephanie Bird et al. (London; New York: Bloomsbury Publishing, 2016), 33–46.

<sup>14</sup> Stephanie Bird and Mary Fulbrook, 'Introduction: Disturbing Pasts', in *Reverberations of Nazi Violence in Germany and Beyond: Disturbing Pasts*, ed. Stephanie Bird et al. (London; New York: Bloomsbury Publishing, 2016), 8.

demonstrate how emotions matter alongside more conventional factors, such as the motives or memories historians use to assess historical developments regarding war legacies and the various ways of dealing with the consequences of war.

#### The Emotional Turn

The more explicit attention now being paid to emotions in the historiography of World War II and its consequences seems to be a relatively recent development. It coincides with a change generally referred to as the 'emotional turn': an increasing academic interest in emotions in a wide range of fields, from political science and anthropology to sociology, international relations, and history.<sup>15</sup> In these various fields,it seems to be more or less a trend to stress the obscuring of, or negligence shown towards, emotions in the past. For example, in political science, emotions are often introduced as an understudied research object.<sup>16</sup> Also in historical research, authors emphasise that emotions should not be forgotten or neglected, in addition to more traditional factors.<sup>17</sup> Historians of emotions seem to imply that emotions have long been treated by mainstream historiography as a given, without approaching them with critical reflection and questioning. Historian (of emotions) Barbara

<sup>15</sup> Sara Ahmed, *The Cultural Politics of Emotion* (New York: Routledge, 2004); David Redlawsk, ed., *Feeling Politics: Emotion in Political Information Processing* (London and New York: Palgrave Macmillan US, 2006); Jacob Boersema et al., 'Emoties in de Sociologie: De consequenties voor het vak', *Sociologie* 5 (2009): 141–51; Agneta Fischer, Job van der Schalk, and Skyler Hawk, 'Het ontstaan van collectieve emoties via emotionele besmetting', *Sociologie*, no. 2 (2009): 165–79; Paul Hoggett and Simon Thompson, eds., *Politics and the Emotions: The Affective Turn in Contemporary Political Studies* (New York: Continuum, 2012); Alberto Acerbi et al., 'The Expression of Emotions in 20th Century Books', *PLOS ONE* 8, no. 3 (20 March 2013); Andrew A. G. Ross, *Mixed Emotions: Beyond Fear and Hatred in International Conflict* (Chicago: University of Chicago Press, 2013); Thomas Gregory and Linda Åhäll, 'Introduction: Mapping Emotions, Politics and War', in *Emotions,Politics andWar*, ed. Thomas Gregory and Linda Åhäll(New York: Routledge, 2015), 1–14; Todd H. Hall and Andrew A. G. Ross, 'Affective Politics after 9/11', *International Organization* 69, no. 4 (2015): 847–79; Bert N. Bakker, Gijs Schumacher, and Matthijs Rooduijn, 'Hot Politics? Affective Responses to Political Rhetoric', *American Political Science Review*, 9 September 2020, 150–64.

<sup>16</sup> Jeffrey Goodwin, James M. Jasper, and Francesca Polletta, 'Introduction: Why Emotions Matter', in *Passionate Politics: Emotions and Social Movements* (Chicago: University of Chicago Press, 2001), 1–3; David Redlawsk, 'Feeling Politics: New Research into Emotion and Politics', in *Feeling Politics: Emotion in Political Information Processing*, ed. David Redlawsk (London and New York: Palgrave Macmillan US, 2006), 1–2.

<sup>17</sup> Barbara H Rosenwein, *Emotional Communities in the Early Middle Ages* (Ithaca; London: Cornell University Press, 2007), 1; Susan Matt and Peter Stearns, 'Introduction', in *Doing Emotions History*, ed. Susan Matt and Peter Stearns, History of Emotions (Urbana: University of Illinois Press, 2014), 1–5.

Rosenwein, for example, writes that historians should take emotions as seriously as they do other 'invisible' concepts, such as gender.<sup>18</sup>

One possible explanation of the increased attention now being paid to emotions is the advancement of the field of neuropsychology, which has led to a better understanding of the human brain. Traditional distinctions between cognition and emotions are being challenged, thus making space for perspectives in which cognition and emotion are more intertwined.<sup>19</sup> The more general 'emotional turn' has also stimulated the rise of new perspectives and approaches to the study of emotions in history.<sup>20</sup> Increasingly, emotions are seen as part of larger associative networks that also include links to more cognitive components, such as memories, motives, thoughts, images, ideas, and interpretations.<sup>21</sup> Although these are all more or less intangible mental constructs that are the result of internal processes within the brain, most of them have been commonplace considerations in historical research for decades.

<sup>18</sup> Rosenwein, *Emotional Communities*, 1.

<sup>19</sup> Antonio R. Damasio, *De vergissing van Descartes: Gevoel, verstand en het menselijk brein*, 2nd ed. (Amsterdam: Wereldbibliotheek, 1998); Redlawsk, 'Feeling Politics: New Research into Emotion and Politics', 3; Spyros Kosmidis et al., 'Party Competition and Emotive Rhetoric', *Comparative Political Studies* 52, no. 6 (May 2019): 3.

<sup>20</sup> Rosenwein, *Emotional Communities*; Jessica C. E. Gienow-Hecht, ed., *Emotions in American History: An International Assessment*, vol. 3, European Studies in American History (New York: Berghahn Books, 2010); William M. Reddy, *The Navigation of Feeling: A Framework for the History of Emotions* (Cambridge: Cambridge University Press, 2010); Barbara H Rosenwein, 'Problems and Methods in the History of Emotions', *Passions in Context I* 1 (2010): 1–33; Ute Frevert, *Emotions in History: Lost and Found* (Budapest: Central European University Press, 2011); Ute Frevert et al., *Emotional Lexicons: Continuity and Change in the Vocabulary of Feeling 1700–2000*, Emotions in History (Oxford, New York: Oxford University Press, 2014); Rob Boddice, 'The Affective Turn: Historicising the Emotions', in *Psychology and History: Interdisciplinary Explorations*, ed. Cristian Tileaga and Jovan Byford (New York: Cambridge University Press, 2014), 147–65; Peter Stearns and Susan Matt, eds., *Doing Emotions History* (Urbana: University of Illinois Press, 2014); Jan Plamper, *The History of Emotions: An Introduction*, Emotions in History (Oxford: Oxford University Press, 2015); Ute Frevert, 'The History of Emotions', in *Handbook of Emotions*, ed. Lisa Feldman Barrett, Michael Lewis, and Jeannette M. Haviland-Jones, 4th ed. (New York: Guilford Publications, 2016), 49–64; Inger Leemans et al., 'Mining Embodied Emotions: A Comparative Analysis of Sentiment and Emotion in Dutch Texts, 1600–1800.', *Digital Humanities Quarterly* 11, no. 4 (8 January 2018); Ute Frevert, 'Emotional Politics' (WRR Lezing 2019, Den Haag, 24 January 2019).

<sup>21</sup> Fischer, Schalk, and Hawk, 'Het ontstaan van collectieve emoties via emotionele besmetting', 165; Nigel Holt et al., *Psychology: The Science of Mind and Behaviour*, 3rd ed. (London: McGraw-Hill Education Ltd, 2015), 476.

#### History of Emotions

Part of the emotional turn is the rise of a relatively young field known as 'history of emotions'.Emotions as a historical phenomenon are studiedin varied ways and with different approaches. Much of the work aims to analyse the history *of* emotions by tracking developments and changes in certain emotions or the ways in which they are experienced or expressed over time. Such studies often span several centuries. From such studies, we have learnt that emotions are dynamic and inseparably tied to cultural and temporal contexts.<sup>22</sup> Emotions are neither exclusivelyindividual nor exclusively social; neither just culturally, nor naturally determined. Emotions and the ways in which emotions become manifest are highly dependent on the situation, circumstances, and era.This is equally the case in the ways in which they are appraised, perceived, experienced, and expressed.<sup>23</sup> Within this not only rather specific, but also still growing sub-field, a persistent desire seems to exist to make the history of emotions perspective more important,interesting, and relevant to a broader field of research. Rosenwein formulated an 'ultimate goal': '(…) the study of emotions should not (in the end) form a separate strand of history but rather inform every historical inquiry.'<sup>24</sup> Building on Rosenwein, I agree that the history of emotions indeed has the potential to function as an auxiliary field in historical scholarship. Nevertheless, approaching emotions' history only from a history *of* emotions perspective unnecessarily limits its possible impact on mainstream historiography.

The potential of taking into account the role of emotions in history goes beyond the mere tracking of historical developments of emotions as such. Emotions have more to offer history when historians explicitly address the interaction of emotions with historical events, developments, and behaviour.<sup>25</sup> Or, to speak with Ute Frevert, historian and executive director of the Center for the History of Emotions at the

<sup>22</sup> Rosenwein, *Emotional Communities*; Reddy, *The Navigation of Feeling*; Rob Boddice, *Pain and Emotion in Modern History*, Palgrave Studies in the History of Emotions (Houndmills, Basingstoke, Hampshire: Palgrave Macmillan, 2014); Frevert et al., *Emotional Lexicons*; Frevert, *Emotions in History*; Matt and Stearns, 'Introduction'; Plamper, *The History of Emotions*; Janneke van der Zwaan et al., 'HEEM, a Complex Model for Mining Emotions in Historical Text', in *2015 IEEE 11th International Conference on E-Science* (Munich, 2015), 22–30; Susan Broomhall, *Early Modern Emotions: An Introduction* (London and New York: Routledge, 2017); Leemans et al., 'Mining Embodied Emotions'.

<sup>23</sup> Rosenwein, 'Problems and Methods in the History of Emotions', 9; Peter N. Stearns, 'Emotion and Change: Where History Comes In', in *Emotions in International Politics*, ed. Yohan Ariffin, Jean-Marc Coicaud, and Vesselin Popovski (Cambridge: Cambridge University Press, 2016), 49.

<sup>24</sup> Rosenwein, 'Problems and Methods in the History of Emotions', 24.

<sup>25</sup> Redlawsk, 'Feeling Politics: New Research into Emotion and Politics', 2; Pierre Hassner, 'The Dialectic of Rage: How Anger, Fear, Pride and Other Passions Combine, Interact and Fight Each Other in the Post-Cold War World', in *Emotions in International Politics: Beyond Mainstream Inter-*

Max Planck Institute for Human Development: Emotions not only *have* a history, but also *make* history.<sup>26</sup> Inquiring into the mutual interactions between emotions and broader historical developments has the potential to stimulate a more questioning approach to that whichis taken for granted and to commonplace views. In this study, the historical scrutiny of emotions is therefore used to address old questions from new perspectives and to gain a better understanding of (their role in) the course of historical change. This applies, I argue, in particular when dealing with historical events with an evident or generally taken-for-granted emotional charge, as is very much the case with World War II.<sup>27</sup> In the following section, I discuss how emotions play their role as popular side notes in the historiography of the long-term consequences of the German occupation of the Netherlands during World War II.

#### A Cycle of Speaking, Silence, and Renewed Speaking

In Dutch historiography, a rough periodisation exists regarding the ways in which Dutch politics and society dealt with the lasting and late-onset consequences of World War II. The national government's efforts to alleviate the suffering of survivors of the war, such as tailored legislation for war victims, or pension schemes for members of the former anti-Nazi resistance, are prominent in this periodisation. The political discussions and welfare policies it encompassed have become emblematic of a perceived 'emotional history' of the post-war Netherlands. Particularly the legislative history of war-related social benefit schemes seems to have offered a useful diachronic framework of specific dates.<sup>28</sup> Dutch Sociologist Jolande Withuis identified a cyclic post-war periodisation of 'speaking, silence, and renewed speaking' about the consequences of World War II in the Netherlands. The timing of the developments in this cycle roughly mirrors moments of legislative development and change.<sup>29</sup> The moments in which these war-related legislative measures were

*national Relations*, ed. Yohan Ariffin, Jean-Marc Coicaud, and Vesselin Popovski (Cambridge: Cambridge University Press, 2016), 341.

<sup>26</sup> Frevert, *Emotions in History*, 219.

<sup>27</sup> Emotions are, however, not only relevant for the known emotional episodes in history. As social scientists Imrat Verhoeven and Jan Willem Duyvendak emphasised, (appeals to) emotions can even be found in the more mundane processes, such as the processes of policy formation and implementation. Illustrative is their statement that the study of emotions '(…) may (…) prove productive for better understanding (…) problems of failing policy implementation and issues of lacking legitimacy. See Imrat Verhoeven and Jan Willem Duyvendak, 'Enter Emotions: Appealing to Anxiety and Anger in a Process of Municipal Amalgamation', *Critical Policy Studies* 10, no. 4 (October 2016): 482.

<sup>28</sup> The legislative history of benefit schemes for former members of the anti-Nazi resistance and war victims are discussed in the four chapters dealing with case studies 1 and 2.

<sup>29</sup> See e.g., Withuis, *Erkenning*, 16–19, 118–21; Jolande Withuis, 'From Victims Divided To Victims United: The Politics of War Trauma in the Netherlands, 1945–1980', in *A Quest for Alter-*

established, are also connected to broader historical developments and changes in public ways of dealing with the legacies of the war. Withuis described how the periodisation she observed was mirrored in the publication of war-related articles in *Maandblad Geestelijke volksgezondheid,* a monthly magazine for mental health professionals.<sup>30</sup> In her book *Erkenning* (2002), Withuis focuses primarily on (professional) engagement with mental health problems related to or caused by World War II experiences. Obviously, dealing with trauma as a lasting or late-onset consequence of the war is not the same as addressing emotions. Withuis, however, emphasised how the scope of this periodisation reached further than the realm of professional mental health care. More importantly, she points to changes in the 'mental climate' and to the acceptance of discussing and expressing emotions in public ('the emancipation of emotions') as one of the explanatory factors for the observed historical developments in the cycle.This is where emotions enter the equation and play a role in the aforementioned cycle of 'speaking, silence, and renewed speaking'.<sup>31</sup>

The first phase in the cycle encompassed the immediate post-war years (1945–1948). This was a period in which elaborate attention was paid to (personal) war experiences, their consequences, and also to related emotions and expressions thereof in Dutch society.<sup>32</sup> Withuis goes on to identify the following period as a contrasting phase of relative silence.<sup>33</sup> Although the exact dates of the various phases vary between authors, the idea that a 'silent' phase started around 1948 and included most of the 1950s is widespread. In the relevant literature, 'silence' is not always clearly defined or demarcated.The term does not seem, for example, to refer to what American historian Jay Winter termed 'liturgical silences', which encompasses the kind of silence that is an essential part of mourning practices.<sup>34</sup> Instead, the various authors seem to use 'silence' to refer to an observation that there was little or no space for talking publicly about the war and its lasting consequences in the (contemporary) present.<sup>35</sup> This idea of a silent phase regarding the war past resembles

*native Sociology*, ed. Kenji Kōsaka and Masahiro Ogino (Melbourne: Trans Pacific Press, 2008), 124–40; Withuis and Mooij, 'From Totalitarianism to Trauma', 198–206; Oosterhuis, 'Mental Health, Citizenship, and the Memory of World War II', 20, 34.

<sup>30</sup> Withuis, *Erkenning*, 17–18.

<sup>31</sup> Withuis, 118–19.

<sup>32</sup> Withuis, 118–21; Withuis, 'From Victims Divided To Victims United', 124–29; Withuis and Mooij, 'From Totalitarianism to Trauma', 199–202.

<sup>33</sup> Withuis, 'From Victims Divided To Victims United', 129–30, 138; Withuis and Mooij, 'From Totalitarianism to Trauma', 203–4.

<sup>34</sup> JayWinter, 'Thinking about Silence', in *Shadows ofWar: A SocialHistory of Silence in the Twentieth Century*, ed. Efrat Ben-Ze'ev, Jay Winter, and Ruth Ginio (Cambridge: Cambridge University Press, 2010), 4–5.

<sup>35</sup> Withuis, *Erkenning*, 118; Beunders, *Publieke tranen*, 37; Marij Leenders, ''Gesol met de rechten van vervolgde mensen'. Emoties in het asieldebat 1938–1999', in *Jaarboek Parlementaire Geschiedenis 2003: Emotie in de politiek*, ed. Carla van Baalen et al. (Nijmegen, Den Haag: Centrum

the more general definition of 'silence' provided by Winter. He considered silence in general as '(…) a socially constructed space in which and about which subjects and words normally used in everyday life are not spoken'.<sup>36</sup> Yet where did this idea of such a silence in the 1950s came from?

Silence regarding the war seems to parallel a persistent perspective of the 1950s in which it is considered to be a disciplined and forward-looking period of hard work, recovery, and reconstruction after the war.<sup>37</sup> That such a characterisation is so readily accepted has its origins in the persistence of an almost caricatural perception of the 1950s.<sup>38</sup> *Spruitjeslucht*, as displayed in Gerard Reve's well-known novel about post-war life*De Avonden* (1947)*,* became a popular characterisation of the 1950s in the Netherlands. Whilst the term literally refers to the smell of overcooked Brussels sprouts, it is here associated with a dull, narrow-minded Dutch civil society.<sup>39</sup> The silence outlined above resonates with this characterisation, and in Dutch history the 1950s are seen as a period in which little attention was paid to the most recent past and the (then ongoing) consequences of World War II.<sup>40</sup> Although the term 'silence' perhaps carries a different implication, the lack of (professional and public) attention paid to the (mental health) consequences of the war does not find its parallel in a total silence regarding anything war-related in society or in the political arena. When it comes to politics, for example, the Dutch national government still had to discuss, address, and solve many directly war-related issues in the 1950s.<sup>41</sup> Talking (or not talking) about the war, however, does not necessarily also resembles (witholding from) expressing emotions. Yet what are the implications of this idea of

voor Parlementaire Geschiedenis, SDU Uitgevers, 2003), 65; Jolande Withuis, 'Opkomst en neergang van PTSS', *Directieve therapie* 30, no. 3 (1 October 2010): 149–59; Withuis and Mooij, 'From Totalitarianism to Trauma', 198–99, 203–4; Oosterhuis, 'Mental Health, Citizenship, and the Memory of World War II'.

<sup>36</sup> Winter, 'Thinking about Silence', 4.

<sup>37</sup> In historiography, often references are made to the work by Dutch historian Hans Blom, although his well-known statement on 'years of discipline and asceticism' initially concerned mostly the period 1945–1950. See Hans Blom, 'Jaren van tucht en ascese. Enige beschouwingen over de stemming in herrijzend Nederland (1945–1950)', *BMGN − Low Countries Historical Review* 96, no. 2 (1 January 1981): 331–32; Withuis, *Erkenning*, 17–18, 31–32; Beunders, *Publieke tranen*, 18–19; Withuis, 'From Victims Divided To Victims United', 129–30, 138.

<sup>38</sup> Doeko Bosscher, 'De jaren vijftig epischer geduid', *BMGN − Low Countries Historical Review* 112, no. 2 (1997): 209–26.

<sup>39</sup> Simon van het Reve, *De Avonden*, 1st ed. (Amsterdam: De Bezige Bij, 1947); Jelle Leenes, *Hollandse luchten: ruiken aan Nederland* (Amsterdam: Atlas Contact, 2011).

<sup>40</sup> Beunders, *Publieke tranen*, 37; Leenders, 'Emoties in het asieldebat', 65; Withuis, 'Opkomst en neergang van PTSS'; Withuis and Mooij, 'From Totalitarianism to Trauma', 203–4.

<sup>41</sup> Ismee Tames and Jolande Withuis, 'Hoe de Koude Oorlog een gezamenlijke oorlogsherdenking onmogelijk maakte', *Historisch Nieuwsblad* 2010, no. 9 (2010).

'silence' for perspectives on publicly talking about or expressing war-related emotions?

Parallel to the myth of a 'silence', many historians did later identify an 'unemotional style' in dealing with or talking about anything war-related in the public domain. As a decade, the 1950s is often also associated with an unemotional way of dealing with the war and its consequences in politics. Talking about or expressing emotions related to this recent past was subject to a process of collective concealment. Dutch historian Remieg Aerts, writing about the parliamentary history of the Netherlands, described the political style of dealing with consequences of the war in the first two post-war decades as 'legal' and 'administrative'.<sup>42</sup> Politicians, especially those active during the 1950s, are described as having a 'stiff upper lip'. It is stated that, at that time, expressing emotions was not considered part of the standard etiquette.<sup>43</sup> However, this silent phase is not so much characterised by these particularities, as it is by its contrast with later periods.<sup>44</sup>

Many Dutch historians studying World War II in the Netherlands agree that something had changed in the 1960s in the way the consequences of the war were dealt with. A consensus seems to exist that the legal and distant style of dealing with the war's legacy that had prevailed throughout the 1950s, changed. Aerts wrote, in support of the idea of this 'break', that '[t]he right to emotion and recognition of individual suffering was not claimed until the late 1960s, when the debate on the war acquired a strong moral character'.<sup>45</sup> Amongst others, the historians Ido de Haan, Chris van der Heijden, and Frank van Vree attributed breaking 'the silence of the 1950s' to particular, concrete publications and events. A much-used example is the popular television series *De Bezetting* (1960) by Loe de Jong. De Jong was not only the first director of the former RVO and RIOD (now known as NIOD, Institute for War, Holocaust, and Genocide Studies), but also author of the standard reference on the history of the Kingdom of the Netherlands during World War II (*Het Koninkrijk der Nederlandeninde TweedeWereldoorlog*).<sup>46</sup> Other examples are the Eichmann trial(1961), and the publication of Jacques Presser's book *De Ondergang* (1965) on the persecution of Jews in the Netherlands.<sup>47</sup> These developments affirm a third phase of 're-

<sup>42</sup> Aerts, 'Emotie in de politiek', 21.

<sup>43</sup> Leenders, 'Emoties in het asieldebat', 65; Beunders, *Publieke tranen*, 196–97.

<sup>44</sup> Bosscher, 'De jaren vijftig epischer geduid', 209; Cas Wouters, 'The Slippery Slope and the Emancipation of Emotions', in *Politics and the Emotions: The Affective Turn in Contemporary Political Studies*, ed. Simon Thompson and Paul Hoggett (New York: Continuum, 2012), 213.

<sup>45</sup> Aerts, 'Emotie in de politiek', 21.

<sup>46</sup> Loe de Jong, *Het Koninkrijk der Nederlanden in de Tweede Wereldoorlog* (Den Haag: SDU Uitgevers, 1969).

<sup>47</sup> Frank van Vree, 'Televisie en de geschiedschrijving van de Tweede Wereldoorlog', *Theoretische Geschiedenis* 22, no. 1 (1995): 18–19; Ido de Haan, *Na de ondergang: De herinnering aan de Jodenvervolging in Nederland 1945–1995* (Den Haag: SDU Uitgevers, 1997), 116–18; Conny Kristel, *Ge-*

newed attention' that is also present in Withuis' cycle outlined above. Emotions have a prominent role in this presumed transformation. Van der Heijden, for example, stated that the 1960s provided the 'emotional space' to consciously recall and discuss the war past.<sup>48</sup> According to Dutch historian Hinke Piersma, emotions started to play a major role, not only in societal, but also in political engagement with the consequences of the war from the 1960s onwards.<sup>49</sup> Also according to Withuis, compassion for victims had increased, as had talking about, understanding, exchanging, and expressing related emotions.<sup>50</sup> Yet it is not the 1960s, but the beginning of the 1970s that is even stronger contrasted with the 1950s.

In Withuis' cycle, the phase covering the period 1968–1980 is mentioned as the one in which the consequences of 'breaking the past silence' really became visible.<sup>51</sup> During this time, the national Dutch government established new and elaborate legislative schemes. These schemes were not only tailor made, but also explicitly aimed at alleviating the immaterial consequences of World War II (e.g., trauma) for an ever-expanding group of victims. Withuis connected these developments to the so-called proto-professionalisation of trauma and the de-stigmatisation of mental illness. She considered the changes as an integral part of a wider, more generic democratisation movement that had started in the 1960s. Withuis identified this phase as one of an 'emancipation of emotions' because of a 'changing mental climate' that paralleled the 'renewed speaking' about personal experiences, suffering, consequences, and emotions related to World War II.<sup>52</sup> This development was often depicted by contemporaries as 'finally' addressing previously 'forgotten' war victims.<sup>53</sup> Piersma wrote how this perceived 'late action' was often viewed with astonishment

*schiedschrijving als opdracht: Abel Herzberg, Jacques Presser en Loe de Jong over de jodenvervolging* (Amsterdam: Meulenhoff, 1998); Heleen Pott, 'Onherkenbaar in de mist. Over herinneringen, emoties en de holocaust', *Krisis* 72 (1998): 46; Withuis, *Erkenning*, 17–18; Beunders, *Publieke tranen*, 125; Van der Heijden, *Dat nooit meer*, 329, 365–86.

<sup>48</sup> Van der Heijden, *Dat nooit meer*, 329.

<sup>49</sup> Piersma, *De Drie van Breda*, 193.

<sup>50</sup> Withuis, *Erkenning*, 118–19; Withuis, 'From Victims Divided To Victims United', 130–39; Withuis and Mooij, 'From Totalitarianism to Trauma', 204–9.

<sup>51</sup> Jolande Withuis and Annet Mooij, 'Conclusion', in *The Politics of War Trauma: The Aftermath of World War II in Eleven European Countries*, ed. Jolande Withuis and Annet Mooij (Amsterdam University Press, 2010), 328.

<sup>52</sup> Withuis, *Erkenning*, 118–20; Jolande Withuis, 'De veranderende omgang met de oorlog: van politiek naar psychologie', *Cogiscope: tijdschrift over gevolgen van oorlog en geweld* 3, no. 2 (2007): 28; Withuis, 'Opkomst en neergang van PTSS', 158; Withuis and Mooij, 'Conclusion: The Politics of War Trauma', 328; Oosterhuis, 'Mental Health, Citizenship, and the Memory of World War II', 20, 31.

<sup>53</sup> Chapter 9 of this book deals with a more elaborate discussion of this phenomenon. See also 'Handelingen Tweede Kamer 1969–1970', n.d., 105; 'Handelingen Tweede Kamer 1970–1971', n.d., 3477, 3485, 3495.

or, at the very least, understood as exceptional and overdue – both at the time and by later observers. The idea even resonates in the Dutch title of her book (*Bevochten recht*), with the layered meaning of 'obtained' and 'conquered' (*bevochten*) and 'justice' and 'rights' (*recht*). This implies that the right to welfare legislation and justice was something that had to be fought for, as it had to be obtained from an unwilling government.<sup>54</sup> To conclude, this third phase of changes in the public and political acknowledgement of trauma and related legislative developments was seen byWithuis as the result of an 'emancipation of emotions'. This emancipation, in its turn, was part of a broader historical and societal development in the 'mental climate' of the Netherlands that '(…) had changed in an unrecognisable manner'.<sup>55</sup>

#### Emotionalisation and Emotion Culture

Had such a shift actually occurred? Ideas of changing mental climates and parallel developments in dealing with emotion seem to build on a persisting idea that a so-called 'emotion culture' has been established after a process of 'emotionalisation' in the second half of the twentieth century.<sup>56</sup> Particularly since the turn of the century, this 'emotionalisation' presumption has become widespread: Not only in the Dutch context, but also in academic literature on the culture of, for example, the United States and the United Kingdom, the idea is prevalent.<sup>57</sup> Although explanations of this phenomenon vary between scholars and their particular perspectives, disciplines, and research interests, the general idea is broadly similar. Authors refer to a more or less generic historical development in which public displays of emotion have become more common, more socially accepted, and more frequent. Not only within academia, but also in broader public and press spheres, this idea of an emotion culture that has evolved in the last quarter of the twentieth century remains omnipresent.<sup>58</sup>

<sup>54</sup> Piersma, *Bevochten recht*.

<sup>55</sup> Withuis, *Erkenning*, 118–19.

<sup>56</sup> Beunders, *Publieke tranen*, 10, 147;Wouters, 'The Slippery Slope and the Emancipation of Emotions', 199–200; Frevert, 'The History of Emotions', 62; Frevert, 'Emotional Politics'.

<sup>57</sup> Beunders, *Publieke tranen*; M. D. Kimmich, 'Emotional Culture? An Investigation into the Emotional Coverage of Televised Leader Debates in Newspapers and Twitter' (PhD thesis, London South Bank University, 2016).

<sup>58</sup> See e.g., René Moerland, 'Nu ook nog in de politiek: emoties!', *NRC Handelsblad*, 24 December 2003; Gerben van Kleef, 'Emoties lijsttrekkers geven de doorslag in "emocratie"', *Trouw*, 11 September 2012; Marjolijn de Cocq, 'De Aantrekkingskracht van Een TRAGISCHE FIGUUR', *Limburgs Dagblad*, 25 March 2015; Charles Beckett, 'How Journalism Is Turning Emotional and What That Might Mean for News', *Polis* (blog), 10 September 2015, https://blogs.lse.ac.uk/polis/2015/09/10/how-journalism-is-turning-emotiona l-and-what-that-might-mean-for-news/; Xandra Schutte, 'Het Opdringerige Heden; Essay − De Emotionalisering van de Media', *De Groene Amsterdammer*, 5 April 2016; Char-

The different terms used to describe the aforementioned historical developments and phenomena seem semantically similar: 'emotionalisation', 'changing mental climate', 'emotion culture', etc. Nevertheless, there are differences in the explanations of the broader societal developments to which they refer. Withuis, for example, used 'mental climate' to refer to a broad societal and historical phenomenon. She considered 'emancipation of emotion' as a result, as an aspect of changes in this 'climate'. Withuis regarded 'emancipation of emotion' primarily as an increasing acknowledgement of, and attention for, emotions (of others), emotional damage, mental illness, and trauma. Dutch media historian Henri Beunders used a much broader interpretation of the term. In his book *Publieke tranen* (2002), Beunders described a more general increase in emotional expression and behaviour. The book deals mostly with the ways in which emotions were expressed and appreciated in public. Beunders described a historical development that started with a period of repressing or withholding emotions that later developed into the emergence of an 'emotion culture'. From that point onwards, an overwhelming openness and freedom to express emotions in the public space became established. He pointed to 'the emancipation of (mainly unpleasant) emotions' from the 1970s onwards, which developed into a general and overwhelming 'emotionalisation of [Dutch] society'.<sup>59</sup> According to Beunders, this development ultimately peaked around the turn of the century.<sup>60</sup> He observed emotionalisation in Dutch society, politics, television, (news) media,<sup>61</sup> and penal justice.<sup>62</sup>

Beunders depicted the course of this process not as a sudden break, but as a curve, which started to flatten out from 1900 onwards, reaching an absolute trough between 1955 and 1960. However, the curve started to rise again from the early 1970s onwards.<sup>63</sup> Similarly, Remieg Aerts emphasised the impact on political discourse and parliamentary debate. He interpreted emotionalisation as part of an ongoing 'wave motion' with, from 1848 onwards, periods of higher and lower emotionality

les Beckett, 'Media and the Manchester Attacks: Evil and Emotion', *Polis* (blog), 24 May 2017, https://blogs.lse.ac.uk/polis/2017/05/24/media-and-the-manchester-attacks-thegood-bad-and-propaganda/; Sander van Walsum, 'Vroeger hoorde het erbij, nu is het goed voor code rood', *De Volkskrant*, 16 February 2021.

<sup>59</sup> Beunders, *Publieke tranen*, 10, 39, 48.

<sup>60</sup> Beunders, *Publieke tranen*; 10, 39, 48. The predominancy of negative emotions was also emphasised in Bas van Stokkom, 'Morele woede en punitieve verlangens: Over de verrechtsing van het strafklimaat', in *Het hart op de tong: Emoties in de politiek*, ed. René Gabriëls, Sjaak Koenis, and Tsjalling Swierstra (Amsterdam: Amsterdam University Press, 2018), 83–106.

<sup>61</sup> Beunders, *Publieke tranen*; Schutte, 'Het Opdringerige Heden'.

<sup>62</sup> Henri Beunders, *Hoeveel recht heeft de emotie? Straffen in de slachtoffercultuur* (Amsterdam: Amsterdam University Press, 2018).

<sup>63</sup> Beunders, *Publieke tranen*, 41, 48.

– in his case regarding Dutch political debate. He also depicted the 1950s as a period of 'legal'and 'administrative' politics. In accordance with Beunders andWithuis, Aerts described the period from the early 1970s onwards as a period of increasingly emotional politics.<sup>64</sup> Subsequently, in *Macht der gewoonte*(2018)*,* the Dutch historian Carla Hoetink identified the 1980s as a period in which expressions of emotion were increasingly seen as unprofessional. Professional politicians, wrote Hoetink, distanced themselves from strong emotions or emotional expressions. She attributed these developments to increasingly technocratic and professional ways of doing politics, which was reflected in the tone of debate in parliament.<sup>65</sup>

Although Aerts and Hoetink dealt mostly with the political realm, and Beunders looked mainly at the media landscape, their curves of emotionalisation unmistakably mimic the 'changes in mental climate' that underlie Withuis' cycle in the historical development of the national coming to terms with the consequences of the German occupation of the Netherlands in World War II.

#### Received Wisdom

Emotionalisation as a historical phenomenon has gained the status of a so-called 'received wisdom'.Historian and political scientist Ismee Tames writes that theinvestigation of this type of knowledge is particularly interesting: 'Received wisdoms' offer (historical) researcher an opportunity to rethink issues that have been widely taken for granted for a long time.<sup>66</sup> The publications mentioned earlier offer a concrete diachronic framework of emotionalisation curves. Despite the different thematic approaches, scopes, and sources used, a consensus seems to exist that Dutch society, culture, and politics have undergone a process of emotionalisation. The belief that (expressions of) emotions have proliferated in the public domain since the 1970s is, also beyond the academic realm, hardly controversial. It is often mentioned in passing, and by constantly repeating the same statement, its acceptance has grown.<sup>67</sup>

<sup>64</sup> Aerts, 'Emotie in de politiek', 21.

<sup>65</sup> Carla Hoetink, *Macht der gewoonte: regels en rituelen in de Tweede Kamer na 1945* (Nijmegen: Vantilt, 2018), 421–98.

<sup>66</sup> The concept of 'received wisdoms bring us back to the start of this chapter', where I stated that the close relationship between emotions and the post-war dealing with the consequences of war and wartime experience seems self-evident. The link between emotions and war is frequently noticed, but seems at the same time so obvious that it is taken for granted and rarely subjected to any further investigation. This also goes for received wisdom regarding the broad societal development of 'emotionalisation'. The idea of 'received wisdom' is translated from the Dutch '*geaccepteerde wijsheden*' and is derived from Tames, *Doorn in het vlees*, 7–8.

<sup>67</sup> Examples in the academic literature are Oosterhuis, 'Mental Health, Citizenship, and the Memory of World War II', 20, 34; van Stokkom, 'Morele woede', 99.

This received wisdom resonates in the caricatural characterisation of the 1950s*,* in the 'cycle of speaking, silence, and renewed speaking', and in a persistent opposition between 'silence' (in the 1950s) and 'emotionalisation' (from the 1970s onwards) in academic literature. This opposition is not only taken for granted, but also remains uncontested and unquestioned. Consider, for example, *Publieke tranen* once more. Instead of first questioning the process of emotionalisation, the book primarily focusses on the motives (*drijfveren)* behind the emergence of a supposed emotion culture in print, media, and national television.<sup>68</sup> That such a process of emotionalisation in the second half of the twentieth century seems highly plausible, or easily relatable to personal experiences, does not necessarily makes it true.

The received wisdom of 'emotionalisation' is far from insignificant. Here, two observations are important. First, developments in dealing with emotions in public are used as explanatory factors for more specific historical developments. In this investigation, the role of this received wisdom in explaining the way the consequences of the war have been dealt with is particularly important. Central to this particular history are the national government's legislative schemes aimed at victims of the German occupation. Second, the developments of legislative schemes seem to have offered a useful diachronic framework of specific dates – in other words, a periodisation. Withuis' cycle, for example, displays how this framework of legislative changes is often connected with 'emotionalisation' as a historical process. These observations show how the history of war-related national legislative schemes is linked to the historiographic consensus on a broad societal development in talking about and expressing emotions. However plausible a link between the changes in postwar legislation or political dealing withconsequences of the war and 'emotionalisation' may seem, we do not know precisely how this relationship has developed over time, and if and how the role of emotions affected historical war-related legislative changes. How this relates to epistemological questions, common research practices in historiography, and empirical support, is discussed later in this chapter.

#### **1.3 Questioning the Role of Historical Emotions**

The ways in which the consequences of the war were subsequently dealt with in the Netherlands is neither a forgotten nor an overlooked theme in historiography. Indeed, the aim of this investigation is not merely to point at another lacuna or to fill in another gap in our knowledge of the past, nor is its primary concern to debunk myths. Instead, this investigation builds on an existing historiographic consensus and starts from a point of curiosity towards what kind of insights a different approach, or a new perspective, might provide. A re-evaluative analysis of

<sup>68</sup> Beunders, *Publieke tranen*.

the topic of the government's post-war engagement with the war's consequences is much needed and justified by the epistemological issues raised by received wisdoms regarding broader historical phenomena. How, for example, are historians able to investigate the course of broad historical trends and developments in (the expressions of) emotions in societies or national politics at all? How can we support the resulting historical claims with empirically convincing evidence that is both traceable and diachronic? In addition, there is also the epistemological question of how knowable past emotions are, and how we can investigate the role of emotions in the course of history. The particular theoretical aspects and difficulties that accompany any historical investigation of emotions are addressed in more detail in Chapter 2 and Chapter 4.

As 'emotionalisation' seems so fundamental, not only to the aforementioned historical claims, but also as an indispensable background needed for the more in-depth case studies of this book, this will be analysed first. This study assesses whether a process of emotionalisation can be observed in the Netherlands in the second half of the twentieth century. Chapter 5, based on a computer-assisted analysis, assesses whether the process of increasing emotionalisation can be discerned in a diachronic analysis of various historical text collections. Therefore, Dutch parliamentary proceedings, newspaper articles, and the annual speeches of the Dutch queen are analysed. How emotional were these source collections? How did the use of emotional language in these sources develop between 1945 and 1989? How does this rather broad societal phenomenon hold up in the context of these textual sources? The results of the analysis of the more general emotionalisation process forms a baseline for further investigations and the central questions of this study.

As a substantive and concrete historical case, the central question that this study addresses is not only *whether,* but also *how* emotions played a role in the national engagement with the consequences of the war in the Netherlands between 1945 and 1989. The aim is to assess not only whether emotions were manifest, and in what wordings, but also how they were expressed, discussed, and played a role in the parliamentary discussions regarding government interventions aimed at alleviating the suffering of the former members of the anti-Nazi resistance (case study 1) and various groups of war victims (case study 2). In doing so, this investigation aims for establishing a better understanding of how 'emotionalisation', 'changes in mental climate', or 'the emancipation of emotions' interacted with the national post-war dealing with consequences of WorldWar II in the Netherlands, and how this evolved over time. Central to the chapters addressing the case studies is the breakdown of the historical phenomenon to which the literature refers with terms such as 'mental climate', 'emotion culture', 'emotionalisation', or 'emancipation of emotions'. How do these broad and rather generic classifications hold up in the particular context of Dutch parliamentary discussions of war-related legislation? This study builds on

parliamentary discussions on numerous long-lasting cases, all of which are considered consequences of World War II.

#### War and Emotions in Parliament

To investigate if and how emotions may have possibly affected national legislation, this study focuses on the role and usage of emotional language in related parliamentary discussions. The national engagement with consequences of the war was not only debated in parliament, but also shaped, established, voted for, and constantly (re-)evaluated. Important issues in the national dealing with the consequences of the war, such as aid schemes for war victims, resistance pensions, and the prosecution of war criminals, were discussed many times in both houses of the Dutch parliament between 1945 and 1989. Although such clear-cut temporal boundaries always seem a little arbitrary in the investigation of diachronic developments over decades, for obvious reasons, such criticism can hardly be applied to the starting point of 1945. Not only did the German occupation of the Netherlands end that year, it also marks the first post-war phase of (re)starting the parliamentary democratic process in the Netherlands. The year 1989 seems less obvious. However, 1989 not only marks the end of the Cold War and the dissolution of the Communist Party of the Netherlands (CPN), but can also be considered the yearin which one of the most long-lasting warrelated Dutch controversies came to an end, as the last two foreign war criminals incarcerated in the Netherlands were finally released from prison.

Parliamentary proceedings are the main source used in this investigation. They are, in the first place, a relatively demarcated environment. As a historical source collection, Dutch parliamentary proceedings stand out in their completeness: Verbatim minutes of all discussions in both houses of parliament are included. They are not only readily available, but also digitised, structured, enriched, and machinereadable.<sup>69</sup> This dataset allows for detailed, sustained scrutiny (close reading), as well as for the application of computational methods for text analysis (text mining). In addition, the government interventions and legislative schemes discussed in parliament play a prominent role in both the historiography of 'emotionalisation' and in that of the aftermath of the war.

Even within the realm of post-war parliamentary discussion, it is difficult to establish a concrete demarcation or conceptualisation of what 'engaging with the

<sup>69</sup> Machine-readable means that a computer can process the information (characters, words, etc.) contained in the documents. This dataset is described in more detail in Chapter 3. See also M. Marx et al., 'Thematic collection: Political Mashup and Dutch Parliamentary Proceedings 1814–2013' (Data Archiving and Networked Services (DANS), 2012); Hinke Piersma et al., 'War in Parliament: What a Digital Approach Can Add to the Study of Parliamentary History', *Digital Humanities Quarterly* 8, no. 1 (1 April 2014).

consequences of the war' fully encompasses.The war is everywhere and 'everywhen'. Therefore, this book focusses mainly on demarcated case studies, dealing with groups of people that are relatable to emotions and the 'emotionalisation' literature. Each case study corresponds to a different subset of parliamentary debates. The analyses in Chapters 5 to 9 rely predominantly on the parliamentary discussion of topics related to three different groups of 'war actors': people that survived the war, or the descendants of people who did not survive. I focus on Dutch parliamentary debates dealing with the real and lived consequences for these different groups of war survivors. At some point, all of these people had been affected by the German occupation of the Netherlands, which took place between May 1940 and May 1945. Hence, they were the subject of war-related post-war national welfare legislation. A first subset of debates deals with former Nazi collaborators, perpetrators, war criminals, or people whose (alleged) wartime past caused controversy. Analysis of these discussions in Chapter 5 primarily supports the evaluation of the validity of the research methods. Debates on two other groups of war survivors are used for more in-depth scrutiny in Chapters 6 and 7 (case study 1) and in Chapters 8 and 9 (case study 2). Case study 1 deals with the so-called 'extraordinary government employees': people who had actively resisted the Nazi occupier between 1940 and 1945. Case study 2 follows the parliamentary discussion of a rather heterogeneous group of victims of the war who suffered, for example, from Nazi persecution, violence, repression, or warfare.

#### Tensions from Two Sides

This investigation brings together two tensions that originate in different academic disciplines and fields. First, there is the history of emotions. In 2014, historians Susan Matt and Peter Stearns wrote about their future expectations for the history of emotions.They anticipated that the digitisation of sources and development of computational methods '(…) can establish some analytical parameters in ways heretofore impossible'.<sup>70</sup> Concerning the investigation of emotions in history, however, there have been only a handful of attempts made at developing new, more systematic research methodologies to date. However, the interdisciplinary work by Inger Leemans, Erika Kuijpers, Janneke van der Zwaan, and Isa Maks represents a positive exception.<sup>71</sup> Their work is rather exceptional in the context of another tension that originates in another field: the digital humanities. Although the digital humanities, and digital history in particular, builds on decades of using computers in research, the more recent digital humanities 'boom' was at first primarily committed

<sup>70</sup> Susan Matt and Peter Stearns, 'Afterword', in *Doing Emotions History*, ed. Susan Matt and Peter Stearns, History of Emotions (Urbana: University of Illinois Press, 2014), 206.

<sup>71</sup> Zwaan et al., 'HEEM'; Leemans et al., 'Mining Embodied Emotions'.

to reflection,<sup>72</sup> experimentation, exploration, invention, and displaying the potential of computational methods.<sup>73</sup> The field is now entering a phase in which applying those same methods in pursuing new insights relevant to more traditional academic historical fields becomes more important.<sup>74</sup> Computational methods and digital or digitised sources are being used to advance existing humanities problems, questions, and debates.This is not only the case for the digital humanities in general, but for digital history in particular as well. Some examples of the growing impact of, for example, computational text mining in digital history are, amongst others, the investigation of political discussions by historians Kaspar Beelen<sup>75</sup> and Luke Blaxill,<sup>76</sup> or Melvin Wevers' PhD dissertation based on the computational analysis of digitised historical newspapers.<sup>77</sup>

#### **1.4 Emotions in Society and Politics: A Difficult Subject of Historical Inquiry?**

As outlined above, the history of the public and political engagement with the consequences of World War II builds on an elaborate historiography of various historical studies.The nature of these studies is often concrete and demarcated, and thus they permit a close, detailed, and careful approach to historical analysis and evidence based on a relatively limited number of sources. Consequently, most of the resulting academic literature is empirically rather strong. However, this does not apply to the study of the role of emotions, or the broad historical statements and claims underlying them. Whilst the related studies are generic and broad in scope, their empirical fundament is often rather weak. The attention paid to epistemological and methodological issues concerning the investigation of broader societal developments is limited. These issues, I argue, underlie the current consensual status of 'received wisdom' in historiography.The next section explains how the investigation

<sup>72</sup> Whenever trying something new with computers, there is always the risk of unsubstantiated 'tech optimism'. This is, I believe, not applicable to the field of digital humanities. This seems the result of elaborate reflection (both on and within the field), which is now followed by increasingly more work that is dedicated to the actual application of computational methods in historical and humanities research.

<sup>73</sup> Zaagsma, 'On Digital History', 7–11.

<sup>74</sup> Lucas van der Deijl, 'The Dutch Translation and Circulation of Spinoza's Tractatus Theologico-Politicus in Manuscript and Print (1670–1694): A Computational Reconstruction', *Quaerendo* 50, no. 1–2 (4 June 2020): 207–37; Romein et al., 'State of the Field'.

<sup>75</sup> Rheault et al., 'Measuring Emotion in Parliamentary Debates with Automated Textual Analysis'; Blaxill and Beelen, 'A Feminized Language of Democracy?'

<sup>76</sup> Blaxill, *The War of Words*.

<sup>77</sup> Wevers, 'Consuming America'.

of emotionalisation as a diachronic historical process links to some fundamental issues that are typical of this type of historical claim. In the subsequent paragraphs, the opportunities and limitations of a new hybrid approach to the sources, based on computational methods, are discussed.

#### Semi-quantitative Claims and Tolerance for Fuzziness

Like many societal developments, the 'emotionalisation of society' originates in what can be called a semi-quantitative claim that involves comparison over time.<sup>78</sup> In such comparative acts, historians tend to come up with conclusions such as 'typewriting as a skill has become *less* important', or 'bicycles became *more* popular as a way of transportation'. Although not always problematic, it can be an issue to make such semi-quantitative statements based on predominantly qualitative historiographic approaches to sources. This is reflected in the way in which historians sometimes write about historical change over time. Despite the semi-quantitative nature of the claim, the case for 'emotionalisation' is made without actually showing an empirically based diachronic historical pattern, curve, or trend as argument. Quantitative characteristics of developments or (collections of) sources are described with adjectives, rather than by using actual empirical measurements or numbers that are based on large(r)-scale source collections. Historian of British politics Luke Blaxill wrote that as a result of this shortcoming, '(…) we cannot be entirely sure – or prove – whether we have discovered an anthill, a hillock or a mountain'.<sup>79</sup>

Thus, historiographic claims regarding broad societal developments and their supporting arguments rarely excel in terms of precision or clarity.<sup>80</sup> Such claims often seem to suffer from a high tolerance for fuzziness.This insight is not new. As far back as 1978, social scientists Christien Brinkgreve and Michel Korzec had already identified this as a more general problem of historical research into broad societal shifts and changes.<sup>81</sup> The nature of this issue lies in the difficulty of investigating broad, heterogeneous entities such as 'society' or 'national politics' as a comprehensive whole. Even – or perhaps especially – in the case of abundant relevant and representative historical sources, it is unfeasible for an individual researcher to take everything into account.There is a so-called comprehensiveness issue, as individual

<sup>78</sup> Geoffrey Rockwell and Stéfan Sinclair, *Hermeneutica: Computer-Assisted Interpretation in the Humanities* (Cambridge, Massachusetts and London: The MIT Press, 2016), 41.

<sup>79</sup> Blaxill, *The War of Words*, 32–34.

<sup>80</sup> This is an observation that comes to the surface particularly in the process of formalising the different aspects of such historical claims to make them suitable for quantitative or computational analysis.

<sup>81</sup> Michel Korzec and Christine Brinkgreve, *'Margriet weet raad'. Gevoel, gedrag, moraal in Nederland 1938–1978* (Utrecht and Antwerpen: Het Spectrum, 1978), 10.

researchers cannot (close) read, for example, *all* parliamentary proceedings, or *all* newspaper articles of fifty years or more.<sup>82</sup> As a consequence, the 'great unread' is left unconsidered.<sup>83</sup>

This tolerance for fuzziness also resonates in the supporting arguments or evidence for broad societal claims. The broader the range of the claim, the weaker the evidence seems. Our brains bother us here. According to Blaxill, the issue originates in the fact that humans are,in contrast to computers, poor judges of quantity. For example, without computers, a researcher's ability to estimate frequencies throughout extended texts is inherently unreliable. This becomes problematic when historians investigate or make comparisons between long-term diachronic developments and their manifestation in large text data collections. In addition to the issue of quantity, complexity also plays a significant role. It is difficult for an individual researcher to empirically keep track of multiple emerging or contradicting patterns over time when reading, especially when patterns emerge that are not considered relevant at the onset of the investigation. This carries a risk of outweighing the importance of material that seems relevant to what the researcher is interested in, or already believes to be true, and to undermine other materials (which may not seem relevant at first sight). The issue originates in common practices within the historical discipline as well. Historians tend to illustrate the typicality of historical phenomena or events with supposedly representative examples, such as anecdotes or quotations. Although this is not always an issue,in the case of investigating demarcated, specific historical events or phenomena, in cases where research subjects are underrepresented, or in dealing with a scarcity of sources, it becomes problematic when making historical claims that consider long-term, broad societal developments.<sup>84</sup> The use of quotations or 'anecdotal' evidence, I argue, simply does not scale very well. It causes a mismatch between (the scale of) evidence and claim when used in supporting broad societal historical developments in the long term.

In the case of the post-war history of the Netherlands, scarcity of 'societal' sources, such as newspapers, magazines, radio bulletins, or political proceedings, is not an issue. Therefore, it is unnecessary to support diachronic historical claims that are the result of the investigation of accessible, large-scale collections of sources with 'anecdotal' evidence or supposedly 'representative' quotations alone. Despite an abundance of available and representative sources, '(…) none of us really know

<sup>82</sup> Joris van Eijnatten, Toine Pieters, and Jaap Verheul, 'Big Data for Global History: The Transformative Promise of Digital Humanities', *BMGN − Low Countries Historical Review* 128, no. 4 (16 December 2013): 58–59; Wevers, 'Consuming America', 41–42, 422–23.

<sup>83</sup> For more information on the origin of this term, see Darren Reid, 'Distant Reading, "The Great Unread", and 19th-Century British Conceptualizations of the Civilizing Mission. A Case Study', *Journal of Interdisciplinary History of Ideas* 8, no. 15 (15 December 2019): 1–15.

<sup>84</sup> Blaxill, *The War of Words*, 30–31.

what's in there yet'.<sup>85</sup> This is even more so in the case of providing evidence for silence or the absence of certain trends in such sources.<sup>86</sup> A combination of the unreliability of dealing with frequencies in long texts as an individual researcher, the tolerance for fuzziness, and a comprehensiveness issue all quite possibly impacted the empirical evidence of 'silence', 'emotionalisation', and 'emotion culture' in the available academic literature. Instead of approaching this semi-quantitative statement with both qualitative and quantitative evidence, incidents in the shape of well-known emotional peak moments, notorious historical actors and events, controversies, incidents, and scandals are over-represented.<sup>87</sup> They not only hold a predominant position, but also remain mostly uncontested. As mentioned earlier, examples of World War II-related issues are remarkably prominent. If there is one favourite example in academic literature, the so-called 'Breda Three' controversy from 1972 seems to take the lead.

#### A Favourite Example: The Breda Three

The Breda Three controversy centred on the last three surviving foreign war criminals incarcerated in the Netherlands. The parliamentary hearing on the 29th of February 1972 about a possible (early) release of the Breda Three (following a commuted death sentence) was characterised by many personal stories and the strong emotional involvement of all speakers present. It is generally considered one of the most emotional episodes in Dutch (political) history. The debate lasted for 13 hours.<sup>88</sup> Around this debate, '(…) emotions regarding the possible release [of the Breda Three] reached a peak', wrote Piersma.<sup>89</sup> Indeed, one of the leading actors in the case, and also the instigator of public anger, Minister of Justice Dries van Agt later described this parliamentary hearing as 'the most emotional hours' of his political life.<sup>90</sup> The case clearly demonstrates how war criminals remained a trigger for emotional responses long after the war had ended.This, and similar well-known

<sup>85</sup> Ted Underwood, 'A Genealogy of Distant Reading', *Digital Humanities Quarterly* 11, no. 2 (27 June 2017): 27.

<sup>86</sup> Rockwell and Sinclair, *Hermeneutica*, 41; Blaxill, *The War of Words*, 32–34.

<sup>87</sup> Examples include Beunders, *Publieke tranen*; Aerts, 'Emotie in de politiek'; Beunders, *Hoeveel recht heeft de emotie?*; An exception to this tendency is provided by Jolande Withuis' study of articles in decades of issues from the monthly magazine *Maandblad Geestelijke Volksgezondheid*. See Withuis, *Erkenning*.

<sup>88</sup> Beunders, *Publieke tranen*, 125, 205; Piersma, *De Drie van Breda*, 117–35; Van der Heijden, *Dat nooit meer*, 462–65; Oosterhuis, 'Mental Health, Citizenship, and the Memory of World War II', 28.

<sup>89</sup> Piersma, *De Drie van Breda*, 7.

<sup>90</sup> Carla van Baalen et al., eds., 'Ten geleide', in *Jaarboek Parlementaire Geschiedenis 2003: Emotie in de politiek* (Nijmegen, Den Haag: Centrum voor Parlementaire Geschiedenis, SDU Uitgevers, 2003), 8–9.

public and political controversies, seem to have served as perfect exemplars in supporting and proving the broader claim of the emotionalisation of society in the early 1970s.

However, in the context of investigating broader, relatively long-term developments in history, these incidents alone are not sufficient, as they offer rather limited insight or evidence. In addition, there are convincing historical counter-cases available that lend support to opposing claims about earlier emotional peak moments. For example, in 1952, more than 15.000 people marched through the streets of Amsterdam, protesting against the decision to give clemency to German war criminal Willy Lages (sometimes known as the 'fourth' of the Breda Three).<sup>91</sup> Although the characteristics of this protest differed from the events of 1972, such a public display of emotion in the early 1950s conflicts with the generally accepted idea that emotions were generally silenced and concealed at that time.Whilst particular events *can* offer convincing and transparent historical evidence regarding specific histories and/or historical tipping points, including, of course, for the cases themselves, in supporting claims about broader societal change, they lack scalability. Such cases do not provide verifiable support for claims regarding broad, complex, multifaceted, diachronic phenomena or longer-term changes. This is not only an issue in providing arguments for an increasing or decreasing trend in emotionality; it is also particularly problematic when historians want to prove silence or the absence of such a development in their sources.<sup>92</sup> This investigation does not discard the particular or the incidental, but connects them to a quantitative perspective on the sources based on a computer-assisted methodology.

#### Talking About Emotions

Another fundamental issue is the differentiation between speaking in an emotional *manner* or*style* and speaking *about* emotions.Withuis, for example, used 'emancipation of emotions' and historical changes in the 'mental climate' of the Netherlands as explanatory factors for increasing recognition and acknowledgement of war victims and World War II-related trauma and misery in the 1970s.<sup>93</sup> This implies that speaking *about* emotions increased alongside a corresponding and powerful increase in the expression *of* emotions.Of course,it is plausible that increasing public openness to the expression of emotions also raised more understanding of war-related mental problems, trauma, and the emotions of others.That this development eventually

<sup>91</sup> Piersma, *De Drie van Breda*, 51.

<sup>92</sup> Rockwell and Sinclair, *Hermeneutica*, 41; Blaxill, *The War of Words*, 33–34.

<sup>93</sup> See e.g., Withuis, *Erkenning*, 118–21; Oosterhuis, 'Mental Health, Citizenship, and the Memory of World War II'.

stimulated the establishment of tailored national legislation, also seems likely.However, whilst such a connection is plausible,it should not necessarily be assumed to be either true or justified. Expressing emotion or behaving in an emotional way ('emotionalisation', 'emotion culture') is not necessarily the same as talking about emotions – either one's own or those of others ('emancipation of emotions'). Conversely, an increase in the discussion of emotion itself does not necessarily have to be subject to processes of emotionalisation. It is of course perfectly possible to acknowledge (other people's) emotions, without acting or expressing oneself in an emotional way.

#### A Twofold Aim

The difficulties of dealing with large-scale text collections, comprehensiveness issues, tolerance for fuzziness, and the predominance of qualitative evidence to support semi-quantitative claims all indicate the pressing need for new approaches.<sup>94</sup> They inspire a re-evaluative investigation in which the epistemological and methodological aspects of current historiography, new approaches, and historiographic practices are combined and evaluated. This investigation builds on a hybrid and iterative analytical cycle, in which traditional historiographic practices are combined with computational methods. Overall, the approach taken in this investigation can be summarised as a 'computer-assisted research methodology'. The application of emotion mining in a computer-assisted research methodology makes the aim of this study twofold: In addition to the substantive historical questions and a re-evaluation of historiography, the study reflects on the integration of computational methods for analysing texts in the context of historical research practice. The work of 'digital historians' mentioned earlier (Beelen, Blaxill, and Wevers) illustrates how methods from disciplines such as computational linguistics can be integrated into historical research practice and,in the process, adapted to the questions, interests, and needs of the historian.<sup>95</sup> The application of computational techniques to the investigation of digitised historical resources forms an important part of the emergent field of the digital humanities – a development that is not only relatively recent, but also still very much in progress.<sup>96</sup>

When investigating the role of emotions in historical texts, the integration of computational techniques for emotion mining certainly seems a fruitful path to ex-

<sup>94</sup> Additional epistemological questions regarding the investigation of historical emotions will be addressed in Chapter 2.

<sup>95</sup> Rheault et al., 'Measuring Emotion in Parliamentary Debates with Automated Textual Analysis'; Blaxill and Beelen, 'A Feminized Language of Democracy?'; Wevers, 'Consuming America'; Blaxill, *The War of Words*.

<sup>96</sup> For a brief overview of the emergence of the digital humanities, see Gregor Wiedemann, *Text Mining for Qualitative Data Analysis in the Social Sciences: A Study on Democratic Discourse in Germany* (New York: Springer, 2016), 3–7.

plore. Emotion mining is the use of Natural Language Processing (NLP) or quantitative computational text analysis to systematically identify, extract, quantify, and analyse emotions in (large collections of) texts. The technicalities of the research methods, materials, and techniques used in this study are presented in more detail in Chapter 4. As this approach currently remains uncommon in historical research, this study also constitutes an evaluation of the application of existing computational methods to emotion mining in history. Emotion mining, however, does not represent a novel technological invention. Rather, it is a methodological innovation that consists of the application of existing methods and resources within the realm of historical research.

#### **1.5 Digitised Sources and a Computer-assisted Methodology**

The availability of large-scale, digitised, structured, and machine-readable historical text sources allows historians to integrate computational methods and techniques into their research practice.The size, structure, and format of such digitised source collections create the possibility of computer-assisted quantitative and qualitative analyses hitherto considered impossible.<sup>97</sup> The popularity of the digital humanities and the application of computer-assisted methodology in academic history now benefits from a reciprocal relationship with the large-scale digitisation of comprehensive historical source collections. Examples are the digital collections of political proceedings from all over the world (Political Mashup, DiLiPaD), or the ever-growing newspaper collections digitised in the context of projects such as *Delpher* (Royal Library of the Netherlands) or the collaborative project in Luxembourg and Switzerland known as *impresso*. <sup>98</sup> These initiatives and their resulting websites have already altered the way in which historians interact with primary

<sup>97</sup> Wevers, 'Consuming America', 41–43.

<sup>98</sup> Marx et al., 'Thematic collection'; 'Delpher − Kranten, Boeken & Tijdschriften', accessed 1 September 2020, https://www.delpher.nl/; 'Impresso − Media Monitoring of the Past', impresso (impresso), accessed 1 September 2020, https://impresso-project.ch/; Michiel van Groesen, 'Digital Gatekeeper of the Past: Delpher and the Emergence of the Press in the Dutch Golden Age', *Tijdschrift Voor Tijdschriftstudies* 38(December 2015): 9–19; Maud Ehrmann, Estelle Bunout, and Marten Düring, 'Historical Newspaper User Interfaces: A Review', in *Proceedings of IFLA WLIC 2019*, 2017; Marten Düring et al., 'A Digital Future for Yesterday's News', in *Proceedings of Digital Humanities Benelux 2018*, 2018, 3.

sources.<sup>99</sup> Digitised and machine-readable historical text collections have, however, more to offer.

#### The Computer and the Historian's Macroscope

Scholars Shawn Graham, Ian Milligan, and Scott Weingart coined the term 'the historian's macroscope' to describe and understand the role of computational techniques in historical research methodology.They consider this macroscope as similar to a microscope or a telescope, however, rather than seeing things that are either very small, or very far away, the macroscope helps the historian to see things also at large.<sup>100</sup> This relates, first and foremost, to the most obvious advantages of using a computer: speed and scale. A computer can read and analyse millions of words and thousands of documents, topics, or keywords in seconds. Especially when investigating large-scale historical text collections, such as decades of parliamentary proceedings, simply speaking, the computer helps to speed things up. This makes a large-scale, long-term perspective on contemporary history feasible in terms of research time and resources. Then there is scale. Computers are reliable assessors when it comes to large quantities. They can be easily programmed to assess large volumes of data. Computers can execute long and complex instructions with ease, efficiently completing large-scale tasks that would be difficult and tedious to achieve manually.<sup>101</sup> Thus, using computers can help to make complex and largescale analyses feasible, even for individual researchers or small research groups. The use of computers in historical research addresses the fundamental problem of comprehensiveness in historiographic practice: the inability of researchers to read *all* parliamentary debates, or *all* national newspapers. The macroscope helps historians to attempt empirical problems at a different speed and scale, and to go beyond the qualitative assessment of a limited amount of sources or evidence when investigating broad historical phenomena.<sup>102</sup>

Graham, Milligan, and Weingart's definition of the macroscope can best be understood as a metaphor that refers as much to a (set of) tool(s), as to a new and differ-

<sup>99</sup> Graham, Milligan, and Weingart, *Exploring Big Historical Data*, 267; Estelle Bunout and Milan van Lange, 'Nibbling at Text: Identifying Discourses on Europe in a Large Collection of Historical Newspapers Using Topic Modelling', *C2DH | Luxembourg Centre for Contemporary and Digital History* (blog), 21 June 2019, https://www.c2dh.uni.lu/thinkering/nibbling-text-identi fying-discourses-europe-large-collection-historical-newspapers-using.

<sup>100</sup> Graham, Milligan, and Weingart, *Exploring Big Historical Data*, xvi, 1.

<sup>101</sup> Rockwell and Sinclair, *Hermeneutica*, 43.

<sup>102</sup> Graham, Milligan, and Weingart, *Exploring Big Historical Data*, 70–71; Blaxill, *The War of Words*, 31.

ent additional perspective on historical sources.<sup>103</sup> As the macroscope approaches sources also 'at large', its perspective is considered also as more 'distant', contrary to so-called 'close reading' – the careful consideration of texts and sources that is more common to historical research. However, they stress that the use of computational methods is not necessarily the only way historians will do their work in the future. Rather, computational methods are complementary, as they are an addition to more traditional ways in which historians deal with their sources. Yet computational methods in no way replace detailed scrutiny based on the close reading of sources. Furthermore, computational methods in no way discard going back and forth, zooming in and out, and approaching the same historical source or collection from different perspectives. Central to this study is the belief that this back-andforth between qualitative and quantitative perspectives, between close and distant reading, between traditional historiographical and innovative computational methods, not only has the potential to operationalise the central research questions, but also offers promising opportunities and new perspectives to the historical field.

Historical research can be broad in scope and comprehensive in terms of scale. The latter particularly applies when there is an abundance of source materials available, as is often the case in the history of the second half of the twentieth century. The macroscope facilitates the analysis of sources in a perspective that is not only quantitative, but also systematic, comprehensive, and diachronic. This means that not only the incidental or well-known flashpoints of history are taken into account. The macroscope offers a perspective that is comprehensive and comparative over time, as it also takes into account what has been described as 'the great unread' or, in the words of Blaxill, 'the dull sea of grey'.<sup>104</sup> Identifying quantitative trends, presences, or absences in texts can be done more rigorously, and can even provide evidence for what is lacking in the sources.<sup>105</sup> Making that which is absent in the sources more visible, implies that a concept such as 'silence' has the potential to become a more empirically observable analytical concept. How that works in practice is the subject of an ongoing evaluation throughout this book. The need for a comparative, diachronic, and quantitative perspective on comprehensive digitised historical datasets makes the case for the integration of systematic methods based on computational analysis techniques in historical research practice.

<sup>103</sup> Graham, Milligan, and Weingart, *Exploring Big Historical Data*, xv, xviii, 1; Kristoffer Nielbo, 'Mind the Text: Traces of Mental States in Unstructured Historical Data', in *Evolution, Cognition, and the History of Religion: A New Synthesis*, ed. Anders Klostergaard Petersen et al. (Leiden and Boston: Brill, 2019), 248.

<sup>104</sup> Blaxill, *The War of Words*, 31.

<sup>105</sup> Graham, Milligan, and Weingart, *Exploring Big Historical Data*, 70–71; Reid, 'Distant Reading'; Blaxill, *The War of Words*, 31.

#### Text Mining

Most computational techniques used in this investigation have not been developed within the domain of historical research. Existing Natural Language Processing (NLP) techniques and resources come from fields such as computer or corpus linguistics, computer science, and (social) psychological research (for more details, see Chapters 3 and 4). These techniques facilitate a computational quantitative analysis of language. The process of using a computer to derive information from (large bodies of) texts is also known as text mining. Text mining is a general term that refers to a wide range of techniques for the computational processing, analysis, and (statistical) evaluation of (usually) large text corpora. Text mining facilitates the discovery, retrieval, identification, measurement, comparison, or statistical evaluation of patterns of language use.<sup>106</sup> Relatively recent technical innovations in the field of text mining have paved the way for its integration into various academic disciplines.<sup>107</sup> What kind of language is (or can be) 'mined' is highly dependent on disciplinary practices, available text collections, resources, research interests, and the questions being asked. This can be, to give some examples, the use of pronouns in personal writings,<sup>108</sup> the sentiment polarity or emotional charge of words used

<sup>106</sup> Eijnatten, Pieters, and Verheul, 'Big Data for Global History', 59–60; Johanna Drucker et al., 'Distant Reading and Cultural Analytics', in *Introduction to Digital Humanities: Concepts, Methods, and Tutorials for Students and Instructors* (Los Angeles: UCLA, 2014), 43.

<sup>107</sup> Alessandro Marchetti, Rachele Sprugnoli, and Sara Tonelli, 'Sentiment Analysis for the Humanities: The Case of Historical Texts' (Digital Humanities, Lausanne, 2014); Matthew Jockers, *Text Analysis with R for Students of Literature*, Quantitative Methods in the Humanities and Social Sciences (New York: Springer International Publishing, 2014); Martijn Schoonvelde, Christian Pipal, and Gijs Schumacher, 'Text as Data in Political Psychology', in *Atlas of Language Analysis in Psychology*, ed. Ryan Boyd and Morteza Deghani (Guilford Press, 2021).

<sup>108</sup> James W Pennebaker, *The Secret Life of Pronouns: What Our Words Say about Us* (New York: Bloomsbury Press, 2013).

in consumer reviews,<sup>109</sup> literary texts,<sup>110</sup> tweets,<sup>111</sup> plays,<sup>112</sup> songs,<sup>113</sup> and political texts.<sup>114</sup> It can also be used to investigate the mental states of individuals, groups, or even entire societies.<sup>115</sup> In the specific context of historical research, text mining has already been applied to facilitate, for example, historical comparisons between groups, countries, or political parties, and to the investigation of diachronic historical trends in political and public discourse.<sup>116</sup>

A fundamental aspect of text mining is its focus on natural language. When text mining is applied in a research context, this is often operationalised by analysing the occurrences of very particular words (or sets thereof), word combinations, or

110 Saif M. Mohammad, 'From Once Upon a Time to Happily Ever After: Tracking Emotions in Novels and Fairy Tales', *Proceedings of the 5th ACL-HLT Workshop on Language Technology for Cultural Heritage, Social Sciences, and Humanities*, 24 June 2011, 105–14; Saif M. Mohammad, 'From Once upon a Time to Happily Ever after: Tracking Emotions in Mail and Books',*Decision Support Systems* 53, no. 4 (November 2012): 730–41; Acerbi et al., 'The Expression of Emotions in 20th Century Books'; Olivier Morin and Alberto Acerbi, 'Birth of the Cool: A Two-Centuries Decline in Emotional Expression in Anglophone Fiction', *Cognition and Emotion* 31, no. 8 (12 February 2016): 1663–75.

<sup>109</sup> Ashlee Humphreys and Rebecca Jen-Hui Wang, 'Automated Text Analysis for Consumer Research', *Journal of Consumer Research* 44, no. 6 (1 April 2018): 1274–1306.

<sup>111</sup> David Robinson, 'Text Analysis of Trump's Tweets Confirms He Writes Only the (Angrier) Android Half', *Variance Explained* (blog), 9 August 2016, http://varianceexplained.org/r/trump-t weets/.

<sup>112</sup> Zwaan et al., 'HEEM'; Leemans et al., 'Mining Embodied Emotions'.

<sup>113</sup> Sara Locatelli, 'Statistics Sunday: Welcome to Sentiment Analysis with "Hotel California"', *R-Bloggers* (blog), 20 May 2018, https://www.r-bloggers.com/statistics-sunday-welcome-tosentiment-analysis-with-hotel-california/.

<sup>114</sup> Steven Grijzenhout, Maarten Marx, and Valentin Jijkoun, 'Sentiment Analysis in Parliamentary Proceedings', in *From Text to Political Positions: Text Analysis across Disciplines*, ed. Bertie Kaal, Isa Maks, and Annemarie van Elfrinkhof (Amsterdam: John Benjamins Publishing Company, 2014), 117–34; Rheault et al., 'Measuring Emotion in Parliamentary Debates with Automated Textual Analysis'; Denise Traber, Martijn Schoonvelde, and Gijs Schumacher, 'Errors Have Been Made, Others Will Be Blamed: Issue Engagement and Blame Shifting in Prime Minister Speeches during the Economic Crisis in Europe', *European Journal of Political Research* 59, no. 1 (2020): 45–67; Schoonvelde, Pipal, and Schumacher, 'Text as Data in Political Psychology'; Milan van Lange and Ralf Futselaar, 'Vehemence and Victims: Emotion Mining Historical Parliamentary Debates on War Victims in the Netherlands', *DH Benelux Journal* 3 (2021).

<sup>115</sup> Pennebaker, *The Secret Life of Pronouns*; Mark Dechesne and Bryn Bandt-Law, 'Terror in Time: Extending Culturomics to Address Basic Terror Management Mechanisms', *Cognition and Emotion* 33, no. 3 (3 April 2019): 492–511; Nielbo, 'Mind the Text'.

<sup>116</sup> Rheault et al., 'Measuring Emotion in Parliamentary Debates with Automated Textual Analysis'; Blaxill and Beelen, 'A Feminized Language of Democracy?'; Wevers, 'Consuming America'; Milan van Lange and Ralf Futselaar, 'Debating Evil: Using Word Embeddings to Analyse Parliamentary Debates on War Criminals in the Netherlands', *Contributions to Contemporary History* 59, no. 1 (7 June 2019): 140–56; Blaxill, *The War of Words*; Van Lange and Futselaar, 'Vehemence and Victims'.

sentences in language use. In general, the study of language use is, in the words of computational social scientist GregorWiedemann, '(…) like a window through which they [researchers] try to reconstruct the ways speaking actors perceive themselves and the world around them'.<sup>117</sup> Also for historians, the study of language is a familiar approach. Examples include discourse analysis and conceptual history or *Begriffsgeschichte*. <sup>118</sup> Whether implicitly or explicitly addressed, language provides insights into the perceptions of the worlds surrounding particular historical actors – individuals, groups, societies, or communities representing a specific place or period. The focus on language contributes to a deeper understanding of social phenomena in their contemporary context in addition to investigating actual social or historical phenomena, events, or the outcomes of historical processes.<sup>119</sup> How this works for emotional expressions, is addressed in Chapter 2. Before natural language or texts can be 'mined' with computational text mining techniques, however, the computer needs a basic understanding of the language itself, not least as a computer cannot work with a broad or ambiguous understanding of phenomena as they are used in historical research, such as 'technocracy', 'fascism', 'censorship', 'hatred', or 'femininity'.

#### Formalisation

To investigate historical phenomena in text corpora using a computer, a researcher has to make explicit what it is that the computer has to 'mine'; what it is that the computer has to look for in the sources; and how this is possibly manifest in natural language use or the choice of words. It is a necessity to 'tell' the computer in an unambiguous, formalised way what has to be analysed and how this can be manifest in written sources. To give some examples: What kind of words define 'feminist' discourse in 1970s Britain?<sup>120</sup> What makes a political speech emotional in 1950? What language represents 'fascism'? In other words, a strict formalisation is necessary in

<sup>117</sup> Wiedemann, *Text Mining for Qualitative Data Analysis*, 17–18.

<sup>118</sup> This was already suggested by Pim Huijnen et al., 'A Digital Humanities Approach to the History of Science Eugenics Revisited in Hidden Debates by Means of Semantic Text Mining', *SO-CIAL INFORMATICS, SOCINFO 2013* 8359 (2014): 70–84; see also Tom Kenter et al., 'Ad Hoc Monitoring of Vocabulary Shifts over Time', in *Proceedings of the 24th ACM International on Conference on Information and Knowledge Management* (CIKM'15: 24th ACM International Conference on Information and Knowledge Management, Melbourne Australia: ACM, 2015), 1191–1200; Djouaria Ghilani et al., 'Looking Forward to the Past: An Interdisciplinary Discussion on the Use of Historical Analogies and Their Effects', *Memory Studies* 10, no. 3 (July 2017): 274–85; Michael Piotrowski and Mateusz Fafinski, 'Nothing New Under the Sun? Computational Humanities and the Methodology of History', in *Computational Humanities Research*, 2020; Wevers and Koolen, 'Digital Begriffsgeschichte'.

<sup>119</sup> Wiedemann, *Text Mining for Qualitative Data Analysis*, 17–18.

<sup>120</sup> Blaxill and Beelen, 'A Feminized Language of Democracy?'

order to sharpen the objects of investigation, in particular in the context of the computational methods used in this investigation.

However, the application of such processes of formalisation has some downsides. The use of text mining forces researchers to make methodological adjustments to their research objects. The process of defining a phenomenon such as 'hatred' into a relatively limited set of words, for example, is rather reductionist. Some of the complexity and ambiguity of language is undeniably sacrificed in the process.<sup>121</sup> Thankfully, the full complexity of language is often unnecessary to analyse texts effectively. This applies, at least, to language in bulk.<sup>122</sup> The reciprocal systematised or formalised approach has important advantages, especially when 'fuzzy phenomena' or hard-to-grasp concepts, such as emotion, are to be analysed.

In fact, I consider it an advantage that computational techniques encourage historians to frame the objects of their research in systematic and formalised ways. Commonly, historians rely on personal judgement and interpretation of the sources concerned. This does not have to be problematic, as it is often via personal and empathetic encounters with the sources that historians gain a connection to, and understanding of, the past. Nevertheless, especially when studying ethically charged controversies in the relatively recent past, there is a vulnerability to the introduction of bias.<sup>123</sup> The application of text mining forces researchers to formalise the grounds on which they base their claims, as they have to make the precise criteria they use in their analytical process very explicit.<sup>124</sup> This leads to greater transparency regarding the choices made in the analytical process and has the potential to make the possible introduction of unwanted biases more visible. This stimulates increased reflection on how analytical processes are possibly affected by what we analyse, our existing presumptions as researchers, common knowledge, or received wisdoms.<sup>125</sup>

#### Verifiability

The formalisation of criteria in text mining makes (parts of) the historical analysis of large datasets not only transparent, but also replicable. This makes the text-mining aspects of the analytical process more verifiable compared to traditional historiographical practice. Geoffrey Rockwell and Stéfan Sinclair, both professors of digital

<sup>121</sup> As the performance of computational text mining can never be guaranteed, validation by more traditional close reading remains essential. See also Justin Grimmer and Brandon Stewart, 'Text as Data: The Promise and Pitfalls of Automatic Content Analysis Methods for Political Texts', *Political Analysis* 21, no. 3 (2013): 268.

<sup>122</sup> Grimmer and Stewart, 272.

<sup>123</sup> Van Lange and Futselaar, 'Debating Evil', 143.

<sup>124</sup> Rockwell and Sinclair, *Hermeneutica*, 41; Underwood, 'Distant Reading'; Blaxill, *The War of Words*, 34.

<sup>125</sup> Rockwell and Sinclair, *Hermeneutica*, 43.

humanities, state in their book *Hermeneutica* that text mining even holds the promising potential of providing more certainty in historical claims that are made based on the analysis of large collections of texts.<sup>126</sup> However, they do not believe that this will lead to the same form of certainty found in other (exact) scientific processes. I consider not so much certainty, but replicability and thus verifiability a huge advantage of the text mining approach. To put it in Rockwell and Sinclair's words:

'If you check my interpretation and disagree with my claims, you can criticize the analysis, the choices of texts, the techniques applied, and the interpretations of results – something you can't do if my interpretation is based on anecdotal quotes or on implied authority.'<sup>127</sup>

This becomes especially relevant when making historical claims that concern broad societal developments over an extended period. The application of text mining in historical analysis provides not only researchers, but also their readers with the possibility to verify whether, for example, an observation is part of a broader trend, a significant change, or a minor development occurring in only a handful of instances. This is especially valuable when the output is used to support claims about diachronic trends or in comprehensive comparisons.<sup>128</sup>

#### Objectivity

The integration of text mining in historical research adds a quantitative perspective. It supports historical comparability and makes analytical processes more verifiable. Nevertheless, this does not mean that text-mining results are automatically also more reliable, objective, or true. With text mining comes quantification, and with quantification comes the pitfall of considering the resulting 'hard measurements' or numeric outputs as objective truth. In the first place, capturing complex, ambiguous, or abstract phenomena in numeric terms can hardly be very precise. This also applies to the application of text mining in the study of historical emotions in the language of parliamentary debates, as computational methods also rely on assumptions and strategies, and come with (sometimes unknown) implications or fallacies.<sup>129</sup> Quantitative output, such as numeric measurements, graphs, or statistics, can be verifiable, insightful, and even reliable, but should never be considered

<sup>126</sup> Rockwell and Sinclair, 187.

<sup>127</sup> Rockwell and Sinclair, 188.

<sup>128</sup> Blaxill, *The War of Words*, 32–33.

<sup>129</sup> Bernhard Rieder and Theo Röhle, 'Digital Methods: Five Challenges', in *Understanding Digital Humanities*, ed. David M. Berry (London: Palgrave Macmillan UK, 2012), 67–85, 70, cited in ; Zaagsma, 'On Digital History', 23; Isabella Sarto-Jackson, Robert J. Richards, and Richard R. Nelson, 'Quality & Quantity: Limits of Quantification in the Sciences', *Biological Theory* 10, no. 3 (1 September 2015): 185.

as the final answer or objective truth. It is important to be aware of the reification – the fallacy of misplaced concreteness – that the quantitative outputs of text mining (numbers) can carry.<sup>130</sup>

#### Beyond Experimentation and Performance

The final part of this introduction discussed some of the pros and cons of 'the historian's macroscope' in theory. As part of an analytical workflow, it potentially allows for a 'back-and-forth' between a computational quantitative perspective to the sources that is 'distant', and a 'closer' reading of the sources that is qualitative and more common to historiographical practices.The true value of themacroscopein the practice of historical research, however, remains rather uncertain. A central aim is therefore to evaluate the value of a computer-assisted approach in this investigation – one that is primarily historical. The critical evaluation of existing historiography with the application of computational techniques builds on the experimentation, exploration, and the display of potential typical to the digital humanities.This study is about the application of those techniques in the particular practice of historical research, rather than concerning a technical evaluation of the performance of the techniques or methods as such.This brings together the tensions between common historiographic practices; how they have affected our understanding of the role of emotions in the post-war history of dealing with the war; and the possibilities and pitfalls of a novel computer-assisted approach. Rethinking the ways in which we look at the role of emotions in the parliamentary post-war dealing with the consequences of the German occupation in the Netherlands serves as the substantive historical case.

To conclude, this chapter has left open two issues fundamental to this investigation. The first regards methodology: The macroscope sounds nice, but what is needed to make it happen? How to shape a computer-assisted methodology in historical research that,inevitable, always relies to some extent on old paper?The materials and data used are discussed in Chapter 3. Chapter 4 discusses what the macroscope entails in this study. First, however, Chapter 2 is dedicated to fundamental epistemological questions: How to define emotions? How to identify (changes in) emotions? How knowable are historical emotions? And how to analyse emotions historically?

<sup>130</sup> Sanne Blauw, *Het best verkochte boek ooit\*: \*met deze titel: hoe cijfers ons leiden, verleiden en misleiden* (Amsterdam: De Correspondent, 2018), 49, 89.

## **2. Emotions**

#### **2.1 Defining Emotion**

'Emotion' is like 'language' or 'freedom': These concepts are supposedly well known to the extent that their everyday use is often taken for granted.<sup>1</sup> As human beings, we are quite familiar with emotional states, ranging from negative ones associated with anger, fear, disgust, and sadness to positive states like joy, trust, and love. Yet what is emotion? Some people say that football is emotion. Defining emotion more precisely is however difficult. If there is one thing that the many attempts to arrive at a definition of emotion in various academic disciplines show, it is that definitions of the words 'emotion', 'emotions', and 'emotionality' are subject to disagreement, changing perspectives, and cultural and linguistic differences. Debates about emotions are complex and filled with ambiguities and diverging nuances. In addition, it remains difficult to distinguish between emotion and feeling, or between emotion and mood, affect, passion, or sentiment. These distinctions are discussed within a wide range of disciplines, and show how difficult it is to provide a fixed definition.<sup>2</sup> Before investigating the role of emotions in historical sources, this chapter explores the definition of emotion as an object of research in more detail.

Emotions are triggered by external or internal stimuli. On the individual level, emotional responses and expressions result from appraisals of these stimuli.<sup>3</sup> Personal emotions are a complex of neuro-physical, social, and cultural actions and reactions.<sup>4</sup> In psychology, 'affect'is often considered as unconscious and embodied (in facial expressions, for example), whereas 'emotion' is more consciously anchored in

<sup>1</sup> Lucia Omondi, 'Dholuo Emotional Language: An Overview', in *The Language of Emotions*, ed. Susanne Niemeier and René Dirven (Amsterdam: Benjamins, 1997), 88; Gregory and Åhäll, 'Introduction: Mapping Emotions, Politics and War', 3.

<sup>2</sup> Tsjalling Swierstra, René Gabriëls, and Sjaak Koenis, 'Het hart op de tong', in *Het hart op de tong: Emoties in de politiek*, ed. René Gabriëls, Sjaak Koenis, and Tsjalling Swierstra (Amsterdam: Amsterdam University Press, 2018), 8–9; Damasio, *De vergissing van Descartes*, 150; Holt et al., *Psychology: The Science of Mind and Behaviour*, 474–75; Kimmich, 'Emotional Culture?', 4.

<sup>3</sup> Holt et al., *Psychology: The Science of Mind and Behaviour*, 475.

<sup>4</sup> Leemans et al., 'Mining Embodied Emotions', 2.

language and meaning.<sup>5</sup> Political scientists Todd Hall and Andrew Ross reserve the term 'emotion' for the more socially recognized and structured episodes of what they describe as 'affective responses'. In this definition, emotions are seen as a subcategory of the more general term affect, where emotions represent the more structured or patterned elements, such as joy, fear, or anger.<sup>6</sup> Based on a multidisciplinary perspective, Morgane Kimmich conceives emotion in her PhD-thesis 'Emotional Culture?' as an umbrella term that summarizes a broad range of various mental states. These states can be termed feelings, moods, affect, passions, or sentiments.<sup>7</sup> In order to come up with a viable and feasible working definition of emotion, I adopt the aspects or characteristics of emotion that are constructive in the context of this study. Emotion is considered as an umbrella term for the more structured or patterned elements of a broad range of mental states – feelings, moods, affect, passions, or sentiments – that people can experience more-or-less consciously and anchor in language and meaning in their communications and social interactions.

Without claiming to be exhaustive, this chapter proceeds with addressing some of the essential characteristics of emotions. A breakdown of the concept of emotion can be established alongside many different lines. These lines are not always distinctive or strictly separated from one another. First, emotions are more than only hysterical or purely irrational responses. Emotions can be the result of (a combination of) spontaneous and cognitive processes. Second, emotional manifestations result from different kinds of stimuli or motivations. Emotions are actually felt and experienced, or they are expressed with a certain goal or impact in mind. The latter category, the 'strategic' emotions, are of special interest when investigating 'public' sources, such as political speeches, discussions, or news media. Professional language users, such as politicians, parliamentarians, journalists, etc., create and shape these sources. We may assume that they write or speak consciously, and with an audience in mind. Their language use has thus a high potential for containing strategic expressions of emotionality. Third, although emotions are often considered as deeply personal, emotions have the potential to 'circulate' beyond the individual level by various kinds of social interactions between people.How politicians or media interact with these 'circulations of emotions', and how 'emotional contagion' can take place will be addressed next. The chapter concludes with an epistemological reflection on the particularities of empirically investigating emotions in historical textual sources.

<sup>5</sup> Kimmich, 'Emotional Culture?', 4.

<sup>6</sup> Hall and Ross, 'Affective Politics', 848–49.

<sup>7</sup> Kimmich, 'Emotional Culture?', 4, 6.

#### **2.2 Emotions: Where They Come From**

It seems insufficient to state that emotion simply signifies a feeling that is completely detached from cognition. Human beings cannot think – process information cognitively – without emotions. Emotions play an important role in shaping and understanding human behaviour, whether this results from spontaneous or cognitive processes. Experiencing and expressing emotions in relation to contextual circumstances involves also cognitive appraisal of those circumstances. Emotions are, in the words of philosopher Martha Nussbaum, ways of viewing the world. They are important forms of knowledge.<sup>8</sup> For a long time, however, researchers in the field of neuropsychology stressed the distinction between cognition and emotion. Following the results of more recent neuropsychological research, this interpretation makes place for a perspective wherein cognition and emotions are intertwined.<sup>9</sup>As a result,itis nowincreasingly acceptedin a wide range of academic fields to consider emotions as much more than just spontaneous reactions to stimuli.<sup>10</sup> Focussing on a strict dichotomy between emotions and ratio, or between emotional and cognitive processes, is not constructive for understanding the role of emotions in history, politics, or in our daily lives.<sup>11</sup>

Philosopher Robert Solomon emphasised how he considers emotions (also) as cognitive phenomena: They consist of judgements that tell us something about our situation, and that of others.<sup>12</sup> Therefore, emotions are part of a larger associative network that also includes links to cognitive components such as memories, motives, thoughts, images, ideas, and interpretations.<sup>13</sup> As mentioned in the previous

<sup>8</sup> Martha Nussbaum, 'Emotions andWomen's Capabilities', in *Women, Culture, and Development: A Study of Human Capabilities*, ed. Martha Nussbaum and Jonathan Glover (New York: Oxford University Press, 1995), 374; Martha Nussbaum, *Upheavals of Thought: The Intelligence of Emotions* (Cambridge: Cambridge University Press, 2001), 1–22; Robert Solomon, *Not Passion's Slave: Emotions and Choice*, *Not Passion's Slave* (Oxford: Oxford University Press, 2003); Cited in Roland Bleiker and Emma Hutchison, 'Grief and the Transformation of Emotions after War', in *Emotions, Politics and War*, ed. Linda Åhäll and Thomas Gregory (New York: Routledge, 2015), 215–16.

<sup>9</sup> Damasio, *De vergissing van Descartes*; Redlawsk, *Feeling Politics*, 3; Boersema et al., 'Emoties in de Sociologie', 145–46; Frevert, 'The History of Emotions', 62.

<sup>10</sup> Bleiker and Hutchison, 'Grief and the Transformation of Emotions after War', 215.

<sup>11</sup> Swierstra, Gabriëls, and Koenis, 'Het hart op de tong', 13–16.

<sup>12</sup> Robert Solomon, *The Passions. Emotions and the Meaning of Life*(Princeton: Princeton University Press, 1977), 187; Cited in Swierstra, Gabriëls, and Koenis, 'Het hart op de tong', 15.

<sup>13</sup> These are all more or less intangible mental constructs that have already become commonplace in historical research nowadays. The next step is to integrate emotions to historical inquiry in the same way. See also Fischer, Schalk, and Hawk, 'Het ontstaan van collectieve emoties via emotionele besmetting', 165; Rosenwein, 'Problems and Methods in the History of Emotions', 24; Holt et al., *Psychology: The Science of Mind and Behaviour*, 476.

chapter, these more or less intangible mental constructs have already become commonplacein historical research a long time ago.<sup>14</sup> Moreimportantly here, however,is that this also implies that emotions are not experienced nor expressed in a vacuum. The stimulation of one of those mental constructs can trigger other elements in the associative network, depending on the nature and strength of the associative links. Emotions can be considered as equally important as cognitive mental constructs are in (understanding) human behaviour and history.<sup>15</sup>

This does not mean that every experienced emotion is equally much cognitively processed. Emotions show great variation. Some emotions, such as surprise or joy, are more spontaneous and universal than emotions such as shame, guilt, trust, or anger. These emotions are generally the result of cognitive processing of stimuli. It is also possible that the same emotion comes in different forms and takes different shapes depending on variation in context, circumstances, or experiences from the past. The anger that one feels after a hit in the face for no reason differs from the anger felt when the city council plans to build a waste dump in the backyard. The latter emotion is more constructed by its nature, involves cognitive processing, and can therefore be considered as more of a cognitive nature than the first one.<sup>16</sup> There is little cognitive processing required to fear the sound of a rumbling burglar smashing your kitchen window, whereas it needs much more cognitive processing before you experience fear of climate change. Even when an emotion like the fear for the consequences of global warming is more the result of an intuitive than an elaborately cognitive process, this fear strongly depends on a deeper understanding of the world around us. It is more than just a spontaneously automatic reaction to an external event – as fear of the stumbling burglar can be.<sup>17</sup>

The majority of the perceptible historical emotions, the ones recorded from the past, are expected to be the traces or imprints left by expressions of cognitively processed emotions. Spontaneous emotional responses are typically expressed in nonverbal or facial expressions – types of expressions that often remain unrecorded. An important carrier of past emotions, are the recordings of emotional responses and expressions in written texts or minutes of historical conversations, speeches, or discussions. Because of the cognitive processing that necessarily precedes the creation of spoken or written language, emotions manifest in language are almost inevitably of a more cognitively processed nature than spontaneous emotional reactions.What

<sup>14</sup> Fischer, Schalk, and Hawk, 'Het ontstaan van collectieve emoties via emotionele besmetting', 165; Holt et al., *Psychology: The Science of Mind and Behaviour*, 476.

<sup>15</sup> Goodwin, Jasper, and Polletta, 'Introduction: Why Emotions Matter', 9.

<sup>16</sup> Goodwin, Jasper, and Polletta, 12–13; Beunders, *Hoeveel recht heeft de emotie?*, 31.

<sup>17</sup> Nussbaum, 'Emotions and Women's Capabilities', 374; Nussbaum, *Upheavals of Thought*, 1–22; Solomon, *Not Passion's Slave*; Cited in Bleiker and Hutchison, 'Grief and the Transformation of Emotions after War', 215–16; Goodwin, Jasper, and Polletta, 'Introduction: Why Emotions Matter', 12–13.

kind of processes underlie the processing, evoking, experiencing, and – eventually – expression of those emotions?

#### **2.3 Emotions: Why They are Manifest**

It is difficult to gain access to the personal origins of emotions, even when someone is crying right in front of you. As discussed in Chapter 1, an aim of this study is to evaluate what historical imprints of the manifestations of emotions can tell us.Were emotions present at all? How were these emotions expressed? In what wordings? In which context or circumstances? Emotions can be compared, for example by looking at the different contexts and circumstances they are expressed in, or by addressing the emotions surrounding particular people, events, or topics. Historians can make inferences about why these emotions are expressed, and what kind of long-term patterns, dynamics, and shifts become apparent by studying them in their historical contexts. To help understanding the possible meanings or underlying motives of expressed emotions, this chapter pays some attention to the more theoretical basics of possible stimuli or triggers for the expression of emotions. Building on theories derived from both political science and historical research, the manifestation of emotions observable in the context of public speech, media, and political debate can be roughly distinguished in two different but partly overlapping categories. First, there are the manifestations of emotions that are an expression of emotions actually experienced by – individual – participating actors: lived emotions. The second category consists of the strategically expressed emotions: The emotions that are the result of intentional or strategic actions by the speaker or actor involved. They are expressed with a goal or impact in mind. Such strategic emotions are deliberately displayed when used in rhetorical arguments.

#### Lived Emotions

In the first place, lived emotions are typically triggered by situations that touch upon things humans care about; concerns. Emotions highlight something that requires (special) attention.<sup>18</sup> There is always a potential for evoking emotions when things are said, happen, or seem to be about to happen relevant for people's concerns. Concerns have the potential to evoke a wide variety – both in degree and character – of emotional responses.<sup>19</sup> These concerns regard not only direct personal threats, nor do they necessarily relate to tangible objects, such as family members, money, or possessions. These concerns also encompass intangible mental constructs. People

<sup>18</sup> Swierstra, Gabriëls, and Koenis, 'Het hart op de tong', 8.

<sup>19</sup> Nico Henri Frijda, *De wetten der emoties* (Amsterdam: Prometheus, 2008), 18–19, 141–42.

express emotions to signify, for example, normative attitudes or value judgements about what they consider right or wrong. Identities, values, normative beliefs, attitudes, and standards become potentially visible through studying emotions. These are all mental constructs that otherwise have the general tendency to be expressed in implicit or unconscious ways.<sup>20</sup>

Second, dispositions can also trigger emotional responses*.* Dispositions are, basically, more latent and persistent than concerns. Political scientist Andrew Ross stated that dispositions are '(…) not a felt state in itself but a particular internalised sensitivity.'<sup>21</sup> Psychologist Nico Frijda explains the difference between emotions and dispositions as follows: '(…) being frightened by a dog and being afraid of dogs is not the same thing. The distinction corresponds with that between occurrent states (…) or dispositions.' These 'occurrent states' are the emotional manifestations, responses, or associations that people express.The 'dispositions' Frijda mentioned are more durable, latent propensities that link certain stimuli to the eliciting of certain emotional responses – or a certain combination of emotional responses.<sup>22</sup> With that, a disposition is a persistent tendency to experience and/or express a specific emotion when confronted with a certain trigger.<sup>23</sup> People can have the disposition to feel an exceptional and strong anxiety towards something. This is, for example, the case with arachnophobia. It only takes the display of a spider to activate the disposition – being afraid of spiders – and corresponding emotional responses are triggered, for example expressing fear by screaming. The degree or intensity of the emotional response can vary depending on context and circumstances. Dispositions and corresponding emotions are not restricted to individual characteristics or experiences. Dispositions can also take shape in a collective way and can therefore also be considered as typical of a specific group, community, or society.<sup>24</sup>

Just as the emotions they potentially evoke, dispositions are often of a rather cognitive nature. Even if a disposition is prompted by, for example, repeated exposure to propaganda, it takes cognitive processing to respond emotionally – whether this disposition is consciously established or not.<sup>25</sup> This does not mean that dispositions are necessarily of a very rational or realistic nature. People can express certain emotional responses without being completely able to account for the existence of

<sup>20</sup> Ross, *Mixed Emotions*, 76, 91; Fischer, Schalk, and Hawk, 'Het ontstaan van collectieve emoties via emotionele besmetting', 165; Swierstra, Gabriëls, and Koenis, 'Het hart op de tong', 9.

<sup>21</sup> Hall and Ross, 'Affective Politics', 854.

<sup>22</sup> Frijda, *De wetten der emoties*, 72; Hall and Ross, 'Affective Politics', 854.

<sup>23</sup> Indrikis Krams, 'Emotional Disposition', in *Encyclopedia of Evolutionary Psychological Science*, ed. Viviana Weekes-Shackelford, Todd K. Shackelford, and Viviana A. Weekes-Shackelford (Cham: Springer International Publishing, 2016), 1–3.

<sup>24</sup> Elzbieta Halas, 'Time and Memory: A Cultural Perspective', *Trames*14, no. 4 (1 December 2010): 318; Dechesne and Bandt-Law, 'Terror in Time', 3.

<sup>25</sup> Hall and Ross, 'Affective Politics', 854–55.

underlying dispositions. Various forms of bias or preconceptions can operate in this fashion. Dispositions and corresponding emotions can be evoked and structured by interaction with social, political, and cultural contexts. Just like emotions, dispositions are not only linked to these contexts, but are often also derived from them.<sup>26</sup> Dispositions can be the result ofmemories of long-ago experienced emotions to specific objects, people, events, circumstances, symbols, etc. An example is the often still strong emotional response when the disposition of the 1988 UEFA Euro finals is triggered within football-loving Dutchmen. This makes dispositions, in the first place, an important emotion-evoking factor to take into account when investigating long-term historical developments in society or politics. Despite the fact they are durable on the long term, they are not static. Dispositions – just as people's concerns – may alter through repeated interaction with emotional manifestations. Changes and shifts can also be the result of prior and ongoing emotional experiences. By repeated exposure to subtle positive experiences and emotions, dispositions can be cultivated and subject to shifts and changes.<sup>27</sup>

#### Strategic Emotions

Manifestations of emotions are not always the direct result of felt or lived emotionality. This implies that expressed emotions can be more than just a consequence of emotions people experience themselves. Emotions can also deliberately and purposefully be displayed, evoked, shaped, modified, or transformed as they are utilized with a certain impact in mind. Displays of emotions can underpin normative standards and collective values of communities and societies. They can be part of important social and political processes, such as the construction of collective identities, processes of demarcation of moral boundaries, moral posturing, or the inclusion or exclusion of certain individuals or groups from communities.<sup>28</sup> Politicians, for example, use emotions in rhetoric to appeal to assumed electoral preferences or standards, to mobilize support, or to draw attention to a certain case, group of people, or topic.<sup>29</sup> This means that expressed emotions have the potential to not only

<sup>26</sup> Nussbaum, *Upheavals of Thought*, 6; Steven Kaindaneh and Andrew Rigby, 'Peace-Building in Sierra Leone: The Emotional Dimension', in *Politics and the Emotions: The Affective Turn in Contemporary Political Studies*, ed. Paul Hoggett and Simon Thompson (New York: Continuum, 2012), 159; Beunders, *Hoeveel recht heeft de emotie?*, 73.

<sup>27</sup> Frijda, *De wetten der emoties*, 72–78; Hall and Ross, 'Affective Politics', 854.

<sup>28</sup> The other way around, emotions themselves are constructed and structured within particular cultural, social, and political environments. See Fischer, Schalk, and Hawk, 'Het ontstaan van collectieve emoties via emotionele besmetting', 165; Ross, *Mixed Emotions*, 76, 91; Swierstra, Gabriëls, and Koenis, 'Het hart op de tong', 9.

<sup>29</sup> Imrat Verhoeven and Tamara Metze, 'Angst, woede en wantrouwen over gaswinning: hoe Groningse politici emotioneel schakelen tussen burgers en Den Haag', in *Het hart op de tong: Emo-*

influence lived emotions, but also evoke or stimulate the emotions experienced by others – and vice versa.<sup>30</sup> Conscious, deliberate, and strategical instrumentalisation of emotions is especially relevant when studying emotion in the context of the relatively demarcated arena of national politics. Professional language users, such as politicians, but also journalists, are well aware of the instrumental power of emotional language. They often use it deliberately, thoughtfully, and purposefully.<sup>31</sup> In what context do they do this? For what reasons?

#### Framing

Emotions often have a role in rhetorical approaches to defining something or someone else, for example in what is known as framing. In the context of this study, framing is considered a discursive practice. It consists of the act or process that steers attention towards persons, organisations, objects, or events by interpreting or representing them in a certain way. An example is the way in which the George Bush administration and the media 'framed' the United States' responses to the 9/11 attacks in 2001 as a 'War on Terrorism'.<sup>32</sup> According to social scientists Tamara Metze and Imrath Verhoeven, 'In this [framing] process framing actors are focused on the selection, accentuation, and characterisation of social reality from their perspective.'<sup>33</sup> With framing, people give substance to their attitudes. In the process of framing, people use words andimages to shape these 'frames'of others. In addition, emotions are often also a necessary component of framing attempts to influence other people's perspectives and perceptions. Emotions play an important role as signposts in

*ties in de politiek*, ed. René Gabriëls, Sjaak Koenis, and Tsjalling Swierstra (Amsterdam: Amsterdam University Press, 2018), 64.

<sup>30</sup> Richard Jackson, *Writing the War on Terrorism: Language, Politics and Counter-Terrorism* (Manchester: Manchester University Press, 2005), 19, 22, 24; Adelheid von Saldern, 'Emotions of Comparisons: Perceptions of European Anti-Americanism in U.S.-Magazines of the 1920s', in *Emotions in American History: An International Assessment*, ed. Jessica C. Gienow-Hecht (New York: Berghahn Books, 2010), 140–41; Spyros Kosmidis et al., 'Party Competition and Emotive Rhetoric', *Comparative Political Studies* 52, no. 6 (1 May 2019): 811–37; Traber, Schoonvelde, and Schumacher, 'Errors Have Been Made, Others Will Be Blamed'; Christian Rauh, Bart Joachim Bes, and Martijn Schoonvelde, 'Undermining, Defusing or Defending European Integration? Assessing Public Communication of European Executives in Times of EU Politicisation', *European Journal of Political Research* 59, no. 2 (2020): 397–423.

<sup>31</sup> The original creators and contributors of the used sources, such as journalists and politicians, can be regarded as professional speakers or language-users. See also Jackson, *Writing the War on Terrorism*.

<sup>32</sup> Michael Ryan, 'Framing the War against Terrorism: US Newspaper Editorials and Military Action in Afghanistan', *Gazette (Leiden, Netherlands)* 66, no. 5 (October 2004): 363–82; Jackson, *Writing the War on Terrorism*, 165–66.

<sup>33</sup> Verhoeven and Metze, 'Angst, woede en wantrouwen', 68.

directing these perceptions. Such perceptions, or, more general, attitudes, towards other people or things become apparent in the study of the emotions used to create, shape, and steer the process of framing.<sup>34</sup>

#### Moral Posturing

Emotions also play a role in defining and explicating one's own position or public stance relative to that of others. This process, moral posturing, consists of taking and expressing a public stance on something. Such a demarcation of moral territory is often done to gain political advantage. Generally, engaging in moral posturing is practiced with a particular external impact or goal in mind. Moral posturing is an obvious example of how emotions are exploited to appeal to assumed electoral preferences or standards, to mobilize support, or to draw attention to a certain case or topic. Not every emotion is however an equally obvious choice in the process of moral posturing.

Because anger is often evoked by a (perceived) failure (of others) to conform to social norms, this particular emotion is key in making a normative ring with moral posturing. Anger in particular highlights the boundaries of moral norms and standards. Manifestations of anger often originate in the recognition of (a perceived threat of) violation of (shared) social norms and values. According to Dutch sociologist Bas van Stokkom, expressions of anger show that the one expressing anger not only recognizes, but also resists evil: The manifestation of anger can assign responsibility or blame to the subject of the expressed anger. Anger can then be seen as a moral protest, in which a particular subject is held responsible.<sup>35</sup> This also goes for the emotion of sadness. Both strong negative emotions are used to highlight the urgency and importance of acting upon certain persons, organisations, or circumstances. Such a use of emotion supports drawing attention towards subjects that should be acted upon, and carries a sense of urgency and relevance. In the practice of political discussion, the (strategic) use of these emotions can therefore be connected to aiming for change, to place political opponents under suspicion,or to build trust.<sup>36</sup>

<sup>34</sup> Verhoeven and Metze, 68; Traber, Schoonvelde, and Schumacher, 'Errors Have Been Made, Others Will Be Blamed'.

<sup>35</sup> van Stokkom, 'Morele woede', 85.

<sup>36</sup> Peggy A. Thoits, 'The Sociology of Emotions', *Annual Review of Sociology* 15 (1989): 325; James Jasper, 'The Emotions of Protest: Affective and Reactive Emotions In and Around Social Movements', *Sociological Forum* 13 (1 September 1998); Markus Wagner, 'Fear and Anger in Great Britain: Blame Assignment and Emotional Reactions to the Financial Crisis', *Political Behavior* 36 (1 September 2014); Cited in Verhoeven and Duyvendak, 'Enter Emotions: Appealing to Anxiety and Anger in a Process of Municipal Amalgamation', 471; Verhoeven and Duyvendak, 471; van Stokkom, 'Morele woede', 85.

The emotion of fear is often exploited to pinpoint deliberately to what is considered important or urgent. Fear steers attention to controversial subjects, to undesired expectations for the future, or to what people fear to lose (or have lost). In striving to emphasise what is considered urgent and important, not only negative emotions play a role. Highlighting can also be achieved through the evoking of sympathy, empathy, and compassion. Studies in, for example,marketing research, show how emotions and emotional responses are mutually related to the liveliness of subjects and examples used. In general, humans are less impressed by numbers, or by the many, as they are by individual, relatable, or close examples.This sympathy bias, or sympathy effect, means that when an identifiable person's suffering is made into a cause, people generally tend to respond more compassionate (or generous, in the case of charitable giving, for example). This is also true for the so-called in-group effect that encompasses the same results but then based on a perceived reduction of social distance to a subject. It is obvious that, again, a negative emotion like sadness is often strategically used to achieve such effects.<sup>37</sup>

#### Lived versus Strategic

The distinction between strategic and lived emotions is sometimes hard to make. A clear distinction does not even necessarily have to exist.<sup>38</sup> Situations where boundaries between strategic and lived emotions become clearly visible are rather exceptional. When these boundaries are visible in a public setting, this often has negative consequences for the actors involved. An example is the former British Prime Minister Gordon Brown. He showed empathy and understanding for a female voter on camera in 2010. But when Brown entered his car, a few seconds later, he called the same woman 'a disaster' and 'a bigoted woman'. Sadly, for him and his reputation, he did this with his microphone still in operation.<sup>39</sup> Not the strategic use of emotion per se, but the fact that this emotion was (too) obviously not truly lived, was reason for controversy that followed. What the Gordon Brown incident also shows is that the situation,in which strategically expressed emotions are obviously lacking a truly felt fundament, is rather exceptional. In general, that emotions are expressed strategically and serve a political purpose does not necessarily mean they are not lived by the actor involved – and vice versa.

<sup>37</sup> K. Sudhir, Subroto Roy, and Mathew Cherian, 'Do Sympathy Biases Induce Charitable Giving? The Effects of Advertising Content', *Marketing Science* 35, no. 6 (November 2016): 8–9, 31–32.

<sup>38</sup> Beunders, *Publieke tranen*, 20–21; Rosenwein, 'Problems and Methods in the History of Emotions', 20–21.

<sup>39 &#</sup>x27;Gordon Brown "bigoted Woman" Comment Caught on Tape', *BBC News*, 28 April 2010, http:/ /news.bbc.co.uk/2/hi/8649012.stm.

It is good to keep in mind that emotional expression is not necessarily the result of lived emotions, but can also be part of a deliberate strategic act with a certain impact in mind. Whether an emotion is actually lived, strategically expressed, or both; its manifestation helps to gain insights into emotions that were substantial and meaningful in a contemporary context of the past.<sup>40</sup> Emotions, strategic or lived, shape as much as they are shaped*.* <sup>41</sup> The Gordon Brown incident shows how contemporary contexts, power dynamics, and social, cultural, and political practices are reflected in manifestations of emotions. This investigation attempts to assess how both categories of 'motivations' underlying emotional expressions, the lived and the strategic one, are connected with contemporary social, cultural, and political contexts. The fact that politicians use emotions strategically already implies that emotions are, in the first place, transferrable through different political, societal, and social interactions.They do not only circulate in the heads of individuals, or only in face-to-face social interactions. Emotions have a collective dimension. They can circulate beyond the individual.

#### **2.4 Circulations of Emotions**

It is obvious that emotions expressed by others can evoke emotional responses.This is, after all, a fundamental motive behind strategic expressions of emotions. Despite the fact that humans sometimes consider emotions deeply private and personal, the emotions people experience and express do not just pop up inside their heads. Emotions seldom come into being in social isolation and are not only the result of individual processes. Emotions can also be derived from interactive and participative social processes. Whether emotions are more cognitively processed and expressed – or not, individuals transmit them through social interaction and communication. When people participate in social interactions and activities, such as discussions, protests, political debates, or rituals, people generate shared emotions.<sup>42</sup> Emotions are triggered by memorized past emotions, or come into being in empathic interaction with emotions others express.When emotions are consciously or unconsciously transmitted within a social environment, I refer to this process as 'circulations of emotions'*.* 43

<sup>40</sup> Dr Garth S. Jowett and Victoria O'Donnell, *Propaganda and Persuasion*, 4th edition (Thousand Oaks, Calif: Sage Publications, Inc, 2005), 209.

<sup>41</sup> Ross, *Mixed Emotions*, 11.

<sup>42</sup> Ross, 16, 91, 152.

<sup>43</sup> This term is based on what political scientist Andrew Ross called 'circulations of affect'. For reasons of consistency, I use the term 'circulations of emotion' here. See Ross, 11, 21.

#### Conscious Circulations and Emotional Contagion

One of the ways in which circulations of emotions can emerge is through deliberate intentions for strategic purposes. Exposure to emotion-evoking social interactions evokes,mirrors, stirs,intensifies, emulates,or otherwise modifies emotions in and between people.The mechanisms contributing to the emergence of circulations of emotions are not always consciously initiated. So-called 'emotional contagion' is an unconscious mechanism contributing to the circulation of emotions. 'Contagion' here does not refer to the spread of emotions as a disease, butis used as metaphor for the more unconscious mechanism by which emotions possibly circulate. In contrast to deliberateintentional expression of emotion, emotional contagionis the unintentional or unconscious transfer of emotions from one to another. Contagious spread of emotions through all kinds of social activities and interactions, such as conversations or mass protests.<sup>44</sup>

The absence of direct social interaction or proximity does not mean emotions cannot be circulated. Emotions can circulate contagiously between people without them being directly exposed to each other, without direct contact. Emotional contagion then takes place in interactions involving serial participation of actors. This means that people do not have to participate in social activities and interaction simultaneously to be part of sharing related emotions. This contagion without contact takes place, for example, by watching others (from a distance, or on television) instead of experiencing emotions together. Contagious circulations of emotions can even occur when people have no personal exposure (or memory) to the original emotion-eliciting event at all. Andrew Ross describes this phenomenon as nonco-present emotional impact.<sup>45</sup> This can take place in physical locations where, for example, people interact with the past, or memories of others. Museum and memorial Auschwitz-Birkenau is an example of a place where people do not necessarily have to be together at the same time to leave the site with a similar, shared emotional response. Such places do not even have to be a physical location per se, as circulations of emotions also come into being by interactions with political speeches, literature, magazines, broadcasts, documentaries, films, newspapers, or social media. These media have the potential to contribute to the circulations of emotions just as they transmit narratives, symbols, or ideologies.<sup>46</sup>

The idea that emotions can pervade the level of social interaction between individuals remains rather vague, but every-day language use shows how familiar theintuition behind this idea actually is: Just think about common phrases such as 'love is

<sup>44</sup> Ross, 25, 54, 153; Fischer, Schalk, and Hawk, 'Het ontstaan van collectieve emoties via emotionele besmetting', 167.

<sup>45</sup> Ross, *Mixed Emotions*, 25, 30, 40.

<sup>46</sup> Ross, 25, 29–31, 40.

in the air' or 'the atmosphere in the classroom was tense'. If circulations of emotions are not specific to a narrowly defined or very specific group or context, but circulating amongst many individuals within a large (national) community, they are sometimes also referred to as societal emotions.<sup>47</sup> The collective dimension of circulations of emotions implies that emotionality can be investigated outside the psychological research environment, outside the lab. As discussed in Chapter 1, the knowability of such emotions is one of the epistemological questions central to this study. The following paragraph discusses how grasping such societal or circulating emotions can be approached from a theoretical perspective. Subsequent Chapters 3 and 4 elaborates on the more concrete steps in the methodological operationalisation of the analysis of emotions in historical texts.

#### **2.5 Empirical and Epistemological Issues**

In a face-to-face setting, when people make an appraisal of the emotions of others surrounding them, this is in first instance based on the manifestation of their emotional expressions. Such a manifestation can take shape in body language, facial expressions, intonation, tone of voice, choice of words, or a combination of those. Together, they are the signifiers that people use, consciously or not, to express themselves and, more importantly here, to take note of the emotions of others. Emotions of others are perceivable and thus knowable because, as human beings, we are familiar with the emotional states of which they are an expression, ranging from joy and happiness to anger, sadness, and disgust. Investigating circulations of emotions in a historical context gives however rise to some fundamental issues.

First, there is an epistemological issue, discussed in Chapter 1, concerning the investigation of emotions from a historical perspective. The emotions of others are difficult to know, investigate, and interpret – and even more so in retrospect. This also applies to the circulations of emotions discussed earlier. Where and how did they circulate? Where and how were these emotions manifest? This investigation is not about the workings or experiences of personal emotionality and internal developments and processes to which it is hard – if not impossible, in a historical context – to gain access.<sup>48</sup> Nevertheless, also the more collective circulations of emotions are sometimes manifest in volatile forms of communication – such as tone of voice,

<sup>47</sup> Aerts, 'Emotie in de politiek', 12–13.

<sup>48</sup> Jörg Nagler, 'The Mobilization of Emotions: Propaganda and Social Violence on the American Home Front during World War I', in *Emotions in American History: An International Assessment*, ed. Jessica C. E. Gienow-Hecht (New York: Berghahn Books, 2010), 67; Britta Waldschmidt-Nelson, 'Does Every Vote Count in America? Emotions, Elections, and the Quest for Black Political Empowerment', in *Emotions in American History: An International Assessment*, ed. Jessica C. E. Gienow-Hecht (New York: Berghahn Books, 2010), 225–26.

hand gestures,or facial expressions.These signifiers are all of a rather ephemeral nature. Historical recordings of these signifiers are exclusive to audio-visual sources. Such sources are only available for the last decades of the period under scrutiny – and if they are available for earlier years, this is in a very unevenly distributed quantity and quality.<sup>49</sup> When analysing emotions in historical context, one has to realize that a large share of – the circulation of – emotions is not recorded at all in the first place. What empirical evidence of past emotions is left for historians?

#### Emotions in Language

One of the observable manifestations of expressed emotions are the ones embedded in language. Language has the ability to produce, stimulate, legitimise, and control emotions, but is primarily a way of expressing them.<sup>50</sup> Emotional language functions as a mediator in purposely transmitting existing or desired emotions from writers, speakers, or discussants to individual or group recipients. In the behavioural sciences, it has become a commonplace view that the words people use can be considered as indicative of their mental or psychological states.<sup>51</sup> This applies to explicit references to emotions (e.g., 'I am boiling with anger!'), as well as to the more implicit manifestations of emotions in language.<sup>52</sup> In this study, however, personal or individual origins or experiences of emotions are not of primary interest. More important is that this idea is not only limited to the everyday words individual people use, or to a clinical setting. As Mark Dechesne and Bryn Bandt-Law pointed out: '(…) one can examine a large sample of words that describe events at a given point in time and make inferences about which mental constructs are active during that time.'<sup>53</sup>

Historical traces or imprints of emotional expressions in the sources are key in this study.They are meaningful and interesting in themselves, as they are reflecting contemporary societal, cultural, and political processes, practices, circumstances, and contexts.This investigation therefore predominantly pays attention to whether, and how, historical sources carry an empirically observable imprint of the (con-

<sup>49</sup> Van Lange and Futselaar, 'Vehemence and Victims'.

<sup>50</sup> Saldern, 'Emotions of Comparisons', 141.

<sup>51</sup> Yla R. Tausczik and James W. Pennebaker, 'The Psychological Meaning of Words: LIWC and Computerized Text Analysis Methods',*Journal of Language and Social Psychology* 29, no. 1 (March 2010): 26.

<sup>52</sup> C. Bröer and J.W. Duyvendak, 'Discursive Opportunities, Feeling Rules and the Rise of Protests against Aircraft Noise', *Mobilization* 14, no. 3 (2009): 340.

<sup>53</sup> Dechesne and Bandt-Law, 'Terror in Time', 3.

temporarily) circulating emotions.<sup>54</sup> The insights from the field of the behavioural sciences outlined above also imply that historians can empirically evaluate historical emotion by the presence or absence of corroborating linguistic traces of emotional manifestations. Emotions in language form the empirical baseline for making seemingly intangible phenomena, such as emotion, observable.<sup>55</sup> Andrew Ross adds that even if circulations of emotions are manifest in language in less direct ways, their credibility as object of research is warranted by the more directly observable utterances in textual sources.<sup>56</sup> Nevertheless, studying emotions through their expressions in language remains a somewhat indirect approach.

Approaching historical emotionality by focussing on emotional words in language may seem indirect and problematic, but historians always have to work with the share of phenomena or events that left traces or was somehow recorded – how implicit or indirect this sometimes might be. Approaching historically expressed emotions through the study of (emotional) language in the sources is therefore not substantively different from mainstream research practices. Examples are the approaches employedin historicalinvestigationsinto other seeminglyintangiblemental constructs such as memories, mentalities, ideology, sentiments, norms, values, identities etc. All these constructs have in common that they can be investigated by looking at their observable and recorded traces. This is equally the case for emotion.<sup>57</sup> Then there is still a risk of assigning one-sidedly emotions to (emotional expressions of) historical actors or circumstances.<sup>58</sup> Ute Frevert does not consider such a potential divergence between emotions and their expressions problematic at all. She writes that the process of identifying an emotion (in oneself) also means defining it. Identifying, defining, putting into words, and expressing emotions transforms lived or purposefully evoked emotions already by instantly giving them direction. Therefore, as Frevert states, '(…) it seems futile as well as unnecessary to draw a clear line between "the emotion itself " and its expression.' Such a distinction, she adds, is counterfactual in that it overestimates the emotion at the expense of its expression. According to Frevert, historians should consider expressions of emotions as emotion *'*tout court' and '(…) not reduce it to the external part of something much greater that unfortunately does not "come before us" so that historians can never reach it.'<sup>59</sup>

<sup>54</sup> Susan Matt, 'Recovering the Invisible: Methods for the Historical Study of the Emotions', in *Doing Emotions History*, ed. Susan Matt and Peter Stearns, History of Emotions (Urbana: University of Illinois Press, 2014), 42.

<sup>55</sup> Saldern, 'Emotions of Comparisons', 141; Hall and Ross, 'Affective Politics', 862.

<sup>56</sup> Ross, *Mixed Emotions*, 42.

<sup>57</sup> Ross, 62; Hall and Ross, 'Affective Politics', 862.

<sup>58</sup> Verhoeven and Duyvendak, 'Enter Emotions: Appealing to Anxiety and Anger in a Process of Municipal Amalgamation'; Verhoeven and Metze, 'Angst, woede en wantrouwen', 68.

<sup>59</sup> Frevert, 'The History of Emotions', 57.

#### Emotion Words

Historians can approach imprints of historical emotional expressions in written (or transcribed) sources, as language serves as one of the vehicles for past emotions. But where to start? As a historian specialized in the historical study of emotions, Frevert again leads the way: 'Evidently, historians cannot conduct their research without relying on written sources, and so their work has to start with emotion words'.<sup>60</sup> This implies that expressions of emotions in language, manifest by the use of emotional words, allow for the empirical observation of the imprints of circulations of emotions in all kinds of historical communication, debates, and discursive media: personal letters, diaries, newspapers, written speeches, and (verbatim minutes of) political discussions. Emotion words are instrumental features that are used in this study as proxies for emotion – something that is, for reasons mentioned earlier, difficult to measure.<sup>61</sup> The epistemological issue of how to know or grasp the traces of historical emotionality (see also Chapter 1) overlaps with the empirical problem of identifying expressions of emotions in texts. The next step in identifying emotional expressions in historical sources, therefore, is to define and demarcate what words count as emotional language and what words don't.

As obvious as it might sound, emotional vocabulary or emotion words give substance to the expression of emotions in language.<sup>62</sup> Emotion words are words that either by their semantic, conceptual, or connotative content express, define, name, evoke, or contribute to the manifestation of emotionality in written texts or spoken language.<sup>63</sup> Chapter 3 'Materials and Data' addresses the particular resources used for defining emotional language use – emotion lexicons – in more detail. For now, this chapter concludes with addressing the fundamental structuring that is at the core of the emotion lexicons used in this investigation. How to define, identify, classify, and categorize (different kinds of) emotional expressions?

#### Basic Emotion Categories

In real-life, face-to-face situations, identifying emotions or distinguishing between different kinds of emotions is already complex, implicit, multi-faceted, and often heavily dependent on people's personal appraisals, preconceptions, and interpretations. Different emotions can be made perceivable in a research context by build-

<sup>60</sup> Frevert, 56.

<sup>61</sup> Graham Alexander Sack, 'NEH Panel Simulating Plot: Towards a Generative Model of Narrative Structure', in *Papers from the AAAI Fall Symposium*, 2011, 10; Cited in Franco Moretti, '"Operationalizing": Or, the Function of Measurement in Modern Literary Theory', Pamphlets of the StanfordLiteraryLab, 60, no. 1 (March 2014): 4.

<sup>62</sup> Ross, *Mixed Emotions*, 63.

<sup>63</sup> Omondi, 'Dholuo Emotional Language: An Overview', 89–90.

ing on a structured basic categorization of emotions. The idea that all emotions are derived from a set of innate and universal emotions underlies this categorisation. It is an attempt to transform the dense and ambiguous concept of emotion into researchable and perceivable objects. It is important to realize that emotions, although they have been understood as universal for a long time, are context- and time-dependent, or culture-dependent. This makes all the forms of categorization of emotion also dependent on the time, place, context and culture wherein they have been established.Using basic categories of emotions is a simplification of a complex and ambiguous reality. Investigating emotion forces researchers into epistemological sacrifices – as isolating emotions from their complex and ambiguous characteristics is. Methodological adjustments are however required in making something as complex and multi-faceted as emotion investigable historically.<sup>64</sup>

Categorizations of emotions originate in the presumption that most emotions are primarily variations or an expression of different degrees of intensity of one or several basic emotions. What desperation, for example, is to the basic emotion of sadness, is delight to joy.<sup>65</sup> On the one hand, basic emotions and their varieties are nested together in very complex social emotions, such as pride, envy, or guilt.<sup>66</sup> On the other hand, basic emotion categories are not as rough as they seem at first sight. Apparent basic emotions, such as sadness, can take multifaceted forms. Sadness may consist of love for a person on the one hand, and of mourning about the loss of this particular person on the other.<sup>67</sup> Emotions can thus be a complex interplay of emotional categories, compiled out of various basic emotions.

Although different theoretical approaches to categorizing (basic) emotions vary in form, demarcation, and implication, the basic principle is often similar. Neuroscientist Antonio Damasio indicated a categorization with at least three subtypes. He considered the first, basic universal emotions, as innate (e.g., joy, anger, sadness, fear, disgust).The second subtype encompasses the subtle universal but not innate emotions, which are variations of the basic feelings: euphoria and ecstasy as variations of happiness, melancholy and wistfulness as variations of sadness, etc. The third subtype consists of the background emotions, related to mood.<sup>68</sup> Psychologist and pioneering emotion researcher Paul Ekman argued that the first subtype consists of six basic emotions: joy, sadness, anger, fear, disgust, and

<sup>64</sup> Friedrich Ungerer, 'Emotions and Emotional Language in English and German News Stories', in *The Language of Emotions*, ed. Susanne Niemeier and René Dirven (Amsterdam: John Benjamins Publishing Company, 1997), 319; Frevert, 'The History of Emotions', 53–55.

<sup>65</sup> Ungerer, 'Emotions and Emotional Language', 319.

<sup>66</sup> Ross, *Mixed Emotions*, 21.

<sup>67</sup> Ross, 19.

<sup>68</sup> In his publication, Damasio uses the term feeling instead of emotion. For the sake of consistency, I prefer to use emotion instead of feeling here. Damasio, *De vergissing van Descartes*; Ad Foolen, 'The Expressive Function of Language: Towards a Cognitive Semantic Approach', in

surprise. In addition, psychologist Robert Plutchik proposed that there are eight basic emotions. These include the six basic emotions of Ekman and the additional emotions trust and anticipation. He structured his categorization in 'Plutchik's Wheel of Emotions'.<sup>69</sup> When I write about basic emotions from now on, I refer to the categorisation of the second ring of Plutchik's Wheel of Emotions (see Figure 1).

Plutchik's Wheel of Emotions is used for establishing the structure and categorisation in the creation of so-called emotion lexicons. An emotion lexicon is, in this study, used as the predominant resource for reference and identification of emotional language use. Emotion lexicons are lists of words that make emotional language empirically identifiable, quantifiable, measurable, and comparable.<sup>70</sup> The question is how these (basic) emotion categories can function as an anchor to define and distinguish the hundreds of different words loaded with various (emotional) meanings and associations.<sup>71</sup> In this investigation, a generic emotion lexicon is used. It is based on the categorizations of basic emotions to identify, observe, interpret, and understand the hundreds of emotional words in historical sources. The more technical and methodological details of the application of such a lexicon, the NRC Word-Emotion Association Lexicon, will be further discussed in Chapter 3. This next chapter also explains how the analysis of emotions in historical textual sources is operationalised by using lists of emotional words that are based on, amongst other things, Plutchik's Wheel of Emotions.

*The Language of Emotions: Conceptualization, Expression, and Theoretical Foundation*, ed. Susanne Niemeier and René Dirven (Amsterdam: John Benjamins, 1997), 17.

<sup>69</sup> Paul Ekman, 'An Argument for Basic Emotions', *Cognition and Emotion* 6, no. 3–4 (1 May 1992): 169–200; Saif M. Mohammad and Peter D. Turney, 'Crowdsourcing a Word–Emotion Association Lexicon', *Computational Intelligence* 29, no. 3 (2013): 1.

<sup>70</sup> Mohammad, 'From Once Upon a Time to Happily Ever After: Tracking Emotions in Novels and Fairy Tales'; Mohammad and Turney, 'Crowdsourcing a Word–Emotion Association Lexicon'.

<sup>71</sup> Foolen, 'The Expressive Function of Language', 17; Mohammad and Turney, 'Crowdsourcing a Word–Emotion Association Lexicon'; Holt et al., *Psychology: The Science of Mind and Behaviour*, 475.

*Figure 1: Plutchik's Wheel of Emotions. Source: Wikimedia Commons, 2011.*

## **3. Materials and Data | Digitised Sources and a Lexicon**

The computational analysis of emotion in historical texts undertaken in this study relies on an external digital resource: the NRC Word-Emotion Association Lexicon (EmoLex). This emotion lexicon is the pre-eminent external (re)source used in this investigation and its use is explained in this chapter. The lexicon is used in a computer-assisted analysis of digitised historical sources, primarily the *Handelingen der Staten-Generaal* dataset, which is comprised of parliamentary proceedings. Historians already work with digitised collections, digital search tools, or explorative digital user interfaces. Indeed, websites that give access to digitised source collections with user-friendly (search) interfaces are already popular with those conducting research.<sup>1</sup> Many historians incorporate tools and websites based on these materials in their analytical processes, although they often do not even consider themselves to be 'working computationally' or 'digitally'.<sup>2</sup> Historians today seem open to experimentation with the possibilities that the mass digitisation of historical sources has to offer. With a single keyword, such as 'war', 'Holocaust', or 'genocide', hundreds of possibly relevant historical sources now appear on our screens. However, it is important to remember that the digitisation processes underlying these datasets not only has much to offer, but also alters our relationship and interactions with the sources.<sup>3</sup> In what follows, the underlying digitised historical sources and machinereadable data are therefore addressed in more detail.<sup>4</sup> They are an inevitable part of the computer-assisted methodology discussed in Chapter 4.

<sup>1 &#</sup>x27;Delpher − Kranten, Boeken & Tijdschriften'; 'Impresso Website'; Romein et al., 'State of the Field', 304–6.

<sup>2</sup> Zaagsma, 'On Digital History', 17.

<sup>3</sup> Joris van Zundert and Tara Andrews, 'Qu'est-Ce Qu'un Texte Numérique? A New Rationale for the Digital Representation of Text', *Digital Scholarship in the Humanities* 32, no. suppl\_2 (1 December 2017): ii78–88.

<sup>4</sup> Machine-readable means that a computer can process the information (characters, words, etc.) contained in the documents.

#### **3.1 Emotion Lexicons as 'Outside Method'**

Chapter 2 showed that, in the computer-assisted methodology of this study, a resource for the identification of emotional language use is needed. A computer cannot identify emotions by itself; it needs some point of reference. Researchers have to make explicit what it is that they wish the computer to 'mine' – what has to be identified and analysed in the sources. In the context of emotion mining texts, the identification or retrieval of emotional language is always based on some existing body of knowledge. This knowledge can take the form of expert knowledge of the (type of) texts under scrutiny, of emotions, judgements of human annotators or other contributors, or be based on external resources. Several options are possible for the identification and measurement of emotional words in language. In the context of a study like this one, there are roughly two ways to approach this: an 'inside' and an 'outside' approach.However, a fundamental difference between both approaches exists in the ways in which emotionality is defined and identified, how it relates to the sources under scrutiny, and the waysin which the assessment of retrieving emotions from texts is evaluated.The difference lies in whether the judgement of emotionality (and its evaluation) is internal and specific to the sources under scrutiny, or external and based on an existing, generic emotion lexicon.

#### The Inside Approach

A first approach would be an internal method that is domain- or source-specific. Researchers assign (fragments of) the texts under scrutiny to so-called human annotators. Generally, the annotators label text (fragments) to identify manifestations of emotional language. In an ideal situation, and if annotators agree (i.e., there is inter-annotator-agreement), their labelled fragments can be used to 'train' the computer to learn the distinctive features of the emotional text concerned, resulting in a source-specific list of words or (text) characteristics that can be used for the identification, measurement, and/or analysis of emotions in (other) parts of the text collection under scrutiny.<sup>5</sup> Many variations to this process are possible. There are, for example, hybrid formats, where outcomes generated with (existing) generic emotion lexicons are enriched with, compared to, or evaluated with the results of evalu-

<sup>5</sup> Examples of 'mining' historical emotions based on human annotations are Ryan Heuser, Franco Moretti, and Erik Steiner, 'The Emotions of London', *Pamphlets of the Stanford Literary Lab*, October 2016; Ryan Heuser, Mark Algee-Hewitt, and Annalise Lockhart, 'Mapping the Emotions of London in Fiction, 1700–1900: A Crowdsourcing Experiment', in *Literary Mapping in the Digital Age*, ed. David Cooper, Christopher Donaldson, and Patricia Murrieta-Flores(London: Routledge, 2016), 24–46; For an elaborate description of the annotation and evaluation process, see Heuser, Algee-Hewitt, and Lockhart, 31–34.

ations by human annotators.<sup>6</sup> Other variations are also possible, such as using distributional semantics to 'train' a domain-specific lexicon that is based on a generic resource and the particular characteristics of the dataset under scrutiny.<sup>7</sup>

Most internal approaches require blinding, since annotators should not be aware of the results of the generic approach, nor have a strong bias with regard to the original sources or specific subject matters under analysis. In the context of this study, however, this creates a problem, as the subject matter is ethically charged, relatively recent, highly emotive, and well-known, certainly amongst those with sufficient expertise to evaluate historical Dutch texts. An evaluation of emotion-mining output using human annotators therefore would contravene (or even threaten) the aims of this investigation. As discussed in Chapter 1,many people hold strong views regarding World War II. Indeed, it is difficult, if not impossible to grow up in the Netherlands without developing strong ideas about the war and its aftermath. If (only) human annotators are used, it is inevitable that these views or personal biases will be introduced in the process. The use of human annotators, whether experts in the domain or not, might lead to circular reasoning: Integrating them in the computational workflow (re-)introduces, first, personal preconceptions; and second, is likely to reproduce commonplace views in the evaluative and analytical process. Clearly, this is problematic, as these received wisdoms are, at least in part, a constituent of the very thing under investigation. Yet the reverse is also true: People without sufficient (basic) knowledge are expected to be unreliable assessors of historical parliamentary debates from, for example, the 1950s. There is a high risk of misinterpretation of the (un)emotionality of historical Dutch texts from the earlier years of the 1945–1989 datasets. Put simply, no one can be relied upon to judge the rather ceremonial and formal texts from the second half of the twentieth century and, at the same time, not be biased with regard to the subject matter of the discussions they are being asked to assess.<sup>8</sup>

<sup>6</sup> Marchetti, Sprugnoli, and Tonelli, 'Sentiment Analysis for the Humanities'.

<sup>7</sup> Rheault et al., 'Measuring Emotion in Parliamentary Debates with Automated Textual Analysis'; Sven Buechel, Johannes Hellrich, and Udo Hahn, 'Feelings from the Past: Adapting Affective Lexicons for Historical Emotion Analysis', *Proceedings of the Workshop on Language Technology Resources and Tools for Digital Humanities*, December 2016, 8.

<sup>8</sup> Van Lange and Futselaar, 'Vehemence and Victims'.

#### The Outside Approach

A second approach is a more generic, or 'outside' method. The lexicographical approach chosen here builds on the semantics (the meaning of words) and relies on word lists that have been created independently from the project at hand and the data used.<sup>9</sup> By using an existing emotion lexicon that has generated good outcomes in other investigations, a rough yet useful measuring tool can be created.<sup>10</sup> Emotion lexicons are structured lists of words that make emotional language empirically identifiable, retrievable, quantifiable, measurable, and comparable. Such lexicons are often organised into positive and negative polarity, or their structure builds on the categorisation of words into basic emotions (see Chapter 2). Since it is considered problematic to use human annotators in this specific case, generic lexicons offer a good, if second best, option.

The computational analysis of historical sources in this investigation relies on the NRC Word-Emotion Association Lexicon (EmoLex), created in 2010 by computational linguists Saif Mohammad and Peter Turney. The EmoLex consists of hundreds of words that are, either through denotation (meaning) or connotation, associated with different emotions. Due to the large number of words included, the EmoLex is considered a high-coverage lexicon.This linguistic resource originates in a large-scale crowdsourcing project using human annotators. The annotators were asked how they perceive the general emotion association of individual words. Mohammad and Turney based the creation process on the assessment of words, rather than sentences or longer texts.This lexicon is selected because not only polarity (positive and negative) is taken into account, but also different diverging categories of basic emotions (joy, anger, sadness, etc.) are represented. For this investigation, six basic emotions are used (see Table 1). The EmoLex has been translated by machine translation from English into multiple other languages.<sup>11</sup> Whilst this process may have introduced some errors, several studies have shown that generally most of the word-emotion associations of the EmoLex remain relatively consistent across differ-

<sup>9</sup> Morin and Acerbi, 'Birth of the Cool'; Acerbi et al., 'The Expression of Emotions in 20th Century Books'; Marchetti, Sprugnoli, and Tonelli, 'Sentiment Analysis for the Humanities'.

<sup>10</sup> Mohammad, 'From Once Upon a Time to Happily Ever After: Tracking Emotions in Novels and Fairy Tales'; Mohammad, 'From Once upon a Time to Happily Ever After'; Saif M. Mohammad, 'Ten Years of the NRC Word-Emotion Association Lexicon', Medium, 17 May 2020, https://medium.com/@nlpscholar/ten-years-of-the-nrc-word-emotion-associationlexicon-eaa47a8dd03e.

<sup>11</sup> Mohammad and Turney, 'Crowdsourcing a Word–Emotion Association Lexicon'; Saif M Mohammad, 'Practical and Ethical Considerations in the Effective Use of Emotion and Sentiment Lexicons', 2020, 1–6, https://arxiv.org/abs/2011.03492v2.

ent languages.<sup>12</sup> Although the EmoLex is based on human annotations, at the core lies the idea that the lexicon has to be applicable to different kinds of texts, domains, topics, and applications.The NRC EmoLex is generic in the sense that it is not based on a single type of source or dataset, nor is it developed for a particular research application.<sup>13</sup> This has important advantages.


*Table 1: Fragment of the NRC Word-Emotion Association Lexicon (EmoLex).*

#### Bias and the NRC EmoLex

As a resource for defining whether language is emotional or not, generic emotion lexicons are not only independent of the text sources on which they are used; they are also independent of the context and subject matter of this study, and myself as a researcher.They thus constitute a so-called 'outside'method. Indeed the EmoLex is entirely external to the field of historical research as a whole and is free from personal preconceptions about domain-specific historical knowledge. As the study of ethically charged topics from the relatively recent past is particularly vulnerable to the introduction of personal biases, preconceptions, commonplace views, or received wisdoms, this is an important advantage in this study. The generic lexicon offers an additional, outside perspective, as it does not rely on personal judgements and interpretations by the investigator involved. In addition, the application of emotion lexicons as an auxiliary method not only helps to make historical claims more verifiable, but also has the potential to make (personal) researcher preconceptions, ideas,

<sup>12</sup> Mohammad, 'Practical and Ethical Considerations in the Effective Use of Emotion and Sentiment Lexicons', 3; Peter Boot, Hanna Zijlstra, and Rinie Geenen, 'The Dutch Translation of the Linguistic Inquiry and Word Count (LIWC) 2007 Dictionary', *Dutch Journal of Applied Linguistics* 6, no. 1 (19 October 2017): 65–76.

<sup>13</sup> Mohammad and Turney, 'Crowdsourcing a Word–Emotion Association Lexicon'; For an overview of ten years of using the NRC EmoLex in research and other applications, see Mohammad, 'Ten Years of the NRC Word-Emotion Association Lexicon'.

and interpretations more explicit and visible in the analytical process.<sup>14</sup> A valuable advantage of generic emotion lexicons is that the judegment of what is emotional and what is not, is established independently of the particular sources and subject matters under examination.<sup>15</sup>

However, the fact that, in the context of this study, the externally created and validated EmoLex is preferred over an internal method does not mean it is without disadvantages. Emotion lexicons are, whether in a laboratory setting or in large crowdsourcing projects, also created by humans. Consequently, they are inevitably based on some form of judgement.The decisions and choices made by their creators and contributors necessarily influence the resulting outputs. There is always a risk that unknown or invisible biases were introduced in this process. Thus, resources such as emotion lexicons – an inevitable part of the computer-assisted methodology of this study – are never entirely free of bias. Lexicons are always to some extent space- and time-specific to the context in which they are created. How such biases operate and influence results is not entirely knowable to the users of EmoLex. Saif Mohammad, one of the creators of EmoLex, is aware of these limitations and therefore provides users with some guidelines in order to minimise the risk of introducing unknown or invisible biases.<sup>16</sup> These guidelines are taken into consideration and evoke evaluation, curation, and reflection on applying the EmoLex in a particular (genre-specific) context. The implications of these processes underlie the methodological choices discussed in Chapter 4. Importantly, possible biases in the EmoLex are not necessarily problematic in the context of this investigation, as they are external to the sources used, consistent throughout the entire analysis, and expected to be barely relevant to the specific historical themes under scrutiny.

#### Language Change and Other Issues

A more problematic aspect of emotion lexicons in historical research is that they have all been created relatively recently, whereas the languages they represent change over time. For example, spelling, as well as the meaning of words themselves, change and vary. In the Netherlands, for example, there were major updates in official spelling in 1934, 1947, and 1954. However, it was many years before the new official Dutch spelling became widely accepted. The issue of spelling variation – the most obvious change in the case of the Dutch language – can be largely

<sup>14</sup> Van Lange and Futselaar, 'Debating Evil', 143.

<sup>15</sup> Boot, Zijlstra, and Geenen, 'The Dutch Translation of the Linguistic Inquiry and Word Count', 66; Andrew Piper, 'Fictionality', *Journal of Cultural Analytics* 1, no. 1 (20 December 2016), cited in; Floor Naber and Peter Boot, 'Exploring the Features of Naturalist Prose Using LIWC in Nederlab', *Journal of Dutch Literature* 10, no. 1 (2019): 54.

<sup>16</sup> Mohammad, 'Practical and Ethical Considerations in the Effective Use of Emotion and Sentiment Lexicons'.

mitigated by the application of modern language technology in the pre-processing of the sources.<sup>17</sup> The potential problem of historical semantic change, however, remains. Semantic change is, at least in theory, a serious and seemingly unresolvable drawback of using generic emotion lexicons in diachronic historical research.

Words can fall into disuse, new words appear, or words gain new or additional meanings over time. Take, for example, the word 'trauma' (spelt the same in both Dutch and English). In the second half of the twentieth century, the meaning of this word grew from its initial meaning, 'physical wound', to also encompass psychological damage.<sup>18</sup> This has obvious implications for diachronic historical research based on historical texts and word use.<sup>19</sup> It is, for example, possible that emotion words, and the emotions to which they refer or are associated with, change. Such words do not necessarily have strict, fixed meanings.<sup>20</sup> Even – or perhaps, particularly – when using human annotators (the 'inside' approach), this problem arises. The modernday origins and hence limitations of generic lexicons are neither solved, nor evaluated by using modern-day human annotators to evaluate results on historical text materials. Thus, one must ask, how large an issue is this for historical or diachronic emotion mining?

During a multidisciplinary workshop at the DH Benelux conference in 2018, this issue was discussed with experts (on the Dutch language) from the fields of computer linguistics, literary studies, digital humanities, and the history of emotions. Although consensus existed regarding the existence and magnitude of the problem, the participants present in 2018 agreed on the usability of modern-day generic emotion lexicons toinvestigate textin the context of contemporary Dutch history. For the investigation of Dutch, using generic lexicons on texts dating back to 1850 or later was not considered problematic.<sup>21</sup> This consensus is perhaps unsurprising, given the fact that the semantic meaning of the vast majority of words used in every-day language remains stable.

<sup>17</sup> This problem is discussed in Chapter 4 and in M. Reynaert et al., 'PICCL: Philosophical Integrator of Computational and Corpus Libraries', in *Proceedings of CLARIN Annual Conference 2015* (CLARIN Annual Conference, Wrocław, 2015), 75–79.

<sup>18</sup> Withuis, *Erkenning*, 9–10.

<sup>19</sup> Rockwell and Sinclair, *Hermeneutica*, 132; Vaclav Brezina, *Statistics in Corpus Linguistics: A Practical Guide* (Cambridge University Press, 2018), 219–20.

<sup>20</sup> Frevert, 'The History of Emotions', 57.

<sup>21</sup> This issue was discussed during the 'LIWC in the Digital Humanities' workshop at the DHBenelux 2018 conference on 6 June 2018 with, amongst others, Peter Boot, Inger Leemans, Erika Kuijpers, Hennie Brugman, Isa Maks, and the author. See also Peter Boot et al., 'Workshop LIWC in the Digital Humanities' (Workshop, DH Benelux 2018, Amsterdam, 6 June 2018), http://2018.dhbenelux.org/workshops/#liwc.

As the linguist Vaclav Brezina points out,many studies focus on what changes in language, rather than on (the possibility of) stability.<sup>22</sup> Whilst one can come up with examples of words such as 'awful' or 'gay' in English that have drastically changed their meaning, the number of words that have completely changed meaning or shifted from one emotion (e.g., 'joy') to another (e.g., 'anger') in the second half of the twentieth century is negligibly small. Although they certainly exist, they are the exception rather than the rule.<sup>23</sup> Brezina compares language (change) with a large brick house: The many levels of such a structure undergo constant renovation. Some bricks need replacement now and then, but many need to remain part of the construction to keep the building standing.<sup>24</sup> In addition, a study in 2016 by William Hamilton et al. showed that, in English historical texts over an extended period of 150 years, only around 5% of non-neutral words switched polarity (from positive to negative, or vice versa).<sup>25</sup> Additionally, I argue, semantic change is not necessarily a major issue in the analysis of relatively recent historical Dutch text collections. Especially when encompassing the relatively short period of approximately fifty years, as is the case in this study, it is unlikely that many words changed so much that they moved from one emotion category to another, or gained a completely opposite meaning or polarity.<sup>26</sup>

Nonetheless, the use of generic lexicons to identify, analyse, and evaluate manifestations of emotions undeniably remains a rather crude method. Hamilton et al., for example, highlighted how lexicons scores varied between text collections derived from different communities.<sup>27</sup> Due to its rather rough and generic categorisation, the more complex or layered emotions such as shame, compassion, or vulnerability are not explicitly taken into consideration in the emotion-mining process. In addition, for the emotions that are incorporated in the lexicon, it is not always a very reliable method for identifying a certain emotion on the level of a single term or sentence. This lack of precision, however, is less important, as I am predominantly interested in general trends and proportionality in the investigation of large-scale digitised historical sources.<sup>28</sup>

First, there is the advantage of bulk. In this investigation, more than 300 million words are analysed. The Swiss linguist Gerold Schneider (amongst others) points

<sup>22</sup> Brezina, *Statistics in Corpus Linguistics*, 220.

<sup>23</sup> Boot et al., 'Workshop LIWC in the Digital Humanities'; Morin and Acerbi, 'Birth of the Cool', 1669.

<sup>24</sup> Brezina, *Statistics in Corpus Linguistics*, 220.

<sup>25</sup> William L. Hamilton et al., 'Inducing Domain-Specific Sentiment Lexicons from Unlabeled Corpora', *ArXiv:1606.02820 [Cs]*, 9 June 2016.

<sup>26</sup> See also Boot et al., 'Workshop LIWC in the Digital Humanities'.

<sup>27</sup> Hamilton et al., 'Inducing Domain-Specific Sentiment Lexicons from Unlabeled Corpora'.

<sup>28</sup> Wiedemann, *Text Mining for Qualitative Data Analysis*, 130–31; Mohammad, 'Practical and Ethical Considerations in the Effective Use of Emotion and Sentiment Lexicons', 4.

specifically to the advantage of large quantities. Whilst an individual word may not be a very powerful 'discriminator', as it can be susceptible to language change, many of those 'weak discriminators' together can help achieve high accuracy.<sup>29</sup> The fact that emotion lexicons contain many terms makes them less sensitive to incidental errors or the historical changes caused by individual words. Moreover, when a lexicon is applied to larger text collections, this method creates an opportunity for a systematic comparison between texts of different origin or from different periods. Even if the lexicon is not always precisely accurate,its value is in the score of one document or text unit relative to all others. Emotion lexicons are a means to compare relative emotionality, rather than to provide an absolute measure. Or, in the words of Mohammad: 'Comparative analysis is your friend.'<sup>30</sup>

The limitations of EmoLex do not overshadow its advantages.The more 'distant' perspective, for example, is less reliant on the personal preconceptions, ideas, and interpretations of the historian. In addition, the identification of emotional language in a large body of text is transparent, traceable, and replicable using this approach because the emotion lexicon used is open source. In this investigation, lexicon-based emotion mining is used to provide an additional perspective. Using the historian's macroscope still holds the possibility to go back and forth, from distant to close reading, to zoom in and carefully read the sources. In the following section, I look at the specific historical sources used in this investigation.

#### **3.2 Historical Sources as Digital Datasets**

#### The Dutch Parliament

Chapter 1 already outlined the role of Dutch parliamentary debates as the main object of study in this investigation. These parliamentary proceedings are the minutes of the meetings of the Dutch parliament, officially known as the States General (*Staten-Generaal*). Since 1814, the Dutch parliament has operated as a bicameral body in a parliamentary democracy. Universal suffrage and proportional representation characterise the Dutch democratic system. Initially, the Senate,or*EersteKamer* (First Chamber), had 50 seats, but in 1956, the number of seats expanded to 75. Similarly, the House of Representatives, or *Tweede Kamer* (Second Chamber), now fulfils its role

<sup>29</sup> Morin and Acerbi, 'Birth of the Cool', 1669; Gerold Schneider, 'Changes in Society and Language: Charting Poverty', in *Corpora and the Changing Society: Studies in the Evolution of English*, ed. Paula Rautionaho, Arja Nurmi, and Juhani Klemola (Amsterdam: John Benjamins Publishing Company, 2020), 37.

<sup>30</sup> Drucker et al., 'Distant Reading and Cultural Analytics'; Mohammad, 'Practical and Ethical Considerations in the Effective Use of Emotion and Sentiment Lexicons', 4.

with 150 members/seats, whereas prior to 1956, the number was 100.The Dutch parliament and the government are housed within the area known as the *Binnenhof* in the city centre of The Hague. Members of the House of Representatives are elected every four years, and the formation of the government is based on the election results.

The tasks (and power) of the Dutch parliament encompass not only checking and deciding upon legislation, but also regulating the activities of the government and government ministers. As a legislative body, parliamentarians discuss and review government proposals for new or modified national legislation. After governmental bills pass the House of Representatives, the Senate reconsiders the legislation. Both the government and the House of Representatives have the right to propose or amend legislation, whereas the Senate can only discuss, approve, or disapprove legislative proposals. Members of parliament (MPs) question responsible government ministers, question bills or existing policies, or call for action. With an interpellation request, for example, members can demand an explanation from government politicians. To request an interpellation debate, more than thirty other MPs have to support the request.<sup>31</sup> Especially the right to initiate legislation was, as will be discussed later in Chapter 8, crucial to the establishment of more elaborate or tailormade legislation for victims of World War II in the 1970s.

#### Parliamentary Debate

Although the minutes of parliamentary discussions are a rather comprehensive source when writing an institutional history of the Dutch parliament, the perspectives they offer on the government and the parliament's engagement with the consequences of World War II in Dutch post-war society are limited. Both houses of the Dutch parliament form a demarcated environment. As a historical source, the minutes of debates in parliament do not provide a reliable sample for farreaching generalisations about (national) cultures or entire societies. Nevertheless, the historical proceedings of both houses of the Dutch parliament offer a valuable, albeit limited mirror on the politics and broader society of the past. Whilst the parliamentary arena is demarcated, it is not isolated. Parliament can best be seen as a microcosm that operates in the context of a mutual interaction with wider society. The parliamentary agenda is not only influenced by (and influences) the government, but also by issues in society. MPs in the House of Representatives are elected by the people. As Dutch parliamentarians are (and often present themselves literally as) representatives of their constituencies, they put the concerns of those

<sup>31 &#</sup>x27;The Right to Question the Cabinet', Text, Website of the Dutch House of Representatives, 16 January 2018, https://www.houseofrepresentatives.nl/how-parliament-works/democracynetherlands/duties-and-rights/right-question-cabinet.

constituencies (often explicitly) on the parliamentary agenda.<sup>32</sup> On the one hand, they react to pressing social and societal issues, whilst, on the other, parliamentary debates can encourage or boost current affairs, social unrest, pressing issues, etc.<sup>33</sup> As various controversies later in this book will demonstrate, the parliament also interacts with various media, including print media, and sometimes media attention itself leads to parliamentary discussion.

#### Minutes of Debates

This investigation uses the minutes of parliamentary debates. In Dutch, this collection of texts containing the verbatim minutes of the parliamentary debates is known as the*Handelingender Staten-Generaal*(hereafter referred to simply as the*Handelingen).* Nevertheless, not all words spoken in parliament are incorporated in the *Handelingen*, as insults and inadmissible language use are excluded from the official minutes.<sup>34</sup> In addition, any such source is inevitably limited to the text that is actually spoken. This may seem obvious, but it is important to note that a great deal of nuance that was available to those present at the time remains invisible, such as the sly smile, the red face, the flying saliva, etc.The role of the *Handelingen* in this investigation is primarily that of a case study – one that is eminently fit for purpose, as both houses of parliament consistently discussed legislation that was aimed at perpetrators, or at alleviating the suffering of survivors ofWorldWar II.The national government's treatment of and engagement with these people was not only debated, but also shaped, established, voted for, and constantly (re-)evaluated in the Dutch parliament between 1945 and 1989. These debates included developments, deviations, and various (re-)considerations of parliamentary processes.The proceedings of such debates as sources therefore often differ from the legislation and policy (documents) to which they might lead. The latter offers only the result; the former also the varied developments and diverging viewpoints eventually leading to the establishment of new legislation – or not. In addition, parliamentary debates are polyvocal, as they represent the different and often contradictory voices present in the discussions. Nevertheless, the individual contributions to the debates or differences in speaking style are not taken into account in the distant reading aspects of the computer-assisted methodology (in Chapters 5, 6, and 8). For the sake of comparison over time and between different themes and debates, in computational analysis, parliamentary debate is mostly treated as a single analytical entity, whereas individual voices are addressed by the close-reading analysis that is discussed in Chapters 7 and 9.

<sup>32</sup> Hoetink, *Macht der gewoonte*, 25, 135, 446.

<sup>33</sup> Beunders, *Hoeveel recht heeft de emotie?*, 90–91.

<sup>34</sup> Peter Bootsma and Carla Hoetink, *Over lijken: ontoelaatbaar taalgebruik in de Tweede Kamer* (Amsterdam: Boom, 2006).

#### The Handelingen Dataset

As a digitised historical dataset, the verbatim minutes of both houses of the Dutch parliament offer an exceptional opportunity for historical research into long-term developments. The *Handelingen* are not only open source available, but also offer a structured, demarcated, and,most importantly, continuous source for historical research.This continuity relates to the nature of the source itself, its availability, quality, and quantity. As a historical source, the dataset also stands out in its completeness: The verbatim minutes of all discussions in both houses of Dutch parliament are incorporated. Decades of debates are transcribed in the original records that are kept in the *Handelingenkamer* in The Hague.<sup>35</sup> The Royal Library (Koninklijke Blibliotheek, KB) of the Netherlands digitised the *Handelingen* and made them available online as a digital and machine-readable dataset.<sup>36</sup> For the process of turning scans of physical documents into machine-readable text, an automated process known as optical character recognition (OCR) was used.<sup>37</sup> Although this process is vulnerable to the introduction of flaws and errors caused by imperfections in the original sources, or by limitations of the OCR software, the *Handelingen* dataset is of a particularly high standard.<sup>38</sup> As part of the Political Mashup project (and later also the Digging into Linked Parliamentary Data project, DiLiPaD), the historical digitised proceedings (1814–2013) were structured and enriched.<sup>39</sup> Information about, for example, the dates of debates, the names of the MPs speaking, or the parties to which they belong, was added in structured metadata schemes.<sup>40</sup> Constituent parts of the debate (individual speeches, announcements by the chair, etc.), titles of debates, and topics of discussion were also included.<sup>41</sup> The resulting dataset allows for de-

<sup>35 &#</sup>x27;Handelingenkamer', Text, Website of the Dutch House of Representatives, 22 August 2018, https://www.tweedekamer.nl/contact\_en\_bezoek/de\_tweede\_kamer\_in\_beeld/handelingenkamer.

<sup>36</sup> Piersma et al., 'War in Parliament'; Koninklijke Bibliotheek, 'Staten-Generaal Digitaal', accessed 6 May 2021, https://www.kb.nl/bronnen-zoekwijzers/dataservices-en-apis/statengeneraal-digitaal.

<sup>37</sup> Romein et al., 'State of the Field', 294–95.

<sup>38</sup> Wevers, 'Consuming America', 80–81.

<sup>39</sup> Maarten Marx and Maarten de Rijke, 'Political Mashup Position Paper', 2008, 2; Marx et al., 'Thematic collection'; Bart de Goede et al., 'PoliticalMashup Ngramviewer', in *Research and Advanced Technology for Digital Libraries*, ed. Trond Aalberg et al., Lecture Notes in Computer Science (Berlin, Heidelberg: Springer, 2013), 446–49.

<sup>40</sup> Richard Gartner, 'A Metadata Infrastructure for the Analysis of Parliamentary Proceedings', in *2014 IEEE International Conference on Big Data (Big Data)*, 2014, 47–50.

<sup>41</sup> Alex Olieman et al., 'Good Applications for Crummy Entity Linkers? The Case of Corpus Selection in Digital Humanities', in *Proceedings of the 13th International Conference on Semantic Systems*, vol. abs/1708.01162 (Semantics 2017, New York: Association for Computing Machinery, 2017), 84.

tailed, sustained scrutiny (close reading), as well as for the use of search tools and text mining. The additional metadata facilitates the retrieval and selection of relevant subsets of debates in the approximately 2.5 million pages of parliamentary discussions.<sup>42</sup> Of course, this raises the question of what makes a debate relevant and how it might be retrieved.

#### Subsets

Before any relevant debate can be retrieved and selected from the collection, some criteria are needed. As discussed in Chapter 1,it is hard to establish a concrete working definition of the post-war parliamentary engagement with the consequences of the German occupation during World War II in the Netherlands. However, such a definition is required if we are to formulate selection criteria for retrieving relevant parliamentary discussions. To establish a somewhat systematic approach, I have identified three different diverging subsets of discussions. These are used as a first step in the identification and organisation of relevant debates from the *Handelingen* dataset. Each subset deals with a concrete topic and a group of people. In this investigation, parliamentary discussions related to three different groups of 'war survivors' are identified in the collection. Every group consists of people who either survived the war, or were descendants of those who had not. Hence, the selected relevant parliamentary debates deal with the real and lived consequences for people who were, at some point and in some way in their lives, affected by the German occupation of the Netherlands. This could be as a war victim, as a former member of the anti-Nazi resistance, but also as collaborator, perpetrator, or war criminal.These people all became the subject of war-related, post-war parliamentary engagement, be it in debates, national legislation, or political controversy. The first subset, 'perpetrators and collaboration', deals with the discussions about perpetrators,(foreign) war criminals, former Nazi collaborators,and public figures who caused controversy because of their (alleged) suspicious war pasts.This subset is used to explore the validity of the research instruments used. The subset 'extraordinary government employees'is discussed in Chapters 6 and 7 (case study 1), whilst the 'war victims' subset is the subject of an in-depth analysis in Chapters 8 and 9 (case study 2). A next step is to find,identify, and retrieve as many relevant parliamentary discussions as possible within the *Handelingen* dataset to give substance to these subsets.

<sup>42</sup> Piersma et al., 'War in Parliament'; Olieman et al., 'Good Applications for Crummy Entity Linkers? The Case of Corpus Selection in Digital Humanities'; Alex Olieman et al., 'Riches of the Poor: Using Crummy Entity Linkers for Interactive Search in Digital Humanities', *ArXiv:1708.01162 [Cs]*, 3 August 2017, http://arxiv.org/abs/1708.01162.

#### Retrieval and Selection

Obviously, the desired subsets of debates are not already identified in the metadata of the *Handelingen* dataset. Nevertheless, the Political Mashup metadata does present a viable starting point. Before each historical parliamentary debate took place, the chair initially formulated the topic of discussion for that (part of the) day. Each document in the original minutes of the Dutch parliament has been provided with a topic-description of the scheduled discussion. In the Political Mashup project, these debate topics were structured and made accessible in an organised way. The topic descriptions are an integral part of the metadata of each enriched XML-file in the dataset.<sup>43</sup> Based on these XML-files, metadata can be retrieved (or 'parsed') from the document collection. A table with a unique document-identifier, the date of the meeting, and the (original) topic descriptions were derived from the collection using XML-parsing packages in the R-programming environment. The XML-files also allowed for the retrieval of the number of speeches (constituent parts) per document.<sup>44</sup> Table 2 displays some examples of the metadata available in the Political Mashup collection.

The complete version of this table, containing all parliamentary debates from the 1945–1989 period, was used for the next step: the selection of relevant subsets. The first step in the selection procedure relied on the following selection criteria:


<sup>43</sup> For more information about the use of XML in digitised or digital historical collections, see Romein et al., 'State of the Field', 300.

<sup>44</sup> This table was automatically generated from the XML-files using an XML parser in the R-programming environment. This software will be discussed later on in this chapter. For the XMLparsers in R, see Hadley Wickham, Jim Hester, and Jeroen Ooms, *Xml2: Parse XML*, version 1.3.2, 2020, https://CRAN.R-project.org/package=xml2; CRAN Team, Duncan Temple Lang, and Tomas Kalibera, *XML: Tools for Parsing and Generating XML Within R and S-Plus*, version 3.99-0.6, 2021, https://CRAN.R-project.org/package=XML.


*Table 2: Handelingen metadata for some example documents. The data is parsed from the Political Mashup XMLs and used for the manual retrieval of relevant debates.*

**<sup>a</sup>** The topic descriptions contain errors as they are automatically retrieved from scans of the original documents by using OCR in the Political Mashup project. See also Maarten Marx, 'The University of Amsterdam's PoliticalMashup', On History, 23 April 2014, https://blog.history.ac.u k/2014/04/the-university-of-amsterdams-politicalmashup/; Piersma et al., 'War in Parliament'; Martin Reynaert, 'On OCR Ground Truths and OCR Post-Correction Gold Standards, Tools and Formats', in *Proceedings of the First International Conference on Digital Access to Textual Cultural Heritage − DATeCH '14* (DATeCH '14, Madrid, Spain: ACM Press, 2014), 164.

Furthermore, the remaining parliamentary debates were manually annotated based on careful consideration of their topic descriptions. This step included much more (personal) interpretation and judgement than those that preceded it.The process was performed according to the following criteria:


The metadata was not always complete, even in the enriched Political Mashup version of the dataset. In very rare cases, metadata on the topic of discussion was entirely missing. Therefore, in addition to the systematic manual annotation of the parliamentary debates based on the criteria outlined above, secondary literature dealing with the history of specific (types of) legislative schemes in the Netherlands was consulted to check for (references to) parliamentary debates missing in the subsets.<sup>45</sup> This led to the identification of some additional relevant debates that were added to the subsets as well. As a final check in the retrieval procedure, a fulltext keyword search was performed. Based on the debates previously retrieved via secondary literature, expected debates on known relevant political issues, and the NIOD World War II-thesaurus, a checklist of potentially relevant search terms was created (see also Supplement 1.2).<sup>46</sup> This checklist was then used to keyword search the *Handelingen* database using the online Political Mashup search interface. Unfortunately, the online parliamentary debate search engine is no longer operational since early 2020.<sup>47</sup> However, this keyword-search method only identified a small number of additional debates, as the majority of relevant debates identified with keyword search had already been retrieved in previous steps.<sup>48</sup>

#### Controlled Vocabulary

Based on the criteria and procedures outlined above, a subset of war-related parliamentary debates was selected from the Political Mashup collection. The next step in organising these source materials encompassed the identification and annotation of debates relevant to each of the three different groups of war survivors by using a

<sup>45</sup> Romijn, *Snel, streng en rechtvaardig*; Bossenbroek, *De Meelstreep*; Withuis, *Erkenning*; Fühner, *Nachspiel*; Ludi, 'Who Is a Nazi Victim?'; Piersma, *De Drie van Breda*; Tames, *Besmette jeugd*; Piersma, *Bevochten recht*; Touwen-Bouwsma, *Op zoek naar grenzen*; Withuis and Mooij, 'From Totalitarianism to Trauma'; Van der Heijden, *Dat nooit meer*; Grevers, *Van landverraders tot goede vaderlanders*; Tames, *Doorn in het vlees*; Tames, 'Ashamed About the Past: The Case of Nazi Collaborators and Their Families in Postwar Dutch Society'; Tames and Romijn, 'Transnational Identities of Dutch Nazi-Collaborators and Their Struggle for Integration into the National Community'.

<sup>46</sup> NIOD Institute for War, Holocaust, and Genocide Studies, 'WO2-thesaurus', NIOD-website, 5 September 2017, https://www.niod.nl/nl/wo2-thesaurus; Annelies van Nispen and Lizzy Jongma, 'Holocaust and World War Two Linked Open Data Developments in the Netherlands', *Umanistica Digitale* 3, no. 4 (1 March 2019).

<sup>47</sup> A 2019 snapshot of the website can be consulted via the Wayback Machine of the Internet Archive. Unfortunately, the server of the Political Mashup website has been offline since February 2020.

<sup>48</sup> The success of the task and the relative ease with which the relevant debates were identified, led me to decide not to train or further experiment with a (lexicon-based) automatic classifier, or an approach based on machine learning.

so-called controlled vocabulary.<sup>49</sup> This controlled vocabulary was based, initially, on the concepts referring to each one of the three subsets: 'perpetrators and collaboration', 'extraordinary government employees', and 'war victims'. These terms served as an organised arrangement to structure the documents retrieved and selected (see also Supplement 1.1). This approach was taken a step further by organising the documents on a more detailed level. Subsequently, and based on the variation in the contents, topics, and legislative schemes discussed, the individual documents were organised into different thematic clusters of debates within the 'perpetrators and collaboration' subset and 'war victims' subset.<sup>50</sup> The 'extraordinary government employees' subset contained only very little variation in topics, as most debates related to only two different legislative schemes. I therefore consider the organisation into thematic clusters superfluous for that particular subset. For every individual debate identified as being relevant to this study, related tags from the controlled vocabulary were connected to the Political Mashup metadata and the machine-readable transcripts of the actual speeches and debates.The transcribed texts were automatically derived from the Political Mashup XML-files using XML-parsers in R-programming.<sup>51</sup>

Table 3 shows the subsets and examples of thematic debate clusters and topics. Following the selection criteria outlined above, four debates were found that fitted both the 'extraordinary government employees' subset and the 'war victims' subset. These debates dealt with the so-called *Algemeen Verdrag* ('General Treaty') between the governments of West Germany and the Netherlands (see also Chapter 7). As they are taken into account in the analysis of both subsets and in both case studies, they are included in the counts of the 'Number of documents' columns of both subsets in Table 3. The entire selection procedure led to three different subsets, containing 180 unique documents in total. Supplement 1.1 displays the complete results of using the controlled vocabulary: It represents the organised and annotated subsets as they were used in further pre-processing and analysis. Now a substantive subset of debates has been selected and retrieved, a next step was to prepare not only the texts of the subsets, but also the entire 1945–1989 collection of parliamentary proceedings for computational text-mining analysis. This process of preparing a text collection for text mining is also known as pre-processing and is discussed in more detail in Chapter 4.

<sup>49</sup> I am grateful to Douglas Lambert, who introduced me to the concept of 'controlled vocabularies'. See also Douglas Lambert and Michael Frisch, 'Digital Curation through Information Cartography: A Commentary on Oral History in the Digital Age from a Content Management Point of View', *The Oral History Review* 40, no. 1 (1 January 2013): 135–53.

<sup>50</sup> An overview of relevant legislative schemes is provided in Chapter 6 (Extraordinary government employees) and Chapter 8 (War victims).

<sup>51</sup> Wickham, Hester, and Ooms, *Xml2*, 2; CRAN Team, Lang, and Kalibera, *XML*.



#### Complementary Text Collections

The computer-assisted methodology of this study makes applying computational methods for text mining to additional datasets relatively easy. In creating a more general baseline of emotional language use, there is no good reason to rely on the rather limited perspective of the *Handelingen* dataset alone. I briefly discuss two additional text collections that are therefore also used in this study. They are complementary to the parliamentary proceedings. In the investigation of 'emotionalisation' in Chapter 5, a sample of historical newspaper articles from two different national Dutch newspapers (*De Volkskrant* and *De Telegraaf* ) and the *Troonreden,* the annual speeches by the Dutch queen, are also analysed. All these sources addressed a broad public and thus imply a wider audience – and potential broad societal impact. In addition, they were created and established in direct interaction with society.<sup>52</sup> Yet these complementary sources are selected primarily because they are historically stable, rather than for the historical reality they represent. All historical text collections used are consistently available in the 1945–1989 period, and relatively constant through time in terms of their origin, nature, size, and form.The text collections act as samples of language use with a certain emotionality.The selection of these different collections is based on the presumption that if the received wisdom of 'emotionalisation' is true, it is likely to be visible in these different text collections.

#### Newspaper Articles

The newspapers *De Volkskrant* and *De Telegraaf* are used in this study. They were selected because these papers were nationally distributed across the Netherlands.The two newspapers are also consistently available under the same title and in a digital format for (almost) the entire period under scrutiny.<sup>53</sup> *De Volkskrant* initially had a left-of-centre, Catholic identity, whereas *De Telegraaf* is not consistently aligned to any particular pillar in the politically segregated (or 'pillarised') Dutch society.<sup>54</sup> However, non-alignment to one of the established pillars does not mean political neutrality, as *De Telegraaf* can be considered as adopting a highly partisan stance on political issues. The Royal Library of the Netherlands digitised millions of historical Dutch newspaper pages, making them available and readily accessible via an online interface on the Delpher website.<sup>55</sup> Large-scale newspaper collections from the seventeenth to the twenty-first century have been enriched and made searchable via this online portal.

The various functions of the Delpher website were made possible primarily by transforming scans of the original newspaper pages into machine-readable texts with OCR.<sup>56</sup> The quality of the outputs of this process is not always perfect. The age and the quality of the original material is an important determinant of the performance of the OCR process. Older newspapers generally generate more errors in the

<sup>52</sup> Rosenwein, Problems and Methods in the History of Emotions, 12.

<sup>53</sup> This consistency is preferred in the investigation of long-term developments, such as the presumed 'emotionalisation' from the 1970s onwards. The national Dutch newspaper *NRC Handelsblad,* for example, is less suitable here, as this newspaper was established after a fusion of two existing newspapers in 1970.

<sup>54</sup> Frank van Vree, *De metamorfose van een dagblad. Een journalistieke geschiedenis van De Volkskrant* (Amsterdam: Meulenhoff, 1996); 'Archief De Volkskrant', Katholiek Documentatie Centrum, 2013, https://www.ru.nl/kdc/bladeren/archieven-thema/subpagina-archieven-thema /cultuur-vrije-tijd-ontspanning/archieven\_van/archieven/volkskrant/; Wevers, 'Consuming America', 77.

<sup>55 &#</sup>x27;Delpher − Kranten, Boeken & Tijdschriften'.

<sup>56</sup> M.H. Beals and Emily Bell, 'Delpher', *The Atlas of Digitised Newspapers and Metadata: Reports from Oceanic Exchanges*, 2020.

automated recognition process than more recent publications. Taking into account the OCR outputs of the entire Delpher collection, the quality of the OCR is relatively good for the post-WorldWar II period.<sup>57</sup> The enriched Delpher newspaper collection not only contains metadata on the newspaper title, issue, and publishing date, but also on the type of text element (article, advertisement, etc.) in the collection.<sup>58</sup> In addition to the ability to read and analyse the newspapers on the Delpher website, the files themselves are also downloadable (or 'harvestable') from the KB's computer server, including the OCR-generated transcripts and metadata.<sup>59</sup> This allows one to select articles and exclude, for example, advertisements, and for the creation of a subset of a random sample of 1000 articles per month (between 1945 and 1989) for both newspaper titles.<sup>60</sup> More details of the newspaper dataset can be found in Table 4 under 'Facts and Figures'.

#### The Queen's Speeches

At the start of each parliamentary year, the queen (latterly, king) of the Netherlands gives a ceremonial speech. In this annual address,in Dutch known as *Troonrede*('Oration of the Throne'), the national government's policy ambitions for the coming year are presented. However, successive queens did not write these speeches themselves. Rather, coalition government officials define the plans, topics, statements, and formulations of the speeches. The speech is often both reflective and anticipatory in nature. Major events, politics, and developments of the past year are discussed, just as the ambitions, plans, and policy goals for the year to come. Gerard Breeman et al. point out that the language of the speeches is, in the first instance, quite symbolic.<sup>61</sup> The hopes and fears for the country's future, as they are expressed in the speeches, are usuallymostly addressedin a symbolicmanner.However,according to Breeman, the speeches are not merely ceremonial. The queen's speeches also contain the anticipation of concrete public policy proposals and announce new legislative plans. They contain reflection on actual events, real national problems, and anticipate the government's budget and the future of Dutch politics and society.<sup>62</sup> Digitised ver-

<sup>57</sup> Kenter et al., 'Ad Hoc Monitoring of Vocabulary Shifts over Time', 1196; Wevers, 'Consuming America', 80–81.

<sup>58 &#</sup>x27;Handleiding Delpher', accessed 26 April 2021, https://www.delpher.nl/nl/platform/pages/he lpitems?nid=373; 'Delpher − Kranten, Boeken & Tijdschriften'; Wevers, 'Consuming America', 84–85; Beals and Bell, 'Delpher'.

<sup>59 &#</sup>x27;Workshop "Mining Delpher Data" at DHBenelux', KB Lab, 2017, https://lab.kb.nl/agenda/wo rkshop-mining-delpher-data-dhbenelux..

<sup>60</sup> I am grateful to Lizzy Jongma and Melvin Wevers for helping out with harvesting the newspapers.

<sup>61</sup> Gerard Breeman et al., 'Political Attention in a Coalition System: Analysing Queen's Speeches in the Netherlands 1945–2007', *Acta Politica* 44, no. 1 (16 March 2009): 1, 4.

<sup>62</sup> Breeman et al., 1, 4.

sions of the annual queen's speeches are formally incorporated in the *Handelingen der Staten-Generaal* archival collection*,* but are here treated as separate dataset. Furthermore, the *Troonreden* are in the public domain and as digitised dataset available on various (online) platforms.<sup>63</sup> The dataset with machine-readable speeches from the 1945–1989 period used in this investigation has been collected, structured, and made available by Gerard Breeman.<sup>64</sup>

#### Facts and Figures

Table 4 summarises the information on the contents of the *Handelingen* dataset and the additional datasets used in Chapter 5.


*Table 4: Facts and figures of the datasets used in this investigation.*

<sup>63</sup> See for example 'Troonredes.nl: alle troonredes van 1814 tot nu op één plek', accessed 26 April 2021, https://troonredes.nl/; 'Troonrede in Wikisource', Wikisource, accessed 26 April 2021, h ttps://nl.wikisource.org/w/index.php?search=troonrede&ns0=1&ns102=1; 'Troonrede', Parlement.com, accessed 26 April 2021, https://www.parlement.com/id/vh8lnhrptxwq/troonrede

<sup>64</sup> Breeman et al., 'Political Attention in a Coalition System'.

<sup>65</sup> Due to the wartime collaboration with the Nazi occupier, *De Telegraaf* was not allowed to publish between 1945 and 1949.

<sup>66</sup> This number contains only the complete entries in the *Handelingen* set, with at least a document-identifier, a timestamp/date, and text. Nevertheless, not all of the documents counted have a topic description.

<sup>67</sup> Applies only to the newspaper datasets.


The digitised historical datasets outlined above offer historians the possibility to incorporate computational methods, and with it the potential to analyse historical questions from a macroscopic perspective.<sup>70</sup> How this is done, is the topic of Chapter 4.

<sup>68</sup> This number can be lower than the number of documents in a set or sample because some documents contain less than 250 words.

<sup>69</sup> These numbers are the approximate word counts of the documents as they are used in the analysis. These word counts are based on pre-processed texts. Stop words (e.g., 'the', 'a') are already removed from the texts and not counted. This pre-processing will be discussed in Chapter 4.

<sup>70</sup> Van Lange and Futselaar, 'Debating Evil'; Romein et al., 'State of the Field', 304–6.

## **4. Methods and Operationalisation | A Computerassisted Approach to the Analysis of Digitised Historical Texts**

'Operationalizing means building a bridge from concepts to measurement (…)', wrote the Italian literary historian Franco Moretti.<sup>1</sup> Operationalisation is the primary aim of this chapter. What operations are needed to assess the questions, issues, and concepts presented in Chapters 1 and 2? The application of computational analysis on historical sources not only requires a different relationship with those sources, but also demands the pre-processing of historical materials to make them fit for computational analysis. Chapter 3 described the primary source materials used in this study, namely digitised historical text collections. The study of emotion in these sources, achieved via quantification, is operationalised by using an external digital resource (the NRC Word-Emotion Association Lexicon) in this study.

This lexicon, however, cannot just be 'thrown' at a dataset. To curate, manage, modify, pre-process, and analyse data with the computer, software and a systematic working plan are required. This working plan, the different steps in analysing the materials, and the statistical evaluation and visualisation of results, form a workflow. In what follows, the different elements of this workflow, including the software packages used in each step, are explained. After pre-processing the materials and data, the mixed methods on which this study relies are discussed. They give substance to a hybrid historical research process, combining common historiographic practices with computational methods.<sup>2</sup> Many considerations, choices, and decisions underlie the computer-assisted research methodology. This chapter not only discusses the sources and resources used, but also the choices and decisions made regarding software, statistics, and the visualisations of results. The application of relatively simple, off-the-shelf software and resources are preferred over 'fancy' tools, interfaces, or indeed, computational or statistical glitz. This leads

<sup>1</sup> Moretti, '"Operationalizing"', 3.

<sup>2</sup> Zaagsma, 'On Digital History', 14–19.

to an operational workflow for the computer-assisted part of the analytical process. The outcomes and results of both distant reading using computational methods and close reading are discussed in the chapters that follow.

### **4.1 A Pre-processing Workflow**

### Preparing the *Handelingen*

The historical text collections used in this investigation have already undergone a transformation, from a physical archival collection to a set of digital scans, via OCR, to an enriched and structured digital dataset. As discussed above, the quality of OCR outputs is not always perfect. Therefore, the Philosophical Integrator of Computational and Corpus Libraries (PICCL) is used here to further prepare the *Handelingen* for computational analysis. The use of this application can be considered as a process leading to the greater unification of the contents of the textual data sets. PICCL helps to improve the recognition or identification of emotion lexicon words in the *Handelingen* dataset*.* Text-Induced Corpus Clean-up (TICCL) is part of the PICCL pipeline and allows for the automatic post-correction of OCR-related errors, typographical mistakes, historical spelling variation, and spelling errors that are present in the texts.<sup>3</sup> In an additional process known as lemmatisation, words are brought back to their linguistic basic form. *Waren* ('were') and *was* ('was'), for example, are transformed into *zijn* ('to be'). This also applies to, for example, plural forms and to the normalisation of historical spelling variation.<sup>4</sup> Outputs were stored as structured FoLiA-XML files.<sup>5</sup> The prepared textual data now needs to be loaded into an environment that allows not only for further pre-processing, but also for its analysis, visualisation, and statistical evaluation.

### Flexibility, R, and Packages

Today, an increasing number of tailored applications, tools, online and virtual labs, and media suites are available to the historical research community. User-friendly interfaces and tools are often incorporated in the websites that are the result of such

<sup>3</sup> I am grateful to Martin Reynaert for his assistance in pre-processing the data. Reynaert et al., 'PICCL'.

<sup>4</sup> Grimmer and Stewart, 'Text as Data', 272; Reynaert et al., 'PICCL'; Reynaert, 'On OCR Ground Truths and OCR Post-Correction Gold Standards, Tools and Formats'; Martin Reynaert et al., 'Text-Induced Corpus Clean-up: Current State-of-the-Art' (Workshop Processing of Historical Corpora, Donostia, Spain, 2018), 26.

<sup>5 &#</sup>x27;FoLiA:: Format for Linguistic Annotation', accessed 27 April 2021, http://proycon.github.io/folia/.

endeavours.<sup>6</sup> This did not lead to widely accepted best practices or standard tools for text mining in historical research. I consider this neither necessary, nor desirable. Preselected and predefined computational analysis techniques in (online) interfaces make it difficult to subject different (kinds of) datasets (or alternative [versions of] data) to computational analysis. In addition, standardised or pre-defined operations are not always fit for particular research purposes.<sup>7</sup> In addition, it remains unclear how durable the accessibility and usability of such web interfaces and (online) analytical tools will remain in the longer term. A better alternative is offered by using off-the-shelf open-source software. This allows for a flexible and versatile application of various software packages and analytical computational techniques. To bring together digital datasets, emotion lexicons, and techniques for curation,management, modification, pre-processing, analysis, evaluation, and visualisation, this study relies on R-programming.

Ris,first and foremost, versatile and flexible.R can work withmany types of data and offers the possibility to use various complementary software packages in combination. This means that varied tasks and operations can be performed within a single environment. R-programming has the additional benefit of being accessible, open-source, widely used (also for text mining), and therefore well documented.<sup>8</sup> In this study, R is used in combination with the RStudio integrated development environment (IDE). RStudio allows for a clear, structured, and accessible way of writing, editing, and executing R-code. It also functions as an environment for plotting and debugging, as well as offering data and workspace management.<sup>9</sup> One of the main advantages of using R is the overwhelming availability of different kinds of packages. Currently, more than 10.000 packages are available in R's official repository CRAN, and new ones are added almost every day. Packages contain collections of functions, operations, and datasets developed by fellow R-users and developers within the community. Packages usually come with helpful instruction manuals. They increase the functionality, usability, and power of R as a programming environment, and offer accessible, structured, and standardised ways of working with specific types of data, tables, statistics, or visualisations.<sup>10</sup>

<sup>6</sup> Graham, Milligan, and Weingart, *Exploring Big Historical Data,* 158.

<sup>7</sup> Grimmer and Stewart, 'Text as Data', 270–71.

<sup>8</sup> See for example Jockers, *Text Analysis with R for Students of Literature*; Julia Silge and David Robinson, *Text Mining with R: A Tidy Approach* (Sebastopol, CA: O'Reilly Media, 2017).

<sup>9</sup> RStudio Team, *RStudio: Integrated Development Environment for R* (Boston, MA, 2018), http://w ww.rstudio.com; R Core Team, *R: A Language and Environmentfor Statistical Computing* (Vienna: R Foundation for Statistical Computing, 2019), https://www.R-project.org.

<sup>10 &#</sup>x27;R Packages: A Beginner's Guide', DataCamp Community, 25 March 2019, https://www.datacamp.com/community/tutorials/r-packages-guide.

#### Loading and Pre-processing

To use R for data analysis, the resources and source materials first need to be loaded into the RStudio working space. The formats or file types of the different datasets vary, but they share the characteristic that they are all machine-readable and, to some extent, organised. The structured data files allow loading the documents and their structured elements, such as metadata tags, from local folders into the RStudio workspace. All files of each dataset, whether they are stored as a .tsv, .csv, .xlsx, or .xml file type, are loaded into RStudio and turned into a so-called data frame; a basic table that R can work with and upon which it can run its functions.<sup>11</sup> This step is also applied to the .csv-files that were derived from the NRC EmoLex.<sup>12</sup>


#### Cleaning the Texts

Machine-readable datasets, even when loaded successfully in RStudio as a data frame, are not readily suitable for text analysis although the computer can now recognise and distinguish between different text elements, characters, symbols, numbers, and words. In the case of the *Handelingen,* different data from the XML-

<sup>11 &#</sup>x27;Data.Frame Function', RDocumentation, accessed 27 April 2021, https://www.rdocumentation.org/packages/base/versions/3.6.2/topics/data.frame.

<sup>12</sup> The EmoLex was downloaded first to remove blocklisted words and check the outputs before using them in analysis (see also Chapter 3 and 4). This step is, however, not always necessary as multiple R packages allow for directly importing the EmoLex into R. See for example Matthew Jockers, *Syuzhet: Extracts Sentiment and Sentiment-Derived Plot Arcs from Text*, version 1.0.6, 2020, https://CRAN.R-project.org/package=syuzhet.

<sup>13</sup> Wickham, Hester, and Ooms, *Xml2*.

<sup>14</sup> R Core Team, *R: The R Utils Package*, accessed 27 April 2021, https://stat.ethz.ch/R-manual/R -devel/library/utils/html/00Index.html; Hadley Wickham et al., *Readr: Read Rectangular Text Data*, version 1.4.0, 2020, https://CRAN.R-project.org/package=readr.

<sup>15</sup> Hadley Wickham and Jennifer Bryan, *Readxl: Read Excel Files*, version 1.3.1, 2019, https://CRA N.R-project.org/package=readxl.

<sup>16</sup> Lionel Henry, Hadley Wickham, and RStudio, *Purrr: Functional Programming Tools*, version 0.3.4, 2020, https://CRAN.R-project.org/package=purrr.

files are recognised and structured in data frames with rows and columns. Nevertheless, further pre-processing of the contents is still required. The next steps in pre-processing are largely standardised procedures of unification and cleaning of the texts for text mining. First, this encompasses a further unification of lexical units (words) in the texts. Texts are also 'cleaned': Unwanted or unnecessary elements in text-mining analysis are removed from the texts. The full complexity and variability of language is unnecessary to analyse texts effectively, and not all lexical units are meaningful. So-called stop words (e.g., articles, some verbs, or pronouns), for example, do not always convey any relevant meaning. These words are 'semantically poor'. By their numbers, stop words are the words that texts have most in common with one another, and therefore need to be removed for effective text mining.<sup>17</sup> Another, very practical reason to remove them is that stop words occur very frequently and thus use a lot of computer-processing capacity. To make textual datasets suitable for text mining, their removal helps by reducing working memory usage.<sup>18</sup> This step is often inevitable when working on office-grade computers.


To illustrate the effects of the pre-processing steps detailed above, Table 5 displays the examples from the*Handelingen* dataset usedin the previous chapterin their new form.

<sup>17</sup> Nan Z. Da, 'The Computational Case against Computational Literary Studies', *Critical Inquiry* 45, no. 3 (1 March 2019): 623.

<sup>18</sup> Andrew Goldstone, 'Producing a Segmented Corpus of Novels', 19 April 2015, 1; Wiedemann, *Text Mining for Qualitative Data Analysis*, 27.

<sup>19</sup> Ingo Feinerer and Kurt Hornik, *Tm: Text Mining Package*, 2018, https://CRAN.R-project.org/pa ckage=tm.

<sup>20</sup> Kenneth Benoit et al., 'Quanteda: An R Package for the Quantitative Analysis of Textual Data', *Journal of Open Source Software* 3, no. 30 (2018): 774.


*Table 5: Fragment of pre-processed text in the Handelingen dataset.*

With the exception of stop word removal, the abovementioned steps are applied to the emotion lexicon as well. Just as the text files, the EmoLex is now also loaded as a data frame into the RStudio-working space. There is, however, another process of cleaning applied before the actual analysis: removing so-called 'blocklisted words' from the standard NRC EmoLex.

#### Cleaning the Lexicon

A generic resource is seldom a perfect fit for a particular investigation or dataset. There are, also in the EmoLex, (modern-day) words that are not present in the historical text collections analysed. This is not necessarily problematic in this study, as those words would also never be identified in the historical texts. More problematic, however, are words in EmoLex that are evidently not emotionally charged in the context of the sources used. This can be due to a mismatch in sense or meaning, inappropriate bias in the lexicons, or because of context-specificity of certain words.<sup>21</sup> Therefore, the creators of the EmoLex also recommend experimentation with the

<sup>21</sup> Mohammad, 'Practical and Ethical Considerations in the Effective Use of Emotion and Sentiment Lexicons', 2–3.

application of the lexicon to particular sources, and manual examination of emotion-mining results and frequently occurring words in the documents and lexicon outputs.<sup>22</sup> Because parliamentary proceedings are the most important source used here, this lexicon-cleaning procedure was applied to the *Handelingen* dataset in particular. Based on preliminary experimentation, which was performed prior to the actual analysis of emotion-mining outputs, the lexicon was slightly adjusted. Very frequent lexicon words (top 100) in each session year between 1945 and 1989 of the *Handelingen* that evidently have no emotional charge in the context of parliamentary discussion were put on a 'block list'. The criteria used for blocklisting words are:


The blocklisted lexicon words that are removed from the EmoLex as part of the preprocessing in this investigation can be found in Supplement 1.3.With unwanted and unnecessary elements removed from both lexicon and text, some more dramatic changes are now applied to the text collections.

#### Chunking the Texts

A next step in pre-processing is dividing each document from the historical text collections into segments that are (approximately) of equal size. These segments, often referred to as (text) chunks, help to overcome the possible effect of variations in text size within the text collections.The more words in a text, the more themes or emotions can run throughout the text.Therefore,in order to generate eventual measurements that are better suited for comparison, it is useful to divide larger texts

<sup>22</sup> Mohammad, 3.

into chunks containing an equal number of words. In that sense, chunking can be understood as a normalisation procedure. To calculate the emotion-word score for particular articles, speeches, or debates consisting of multiple chunks later, mean scores for all the chunks belonging to a given document are calculated. How emotion scores are established will be discussed in part 4.2 of this chapter. For now, it is important to realise that chunking makes the comparison of scores of different texts (of varying lengths) methodologically more reliable.<sup>23</sup> Chunking has an additional advantage: Creating more data points allows for a more detailed tracking of the trajectory of diachronic developments in language use.<sup>24</sup> An important decision here is the actual chunk size employed.

Defining an ideal text-segmentation parameter is difficult. Different tasks demand different chunk-sizes. It is evident that applying text-mining techniques to the analysis of several poems, a single novel, an entire library, or decades of parliamentary debate all require a different level of detail or scale. Trial and experimentation augmented by domain expertise seems to offer the best guide when setting segmentation parameters.<sup>25</sup> This investigation uses a 250-word limit per chunk as its edge parameter.This is not an arbitrary number. Preliminary exploration and experimentation with emotion mining and parliamentary proceedings showed a correlation between text size and emotion scores: The larger the text, the higher the emotion score. This was, however, only the case with very short texts up to approximately 250 words. In texts containing more than 250 words, this correlation was not observed. This was, presumably, also impacted by zero-scores (no emotion words identified), which are more likely to occur in short texts, compared to longer texts.<sup>26</sup>

There is, however, another issue with these edge parameters: What if a document consists of 501 words? This document would be segmented into two neat 250 word chunks, but a third chunk will only contain a single word. Therefore, an extra step is implemented in the chunking process. All the text documents in the corpus are chunked into segments of the same size, but small variation in the chunk size is allowed. This variation does, however, not exceed a 50% chunk size (125-word) limit of maximum variation between chunks.<sup>27</sup> In the case of the example mentioned, this leads to segmenting the document in a 250-word chunk and a 251-word chunk.

<sup>23</sup> Jockers, *Text Analysis with R for Students of Literature*, 120–24, 137–42; Rockwell and Sinclair, *Hermeneutica*, 37; Goldstone, 'Producing a Segmented Corpus of Novels', 1–2.

<sup>24</sup> Brezina, *Statistics in Corpus Linguistics*, 223.

<sup>25</sup> Jockers, *Text Analysis with R for Students of Literature*, 137.

<sup>26</sup> Preliminary experiments with emotion mining on samples of the *Handelingen* data showed that in texts with 0–50 words, the percentage of zero-scores was more than 75%. For texts with 101–150 words, this was almost 20%, whereas this percentage of zero-scores was close to zero in texts with 251–300 words.

<sup>27</sup> Goldstone, 'Producing a Segmented Corpus of Novels', 2; Jockers, *Text Analysis with R for Students of Literature*, 140–41.


After each document was chucked, a unique chunk-identifier was added to each individual text chunk in the dataset. Chunks are also linked to the document-level metadata (date, document-identifier, etc.). From now on, 250-word text chunks are used as basic units for analysis, unless otherwise specified.The subsequent pre-processing steps are not only consequential; they can already be considered part of the analytical process.

#### A 'Bag of Words'

One of the most consequential and sometimes shocking processes in pre-processing text collections for computational analysis is the creation of a so-called 'bag of words'. Central to a bag-of-words approach is the presumption that the original order of words within the texts does not inform the computational analysis. Although there are examples of sentences in which changing word order fundamentally alters their meaning, this is in practice quite rare. A bag of words considers the text as the set of words contained in it. Bag-of-words representations of texts in matrices are suitable as input for further text-mining applications. In many text-mining analyses, a list of words per document (or other text unit) is usually sufficient to capture the meaning of the text – of that which is of interest. Therefore, in this step, word order and the sentence structure of the texts are all discarded.

In the pre-processing of the text collections in this study, the text chunks of 250 words have all been split into individual words. Proceeding, a matrix with frequency counts of each unique word in each chunk was created.<sup>28</sup> The result is formally known as document-term matrix (DTM) or document-feature-matrix (DFM).<sup>29</sup> Usually, the documents or text units (e.g., chunks) correspond with the rows of a DTM, whereas the columns correspond to all the unique words in the text collection. A reversed representation is also possible and is, unsurprisingly, referred to as a term-document matrix, or TDM.

<sup>28</sup> Christopher Manning, Prabhakar Raghavan, and Hinrich Schuetze, *Introduction to Information Retrieval* (Cambridge: Cambridge University Press, 2009), 117; Grimmer and Stewart, 'Text as Data', 272; Hoyt Long and Richard Jean So, 'Literary Pattern Recognition: Modernism between Close Reading and Machine Learning', *Critical Inquiry* 42, no. 2 (1 January 2016): 254.

<sup>29</sup> In the context of this investigation, the terms document-term matrix (DTM) and documentfeature matrix (DFM) can be used interchangeably. In general, in a DFM, the 'features' are not necessarily terms, but may refer to other properties of a document besides terms.

The matrices usually contain absolute frequency counts of word occurrences in each unit within a certain collection of documents. Other schemes for counting or weighting words are also possible, such as relative frequencies or proportions, or term frequency-inverse document frequency (TF-IDF). Especially TF-IDF has some important methodological implications, about which more later. A function to create a DTM or DFM is indispensable in R packages and other applications for text mining.<sup>30</sup> All the different text collections discussed thus far have been transformed into a DTM before any further computational processing was performed. In this investigation, every 'document' in the DTM is a 250-word chunk of text and a TF-IDF weighting has been assigned to each individual word in the matrix. However, what is TF-IDF, and why use these weightings in preference to simple word counts?

#### **4.2 A Computational Analysis of Historical Text Collections**

Transforming each chunk of a text collection into a row in a DTM, and their unique words into columns of that same matrix, is more than simply another part of preprocessing texts.These steps can also be considered part of the analytical process itself.The choices made and parameters set here influence the eventual analytical outputs. The choice to use TF-IDF as a weighting measure in creating DTMs requires some additional explanation, as it is fundamental to the measure of the intensity of emotional language manifest in the texts analysed. As previously discussed, the emotion lexicons are used to identify emotional terms in parliamentary proceedings, queen's speeches, and newspaper articles. The EmoLex allows for the identification,measurement, and comparison of emotional language in each speech, newspaper article, or debate. Emotion-lexicon word lists are used to identify emotion words in the DTMs. The following section shows how DTMs are used to combine the 'weighting' of all lexicon words in order to classify or distinguish a text chunk, or to document it as emotional (or not), and to compare different emotions and documents with one another.

<sup>30</sup> Silge and Robinson, *Text Mining with R*; Benoit et al., 'Quanteda: An R Package for the Quantitative Analysis of Textual Data'; Feinerer and Hornik, *Tm: Text Mining Package*; Cosima Meyer and Cornelius Puschmann, 'Advancing Text Mining with R and Quanteda', *R-Bloggers* (blog), 16 October 2019, https://www.r-bloggers.com/2019/10/advancing-text-mining-with-r-and-q uanteda/.

#### The Problem of 'Just Counting Words'

In defining the emotionality of a given text, simply counting the occurrences of emotion lexicon words seems a straightforward solution to the gathering of 'emotion scores.' However, the typicality or distinctive characteristics of a text are not necessarily best captured by its most frequently occurring words. In general, very common words across text collections are bad discriminators of the distinctive characteristics of a specific document (or text unit, or 250-word chunk, etc.) in comparison with the entire collection.<sup>31</sup> Even when stop words such as 'the', 'and', 'a', 'be', etc. are eliminated, the remaining most frequent terms do not necessarily best capture a document's distinctiveness. Absolute word counts are affected by variations in text length. In this study, for example, with every extra word in a text, there is more chance of that word being one that is associated with emotion.Thus, the use of relative frequencies or proportions would seem to offer a better alternative. However, in the context of this investigation, the use of relative frequencies is largely superfluous. Normalisation by text segmentation (chunking) already corrected for variations in document length.

In addition, relative frequencies, or text chunks of equal size, do not yet solve the otherissuementioned above, namely that the(relative)most frequent terms within a document do not necessarily best represent its distinctiveness.Thus,in this context, using relative word counts has as a consequence that frequently occurring, rather common (emotion) words are assigned the same weight as very exceptional ones. Intuitively, relatively rare words seem to better define the distinctiveness of a text compared to the rest of the text collection. Therefore, a more sophisticated weighting of words is needed. The question is how this might be achieved: How might relative distinctiveness be measured?

#### Weighting Words with TF-IDF

The intuition that a term that occurs in many units across a text collection is not a good discriminator and should be assigned a lower weight than one that occurs only rarely, is implemented in the term frequency-inverse document frequency (TF-IDF) weighting scheme.<sup>32</sup> On a fundamental level, TF-IDF is still based on counting words. However, its implications are more sophisticated. It may seem complicated at first glance, but the intuition behind TF-IDF is easy to understand. TF-IDF assigns a weight to every occurrence of a word in a document that is, roughly speak-

<sup>31</sup> Stephen Robertson, 'Understanding Inverse Document Frequency: On Theoretical Arguments for IDF', *Journal of Documentation* 60, no. 5 (2004): 503; Silge and Robinson, *Text Mining with R*, 31.

<sup>32</sup> Robertson, 'Understanding Inverse Document Frequency', 503.

ing, based on its frequency in that specific document, corrected for its overall occurrence across the complete collection (of documents). In this way, TF-IDF takes into account a word's commonness or its rarity across the entire text collection under scrutiny. The occurrence of a relatively common term is assigned less weight than that of a relatively rare term.The TF-IDF weight of a word measures how distinctive a word is for a specific document in a collection of documents – in the context of the entire collection. The TF-IDF weight of a word approaches the intuitive distinctiveness of words in a text better than when simple word counts or relative frequencies are used. Its use requires only a relatively simple calculation, and hence the scheme has become widely used. TF-IDF has proven to be extraordinarily robust and hard to beat, even by more complex or sophisticated approaches to text-mining tasks.<sup>33</sup> So how (and why) does it work?

#### How TF-IDF works

The TF-IDF weight of a word consists of a combination of two different calculations. The first, the TF-part, assigns weight to the words that occur more frequently in the document, just as a basic word frequency count does.There is, however, a complicating factor. Take, as an example, a fictive text collection containing documents A, B, C, and D. It seems straightforward to consider 13 occurrences of a given word in document A more distinctive for that document compared to document B, containing only 9 occurrences of the same word. Nevertheless, it is less intuitive if two million occurrences of that wordin document C should weigh much more (double!) than one million occurrences of this same word in document D. There is a superficial point in which the assertion that 'the more a term occurs, the more is its distinctiveness' no longer makes much sense. Compared to documents A and B, both documents C and D contain an exceptional high count for the same word.<sup>34</sup> To make the TF-part of the TF-IDF calculation more intuitive and pragmatic, its calculation is therefore logarithmically scaled: Using a base parameter of 10, 0 occurrences weigh as 0, a single occurrence as 1, but 2 is weighted as 1.3, 10 as 2, 1000 as 4, 1.000.000 as 6, and 2.000.000 as 6.3, etc. The possibility to use this logarithmic scaling is incorporated in the parameter settings of the quanteda package for R. TF is defined as: TF = (1 + the log of the counts of word X) / (1 + log of the average word count within the document).<sup>35</sup> TF-IDF is not only based on the term frequency (TF) of a word within the

<sup>33</sup> Robertson, 503; Wevers, 'Consuming America', 57.

<sup>34 &#</sup>x27;Why Is Log Used When Calculating Term Frequency Weight and IDF, Inverse Document Frequency?', Stack Overflow, accessed 29 April 2021, https://stackoverflow.com/questions/27067992/why-is-log-used-when-calculating-term-frequency-weight-and-idf-inversedocument.

<sup>35</sup> Note that the value '1' is added to the log of TF.When TF is equal to 1, the log is 0. By adding the value '1', TF(0) is distinguished from TF(1). For more information, see for example Benoit et al.,

specific document, but also takes into account the number of documents in which that word occurs across the collection. This is the inverse document frequency part, or IDF.

Calculating the IDF for a given word is the part of the weighting scheme in which the rareness (or commonness) of a term within the collection of documents under scrutiny is incorporated.This part is powerful and explains why TF-IDF works so efficiently and became dominant in information retrieval and finding divergent documents or words.<sup>36</sup> The intuition behind IDF is that it diminishes the weight of words that occur in many of the documents in the collection and increases the weight of words that are exceptional. The calculation of IDF works as follows: IDF = log (1 + (number of documents in the collection) / (number of documents containing word X)). Words scoring very high on IDF are words that occur very rarely within the entire text collection. In practice, these words could be indeed quite rare and therefore distinctive, but they could also consist, for example, of incidental OCR errors, very uncommon names, or spelling mistakes. A linear (non-logarithmic) IDF calculation may therefore be boosting the document scores of words that score very high on IDF too much.<sup>37</sup> Just as in the calculation of the TF part, a logarithmic scale is,in the IDFpart, applied to dampen this effect.

The calculation for the complete TF-IDF weight of a word is: TF-IDF = TF \* IDF. In summary, the resulting TF-IDF weights for an individual word in a document are:


<sup>&#</sup>x27;Quanteda: An R Package for the Quantitative Analysis of Textual Data'; Kenneth Benoit and Kohei Watanabe, 'Weight a Dfm by Tf-Idf', quanteda: Quantitative Analysis of Textual Data, accessed 29 April 2021, https://quanteda.io/reference/dfm\_tfidf.html; Kenneth Benoit and Kohei Watanabe, 'Weight the Feature Frequencies in a Dfm', quanteda: Quantitative Analysis of Textual Data, accessed 29 April 2021, https://quanteda.io/reference/dfm\_weight.html.

<sup>36</sup> Robertson, 'Understanding Inverse Document Frequency', 503, 516; Abhijeet Kumar, 'Sentiment Analysis Using Tf-Idf Weighting: Python/Scikit-Learn', *Machine Learning in Action* (blog), 12 February 2017, https://appliedmachinelearning.wordpress.com/2017/02/12/sentiment-an alysis-using-tf-idf-weighting-pythonscikit-learn/; Silge and Robinson, *Text Mining with R*, 31–32.

<sup>37</sup> Manning, Raghavan, and Schuetze, *Introduction to Information Retrieval*, 117–18.

<sup>38</sup> Manning, Raghavan, and Schuetze, 118–20; Kumar, 'Sentiment Analysis Using Tf-Idf Weighting'; Kenneth Benoit and Kohei Watanabe, 'Compute the (Weighted) Document Frequency of a Feature', quanteda: Quantitative Analysis of Textual Data, accessed 29 April 2021, https://quanteda.io/reference/docfreq.html; 'Why Is Log Used When Calculating Term Frequency Weight and IDF, Inverse Document Frequency?'

### A DTM Based on TF-IDF Weighting

TF-IDF weights form the basis for creating structured implementations of bags of wordsin the DTMs. Table 6 shows some rows of an example DTM that has been based on the small sample of *Handelingen* data that was previously used in the earlier tables in this chapter and in Chapter 3. As can be seen in the table, in this investigation, DTMs (and, accordingly, TF-IDF weights) are always based or calculated on all words occurring in the pre-processed text chunks. A next step is to use the DTMs with TF-IDF weights to generate emotion scores.

*Table 6: Fragment of DFM containing three different chunks of a war-related document and the TF-IDF scores for some of its unique words.<sup>39</sup>*


#### From TF-IDF to Emotion Scores

The implementation of TF-IDF approaches the distinctiveness of words in a text more closely than when simple word counts or relative frequencies are used. In the TF-IDF weighting scheme, the TF is an intra-document factor that depends on occurrences of the term within an individual document, and IDF is a factor that is constant across the entire collection of documents. Using TF-IDF mitigates the loss of context information from both each individual document and the complete collection.<sup>40</sup> As a result, however, calculated weights are always based on, and only ap-

40 Kumar, 'Sentiment Analysis Using Tf-Idf Weighting'.

<sup>39</sup> These are not necessarily words from the NRC EmoLex. These words are derived from the first sentence of the first document and only serve showing an example of a DTM here.

plicable to, the information from within the collection of documents used in the calculation. This makes emotion scores based on TF-IDF weights less suitable for an analysis that is primarily concerned with the comparison between different national parliaments, or with comparisons between entirely different datasets. This is because separately established IDF weights in separate DFMs particular to a certain dataset can dampen any mutual differences between that dataset and others. In this investigation, however, rather than considering this as a bug, it is used as an advantageous feature in the analytical process.

This investigation focusses on the relative emotional exceptionality of subsets of documents, thematic clusters, various topics, or periods.This encompasses primarily the comparison of variation within each particular dataset – variation between the scores of particular (clusters of) topics, or over time. It is likely that, in making such comparisons, there is a threshold of emotional language present in (most of) the text collection.This baseline of emotionality is particular to a specific text collection. It is therefore an advantage of TF-IDF weighting that this baseline is taken into account. Take, for example, the *Handelingen*. In the parliamentary setting in which (mostly) problems and political and societal issues are being discussed, the existence of a baseline of (presumably negative) emotional language use is almost inevitable. Relatively common words to describe matters under discussion (e.g., 'victim', 'criminal',or 'war') are omnipresent in the proceedings, but are themselves potentially part of emotion lexicons. By using TF-IDF, a threshold of emotionality is created that lessens the influence of such relatively common emotional terms relative to more strongly emotional words (e.g., 'horrible', or 'disaster') that actually distinguish exceptional emotional discussions from the rest.<sup>41</sup> In this investigation, the TF-IDF weights of words are therefore always calculated within the boundaries of the historical text collection being analysed. TF-IDF weights are definedin a DTM thatis based on individual collections separately. This means that four different DTMs were established in this investigation, based on each of the two newspaper article samples, the queen's speeches, and the parliamentary proceedings. In this way, the TF-IDF weights of the words in individual debates, thematic clusters, or subsets are always established within the context of the entire text collection to which they belong.

The DTMs, based on TF-IDF, assign weights to each unique word. When these weights for a meaningful combination or set of words are added up, we know how characteristic these words are for this specific document in comparison to the other documents in the collection. For each 250-word text chunk of the sources used in this investigation, the TF-IDF weights of each occurrence of a word from the various EmoLex categories are summed. These combined weights measure how distinctive emotion words are for each chunk, relative to all other chunks in the collection. For

<sup>41</sup> Van Lange and Futselaar, 'Vehemence and Victims'.

the non-war-related debates and the newspaper articles, the mean scores of all aggregated 250-word chunks of each day between 1945 and 1989 were calculated. Due to the limited size (in number of words) of the queen's speeches, individual chunk scores were used for that particular dataset.The emotion scores were plotted as data points on timelines (see also Chapter 5). All of the 250-word chunk scores belonging to a particular debate, topic,or thematic cluster used in the 'war survivors' case studies in Chapters 5, 6, and 8 were also averaged to arrive at an 'emotion score' that is particular to that analytical unit of text. The next step in the workflow is the statistical analysis, evaluation, and visualisation of these results.

### **4.3 Statistical Evaluation and Visualisation of Results**

When working with quantitative measurements, the visualisation and statistical evaluation of quantitative data are not only inevitable, but also important if we are to gain an understanding of the meaning of the results and scores. The first part of this section deals with the macro-analysis of emotions in the sources: a diachronic representation of the emotion-mining results that is particularly relevant in the context of the 'emotionalisation'issue discussed in Chapters 1 and 5.The second part deals with a more zoomed-in analysis of particular clusters of war-related themes, debates, and the words behind the emotion scores central to the case studies in Chapters 5, 6, 7, 8, and 9.

#### Timelines, Histograms, and Trends

Diachronic developments in emotion scores can be analysed and visualised over time because all of the datasets used are time coded. Simply plotting scores on a timeline already represents an interpretation of the results.<sup>42</sup> Diachronic graphs are especially helpful in the interpretation of patterns over time. Scores per 250-word chunk are, as discussed earlier, summarised per day to arrive at plots displaying broader trends over time. For each one of the more detailed case studies on the different groups of war survivors, scores are aggregated per document or per cluster of topics. For all of the time series graphs, time is plotted on the x-axis, whereas the emotion scores are plotted on the y-axis. Scores for each analytical unit (chunks, day, thematic cluster, or document) are plotted as a single grey dot on the timeline. As individual data points, these dots have the advantage that they stand out as solid anchors of evidence, whilst overall trends can be somewhat tentative, or sometimes provide an incomplete overview (regarding variation between data points, for ex-

<sup>42</sup> Brezina, *Statistics in Corpus Linguistics*, 225.

ample).<sup>43</sup> Trendlines show overall developments and diachronic trends, and their changes or continuities more broadly.

In this investigation, trendlines are always plotted and smoothed using the LOESS (Local Polynomial Regression Fitting) function. Smoothed trendlines assist in revealing long-term trends in the data. LOESS is integrated in the stat\_smooth and geom\_smooth functions in R package ggplot2. A span parameter setting (alpha) of 0.50 was used. This parameter controls the amount of smoothing applied to the scores (i.e., the 'wiggliness' of the trendline). LOESS smoothing plots a trendline (in black) and a 95% confidence interval (in grey) that shows how (un)certain the development of the trend is. Scores within the upper and lower boundaries of the grey band are the variation of the trendline.<sup>44</sup>

To give an example of this visualisation, fictive linearly increasing trust scores for 440 random days between 25–09-1945 and 28–12-1989 are generated and plotted as grey dots in Graph 1. The smoothed trendline of these fictive scores is plotted in black. To compare the scores of particular subsets or debates with the parliamentary average, and to the scores of other subsets or other case studies, the mean scores of individual debates or thematic clusters are also plotted on the timeline. For the sake of visibility, these means are plotted as coloured bars (rather than dots).Graph 1 shows the scores for the emotion trust of the exemplary debates used in Chapter 3.<sup>45</sup> Of course, smoothed trendlines can also be drawn through the scores of a particular subset of debates to generate a diachronic trendline of the scores of that particular subset.

<sup>43</sup> Brezina, 225.

<sup>44</sup> For more details and parameter information see Hadley Wickham et al., *Ggplot2: Create Elegant Data Visualisations Using the Grammar of Graphics*, accessed 4 May 2021, htt ps://ggplot2.tidyverse.org/; 'Smoothed Conditional Means — Geom\_smooth', accessed 5 May 2021, https://ggplot2.tidyverse.org/reference/geom\_smooth.html; 'Loess: Local Polynomial Regression Fitting', accessed 5 May 2021, https://rdrr.io/r/stats/loess.html; 'R − Understanding the Confidence Band from a Polynomial Regression', Cross Validated, accessed 5 May 2021, https://stats.stackexchange.com/questions/82603/understanding-the-confidenc e-band-from-a-polynomial-regression.

<sup>45</sup> This choice of emotion-specific colours corresponds with Saif Mohammad's colour-emotion association scheme, see Mohammad, 'From Once upon a Time to Happily Ever After', 733.

*Graph 1: Fictive diachronic development in trust scores (grey dots), the smoothed trendline (black) and scores of exemplary debates from Chapter 3 (blue bars) plotted over time (1945–1989).*

#### Linear Regression

Linear regression analysis helps to define whether an observed trend is one of continuity, increase, or decrease, and whether such an observed change is statistically significant or not. Statistical significance here refers to the confidence that an observed effect is unlikely to have occurred merely by chance.<sup>46</sup> This statistic is important in this study, as the result tells whether an observed effect is likely to be the result of chance, or whether the variable 'time' is likely to have an effect on an observed trend. Linear regressions analysis is considered particularly relevant for the analyses in Chapter 5. Linear regression is used to evaluate whether the process of increasing 'emotionalisation' can be discerned in emotion-mining outputs over time. In this study, the lm function from the R stats package is used.<sup>47</sup> To create neat out-

<sup>46</sup> Amy Gallo, 'A Refresher on Statistical Significance', *Harvard Business Review*, 16 February 2016, https://hbr.org/2016/02/a-refresher-on-statistical-significance.

<sup>47</sup> R Core Team, *R: The R Stats Package*, version 4.0.5, R, 2021, https://stat.ethz.ch/R-manual/R-d evel/library/stats/html/stats-package.html.

put tables within the RStudio workspace, functions from the packages broom and gridExtra are applied.<sup>48</sup>

Table 7 displays the linear regression analysis output of the fictive diachronic trust scores (which have also been displayed in Graph 1) as an example. The greycoloured statistical output table is directly derived from the RStudio working environment.

#### *Table 7: Linear regression analysis outputs for fictive trust scores between 1945 and 1989.*


The very low p-values in the right column indicate that there is a significant linear relationship between time (x-axis in Graph 1) and trust scores (y-axis). A p-value lower than 0.05is used here as a threshold of significance. Table 7 shows a significant change over time in the scores. Just as in the other two statistical tests (which will be discussed later), significance is not the only interesting value provided by the test results. Where the p-value indicates that there probably is an effect, the estimate in row 2 of Table 7 shows the magnitude and direction (positive or negative, increasing or decreasing) of that effect. The estimate is a positive value in this example, which indicates a positive relationship between time and trust score (in the fictive data). The output of linear regression analysis shows not only that the observed change over time is statistically significant, but more importantly, that it is also an increasing trend. As could already easily be observed in the trendline in Graph 1, there is a rather strong effect consisting of an increase in trust scores (in the exemplary data) over time.

So far, the statistical tests and visualisations discussed have dealt with time series and diachronic developments. In the case studies of Chapters 5, 6, 7, 8, and 9, however, zooming in on more particular subsets and thematic clusters is also considered important. How then are we to gain insights into the emotion scores of more particular subsets, thematic subsets, or debates?

<sup>48</sup> David Robinson et al., *Broom: Convert Statistical Objects into Tidy Tibbles*, version 0.7.6, 2021, h ttps://CRAN.R-project.org/package=broom; Baptiste Auguie and Anton Antonov, *GridExtra: Miscellaneous Functions for 'Grid' Graphics*, version 2.3, 2017, https://CRAN.R-project.org/packa ge=gridExtra.

#### Summary Statistics

Summary statistics provide insights into characteristics of emotion scores across a particular subset or collection. They are based on relatively simple calculations. Whether the results are grouped per basic emotion, per debate, or per subset, for each of these analytical units the mean, median (middle value), standard deviation, number of observations, coefficient of variation, minimum and maximum emotion scores are calculated and presented in a table.These operations can all be performed with R base functionalities.<sup>49</sup> The descriptive summary statistics are used to compare different subsets or different basic emotion categories on their scores and outputs of emotion mining. The coefficient of variation in the summary statistics tells us something about the variation within the results, but does not tell where this variation in the dataset or subset might be located. Table 8 displays the scores of the exemplary debates used before on language associated with anger (angscore), disgust (disscore), fear (fearscore), sadness (sadscore), joy (joyscore), and trust (trustscore).

*Table 8: Summary statistics for the four example debates' scores sorted by each of the basic emotion categories from the NRC EmoLex.*


To plot summary statistics (represent the results visually), and to find out where (e.g., in which debate) variation in the scores occurred and what discussions stand out or not, boxplots are used as a powerful method to visually assess and present emotion-mining results.

<sup>49</sup> R Core Team, *R: The R Base Package*, accessed 4 May 2021, https://stat.ethz.ch/R-manual/R-de vel/library/base/html/00Index.html.

#### Boxplots

Boxplots offer the opportunity to look in more detail at each emotion score without losing sight of its context. Instead of representing each analytical unit, topic, thematic cluster, or collection of documents as a single statistic (mean, median, etc.), the internal distribution of scores (the variation within) is also made visible. Boxplots display the range of scores: the minimum, the maximum, and the exceptional outlier scores. The box part of the boxplot represents the interquartile range – the 50% of values around the median.The whiskers above and below represent the (25%) lower and (25%) higher values, but exclude extreme values (outliers). Outliers are the dots plotted outside the scope of the box and the whiskers.<sup>50</sup> This allows not only for the generation of insights into internal variety in scores, but also for a comparison of the differences in scores and distributions amongst different debates that form the main analytical units used in the case studies of this study.

The R package ggplot2 offers sophisticated, flexible, aesthetic, and tailored visualisation options and parameter settings in plotting data in boxplots. The ggpubr package is complementary to ggplot2 and assists the creation of publication-ready visualisations within the R and RStudio environment.<sup>51</sup> Boxplot 1 shows the mean score (plotted as ) and median score (plotted as **–**) of the four abovementioned example debates, as well as the distribution of the scores of all individual 250-word chunk scores (•) belonging to each individual debate or document.

<sup>50</sup> Brezina, *Statistics in Corpus Linguistics*, 23, 226.

<sup>51</sup> Wickham et al., *Ggplot2: Create Elegant Data Visualisations Using the Grammar of Graphics*; Alboukadel Kassambara, *Ggpubr: Ggplot2 Based Publication Ready Plots*, accessed 4 May 2021, ht tps://rpkgs.datanovia.com/ggpubr/index.html.

*Boxplot 1: Anger scores for each 250-word chunk in the four example debates of the Handelingen dataset.*

Boxplots allow for a visual assessment of the data, and a comparison between different analytical units. Taller boxes mean more variate scores.This also applies to tall whiskers, or many outliers. Boxplot 1 , for example, shows that the median score of the 1966 debateis lower than the others.The 1955 debate, however, truly stands out, but not so much because of an extreme median value. It is the larger difference between the mean and the median that indicates a higher degree of variation between the chunk scores of this debate (compared to the other debates). The 1955 example stands out with a more skewed distribution and some extreme (high) outlier scores. This means that only a few chunks fuelled the higher mean score. Generally, the difference between two separate boxplots is significant when the boxes do not overlap at all and is likely to be significant if the median line of one box lies outside the range of that of the other box.<sup>52</sup> An example of statistically significant diverging results is given in Boxplot 2, where results for the four example debates on the emotion trust are plotted.

<sup>52</sup> Linh Ngo, 'How to Compare Box Plots', *BioTuring's Blog* (blog), 22 May 2018, https://blog.biotu ring.com/2018/05/22/how-to-compare-box-plots/.

*Boxplot 2: Trust scores for each 250-word chunk in the four example debates of the Handelingen dataset.*

Although this visualisation method is insightful when working with just a few analytical units, boxplots are not always the best way of comparing a large number of observations or debates. This is especially true when looking for the significance of mutual differences between many boxes in a single visualisation. Statistical evaluation of the differences between scores are therefore used to complement the boxplots.

#### Kruskal-Wallis

When looking at the example data used in this chapter, the calculated coefficients of variation (Table 8) and visual representation of the scores (Boxplot 1 and Boxplot 2) indicated differences between the scores of the four debates. This investigation uses the non-parametric Kruskal-Wallis rank sum test from the R stats package to see whether differences (between the mean scores) of multiple debates are statistically significant.<sup>53</sup> Non-parametric tests, such as Kruskal-Wallis test, are methods for statistical analysis that do not require distributions to meet required assumptions, as is required for parametric tests. In other words, the family of non-parametric statistical methods makes fewer assumptions about the data.They therefore

<sup>53</sup> R Core Team, *R: The R Stats Package*.

form a safe, strict,and careful approach to the statistical evaluation of emotion-mining results in this study. As the example results in the boxplots have already shown, the distribution of scores around the mean are not always equal amongst the various debates. In statistical terms, the homogeneity of variances assumption is violated. Moreover, preliminary experiments with larger samples of the data has shown that the scores within the analytical units of this investigation are not always normally distributed, which means that the normality assumption (required for a parametric t-test or ANOVA) is violated.<sup>54</sup>

Non-parametric tests are based on ranking scores in the dataset, rather than using the actual values of the scores for statistical analysis. By using ranks, the effects of strong outliers and variations in the scores are dampened. Imagine that when the two lowest scores in an analysis are 0 and 9 (difference of 9), their rankings (respectively 1 and 2) dampen the anomalous exceptionality of the lowest outlier. By using the rankings, the Kruskal-Wallis test indicates whether there are overall, significant differences between the scores of the debates.<sup>55</sup> In this investigation, the test is used on the per-chunk emotion scores grouped per individual debate and calculates differences between the debates. Table 9 was created using the R packages broom and gridExtra. The table displays outputs for each emotion category of the EmoLex and the four example debates.<sup>56</sup>

*Table 9: Output of the Kruskal-Wallis test on scores of the four example debates, sorted per emotion category.*


<sup>54</sup> Andy Field, *Discovering Statistics Using IBM SPSS Statistics*, 4th edition (London, Thousand Oaks, New Dehli, Singapore: SAGE Publications, Inc, 2014), 213–15, 219–20, 236–38.

<sup>55</sup> Field, 213–15, 219–20.

<sup>56</sup> Auguie and Antonov, *GridExtra*; Robinson et al., *Broom*; 'Introduction to Broom: Let's Tidy up a Bit', cran.r-project.org, 5 April 2021, https://cran.r-project.org/web/packages/broom/vignet tes/broom.html.

As could be expected based on the assessment of the boxplots, the trust scores differ significantly between the example dataset's four different debates. This does not apply in the case of the anger scores. The Kruskal-Wallis test does not indicate which particular debates' scores differ (in terms of statistical significance), and which debates are more divergent from one another. The result of Kruskal-Wallis tests, or the finding that variation exists between debates, can be broken down by mutually comparing all pairs of debates. A pairwise Wilcoxon rank sum test is therefore a next step in the workflow.

#### Pairwise Wilcoxon

To tease apart the finding of significant variation between means of debates, a pairwise comparison using a Wilcoxon rank sum test is used. In this investigation, this test calculates the pairwise or mutual statistical significance of differences between scores for each emotion and for every possible pair of debates.<sup>57</sup> Outputs for emotion scores for trust in the four example debates can be found in Table 10. Variable N in the table represents the number of 250-word chunks in debates 1 (d1) and 2 (d2).

*Table 10: Output of the pairwise Wilcoxon rank sum test on the scores for trust of the four example debates.*


In addition to the significance of the results, the effect size (or magnitude) of each pairwise comparison is also reported in the output of pairwise Wilcoxon tests. This effect magnitude is a value that varies from 0 to (close to) 1. Effect size is meaningful to take into account next to statistical significance (alone). Where the p-value indicates that there is probably an effect (a difference, in this case), the effect size tells us the magnitude of that effect.The interpretation of the variable is usually that a value between 0.10 and 0.30 is considered a small effect, whereas values between

<sup>57</sup> Unless otherwise specified, a confidence interval of 95% (α = 0.05) is used. A difference is statistically significant when (p =< α). Field, *Discovering Statistics*; Alboukadel Kassambara, *Rstatix: Pipe-Friendly Framework for Basic Statistical Tests*, version 0.7.0, 2021, https://CRAN.Rproject.org/package=rstatix; R Core Team, *R: The R Stats Package*.

0.30 and 0.50 are considered moderate. Values over 0.50 are considered large effect magnitudes.<sup>58</sup> Table 10 shows that whilst some mutual differences between debates' scores are significant, this does not necessarily mean that the magnitude of this difference is also large. The fifth row of the table shows that there was only a very moderate effect observed that was very unlikely to be the result of chance. Although these outputs already reveal much about the results of emotion mining, lexicon-based emotion mining has more to offer than quantitative output alone.These quantitative measurements can be combined with the qualitative, linguistic markers that underlie the emotion scores.

#### The Words behind the Scores

The words of the EmoLex identified and measured in the historical text collections not only offer additional insights, but also bridge the gap between a distant, quantitative perspective on emotions, and a qualitative close reading of the historical sources under scrutiny.They are 'the words behind the scores'.<sup>59</sup> To create top (emotion) word lists for every debate (cluster), functions for creating and manipulating DFMs from the quanteda package – the function topfeatures in particular – are used. In addition, the textstat\_frequency function from the quanteda.textstats package is particularly helpful here.<sup>60</sup> From the top word lists created with these packages we learn, for example, that the Dutch words *boos* ('angry') and *geschil* ('conflict') are distinctive of the 1953 example debate mentioned earlier. Resulting top word lists can also be visualised, including TF-IDF scores for each individual word (see Table 11). The TF-IDF scores were not normalised in creating the top word lists and visualisations. These graphs are specific to each debate, and the absolute scores in these tables cannot be readily compared amongst debates or subsets. However, what is considered important here is that the top words tables show the distribution of the unique words within the debate and to what extent different words contribute to a certain debate's emotion score.

<sup>58</sup> Field, *Discovering Statistics*; Kassambara, *Rstatix*; R Core Team, *R: The R Stats Package*.

<sup>59</sup> I am grateful to Peter Boot for making me aware of the value of 'the words behind the scores'. Peter Boot, 'Emotion mining with lexicons, e-mail to the author', 7 July 2020.

<sup>60</sup> Benoit et al., 'Quanteda: An R Package for the Quantitative Analysis of Textual Data'; Kenneth Benoit et al., *Quanteda.Textstats: Textual Statistics for the Quantitative Analysis of Textual Data*, version 0.94, 2021, https://CRAN.R-project.org/package=quanteda.textstats.

*Table 11: Top 10 anger words in two of the example debates.*

#### Method Evaluation in Historical Research

This chapter showed how existing historiography is evaluated by using an 'outside' method in the analytical process. Methods, techniques, and resources derived from fields such as computer science and computational or corpus linguistics are applied to a historical research problem. Computational text-mining techniques are used in measuring and observing trends and proportionality, and in approaching the sources from an alternative 'outside' perspective. Because complete or very neat data are a rarity in the study of history, a lower performance of text mining is expected, compared to the standards of the fields developing the techniques used. In the application of text mining in this study, however, the quantitative outputs do not have to be overly precise or exact according to usual computer or corpus linguistic standards.<sup>61</sup> Wiedemann suggests, in some academic fields, researchers can be less restrictive when evaluating computational techniques in their particular research setting.<sup>62</sup> Perhaps even more important, the text-mining techniques are primarily a means rather than a goal in this investigation. Therefore, the evaluation of the validity of the methods used in this investigation does not follow the standards of their original academic disciplines. The evaluation of the methods, techniques, and resources, expressed in performance measures such as precision or recall, is considered out of the scope of this historical investigation. Nevertheless, applying

<sup>61</sup> Wiedemann, *Text Mining for Qualitative Data Analysis*, 130–31.

<sup>62</sup> Wiedemann, 130–31.

text mining in a field traditionally not overly committed to computational analysis does not mean the validity of the methods used is relied on without reflection.

First, text mining, as used in this investigation, is not the only approach applied to the analysis of the sources. Emotion-mining outputs are not seen as an end in themselves, but rather inspire further close-reading analysis. Second, the evaluation standards of text-mining techniques are reconsidered in the context of the hermeneutic nature of historical research. Methods are evaluated in this study from the standpoint of the historian, in the context of historical knowledge and historiographical practices. Chapter 5 addresses this alternative validation procedure, taking into account the contemporary context of the sources, the historiographical origin of this investigation, and the hermeneutic nature of historical work. This is done, in the first place, by assessing whether the modern-day lexicons are able to identify emotion words in historical sources at all. In a next phase, the results are tested against a well-established principle known as the Pollyanna principle. Finally, a well-known emotional peak issue in Dutch post-war parliamentary history is analysed following the same methodological workflow. The results of this case study are compared to non-related, average parliamentary language use to see whether the exceptional emotional charge of known historical episodes of emotional outbursts is also identified as a peak in emotional language by the application of lexicon-based computational methods.

## **5. Peering Through the Macroscope | Baseline and Background**

This chapter applies the workflow described in Chapter 4 to the analysis of various historical text collections. Placing parliamentary debates centre stage, the chapter contains the initial results of a computational analysis of Dutch historical texts from the period 1945 until 1989. The baselines for further investigations are set, subsequently zooming in on the more particular topics of this investigation. Before addressing (in Chapters 6, 7, 8, and 9) the role of emotions in particular debates on World War II survivors, broader historical and societal developments are assessed with lexicon-based emotion mining. The results of emotion mining are then discussed in relation to more general developments in the manifestation and expression of emotions in Dutch society. From a methodological perspective, the potential value and validity of the computational methods used in addressing emotions in historical sources are evaluated. In the second half of this chapter, the evaluation of lexicon-based emotion mining as a historical research method continues, focussing on the notoriously emotionally charged case study of World War II perpetrators and collaboratoration.<sup>1</sup>

#### **5.1 The 'Emotionalisation' of Society**

#### A Received Wisdom?

There is a widespread consensus that a broad societal development of an 'emotion culture' has occurred in the Western world during the second half of the twentieth century. This historical process, as discussed in Chapter 1, is often described

<sup>1</sup> This process is not performed in an absolute way. In other words, the macroscope is not validated by using some sort of gold standard or ground truth. This process of evaluating methodological validity can, instead, be considered as a relative calibration: Macroscopic output regarding a well-known and notoriously emotional case study is compared relative to the more general or average parliamentary debate scores.

as a process of 'emotionalisation'. This term generally refers to a historical development wherein public displays of emotion have become socially accepted, more common, and therefore more frequent. This idea is reflected in much international literature on, for example, American, British, and Dutch societies.<sup>2</sup> The early 1970s are often considered as the starting point of this presumed increase of emotionalisation. In historiography, this phase is associated with progressive developments and the (consequences of) the democratisation movements of the 1960s.<sup>3</sup> According to the literature, the 1950s in particular represent a striking contrast with the later emergence of this 'emotional culture'. In the Dutch context, the 1950s are generally perceived as a somewhat 'dull' phase, however, this period is also seen as forwardlooking, in which rebuilding and recovery of the nation following World War II were considered a priority.<sup>4</sup> When looking at the Netherlands in particular, this process has been argued to have affected the national media<sup>5</sup> and penal justice,<sup>6</sup> as well as political and parliamentary discourse and debate.<sup>7</sup> Moreover, emotionalisation also features in literature on the national coming to terms with the consequences of the German occupation of the Netherlands in World War II.<sup>8</sup>

As discussed in Chapter 1, both the sociologist Jolande Withuis and the historian Harry Oosterhuis observed a more explicit connection between the broad societal development of emotionalisation and the way survivors of World War II were dealt with in the post-war period. Withuis described that a cycle of speaking (1945–1948), silence (1948–1968), and renewed speaking (1968–1980) played an important role in 'coming to terms' with the consequences of the war, and especially with its victims.<sup>9</sup> After the immediate post-war years, during which there was elaborate yet short-lived attention for war-related experiences and emotions ('speaking'), the subsequent decade is generally characterised as a phase of relative 'silence'. Here, this 'silence' particularly refers to the observation that there was little to no space for the public expressions of emotions in the 1950s – especially not those emotions related to the recent war past. This changed drastically in the third phase during

<sup>2</sup> Beunders, *Publieke tranen*; Beckett, 'How Journalism Is Turning Emotional and What That Might Mean for News'; Kimmich, 'Emotional Culture?'; Beckett, 'Media and the Manchester Attacks: Evil and Emotion'.

<sup>3</sup> Withuis, *Erkenning*, 118; Beunders, *Publieke tranen*, 40–41.

<sup>4</sup> Blom, 'Jaren van tucht en ascese. Enige beschouwingen over de stemming in herrijzend Nederland (1945–1950)'.

<sup>5</sup> Beunders, *Publieke tranen*.

<sup>6</sup> Beunders, *Hoeveel recht heeft de emotie?*

<sup>7</sup> Aerts, 'Emotie in de politiek'.

<sup>8</sup> Withuis, *Erkenning*; Withuis and Mooij, 'From Totalitarianism to Trauma'; Oosterhuis, 'Mental Health, Citizenship, and the Memory of World War II'.

<sup>9</sup> Withuis, *Erkenning*, 118–19; Withuis and Mooij, 'From Totalitarianism to Trauma', 198–99; Oosterhuis, 'Mental Health, Citizenship, and the Memory of World War II', 20.

which, in the words of Withuis, an 'emancipation of emotions' and 'a change in [the] mental climate' took place. These phenomena are also used (amongst others) by the aforementioned authors to explain historical developments in the national legislation related to engaging with war victims in the Netherlands.<sup>10</sup>

The idea of 'emotionalisation' seems to have gained the status of a 'received wisdom'. The understanding of 'things we think we know' regarding broad societal developments can benefit from the complementary 'distant perspective' of macroscopic analysis. Before focussing on the role of emotions in the specific case studies of parliamentary debates on two different groups of survivors of World War II, the process of emotionalisation is investigated, using emotion mining. This chapter proceeds with assessing this broad societal phenomenon by using the same basic instrument – lexicon-based emotion mining – that is used to investigate in more detail the more specific parliamentary debates later on. By investigating all parliamentary debates (not only those related to the war) alongside additional digitised diachronic historical text collections, this chapter sets the baseline as it sketches the background of general developments in the expression of emotions in the Netherlands between 1945 and 1989. This baseline also functions as an empirical norm against which the results of the case studies are assessed in Chapters 6, 7, 8, and 9.

#### A Computational Approach to a Broad Societal Development

The application of text mining makes the complementary analysis of comprehensive datasets feasible and relatively easy. Therefore, as discussed in Chapter 3, there is no good reason to rely on emotion-mining results of the *Handelingen* alone. Although parliamentary debates lie at the heart of this investigation, a sample of historical newspaper articles from two different national Dutch newspapers (*De Volkskrant* and *De Telegraaf* ) and the annual queen's speeches are also analysed.<sup>11</sup> Of course, a complete, comprehensive, or truly representative view of the role of emotions in Dutch contemporary society is practically impossible. However, such a comprehensive view is also unnecessary, as the text collections are used as examples of language use with a certain emotionality. If the received wisdom of emotionalisation is true, one would expect it to be at least observable in these different text collections. In addition, analysing the complementary newspaper articles and queen's speeches provides a more diverse and heterogeneous window on society, compared to analysing parliamentary proceedings alone. The results of the computational analysis of these complementary collections provides a baseline of the manifestation of emotions in the Dutch public sphere between 1945 and 1989.

<sup>10</sup> Withuis and Mooij, 'From Totalitarianism to Trauma', 198–99.

<sup>11</sup> For a more elaborate description of the datasets used, see Chapter 3.

As they offer the possibility for comparison, they serve as a point of departure for the further, more detailed analysis of parliamentary discussions on war criminals, collaborators, the former anti-Nazi resistance, victims of World War II, and the German occupation.

The baseline set out in this chapter is twofold. First, it provides the fundamental methodological asset, enabling a connection to be made between emotion lexicons and historical texts. The analysis of *all* parliamentary debates, the two samples of newspaper articles, and the queen's speeches, made possible by computational textmining techniques, constitutes an important first step in what can be considered an evaluation of the research instrument, the methods, and techniques that are used: Are they able to identify emotional words in historical text sources at all? Can generic emotion lexicons identify words in Dutch historical texts? How do the results compare to more general or universal characteristics of natural language use? Second, the baseline of emotion-mining output provides a historical background. This enables a deeper understanding and interpretation of the analysis of more particular parliamentary debates within a broader context (undertaken later in this investigation).

This chapter addresses the following questions: How emotional were parliamentary discussions, newspaper articles, and queen's speeches between 1945 and 1989? How did the emotional charge of the language used in these text collections develop over time? Can we discern the generally perceived increased emotionalisation in the language used in these sources? Is this emotionalisation replicated by the emotionmining results? Exploring these questions, I make one fundamental assumption: If the received wisdom of an 'emotionalisation of Dutch society'is true, I would anticipate a quantitative increase in the general use of emotional language in parliament, newspapers, and the queen's speeches from the early 1970s onwards.

### **5.2 Results**

The parliamentary debates, a large sample of newspaper articles, and the queen's speeches have been pre-processed and segmented into 250-word text chunks. In line with the workflow described in Chapter 4, words from the EmoLex have been identified, weighted, and these weights are averaged per day or (queen's) speech. Using the NRC EmoLex word lists, the documents were 'scored' on the presence of words associated with various emotions: joy, trust, anger, fear, sadness, and disgust.<sup>12</sup> The

<sup>12</sup> When I speak in the following paragraphs about the 'score' of a specific parliamentary debate on a certain emotion, this score consists of the mean of all TF-IDF weights for all of the 250 word chunks of that debate. However, theoretically, this does not necessarily mean that there is (only) a higher frequency of emotion words in these debates per se. Higher TF-IDF scores

results are presented below. For the sake of clarity and because the types of emotions that are increasing are generally not further specified, only two types of emotions are plotted here: positive and negative emotions. Positive emotion scores here refer to the merged (mean) scores for the EmoLex categories joy and trust, whereas negative scores refer to merged scores for anger, fear, sadness, and disgust. Positive and negative are not used here to refer to opinion polarity, but instead refer to merged groups of scores of the different basic emotions.

#### Queen's Speeches over Time

First, the queen's speeches (1945–1989) are analysed, the results of which are presented below in Graph 2 (negative emotion) and Graph 3 (positive emotion). Because of the small size of the dataset, the scores for each 250-word chunk are plotted individually in grey dots. The trendlines are created using R's smoothing function (LOESS; for more details, see also Chapter 4). Although it is not exactly the same, the meaning of this smoothed line is similar to that of a moving average score (in Graph 2: on negative emotion) over time. A 95% confidence interval is plotted as a grey band around the smoothed trendlines (which are plotted in black). More important here, however,is that the trendlines reveal possible developments and changes in emotion scores over time.

*Graph 2: Negative emotions in the queen's speeches, 1945–1989.*

are also caused by the(relatively) frequent use of(relatively) rarewords. Because of the TF-IDF weighting that is applied to the scores, the (single) occurrence of relatively rare words weighs heavier than the individual occurrence of a very common word. For a detailed description, see also Chapter 4.

*Graph 3: Positive emotions in the queen's speeches, 1945–1989.*

The speeches of QueenWilhelmina in the initial post-war years display relatively high scores compared to the rest of the results (see Graph 2). These speeches were held in 1945, 1946, and 1947. The higher emotional charge seems to support the first phase ofWithuis' cycle of 'speaking, silence,and renewed speaking'.A closer scrutiny of the actual contents of the speeches shows that there was not only reflection on the German occupation, but also words associated with anger, sadness, fear, and disgust were used in relation to the ongoing conflict in Indonesia. For example, in 1947, Minister Louis Beel (who stood in for Queen Wilhelmina) mentioned the 'sacrifices that have been made' and the 'suffering'.<sup>13</sup> The relatively higher scores of 1945–1948 are, however, also visible in the use of positive emotional language in the queen's speeches (see Graph 3). Positive emotion words are expressedin relation to mentions of international support of the liberation of the Netherlands, and by optimistically anticipating the reconstruction and recovery of the war-torn country.<sup>14</sup>

Overall, individual scores (per 250-word chunk) indicate variation amongst the scores of different chunks of the speeches. The smoothed line, however, displays an overall trend with only minor change over time. The curve points towards a modest decrease in negative emotion word scores from 1945 onwards, until the mid-1960s.

<sup>13 &#</sup>x27;Troonrede 1947', 16 September 1947, https://www.parlement.com/id/vjw7eih0q7q3/troonre de\_1945\_volledige\_tekst.

<sup>14 &#</sup>x27;Troonrede 1947'.

The early 1970s display only a very marginal increase, after which the trend in negative emotion score continues to decline. Linear regression analysis confirmed that there was no strong linear trend observed in both negative and positive emotional language use scores in the queen's speeches. If there was a trend visible between 1945 and 1989, it was only a very marginal decreasing trend that was, in addition, not statistically significant (see also Supplement 2.1.1, Table 18 and Table 19). This means that the observed effect is most likely to have been caused merely by chance. However, when the two graphs (above) are compared, the much higher aggregated positive emotion score is remarkable. Its implications are discussed in more detail further on in this chapter.

#### Newspapers over Time

The Dutch queen's speeches are, by their very nature, a relatively small dataset. Whilst they do cover the period under scrutiny, there is of course only one speech per year.<sup>15</sup> In addition, the speeches are carefully considered, and intended to send a specific message to the Dutch population. As a result, it is a source with a rather limited scope. More meaningful, therefore, are the emotion-mining results based on a larger text collection: a large and diachronic sample of articles in two national Dutch newspapers, namely *De Telegraaf* and *De Volkskrant*. This sample is also more heterogeneous in nature, compared to the queen's speeches, as newspapers contain more diverging voices. As the newspaper sample also consists of more documents, output of this analysis consists of many more data points compared to the queen's speeches. Mean scores for each day (for which data is available) between May 1, 1945, and December 31, 1989, are calculated and visualised on a timeline.<sup>16</sup> The results of emotion mining the newspaper sample sets are plotted in Graph 4 and Graph 5 (negative emotions) and Graph 6 and Graph 7 (positive emotions).

<sup>15</sup> More details on the contents and scope of the datasets used can be found in Chapter 3.

<sup>16</sup> In most cases, data is only available after May 1945. For *De Telegraaf*, the dataset starts at 1949 because the publication of this newspaper was discontinued between 1945 and 1949. See also Chapter 3.

*Graph 4: Negative emotions in Dutch national newspaper De Telegraaf (1949–1989).*

*Graph 5: Negative emotions in Dutch national newspaper De Volkskrant (1945–1989).*

*Graph 7: Positive emotions in Dutch national newspaper De Volkskrant (1945–1989).*

Although there is variation between the mean scores per day (indicated by the swarms of outliers around the trendlines), a first observation is that the historical development of the spread of these outliers is stable over time. In addition, the emotion-mining outputs of the newspaper articles indicate neither a strong increasing trend, nor a strong decreasing trend in the use of emotional words in either sample. The smoothed trendlines for both positive and negative language use in *De Telegraaf* and in*De Volkskrant* are relatively stable over a period of more than four decades. Especially the scores in*De Telegraaf* exhibit practically an 'emotional flat liner' between 1949 and 1989 (see Graph 4). A linear regression analysis supports this conclusion of stability of the scores over time. Although statistically significant effects can be reported for both newspapers and for both positive and negative emotions, the effects are so minor that one could also speak of a 'statistically significant straight line' (see also Supplement 2.1.2). Not much happened here, but this lack of activity was very unlikely to be the result of chance.

The queen's speeches and both samples of newspaper articles have in common that their creation is the result of very deliberate and conscious word-choice processes.Their exact formulations were purposefully constructed and thought through before they were committed to text or, in the case of the speeches, were elaborately prepared in advance of being delivered to the nation. As such, neither the speeches nor the articles can be considered to be the result of spontaneous acts by language users. However, the *Handelingen* dataset is somewhat different in nature. Although, in general, parliamentarians deliver well-prepared and well-informed speeches regarding their topics of discussion, this text collection is the result of a more spontaneous and interactive process – at least in relation to the queen's speeches and the newspapers. This raises a new question, namely: Can the stable emotionality found above also be observed in the Dutch parliamentary debates that took place between 1945 and 1989?

#### Parliamentary Debates over Time

All of the discussions of the Dutch Senate and the House of Representatives that took place between the 25th of September 1945 and the 28th of December 1989 were analysed. The resulting outputs, the mean score of all chunks for each day in which debate took place, are plotted in Graph 8 (aggregated scores of the negative emotions) and Graph 9 (aggregated scores of both positive emotions).

*Graph 9: Positive emotions in Dutch parliamentary debates (1945–1989).*

The results of emotion mining the parliamentary debates show no increase in emotion scores over time.The graphs are, at first sight, roughly similar to the graphs

based on the analysis of newspapers and the queen's speeches.The smoothed trendlines plotted in Graph 8 and Graph 9 display a marginal decreasing general trend in the emotionality of parliamentary discussions over time. Based on linear regression analysis of the parliamentary debates' scores for both positive and negative emotions, a significant yet only very marginal decreasing trend can be observed (see Supplement 2.1.3, Table 24 and Table 25). In addition, the graphs illustrate that the difference between the height of the negative and positive scores is also similar to the other two datasets. Nevertheless, the visualisations also indicate a diverging development in the debates compared to the newspapers and speeches. There is a diachronic development visible in the spread of the scores of the parliamentary debate chunks. This visual representation of the scores displays a minor decline in the amount of variation between the different mean scores (per day). This finding can be an indication of a possible development towards a more generally emotional way of speaking in the Dutch parliament, compared to the more extreme role emotions played in the earlier debates, where the emotionality of the debates varies largely between the individual debates (or days). I will pay more detailed attention to how such developments unfolded in the discussions in Chapters 6, 7, 8, and 9.

#### The Pollyanna Principle

All of the abovementioned graphs display higher scores for positive emotions when compared to negative emotion scores. This higher positive emotion score is mainly based on the higher frequency of words belonging to the emotion category 'trust' from the NRC EmoLex (see also the case study results in Supplements 2.2.2, 3.1.4, and 3.2.4). The difference between positive and negative emotion scores connects to the so-called Pollyanna principle. According to this seemingly universal principle, a tendency exists to expect, remember, and express positive emotions in natural language more often than negative ones.<sup>17</sup> Consequently, language associated with positive emotions is more prevalent in natural language than words reflecting negative emotions.This phenomenon is sometimes also referred to as 'positivity bias'.<sup>18</sup> The observations in all the different historical Dutch text collections discussed above follow this principle. When looking at the graphs of positive and negative emotion scores, it becomes obvious that the aggregated positive scores are higher than the aggregated negative emotion scores.

<sup>17</sup> This principle has been well-established since the 1960s in various academic disciplines. See Jerry Boucher and Charles E. Osgood, 'The Pollyanna Hypothesis', *Journal of Verbal Learning and Verbal Behavior* 8, no. 1 (1 February 1969): 1–8.

<sup>18</sup> Peter Sheridan Dodds et al., 'Human Language Reveals a Universal Positivity Bias', *Proceedings of the National Academy of Sciences* 112, no. 8 (24 February 2015): 2389–94.

#### A Stable Trend

One thing that becomes very apparent from the initial results of emotion mining, it is that the emotion scores, both positive and negative, are relatively stable over time. The use of words associated with emotions seems constantin all the sources between 1945 and 1989. That the curves indicate a lack of substantive historical change does not necessarily imply that there was also a lack of (expressions of) emotionality in the speeches, newspapers, and parliamentary discussions. Despite the lack of an indication of a strong historical linear increase, the trendlines of emotional language use in parliamentary debates display that emotionality was persistent.

Turning now to the minor variations found, Graph 2, Graph 3, Graph 5, Graph 7, and Graph 8 display a slightly higher score in the immediate post-war period for negative emotions, positive emotions, or both. If there is a historical development observable in the results of emotion mining, it is a marginal trend of decrease between 1945 and approximately 1950 in the queen's speeches and the parliamentary debates, rather than a strong increase in the manifestation of emotional language from 1970 onwards. In general, however, no more substantive differences between the results of the analysis of the different text collections are found. The average emotionality scores over time are constant in a similar way (i.e., their trends are similar), and mostly agree in terms of the height of scores. Despite the minor variations observed, the emotion-mining outputs of these datasets give no indication for an overwhelming increase in positive or negative emotional language use between 1945 and 1989. Assuming that any 'emotionalisation' of Dutch society would be reflected in an increase in emotional language, emotion-mining results provide no empirical evidence for a general silencing or withholding of emotions in the 1950s, or for an ongoing 'emotionalisation' that started in the 1970s. The computational analysis of emotions did not provide any convincing evidence for the emotionalisation of Dutch politics and media in the second half of the twentieth century, as suggested in the relevant literature discussed in Chapter 1. What, however, does the observed difference between positive and negative emotion scores mean?

#### Peaks, Troughs, and the Macroscope

Previous studies based on computer-assisted analyses of emotional language in large text corpora found convincing empirical evidence in support of the 'Pollyanna principle'.<sup>19</sup> The results of my computational analysis of emotions in different historical Dutch text collections are, in the first place, another case providing empirical support for the principle. Nevertheless, this result is even more meaningful in

<sup>19</sup> Morin and Acerbi, 'Birth of the Cool', 1670; Dodds et al., 'Human Language Reveals a Universal Positivity Bias', 2390.

another way: This observation can also be considered as a first step in gaining trust in the validity of the emotion-mining methods used. The fact that the output of lexicon-based emotion mining in historical texts has often been used as empirical evidence for the principle means that my findings – which are supportive of the principle – can also be considered as supportive of the validity of emotion mining as a method for the macro-analysis of emotions in the particular context of (the sources used in) this historical investigation.

More important for the remainder of this study, however, is that the results of emotion mining have identified a historical development of relative consistency and stability in the manifestation of emotions in three different historical text collections. This raises a new question: Is lexicon-based emotion mining able to identify variations or peak moments in emotion? Answering this question can be seen as a next step in the evaluation of the validity of emotion mining as a research instrument. This next step is performed with a more empirical and historical evaluation of its validity, abilities, boundaries, and the limitations of applying emotion mining in historical research. If emotionality between 1945 and 1989 is not time-bound, as my findings indicate, is it bound to specific cases or topics instead?

#### **5.3 Perpetrators, Collaboration, and Emotion Peaks**

In the following paragraphs, emotion mining as a historical research instrument is evaluated, using a specific case study. This case study will provide more insight into whether the approach is able to capture the exceptional emotional charge of a well-known, controversial, and elaborately investigated historical case. This is performed utilising a subset of parliamentary debates related to this notoriously emotional case, namely the debates dealing with perpetrators, war criminals, and Nazi collaborators.<sup>20</sup> First, however, I offer a brief overview of the historical background of these parliamentary discussions.

#### A Sense of Urgency and Importance

The parliamentary debates in the 'perpetrators and collaboration' subset sought to address the political problems and public controversies related to Nazi collaboration, perpetration, and the war crimes that occurred during the five years of German occupation.These post-war debates dealt, not only with the judicial processes, prosecution, or punishment of alleged Nazi collaborators and war criminals, but also with the so-called 'purge': the 'purification' of Dutch society from people who were

<sup>20</sup> See also Chapter 1 'Introduction' and: Van Baalen et al., 'Ten geleide', 8–9; Piersma, *De Drie van Breda*, 7, 117–35.

considered 'bad elements', or even 'cancerous tumours', in contemporary terms.<sup>21</sup> The former members of the Dutch national-socialist party were also considered guilty: not by virtue of their concrete deeds or behaviour per se, but rather by association.<sup>22</sup> This 'purge' entailed, more concretely, the process of dismissing people who had collaborated with the enemy between 1940–1945 from public professions or other positions in society. Although the issue was complex and multifaceted, it was dealt with (politically) within the relatively short period of only a decade.<sup>23</sup>

Although former national-socialist party members, 'political delinquents', or people who had otherwise collaborated with the Nazis all constituted different kinds of collaborators, and also despite the fact that Dutch and foreign war criminals cannot be equated to each other, they are treated here as more or less belonging to the same group.<sup>24</sup> This is, at least in the context of this investigation, legitimate: As a group, collaborators and perpetrators were initially characterised by simultaneity (in time) of both their offense and/or crime and their judicial process, conviction, or eventual punishment.<sup>25</sup> A common element that connects all the different aspects of the engagement with 'bad elements' in the Netherlands is the urgency with which these were treated in the immediate post-war context. The Ministries of Justice and Internal Affairs took the lead in preparing policies for punishment and 'purification'. Already during the German occupation, the establishment of tribunals and special courts was considered important for the recovery of the country by the government-in-exile. The task of dealing with collaboration and war crimes was given to specialised tribunals and special courts.<sup>26</sup> After the war, the parliamentarians involved stressed how they considered the organisation of these bodies to be one of the most urgent political issues of the moment. In the words of MP Bernhard Verhoeven (Catholic People Party, KVP), it was considered a matter that '(…) does not lie outside the atmosphere of sentiment'.<sup>27</sup> This strong sense of urgency and

<sup>21</sup> Romijn, *Snel, streng en rechtvaardig*, 12.

<sup>22</sup> Tames and Romijn, 'Transnational Identities of Dutch Nazi-Collaborators and Their Struggle for Integration into the National Community', 250.

<sup>23</sup> For a detailed account of how this was achieved, see Romijn, *Snel, streng en rechtvaardig*.

<sup>24</sup> Both 'political delinquents' and 'purification' are frequently used contemporary terms. 'Political delinquents' was used as an umbrella term to refer to war criminals, collaborators, perpetrators, and former Nazi officials. 'Purification' referred to the process of punishment and incarceration of these people, and thus the 'cleaning' of Dutch society and getting rid of undesired, 'bad elements'.

<sup>25</sup> Although the first statement also applies to 'public figures' who became subject of controversies in the 1950s, 1960s, and 1970s, clearly the latter remark does not always apply to the people belonging to this group. See also Table 12.

<sup>26</sup> Romijn, *Snel, streng en rechtvaardig*, 44.

<sup>27</sup> Bernhard Verhoeven was among the many parliamentarians that mentioned the 'sentimental value' of this topic in parliament in 1951. See 'Handelingen Tweede Kamer 1950–1951', n.d., 1749.

importance was incited by, in the first place, an urge to punish the misbehaviour and 'wrong' practices of perpetrators and collaborators. Second, 'swift, severe, and fair' justice was considered an important potential stimulus for the restoration of moral standards in society. It is therefore not surprising that this was one of the first topics elaborately discussed in the House of Representatives in 1945.<sup>28</sup> These perpetrator and collaboration debates were, however, urgent for another reason, which had much to do with pressing contemporary practical issues.

The mass incarceration of alleged perpetrators and collaborators had serious consequences from a practical, social, and financial perspective. More than 100.000 alleged collaborators and war criminals had been interned in the protracted start to liberation in September 1944. During a so-called 'internment wave', which lasted until the summer of 1945, these people were incarcerated in makeshift internment camps throughout the country.<sup>29</sup> After the re-instalment of the Dutch parliament in the autumn of 1945, the miserable circumstances under which most of the suspects lived soon became one of the main topics of discussion.<sup>30</sup> Even after a majority of the (alleged) collaborators had been released from internment, not all concerns, problems, or potential triggers for discord and controversy were gone. The consequences and implications of the measures aimed at 'political delinquents' repeatedly led to the rise of new political issues. For example, in the late 1940s, it was the issue of statelessness that incited political discord and worry.<sup>31</sup> Over the years, the issue of political delinquents became a recurring theme on the political agendain the Netherlands. The overwhelming majority of the 100.000 (alleged) collaborators had either been acquitted or been released relatively quickly.<sup>32</sup> Contemporaries considered too strict a punishment – especially for domestic collaborators – undesirable and counter-productive in supporting national unity, harmony, and the restoration of family life. Consequently, the number of people still incarcerated had decreased spectacularly by the first half of the 1950s. Nonetheless, a small yet dwindling number of people remained behind bars for decades.<sup>33</sup>

#### Symbolic and Abstract

The last remaining war criminals behind bars in the Netherlands were, in the words of historian Peter Romijn, treated as a 'category *sui generis*': The official and national

<sup>28 &#</sup>x27;Handelingen Tweede Kamer 1945–1946', n.d., 83–102; Romijn, *Snel, streng en rechtvaardig*, 9–11.

<sup>29</sup> Grevers, *Van landverraders tot goede vaderlanders*, 138–39.

<sup>30</sup> Tames, *Doorn in het vlees*, 87; 'Handelingen Tweede Kamer 1946', n.d., 73–88.

<sup>31</sup> Tames, *Doorn in het vlees*, 87–88.

<sup>32</sup> Tames, *Besmette jeugd*, 14.

<sup>33</sup> Ralf Futselaar, *Gevangenissen in oorlogstijd: 1940–1945*, 1st ed. (Amsterdam: Boom, 2015), 230, 235–36; Van Lange and Futselaar, 'Debating Evil'.

engagement with these people was (and remained) exceptional. For example, the ministers responsible needed to discuss plans for the interruption of punishments or proposals for individual clemency with the Dutch parliament.<sup>34</sup> Discussing such plans often became a flashpoint for extensive media attention, public controversy, and protests. The fundamental nature of the issue, however, had changed. In contrast with the immediate post-war period, when the thousands of incarcerated suspects represented a very present societal problem and an immediate crisis, the topic became a matter that was predominantly symbolic in its meaning and importance. According to the German historian Harald Fühner, dealing with these issues from the late 1950s onwards in the Netherlands had become less about the war criminals themselves, and more about the values they symbolised.<sup>35</sup> The possible release of the last few 'tough guys' could,in the 1970s, not be considered as a serious societal threat anymore. The last foreign war criminals incarcerated in the Netherlands – since the 1970s known as the 'Breda Three' – became the symbolic embodiment of wartime evil.

The symbolic meaning attached to Nazi war criminals seems to have also affected public and political engagements with 'milder cases': former collaborators, or former members of the Dutch national-socialist party. Wartime collaboration was increasingly understood in more comprehensive, broader, and abstract terms in subsequent decades. From the second half of the 1960s onwards, people who were being 'exposed' as former national-socialists because of their past behaviour or crimes, or even only by affiliation or attitude, were increasingly directly associated with the horrors of the Holocaust and concentration camps.<sup>36</sup> This often involved individuals in positions of high public office. These were people 'in the spotlight' who either had something to lose or were seen as part of contemporary 'elites'. When new information came to light (sometimes deliberately, sometimes more or less accidentally), this often led to controversy, public protests, and elaborate media coverage. Often such cases also led to debate in the Dutch parliament, as was the case for, amongst others, Mayor Frans Schokking (1956), the Dutch art collector Pieter Menten (1976–1980), and the politicians Joseph Luns (1978–1979) and Willem Aantjes (1979). Not only by later commentators, but also at the time, these cases were often and repeatedly classified as exceptional emotional peak moments. The parliamentary discussion of, for example, the Menten case (an art collector suspected of committing war crimes) is generally considered to be not only one of the most elaborate, but also one of the most spectacular debates in Dutch

<sup>34</sup> Romijn, *Snel, streng en rechtvaardig*, 271.

<sup>35</sup> Fühner, *Nachspiel*, 441–43.

<sup>36</sup> Tames, *Doorn in het vlees*, 280–81, 353–55.

parliamentary history.<sup>37</sup> Dries van Agt, minister of justice (at that time), considered the incident one of the most difficult parliamentary discussions of his career. Later reflecting on his job and on such cases, Van Agt stated that '(…) anything that makes this country and its population emotional (…) always takes place within the domain of the Ministry of Justice'.<sup>38</sup>

#### The Breda Four, Three, and Two

The 'tough guys' mentioned previously were actually a small group of foreign war criminals incarcerated in the panopticon prison in Breda. As mentioned in Chapter 1, these prisoners repeatedly became flashpoints of intense public unrest, media attention, and political discord. For as long as these notorious former Nazi officials and officers remained behind bars, plans for their release continued to resurface and cause controversy.<sup>39</sup> In 1952, German war criminal Willy Lages requested – and received – clemency. Lages did not become a free man, but his death sentence was commuted to life imprisonment. The commutation of Lages' sentence came about under pressure from Queen Juliana, who strongly opposed the death penalty.<sup>40</sup> In response, more than 15.000 people protested in Amsterdam against the ministerial decision.<sup>41</sup> An interpellation debate in the House of Representatives followed. Many MPs deemed the decision a disgrace. The communists in particular used the case to mobilise support, and to profile itself as the 'resistance party'.<sup>42</sup> In the 1952 debate, the communist parliamentarian Henk Gortzak exclaimed that Lages was 'the greatest of all villains!'<sup>43</sup> It was, however, too little, too late. Lages had already received clemency. A few weeks after the Lages controversy, seven other war criminals escaped the Breda prison during the Christmas dinner.<sup>44</sup> This again led to parliamentary debate, as Dutch parliament demanded transparency and clarification about the circumstances that had made the escape of the seven Dutch collaborators and war criminals possible.<sup>45</sup> The Lages controversy and the prison escape of 1952 are characteristic for war criminal-related controversies in the 1950s.<sup>46</sup> In the years that

<sup>37</sup> Johan van Merriënboer, Peter Bootsma, and Peter van Griensven, *Van Agt biografie: tour deforce* (Amsterdam: Boom, 2008), 191.

<sup>38</sup> Van Merriënboer, Bootsma, and van Griensven, 199.

<sup>39</sup> Van Lange and Futselaar, 'Debating Evil', 141.

<sup>40</sup> De Haan, *Na de ondergang*, 105.

<sup>41</sup> Piersma, *De Drie van Breda*, 51.

<sup>42</sup> Fühner, *Nachspiel*, 257–59.

<sup>43 &#</sup>x27;Handelingen Tweede Kamer 1952–1953', n.d., 2024.

<sup>44</sup> Tames, *Besmette jeugd*, 190–91.

<sup>45 &#</sup>x27;HTK 1952–53', 2025.

<sup>46</sup> Tames, *Besmette jeugd*, 190–92.

followed, other former German officials incarcerated in the Breda prison became flashpoints of war criminal-related controversies.

In the summer of 1966, Lages again triggered public turmoil. He had fallen ill. Therefore, the Dutch Minister of Justice Ivo Samkalden (Labour party, PvdA) granted him a suspension of his prison sentence. In response, the following day, people once more marched through the streets of Amsterdam to demonstrate against the decision.<sup>47</sup> The communist MP Marcus Bakker submitted a request for an interpellation debate. Lages, however, had already left the country to undergomedical treatmentin Germany. Just as in 1952,many protesters marched the streets.<sup>48</sup> And Lages? Against all expectations, Lages recovered and never returned to the Breda prison, living in Germany until his death in 1971. Following Lages' departure, the three last remaining foreign war criminals in the Netherlands – Ferdinand aus den Fünten, Franz Fischer, and Joseph Kotälla – became known as the 'Breda Three'. They became central to one of the best-known and much-discussed World War II-related cases in the Netherlands.<sup>49</sup>

The then minister of justice, Dries van Agt, considered clemency (which now meant release from prison) for the Breda Three in 1972. Just as in 1952 and 1966, controversy arose. Not only did Van Agt's proposal stir elaborate media attention; it also triggered a latent disposition as it led to powerful emotional responses. As mentioned in Chapter 1, this again led to mass protests and strong expressions of public emotion.<sup>50</sup> It was apparent that the emotions (still) circulating in society did not resonate with Minister Van Agt's proposal to release the Breda Three. In the parliamentary discussion that followed, these circulating emotions were echoed, not only by MPs opposing Van Agt's proposal, but also in the active interventions into the discussion by mental health professionals specialised in war trauma. The debate, which lasted for 13 hours, was characterised by heart-breaking and emotional personal stories.<sup>51</sup> In 1952 and 1966, it had been too little, too late: Clemency had already been granted to Lages in 1952, and he had already left the country in 1966 by the time MPs gathered for discussion. In 1972, however, things took a different turn. Mass protests, interventions from mental health professionals and interest groups for victims, and the elaborate and emotional parliamentary discussion itself eventually led to the cancellation of the release of the Breda Three in 1972. As a key figure in the decision-making process and, as a result, also instigator of public anger, Van Agt

51 Piersma, 117–35.

<sup>47 &#</sup>x27;Protesttelegrammen en demonstratie: Interpellatie over Lages in Kamer', *Friese koerier: onafhankelijk dagblad voor Friesland en aangrenzende gebieden*, 10 June 1966, Day edition, https://re solver.kb.nl/resolve?urn=ddd:010690209.

<sup>48</sup> Fühner, *Nachspiel*, 257–59.

<sup>49</sup> Piersma, *De Drie van Breda*, 7–12.

<sup>50</sup> Piersma, 7.

later described the parliamentary hearing on the BredaThree as 'the most emotional hours of his political life'.<sup>52</sup>

After Kotälla died in 1979, it took another ten years before the remaining Breda Two were again elaborately debated in parliament. Piersma stated that the 1989 parliamentary discussion on their possible release initially resembled that of 1972: The argument of the rule of law was juxtaposed with a disposition of general solidarity with war victims. The latter once more became apparent in the form of protests demonstrating strong public expressions of emotions that were still circulating in Dutch society. However, this time both the government and those supportive of the release in the Dutch parliament did everything to avoid any escalation of the issue. Aus den Fünten and Fisher were eventually released from the Breda prison on the 27th of January, 1989. With this decision, political and parliamentary discussions on World War II criminals incarcerated in the Netherlands came to an end.<sup>53</sup>

As discussed in Chapter 1, the extensive contemporary media attention, emotional public protests, and heated political debate of the BredaThree case is generally considered to be one of the most notable and emotional World War II-related issues experienced the Netherlands during the second half of the twentieth century. Not only does this topic frequently resurface in elaborate and detailed historical studies of the multiple collaborator- or war criminal-related cases and controversies, it is also regarded as a particularly emotional episode in the contemporary history of the Netherlands. This is also true of the associated parliamentary discussions. Not only do the testimonies of those who directly experienced these events support this characterisation, it is also justified by the results of elaborate historical research on this particular theme.<sup>54</sup> It is therefore evident that these debates can be considered emotional. The following short section focuses on how the emotionality of the issue, as measured with emotion mining, developed over time. More importantly, however, is that the case serves as an important resource in the evaluation of the validity of the measuring instrument itself. Central is the question whether these notoriously emotionally charged debates also score or can be quantified as emotional with the lexicon-based approach to mining historical emotions. Are the expected patterns confirmed by the results of computational methods? Does the output of emotion mining replicate the supposed emotional charge experienced at the time when collaborators and war criminals in post-war Dutch debates were discussed?

<sup>52</sup> Van Baalen et al., 'Ten geleide', 8–9.

<sup>53</sup> Piersma, *De Drie van Breda*, 179, 188.

<sup>54</sup> See Harald Fühner and Johan Willem Friso Wielenga, 'De Drie van Breda: strijd tussen leed en genade', in *Zimmer Frei: Nederland-Duitsland na 1945*, 2001, 42–45; Piersma, *De Drie van Breda*; Fühner, *Nachspiel*; Tames, *Doorn in het vlees*; Grevers, *Van landverraders tot goede vaderlanders*.

#### **5.4 Results**

#### The 'Perpetrators and Collaboration' Subset

The debates related to the topics named in the schematic overview below are separated from the entire *Handelingen* dataset. Debates belonging to this subset are scored on the manifestation of words associated with the positive emotions joy and trust, and the negative emotions anger, fear, disgust, and sadness from the NRC EmoLex. For more details on the methodological workflow, see Chapter 4.


*Table 12: Thematic clusters within the 'perpetrators and collaboration' subset of parliamentary debates, 1945–1989.*

The 'perpetrators and collaboration' subset is separated from the bulk of parliamentary debates from the 1945–1989 period. The selection procedure described in Chapter 3 is used. All other debates, unrelated to the topics in the scheme, are also taken into account in the results presented below. From now on, the unrelated majority of parliamentary debates will be referred to as the 'non-war-related', 'non-related' or 'non-relevant' debates set. This text collection forms the background of 'average parliamentary discussion'. The scores of this set are considered a baseline of emotional language, against which the scores of the case studies or subsets are contrasted. The scores of particular subsets of debates that diverge from this baseline are higher or lower and are thus considered exceptional in comparison to the overall score of 'average parliamentary discussion' in the 1945–1989 period. How do the debates of the subsets score on emotional language?

#### Summary Statistics and Scores over Time

The 98 different documents containing the retrieved perpetrators and collaboration debates are all scored. The summary statistics of the results of emotion mining are calculated according to the procedure described in Chapter 4. The output is presented below in Table 13.

*Table 13: Summary statistics of emotion scores of the 'perpetrators and collaboratoration' subset (1945–1989).*


The (same) results of emotion mining on this particular subset are also plotted together with the average, non-war-related parliamentary debate scores in Graphs 9 (joy) and 10 (anger). Plots on all different NRC EmoLex emotion categories can be found in Supplement 2.2.2. Just as in the previous paragraphs on 'emotionalisation', the grey dots represent the mean scores (per day in which debate took place) of 'average parliamentary discussion' between 1945 and 1989. The black line drawn through the scatterplot in Graph 10 represents the smoothed scores of all these non-war-related debates. The trendline is, again, created using LOESS smoothing in R. In contrast to the earlier graphsin section 5.2,Graph 10 also displays the mean scores of the particular debates on collaborators and war criminals. The mean joy scores for the

war criminal and collaborator debates are plotted as yellow bars. The general trend over time (LOESS) of these scores is plotted as a yellow line.

*Graph 10: Joy scores of the 'perpetrators and collaboration' debates subset (yellow) compared to average parliamentary debate (light grey and black), 1945–1989.*

Graph 10 displays a relative stable course in the development of language use associated with joy in debates on 'perpetrators and collaboration'in Dutch parliament. This also applies in the case of trust scores (see Supplement 2.2.2). As this visualisation also shows, the height of the joy scores of these thematic debates roughly corresponds to the average parliamentary discussions on other issues.The debates on war criminals, perpetrators, and collaboration do not stand out in terms of the manifestation of language associated with positive emotions. Joy and trust were, however, presumably not the emotions that Minister Van Agt referred to when he called the Breda Three debate the most emotional episode of his political career. In contrast to positive emotions, an emotion such as anger is generally expected to be a stronger marker in distinguishing debates on war, atrocities, perpetration, and victimhood from more general political issues. As with sadness, anger is generally more strongly

connected to narratives of war and mass violence.<sup>55</sup> What about the negative emotion scores of collaborator and war criminal debates?

#### The Anger Anomaly

In Graph 11, the scores of language use in the 'perpetrators and collaboration' debates associated with the negative emotion anger are plotted. The graphs with results of the analysis of sadness, fear, and disgust can be found in Supplement 2.2.2.

*Graph 11: Anger scores of the 'perpetrators and collaboration' subset (red) compared to average parliamentary debate (light grey and black), 1945–1989.*

Some distinguishing characteristics of this subset of parliamentary debates become apparent from these visual representations. First, the difference between scores for joy and anger is immediately obvious. Second, the anger scores for the debates on 'perpetrators and collaboration' (red bars) are significantly higher than most non-war-related debates from the period 1945–1989. The smoothed trendline of the anger scores of this subset (red line) is significantly higher than the average anger score over time of the non-war-related parliamentary topics (black line). Graph 11 is indicative of the exceptionality of discussing topics related to war

<sup>55</sup> Thomas Brudholm and Johannes Lang, 'Introduction: Emotions and Mass Atrocity', in *Emotions and Mass Atrocity: Philosophical and Theoretical Explorations*, ed. Thomas Brudholm and Johannes Lang (Cambridge and New York: Cambridge University Press, 2019), 14.

criminals, perpetrators, and collaboration during the German occupation of the Netherlands in the post-war parliamentary context. Almost all debates on this topic display negative emotionality scores far exceeding the average for parliamentary discussions on other (unrelated) issues. The high anger scores of the perpetrator and collaboration debates are amongst the highest scores of the whole period from 1945 to 1989, as is the case for all negative emotions. In addition, the scores for language use associated with the negative emotions fear and sadness are exceptionally high in the Dutch parliamentary debates on 'perpetrators and collaboration' held between 1945 and 1989. The same applies in the case of disgust – an emotion that is usually much lower scoring than the other negative emotions (for graphs, see also Supplement 2.2.2).

A historical peak in the *number* of debates on this issue in parliament is reflected inGraph 10 andGraph 11.The coloured bars, each representing anindividual collaborator or war criminal debate, are not evenly distributed over the 1945–1989 timeline. Whereas debates of the subset in the initial post-war years potentially dealt with more than 100.000 'bad elements' who were incarcerated, the later discussions were mostly centred around individual (alleged) collaborators and war criminals.The distribution of debates is a consequence of this change – from ongoing discussions on a fundamental and pressing issue right after 1945, to incidental and regularly recurring debates on an issue that would later acquire a more symbolic meaning. The German war criminal Willy Lages, for example, stirred turmoil and discussions in 1952 and 1966. His fellow inmates did the same in 1972 and 1989. This is also true of the controversies regarding (unmasking) the war pasts of prominent individuals, such as Schokking (1956), Menten (1976–1980), Luns (1978–1979), and Aantjes (1979). Compared to the immediate crisis management of the immediate post-war years, these later parliamentary discussions arose as incidents, rather than being considered part of protracted ongoing parliamentary discussion on a major societal problem.

#### Final Observations: Emotion Mining and 'Bad Elements'

Regarding the manifestation of the positive emotions of joy and trust, the 'perpetrators and collaboration' debates score in the same range as non-relevant parliamentary discussions in the second half of the twentieth century. Although these scores are not exceptional,it stillmeans those debates contain a substantive amount of positive emotional language.When comparing this particular set of debates to the nonrelevant parliamentary discussionin terms of negative emotion scores, another conclusion emerges. Negative emotion scores are exceptionally high. Results confirm the expected pattern(based on the elaborate historical research on theissue). In general, I consider the Dutch parliamentary debates on 'perpetrators and collaboration' extreme outliers regarding the manifestation of language associated with negative emotions. These results of emotion mining suggest that these debates made extensive use of words associated with anger, fear, sadness, and disgust.

Graph 10 and Graph 11 both support the aforementioned finding that the emotionality of the debates, also in those related to the particular 'perpetrators and collaboration' subset, remained relatively stable over time.This applies to both positive and negative emotions. The only exception is the 1989 debate on the release of the Breda Two, which revealed high-scoring negative outliers compared to earlier debates. Nevertheless, linear regression analysis of the results of the 'perpetrators and collaboration' debates points overwhelmingly at only very marginal,mostly non-significant and negative effects (decreasing trends). These results of linear regression analysis are shown in the tables in Supplement 2.2.3.When the Breda Three debates of 1972 are compared to both the preceding and subsequent debates on collaborators and war criminals, the case was not so unique in terms of exceptional high scores for negative emotional language use. It seems that the notoriously emotional charge of the 1972 Breda Three debate was strengthened by retrospective processes of meaning-making – as former minister Van Agt illustrated when, on reflection, he called this 'the most emotional hours of his career'.<sup>56</sup>

Emotion mining revealed that most debates on political delinquents between 1945 and 1989 are exceptional high peaks when it comes to negative emotion scores. The quantitative output of emotion mining supports the case for negative emotion as a relative constant, as, also in the debates on political delinquents, these negative emotion scores were a constant high. These results indicate that earlier debates on war criminals and collaboration in the late 1940s, the 1950s, and the 1960s were also imbued with a negative emotional charge. Yet they were no less emotional than the well-known controversies that were debated in the Dutch parliament in the early 1970s. This is also true of the controversies surrounding public figures in the late 1970s and 1980s, and the 'conclusive' debate on the Breda Two in 1989.

#### **5.5 Conclusion**

My analysis of the use of emotional language in the Dutch parliament, newspapers, and queen's speeches provided the baselines of emotion scores for the period from 1945 to 1989. In this chapter, two important observations lived up to my expectations. These findings are important signifiers of the validity of the method for the investigation of historical emotions in the chapters that follow. A third observation in this chapter was, however, less congruent with existing knowledge.

First, the results of the lexicon-based emotion mining found support for the universal tendency of human communication to express positive terms more fre-

<sup>56</sup> Van Baalen et al., 'Ten geleide', 8–9.

quently than negative ones. Quantitative empirical evidence was found for the so-called 'Pollyanna principle'. I argue that this outcome corroborates the validity of emotion mining as a technique for distant and quantitative historical text analysis. Multiple studies based on a computational analysis of emotional language in largescale textual databases had already provided quantitative empirical evidence supporting the Pollyanna principle.<sup>57</sup> The fact that the output of lexicon-based emotion mining on various historical Dutch text collections produced further empirical evidence means the findings support the validity of emotion mining as a method for the analysis of emotions in this particular historical research context.

Second, the question of whether lexicon-based emotion mining would be able to identify well-known peaks in emotionality in historical Dutch texts at all, was addressed.The notoriously emotional parliamentary discussion of post-war cases and controversies related to 'perpetrators and collaboration' was used to further evaluate emotion mining as a method for historical research. Resurfacing controversies, and the notorious 1952 Lages and 1972 Breda Three debates in particular, are not only well-known, but also elaborately studied and discussed in historiography. Their emotional charge can be considered evident. In the first place, the emotionmining results supported the claim of the exceptional emotionality of these discussions when it comes to negative emotions such as anger, sadness, fear, and disgust. Lexicon-based emotion mining was thus able to capture the exceptional nature of the emotional charge of this well-known case study in comparison with other, nonrelated parliamentary debates(i.e., the baseline of emotionin parliament).Although the BredaThree case of 1972 in particular is one of the best-known controversies, this did not mean it was also the most emotional of the discussions in Dutch parliament. The 1972 case was not unique in its high emotional charge. Multiple parliamentary discussions on 'purification' in the 1940s, on the Lages cases in 1952 and 1966, on several prominent 'public figures', and those on the eventual release of the final two foreign war criminals in 1989 resemble the 1972 case regarding the exceptional high scores for language use associated with negative emotions.

A third observation is less congruent with received wisdoms or commonplace views. The results of applying emotion mining to different text collections did not provide convincing evidence for the claim of a strong and ongoing emotionalisation of Dutch society from the 1970s onwards. Assuming that such a change would be reflected in an increase in emotional language use, emotion-mining results provided no evidence for the emotionalisation of Dutch society and politics, or the sudden and ongoing emergence of an emotion culture in the early 1970s, as commonly described in historiography. These results also contradict received wisdom regarding a general silencing or withholding of emotions in the 1950s. If there was any peak

<sup>57</sup> Morin and Acerbi, 'Birth of the Cool', 1670; Dodds et al., 'Human Language Reveals a Universal Positivity Bias', 2390.

observable in the use of emotional words, this took place in the late 1940s, rather than starting and taking off from the early 1970s onwards. In addition, based on a closer scrutiny of the queen's speeches, this peak had more to do with the Dutch-Indonesian War, than with the consequences of World War II in particular.The results of emotion mining showed that emotion was a relative constant in parliamentary debates, newspapers, and queen's speeches between 1945 and 1989.

Now the validity of lexicon-based emotion mining has been evaluated in a twostep process, the contours of the historical background are visible, and the baselines have been drawn, it is time to move on to the parliamentary issues that are central to this investigation. In the next chapters, I will address the role of emotions in debates on the so-called 'extraordinary government employees' – the different groups that actively worked against the Nazi occupier – and diverse groups of war victims following war, persecution, and occupation under Nazi rule in the Netherlands (1940–1945).

## **Case Study 1 | 'The Resistance'**

The computer-assisted workflow, based on emotion mining to identify, measure, and compare emotionality of texts, is applied to two different case studies in the following chapters. The baselines of emotional language use in the Dutch parliament have been set in Chapter 5. For each thematic case study, a subset of parliamentary discussions from the period 1945–1989 was retrieved from the *Handelingen* (see also Chapter 3). Chapters 6 and 7 address a first case study and deal with the parliamentary discussion of national legislative schemes aimed at a group of people that were retrospectively regarded as 'soldiers' outside the regular national army. The former anti-Nazi resistance received much public and governmental attention from the first years after liberation onwards.These people had been actively involved with resistance activities or had been working against the Nazi occupier in other ways. Sometimes, this led to their death, or to lasting illness that affected their circumstances (or those of their families) long after the war. They were, by the postwar government, treated as veterans or former soldiers. The Resistance was used to legitimise the still precarious post-war political order. To whom did the Resistance belong? Who had been part of this group?

Although it rolls off the tongue, the actual groups of people which 'the Resistance' refers to is (and was) not easy to delineate. In wartime, there had never been a singular, homogenous, or unanimous group that acted united under the term 'the Resistance'. Resistance against the Nazi occupation in the Netherlands had consisted of various acts of a heterogeneous conglomerate ofindividuals,initiatives, and groups. The notion of a homogeneous unity should be considered as constructed. 'The Resistance' was used as a reference to position oneself – and others. Unity of the notion of 'the Resistance' was, both during and after the war, needed for the government and various other groups of people to express themselves or to claim legitimacy and power. This, however, by no means diminishes the actual deeds and actions of a whole range of people obstructing the occupation and its persecution of groups within society – most notably those people labelled as Jewish.

Each case study in this investigation consists of three parts. The first part introduces the political and societal background of the history of national legislation aimed at 'the Resistance' and, in the second case study, war victim legislation. Legislative schemes were the primary reason why parliamentary debate on 'war survivors' took place. The establishment and/or modification of legislative schemes are also an opportunity to gain insights in how things worked at certain moments in time. Thereby, the legislation itself is not considered as an expression of emotion, or as representative of a contemporary 'mental climate'. It is expected the role of emotions can be encountered in the debates that dealt with relevant legislation over the years. The second part of each case study proceeds with the quantitative results of applying emotion mining to the particular subset of parliamentary debates. This distant perspective, in which the NRC EmoLex is used to identify and score emotional language use, allows for a comparison with the emotion scores of other parliamentary discussions and case studies. It also offers the opportunity to evaluate variation between the debates of thematic clusters within each subset and, to conclude, to gain insights in the historical development of emotion scores over time. The emotion-mining results are presented in Chapter 6 (for case study 1) and Chapter 8 (for case study 2). The third part of each case study, the more detailed, closereading analysis of the role of emotions in these parliamentary discussions, will be discussed in subsequent Chapter 7 for the first case study and in Chapter 9 for the second.

#### **A History of Resistance Legislation (1947–1985)**

The Dutch government had resided in exile in London during the occupation years, in contrast to the countries where the government had remained more or less in charge of the country, albeit under control of the Third Reich, such as Denmark. Therefore, the Dutch government and monarchy tried to keep in touch with underground resistance organisations back in the occupied country, and with so-called *Engelandvaarders.*<sup>1</sup> In order to continue the war against Nazi-Germany on the side of the allies, anticipating an allied liberation and possible administrative chaos, the Dutch government attempted to merge a diverse conglomerate of different resistance groups into some sort of national army. After D-day in 1944, some major resistance organisationsmergedinto the*Binnenlandse Strijdkrachten*(BS, 'Internal Forces') under command of Prince Bernhard of the Netherlands. Although even the BS was hardly a unity in practice, it symbolised the assignment of some sort of national military status to the members of the underground resistance organisations in the Netherlands.

On May 5th, 1945, Liberation Day, the German forces in the Netherlands surrendered and the entire country was officially liberated. Prime Minister Pieter Sjoerds

<sup>1</sup> *Engelandvaarders*were Dutch people who had left the occupied country for the UK by successfully crossing the channel.

Gerbrandy handed in the resignation of the London government. Influenced by Dutch Queen Wilhelmina, he had made a promise during the occupation: after liberation, the government-in-exile had to make place for leaders who had experienced the occupation years in the Netherlands themselves. Queen Wilhelmina was a supporter of eliminating the 'pillarised' pre-war social and political structures, which were organised alongside religious and ideological lines. The 'pillars' were perceived by many as fragmenting and weakening the nation.<sup>2</sup> The Queen brought together former resistance fighters and leaders, most of them willing to work towards 'renewal'.<sup>3</sup> Representatives of most pre-war political parties were invited to shape this government. Dutch communists, who had had an exceptional high share in the anti-Nazi resistance, were included in the negotiations. Soon however, tensions arose. The Communist Party of the Netherlands (CPN) and the Orthodox Protestant Party (ARP) dropped out of the coalition negotiations. In June 1945, Queen Wilhelmina nevertheless appointed a temporary government, led by social democrats Wim Schermerhorn and Willem Drees. <sup>4</sup> It is assumed that the former represented a more progressive perspective, the latter a traditional approach to politics.<sup>5</sup> This first post-war government was a mix between members in favour of renewal, and more conservative spirits that harked back to the pre-war situation.<sup>6</sup> The motto of this government was: 'reconstruction where necessary and renewal where possible'.<sup>7</sup> Nevertheless, this did not lead to the immediate establishment of a stable post-war political order, or a functioning parliamentary democracy.

There was another issue at stake as well. This issue was similar to that of other former occupied countries in Europa, liberated by the allied forces, such as Norway, Denmark, Belgium, and France: the resurgent post-war political order was in need of legitimacy. These national governments therefore all strived to associate themselves with 'internal victors', with those people representing – what was considered as – a just normative behavioural attitude during the recent years of enemy occupation. A group of former resistance fighters seemed to offer an opportunity, as they symbolised something new, something internal or national, and had the cultural

<sup>2</sup> Pieter Lagrou, *The Legacy of Nazi Occupation: Patriotic Memory and National Recovery in Western Europe, 1945–1965* (Camrbidge: Cambridge University Press, 1999), 23.

<sup>3</sup> Jac Bosmans and Alexander van Kessel, *Parlementaire geschiedenis van Nederland* (Amsterdam: Boom, 2011), 54.

<sup>4</sup> Schermerhorn became Prime Minister because Drees declined the position on his own accord. On June 25, 1945 the Schermerhorn-Drees administration took office.

<sup>5</sup> Schermerhorn was a professor from Delft, who worked on his progressive plans for a renewed country already during the war in the early 1940s. See also Bosmans and Van Kessel, *Parlementaire geschiedenis van Nederland*, 55.

<sup>6</sup> A more elaborate history of the first post-war Dutch government can be found in Bosmans and Van Kessel, *Parlementaire geschiedenis van Nederland*.

<sup>7</sup> Bosmans and Van Kessel, 56.

capital necessary to promote legitimacy of the new post-war political order.<sup>8</sup> This meant that fostering close links to former resistance networks, or including them in the political recovery could signal how (new) political leaders were both on the side of the winners of this war and on the side of 'the best elements' of the nation. On the other hand, just as in abovementioned countries, the Netherlands had a conglomerate ofindividuals,initiatives,and groups of formermembers of the anti-Nazi resistance that was eager to gain political power and influence.<sup>9</sup> There had to be negotiated in the Netherlands. What position should the former resistance get in the post-war society?

#### 'The Resistance' in Post-war Politics

The need for legitimacy led to a desire to incorporate the 'spirit of resistance' in the post-war governmental order. In early post-war years, there had initially even been thoughts about appointing a special 'Minister of the Resistance'.The Minister would have dealt with the post-war care for former members of the resistance and their descendants – just as happened in Denmark.<sup>10</sup> In the Netherlands,it never got that far. Assigning substantive political power to former resisters (merely) because of their behaviour and attitude during the occupation was out of the question for Schermerhorn. Distrust regarding an alleged adventurous or 'on the loose' spirit in the resistance may have influenced this decision. Schermerhorn said that participation in national politics simply demanded different personal qualities than working against the Nazi occupier had demanded: it did not make people automatically fit for doing politics.<sup>11</sup> There had to be another way in which the post-war political order could associate itself with the former resistance movements. What would appropriate their spirit, but limit the actual political power assigned to the former resistance?

A possible answer was already developed during the occupation.The Dutch government-in-exile had, back then, proposed arrangements for people suffering because of anti-Nazi activities. The Dutch government had promised that no family

<sup>8</sup> Lagrou, *The Legacy of Nazi Occupation*, 22; Ralf Futselaar, 'From Camp to Claim − The KZ Syndrome and PTSD in Scandinavia, 1945–2010', in *The Politics of War Trauma: The Aftermath of World War II in Eleven European Countries*, ed. Jolande Withuis and Annet Mooij (Amsterdam: Amsterdam University Press, 2010), 249; Robert Gildea and Olga Manojlović Pintar, 'Afterlives and Memories', in *Fighters across Frontiers Transnational Resistance in Europe, 1936–48*, ed. Robert Gildea and Ismee Tames, 1st ed. (Manchester: Manchester University Press, 2020), 215–18.

<sup>9</sup> Futselaar, 'From Camp to Claim', 248–49.

<sup>10</sup> Futselaar, 248–55.

<sup>11</sup> Because this issue of political power for the resistance was mainly discussed in the government, it took place outside the parliamentary context discussed further on in this chapter. See Piersma, *Bevochten recht*, 24–25.

should worry about the future, as the nation held an *ereschuld* ('debt of honour') towards those who sacrificed their health or their lives for their country.<sup>12</sup> Louis Beel, Minister of Internal Affairs(Catholic People Party,KVP)in the first post-war cabinet, was responsible for the execution of these intentions. According to the standards of those days, Beel felt that what private initiatives and non-governmental organisations could solve, should not become a task of the national government.<sup>13</sup> There was one exception. With his colleague, Minister of Finance Piet Lieftink (Christian Historical Union, CHU), Beel thought there should be some sort of national arrangement for those who '(…) in the conduct of resistance activities, i.e. as a direct consequence of their active fight against the enemy, or as a result of ill-treatment by the enemy, have undergone wholly or partly invalidity on the account of their resistance attitude'.<sup>14</sup> This belief, combined with the quest for political legitimacy outlined above, lay the foundations for a proposal for an extraordinary social benefit scheme named*Wet Buitengewoon Pensioen 1940–1945* (WBP),or Extraordinary Pension Act 1940–1945.

#### Extraordinary Government Employees

The Extraordinary Pension Act 1940–1945 was indeed something extraordinary.This type of social welfare legislation was not at all common in the late 1940s. Usually, people who could not earn a living because of invalidity were dependent on charity organisations or the church. The establishment of the WBP, which guaranteed an income for those people (or their descendants) who had become unable to support themselves because of invalidity or death caused by anti-Nazi resistance activities, was considered as a pay-off of a 'debt of honour'. The post-war Dutch government took on this debt to create a connection between the sacrifices these people had made for the fatherland and the recovering post-war political order. The proposal for this extraordinary act can therefore also be understood as a result of the felt need for political legitimacy, by appropriating the former resistance and by highlighting a specific notion of moral and political virtue.<sup>15</sup> The WBP proposed to award a government pension to people without the necessity of formal government employment. There had been no official contract. No salary had been paid. The pension scheme built on the idea of a fictive and retroactive state employment. Encouraged by the clandestine *Radio Oranje* ('Radio Orange') the resistance had acted voluntarily, with-

<sup>12</sup> Withuis and Mooij, 'From Totalitarianism to Trauma', 195.

<sup>13</sup> Piersma, *Bevochten recht*, 26.

<sup>14</sup> Piersma, 30.

<sup>15</sup> Ludi, 'Who Is a Nazi Victim?', 13.

out being in formal military service in the occupation years.<sup>16</sup> Former members of the resistance were treated as being retroactively in military service of the kingdom of the Netherlands during the German occupation.<sup>17</sup> This status did not convey actual political power, but provided the former resistance with rights to – by contemporary standards – exceptional social legislation.

According to the official legislative formulations of the WBP act, the criterion for eligibility was 'having acted against the Nazi occupier'. This 'virtue' was used as common denominator in an attempt to bridge gaps between different resistance individuals, groups, 'pillars', ideologies, and between the former resistance and the contemporary post-war political order. Such a 'shared virtue' not only had to bridge differences, it also had to promote reconciliation and national unity in the post-war Netherlands. Generally, however, in the practical implementation of acts such as the WBP, such attempts often collided with a historical reality that had been more unruly.<sup>18</sup> A homogeneous notion of 'the Resistance' was therefore deemed necessary to bring together the disparately organised and unlinked conglomerate of individuals and groups that had been resisting the Nazi occupational regime – and whose visions of the future did not necessarily align with those of the post-war Dutch government.This was attempted by emphasising acts of resistance in the legislative formulation of the act, rather than intentions, political motivations, or orientations of the applicant.<sup>19</sup> That particular behavioural characteristics provided the former resistance with the right to extraordinary welfare benefits also had its implications: It created a dichotomy between 'active' victims of the former resistance and 'passive' groups of war survivors, such as victims of persecution.<sup>20</sup>

Yet there was more. The act was, first and foremost, considered as recognition and pay-off of the 'debt of honour', held by the post-war Dutch government*.* However, as death or continuing invalidity was a formal precondition for eligibility, there was a similarity between the WBP and a regular pension scheme. There had been a trade-off: only people who had sacrificed their health or even their lives for their country were eligible for a resistance pension. A causality clause was therefore included with the act: Applicants had to provide proof that damage resulted from activities in the context of resistance against the occupier. In establishing this legislation, the government consulted *Stichting 1940–1945* ('Foundation 1940–1945'). This interest group for former members of the resistance and their descendants had been

<sup>16</sup> Piersma, 15, 43; Martin Bossenbroek, *De Meelstreep* (Amsterdam: Bakker, 2001), 420; 'Handelingen Tweede Kamer 1946–1947', n.d., 1954.

<sup>17</sup> Withuis and Mooij, 'From Totalitarianism to Trauma', 196.

<sup>18</sup> Ross, *Mixed Emotions*, 136–37.

<sup>19</sup> Piersma, *Bevochten recht*, 62.

<sup>20</sup> This is also discussed in Chapter 7. See also Ludi, 'Who Is a Nazi Victim?', 10–11.

founded by foremen of the resistance in the autumn of 1944. *Stichting 1940–1945* became eventually responsible for the implementation of the WBP.<sup>21</sup>

#### Sailor-War Victims

The WBP was extraordinary, but not unique. Parallel to the WBP, there was the *Wet Buitengewoon Pensioen Zeelieden-Oorlogsslachtoffers 1940–1945* (WBPZ), or Extraordinary Pension Act Sailor-War Victims. This act, established in December 1947, dealt with the sailors and personnel of merchant ships and fishing boats confiscated between September 1939 and December 1945 (for fishermen) and March 1946 (for merchants) by the Dutch government for the war effort. Despite the naming, the WBPZ was not so exceptional, as sailor-war victims had actually been in government service. This was because of *vaarplicht* – an injunction to continue sailing for the allied effort. Approximately 2700 sailors had died, and several hundred were wounded or had become disabled due to war-related circumstances. The demarcation of who was eligible – and who was not – was easy, compared to that in the WBP. In contrast to the former members of the resistance, the sailors had been in actual military service of the government after all.<sup>22</sup>

#### A Question of Demarcation

That eligibility for the WBP was based on a 'demarcation of righteousness', was almost self-evident: Only people belonging to the 'right' category of war survivors were eligible, as the pension scheme was not open to, for example, former collaborators who had lost a limb.The*Raad van State*('Council of State'), however,identified already in 1947 a fundamental problem in the demarcation of who was eligible for the WBP, and who was not. In their assessment of the proposed act they emphasised that it might have become difficult to claim a resistance past for applicants, for example. An even more difficult issue was the burden of proof. How to prove that invalidity or death – both criteria for eligibility – had been the result of resistance activities? The council felt the proposed criteria in the bill were vaguely formulated, and open to too broad an interpretation. This complicated the practical feasibility of implementing the WBP. Minister Beel disagreed with these objections. He even accused the council of a lack of understanding of what it had meant to have been resisting the Nazi occupier.<sup>23</sup> As the government had chosen a pension scheme, this was and remained the point of departure in the first WBP and WBPZ debates in 1947. Objections, already raised in this very early stage by the council, proved to be lasting

<sup>21</sup> Withuis and Mooij, 'From Totalitarianism to Trauma', 196.

<sup>22</sup> Bossenbroek, *De Meelstreep*, 418–19; Piersma, *Bevochten recht*, 29.

<sup>23</sup> Piersma, *Bevochten recht*, 39–42.

and recurring. Was the WBP about suffering, care, compensation, or recognition and acknowledgement? And in what form? For whom? These questions, objections similar to that of the council, and dissatisfaction with certain aspects of the legislation kept triggering parliamentary discussions over the years. It was not that new hurdles came to the surface, as similar questions about in- and exclusion and the demarcation of who was eligible (or not) continued to resurface over the decades.

#### Trauma and Applications on the Rise

Years later, in the early 1970s, the application of the WBP in daily practice changed drastically. To the relief of the members of the Dutch parliament, the long-awaited *omgekeerde bewijslast* ('reversed burden of proof ') was introduced in 1971.<sup>24</sup> The burden of proof that one's ailments were a consequence of acts of resistance no longer rested on the applicant. Instead, the *Buitengewone Pensioenraad* ('Extraordinary Pension Council') had to prove the (un)eligibility of an applicant.<sup>25</sup> Withuis and Mooij write that this shows how a suspicious approach towards former members of the resistance had become socially unacceptable in the early 1970s. Consequently, it became, for example, taboo to take into account the (pre-war) medical history of applicants.<sup>26</sup> Whilst the actual number of potential beneficiaries dwindled, the number of applicants increased. Whereas between 1950 and 1970 approximately 6000 new clients requested a resistance pension, only between 1970 and 1980 some 8000 people applied. The year 1977 was a peak, when almost 1300 applications had to be dealt with. Also the proportion of invalids compared to descendants of resisters killed in action grew.<sup>27</sup> What was going on?

A first explanation is an increasing demand for recognition and acknowledgement of acts of resistance. Many people considered the attribution of a resistance pension as part of this process. *Ereschuld* ('debt of honour') became *erezaak* ('matter of honour') for many individuals with a personal past of resisting the Nazi occupier, noted the chair of *Stichting 1940–1945* in retrospect.<sup>28</sup> These calls for acknowledgement often remained unanswered. Next to acts of resistance, causal death, illness, or invalidity remained prerequisites for WBP eligibility.<sup>29</sup> Another develop-

<sup>24 &#</sup>x27;Handelingen Tweede Kamer 1973–1974', n.d., 3838.

<sup>25</sup> Withuis and Mooij, 'From Totalitarianism to Trauma', 205; Bossenbroek, *De Meelstreep*, 423–24.

<sup>26</sup> Withuis and Mooij, 'From Totalitarianism to Trauma', 205.

<sup>27</sup> A.J. van der Leeuw, 'De ontwikkeling van de overheidszorg voor oorlogsgetroffenen', in *Oorlogstrauma's na 45 jaar?: politiek en psychiatrisch ongeduld*, ed. A. Engelsman (Amsterdam: Van Gennep, 1989), 95; Bossenbroek, *De Meelstreep*, 423–24; Withuis and Mooij, 'From Totalitarianism to Trauma', 204.

<sup>28</sup> Cited in Piersma, *Bevochten recht*, 202.

<sup>29</sup> Piersma, 201–2.

ment pushing the number of applications was growing attention for so-called 'late consequences' of the war. This development was stimulated by the simple fact that 'late consequences' take some time to emerge and be recognized, but can also be attributed to the increasing professionalisation of the medical profession in psychotraumatic research and treatment. Medical professionals now recognised the possibility that a healthy and mentally stable person could become traumatised and thus ill because of exceptional negative experiences. Such experiences were typically associated with war, violence, repression, concentration camps, and resistance activities. As an external effect, psychological illness and mental health problems became widely known and discussed in public.<sup>30</sup>

Former members of the resistance increasingly started to think about themselves in terms of 'trauma'. Abram de Swaan described this phenomenon as 'protoprofessionalisation': people take over insights and terminology from medical professionals and interpret, formulate, and present their own complaints and diseases accordingly.<sup>31</sup> For people who had not died, had not lost a limb, or had no otherwise demonstrable physical damage, the professional and public acknowledgement of trauma created an opportunity to seek help. Withuis stated that for these people the step towards acknowledging their mental health issues and psychological illnesses became easier. General acceptance of trauma accommodated visiting a doctor with a 'real illness': a problem that was caused by an external stressor and that was publicly acknowledged.<sup>32</sup> As discussed in Chapters 1 and 5, sociologist Withuis attributed this public emergence of trauma to developments of secularisation and emancipation of emotions.This newly recognised illness contributed to the growing number of applications, as it opened new doors to the resistance pension. Whether – and if so, how – this affected parliamentary discussions of the WBP and WBPZ, and the role of emotions in those, will be addressed in Chapter 7.

#### Emotion Mining and the Resistance

Based on the topic descriptions in the *Handelingen* dataset, following the procedures described in Chapter 3, 25 relevant debates were identified. They cover the period between 1947 and 1985. All documents contain substantive discussions related to the

<sup>30</sup> Withuis, *Erkenning*, 118; Withuis and Mooij, 'From Totalitarianism to Trauma', 210; Withuis, 'Opkomst en neergang van PTSS', 158; JolandeWithuis, 'Introduction: The Politics ofWar Trauma', in *The Politics of War Trauma: The Aftermath of World War II in Eleven European Countries*, ed. Jolande Withuis and Annet Mooij (Amsterdam: Amsterdam University Press, 2010), 2.

<sup>31</sup> Abram de Swaan and Herman Hendriks, *Zorg en de Staat: welzijn, onderwijs en gezondheidszorg in Europa en de Verenigde Staten in de nieuwe tijd* (Amsterdam: Bakker, 1996), 249–52; Withuis and Mooij, 'Conclusion: The Politics of War Trauma', 328.

<sup>32</sup> Withuis and Mooij, 'From Totalitarianism to Trauma', 211.

WBP and WBPZ. In what follows, the outputs and results of a computational analysis of the emotions expressed in these parliamentary debates on WBP and WBPZ are discussed. How emotional were these debates? What pattern follow the manifestations of emotions in the resistance debates over the decades? How do these results relate to the general trends in emotions in parliamentary language use in the second half of the twentieth century? How do the 'extraordinary government employees' debates on the WBP and WBPZ relate to the 'perpetrators and collaboration' debates addressed in Chapter 5?The answers to these questions, that predominantly encompass quantitative aspects of the debates, are synthesised in Chapter 7 with a qualitative, more 'traditional' historiographical approach to the parliamentary debates. First, however, Chapter 6 presents the results of a distant-reading analysis of emotions in the parliamentary debates dealing with the WBP and WBPZ.

## **6. Erratic Emotions | Mining the Underground in the Dutch Parliament**

Most debates in the 'extraordinary government employees' subset deal with the establishment, modification, or discussion of WBP or WBPZ legislative schemes (see also Chapter 7). Some debates deal with both schemes at once. Words associated with the six emotion categories from the NRC EmoLex are identified, weighted, and thus 'scored' in all 250-word text chunks of the documents within this subset. Chapter 4 gives a more detailed description of this procedure. The complete overview of the results of computational analysis of emotional word use in the debates on the 'extraordinary government employees' (the beneficiaries of WBP and WBPZ) can be found in Supplement 3.1.<sup>1</sup>

### **6.1 Emotion-mining Results and the Resistance**

#### Variation between Emotions

Summary statistics are calculated for the scores of all 'extraordinary government employees' debates on each basic emotion category from the NRC EmoLex. All individual documents within this particular subset of parliamentary debates are taken into account. As the output in Table 14 shows, mean and median score for the emotion of trust is much higher than all other emotion scores. There are no zero scores for trust. This means language use associated with trust is not only prominent, but

<sup>1</sup> Just as in the previous chapter, when I mention the 'score' of a specific parliamentary debate on a certain emotion here, this score consists of the mean of all TF-IDF weights for all of the 250-word chunks of which that debate consists. Theoretically this does, however, not have to mean that there is (only) a higher frequency of emotion words in these debates per se. Higher TF-IDF scores are also caused by the (relative) frequent use of (relative) rare, uncommon occurring anger words. Because of the TF-IDF weighting that is applied to the scores, the (single) occurrence of relatively rare words weighs heavier than the individual occurrence of a very common anger word. For a detailed description of the weighing measure, see also Chapter 4.

also ubiquitous in the WBP and WBPZ discussions. The output corroborates earlier findings. All previous results of applying emotion mining to historical text collections in this study have already found evidence for the Pollyanna principle (see Chapter 5). The table shows that this effect is observed in the average parliamentary debates set, but also in this subset. Yet what about the negative emotions anger, fear, disgust, and sadness? And what about the scores of the debates particularly dealing with WBP and WBPZ? How do these debates relate to the baselines set in Chapter 5? Is there variation between these emotions as they were manifest in these different discussions, or were discussions consistent in emotional charge over time?


*Table 14: Summary statistics of emotion scores of all WBP and WBPZ debates (1945–1989).*

## Variation Between and Within Debates

The coefficients of variation give a first insight in differences between the scores of the 25 debates on the six emotion categories.<sup>2</sup> This measure gives an indication of the spread, or the relative differences in scores between the individual debates. The higher this measure, the more mutual variation between the debates within the thematic subset – independent of the absolute height of the emotion scores.The coefficient of variation of the trust scores is relatively low compared to the other emotions. Summary statistics of the scores (see Table 14) on anger, disgust, fear, and sadness display higher coefficients of variation. There is more variation between the individual debates for these emotion categories – compared to the emotion of trust. It does, however, still not tell much about where this variation within the subset can be located. To put it differently, we still need to find out what debates possibly stand out – and which ones do not.

<sup>2</sup> The coefficient of variation is also known as relative standard deviation. This measure is defined as the ratio of a standard deviation to the mean.

As all scores are in first instance based on the means per 250-word text chunk, the scores per individual chunk can also be plotted and analysed for each individual debate. This step can be considered as zooming in the macroscope a little bit. The scores-per-chunk are plotted in boxplots for each one of the debates in the 'extraordinary government employees' subset.The results for the emotion of anger are plotted in Boxplot 3, additional outputs regarding the five other basic emotions can be found in Supplement 3.1.1.

*Boxplot 3: Anger scores for each 250-word chunk in the WBP and WBPZ debate subset.*

Boxplot 3 gives a visual representation of the spread of the chunk scores within each debate in the subset. This can, in the first place, be derived from the difference between means (plotted as ) and median (plotted as **–**) score. Boxplot 3 also displays the distributions of individual chunk scores (•) within each individual debate.There are some peak debates that have both a high mean and median score (e.g., November, 2nd, 1955). In the debates where median and mean coincide, the scores per chunk are normally distributed. These debates (e.g., February 19th, 1963) are characterised by only a few outliers or extreme values. Other debates score very low on anger (e.g., 27th of November 1947) or have a rather skewed distribution of scores per chunk (e.g., 27th June 1957). In general, these distributions tell us something about fundamental characteristics of the parliamentary discussions: emotions were, also in high scoring debates, not a constant factor over the course of discussions. Episodes of strong emotionality alternated with ones that were more neutral. How this emerged in practice in the actual discussions in parliament,is described in more detail in Chapter 7.

#### Significance of Variance

The coefficients of variation in Table 14 and scores plotted in Boxplot 3 are an indication of differences between the scores of the debates. To see whether the differences (between themeans) of theseirregularly distributed groups of scores(for each chunk in each debate) are statistically significant, a non-parametric Kruskal-Wallis rank sum test is used.<sup>3</sup> The Kruskal-Wallis test tells whether, overall, group scores (in this case, chunk scores grouped per individual debate) differ significantly. Output and statistics of the test can be found in Supplement 3.1.2. The results indicate a statistically significant difference in the anger scores of the 25 debates on 'extraordinary government employees'. A significant variation between the different debates on the negative emotions of sadness, disgust, and fear is also found. Although the coefficients of variation for joy and trust are relatively low, the Kruskal-Wallis test also found a statistically significant difference for scores for word use associated with these two positive emotions.

The Kruskal-Wallis shows a difference exists, but does not indicate which particular debates' scores differ significantly (statistically speaking), and which debates are thus more divergent from one another.The overall finding that there is variation between debates, can be broken down by mutually comparing all pairs of individual debates. To tease apart the overall finding of significant variation between means of debates, a pairwise comparison using aWilcoxon rank sum test is used.This test calculates the pairwise statistical significance of scores for each emotion and for every possible pair of debates in the subset.<sup>4</sup> In addition to significance of the results, the effect sizes of each pairwise comparison are also reported.Complete output tables of the pairwise Wilcoxon tests for all other emotions and other relevant debates can be found in the online repository (see Supplement 3.1.3). Overall, most pairwise comparisons show no significant differences betweenindividual debates.Effect sizes are also predominantly small. Yet there are some exceptions. There are only a few, but very obvious and exceptional outliers: peak debates that caused the observed variation within the data. These findings will be discussed in further detail later on in this chapter and in Chapter 7.

The statistical evaluation of the results shows significant variation between the means of the debates in the 'extraordinary government employees' subset.This find-

<sup>3</sup> Because it cannot be assumed that all individual debates or subsets of debates have normally distributed scores, a non-parametric Kruskal-Wallis test is used here. See also Chapter 4.

<sup>4</sup> Unless otherwise specified a confidence interval of 95% (α = 0.05) is used. A difference is statistically significant when (p =< α). See also Field, *Discovering Statistics*.

ing applies to the negative emotion scores in particular. As also discussed in Chapter 5, variation between emotion scores in debates does not necessarily mean there is also a linear relationship between emotionality and time. In other words, differences do not necessarily implicate the existence of historical change that evolved as an increase or decrease in emotionality of the debates over time.

#### Diachronic Developments in Emotionality

To compare the particular historical development of emotional charge in debates dealing with 'extraordinary government employees' with the average parliamentary discussion, and the other case studies, the mean scores for each individual debate are plotted on timelines. These visualisations should look familiar by now, as their design is based on the ones presented in Chapters 4 and 5. To find out how variation is distributed over time, the results of emotion mining in the debates are not only compared pairwise, but also plotted on a timeline and compared to the average parliamentary discussions in the 1945–1989 period (see Supplement 3.1.3 and 3.1.4).The mean score of each individual debate of the subset is plotted as a coloured bar, complementary to the timeline of non-war-related parliamentary debate scores.<sup>5</sup> The moving average (using LOESS smoothing) is plotted as a coloured trendline. Just as the timelines in Chapters 4 and 5, the mean scores per day of non-war-related discussions are plotted as grey dots. The black line represents (again) the overall trend in scores of the average parliamentary debate between 1945 and 1989.This visualisation allows for a diachronic overview of the scores of the debates on 'extraordinary government employees', and a comparison with non-war-related discussion. Graph 12 displays the results for the emotion of anger.

<sup>5</sup> These colours follow Saif Mohammad's colour-emotion scheme. See also Saif Mohammad, 'Colourful Language: Measuring Word-Colour Associations', in *Proceedings of the ACL 2011 Workshop on Cognitive Modeling and Computational Linguistics (CMCL)* (Portland, OR, 2011); Mohammad, 'From Once upon a Time to Happily Ever After', 733.

*Graph 12: Anger scores of parliamentary debates dealing with WBP and WBPZ (red) compared to average parliamentary debates' scores (light grey and black) between 1945 and 1989.*

*Graph 13: Sadness scores of parliamentary debates dealing with WBP and WBPZ (grey) compared to average parliamentary debates' scores (light grey and black) between 1945 and 1989.*

The scores for the use of language associated with the emotion of sadness in the 'extraordinary government employees' debates are plotted in Graph 13 in grey bars. The results of analysing the other emotions in this subset can be found in Supplement 3.1.4.

#### Post-war Neutrality?

A first visual evaluation of the outcomes of emotion mining of the positive emotions displays a relative stable general course over time. Although there is, also in the case of positive emotions, some variation, the general trend seems to be one of stability (see Supplement 3.1.4). As could be expected from the statistical output addressed earlier, the negative emotion scores display more variation over time. The initial post-war debates on establishing the WBP and WBPZ in 1947 do not score exceptionally high on negative emotions. These debates score lower than subsequent discussions in the early 1950s. The debate in July 1947 scores even lower for the emotion of anger than the mean of non-war-related discussions of those days (see Graph 12 and Supplement 3.1.4). This applies to all negative emotions, except sadness (see Graph 13).<sup>6</sup> The sadness score of 1947 is higher than the non-war-related average score. To conclude, the immediate post-war years, discussions seem to have evolved in a relative unemotional way. All negative emotions score relatively low compared to other discussions and later debates in the subset. Only sadness in 1947 is an exception.

#### Peaks in the Mid-1950s

Besides the high sadness score of 1947, the subsequent debates in the early 1950s stand out as outliers regarding negative emotions (compared to other debates in the first 15 years after the war). Multiple 'extraordinary government employees' debates in the early 1950s score anomalously high on anger compared to the first debate of this kind, which was held on July 8, 1947.<sup>7</sup> Other peaks are, for example, the debates held on April 1st and 2nd in 1954 (see also Graph 12, Graph 13, and the outcomes of the Wilcoxon test in Supplement 3.1.3). Another anomaly is the debate that took place on the 2nd of November, 1955. In this discussion, the members of the Dutch House of Representatives talked with the responsible minister about the implementation of the WBPZ act.<sup>8</sup> Although this particular debate scores significantly much higher for

<sup>6</sup> This is revealed when, for example, sadness scores of this 1947 debate (n = 48) are compared to that of the May, 18th debate of 1951 (n = 28). No significantly lower score is reported for 1947 (*p =* 0.99, *r* =< 0.01).

<sup>7</sup> For example, the April 1st debate of 1954 (n = 28) scores significantly higher than the 1947 debate (n = 48) on anger word use (*p =<* 0.01). A moderate effect size can also be reported (*r =* 0.31).

<sup>8</sup> See also Chapter 7 and 'Handelingen Tweede Kamer 1955–1956', n.d., 3055–77.

the other negative emotion categories,it is especially the fear score that stands out in the statistical evaluation of the results.<sup>9</sup> The November 2, 1955 debate is an all-time high score for sadness within the subset of debates on the thematic category of 'extraordinary government employees'.<sup>10</sup> Graph 12 and Graph 13 display another peak debate. This emotional peak belongs to the debate held in parliament on February, 19 in 1963.<sup>11</sup>

#### The Parliamentary Gap Years

The graphs discussed above show also that, compared to the early-1950s and 1963 peak debates, the debates in the late 1960s, 1970s, and 1980s score all relatively lower for anger, fear, and disgust word use.<sup>12</sup> Based on these findings, and the general parliamentary trend, I argue that negative emotion peaks in both the 1963 and the mid-1950s debates can be considered exceptional. An important additional observation is that there are not many parliamentary debates on WBP and WBPZ scheduled in the following decade (1970s). This lacuna was not so much the result of a lack in emotional language use in the discussions, as it was the consequence of the absence of substantive parliamentary discussion on the topic. This did not mean there was no parliamentary engagement with the former resistance in this period.On the contrary. Multiple proposals for legislative adjustments were made by the government and scheduled for parliamentary reflection. These proposals, however, were mostly accepted by MPs without any discussion. Often, they even passed without a vote in

<sup>9</sup> For example, the November, 2 debate of 1955 (n = 17) scores significantly higher on fear word use than the July, 8 debate of 1947 (n = 48, *p =<* 0.01). The effect size is large with *r =* 0.60. Also in anger word use this debate differs significantly from the first resistance debate held in July 1947 (*p* = < 0.01). The effect size (*r* = 0,39) is however lower compared to the fear scores (*r* = 0.60). The scores for fear really stand out.

<sup>10</sup> On sadness this November 2nd debate in 1955 scores significantly higher than the July 1947 debate mentioned earlier (*p =* 0.03,*r* = 0.36). This 1955 debate is the only debate of the 1950–1959 period that differs significantly on sadness from the first WBP debate (July 1947).

<sup>11</sup> When this debate (n = 82) is compared to the first debate of July 1947 (n = 48), a significant difference can be reported (*p =*< 0.01). Size of this effect is moderate with (*r =* 0.47).

<sup>12</sup> For the anger scores, for example, a significant lower score of the debate on the 5th of July 1966 (n = 34) can be reported when compared to the May 18th debate of 1951 (n = 28, *p =<* 0.01, *r* = 0.46). This also applies when the 1951 debate is compared to the later May, 21 debate in 1985 (n = 15, *p =<* 0.01, *r =* 0.56). When these 1966 and 1985 debates are compared to the abovementioned April 1st debate of 1954, moderate and large effect sizes (resp. *r =* 0.38 and *r =* 0.50) and significant differences can be reported (resp. *p =* 0.02 and *p =* 0.01). The anger scores of these debates are more similar to those of 1947. When these 1966 (*p =* 0.68, *r* = 0.07) and 1985 (*p =* 0.19, *r* = 0.22) debates are compared to the July 8th 1947 debate, no significant differences in anger scores are revealed.

the House of Representatives.<sup>13</sup> No debate means no speech – and thus no substantive body of language use that can be 'mined'.

Graph 12 and Graph 13 display emerging patterns in the emotion-mining results of the debates in the 1945–1989 period on the 'extraordinary government employees'. These patterns can be considered as erratic for two different reasons. First, there was not very much continuity in the elaborate nature or frequency of discussions – see also the 'gap years'.Contrastingly,in the late 1940s and early 1950s frequent and elaborate discussions were identified. Second, the variation between the emotion scores of debates that did take place is indicative of a rather erratic course in the emotional charge of debating the 'extraordinary government employees' in the Dutch parliament over the years. Discussions on legislation for the former resistance and the sailor-war victims displayed emotional peaks between 1950 and 1955, and in 1963. MostWBP orWBPZ-themed debates after 1955 score lower on words associated with all four negative emotion categories compared to early-1950s peaks. These findings do not point at a clear pattern or an obvious increase or decrease in emotionality. Most importantly, these findings contravene the idea of a strong increase in emotionality in these debates.This is also backed by linear regression analysis on the debates belonging to the specific subset. Only a very weak decrease was found (see also Supplement 3.1.5). As these results only relate to quantitative characteristics of the role of emotions in the debates, the question remains whether the qualitative characteristics of emotions as they were manifest in the language use in peak debates differ from that of the other discussions. What was going on in those debates?

#### **6.2 The Emotion Words behind the Peaks**

Statistics and graphs only display the quantitative characteristics of emotional language in the debates.The historian's macroscope can be zoomed in and come closer. Emotion-mining output is not only about numbers. It also provides a first glance of the more qualitative characteristics of the parliamentary discussion of 'extraordinary government employees'.<sup>14</sup> Based on the DTMs used in 'scoring' the texts (see Chapter 4), an overview of the words that are behind the scores can be generated.The top 20 highest-scoring unique words in the debates for each particular emotion category can be found in Supplement 3.1.6. A closer look at the list of these high-scoring words gives a first impression of predominant aspects of the manifestations of emotional language in the debates.When looking in more detail at the EmoLex words in

<sup>13</sup> Proposals for adjustments to the WBP and WBPZ were scheduled – but not debated – on 8th of May, 29th of June, 6th of July 1971, 28th of August and 9th of September 1975, and the 23th of November and the 7th of December 1976.

<sup>14</sup> Boot, 'Emotion mining with lexicons', 7 July 2020.

the parliamentary proceedings, it becomes obvious that most of the high-scoring words in the November 1955 peak debate, for example, are associated with physical suffering or illness. Words like 'cancer' and 'illness' frequented the speeches. These words are not only in the NRC lexicon for sadness, but are also part of the lexicons for fear, anger, and disgust. In addition, the words 'disabled' and 'stroke' are distinctive of this particular discussion.These words are strongly associated with the emotions of sadness and fear, according to the EmoLex.

Many other high-scoring emotion words seem, at least at first sight, pretty straightforward. Take some of the top words in the 1947 discussion, or in the 1951 and 1954 debates. First, these are words almost inevitably used to refer to the topic under scrutiny, such as 'fighter' (as in: 'resistance fighter'), or 'victim'. When taking into account the top 20 of anger words for the debates in 1947, 1951, and 1954, some developments in these word lists stand out. Although these top words do not tell a comprehensive story, they draw the contours and invite further investigation. This will be elaborated in Chapter 7. A remarkable observation is the emergence of the word *'*discrimination' in the top list of a 1951 debate. This word also occurred frequently in both the 'peak emotional' debates of April 1954. Another point of interest is the occurrence of words associated with anger that are referring to individuals. Examples are nouns such as 'criminal' and 'murderer', or the judgemental adjective 'brute'. The word 'enemy' also climbs the top 20, when 1954 debates are compared to the one in 1947. Very strong negative acts (e.g., 'treason') or emotions (e.g., 'angry' and 'hatred') also appear for the first time in the top 20 in those 1954 peak discussions.

Emotion mining scores are the result of a rather crude measurement of a 'fuzzy concept'. The top word lists do not tell the whole story. Therefore, Chapter 7 pays more detailed attention to the historical context in which these findings can be placed, understood, and interpreted. How were words associated with emotions used in the parliamentarians' speeches and debates? In what context did they appear? What does this say about the role of emotions in the parliamentary engagement with the so-called 'extraordinary government employees'? The next chapter discusses the results of close-reading analysis of the parliamentary debates on these people. Both WBP and WBPZ were elaborately discussed over the years, although the WBP in particular made up the majority of the parliamentary debates. This was not only the case in the 1940s, but also in later decades when adjustments and corrections were proposed and discussed.What can we say about the role that emotions played in the discussion and establishment of the WBP and WBPZ legislation in parliament? And what explains the variation in scores between the debates, and the multiple emotional peaks observed in the emotion-mining results?

## **7. A Strong Disposition | Discussing the Pension Act for Extraordinary Government Employees**

The Dutch parliament discussed legislative proposals for a tailored resistance pension scheme in 1947. Despite the relatively early post-war establishment of the elaborate social welfare acts known asWBP andWBPZ, Chapter 6 displayed how the issue kept recurring on the parliamentary agenda over the years. Emotion-mining output indicated a rather erratic course of debates over time. This does not only relate to the distribution of the debates over the timeline, but also to the trend in emotion scores over time. Discussions on legislation for the former resistance and the sailorwar victims displayed some exceptional emotion peaks between 1950 and 1955. After 1955, however, except for a peak in 1963, most debates on the WBP or WBPZ score much lower on words associated with negative emotions than the early-1950s peaks. What was going on in the emotional peak debates? Why was this course of debates so erratic over time? And what does this say about a more general role of emotions in the parliamentary engagement with the 'extraordinary government employees'? This chapter addresses these questions with a closer, detailed, more 'traditional' historical analysis of these same WBP and WBPZ debates over the years.

#### **7.1 A Resistance Disposition**

'A rigorous governmental consensus'. That is how the Belgian historian Pieter Lagrou described the post-war political memory of the resistance in the Netherlands.<sup>1</sup> This characterisation also fits the first parliamentary discussion on this heterogeneous group of people on July 8th, 1947. Parliamentarian Anton Roosjen (ARP) quoted Queen Wilhelmina by saying '(…) it would be a debt of honour (…) not to forget those left behind by our heroes.'<sup>2</sup> Roosjen seemed to speak for most of his colleagues when

<sup>1</sup> Pieter Lagrou, 'The Politics of Memory: Resistance as a Collective Myth in Post-War France, Belgium and the Netherlands, 1945–1965', *Focus: History and Memory* 11, no. 4 (2003): 545.

<sup>2 &#</sup>x27;Handelingen Tweede Kamer 1946–1947', n.d., 1953.

he said he admired the fundamental idea behind this act.<sup>3</sup> The results of computerassisted analysis, discussed in Chapter 6, indicated a relative high score for words associated with positive emotions joy and trust in this debate. MPs used them in expressing their consent with the act. Relative to the very few expressions of other negative emotions, sadness scored quite high. A closer look at the course of the debates shows that words associated with sadness were not used to refer to past experiences of suffering. Although invalidity or death was an explicit criterion for eligibility for the WBP, historical suffering itself did not play a significant role in the way this act was discussed. Many of the identified words associated with sadness can be considered as inevitable in debating and describing this topic, such as 'deceased', 'disabled', 'victim', and 'widow'. Such choices of words are not meaningless: They shed a light on how MPs envisaged the current circumstances of this group of beneficiaries.With using such terms, speakers emphasised the contemporary identity and current suffering of the former members of the resistance over a role as, for example, heroic resisters of the past. Without being subject to heated debate, I argue these formulations were used to give expression to feelings of sympathy and compassion within the MPs.

As for the other negative emotions, relative low scores indicate an unemotional tone of discussion, especially in comparison with contemporary issues and later WBP debates (see Chapter 6). It is striking how parliamentarians referred to 'the Resistance' as if it was a clear-cut unity.The discussants that were present in those early post-war days seemed already imbued with the importance of this issue. Emotional provocation or the strategic usage of emotions such as anger was deemed unnecessary. Even opposition-party MPs felt apparently no urge to resist. Closer scrutiny of the debates indicates how most parliamentarians involved were reasonably satisfied by the proposal. Communists, for example, believed that the proposed legislation would also benefit their constituency sufficiently. It turned out that there was, back in 1947, simply not much to fight about. Consensus about the necessity of the establishment of an extraordinary pension scheme is reflected by a tendency of relative moderate expression of negative emotions in these parliamentary discussions. It shows that what is considered important, does not necessarily lead to heated discussion full of emotional utterances. Nevertheless, the MPs debated for more than four hours. What took them so long?

#### The Devil is in the Detail

A question remaining was how suffering, care, and the 'debt of honour' were related. Providing care for former resisters and their families was, as multiple parliamentary

<sup>3 &#</sup>x27;HTK 1946–47', 1951–59, 1982–2000.

speeches display, considered by many as an honourable pay-off of the 'debt of honour' by the nation.<sup>4</sup> These fundamental questions were however not solved in parliament. It proved to be fertile ground for discord – even decades later. In 1947, however,much of the discussion took place on the level of details. Parliamentarian Harm van Sleen (Labour party, PvdA) had '(…) spotted some stains (…)'in the legislative text, for example.<sup>5</sup> Van Sleen and his colleagues therefore submitted amendments on the amounts of money, tax deductibility, and the relationship between WBP payments and other sources of income. Yet there was more. The broad-felt 'debt of honour' made that for many the national government's engagement and intervention with this matter should ideally have been a formality without any discussion at all.

Despite relative low (negative) emotion scores, there was quite some discussion. The 1947 debates display how a high intensity of debate did not necessarily went hand in hand with a strong emotional charge. Nevertheless, regarding the emotions that were expressed by the MPs involved, sadness and grievance were predominant. The manifestation of these emotions was often related to the experience that debate about this legislation proved to be inevitable. MPs considered especially discussions related to money as disrespectful. In response to the lengthy discussion of financial details in the House of Representatives, Senator Hendrik Algra (ARP) mentioned that '(…) the financial motivation should not play any part in this at all'.<sup>6</sup> Also expectations about the future practical execution of the act triggered responses. Senator Baron Herman van Voorst tot Voorst (KVP) feared a load of administrative paperwork. He considered the fact that a bureaucratic institution had to intervene with the private lives of beneficiaries unnecessarily and disrespectfully onerous. Even stronger was the fear that beneficiaries might be associated with other recipients of financial (state) support, such as the poor or the unemployed.<sup>7</sup>

The legitimising qualities of former members of the resistance in post-war politics, the personal involvement of several MPs, and what was called a 'debt of honour' resulted in a consensual parliamentary attitude and a seemingly strong-felt solidarity. Regarding the position of those who had worked against the occupier, an internalised sensitivity emerged, that consisted of almost everything related to the people now officially considered as 'extraordinary government employees'. Or, in other words, a strong and durable 'resistance disposition' was established in parliament. This happened already soon after the German occupation had come to an end. Even the slightest disinterest in this virtuous group, a threat to solidarity, or tampering of the exceptional position of the resistance, seemingly touched upon MPs' concerns.

6 'HTK 1946–47', 1991–92.

<sup>4 &#</sup>x27;HTK 1946–47', 1956.

<sup>5 &#</sup>x27;HTK 1946–47', 1984.

<sup>7 &#</sup>x27;Handelingen Eerste Kamer 1946–1947', n.d., 984.

The parliamentary engagement with the WBP shows, in the first place, how this discussion was about much more than just a pension scheme. As a result of the 'resistance disposition', displaying any disinterest or a *laissez-faire* attitude regarding former members of the resistance was out of the question.<sup>8</sup> In 1947, this triggered elaborate debate on the level of details and, albeit very moderate, also emotional responses. Together with the number of amendments submitted by MPs, this made very clear that there was an honour here that had to be upheld.

MPs' responses barely affected the resulting formulations of legislation after all. The parliamentary discussion of details did not substantively affected the act, and Minister Beel declared amendments submitted by the parliamentarians in these debates unacceptable.The debates nevertheless show that MPs from both houses of the parliament were very willing to put pressure on the government when it concerned resistance-related legislation. Despite the 'stains' already mentioned by Van Sleen in the House of Representatives, *Stichting 1940–1945* advised the Senate as well to accept the bill. They expected that complete withdrawal of the proposal might have worse consequences. A majority of members of the House of Representatives eventually accepted the WBP on August 28, 1947. It also passed the Senate later that year.

#### Poignant Examples and Strategic Emotions

Following the establishment of the WBP, debates were put repeatedly on the parliamentary agenda to discuss propositions for modifications made by the government. In May 1951, for example, a legislative proposal was discussed concerning a raise of the pension payments. Main reason for this intervention was the increase of the costs of living in the recovering post-war economy of the Netherlands.<sup>9</sup> In general, the Dutch parliamentarians showed a welcoming attitude towards such improvements. MPs emphasised the fact that care for this group remained an important duty of the Dutch state.<sup>10</sup> Nevertheless, the issue provided enough food for debate again.The results of emotion mining (in Chapter 6) in the discussions on 'extraordinary government employees'in the 1950s displayed a significantincreasein themanifestation of words associated with negative emotions. How does this relate to the aforementioned 'resistance disposition'? What was going on in these discussions?

As the Council of State had already foreseen in 1947, defining who was eligible, and to what extent, proved to be difficult. Resistance activities had been, by their nature, clandestine and thus often badly documented. In addition, the group of peo-

<sup>8</sup> Historian Henri Beunders also observed this phenomenon in the context of more general acknowledgement of victimhood. He, however, placed this development in the 1990s. See also Beunders, *Publieke tranen*, 204.

<sup>9 &#</sup>x27;HTK 1950–51', 1720–29.

<sup>10 &#</sup>x27;HTK 1950–51', 1724, 1726–29.

ple that was now 'united' in the formulation of the WBP was, both historically and in the contemporary context, heterogeneous.This made the practical application of the act difficult. First and foremost, practical problems in dealing with the contemporary bureaucratic reality caused confusion and friction. As a result, stories of inadequate applications of the act, rejected applicants, and resulting poignant cases of contemporary suffering soon reached parliamentarians via close connections and personal networks.<sup>11</sup> MPs did not let this go. They used the parliamentary discussions on legislative modifications mainly as an opportunity to address demands or complaints about the practical execution of the WBP.<sup>12</sup> Communist Jan Haken, for example, said the legislation had to be structurally adjusted in such a way that nobody with an active resistance past should bein need of further support. Just asmany of his colleagues, he advocated for a more inclusive, elaborate, or generous benefit legislation.<sup>13</sup> This was easily said for parliamentarians, as practical issues and actual consequences of demarcation were placed at safe distance.

*Stichting 1940–1945* was responsible for the execution of the WBP and WBPZ schemes in practice.<sup>14</sup> In designing the WBP, the Dutch government had more or less outsourced the practical implementations of difficult issues and decisionmaking about individual cases to a non-governmental organisation. MPs could therefore opt for more inclusive eligibility criteria and elaborate legislation without far-reaching practical or political consequences. Although they were responsible for the budgets, they were, after all, not the ones that had to make the decisions about individual applications. In this way, parliamentarians could relatively easily claim the emotive qualities of the former resistance in their speeches. It was in such contributions to the debates that language associated with sadness and fear was manifest.The words associated with sadness, also mentioned in Chapter 6, were not so much used to give expression to mourning of the MPs themselves. These words were manifest in elaborate descriptions of poignant examples of contemporary misery of 'resistance victims' – as the WBP beneficiaries were now often referred to.<sup>15</sup>

Advocating for better benefit legislation through emphasising contemporary misery was not unique to discussing the WBP. This was also done in discussions on the WBPZ for sailor-war victims. A rather extravagant example of this strategic expression of emotions is the WBPZ discussion of November 2 and 3, 1955 – one of

<sup>11</sup> Many of the MPs prominently involved in these debates had strong personal connections to networks of former members of the resistance. Not only social democrat In 't Veld, for example, but also his fellow party member Scheps and communists Haken and Wagenaar had a personal past as underground resistance worker.

<sup>12 &#</sup>x27;HTK 1950–51', 1720–21.

<sup>13 &#</sup>x27;HTK 1950–51', 1723.

<sup>14</sup> Ismee Tames, 'Stichting 1940–1945 en "bijzondere solidariteit"', *Contactblad '40-'45*, 2015, 13.

<sup>15 &#</sup>x27;HTK 1950–51', 1726.

the peak debates discussed in the previous chapter.<sup>16</sup> After asking the chair for explicit permission to do so, liberal MP Govert Ritmeester (People's Party for Freedom and Democracy, VVD) summed up many individual examples, including people's names, illnesses, and a description of their contemporary suffering.<sup>17</sup> Ritmeester mentioned all possible diseases these people were suffering from, such as cancer, strokes, disability, or illness in a more general sense. The use of these terms fuelled the anomalously high manifestation of sadness in this debate (that already stood out in the evaluation of results discussed in Chapter 6).This debate is exemplary for WBPZ and WBP debates in this period, where emotionally charged descriptions of individual suffering of identifiable victims were made into a cause. This had to evoke and invoke feelings of compassion for specific groups that were apparently left out in the current legislative arrangements according to the MPs involved. This rhetorical practice seemed like a deliberate political strategy, or the result of a strong urge to represent or support a constituency. These formulations were not only used to demonstrate a fear of falling short or deterioration of post-war living conditions for some, it also shows how what I have called a 'resistance disposition' before remained strong and persistent.

There was another parliamentarian very well aware of the 'poignant cases' amongst former resisters. Senator Joris in 't Veld (PvdA) knew these cases very well because of his membership of the Administrative High Court. There he dealt with individuals appealing to decisions made by *Stichting 1940–1945*. In contrast to many of his colleagues in the House of Representatives, the senator did not blame all misery on defective legislation. In June 1954, In 't Veld mentioned that some former resisters had 'lived the high life during the occupation'. According to In 't Veld, this was the reason that those people now experienced problems in fitting in the post-war societal order. He felt these people, or 'tragic cases', as he called them, might benefit more from mental support instead of financial aid. He considered his speech that day therefore also as a call to the churches and the *Humanistisch Verbond* (Humanistic Union) to pay more attention to these people within their networks.<sup>18</sup> This statement is not only very typical of the 'pillarised' society of the Netherlands, it also characterises a common current of those days: In 't Veld made very clear that any mental support or psychological care was not considered integral to national resistance legislation.<sup>19</sup> In 't Veld also recalled some of the cases that

<sup>16 &#</sup>x27;HTK 1955–56', 3047–78.

<sup>17 &#</sup>x27;HTK 1955–56', 3047–49.

<sup>18</sup> Joris in 't Veld had been involved in the establishment of the humanistic foundation *Humanitas* in 1945. In 1954 he was chairman of *Humanitas* and also active as board member of the Humanistic Union. With this, he emphasised the (desired) legal similarity of the *Verbond* and the churches in Dutch society.

<sup>19 &#</sup>x27;Handelingen Eerste Kamer 1953–1954', n.d., 3387.

had been mentioned in the House of Representatives: Not as argument for more elaborative legislation, but to illustrate the remaining difficulty of the question of demarcation for WBP eligibility. Every criterion created a possibility that individual applicants would be excluded because of the specific nature of their resistance activities, current income, degree of disability, etc.<sup>20</sup> Minister Frans van Thiel was also present. His newly established Ministry of Social Work (1952) had become responsible for social welfare legislation such as the WBP.<sup>21</sup> Van Thiel stated that this problem might never become easy. No matter how legislation was formulated, there would always be distressing cases of applicants falling through the cracks.<sup>22</sup>

In contrast to this resigned attitude in the Senate, the possibility of former resisters feeling left out was unacceptable for the members of the House of Representatives. It fuelled a second issue regarding the execution of the WBP:The emergence of a strong fear in MPs that their own group, their constituencies, or their supposed electorate would be left out. It was in this context that the word *discriminatie* ('discrimination') popped up in 1951 – a word strongly associated with emotions anger and fear (see also Chapter 6). Although the all-inclusive formulation of 'the Resistance' in the WBP suggested otherwise, communist MP Haken stated that some applicants had experienced discrimination. He could even mention 'multiple personal examples' of how mainly, but not exclusively, communist resistance fighters experienced exclusion from the pension scheme.<sup>23</sup> In other words, Haken stood up for 'his' group by claiming that they should as well be part of the umbrella term of 'the Resistance' as it was used in the legislation's formulations. Nevertheless, the responsible minister<sup>24</sup> assured Haken that modern-day oppositions and political affiliations of applicants would always be kept outside the WBP.<sup>25</sup> How tenable was this claim?

#### **7.2 The Cold War Enters the Stage**

The substantive adjustments to the WBP and WBPZ legislation proposed in 1951 were discussed in parliament three years later.<sup>26</sup> The results of emotion mining (see Chapter 6, Boxplot 3, and Supplement 3.1.1) indicated higher emotion scores in 1951

<sup>20 &#</sup>x27;HEK 1953–1954', 3388.

<sup>21</sup> For an elaborate description of this ministry and its influence on the WBP, see Piersma, *Bevochten recht*, 53–58.

<sup>22 &#</sup>x27;HEK 1953–1954', 3389.

<sup>23 &#</sup>x27;HTK 1950–51', 1722–23.

<sup>24</sup> In 1951 this was still the Minister of Internal Affairs, Johannes van Maarseveen from the KVP.

<sup>25 &#</sup>x27;HTK 1950–51', 1726.

<sup>26</sup> These adjustments were discussed in the House of Representatives on April 1st and 2nd, 1954. On June 29 that year, it was the Senate's turn.

and 1954 discussions. Something was going on in these debates; these scores indicate a change compared to 1947. At first glance, the debates display again a consensual and welcoming attitude of MPs towards the governmental proposals.<sup>27</sup> MP Johan Scheps opened his speech in 1954 even with recalling the 'mild atmosphere' of the 1947 debates.<sup>28</sup> Nevertheless, national and international political developments had been turbulent in the past years. Take, for example, the Czechoslovak *coup d'état* (1948), the start of the Korean War (1950), and the East Berlin uprising (1953).<sup>29</sup> Such global events had also put relations on edge within the Dutch political landscape. Members of the Dutch communist party, that, at that time, moved towards stronger Moscow-oriented policies, were officially banned from government and civil service positions.This included even schoolteachers and postal workers.<sup>30</sup> Also in the Dutch parliament,MPs isolated their communist colleagues by rejecting their motions and amendments on principle. In the Dutch context, Cold War tensions became most clearly manifest in political discord on the left. Labour (PvdA) led the way in hardening of the relations and the isolation of communists.<sup>31</sup> Especially Johan Scheps – described by Piersma as 'the communist-eater' – was notorious.<sup>32</sup>

The contemporary reality of the Cold War took centre stage in engaging with the heroes (or the villains) of the past. Contemporary events made the ally of the war past the enemy of today (or tomorrow).This affected relations between different interest groups of former resistance members. Some of these organisations decided to exclude communists from their activities, or from their management boards.<sup>33</sup> Communists were also banned from the association for former political prisoners, *Expogé,* in 1949. *Stichting 1940–1945* broke all ties with their communist counterpart association *Verenigd Verzet 1940–1945* ('Resistance United 1940–1945').<sup>34</sup> These developments and contemporary events in the national and international political arena put pressure on the post-war arrangements for the former resistance.This raised the question who, in the Cold War context, remained (still) eligible for the WBP scheme. How feasible was Van Maarseveen's claim of political neutrality regarding the WBP, against a background of rising Cold War tensions?


<sup>27 &#</sup>x27;Handelingen Tweede Kamer 1953–1954', n.d., 3745.

<sup>28 &#</sup>x27;HTK 1953–54', 3745.

<sup>29</sup> Piersma, *Bevochten recht*, 63.

<sup>30</sup> Bossenbroek, *De Meelstreep*, 332.

<sup>33</sup> These organisations were not only supported by the government, but, in the case of *Stichting 1940–1945*, also responsible for the implementation of the WBP. The *Stichting 1940–1945*, however, stated that intentions of resistance, or contemporary political position did not play a role in their awarding procedures of resistance pensions.

<sup>34</sup> Piersma, *Bevochten recht*, 62–63; Tames and Withuis, 'Hoe de Koude Oorlog'.

Most MPs involved understood very well that the historically heterogeneous variety of the group of wartime resisters clashed with the strive for a consensual or reconciling attitude on the 'debt of honour'outlined before,or with the use of 'the Resistance' as a unifying umbrella term*.* <sup>35</sup> The solidarity that came with the 'resistance disposition' discussed before was far-reaching, but proved to have its limits. Scheps, one of the prominent speakersin a fierce debatein 1954, showed where he envisioned the boundaries. He fundamentally opposed the idea of distributing resistance pensions to a group of people that, as Scheps feared, had the potential to become a fifth column in post-war Dutch society. The 'communist-eater' was a strong opponent of communists' eligibility for the WBP. As his main argument, Scheps used the fact that the extraordinary status of former resisters was based on service for the Dutch government in times of alien occupation. Scheps adhered to his belief that the communist resistance had acted out of loyalty to the Soviet Union, and not out of loyalty to the Dutch fatherland. For him their resistance did not originate in a true 'national attitude'. Nor had it been worthy, or right.Their activities had been clandestine, but had not been, in his words, 'patriotic resistance'. Former members of the resistance with a communist worldview had acted out of other intentions, Scheps said.<sup>36</sup>

#### Mud-slinging and the 'Chicken Legs of Intimate Friendship'

Scheps' statements in the 1950s debates can be considered as contradictory to the official formulation of the WBP on a fundamental level: He now placed intentions (behind working against the Nazi occupier) above people's actual acts.The demarcation of moral territory seemed his main objective: Communists should be excluded, in retrospect, from the occupation-time history. In this debate, it became also clear how and why Scheps thought communists were to be excluded from the umbrella term 'the Resistance'. They did not belong to his desired 'resistance' – and were thus not eligible for a resistance pension. Scheps could say this, make this point, without it having major practical consequences. After all, the responsibility of making such decisions of in- and exclusion on the level of individual applications was still with *Stichting 1940–1945*. Scheps' contribution to the debate in 1954 turned out to be merely a performative statement. How did this influence the role and expression of emotions in the discussions?

To support his viewpoint, Scheps quoted texts from an early wartime copy of the clandestine communist newspaper *Noorderlicht*. He used an article where the British were described as 'mortal enemy' of the nation. Scheps used this to emphasise how communists agitated against an important ally that was already on the Dutch side '(…) when the Germans and the Russians were eating the chicken legs

<sup>35</sup> Ross, *Mixed Emotions*, 136–37.

<sup>36 &#</sup>x27;HTK 1953–54', 3762–63.

of intimate friendship together.'<sup>37</sup> With this remarkable metaphor, he referred to the fact that the Dutch communist party had not distanced itself from Nazi Germany from the start of the occupation in May 1940. Because of the Molotov-Ribbentrop-Pact, overly hostile attitudes towards Nazi Germany had been an uncomfortable internal dilemma for the Soviet-oriented CPN. After Hitler started Operation Barbarossa against the Soviet-Union in June 1941, at least this internal struggle was soothed.<sup>38</sup> Scheps felt this pre-June-1941 publication could still serve as a strong argument in 1954.<sup>39</sup>

The communist response was swift. Communist MP Wagenaar expressed his disgust when he called it inappropriate of Scheps to dig up this part of history.<sup>40</sup> He and his colleagues felt as if, with retroactive force, the communists were thrown out of 'the Resistance'. This was painful, as the losses in the communist anti-Nazi resistance had been exceptionally heavy. Therefore, with the horrors of the German occupation still in mind, communists considered exclusion from the WBP as discrimination. In response to Scheps' 'history lesson' with quotations from *Noorderlicht*, Haken said:

'Despite the fact, that the honourable representative in the early days of the occupation felt that it would be possible to cooperate with the Zwart Front of Arnold Meijer, we would, if the honourable representative, because of circumstances resulting from his act of resistance, would also apply for this legislation, see no reason for rejection.'<sup>41</sup>

Where did this came from? Haken alluded to a pamphlet from 1940, in which Scheps had written about possible a cooperation with the Dutch fascist party *Zwart Front* ('Black Front').<sup>42</sup> CPN parliamentarian Haken took the moral high ground here. Scheps responded that bringing up this past was pure insinuation, as *Zwart Front* had nothing to do with this. His quote from *Noorderlicht* did indeed, he claimed.<sup>43</sup>

<sup>37 &#</sup>x27;HTK 1953–54', 3745.

<sup>38</sup> See also Piersma, *Bevochten recht*, 61; Mooij, *De Strijd Om de Februaristaking*.

<sup>39 &#</sup>x27;HTK 1953–54', 3745.

<sup>40 &#</sup>x27;HTK 1953–54', 3745.

<sup>41</sup> Although it is difficult to reconstruct, this ad hominem style change in his speech indicates that Haken's response was not the result of well-prepared rhetoric. Presumably, this speech was improvisation work. See also 'HTK 1953–54', 3754.

<sup>42</sup> Haken implicitly referred to a pamphlet Scheps had written in 1940, where he discussed a possible synthesis of the oppositions between, amongst others, Labour/social democrats and the*Nationaal Front* (National Front), the successor of the *Black Front* (both led by Dutch fascist Arnold Meijer). Trivialising Scheps' resistance work, based on this pamphlet, was for Haken similar to rejecting communist illegal workers from the*Noorderlicht* group for the WBP based on what they had published in 1941.

<sup>43 &#</sup>x27;HTK 1953–54', 3762.

#### Moral Posturing and Emotions

The fierce debate outlined above illustrates how Scheps, Wagenaar, and Haken ended up in rather fierce and personal Cold War-related mud-slinging. It is an illuminating example of how contemporary political tensions and fears came to dominate the post-war engagement with the group known as 'extraordinary government employees' during the Cold War. Associating oneself by expressing solidarity with the 'morally just' group of members of former resistance was, as the analysis of the 1947 debates already showed, a welcomed political instrument. It was seen as a way of promoting political virtue and legitimacy. Despite the umbrellaterm formulation of 'the Resistance' in the WBP, ideological differences between groups and individuals in the wartime resistance had been prominent. This now made the issue come in handy as an instrument in doing Cold War party-politics. As such, discussing the WBP gained an additional potential to be used as a mechanism for inclusion, exclusion, and framing.<sup>44</sup> As Piersma already noted, creating contemporary political suspicion was an important part of this.<sup>45</sup> In 1954, Scheps turned the 'resistance disposition' around by excluding his supposed enemy from the WBP and from his interpretation of 'the Resistance'. Communists, in response, alluded to the same disposition, by claiming how unfair and discriminating this was. How does this contemporary issue, that was fought out in a personal feud over the WBP legislation, relate to the higher emotion scores in these debates?

The 'resistance disposition', already observed in 1947, created not only a widespread sense of solidarity. This disposition also attached almost instantly a sense of urgency and importance to the issue, and forced political opponents to always take a stance in debates on the WBP. As the discussion between Scheps and his communist colleagues shows, not responding was not an option. In this way, the WBP discussions in parliament also became a vehicle to address other issues. In 1954, with Cold War tensions running high, the WBP debate proved to be a moment to fight over political oppositions. This demarcation of one's moral territory, or moral posturing, was done alongside the lines of the boundaries or scope of the WBP legislation. Because of the broadly-shared attitude of importance and political virtue connected to 'the Resistance', emotions related to this disposition could be re-directed and re-used in doing contemporary politics. The emotions connected to the 'extraordinary government employees' were given a role in the discussion of something else – in this case, in doing Cold War politics.

The results of emotion mining already indicated that particularly the debates in April 1954 scored higher than before on negative emotions. In 1954, however, not so much the emotions related to wartime history, let alone those of former resisters

<sup>44</sup> Ludi, 'Who Is a Nazi Victim?', 19.

<sup>45</sup> Piersma, *Bevochten recht*, 63.

themselves, became manifest in the WBP discussions. Rather, in the early 1950s, Cold War oppositions fuelled the expression of emotions. Words that are strongly associated with negative emotions (e.g., 'enemy', 'brute', and 'hatred') fuelled the peak negative emotion scores of these debates.<sup>46</sup> With these words, associated with anger, fear, and disgust, MPs made the debate emotional and personal. In this process of moral posturing, of demarcating what they considered just – and what not, they framed their opponents in a very hostile way. What the early-1950s discussions of theWBP show,is how 'extraordinary government employees' became a mean, and not a goal, in aiming for a contemporary political purpose.

In the early-1950s, moral posturing by means of discussing resistance legislation, turned out to be a merely performative practice. Despite the Cold War oppositions fought out in the resistance pension debates, the discord did neither substantively affect the actual act, nor did it influence the 'rigorous government consensus' about its necessity. The legislative proposal of 1954 became eventually even accepted without a vote by the House of Representatives. As the *Stichting 1940–1945* remained responsible for executing the WBP in practice – based on the same old legislative formulations – also communists retained eligible. In later debates, even Scheps and Haken, for all their mud-slinging, teamed-up again against government policies regarding 'extraordinary government employees' that were, in their eyes, disrespectful.<sup>47</sup>

#### **7.3 Wiedergutmachung. But for Whom?**

Not many other topics beside the WBP or WBPZ made the 'extraordinary government employees' appear on the parliamentary agenda. In the early 1960s, however, the negotiations about *Wiedergutmachung* payments from West Germany did. These reparation payments and their internal distribution had little to do with the existing resistance pension scheme. In contrast to the WBP, they consisted of one-off payments and eligibility was unrelated to invalidity or death of potential beneficiaries. Nevertheless, the issue caused friction with the established 'resistance disposition' described above.The governments of West Germany and the Netherlands signed an *Algemeen Verdrag* ('General Treaty') on the 8th of April 1960. Three years later – again! – Scheps stated in parliament that the Dutch had not negotiated very well and the Federal Republic of Germany (FRG) had showed an unwilling attitude.<sup>48</sup> Of the 382 million mark the Dutch demanded in 1957, they ended up with only 280 million in 1960. 125 million marks were intended as one-time reparation payments for victims

<sup>46</sup> See also Chapter 6.

<sup>47 &#</sup>x27;Handelingen Tweede Kamer 1954–1955', n.d., 1–5.

<sup>48 &#</sup>x27;Handelingen Tweede Kamer 1962–1963', n.d., 694, 699, 738.

of the Nazi occupation who had been persecuted because of 'race, religion, or worldview' – as the treaty stated.<sup>49</sup> In contrast to other countries, the Dutch had not negotiated withWest Germany to incorporate the former members of the resistance in this definition. Instead, the Dutch delegates had negotiated settlements for – predominantly – Jewish victims of persecution. They had not included people who had actively resisted the occupier. This created a problem.

A government-initiated commission led by former Prime Minister Drees (PvdA) examined the internal distribution of the millions of marks in the Netherlands. The commission decided that people who actively worked against the Nazis should also receive their share. They expected that this would also become a strong parliamentary demand.<sup>50</sup> The initial government delegates, however, had kept the omission in the negotiations silent. According to Piersma, in this way, the controlling duties of the parliament were deliberately set aside.<sup>51</sup> Piersma also pointed at the possibility that some parliamentarians (Scheps and Goedhart) might even have known about this during the debate in 1963. Although it is hard to prove, she implies that these MPs left the omission deliberately out of the debate to prevent exclusion of former resisters. This seems likely, given the strong and persistent disposition in favour of 'the Resistance' identified in the earlier debates.<sup>52</sup> The commission led by Drees anticipated well to the situation, as their expectations turned out to be right. Without exception, MPs demanded in February 1963 that the payments should not only benefit victims of persecution, but also former members of the resistance.<sup>53</sup>

As expected, the internal distribution of the West-German payments were indeed elaborately discussed in parliament. This topic also accounted for the peaks in emotion scores in 1963 that were shown in Chapter 6. In those discussions, many of his colleagues supported Isaac Diepenhorst (CHU) when he called a possible exclusion of the resistance 'intolerable discrimination'.<sup>54</sup> Other emotional contributions to the debate came, for example, from communist Bakker. He repeatedly called the whole state of affairs 'a disgrace'. It did not do justice to the 'sacrifices of the resistance', he said.<sup>55</sup> The shared attitude of disapproval, illustrated by these quotes, accounts for much of the emotional expressions manifest in the discussions of 1963.

<sup>49 &#</sup>x27;HTK 1962–63', 700.

<sup>50</sup> Piersma, *Bevochten recht*, 90–93.

<sup>51</sup> The government had kept this deficiency in the negotiations silent because they excpected pressure and criticism from the parliament, if it would come out the resistance was left out of the distribution of the Wiedergutmachungs payments in the initial agreement. See also Piersma, 93.

<sup>52</sup> Piersma, 106–7.

<sup>53 &#</sup>x27;HTK 1962–63', 689, 691, 697, 704.

<sup>54 &#</sup>x27;HTK 1962–63', 689, 691, 697, 704.

<sup>55 &#</sup>x27;HTK 1962–63', 689.

This debate made obvious that only a threat of doing short the 'extraordinary government employees' already triggered a disposition amongst parliamentarians. As a result, emotional responses were evoked in the House of Representatives. Now MPs expressed their own grievance, disappointment, and even anger, in anticipation of possible shortcomings regarding the former resistance. There was, however, one other current in the debate that accounted for the exceptional high emotion scores in the 1963 discussions.

#### Implications for Others

The main prerequisite for eligibility for the German reparation payments was 'victimhood because of persecution'. An active-passive dichotomy, that distinguished 'the Resistance' from other groups of war victims, had not only been prominent in parliament: It was also a formal criterion of the WBP since 1947. This now seemed to exclude former members of the anti-Nazi resistance from the *Wiedergutmachung,* as the Dutch negotiators, possible as a result of this dichotomy that persisted in the post-war Dutch context, had not included them in their definition of 'victims of persecution'.<sup>56</sup> From a German perspective, this made sense. Since the liberation in 1945, former resisters were treated by the Dutch government as being retroactively in military service. Why would the FRG financially compensate retroactive soldiers of a former enemy army?<sup>57</sup> Amongst Dutch MPs in 1963, however, the idea that the resistance might fall through the cracks of the General Treaty evoked their worst fears.

The logic of how the Dutch government had engaged with 'the Resistance' since 1945, frustrated what MPs were trying to achieve in this particular case. As a result, the notion of the 'active' former resister now had to be reframed into that of a 'passive' victim of Nazi persecution*.* <sup>58</sup> As a result, MPs worked their way towards including the resistance in their definition of victimhood, instead of emphasising their exceptionality, deeds, or moral virtue. It shows, in the first place, how strong the 'resistance disposition' was once again. Where the 1951 and 1954 debates showed that discussing legislation for 'the Resistance' was almost instantly associated with emotion, the 1963 debate again triggered deliberate and strategic use of emotions. MPs brought the contemporary misery of former resisters again to the surface in

<sup>56</sup> This is more elaborately discussed in the next two chapters. See also Piersma, *Bevochten recht*, 93.

<sup>57</sup> This was in line with other German reparation arrangements. See Ludi, 'Who Is a Nazi Victim?', 15–16.

<sup>58</sup> The following observation is illustrative of this development: The word *ereschuld* ('debt of honour') was only mentioned once in the 1963 debates. In comparison, the word occurs eleven times in the 1954 WBP discussion.

the debates that followed in parliament – just as happened in 1955. Bringing up contemporary misery in parliament now had to put emphasis on the victimhood of this group. Many of the words associated with emotions were used in this process and underlie the high emotion scores reported in Chapter 6.

The remarkable course of events regarding the parliamentary discussion of the internal distribution of the payments did however not mean that other victim groups were entirely forgotten in the process. Closer scrutiny of the debates of 1963 displayed that a share of the payments for – mostly – Jewish victims of persecution was explicitly and repeatedly considered as irrefutable, legitimate, and necessary.<sup>59</sup> The occasional equation of the resistance with other groups of victims had, however, some serious implications. Retrospectively including 'the Resistance' in the General Treaty – whilst there had been no negotiations with the FRG regarding payments for this group of war survivors – inevitably meant that victims of persecution would receive a smaller share. This consequence was, in 1963, left unmentioned in parliament.

#### Routine Discussions

The discussions on the WBP and WBPZ in the House of Representatives in the subsequent years all followed a similar routine: Improvements were proposed by the government, MPs acknowledged them, but stressed at the same time that still more had to be done. Often, MPs mentioned their close personal contact and engagement with *Stichting 1940–1945* and other interest groups.<sup>60</sup> Communist MP Annie van Ommeren-Averink stated on July 5, 1966 that proposed adjustments were praiseworthy, but '(…) unsatisfactory on a number of issues'.<sup>61</sup> Especially the causality clause proved to be a recurring subject of such dissatisfaction.<sup>62</sup> Almost as if it was a prophecy, Secretary of the State Hein van de Poel (KVP) said in 1967 that there would be always something left to desire regarding resistance legislation.<sup>63</sup> Although these discussions do not identify as particularly emotional, emotional language was observed in weighty formulations on the nature of resistance work. Expressions of negative emotions like anger, disgust, and sadness were not as prominent as they had been (in the choice of words) in the performative discussions of the mid-1950s. Given earlier experiences it is not surprising that especially communist MPs increasingly emphasised the discussion of government care for former

<sup>59 &#</sup>x27;HTK 1962–63', 684, 700; Tames, 'Stichting 1940–1945', 14.

<sup>60 &#</sup>x27;Handelingen Tweede Kamer 1965–1966', n.d., 2234–35.

<sup>61 &#</sup>x27;HTK 1965–66', 2233.

<sup>62 &#</sup>x27;Handelingen Tweede Kamer 1967–1968', n.d., 100–101.

<sup>63 &#</sup>x27;HTK 1967–68', 102.

members of the resistance should never (again) serve contemporary politics.<sup>64</sup> In the late 1960s and early 1970s, the WBP and the WBPZ kept recurring regularly on the parliamentary agenda. Was this desire feasible?

### **7.4 A Formality?**

Some substantive changes in the implementation and execution of the act were made in the early 1970s. Most of the government's legislative proposals and modifications were, again, accepted by parliament without deliberation, and often even without a vote.<sup>65</sup> As graphs in Chapter 6 already indicated, the May 22nd debate of 1974 in the House of Representatives was a break from this trend. Apparently, not everyone was satisfied with the habit that had been developed, to let modifications to the WBP and WBPZ pass without elaborate parliamentary discussion. Joop Wolff (CPN) opened with commenting that it would be 'very unjust to let legislative adjustments like this pass as a *hamerstuk* ('formality') now'.<sup>66</sup> This discussion makes a rather distanced, formal, and therefore professional impression. The once so prominent tendency of elaborately describing examples of individual misery came to an end. The more performative practices of moral posturing and political suspicion making also seemed to be omitted. Maybe this was the result of the relaxing Cold War oppositions, as communist Wolff, for example, could now count on support of many colleagues when he submitted amendments on modifications to the WBP in 1974.<sup>67</sup>

Based on closer scrutiny of the 1974 debate, I argue the emergence of trauma as acknowledged illness and its integration in the execution of the WBP only moderately impacted parliamentary discussions. Trauma was now named in one breath with disabilities or death. Words associated with emotions were often referring to 'illness' or 'distress' and were only moderately used, but all the more often in the context of mentioning mental or psychological illness and distress. In addition, words referring to a more or less generic interpretation of victimhood seem also to have become more prominent. Although these words are associated with the emotions of fear and sadness, the parliamentary debate in 1974 was in general low scoring on emotional word use (see Chapter 6).This observation becomes especially clear when

<sup>64 &#</sup>x27;HTK 1965–66', 2233.

<sup>65</sup> Proposals for adjustments to the WBP and WBPZ were scheduled – but not debated – on 8th of May, 29th of June, 6th of July 1971, 28th of August and 9th of September 1975, and the 23th of November and the 7th of December 1976.

<sup>66 &#</sup>x27;HTK 1973–74', 3838.

<sup>67 &#</sup>x27;HTK 1973–74', 3843.

this particular debate is compared to the debates of the mid-1950s. My analysis indicates that the acknowledgement of trauma did not lead to a significant increase in public expressions of strong negative emotions in the parliamentary discussion on the resistance pensions in the Dutch House of Representatives. In the early 1970s, more and more social legislation for other groups of war victims was established. Discussing these arrangementsin parliament wasincreasingly merged with the discussion of the resistance and sailor-war victims benefit legislation. Chapters 8 and 9 deal more elaborately with the question how developments in these merged discussions relate to the historical phenomenon of 'the emancipation of emotions'.

#### **7.5 Conclusion**

The establishment of a resistance pension, which guaranteed an income for those people (or their descendants) who had become unable to support themselves because of invalidity or death caused by anti-Nazi resistance activities, was considered a payment of a 'debt of honour'. The post-war Dutch government took on this debt to create a connection between the sacrifices these people had made for the fatherland and the recovering post-war political order. The heterogeneous conglomerate of individuals, groups, and activities, united under the umbrella-term of 'the Resistance', stood for what had been morally just during the war. Post-war Dutch politicians were willing to associate the political status quo, their parties, and themselves with the former anti-Nazi resistance. Not just appealing to a personal or party resistance past, but also the appropriation the former anti-Nazi resistance seemed a welcome instrument for promoting political legitimacy and moral virtue right after the war. Compared to other groups of war victims, the heterogeneous group that can be summarized as 'extraordinary government employees' held an exceptional position until the 1970s.Their tailored pension act was established in 1947, and modifications and improvements to this legislation were discussed repeatedly over the years.

The emotion-mining results of Chapter 6 showed there was, in the first place, a rather erratic coursein the elaborateness of parliamentary debates over the decades. Discussion was not scheduled on a regular basis. In addition, the scores for language use associated with emotions in the Dutch parliamentary discussions on 'extraordinary government employees' showed great variation.This particularly encompassed the scores for word use associated with anger, fear, and disgust. The observed variation in emotion scores over time was irregular, as neither a linear increase nor decrease was observed. Close reading of the discussions indicated that from the late 1940s onwards, the parliament displayed a fundamentally consensual attitude towards the welfare legislation for this particular group. What I have called a 'resistance disposition' before, was reflected by relative resigned tone of voice in the first parliamentary debate on the issue in 1947. The analysis of the emotional language

use in the first debate engaging with this act indicated indeed that strong negative emotions, such as anger, fear, and disgust, were less present in these debates, compared to the average parliamentary debate of that era, or to later discussions on the same topic. That (the subject of) a legislative scheme or parliamentary issue had an evidently emotional nature did not mean emotions also always played a predominant role in its discussion. The results of applying emotion mining to the 1947 debates demonstrate how that what was considered important, was not necessarily always accompanied by strong (negative) emotional expressions.

In contrast to the calm debates of 1947, emotion-mining results (Chapter 6) indicated peak emotion scores in the 1950s. The results of close-reading analysis of these debates in this chapter showed that these emotional expressions were evoked by contemporary issues, rather than by reflection on the recent past. Suffering, sacrifices, or heroic acts of resistance during the occupation were not what was discussed. As the application of the WBP and WBPZ did not always meet all expectations, concerns about shortcomingsin the post-war living conditions of these people touched upon the 'resistance disposition'. Concerns about a possible impairment of the position of former resisters evoked the expression of emotions, as was observed in the manifestation of language use associated with negative emotions. The fear that people were falling through the cracks of the welfare legislation was a stimulus for eliciting sadness in particular.Words associated with sadness were, for example, prominent in elaborate descriptions of poignant examples of suffering of beneficiaries in 1955.These examples of contemporary circumstances were used to emphasise misery in the contemporary present, and account for much of the emotional language use in these debates. Nevertheless, the discussion between MPs was still of a rather consensual nature, as nobody in the House of Representatives fundamentally disagreed with a shared dislike for all too elaborate discussion in parliament, talking about money, or too narrow-defined eligibility. Advocating for more elaborate legislation, or for the exclusion of ideological adversaries, could be done by MPs without having to carry far-reaching practical or political consequences – as the difficult task of making the actual decisions on individual cases was outsourced to a non-governmental organisation.

The observed emotional peaks in the 1950s not only originated in the discussion of the modifications to the WBP and WBPZ legislation itself. High emotion scores for this period corroborate an existing idea that global politics in the Cold War influenced the engagement with the former resistance in the Netherlands. Rising Cold War tensions, for example, made discussions about the WBP a forum for casting aspersions on political opponents, or playing to the gallery. These discussions were more or less exploited for addressing contemporary political discord, interests, and concerns. Not despite, but merely because of the broadly shared consensual attitude and 'resistance disposition' the issue became a vehicle to 'frame' others and for moral posturing. Because of this disposition, displaying any disinterest in discussing 'the Resistance' seemed out of the question. In this way, the discussion of this topic in parliament forced engagement, also for political opponents. Observed developments in the role of emotions in these debates were, to conclude,more a consequence of doing contemporary politics, than resulting from shifting parliamentary attitudes towards the group of 'extraordinary government employees'.

The emotion-mining results gave a first indication of anomalies in emotion scores in the 1950s. The results of close reading the subset of WBP and WBPZ debates brought me to a conclusion regarding the role of emotions: Emotions in this subset were especially prominent in specific cases, and when they were, they were serving a contemporary aim. Historically lived or experienced emotions (of others) seldom came to the surface in the parliamentary discussions on WBP and WBPZ. Mentions of lasting (contemporary) suffering of beneficiaries, as was observed in 1955 and 1963, led to much negative emotional word use. In other cases, such as in 1954, the pressing political issues of the day evoked strong emotional expressions. The modifications and improvements of WBP and WBPZ legislation in subsequent years were often accepted by parliament without any discussion or vote at all. This caused, in the first place, a lacuna in the emotion mining measurements (see also Chapter 6). After close reading the few debates of the late 1960s and 1970s in the case study's subset, the acknowledgement of (war) trauma in both the professional and public sphere seemed to have only limited impact on emotional manifestations in parliamentary discussion. In the 1974 discussion of the WBP and WBPZ – one of the last debates where these acts were discussed separately from other victim legislation – this led to the striking increase in mentioning people's mental distress in parliament.

From the early 1970s onwards, Jews and other groups of victims of persecution, civilian war victims, and even second-generation victims also gained the rights to tailored social welfare legislation. As a result, the exceptional position of the 'extraordinary government employees', compared to other groups of war victims, was not so exceptional anymore. And so their quality as a vehicle for addressing contemporary political issue dwindled. How this developed in the case of other groups of war survivors, is discussed in Chapters 8 and 9 by addressing how benefit acts for other groups of war victims came into being. I do not consider it exceptional that these acts were developed relatively 'late' in the early 1970s. Such victim legislation emerged against a background of a rapid and elaborate expansion of the scope of national welfare state policies at that time. The establishment of the WBP and WBPZ in 1947 had been, in the pre-welfare state context, the true exception.

## **Case Study 2 | 'War Victims'**

The second case study deals with the parliamentary discussions about legislation aimed at a greater diversity of different kinds of war survivors.The common denominator here is that they were all more or less considered victims of the German occupation of the Netherlands during World War II.The characteristics of the debates in this second case study are more varied, compared to those assessed in the first case study. Contrasting the 'extraordinary government employees', who had their pension scheme from 1947 onwards, 'war victims' cannot only be considered as an even more heterogeneous group of people, also the definition of who was considered as a 'war victim' was subject to change over time. As a result, the legislation aimed at alleviating their suffering shows more variety in different kinds of legislation, acts, and social welfare schemes over the years.The term 'war victims' is here broadly understood and refers to an incoherent and changing group of people with divergent experiences of suffering related to,amongst other things,material damage,warfare, organised violence, Nazi violence, persecution, and repression.

This introduction of the second case study proceeds with the legislative history of the post-war engagement with 'war victims' in Dutch politics between 1945 and 1989. Central to this first part is the history of different legislative acts and proposals over the years, as this is the first and foremost reason why parliamentary discussion on 'war victims' took place. I do not see the legislation as such as a potential expression of emotion,or as representative of the contemporary 'mental climate'or societal engagement with war-related emotions. Manifestations of emotions can, however, be encountered in the parliamentary debates dealing with these legislative developments over the years. The results of emotion mining of these various 'war victim' debates are presented in Chapter 8. Just as in Chapter 6, the results of applying emotion mining to a specific case are statistically evaluated and compared to the 'average parliamentary debate' that was analysed and discussed in Chapter 5. Furthermore, Chapter 8 makes a comparison between the results of all debates in the 'war victims' subset. Diachronic developments of emotions over time are assessed as well.The final part of the case study, discussed in Chapter 9, contains a more detailed analysis and interpretation of the role of emotions in the sources.

### **A History of Alleviating War Victims' Suffering in Parliament (1945–1989)**

Just as in the national engagement with 'the Resistance', Dutch government anticipated a role in assisting victims of the war already before the whole country was liberated from the Nazi occupier. Contemporary estimates stated that more than 800.000 people could be considered as 'direct war victims' in the summer of 1945. These people were in immediate need of food, water, clothing, housing, and medical supplies.<sup>1</sup> Dutch ministers-in-exile founded a central aid organisation (*Centraal Bureau voor de Verzorging der Oorlogsslachtoffers,* CBVO) by Royal Decree in the spring of 1945.<sup>2</sup> With the foundation of the CBVO, the Dutch government accepted at least some role in the care for basic needs of the thousands of war victims in the country.<sup>3</sup> The CBVO was gradually abolished in 1947 and 1948. Its tasks were redirected towards the municipal social services.<sup>4</sup> I see the national government's involvement with the aid for victims in these immediate post-war years primarily as an act of crisis management.

In the case of the former resisters and the sailor-war victims, tailored national legislation was already established in 1947. The position of these people, embedded in extraordinary pension acts known as WBP and WBPZ, was exceptional when compared to the national government's post-war engagement with other groups of survivors of the German occupation of the Netherlands. From 1947 onwards theWBP and WBPZ were subject to modifications, additions, and extensions.Therefore, they were recurrently discussed in parliament. In contrast, the historical development of national war victim legislation followed a more erratic, varied, and layered course. The legislative history of treating war victims in the Netherlands can, I argue, be compared to tossing a stone in a pond: After the stone disturbed the water surface, the following ripple undulated.This then triggered a next ripple, and drew new, ever larger rings in the water.

The CBVO is the stone. The *Wet Materiële Oorlogsschaden* (MOS), or Material War Damage Act, followed. This legislative scheme was aimed at Dutch citizens in the Netherlands who suffered material damage or losses because of the war. The first post-war governmental interventions regarding war victims were, according to the Dutch historian Martin Bossenbroek, mostly characterised by a focus on collectivity and materiality. Although there were some exceptions, in general, individual vic-

<sup>1</sup> Bossenbroek, *De Meelstreep*, 183.

<sup>2</sup> This bureau was founded by Royal Decree before the first post-war parliament took office. As a result, the discussion on its establishment is not recorded in the *Handelingen* dataset used in this investigation. See also Bossenbroek, 80–89.

<sup>3</sup> Bossenbroek, 131; Piersma, *Bevochten recht*, 30.

<sup>4</sup> Bossenbroek, *De Meelstreep*, 367–68.

tims, or particular victim groups did not receive any specific treatment.<sup>5</sup> Every victim was considered another 'wound' to the 'popular body'. The first national government aid schemes were mainly aimed at recovery, and not at care or compensation. Thus, neither individual nor divergent experiences of occupation, but current material needs and facilitating the reconstruction of the country were key in the MOS. Basic needs – a roof over people's heads – were urgent and directly visible necessities.Government and parliament considered alleviating these needs not only as priority, but also as their main responsibility. Bossenbroek pointed out how this shaped – perhaps even unconsciously – the general understanding and definition of 'war victim'.<sup>6</sup> The act was elaborately discussed and eventually established on the 9th of February 1950.<sup>7</sup> Although full restitution of all material losses, of every damaged house, or every lost armchair, was out of the question, many people affected by the occupation were back on their feet with the implementation of the MOS.

#### The Scope of National Legislation

Against a background of industrialisation, Marshall Payments, European cooperation, and the independence of Indonesia, Dutch government was re-orienting its roles and tasks. The Dutch economy not only recovered in the 1950s, the Dutch state also drastically restructured the national economy. Employment was on the rise and the economy grew. These circumstances created space for, not only, decisions regarding increasing government spending, but also for a broader perception of the government's fields of policy.<sup>8</sup> The Dutch government developed policies with

<sup>5</sup> There was one exception in the 1950s: The case of the so-called *Oorlogspleegkinderen* (foster children, or orphans of war). These children – mostly Jewish – had survived the war in hiding with clandestine workers or foster parents. Especially the orphans left behind by parents that did not return after the Liberation, were at the core of public concern and commotion. A special temporary committee was established in 1945, and the contemporary dealing with these children caused much public discord. For an elaborate account of the course of events, see Bossenbroek, 247–57.

<sup>6</sup> As the name 'Material War Damage Act' already suggested, physical wounds, invalidity, psychic suffering, or other consequences of the war for individuals were not part of this. Neither individual experiences of war (e.g., persecution, or Arbeitseisatz), nor belonging to a particular group of victims (e.g., Jewish victims of persecution, victims of bombings, etc.), or postwar consequences were playing a role in defining eligibility for this compensation scheme. See Bossenbroek, 400.

<sup>7</sup> With a national income of 10 billion guilders in 1946, the total MOS payment of more than 6 billion was a decent expense in favour of the reconstruction of the war-torn society. See Bossenbroek, 277–80, 397–400.

<sup>8</sup> Sjoerd Keulen, *Monumenten van beleid: De wisselwerking tussen Nederlands rijksoverheidsbeleid, sociale wetenschappen en politieke cultuur, 1945–2002* (Hilversum: Uitgeverij Verloren, 2014), 152, 293.

a more active role in the private lives and welfare of its citizens. One example was the establishment of the Dutch Ministry of Social Work in 1952. The Dutch parliament also grew – quite literally: The number of seats in the Senate went from 50 to 75, and the House of Representatives grew from 100 to 150 seats.<sup>9</sup> Not much later, in the 1960s, a relatively modern, elaborate, and expensive welfare system was put together. Government spending boomed as the *Kinderbijslagwet* (General Child Benefit Act), *Algemene Bijstandswet* (General Social Assistance Act, ABW), *Arbeidsongeschiktheidswet* (Incapacity for Work Act), and *Ziektewet* (Sickness Benefits Act) were all established and implemented between 1963 and 1967. With the *Bijstandswet,* government acknowledged primary responsibility for general poverty relief. Social assistance benefits for people in need became a right, and a direct task of the national government.<sup>10</sup>

The re-orientation of the government resulted in increasing government intervention and spending. It was an essential building block in the construction of more elaborate and tailored war victim legislation. Arrangements for war victims were transformed into a so-called *Rijksgroepsregeling Oorlogsslachtoffers* (National Group Arrangement for War Victims, RO). Following the evolution of social legislation, war victims were incorporated in the establishment of the ABW (or *Bijstandswet*) in 1965. Although they now received a special status within this legislation*,* the practical implementation of governmental care for war victims was still organised on the level of municipalities – just as this was the case for the 'regular' ABW beneficiaries. Being incorporated in general social benefit legislation became a source of parliamentary opposition towards government policy in the 1960s. Although responsible Minister Marga Klompé considered it 'unimaginable' that people demanded more, opposition-party MPs considered the incorporation of war victims in the *Bijstandswet* 'an unfair degradation of the status of war victims'.<sup>11</sup> Especially MPs of opposition parties made clear they considered incorporation in the *Bijstandswet* as dishonourable, shameful, and painful. This resulted in a persistent demand for tailored and autonomous legislative schemes for particular groups of war victims.<sup>12</sup> There was, however, more needed before such a scheme was broadly accepted and eventually established.

<sup>9</sup> Bas Kromhout, 'Kamer van honderd kon werkdruk niet aan', *Historisch Nieuwsblad*, 2010, 29.

<sup>10</sup> Jan Luiten van Zanden and Ed Lof, *Een klein land in de 20e eeuw: economische geschiedenis van Nederland 1914–1995* (Utrecht: Het Spectrum, 1997), 93–94, 96.

<sup>11 &#</sup>x27;Handelingen Tweede Kamer 1961–1962', n.d., 3224–25; 'HTK 1962–63', 3070–71; Piersma, *Bevochten recht*, 143.

<sup>12 &#</sup>x27;Handelingen Tweede Kamer 1960–1961', n.d., 3243; 'HTK 1969–70', 105; 'HTK 1970–71', 3477, 3485, 3495.

#### Ever Larger Rings in the Water

Developments in the medical profession and the so-called proto-professionalisation of trauma paralleled with a growing public attention for 'late consequences' ofWorld War II (see also Chapters 1, 6, and 7).<sup>13</sup> Ido de Haan already pointed out that developments in the public attention for, in the first place, the persecution of the Jews under Nazi rule, was not necessarily subject to a strong quantitative change in the 1960s. De Haan referred for example to the sales figures of books on the persecution of the Jews of the 1950s compared to those of the 1960s.<sup>14</sup> On different levels, however, things seemed to be changing. War victims, for example, increasingly managed to find each other and were now organising themselves ever more successfully. Interest groups of particular groups of war victims had become official and established organisations. Together with journalists, activists, writers, psychologists, and victims themselves these groups actively participated in public debates.<sup>15</sup> From the late 1960s onwards, the organisation and centralisation of these victims' organisations gave a final push towards the development of tailored and autonomous legislative schemes.

The development of interest groups, and the increased attention for 'late consequences' and related late-onset suffering also resonated with the Dutch parliament.<sup>16</sup> The demand for tailored legislation, separated from the *Bijstandswet,* lasted. Influenced by personal connections with and active lobbying of the foundation *Joods MaatschappelijkWerk* (Jewish SocialWork, JMW),MP Joop Voogd became particularly concerned with the circumstances of war victims.The opposition-party MP became one of the key figures advocating for more elaborate national arrangements. He was especially focused on the care for victims of persecution under the Nazi occupational regime.<sup>17</sup> Voogd had a scheme similar to the WBP in mind. He successfully put the issue on the parliamentary agenda by submitting multiple parliamentary motions in the early 1970s.<sup>18</sup> Although Voogd had paved the way,it was eventually a (stripped-

<sup>13</sup> De Swaan and Hendriks, *Zorg en de Staat*, 249–52; Withuis and Mooij, 'From Totalitarianism to Trauma', 211; Withuis and Mooij, 'Conclusion: The Politics of War Trauma', 328.

<sup>14</sup> De Haan, *Na de ondergang*, 116–18.

<sup>15</sup> See also the role of these actors in the Breda Three controversy in 1972, discussed in Chapters 1 and 5 and in Piersma, *De Drie van Breda*.

<sup>16</sup> In February 1970, for example, an open letter, published in national newspapers and signed by social workers involved in the care for Jewish victims of war and persecution, eventually led to discussion in the Dutch parliament. See also 'HTK 1969–70', 101.

<sup>17</sup> Joop Voogd (PvdA) was an opposition-party member of parliament that had several personal meetings and encounters with representatives of the JMW. They convinced him that current arrangements for Jewish victims of persecution were not sufficient. See also Piersma, *Bevochten recht*, 136–39.

<sup>18 &#</sup>x27;HTK 1969–70'; 'HTK 1970–71'; 'Handelingen Tweede Kamer 1971–1972', n.d.

down) motion by his coalition-party colleague Jacques Baruch (Democratic Socialists '70, DS'70) that led to the establishment of the *Wet Uitkeringen Vervolgingsslachtoffers 1940–1945* (Benefit Act for Victims of Persecution 1940–1945, or WUV) in 1972. <sup>19</sup>

After increasing attention, and a higher level of organisation, a first group of war victims had gained a status that was now confirmed by national recognition of their victimhood – shaped as a tailored welfare legislation. The path towards the WUV had been, at first, preceded by a process of centralisation and organisation of victim groups. In a next phase, these interest groups had found their way to opposition-party MPs. The parliamentarians that were imbued with the importance of a tailored welfare scheme supported and advocated for the war victims' case. In this process, the people involved referred to existing legislation (in the case of the WUV, this was mainly the WBP) to advocate for a tailored act for a 'new' group of war survivors. Finally, coalition-party MPs and responsible ministers and secretaries of the state were forced into a position where they could not say no anymore. Looking at the legislation and decision-making processes that followed theWUV, the establishment of this particular act can, in retrospect, almost be considered as some sort of blueprint: Each following group of war victims gained their rights to tailored welfare legislation by following an almost similar pattern.

#### Another Group, another Act, another Joop

A next ring in the water surface was the legislative scheme aimed at alleviating the contemporary suffering of so-called civilian-war victims. Again, increasing public and media attention for a particular group of victims, combined with a higher degree of organisation of the (interests of) the people involved, led to parliamentary attention for their case.<sup>20</sup> The group of civilian-war victims had not acted against the Nazi occupier, nor had they actively been persecuted. The group consisted, for example, of victims of (Nazi) violence, or survivors of (allied) bombings.The civilianwar victims were, generally speaking, people who had been at the wrong place at the wrong time.What Joop Voogd had been for the victims of persecution, his namesake Joop Worrell (PvdA) became in the case of civilian-war victims.The act aimed specifically at their problems and suffering, the *Wet Uitkeringen Burger-oorlogsslachtoffers*

<sup>19</sup> Van der Heijden, *Dat nooit meer*, 466–67.

<sup>20</sup> In the late 1970s an initiative by a victim of the allied bombing of Haaksbergen in 1945, Dirk Reimelink, led to the establishment of *Stichting Burger-oorlogsgetroffenen* (Foundation Civilian-War Victims, SBO). More public and media attention for this particular group of victims, outcomes of a governmental committee report, combined with an increasing degree of organization of the (interests of) stakeholders, led eventually to political and parliamentary attention for their case. See also Piersma, *Bevochten recht*, 187–90.

*1940–1945* (Benefit Act for Civilian-War Victims 1940–1945, WUBO), was established in 1983.<sup>21</sup>

Compared to the time when the WUV had been established, however, economic and political precedents were less favourable for the establishment of (new) elaborate social welfare schemes in the 1980s. The crumbling of important building blocks of this type of social benefit legislation affected the scope and generosity of the WUBO.<sup>22</sup> Where the WUV can be considered as a stripped-down version of the WBP, the WUBO was a stripped-down version of the WUV.

Yet not all odds were against the WUBO. In 1975, an advisory committee was established. This committee, known as WAC, focused on the immaterial assistance of victims of persecution, bombardments, *Arbeitseinsatz,* etc*.*The WAC worked until 1978 and consisted of domain experts and specialists on the issue.<sup>23</sup> The WAC's recommendations were followed up by, for example, the official establishment of the *Informatie- en Coördinatieorgaan Dienstverlening Oorlogsgetroffenen* (ICODO), a foundation for permanent assistance and a source of information for a broad range of people who were victimised during World War II.<sup>24</sup> A schematic overview of these and all other acts taken into account in this introduction to the second case study can be found in Table 15.

The establishment of the WAC and ICODO were a first and foremost result of more general political changes. In the late 1970s and early 1980s, Dutch national governance in general had been subject to developments of professionalisation and technocratisation. In the context of dealing with war victims, the establishment of bodies like theWAC (1975–1978) and ICODO (1980–2005) can be considered as a concrete result of these developments.<sup>25</sup> As the WAC's recommendations played a supporting role in the establishment of the WUBO, this act was not only negatively impacted by contemporary social and political context.

<sup>21</sup> Van der Heijden, *Dat nooit meer*, 469.

<sup>22</sup> Piersma, *Bevochten recht*, 191–93.

<sup>23</sup> *Werk- en Adviescollege Immateriele Hulp aan Oorlogsgetroffenen* (WAC) was an advisory committee, established in 1975. Its primary aim was to advise the government on the problems related to immaterial 'late consequences' of Word War II. It led eventually to the establishment of ICODO, *Informatie- en Coordinatororgaan Dienstverlening Oorlogsgetroffenen,* a permanent organisation providing information on the issue to the government and to organisations responsible for implementing war-related (social) legislation.

<sup>24 &#</sup>x27;Onderzoeksgids Oorlogsgetroffenen WO2 − ICODO', accessed 29 July 2020, http://www.oorlogsgetroffenen.nl/archiefvormer/Stichting\_ICODO.

<sup>25 &#</sup>x27;Onderzoeksgids Oorlogsgetroffenen WO2 − WAC', accessed 22 July 2020, http://www.oorlogsgetroffenen.nl/archiefvormer/WAC.

*Table 15: A schematic overview of the national government's engagement with war victims in the Netherlands between 1945 and 1989.*


The Dutch historian Carla Hoetink described in her book *Macht der gewoonte* (*Force of Habit*, 2018) that increasing technocratisation was also reflected in the tone of voice in debates in the parliamentary arena.<sup>26</sup> Professional politicians, wrote Hoetink, increasingly distanced themselves from strong emotions. Expressions of emotions became, according to her, considered as less professional.<sup>27</sup> This brings us back to the role of emotions in the debates. Chapter 8 is the result of the first emotion-mining assessment of the question whether expressions of emotions were manifest in parliamentary discussions on these various war victim-related acts at all. Were emotions present in the language of the MOS discussions (that took place within the period often considered as one of 'silence' and 'concealment' regarding the expression of emotions)? What about emotions in the elaborate debates of the 1970s, when the WUV was established and discussed? Did an increasing professionalisation of Dutch politics paralleled with a more neutral or concealed tone of voice in parliamentary debates, as Hoetink suggested?

<sup>26</sup> Hoetink, *Macht der gewoonte*, 421–98.

<sup>27</sup> Hoetink, 425.

## **8. Emotional Consistency | A Macroscopic View on War Victim Debates**

National legislation aimed at alleviating the suffering of various groups of victims of the German occupation of the Netherlands was subject to a more erratic legislative development when compared to the resistance pension of 1947. Compared to the debates on the WBP or WBPZ legislation, the subset of parliamentary debates on war victims is therefore also of a more heterogeneous nature. National war victim schemes were developed in interaction with changing ideas about who was considered to be a 'war victim', shifting interpretations ofgovernmental responsibilities, interventions, and fields of policy, economic developments, and the establishment of the welfare state in the Netherlands. Most debates in case study 1 dealt with two closely related welfare schemes (WBP and WBPZ) that were repeatedly subject to modifications. The subset of parliamentary debates in the current case study on 'war victims' consists of various discussions on different and diverging legislative schemes over the years. Because of this course of historical developments, the debates dealing with war victim-related legislation also cover a wider variety of different groups of people over time. As a result, not only the legislative schemes, the groups of people that they dealt with, but also the corresponding and related parliamentary discussions over the years can all be considered part of the chronology of various initiatives and acts aimed at alleviating the suffering of victims of World War II in the Netherlands.

#### **8.1 Emotions and War Victims in Parliament**

#### Thematic Clusters

All debates in the 'war victims' subset deal with the establishment, modification, or discussion of different legislative schemes. These debates are connected by the fact that they are all related to and/or aimed at alleviating the suffering of victims of the German occupation of the Netherlands during World War II. As the debate subset in this case study deals with various themes, all individual debates within the 'war

victims' subset are merged into multiple thematic clusters (see Table 16). This process is more elaborately described in Chapter 3. In addition, the 'war victims' subset also contains more than two times the number of debates, compared to the 'extraordinary government employees' subset used before. The thematic 'war victims' clusters are based on various debates on a single law, act, proposal, or issue. They follow – where possible – the original historical topic descriptions in the *Handelingen* dataset. I want to stress that eventual differentiation between types or categories of 'war victims' or related legislative schemes that are used here are not primarily my construct. Categorisations are, where possible, derived from different post-war legislative schemes, or stages in the post-war governmental engagement with this rather heterogeneous group.


*Table 16: Thematic clusters of debates dealing with war victim-related legislation in the Dutch parliament between 1945 and 1989.*


Table 16 shows that not every war victim-related topic discussed in the Dutch parliament was directly related to a particular act or welfare scheme. In the 1950s, for example, several issues regarding individual cases of war victims were debated in the parliamentary arena. Because these topics, however, cannot be considered as part of the historical processes of the establishment and discussion of the various acts aimed at alleviating the suffering of war victims, these issues are only taken into account in the emotion-mining analysis of this chapter. They are not further analysed in Chapter 9. The thematic clusters of debates in Table 16 are chronologically ordered.This applies also to all other tables and figures used from now on.This order is based on the first occurrence of the theme on the parliamentary agenda.

This chapter proceeds with the results of the computational analysis of emotions in the abovementioned war victim debates in the Dutch parliament. The quantitative results of this chapter are synthesised with a qualitative 'close reading' historiographic approach to the parliamentary debates in Chapter 9. There, detailed attention is also paid to the historical processes of the establishment and discussion of the various acts.

#### Variation and the Pollyanna Principle

The words associated with each of the six emotion categories from the NRC EmoLex are identified, weighted, and thus 'scored' in all 250-word text chunks of the documents within this subset. A more detailed description of this procedure can be found in Chapter 4.The complete overview of the results of computational analysis of emotion word use in the debates (belonging to this subset) can be found in Supplement 3.2. This also applies to the statistical evaluation and the output.<sup>1</sup> Summary statis-

<sup>1</sup> Just as in Chapters 5 and 6, when I mention the 'score' of a specific parliamentary debate on a certain emotion, this score consists of the mean of all TF-IDF weights for all of the 250-word chunks of which that debate consists. Theoretically, this does however not has to mean that there is (only) a higher frequency of emotion words in these debates. Higher TF-IDF scores are also caused by the (relative) frequent use of (relative) rare, uncommon occurring emotion words. Because of the TF-IDF weighting that is applied to the scores, the (single) occurrence of relatively rare words weighs heavier than the individual occurrence of a very common emotion word. For a detailed description, see Chapter 4.

tics of the scores of the subset are calculated and presented in Table 17. Summary statistics show that the mean and median score of trust is again much higher than all other emotion scores. Corroborating the earlier findings of this study, the emotion-mining results again provide empirical evidence for the Pollyanna principle. The table shows that this also goes for the particular subset of debates related to war victims.


*Table 17: Summary statistics of emotion scores of all war victim debates (1945–1989).*

To put the results in Table 17 in perspective, the war victim debates can be compared to the scores of the debates on 'extraordinary government employees'of Chapter 6. When Table 17 is compared to the summary statistics of the results of the 'extraordinary government employees' debates(see Supplement 3.1),a first observation is that most of the mean (and median) emotion scores of the subset are higher in the war victim debates. Only the emotion joy scores slightly higher in the resistance debates. More meaningful, however, is the observation that the war victim debates have much lower coefficients of variation compared to the resistance debates discussed earlier on.This applies to scores for words associated with all basic emotions. Despite the more heterogeneous nature of the topics discussed, these results indicate a more even distribution of emotion scores over the different individual debates within the victim subset. To put it differently: The variation between the (individual) debates in the 'war victims' subset is considerably lower, compared to the variation between the (individual) resistance debates.These findings suggest that the debates about war victims are mutually more similar, in contrast to the resistance debates that show greater variation. As discussed in Chapters 6 and 7, the resistance subset was characterised by (a few) anomalously high peak debates.

#### Little Variation between the Clusters

Boxplot 4 displays the scores for all 250-word text chunks in each of the debates belonging to the thematic debate clusters related to the discussion of war victims.The boxplot gives, in the first place, a visual representation of the elaborateness of the debates on various issues in terms of number of words (in Boxplot 4 represented as 250-word text chunks that are each plotted as •). As can be observed, especially the theme of material war damage was discussed more elaborately than the other themes. More important, however, is the distribution of the emotion scores, and the comparison between the means () and medians (**–**).Boxplot 4 shows the scores for sadness, within this subset the emotion with a relative high coefficient of variation (see Table 17). Nevertheless, this boxplot gives a visual representation that shows only marginal variation. Not only is the variation between the different individual debates already relatively small, this also applies to the variation in scores between the thematic clusters of debates. Although there is some minor variation within the clusters, and the one thematic cluster displays more skewed scores than the other, most of the boxes are in the same range.This becomes even clearer when comparing these boxplots to the ones displaying the scores of the 'extraordinary government employees' debates discussed earlier (see supplement 3.1 and Chapter 6).

#### *Boxplot 4: Sadness scores for each 250-word chunk in the 'war victims' debates subset.*

Based on Boxplot 4, no anomalies within the subset of case study 2 can be reported. This is not only true for the scores for the emotion of sadness, but also for all other negative and positive emotion categories (see Supplement 3.2.1). Based on these emotion-mining results, there is no indication to assume that there were thematic clusters of debates really standing out in terms of emotional charge compared to others within the 'war victims' subset.

#### Significance and Effect Size

Just as in the case study of 'extraordinary government employees', the scores of the war victim debates are subjected to a non-parametric Kruskal-Wallis test. Despite the low coefficients of variation, based on these test results, statistically significant differences between the means of all debates can be reported (see Table 42 in Supplement 3.2.2).This is equally the case for the means of each thematic cluster (see Table 43 in Supplement 3.2.2). These outcomes, however, do not tell which individual debates mutually varied strongest within this subset. Therefore, a pairwise Wilcoxon rank sum test is also applied. A confidence interval of 95% is used.<sup>2</sup> The test is performed for both the individual debates (just as in Chapter 6), but also for the thematic clusters of the 'war victims' subset. This test calculates the pairwise statistical significance of scores for each emotion and for every possible pair of debates. The test was repeated to make the pairwise comparisons between the scores of the debate clusters.

Pairwise comparison of, for example, the anger scores of the thematic clusters shows in the first place that statistically significant differences are only reported in a minority of the cases (12 out of 45 thematic cluster pairs differed significantly on anger). This also goes for the scores for the other emotions fear, sadness, disgust, joy, and trust. More meaningful, however, is that for almost all pairwise comparisons and all negative emotions (very) small effect magnitudes can be reported (see Supplement 3.2.3). This means that, in general, even if there is significant variation between scores of the different thematic clusters within this subset, the size of this effect is often only tiny. This applies to the overwhelming majority of the pairwise comparisons between the thematic clusters on both negative and positive emotions. These results corroborate with the aforementioned findings based on the calculation of summary statistics and visualisation in boxplots.

<sup>2</sup> This means a p-value equal to or smaller than 0.05 is interpreted as statistical significant. For the rstats package, see R Core Team, *R: A Language and Environment for Statistical Computing*.

#### War Victim Debates over Time

To gain more insight in the diachronic development of emotions in the 61 debates on war victims, themean scores of allindividual debates are plotted on a timeline.These visualisations show individual debates,instead of the thematic clusters presented in Boxplot 2. Just as shown in Chapters 5 and 6, each different colour represents a different emotion in Graph 14 and Graph 15. Each bar represents a single debate. The non-war-related parliamentary debates (aggregated daily mean score) are plotted in grey dots. The general trend of the non-war-related debates is plotted in each of the graphs as a black line, whereas the trendline of the war victim debates is plotted according to the emotion-colour scheme. Although not as anomalous as in the 'perpetrators and collaboration' debates (see Chapter 5, Graph 11), Graph 14 and Graph 15 show that the 'war victims' subset scores higher for the negative emotions than the average parliamentary debate between 1945 and 1989. Despite the lack of variation in the scores of the 'war victims' subset observed earlier, the diachronic graphs indicate the existence of some modest peak debates in the subset. These peaks can, for example, be identified in the immediate post-war period, around 1950, in 1963, the early 1970s and in 1983.The graphs, however, also show that these peaks are relatively evenly distributed over the 1945–1989 period.Whether these peaks mark some of the decisive moments in the development and establishment of war victim legislation, will be further discussed in Chapter 9.

*Graph 14: Anger scores of parliamentary debates dealing with war victims (red) compared to average parliamentary debates' scores (light grey and black) between 1945 and 1989.*

*Graph 15: Sadness scores of parliamentary debates dealing with war victims (grey) compared to average parliamentary debates' scores (light grey and black) between 1945 and 1989.*

#### Gap Years All Over Again?

The abovementioned plots display again a gap in the late 1950s and most of the 1960s in the plotted time series with topic-specific bars. The scores of parliamentary debates in 1963 are the only exception.This gap – and the exception – were already observed in the timelines plotted in Chapter 6 (see Graph 12 and Graph 13). Regarding the exception of 1963, this is no surprise, given the fact that discussions on the General Treaty are taken into account in both the subsets of case study 1 and case study 2.These debates dealt not only with the former resistance, but also engaged with the fate of victims of persecution after all. In contrast to the 'extraordinary government employees' debates of Chapters 6 and 7, however, the observed gap was not the result of the fact that debates on acts were scheduled but voted for or accepted without any discussion.This became apparent already in the selection procedure (see Chapter 3). The observed trend – and its sudden disappearance – points at a change. Something was going on. What caused these gaps in the measurements? Here, also the quantitative empirical measurements, plotted in Graph 14 and Graph 15, do not provide us with any further indication of what could have been going on in these gap years.Was there, except from 1963, a parliamentary phase of silence regarding anything related to war victims? Was there no related debate scheduled at all in these years? These questions require more detailed scrutiny.This will be provided in subsequent Chapter 9. First, however, some general characteristics of the 'war victims' debates subset are addressed in comparison with other parliamentary subsets discussed before.

#### A Lack of Change

Based on the visualisations of the emotion-mining results, I consider three observations of special importance here. First, except for the obvious 'gap years', Graph 14 and Graph 15 show that most war victim debates held between 1945 and 1989 are relatively evenly distributed over the timeline. Second, compared to the debates on the 'extraordinary government employees', there is a relatively high number of parliamentary debates on war victims in the period 1970–1989. Yet what about the scores for emotional language use of these debates? Despite a peak in 1983, the coloured lines display a minor decreasing trend in emotion scores. That brings me to a third observation:The graphs indicate that there was no strong increase in emotional language use observed in the Dutch parliamentary debates on war victims between 1945 and 1989. Results of linear regression analysis point in the same direction as the results of this same statistical test on the 'extraordinary government employees' subset (see Supplement 3.2.5). The general trend in the quantity of emotional language used is, in the case of the war victim debates, for most years one of stability and consistency. If there is a general trend observable between 1945 and 1989 in the graphs, it is a minor decrease in the emotion scores.The dwindling of this curve is observed in the first ten post-war years. Based on these results, no ongoing increase in emotionality from the 1970s onwards can be reported. If there was an exceptional peak observable in the trendline, this concerned the scores of the discussions in the immediate post-war years, rather than those of the 1970s (or later).

#### **8.2 Stable Trend, Same Words?**

The observations outlined above mostly encompassed quantitative characteristics of the manifestation of emotions in war victim debates. By zooming in the historian's macroscope, and looking at the lexicon words that are behind the scores, a first indication of the qualitative developments in the actual course of the historical debates can be established. Graph 16 displays the 20 highest-scoring sadness lexicon words in the debates. The top 20 highest-scoring unique emotion lexicon words in the war victim debates for the other emotion categories can be found in Supplement 3.2.6.

Compared to the quantitative stability of the trendlines of the war victim debates, a closer look at the emotion-mining results and 'the words behind the scores' indicate change in the qualitative characteristics of emotional language use over the decades. The types of words in the top 20 word lists of high-scoring emotion words within the subset indicate a semantic shift in emotional language use (see Graph 16). This change can be observed in the 1970s discussions leading up to the establishment of the *Wet Uitkeringen Vervolgingsslachtoffers 1940–1945* (Benefit Act for Victims of Persecution 1940–1945, WUV) in 1972. The observed semantic shift is characterised by a decrease in the use of words used to describe circumstances or events associated

with certain emotions (e.g., 'bombardment', 'suffering', etc.), on the one hand, and an increase in the manifestation of words more explicitly referring to (the expression of) a certain emotion (e.g., 'fear', 'regret', 'bitter', or 'crying'), on the other. This development is most obvious in the lists of words from the emotion categories fear and sadness (see Graph 16 and Supplement 3.2.6).

#### Diverging Distributions of Emotion Words

The top word graphs give a visual representation of each word's share in the emotion score per thematic category of debates.This visualisation also provides a representation of the distribution of those 20 highest-scoring words within each cluster of debates. Especially in the WUBO debates in the 1980s, only one or two unique words fuel the scores for negative emotions anger, fear, and sadness. The number of different words leading to (similarly high) emotion scores is much higher in earlier debates, for example in debates on foster children (1950s), the discussion on the General Treaty (1963), and some of the discussions leading to the establishment of the WUV (1970–1972). The top five of the positive emotions of joy and trust, and the emotion of disgust, display a similar, much more distributed pattern in the WUBO debates compared to the distribution of the top words behind scores for anger, fear, and sadness (see again Graph 16 and Supplement 3.2.6).

The skewed distribution of high-scoring words that can be observed in the WUBO debates means that the relative high emotion scores of these discussions are provoked by only a very limited number of unique words. Closer scrutiny of these debates indicates that these words were, amongst others, mostly the noun 'injury', the adjective 'disabled', and the verb 'to explode' that account for these high emotion scores. These words are associated with negative emotions anger, fear, and sadness according to the NRC EmoLex (see also Supplement 3.2.6). In the first instance, however, such terms seem to point at a limitation of emotion mining as a method: These words were mainly used to refer to the issue at hand, and not in dressing a rhetorical argument, or giving explicit expression to a lived emotion. Nonetheless, that these words were used and apparently predominant, does not say that other emotion words were entirely absent. The question that remains to be answered in Chapter 9, however, is as follows: Do these quantitative aspects of the emotional language use in these discussions mean that these debates can be classified as neutral or unemotional? Or does close-reading analysis and a focus on the more qualitative aspects of the debates shed a different light on the matter?

*Graph 16: Top 20 sadness words for each thematic cluster in the 'war victims' debates subset*

#### Peak Moments, Quantitative Stability, and Qualitative Development

To conclude this chapter, the computational analysis of the manifestation of different emotions in the parliamentary discussions on war victims indicated in the first place a general pattern of relative stability in emotion scores over time. A comparison between emotion-mining results in case studies 1 (Chapter 6) and 2 (Chapter 8) indicates in the first place that the 'war victim' debates, compared to those on 'extraordinary government employees', display fewer anomalies and less variation. Scores of the different thematic clusters related to war victims also do not show strong or exceptional outliers. As an exception, a few particular debates that took place in 1950, 1963, the early 1970s and 1980s, show somewhat higher scores for some of the negative emotions. In comparison, however, the WBP and WBPZ debates of Chapters 6 and 7 displayed a much more erratic pattern and are characterised by strong outlier (high) scores. In this chapter, I have assessed the emotionmining output, and statistically evaluated, visualised and discussed the results on a rather abstract level. Emotion proved a constant in the debates within this case study. Despite the exceptional 'gap years', it seems that emotions played a role over the period between 1945 and 1989. How did this role unfold and develop over the years? A question that remains, in other words, is *how* these findings relate to the role of emotions in post-war parliamentary discussions on the rights of and welfare provisions for various groups of victims of the German occupation in the Netherlands. This will be addressed in more detail in Chapter 9.

## **9. Emotional Scaffolding | The Construction of War Victim Legislation in Parliament**

This chapter deals with a more detailed scrutiny of the parliamentary debates on the rather broad group of people summarised under the term 'war victims'. The results of emotion-mining analysis from Chapter 8 are combined with the results of a close reading of the actual historical parliamentary debates.The quantitative observation of relative stable emotion scores over time, points in the first place at emotionality as a constant in the debates. What formulations, rhetorical practices, speeches, and discussions were underlying these scores? Additional questions are raised by an observed semantic shift in frequently used emotion words.This shift can be considered as one of a decrease in the use of 'words associated with certain emotions' and an increase in 'words more explicitly referring to (the expression of) a certain emotion'. What does this observation mean for the role of emotions in the discussions on the establishment of war victim legislation in parliament? Did that role change? And, if so, how? Were emotions a topic of discussion? Did parliamentarians express their own emotions?This chapter assessesin particular how these questions relate to commonplace ideas, such as a historiographical periodisation (or 'cycle') related to 'silence', 'emotionalisation', and 'emancipation of emotions' regarding the government's engagement with war victims that was discussed in Chapter 1. Were emotions emancipating? And if so, whose emotions were emancipating?

#### **9.1 An 'Excessively Benevolent Willingness'**

One of the first post-liberation parliamentary debates in the subset dealt with the establishment of the Material War Damage Act (MOS). It is striking how elaborate and lengthy these debates were. The members of the Dutch House of Representatives seemed to balance between financial affordability, and fear to fail in providing hundreds of thousands of war-affected Dutch citizens with their basic needs.<sup>1</sup> An at-

<sup>1</sup> In the quantitative analysis of these MOS debates more than 800 chunks (of 250 words each) are taken into account.

mosphere of crisis management is particularly characteristic for these early debates of January 1946. On the one hand, there was a self-proclaimed 'excessively benevolent willingness' amongst MPs to meet people's material needs out of compassion. On the other hand, there was a broadly shared feeling that the Dutch people were simply legally entitled to compensation. MP Chris van den Heuvel (ARP) noted in 1949 that the proposed MOS act was based on these two rather contradictory beliefs. Looking at the eventual legislative formulations, one can conclude that not one or the other side prevailed unambiguously.<sup>2</sup>

As discussed in the previous chapter, the national approach to the matter of alleviating suffering of war victims was characterised by two aspects: collectivity and materialism. These key aspects resonated with the general attitudes of the related parliamentary discussions.<sup>3</sup> Regarding the collective approach, Henk Hofstra (PvdA) explicitly named and praised the approach of breaking with individualism for the good of the national community.<sup>4</sup> Concerning the material characteristic of the MOS, member of the House of Representatives Pieter Zandt (Dutch Reformed Political Party, SGP) noted the following regarding Dutch victims:

'First, we want to bring to the attention of the government, that these people have already suffered much in the war. Many of them are plunged in deep mourning because of the loss of their close relatives. This loss, that many of them still feel every day, the government cannot compensate. No more than the sorrow and the misery that the war victims have undergone because of evacuation or otherwise. We do not ask this from her [the government].'<sup>5</sup>

I consider the quote above to be illustrative of the dominant parliamentary attitude in the MOS debates. First, it shows how compensation of or care for immaterial war damage was not considered as a field of policy of the national government.This was not surprising: The collective and material focus of the MOS matched contemporary political and socio-economic context and traditions.The fact that individual (or particular groups of) victims were not in the spotlights of this first temporary parliament as a field of elaborate policy making, was strongly rooted in pre-war political traditions. In Dutch politics of the 1930s, social work and charity had been mainly the domain of private or confessional organisations and institutions. Government intervention in the individual social and psychological realm was not completely unknown, but definitely not a common practice in the late 1940s.<sup>6</sup> The most visible and pressing problems regarding war victims were of a collective and material nature.

<sup>2 &#</sup>x27;Handelingen Tweede Kamer 1949–1950', n.d., 242.

<sup>3 &#</sup>x27;HTK 1945–46', 388; 'HTK 1949–50', 253.

<sup>4 &#</sup>x27;HTK 1945–46', 388.

<sup>5 &#</sup>x27;HTK 1949–50', 305.

<sup>6</sup> Bossenbroek, *De Meelstreep*, 184.

The immediate post-war years were a period in which crisis management and financial shortages were prominent. Solutions and assistance were therefore also mainly discussed in parliament in such terms.<sup>7</sup> How did this impact the role of emotions in these immediate post-war discussions? Were emotions omnipresent? Overlooked? Suppressed? Or was emotional charge of this very recent past so obvious, that expressing emotion in these debates was considered redundant?

#### Collective + Material ≠ Unemotional

The emotion-mining results in Chapter 8 indicated that the MOS discussions contained more negative emotional word use than other topics discussed in parliament in the late 1940s. Emotional words were, however, not manifest because MPs expressed their own lived feelings regarding the wartime past. A closer analysis of the debates shows emotional language was mostly expressed in object-centred formulations.Wartime devastation, bombings, damaging of property, and other losses were described in great detail.<sup>8</sup> Words associated with anger, sadness, and fear were manifest in such lively descriptions. Here emotional language was,in contrast to the early WBP debates of Chapters 6 and 7, directly connected to the occupation past. Next to descriptions of tangible aspects of this recent past, parliamentarians often dressed their speeches with descriptions of immaterial damage, misery, and personal losses. These descriptions of suffering by the population were behind the expression of many emotional terms in the debates. Whilst MPs highlighted the misery of the recent past, they seldom talked about contemporary consequential suffering, or post-war plight.<sup>9</sup> This was a contrast compared to the WBP and WBPZ debates.

The identifiable, often individual examples of victims or their wartime suffering that were brought to the fore in the debates, show three key aspects of the parliamentary engagement with war victims. First, war victims were considered as people in need of something. This could be compensation of material damage, or care or aid on other levels (that, strikingly, often exceeded the formal scope of the MOS). Despite the collective and material focus of the legislative schemes discussed, my analysis of the debates showed that MPs had an eye for more than only directly visible and material aspects of war victimhood. Second, emotions were strategically expressed by mentioning poignant examples of past suffering of identifiable individual victims. Such examples were mentioned to evoke feelings of grief, and had to

<sup>7</sup> Bossenbroek, 185.

<sup>8</sup> Verhoeven and Duyvendak, 'Enter Emotions: Appealing to Anxiety and Anger in a Process of Municipal Amalgamation'.

<sup>9 &#</sup>x27;HTK 1949–50', 239–312.

appeal to feelings such as sympathy, empathy, and compassion.<sup>10</sup> These emotional accounts, charged with words associated with sadness, fear, and anger, were the emotional scaffolding MPs used to raise or construct their arguments. MPs advocated for a quick, just, fair, and generous compensation of material war damage. The emotional appeals they used to do so, were primarily directed at the government representatives and ministers present – as a majority of the MPs in the House of Representatives already shared a willing attitude towards generous compensation of the material damage.<sup>11</sup> Third, although it seems plausible, a collective and material approach to war damage in the MOS did not mean the MPs involved were debating this matter also in a neutral or unemotional way. On the contrary. The political tone of voice in engaging with the war's consequences, can hardly be considered as 'legal' or 'administrative'.<sup>12</sup> The parliamentary discussions on material war damage were not only elaborate and lengthy, but also full of emotionally dressed examples of suffering that MPs brought to the fore. This parliamentary attitude contrasts with commonplace ideas of the early 1950s as a (solely) forward-looking phase in postwar history.<sup>13</sup> It also corroborates the refutation of the 'silence' that was, according to commonplace views, central to the engagement with war victims in the first postwar decade.<sup>14</sup> What about subsequent discussions in later years?

### **9.2 A 'Silence of the 1960s'?**

Chapter 8 showed 'gap years' in the parliamentary discussion of legislative schemes related to the alleviating of the suffering of war victims (see Graph 14, Graph 15, and Supplement 3.2). The 'gap' in the observations stretched from the end of the 1950s and roughly covered the 1960s. This period forms a lacuna, compared to the otherwise relatively equal distribution of debates on the timeline 1945–1989.<sup>15</sup> Based on this empirical observation alone, it might be tempting to claim there was indeed a period in which 'silence' prevailed, also in the post-war parliamentary engagement with the fate of war victims. The consequences of the war, especially for those considered as or considering themselves as victim of that same war, seem completely out of sight, as they remained mostly undiscussed. This presumed silence would

<sup>10</sup> Leenders, 'Emoties in het asieldebat', 66; Sudhir, Roy, and Cherian, 'Do Sympathy Biases Induce Charitable Giving?', 2, 8–10.

<sup>11</sup> Bossenbroek, *De Meelstreep*, 398.

<sup>12</sup> This was suggested by historian Remieg Aerts, see also Aerts, 'Emotie in de politiek', 21.

<sup>13</sup> Blom, 'Jaren van tucht en ascese. Enige beschouwingen over de stemming in herrijzend Nederland (1945–1950)'.

<sup>14</sup> Withuis, *Erkenning*, 16.

<sup>15</sup> The four scheduled parliamentary debates dealing with the General Treaty in 1963 were an exception. These debates are more elaborately discussed in Chapter 7.

then follow the definition of the American historian Jay Winter, in which silence is considered '(…) a socially constructed space in which and about which subjects and words normally used in everyday life are not spoken.'<sup>16</sup> Although the exact timing is a little different, this observation then also corroborates with the second phase in Withuis' cycle – wherein 'silence' was broken only after the 1960s.<sup>17</sup> The observed lacuna in the measurements, however, can hardly be interpreted as a silence in postwar parliamentary history. Despite the apparent comprehensiveness that the computational analysis of complete and large-scale datasets such as the *Handelingen der Staten-Generaal* seems to offer, the quantitative results do not tell the complete story here. Nor do they provide any further explanation of the observed lacuna. Identifying, counting, and plotting debates only showed something was going on. But what was going on in the parliamentary engagement with and discussion of war victims in the 1960s?

This empirically observed silence has more to do with limitations or boundaries of the emotion-mining analysis and its results, than with a societal change that was reflected in the Dutch parliamentary discussion. A first explanation for the observed lacuna is provided by a political procedure behind the establishment of many of the early national arrangements for war victims, as they were established by Royal Decree. Therefore, following Dutch political procedures, they were simply not scheduled for discussion in (one of) both houses of the parliament.<sup>18</sup> Second, elaborate general social welfare schemes were established in the Netherlands in the 1960s. Since then, care for war victims was accounted for in more general social legislation. As discussed in Chapter 8, this was a consequence of the deliberate decision to incorporate national government care for victims of World War II in generic social benefit schemes. As a result, the parliamentary discussions and decision-making processes regarding war victim-related legislation were not separately scheduled on the parliamentary agenda.War victims were discussed, but not in particular.They were addressed in the context of more general debates on social welfare legislation. Also the practical implementation of the war victim arrangements, as they were part of the *Bijstandswet,* was, until 1969, a responsibility of the municipalities in the Netherlands.<sup>19</sup>

If there is a conclusion to be drawn regarding the observed 'silence' in the parliamentary debates, it is that this lacuna had much more to do with contemporary administrative and organisational political procedures and practices, than it being a

<sup>16</sup> Winter, 'Thinking about Silence', 4.

<sup>17</sup> Withuis, *Erkenning*, 118.

<sup>18 &#</sup>x27;Handelingen Eerste Kamer 1968–1969', n.d., 162.

<sup>19</sup> This end was marked by the installation of a Commissie van Advies inzake Bijstand aan Vervolgden (Committee for Advice on Support for the Persecuted). From then on, this organisation dealt with assessment of applications. See Piersma, *Bevochten recht*, 129.

consequence of developments in attention, attitudes, or emotions connected to the discussion of war victims in parliament.

#### A Lasting Demand

As the aforementioned debates on general social legislation dealt with a variety of topics not directly related to this case study, they were not taken into account in emotion-mining analysis. As outlined above, these debates are behind the lacuna in the emotion score graphs in Chapter 8. It does not mean however that these debates have to be completely set aside. A closer look at these debates shows how the government's role in alleviating the suffering of war victims was incidentally addressed in the more general debates on social welfare legislation.

The demand for disconnecting war victims from the *Bijstandswet* became particularly prominent in the 1960s. MP Took Heroma-Meilink (PvdA), for example, advocated not only for more elaborate arrangements. She suggested that arrangements for war victims should get an autonomous legal fundament, independent of general social legislation.<sup>20</sup> As responsible minister, Marga Klompé did not seem to be unwilling to recognise the particularity of both past and present circumstances of war victims. Klompé insisted, however, that their care and status were already guaranteed, in particular from 1965 onwards. Since then, war victims were included in the general social benefit and poor relief legislation (*Bijstandswet*)*.* Within this act, war victims even had a special status with the *Rijksgroepsregeling*. Minister Klompé – proud as she was of 'her' *Bijstandswet* – could 'not imagine' how people wanted more.<sup>21</sup> What Klompé seemed to overlook, was that the underlying reason for the demands was not primarily based on dissatisfaction with the aid or care that was offered. The demand for tailored national legislation can be understood as a call for official recognition (on the national level) as a victim of the German occupation in World War II. An autonomous war victim act would, according to its advocates, offer this official recognition, by uncoupling war victims from other people in need of (government) support.

The abovementioned developmentt displays, in the first place, how the conception of 'war victim' shifted from 'someone in urgent need' to 'someone (also) in need of recognition and acknowledgement'.The demands of members of opposition parties, such as Heroma-Meilink, turned out to be lasting. Before addressing the consequences of these lasting demands, I focus briefly on the role of emotions in these 1960s debates – although this analysis is, because of the lacuna outlined in Chapter 8, not backed by quantitative results. How does the parliamentary discussion of

<sup>20 &#</sup>x27;HTK 1960–61', 3243.

<sup>21 &#</sup>x27;HTK 1961–62', 3224–25; 'HTK 1962–63', 3070–71; Piersma, *Bevochten recht*, 143.

war victims relate to existing ideas about a 'renewed public attention' and 'late consequences' in the 1960s? What about emotions connected to war victimhood? Were they becoming absent, or did they gain prominence instead?

Based on close-reading analysis of the debates, the tone of voice in the debates seems similar to that of the earlier debates. Nevertheless, there are some noteworthy differences. In the first place, late-onset suffering of war victims was placed centre stage as a parliamentary issue. Increasing professional and public attention for 'late consequences' of the occupation years arose and became more prominent in parliament in the late 1960s. Not only suffering from the past, but also contemporary struggles with this past (trauma) were increasingly cited by those advocating for more elaborate legislation.<sup>22</sup> Second, there were indications that circulations of emotions related to war victimhood, as they were expressed in media and in the public sphere, increasingly also resonated with the parliamentary arena. Emotions connected to war victims and victimhood were, however, not yet considered as productive or constructive elements. Senator Martina Tjeenk-Willink (PvdA), for example, responded to a television film on Auschwitz. She said that 'false sentiments' and 'evoked emotions' had not benefited 'the case'.Who (or what) she referred to here remains unclear, but she thought that a more satisfying arrangement for concentration camp survivors would make such emotion-evoking television completely unnecessary.<sup>23</sup> Secretary of State Van de Poel recognised emotional manifestations in the public sphere, but considered them as disruptive elements that should be kept outside of the political realm:

'I think this is a complicated matter. It is an issue which we – despite everyone's emotional involvement – have to decide about with reason.'<sup>24</sup>

In contrast to the emotional accounts of individual suffering in earlier parliamentary debates on material war damage, Van de Poel apparently considered the explicit discussion of emotions (of others) in parliament as undermining the dignity of, in the case of this example, war victims.<sup>25</sup> How did this develop through the following years? What role did emotions play in these developments? Did this influence the establishment of national government legislation for war victims? And if so, how?

<sup>22 &#</sup>x27;Handelingen Eerste Kamer 1967–1968', n.d., 486.

<sup>23</sup> Tjeenk-Willink responded to the broadcasting of the 1968 documentary by Dick Verkijk on Auschwitz and the Auschwitz-committee: Dick Verkijk, *SPECIALE BERICHTGEVING: Auschwitz en het Auschwitzcomité*, Video, Documentary (VPRO, 1968); 'HEK 1967–68', 479–80.

<sup>24</sup> This idea probably originated in the emotional public responses to the Lages case in 1952 and 1966. That this remained difficult was also proven by the Breda Three case in 1972. "HEK 1967–68," 487.

<sup>25</sup> Tames, *Doorn in het vlees*, 227, 233–34.

#### **9.3 Tailored Legislation for War Victims**

More attention for 'late consequences' coincided with the increasing visibility of interest groups of victims of Nazi persecution in media and public sphere. Such groups became more organised and centralised, and knew how to find their way to national politicians. Just as MPs had close connections with resistance interest groups in the 1950s, some MPs maintained connections with these new victim organisations in the early 1970s.<sup>26</sup> The emotions and emotional appeals of these groups circulated in the written press, on radio and television, and were also shared through personal interactions between interest groups and parliamentarians.<sup>27</sup> The communication lines were short. Take, for example, Joop Voogd (PvdA). This parliamentarian explicitly positioned himself as a link between society, interest groups, and politics. In 1970, Voogd noted that 'late consequences' were more manifest than ever.<sup>28</sup> Voogd took the role of mediator between (representatives of) war victims and his colleagues in parliament.<sup>29</sup> Unsurprisingly, it was also Voogd who submitted motions on the matter. This led to recurring debates in the House of Representatives in the early 1970s, for the first time on the 26th of May 1970. According to his colleagues, Voogd's motions did not excel in terms clarity, demarcation, or unequivocalness.<sup>30</sup> Although they were accepted by the House after voting, the government was not receptive of his WBP-style proposal for a pension scheme for victims of persecution. According to Secretary of State Van de Poel, the plans of opposition-party member Voogd were not in line with government policies. He pointed at budgetary consequences and the potential of excessive government costs.<sup>31</sup> Voogd's demands were not met. Nevertheless, he had firmly established the issue on the parliamentary agenda.

#### Making the Myth

In the debates that followed, communist MP Joop Wolff expressed he 'deeply regretted' the fact that government care schemes for war victims still had to be discussed

<sup>26 &#</sup>x27;HTK 1970–71', 3470–71.

<sup>27</sup> Withuis and Mooij, 'From Totalitarianism to Trauma', 211; Withuis and Mooij, 'Conclusion: The Politics of War Trauma', 328; De Haan, *Na de ondergang*, 116–18.

<sup>28</sup> With this rather generic statement, Voogd presumably referred to the proto-professionalisation of the term '(war) trauma', increased attention in public media for contemporary suffering of and late consequences for war victims, and increased level of organisation and visibility of war victims' interest groups. See also 'HTK 1969–70', 101.

<sup>29</sup> Verhoeven and Metze, 'Angst, woede en wantrouwen', 64, 77.

<sup>30 &#</sup>x27;HTK 1969–70', 101–16; Piersma, *Bevochten recht*, 144.

<sup>31 &#</sup>x27;HTK 1970–71', 3486; Piersma, *Bevochten recht*, 147–48.

in the 1970s.Wolff considered it 'harrowing' that this had 'not been solved satisfactorily' a long time ago.<sup>32</sup> Wolff also used the term 'silence' in the discussions to imply a lack of attention and support for victims in preceding decades. Wolff 's 'silence' was definitely not referring to a 'liturgical silence' that encompasses the solemn and sacred silence common to practices of mourning or commemoration.<sup>33</sup> His 'silence' referred to the opposite, as it was used here as a normative judgement, pointing at supposed negligence of the past. Wolff placed tailored war victim legislation centre stage by emphasising perceived shortcomings of his predecessors from the past. Wolff referred to a 'new definition' of war victims that incorporated late-onset consequences, such as psychological problems and trauma (see also Chapter 8). These were all aspects that, for example, the Material War Damage Act of 1950 had obviously never been intended for. In this process, and with these wordings, Wolff contributed to the creation of a myth that war victims had been completely neglected or forgotten before. Even years later, this myth of 'late action' and previous 'silence' resonated with authors and historians. The term 'silence' was there, however, not used as a normative judgement, but rather as an analytical category to demarcate a specific phase in a cyclic historical development (see Chapter 1).<sup>34</sup>

As paradoxical as it may seem, after his silence or negligence argument, Wolff 's next complaint dealt with the existing arrangements for war victims. The integration of government care into the general *Bijstandswet* was, in his eyes, problematic and caused the 'deprived position of these people'.<sup>35</sup> His colleague Voogd brought up the long-lasting demand again, and called it 'painful' that victims of Nazi persecution in need of support had to turn to the *Bijstandswet*. <sup>36</sup> The term 'painful' is not only associated with sadness and anger according to the EmoLex, it was here also used to refer to the infliction of pain, as if victims were being hurt or victimised again by being abandoned or overlooked by the government. To end this infliction of pain, Voogd wanted national legislation to be more accessible, comprehensive, and generous. Wolff also thought this was the solution to 'elevate' this group out of 'desperate plight'.<sup>37</sup>

Most representatives of left-wing opposition parties supported Voogd and Wolff.<sup>38</sup> Right-wing and coalition-party colleagues were more careful, but did not disagree on a fundamental level. Norma Dettmeijer-Labberton (VVD), for example, feared arbitrary implementation of current prevailing arrangements by the

<sup>32 &#</sup>x27;HTK 1969–70', 105; 'HTK 1970–71', 3477, 3485, 3495.

<sup>33</sup> Winter, 'Thinking about Silence', 4–5.

<sup>34</sup> See Chapter 1 and De Haan, *Na de ondergang*; Bossenbroek, *De Meelstreep*; Withuis and Mooij, 'From Totalitarianism to Trauma'; Piersma, *Bevochten recht*.

<sup>35 &#</sup>x27;HTK 1969–70', 105; 'HTK 1970–71', 3477, 3485, 3495.

<sup>36 &#</sup>x27;HTK 1969–70', 102.

<sup>37 &#</sup>x27;HTK 1969–70', 102–3.

<sup>38 &#</sup>x27;HTK 1969–70', 104–16.

executive municipal institutions.<sup>39</sup> While underlining her strong emotional involvement with the issue, Hannie van Leeuwen (ARP), felt 'fear for discrimination'. She thought an extraordinary arrangement for Jewish victims in particular was inconsistent with the pursuit of equal treatment in general.<sup>40</sup> Others emphasised that not contemporary legislation, or being Jewish anno 1970, but the history of the Nazi regime had created their exceptional position. Creating an autonomous legislative scheme for this group was therefore not considered as discriminatory.<sup>41</sup> These discussions indicate two important things. First, the engagement with the fate of war victims was in essence not about the suffering people had undergone during the occupation in those days. What mattered were consequences of the German occupation that were still suffered on a daily basis especially, so-called late-onset consequences. Second, this post-war, World War II-related suffering was not considered something from the past. The central aim of motions and bills was therefore to alleviate the contemporary plight of those people whose suffering resulted from circumstances of more than 25 years ago.<sup>42</sup>

Eventually it was a motion submitted by MP Jacques Baruch that became the basis for the establishment of a specific act for victims of Nazi persecution. After Baruch, former PvdA member, now affiliated to brand-new coalition party DS'70, dropped the expensive demand for a pension scheme, the WUV came into being. This act was aimed at alleviating the suffering of a relatively narrowly defined group of victims. More importantly, however, the welfare scheme was autonomous, tailored, and disconnected from general social legislation.<sup>43</sup> The question remains, however, how the role of emotions in discussing this legislation possible affected the course of debates, or the development of the legislative schemes themselves.

#### **9.4 The Emancipation of Emotions**

The results of emotion-mining analysis of the 'war victims' subset, discussed in Chapter 8, gave no indication for a strong quantitative increase in emotionality in the 1970s. Nevertheless, a semantic shift in the words underlying these emotion scores was observed. The following paragraphs elaborate on the several qualitative developments in the discussions on tailored welfare schemes for war victims that preceded (and dealt with) the WUV. The 'close reading' of these debates shows that

<sup>39 &#</sup>x27;HTK 1969–70', 106.

<sup>40 &#</sup>x27;HTK 1969–70', 107.

<sup>41</sup> CPN parliamentarian Joop Wolff repeatedly brought up this argument. See 'HTK 1969–70', 105; 'HTK 1970–71', 3476.

<sup>42</sup> Amongst the supporters of this viewpoint was also (responsible) Secretary of State Hein van de Poel. 'HTK 1969–70', 105; 'HTK 1970–71', 3477, 3485, 3495.

<sup>43 &#</sup>x27;HTK 1971–72', 1061–64.

relative stability in emotion scores does not necessarily equals a lack of historical change of development. These paragraphs offer a more detailed breakdown of the historical phenomenon that was described as an 'emancipation of emotions'. I argue that, when taking a closer look at the actual discussions, an increasing focus on contemporary misery is observed, which coincided with a shift in the role of emotions in parliamentary debates in the 1970s.

#### Circulations of (Contemporary) Emotions

The strategically used emotional descriptions of past suffering, which had been omnipresent in the 1950s, fell into disuse. In the early 1970s, it became less common to go into too much detail about suffering that had taken place during the occupation. Emotions in parliament were increasingly related to contemporary late-onset consequences still experienced by war victims. These encompassed not only direct consequences of the German occupation, but also misery related to or resulting from current legislation and practices. This development is interrelated to an increasing interplay between the emotions expressed in parliament, circulations of emotions in society, and the utterances of emotions in the media. This was new to the parliamentary engagement with war victims, but was observed before in, for example, the debates and mass protests related to the Lages cases in 1952 and 1966 (see also Chapter 5). This does not mean that parliamentarians became more emotional. I argue that the emotions that, for example, opposition-party MP Voogd brought up in the debates discussing his many proposals, were not his personal, lived emotions per se. Instead, these emotions had been publicly circulating in media and society or had been expressed by others to Voogd in person.<sup>44</sup> Voogd utilised these emotions to distance himself and his party from the government's line of policy. The expressions of these emotions in parliament can be understood as the result of redirecting publicly circulating emotions into the parliamentary context. MPs acted according to their role as representatives of the people, as they were the ones relaying or re-directing emotions from society into the parliamentary arena. The manifestations of emotions in parliament in this period are an example of a phenomenon that Aerts described as 'sublimed societal emotions'. Circulations of 'societal emotions' were 'sublimed' as they were redirected into the parliamentary sphere.<sup>45</sup>

Increasing parliamentary prominence of circulating 'societal emotions' related to contemporary suffering can be considered as a part of the process that Withuis'

<sup>44</sup> Examples of carriers of these circulating emotions are newspaper articles, television documentaries, and public protests.

<sup>45</sup> Aerts, 'Emotie in de politiek', 13.

described as 'the emancipation of emotions'.<sup>46</sup> In the slipstream of the establishment of the welfare state, explicitly naming emotions of others became more common. An important characteristic of how this phenomenon unfolded in the practice of parliamentary debate is that the people bringing up these emotions were not the people who experienced these emotions first-hand. As the private realm of citizens moved closer into the domain of political interference, the emotions of others emancipated in the parliamentary arena. The next section shows how this development resonated in a shift towards the appraisal of emotions by the parliamentarians involved.<sup>47</sup>

#### Emotional Appraisals

Instead of giving descriptions of the experience or the nature of (past) suffering of war victims to dress up their viewpoint, MPs now frequently made an appraisal of emotions (of other people). These appraisals originated in the circulations of 'societal' emotions.<sup>48</sup> Speakers in the House of Representatives, predominantly Voogd and Wolff, often even named very specific emotions. Voogd, for example, seemed to know what kind of contemporary emotions of others were connected to the matter of war victim legislation, as he summed up that bitterness, disappointment, incomprehension, and frustrations were still experienced every day by Jewish victims of Nazi persecution in particular.<sup>49</sup> Wolff emphasised that he had also observed the presence of similar emotionsin society.<sup>50</sup> This process of emotional appraisal underlies the observed shift in the top list of lexicon words used in the discussions of the 1970s. Instead of describing victims' misery, now MPs named their sadness, bitterness, fearfulness, or grief in debates.This is, as discussed in Chapter 8, backed up by the most frequently used words related to fear and sadness in particular.The nature of the highest scoring words shifted from 'words associated with certain emotions' to 'words explicitly referring to a certain emotion'. Examples of words directly referring to emotion are 'fear', 'regret', 'bitter', 'bitterness', or 'sorrow'. Other words refer


<sup>46</sup> Withuis, *Erkenning*, 118–20; Withuis, 'De veranderende omgang met de oorlog', 28; Withuis, 'Opkomst en neergang van PTSS', 158; Withuis and Mooij, 'Conclusion: The Politics of War Trauma', 328; Oosterhuis, 'Mental Health, Citizenship, and the Memory of World War II', 20, 31.

<sup>47</sup> Although this tendency of frequent 'emotional appraisal' was relatively new to the parliamentary discussion of war victims in the 1970s, I do not want to state that this had never happened before. An illustrative example is mentioned by Dutch historian Marij Leenders in parliamentary debates on Jewish refugees in the Netherlands in the late 1930s. See Leenders, 'Emoties in het asieldebat', 60.

<sup>48</sup> Aerts, 'Emotie in de politiek', 13.

to the typical expressions of emotions, such as 'crying', which is obviously associated with sadness.<sup>51</sup>

The abovementioned shift to emotional appraisal can be seen as a part of 'the emancipation of emotions'. I do not consider the shift as something that evolved coincidentally. The (predominantly) opposition-party MPs were well aware about what they were doing and saying in these debates. It is very likely that the opposition believed that advocating for a publicly articulate and organised group of war victims had electoral potential. It also had a potential delegitimising effect on the established government parties – as these MPs pictured the government in opposition to these 'poor victims'. As members of parliament can be considered as professional language users, it is likely that they deliberately appraised emotions in support of their argument for more generous and elaborate war victim legislation. The role of emotions in these debates can thus be considered that of a supportive structure ('emotional scaffolding') that had to evoke empathy, importance, urgency, and salience within the government representatives present. In its attempts to construct more elaborate and autonomous war victim legislation, the opposition had to mobilise coalition-party colleagues and the government. To achieve this, MPs not only made their political demands urgent, contemporary, and topical, but, above all, also emotional.<sup>52</sup> Yet there was another development in the 1970s that can be considered part of the emancipation of emotions.

#### Emotions Doing Politics

If there is one thing that characterises the role of emotions in the parliamentary engagement with war victims, it is that emotions were never absent. Although emotions related to suffering of victims had been recognised for a long time, they had initially been deliberately kept out of 'doing' politics. In the 1970s, however, emotions became explicitly considered as an actual part of national policy. What changed was that the emotions related to suffering were not only brought up, or used as 'scaffolding', but were now also explicitly mentioned as reason for or motivation behind policy (changes). To put it differently: Circulations of emotions in the public realm were not only recognised, but now also acknowledged. Following increasing mediaattention for the plight of victims of Nazi occupation, emotions became more than only part of political argumentation. Not so much actual war victims, or their suffering, but the related emotions became the political argument as they were deliberately taken into account in decision-making processes. This development links to

<sup>51 &#</sup>x27;HTK 1969–70', 107, 3452; 'HTK 1970–71', 3469, 3471–72.

<sup>52</sup> Idea based on quote from psychologist and Nobel Prize Winner Daniel Kahneman, cited in George Marshall, *Don't Even Think About It: Why Our Brains Are Wired to Ignore Climate Change* (Camden: Bloomsbury Publishing USA, 2014), 57.

what the German historian Ute Frevert describes as the 'politicisation of emotions'. Frevert stated that this is the process in which emotions are no longer (only) serving politics, but also 'doing politics'. Although Frevert discussed a more general development, and places this phenomenon in much later times, the developments in the Dutch parliament in the 1970s are an obvious example of what she describes as 'emotions doing politics'.<sup>53</sup>

I argue this development can be seen as another shift that shaped what Withuis described as 'the emancipation of emotions' in this period. Yet this was not a broad development that unconsciously unfolded in the background. Giving emotions an explicit role in the political process was part of a deliberate political strategy. The 'politicisation of emotions' is not so much contrasting 'emotional scaffolding', but can be considered as resulting from it. In other words, constructing the 'emotional scaffolding' seemed to take effect: Emotions had apparently become so self-evident that even coalition-party MPs and government representatives started considering emotions explicitly as a decisive factor in shaping and establishing war victim legislation. MP Dettmeijer-Labberton (VVD), for example, said that emotions of Jewish victims of persecution not only had to be addressed; they now also had to become 'a decisive element' in the development of future arrangements.<sup>54</sup> Even individual politicians became subject to(or contributed to) this development. Secretary of State Van de Poel, for example, who had clearly stated that he considered emotions as 'disruptive elements' before, now mentioned that he 'accepted emotional arguments as factual ones' in this particular situation.<sup>55</sup>

The 'politicisation of emotions' in the war victim debates had serious implications. It made that emotions, but also the lasting demands for tailor-made legislation could not be set aside anymore. Doing nothing (extra) for (at that moment) mainly victims of Nazi persecution became a very deliberate and charged political choice.<sup>56</sup> With upcoming elections in the spring of 1971 and (early) elections again in November 1972, electoral considerations played an important role. Doing nothing became a choice no one, wherever on the political spectrum, opposition, coalition, or government, wanted to make.<sup>57</sup> As a result, the establishment of a tailor-made war victim scheme, the WUV, became inevitable.

<sup>53</sup> Frevert, 'Emotional Politics'; Ministerie van Algemene Zaken, 'Lezing Ute Frevert over Emotional Politics − Toespraak − WRR', 25 January 2019, 6, https://www.wrr.nl/publicaties/publicaties/2019/01/25/lezing-ute-frevert-over-emotional-politics.

<sup>54 &#</sup>x27;HTK 1970–71', 3473–74, 3490–91.

<sup>55 &#</sup>x27;HTK 1969–70', 3452.

<sup>56</sup> With this development, victims of Nazi persecution gained a political status similar to that of the former members of the resistance. Contrastingly, however, the 'resistance disposition' had been there for almost 25 years. See also Chapter 7.

<sup>57 &#</sup>x27;Handelingen Tweede Kamer 1972', n.d., 716.

#### Recognition and Acknowledgement

Although physical illness, and later, trauma, were prerequisites for eligibility, the WUV act was about more than providing actual support alone. The act served as recognition of victimhood and a national acknowledgement of victims' status. This is best illustrated by the fact that the WUV offered applicants the opportunity to apply for official acknowledgement of victimhood, without applying for accompanying social benefit payments. Without being ill, or being in need, applicants had the possibility to have their status as a victim of persecution officially confirmed on a national level.<sup>58</sup> Representatives of interest groups and MPs involved considered this crucial to the act. This, obviously, also meant that people with a 'questionable war past' were excluded. Not only being ill, but also being 'on the right side of history' was a criterion , just as in the WBP and WBPZ discussed in Chapters 6 and 7. This led, also in a WUV context, to debate about the demarcation of eligibility. According to the initiator of the WUV, Baruch, the new act provided not only Jewish victims of persecution, but also Dutch people who had been interned during the Japanese occupation of the former Dutch East Indies with 'the recognition as a victim that had lacked before'. To Senator Hein van Wijk (Pacifist Socialist Party, PSP), the inclusion of these groups of victims was undesirable. Dutch colonisers in Japanese internment camps did not fit his definition of '(…) victims of persecution because of race, religion, or world view', he said. Their internment had been, according to him, a consequence of the fact that the Dutch had been colonial rulers back then.<sup>59</sup> Wolff agreed with Van Wijk. He added he considered it unacceptable if the bill, in his words, 'equated Dutch conscripts of the colonial wars in Indonesia (1945–1949) with Jewish survivors of the Holocaust'. This was, however, not the case, the minister ensured him, as 'victims' of the conflicts that rose after the end of the Japanese occupation were not even eligible for the WUV.<sup>60</sup>

As self-appointed advocate of victims of Nazi persecution, Voogd was eventually relatively satisfied with the WUV. At the end of the November 1972 debate in the House of Representatives, however, he brought up that he had '(…) the strong feeling that there were other deprived groups of war victims in contemporary Dutch society.'<sup>61</sup> His hunch proved to be justified.

<sup>58 &#</sup>x27;HTK 1972', 718.

<sup>59 &#</sup>x27;Handelingen Eerste Kamer 1972', n.d., 257.

<sup>60 &#</sup>x27;HTK 1972', 711, 716.

<sup>61 &#</sup>x27;HTK 1972', 718.

#### **9.5 Debating the WUBO**

More than ten years after establishing the WUV, another victim group incorporated in the *Bijstandswet* had to be 'elevated' out of their 'deprived' position, in the words of Voogd. His colleague and fellow-social democrat Joop Worrell took the lead this time.<sup>62</sup> Public attention followed after a phase of organisation and centralisation of a particular group of war victims again. In the 1980s, this process concerned the so-called civilian-war victims. Because of clear recommendations of the government advisory board known as WAC<sup>63</sup>, Secretary of State Jeltien Kraaijeveld-Wouters (CDA) was more or less forced to come up with a legislative proposal aimed at alleviating the suffering of the specific group of civilian-war victims.<sup>64</sup> The resulting proposal for a Benefit Act for Civilian-War Victims (WUBO) was discussed in the House of Representatives in the autumn of 1983.<sup>65</sup> What issues were at stake according to the MPs involved? How did the role of emotions in debating war victims developed in the 1980s? How does this relate to both earlier debates and the contemporary political context?

#### Implementation and Implication

Compared to the WUV debates of the 1970s, MPs' main concerns were of a much broader, societal nature ten years later. This is best illustrated by the contributions of Joop Worrel and communist Marius Ernsting, who assigned an important role to the WUBO in preserving memories of the German occupation of the Netherlands.<sup>66</sup> These memories were, according to them, strongly connected to their observations of 'signs of advancing fascism in current society' of the 1980s. The commemorative

<sup>62 &#</sup>x27;Handelingen Tweede Kamer 1983–1984', n.d., 49.

<sup>63</sup> *Werk- en Adviescollege Immateriele Hulp aan Oorlogsgetroffenen* (WAC) was an advisory committee, established in 1975. Its primary aim was to advise the government on the problems related to immaterial 'late consequences' of World War II. It led eventually to the establishment of ICODO, *Informatie- en Coordinatororgaan Dienstverlening Oorlogsgetroffenen,* a permanent organisation providing information on the issue to government and implementing organisations.

<sup>64</sup> Piersma, *Bevochten recht*, 165–66, 190–93; 'HTK 1983–84', 48–49.

<sup>65</sup> For the parliamentary debates, see "HTK 1983–84," 48–61, 69–84, 591–601, 849; A detailed description of the WUBO act can be found via: Sociale Verzekeringsbank, "The Benefit Act for Civilian War Victims 1940–1945 (Wubo) − Former Members of the Resistance and Victims of War − SVB," subject (Sociale Verzekeringsbank, March 19, 2019), https://www.svb.nl/en/members-of-the-resistance-and-victims-of-war/schemes-andlegislation/wubo.

<sup>66</sup> This was concluded by Secretary of State Joop van der Reijden (CDA) after he mentioned the emotional involvement of MPs in the WUBO debate of the 21st of September 1983. 'HTK 1983–84', 69.

function and the fear for reviving fascism were used as arguments in advocating for a more elaborate WUBO than was proposed by the government.<sup>67</sup> This connection between war victims and major contemporary social themes is reminiscent of the 1950s Cold War-style moral posturing discussed earlier in Chapter 7. In the WUBO debates of the 1980s, however, social democrats and communists were on the same side, even when it came to the perceived threat to society that they observed. The 1983 debates score relatively high on use of words associated with emotions. In contrast to the fierce mud-slinging of social democrats and communists in the 1950s, however, it was not only moral posturing that fuelled emotional charge now.

PvdA and CPN parliamentarians stirred debate with their concerns about the implementation and implications of the WUBO. Their concerns related to discrimination, privacy of beneficiaries, and fear for unequal treatment of men and women.<sup>68</sup> Just as in the early 1970s, wartime suffering itself was not elaborately addressed. All these issues and concerns were related to the contemporary or future implementation of the WUBO. What changed in the 1980s, however, was that dressing the debates with emotions related to contemporary misery moved to the background. Appraisals of emotions related to the plight of war victims became significantly less prominent than before. Discussions in the 1980s also lacked the explicit consideration of circulations of emotions. More than before, debates dealt with concrete and formal policy aspects of the WUBO, such as the loss of earning capacity.<sup>69</sup> Especially elements integral to the WBP or WUV, that the WUBO propositions lacked, stirred debate.<sup>70</sup> For the government, the potential costs of the WUBO were the biggest potential problem. This fear was based on the experiences with the WUV that had turned out to be an excessive bite out of the government budget.<sup>71</sup> Several opposition MPs fiercely fought the fear expressed by government representatives about 'uncontrollable expansion' of the group of potential beneficiaries and resulting government costs.<sup>72</sup> Other topics frequenting the WUBO debates, also mostly unrelated to the contemporary circumstances of the potential beneficiaries themselves, were the practical implementation of the WUBO and the position of *Stichting 1940–1945* and the recently established ICODO. The fact that

<sup>67</sup> This was brought up by JoopWorrell (PvdA), and stressed again by newcomer Marius Ernsting (CPN) in his 'maiden speech' on September 21, 1983. See 'HTK 1983–84', 48, 56–57; Piersma, *Bevochten recht*, 195.

<sup>68 &#</sup>x27;HTK 1983–84', 49, 599.

<sup>69 &#</sup>x27;HTK 1983–84', 591, 594.

<sup>70</sup> These were the reversed burden of proof, the anti-hardship clause, and the question what kind of illness or damage was considered enough ground for eligibility. See 'HTK 1983–84', 50–51, 53, 56, 70, 75–76, 80–81, 83.

<sup>71</sup> Van der Heijden, *Dat nooit meer*, 466–67.

<sup>72 &#</sup>x27;HTK 1983–84', 75–76.

these topics were so prominent in the debates, gives the debates a rather professional and technocratic impression.<sup>73</sup> Did this also mean that emotions had left the parliamentary building?

#### Abstract Emotions

Historian Carla Hoetink described how expressions of emotions are generally not considered as part of a professional tone of voice in political debate.<sup>74</sup> Therefore, the more technocratic and professional approach to war victim legislation in the 1980s seems intuitively also associated with an unemotional style of discussion. The emotion-mining results in Chapter 8 showed that although many late-1980s war debates score notmuch higher than average, some peaksin negative emotion scores were observed here. An additional observation is that there was a much more uneven distribution of the scores of frequent individual words associated with negative emotions fear, anger, and sadness in the WUBO debates in the 1980s, compared to the debates in the preceding decades (see Supplement 3.2.6). Closer scrutiny of the discussions led to the observation that it is unfair to conclude there was no role for emotions in these debates anymore.

Instead, the manifestations of emotions in the MPs' language use in the 1980s referred to a much broader and less-specific definition of emotion.The role of emotions developed into one that was more abstract, as the engagement with (perceived) emotions (of others) was less direct. Not only were emotions dealt with in a more abstract manner, they were also approached from a greater distance. Worrell, for example, recalled in 1983 the 'drama of the WUV' without any further descriptions of details of individual examples and without any appraisal of related emotions.<sup>75</sup> Another characteristic of formulations in these debates were the often rather abstract references to 'the emotions' or 'the emotionality' related to the issue at stake.<sup>76</sup> Contemporary MPs did not excel in precision when it came to defining this emotionality. Based on a close scrutiny of the discussions, this emotional charge seems to be mainly a mix of negative emotions: A sense of sadness about what had happened between 1940 and 1945, a still-present fear for repetition or reviving fascism, and anger about perceived unjust treatment of victims and survivors in the contemporary present.

On the one hand, this confirms the claim of an increasing professionalisation and technocratisation of the Dutch political context.<sup>77</sup> This was also the case for an

<sup>73 &#</sup>x27;HTK 1983–84', 54, 71–72, 596–98, 600–601.

<sup>74</sup> Hoetink, *Macht der gewoonte*, 425.

<sup>75 &#</sup>x27;HTK 1983–84', 595, 597.

<sup>76 &#</sup>x27;HTK 1983–84', 69.

<sup>77</sup> Hoetink, *Macht der gewoonte*, 425.

inherently emotional topic as the victims of World War II and the German occupation of the Netherlands. This does not mean, on the other hand, that emotions were entirely absent. Instead, emotions seem to have gained an almost self-evident role in discussing anything war-related. Fundamental consensus that care for war victims was an important task of the national government had only become stronger over the decades. Anything related to the war, especially war victims, spoke for itself in terms of emotional charge. In the 1980s not only parliamentarians, but also government officials and ministers were so imbued with the emotional charge of the topic, that elaborate emotional dressing of debates had become redundant. Apparently, the construction of war victim legislation had become so stable and so strong over the years, that the formerly indispensable emotional scaffolding could be safely torn down.

#### **9.6 Conclusion**

This chapter addressed the role of emotions in the post-war parliamentary engagement with war victims. Specifically the parliamentary debates about welfare provisions for victims of the German occupation in the Netherlands were analysed. Attention was paid to how the role of emotions related to a persistent periodisation in historiography by addressing the following questions: What role did emotions (and their manifestations) play in the discussion and establishment of war victim legislation and national government's engagement with care for victims of World War II? How did this change over time? How is the role of emotions in these debates related to classifications of 'silence', 'neglecting' or 'overlooking' regarding the 1950s? Did contemporary MPs repress their emotions? Was this followed by an 'emancipation of emotions' in parliament? If so, how did this evolve? Whose emotions were emancipating? What role did they play? And how?

In contrast to the crisismanagement of theimmediate post-war period, the general way of engaging with victims of World War II in the 1950s has often been characterised by using words such as 'silence', 'neglecting', and 'overlooking'.These characterisations seem to originate in the lack of the establishment of an elaborate compensation payment or mental health care scheme directed at individual victims in the first two post-war decades. I argue this idea of 'silence' was retrospectively attributed to this period in the 1970s. It was used to point at shortcomings of the past, as a normative judgement in placing the issue centre stage, in making the issue topical. Even years later, this myth of 'past silence' that was followed by 'late action' came back as an analytical category in academic historiography. That the resulting legislation of the 1950s was considered as rather limited in hindsight, did not mean there was a silence, that emotions were concealed,or that contemporary parliamentarians had no eye for war-related suffering in the debates**.**

The government interventions of the 1950s focussed on material war damage from a collective perspective on the recovery of the national community. In this approach, characteristic for pre-war and contemporary Dutch political context, individual or group-specific aspects of victimhood were not considered as a field of policy for the national government. Exceptionality of suffering of certain groups – such as Jewish victims of persecution – was not specifically dealt with. The government policies regarding war victims in the 1950s were shaped by contemporary ideas on the scope of government intervention, demarcation, and feasibility. In discussing this legislation, parliamentarians can hardly be described as being 'stiff upper lip'. This chapter showed how related debates in parliament were anything but free from emotional manifestations. Emotional descriptions of actual misery of identifiable individual victims were a constant in debates on, for example, the Material War Damage Act. Suffering of victims was made into a cause. Such rhetorical practices fuelled the manifestation of emotional language in the discussions. This is supported by relatively high emotion scores for these debates. Emotions were used here to emphasise the gravity of what had happened in the recent past, and to support the salience of the issue. Emotional descriptions as such offered the scaffolding for the argumentation of what MPs considered as 'just' and 'fair'in policy making. The emotional descriptions of suffering from the occupation past can be considered as indicators of an empathetic and committed parliamentary attitude. Victims' misery and related emotions were not absent, silenced, or overlooked. Emotional descriptions of immaterial damage and losses also show how, although it was not considered part of the field of national policy, these aspects of victimhood served advocating for alleviating tangible and directly visible material war damage.

The trendlines and output of emotion mining in Chapter 8 displayed a lacuna in the 1960s regarding debates engaging with war victims. As these 'gap years' were mainly a result of contemporary administrative organisation, this lacuna cannot be interpreted as a consequence of a historical development regarding the (lack of) parliamentary attention, attitudes, or emotions linked to victims of war. What it does show, however, are the boundaries of this type of research material and the computational methods used. It is, in the first place, important to note that economic and financial circumstances in the late 1960s were of transformative influence in the national engagement with war victims. A general trend of expanding the welfare state, and a growing acceptance of far-reaching government involvement in people's personal wellbeing in the second half of the 1960s created the preconditions necessary to expand government care for victims of World War II.These factors can be considered the essential building blocks in constructing more elaborate and autonomous national war victim benefit schemes from the early 1970s onwards. This alone was, however, not enough.

Although the parliamentary discussion of war victim-related legislation was rather limited, I have identified some changes by close reading the more general debates,in which war victims were mentioned or briefly discussed. First,increasing organisation and centralisation of specific groups of war victims in interest groups, placed their demands on the political agenda. Multiple MPs played a key role in this process, as (self-appointed) spokespersons of victims' interest groups. These parliamentarians frequently used a perceived oblivion of victims (in the past) as argument in advocating for more elaborate war victim assistance schemes in the 1970s. By doing so, they (deliberately) contributed to the creation of what I have called a persistent myth. Via them, not only the victims' demands and agendas reached the political realm, also their circulating emotions were redirected into the parliamentary arena. Not only descriptions of wartime-past suffering bore words associated with negative emotions. Parallel to increasing attention in the public sphere for 'late consequences', mentioning the contemporary plight of war victims was underlying the manifestation of emotions in parliamentary discussions. In retrospect, this period can be considered as an intermediate period. It was a prelude to developments observed in the elaborate and frequent discussions of tailor-made legislation for several rather-specific groups of victims of the German occupation in upcoming years.

The quantitative analysis of emotional language in parliamentary debates shows in the first place that 'the emancipation of emotions' did not coincide with an increasing 'emotionalisation' in the parliamentary context (see also Chapters 5, 6, and 8). The results of quantitative analysis of emotional language use in parliamentary debates, however, did indicate proportionally more emotional word use in the war victim debates, compared to the average parliamentary debate. Within the subset of war victim debates over the period 1945 until 1989, my analysis shows a relatively stable course of relatively high emotion scores.This does not mean that nothing had changed. The results of emotion mining gave a first indication of transitions. The most frequent lexicon words behind the scores indicated a shift from 'empathetic' to 'appraising'. This shift became visible in developments in language use, which changed from frequently using 'words associated with emotions' to using 'words explicitly referring to a certain emotion'more often.Emotionmining output had anindicative function here, but did not draw a comprehensive picture of developments, changes, or how 'the emancipation of emotions' unfolded in the debates. Close reading remained essential to understand, interpret, and give meaning to the details of complex and multifaceted developments and changes.

For victims' interest groups, especially opposition-party MPs paved the way to their demand of more autonomous war victim legislation. Some individual MPs stood out in particular in advocating for more elaborate war victim legislation.These particular politicians were well aware of the fact that advocating for a visible and organised group of war victims might have been electorally favourable. It had, at least, a potential delegitimising effect on the incumbent government. To break with the active-passive dichotomy that had been persistent for decades (see Chapter 7), and

to mobilise their coalition colleagues and the government, MPs had to make their (constituencies') demands an emotional issue. A first development that can be linked to what Withuis has described as 'the emancipation of emotions' was the increasing practice of appraising emotions by members of the Dutch parliament. As outlined above, this observation encompassed mostly (the emancipation of) the emotions of others in the 1970s parliamentary debates. These emotions were mostly contemporary emotions of war victims, which were circulating in a public realm. The emotions related not only to suffering or consequences of the war itself, but also to the consequences of (perceived) deficient national legislation in the present. Instead of describing victims'miseryin an evocativemanner, by outlining their 'losses'and 'disasters' in emotional diatribes, in the 1970s MPs' appraisals of related sadness, bitterness, fearfulness, or grief amongst others dominated. In other words, in the 1950s parliamentarians spoke emotionally, whereas in the 1970s they spoke about emotions.

Given the fact that emotions were explicitly considered as unproductive in the 1960s, this development was all the more remarkable. I do not consider this development as a result of what has been described as a change in 'mental climate' or a process of 'emotionalisation'. Changes in the role of emotions are neither an explanatory factor, nor a building block for legislative development. Instead, emotions were strategically used to strengthen what I have called 'emotional scaffolding'. Emancipating emotions in parliament had a supportive role and was part of deliberate political strategies. MPs, as professional language users, created this 'emotional scaffolding' through rhetorical practices in order to support autonomous and elaborate war victim legislation in their confrontation with the actual 'builders' of this construction: The responsible ministers and secretaries of state. Close reading of the war victim debates showed a further development in the role of emotions, as emotions had become a political argument in itself. Ute Frevert described this process as the 'politicisation of emotions'. Coalition-party MPs and government representatives increasingly considered emotions a decisive factor in shaping and establishing war victim legislation.This development also shows how emotion had become a factor of importance in discussing national policy, without necessarily being the opposite of rational reasoning. Professional and technocratic policy makers took emotions they observed in society now explicitly into account. In other words, they made that emotions started 'doing politics'.

The (changing) role of emotions in parliamentary debates seemed to have contributed to the elevation of a strong supporting scaffolding, to stick to the construction-site metaphor. In the second half of the 1970s, this made that responsible politicians stood at a crossroads. Not coming up a with more elaborate, tailor-made, and autonomous legislative scheme had become inconceivable or, at least,it had become a path that no contemporary politician wanted to take. Now they had talked the talk, they had to walk the walk. Although not all initial demands were met in 1972, a tailor-made aid scheme for victims of persecution was established that is known as the WUV.This act was not only the starting point for more parliamentary discussion on its implementation and implications, but also paved the way for national legislation demanded by other groups of war victims.

Although economic precedents had not been favourable in the early 1980s, the WUBO act for civilian-war victims was established in 1983. New and additional autonomous legislative schemes for war victims were being developed and the role of emotions in the related debates also moved along. The all too obvious direct and strategic mentioning of emotions had become superfluous.The connection between emotions and the survivors of the German occupation now seemed self-evident. The debates in the 1980s were not unemotional, but emotions were manifest in a more abstract way and approached with more distance.The circulation of emotions had become so well known, that now MPs only mentioned war-related 'emotions', or 'emotionality' in general terms. These terms had become well-developed forms of response. They were apparently considered sufficient in order to evoke or take into account emotions in parliamentary discussions. A more abstract approach and an attitude of increased distance from emotionality in debates can be considered as part of developments of professionalisation and technocratisation. Where emotional scaffolding had been considered indispensable in the past, the construction of war victim legislation had become so stable that the scaffolding could be safely removed in the 1980s.

## **10. Conclusion | On the Role of Emotions and Computers**

'A re-evaluation of a rather well-investigated historical case' – that is how this study was introduced in Chapter 1. A central aim was to gain a better understanding of the role of emotions in the way the consequences of World War II were dealt with in the Dutch parliament post-war. The central questions were not only *whether,* but also *how* emotions played a role in the manner the Dutch government engaged with the consequences of the war in the Netherlands between 1945 and 1989. Assessing not only *whether* emotions were manifest and in *what* wordings, this study also examined *how* they were expressed, discussed, and played a role in the parliamentary discussions of the national government's interventions aimed at alleviating the suffering of the former members of the anti-Nazi resistance (Chapters 6 and 7) and various groups of war victims (Chapters 8 and 9). In the computer-assisted analytical process, computational techniques for emotion mining were applied to digitised historical sources. The numerical output was statistically evaluated and visualised, linguistic output was presented in top words tables and, inspired by these quantitative results, a selection of the sources were subjected to close reading. This was not only done to gain insights into historical questions, but also to serve an important second aim: to combine, compare, and evaluate the outcomes of both traditional and computational approaches to historical research.

This conclusion reflects, in the first place, on the central questions regarding the more substantive historical aspects of this investigation by discussing the results of combining close and distant reading, and qualitative and quantitative perspectives on the study of newspaper articles, speeches, and, predominantly, parliamentary debates. The chapter reflects first on the more general historical developments in emotionality in Dutch society (10.1). Proceeding, sections 10.2 and 10.3 deal with the two substantive case studies on different groups of war survivors. The chapter concludes with section 10.4, which offers a reflection on the value of computational techniques for emotion mining as part of a computer-assisted methodology for historical research.

#### **10.1 Emotionalisation and Emancipation**

The received wisdom of 'emotionalisation' was addressed in Chapters 1 and 5. Investigating this commonplace view formed the background of this study, as it underpinned later explanations and interpretations in the historiography of post-war developments in the national treatment of the consequences of the German occupation of the Netherlands during World War II.The emergence of an 'emotion culture' as a historical phenomenon was assessed by analysing whether an emotionalisation curve could be discerned in the language use in multiple historical Dutch text collections.The resulting emotion-mining output also set the baseline for the quantitative analysis and comparison of results later in this investigation. Although the (complementary) text collections used do not represent an entire society, they are of diverse origin, covering roughly the second half of the twentieth century. If the received wisdom of emotionalisation were to be true, it would almost certainly be observable in these texts.

The computer-assisted analysis, based on results of computational techniques for emotion mining, revealed a lack of substantive quantitative changes in the manifestation of emotional language use between 1945 and 1989.Whilst I assumed that an 'emotionalisation of society' would be reflected in an increase in emotional language use, emotion mining did not provide supporting quantitative empirical evidence for a strong and ongoing emotionalisation process. Instead, the results of emotionmining analysis showed emotionality was relatively constant in parliamentary debates, newspapers, and queen's speeches between 1945 and 1989. If there was any peak observable in the use of emotional words in the historical texts, it occurred in the late 1940s, rather than from the early 1970s onwards (as suggested in the literature; see Chapter 1). These findings contradict, not only the received wisdom of 'emotionalisation', but also the commonplace view of a 'silencing or withholding of emotions' that occurred in the 1950s.

The application of emotion mining on various text corpora in Chapter 5 also served as a first step in the evaluation of the validation of computational techniques and methods, including their validity in the context of historical research. First, the results provided quantitative empirical evidence for the so-called 'Pollyanna principle' as emotion mining identified more positive terms than negative ones in the texts. The difference was predominantly caused by the (high) frequencies of the words associated with the emotion of trust. The expected patterns were confirmed by the outputs of emotion mining. The fact that emotion-mining output on historical Dutch text collections confirmed this well-established principle was a first indication of the validity of lexicon-based emotion mining as a method for identifying and measuring the expression of emotions in the context of the particular sources used. A remaining question, however, was whether lexicon-based emotion

mining would be able to identify variation in emotionality in the Dutch historical texts at all.

The second part of Chapter 5 dealt with the question whether computational methods for emotion mining would be able to identify known peaks in emotionality. Notoriously emotional parliamentary discussions of post-war controversies related to (alleged) collaborators and war criminals served as a second step in the methodological evaluation.As discussedin Chapters 1 and 5, these resurfacing controversies over the decades are well-known, and elaborately studied and discussed – not only in retrospect or in the relevant historiography, but also in a contemporary context. Academic literature on perpetrators and collaboration was clear: A high emotional charge and association with negative emotions such as anger, sadness, fear, and disgust of this case study was, for the entire period, evident. Emotion-mining results replicated and supported the claim of exceptional emotionality of these discussions, especially (as expected) regarding negative emotions. From a historiographical perspective, I argue, these results further support the validity of emotion mining as a historical research instrument. Lexicon-based emotion mining proved able to capture the exceptionality of the emotional charge of this well-known historical case study in comparison with other, non-related parliamentary debates (i.e., the baseline of emotion in parliament set out in the first part of Chapter 5).

Interestingly, Chapter 5 showed that the 1972 Breda Three case, which is often considered exceptional, turned out not to be unique in its high emotional charge. Whilst it was indeed an emotional peak, other parliamentary discussions – those on war criminals and on 'purification' in the 1940s, on the Willy Lages cases in 1952 and 1966, and on several other public controversies up until 1989 – resemble the 1972 case regarding the exceptionally high scores for language use associated with negative emotions. This finding shows how historical episodes possibly gain additional meanings (or extra emotional charge) in (retrospective) processes of meaning-making, more than can be observed purely in the contemporary sources themselves. In addition, the results of the analyses in Chapter 5 showed how emotionality was primarily not so much time-bound, as it was bound to certain topics.

#### **10.2 The Resistance Disposition and Extraordinary Government Employees**

In Chapter 6, the results of the first in-depth case study of this investigation were described. Employing a computer-assisted approach, the question whether emotions were expressed in parliamentary debates dealing with the so-called 'extraordinary government employees' was addressed. Some of the debates in the subset scored higher on (negative) emotionality than average, non-related parliamentary debates. Overall, this first case study did not score as anomalously high on negative emotions as the 'perpetrators and collaboration' debates of Chapter 5. More interesting, however, is the finding that there was significant variation in emotion scores for the different debates on 'extraordinary government employees' that were held between 1947 and 1985, in particular for the negative emotions anger, sadness, fear, and disgust. The low scores of the early 1947 debates also demonstrated that what was considered important, was not necessarily always accompanied by strong negative emotional expressions in the speeches. At the other end of the spectrum, a source of variation were high-scoring debates in the 1950s. I analysed these observed peaks in Chapter 7, which dealt with the in-depth, close-reading scrutiny of the same subset of debates on the legislative schemes known as WBP and WBPZ.

Expressions of emotion were frequent in the high-scoring debates of the 1950s. This also became apparent from close-reading analysis and corroborates the results of analyses in Chapters 5 and 6: Emotions were a constant, and therefore neither absent, nor concealed, nor silenced in this period. The expression of strong negative emotions in these peak debates, however, was not primarily connected to MPs' expressions of (their own) emotions in their speeches, nor did emotions in these debates relate directly to historical experiences of the former resistance, or to the horrors of the recent war past. Rather, expressions of emotions in these debates were primarily associated with contemporary matters. Descriptions of lasting and contemporary suffering of the potential beneficiaries of the WBP and WBPZ legislation inspired strong expressions of negative emotionsin the debates of the 1950s.MPs recalled the experiences and problems of these people and strategically re-directed the related emotions into the parliamentary arena. Often, the results of emotion mining on the different negative emotion categories (anger, sadness, fear, and disgust) overlapped here. In 1955, however, when a single MP made emotionally charged descriptions of contemporary and individual suffering of identifiable victims into a cause, sadness scores in particular stood out. Scrutiny of the debates also showed that these peaks in sadness-associated word use were not so much used to give expression to lived sadness.Rather, poignant examples of circumstances of someidentifiable individuals and words associated with sadness were used to refer to the sad circumstances of others. This sadness was used both to evoke and invoke feelings of compassion for specific groups that fell (according to the MPs involved) through the cracks of the current legislative arrangements. This is unsurprising – indeed, it makes perfect sense – if we take into account the fact that MPs are professional language users.They chose their words carefully, as they were doing their work: representing the contemporary interests of their electorates or constituencies.

Another source of strong emotional expressionsin the debates lay more removed from the war survivors themselves. This was not a coincidence, as these emotions were much more political or strategic in nature. Closer analysis of debates in the WBP and WBPZ subset showed how pressing political issues of the day evoked very strong emotional language use in the discussions. In the midst of rising Cold War tensions, the role of emotions served, for example, to take a stance regarding the contemporary inclusion or exclusion of communists. In these emotional peak moments, MPs were not so much assessing wartime history, the legislation itself, or its contemporary consequences for beneficiaries. Rather, they used the debates as an opportunity to boost political suspicion in doing contemporary politics. The emotions connected to the 'extraordinary government employees' were given a role in the discussion of something else. Because of a strong and omnipresent 'resistance disposition', the WBP and WBPZ debates seemed a perfect vehicle for Cold War (party) politics. The rhetorical scope of the parliamentary debates went much further than the resistance legislation itself. In practice, however, the actual policy making and voting within parliament was barely affected by this moral posturing. However, this ColdWar politics caused peaks in the emotion scores of the debates in the mid-1950s within the subset on 'extraordinary government employees'.

#### **10.3 Emancipating Emotions in the War Victims Case**

Compared to the 'extraordinary government employees' case study, the emotionmining results of the subset of debates of case study 2 on war victims showed, initially, less variation between the emotion scores of various discussions in the Dutch parliament over the decades. This is remarkable, given the fact that this subset consisted of a much more varied thematic scope compared to the recurring discussion of the WBP and WBPZ addressed in Chapters 6 and 7. The emotion scores of the war victim debates are, for the negative emotions, predominantly higher than the parliamentary average in the 1945–1989 period. When it comes to the distribution of the debates on the 1945–1989 timeline, the years 1949–1950 displayed a peak in the number of debates.

First, following Jay Winter's definition of 'silence' – which describes the circumstances in which and about which subjects and words normally used in everyday life remain unspoken – a commonplace view of a phase of silence does not hold true for this period in the context of Dutch parliamentary discussion.<sup>1</sup> Although the legislation that emerged in the 1950s was often seen as rather limited in hindsight, this did not mean that war victims' suffering was silenced in parliament. The analysis of the debates showed there was no such silence in parliament. There was much debate that took various directions and expressed diverging ideas. It is therefore unfair to state that contemporary parliamentarians had no eye for war-related suffering. Moreover, this finding also shows that legislation itself is not necessarily a very appropriate measure or framework for investigating broad societal developments. The timeline of established legislative schemes, aimed at alleviating the suffering of

<sup>1</sup> Winter, 'Thinking about Silence', 4.

victims of World War II in the Netherlands, is not representative of what Jolande Withuis called 'the mental climate' of the era.<sup>2</sup>

Second, although these debates dealt mostly with alleviating the material consequences of the German occupation, the characterisation of the parliamentary style of this episode in dealing with war victimhood as 'legal' or 'administrative' does not reflect reality. Emotional descriptions of the misery of identifiable individual victims were a constant presence in debates on, for example, the Material War Damage Act in 1949 and 1950. The language use in these contributions fuelled the scores for emotional language use. The role emotions played in these debates can be characterised as strategic. This process was part of a deliberate practice to emphasise the gravity of what had happened in the recent past, and to support the salience of the issue. The emotions, used in emotional and poignant descriptions by many MPs involved, embellished the argumentation for what was considered 'just' and 'fair' in policy making regarding alleviating the material consequences of the German occupation. In terms of the mere presence of emotions in the debates, later parliamentary discussions were less of a break with the 1950s than was often presumed in historiography. Where the idea of the silence of the 1950s originated, was revealed via the close-reading analysis of much later debates in the Dutch House of Representatives. In the early 1970s, a long standing demand for more elaborate and tailored war victim legislation was placed centre stage by emphasising the (perceived) shortcomings of preceding legislation and past politicians. In this process, MPs stimulated the creation of some sort of myth: As an argument for more elaborate legislation, they emphasised the idea that the government had previously neglected or forgotten war victims. They pointed towards a past 'silence' that was not a 'sacred silence', but one of political negligence. Years later, this myth of 'late action' or previous 'silence' resonated in historiography, albeit not as a normative judgement or political argument, but rather as an analytical category underlying a process of historiographical periodisation.<sup>3</sup>

The results of emotion mining on the subsets of the case studies indicated gaps in the quantitative measurements in the 1960s. However, in neither case did this have anything to do with debates scoring a zero on emotional language use. Rather, a lack of substantive debate was behind the gaps in the observations and resulting graphs. The quantitative outputs alone did not clarify what was going on, or what kind of development was behind these lacunae in the measurements. Close reading – by going back to the actual debates, and to the thematic subset retrieval process – offered an explanation. The reason for the gaps differed in both case studies. In the 'extraordinary government employees' subset, the observed gap was a result of

<sup>2</sup> Withuis, *Erkenning*, 118–19.

<sup>3</sup> See Chapter 1 and De Haan, *Na de ondergang*; Bossenbroek, *De Meelstreep*; Withuis and Mooij, 'From Totalitarianism to Trauma'; Piersma, *Bevochten recht*.

the retrieval and selection processes of this study (Chapter 3), caused by the fact that MPs mostly accepted proposals for modification of the WBP and WBPZ schemes without any substantive discussion. Proposals often passed without there even being a vote in the House of Representatives. In the 'war victims' case study, the lacuna had more to do with contemporary administrative and organisational political practices than being a consequence of concealment of emotions,or a lack of talking about war victims in parliament. This finding highlights the limitations of the computational techniques used and that of the rather rigid selection and retrieval procedure of subsets applied to the *Handelingen* dataset. It was not that parliamentary debates did not take place, but rather that the discussion of war victims was incorporated in other, overarching thematic debates. This was, in its turn, a consequence of the policy to incorporate legislative arrangements for war victims in generic social welfare legislation.

In addition to the findings of Chapter 5, emotion-mining results in Chapters 6 and 8 also gave no indication of any overwhelming 'emotionalisation' that was preceded by a phase of concealment of emotional expressions in parliamentary debate. However, the fact that the discussions' emotion scores remained rather stable and consistent did not mean that nothing changed over the decades. Rather, it was that the numeric output of emotion mining simply did not capture this (form of) historical change. Yet the linguistic output of emotion mining did offer a first indication of a qualitative development in the role of emotions, even if these outputs did not provide a comprehensive understanding of developments, changes, or how a possible 'emancipation of emotions' unfolded in the debates.The assessment of the most frequent lexicon words behind the scores gave only a first indication of shifts in the parliamentary discussions. As discussed in Chapters 8 and 9, these word lists indicated (in the 'war victims' subset) a change from an 'empathetic' attitude in parliament, to a more 'appraising' style of dealing with emotions. However, this development of taking emotions more explicitly into account (talking *about* emotions) in political processes did not parallel with an overwhelming increase in the manifestation of emotional language over time (the expression *of* emotions).

Chapter 9 showed, by means of close reading, how the qualitative aspects of the role of emotions in the various discussions that dealt with war victims changed over the decades. Close reading was essential in gaining this better understanding of the meaning and details of the complex and multifaceted developments that ensued.The results of an in-depth study of the actual course of the debates pointed towards more than only a slight difference in nuance regarding Withuis' cyclic phase of emancipating emotions, providing instead a detailed account of how this process evolved in the practice of parliamentary discussions. Chapter 9 illustrated how, in the 1970s, the role of emotions had changed, when compared with that of the first 20 years after the war. I showed how expressing emotion or behaving in an emotional way ('emotionalisation', 'emotion culture') is not (and indeed, never became)

the same as talking about the emotions (of others; 'emancipation of emotions').<sup>4</sup> The observed developments can be dissected into various – sometimes overlapping – shifts in the role of emotions in the context of the Dutch parliament's engagement with war victim legislation in the decades after the war. Chapter 8, and especially Chapter 9, revealed how MPs took the emotions (of others) more explicitly into account in the parliamentary debates, processes, and policy making of the 1970s. This development, at least in the parliamentary context, was not predominantly a consequence of broader societal developments – nor of the mental climate, for that matter – but was instead a political strategy. Emotions were deliberately given another role in discussions.With appraisals of the (societal) emotions of others, and what Frevert has called 'the politicisation of emotions', especially opposition-party MPs tried to mobilise their coalition colleagues and the government to come up with more elaborate and autonomous war victim benefit legislation.<sup>5</sup> In this way, emotions did play a role in establishing legislation aimed at survivors of the war.

Nevertheless, this study concludes that emotions alone do not offer a satisfactory explanation for the legislative developments that were observed. Not only was an 'emotionalisation' process not identified in the debates; the observed changes in the role of emotions alone also offer no sufficient explanation for the quite drastic legislative developments of the 1970s. Social, economic, and political changes offer much more plausible explanations. Whilst emotions undoubtably did play a role, it was more as a necessary condition, rather than as a sufficient condition for the creation of more elaborate and autonomous legislative schemes such as the WUV and the WUBO. Socio-economic and political developments (such as proto-professionalisation and de-stigmatisation of trauma),increased government spending and intervention, together with an expanding welfare state (and its public and political acceptance) can be considered as the essential building blocks.The manifestation of emotionsin parliamentary debates functioned as the necessary scaffolding supporting the construction of more elaborate and tailored national war victim legislation in the 1970s.

That the German occupation during World War II left an enduring emotional imprint in Dutch society is not surprising – and was also the starting point of this study. By quoting Gerdi Verbeet, Chapter 1 highlighted how this history and its lasting consequences continue to elicit emotional responses today. In 2016, Verbeet emphasised 'the sensitivities' and 'feelings' related to a national commemoration of the war. This study showed how the circumstances that Verbeet illustrated, unfolded in parliament in the 1980s: War-related legislation became something about which it was unimaginable, or at least socially unacceptable, to think, speak, or write without at the very least the explicit mention of the supposed emotionality of others.

<sup>4</sup> These three terms are derived from academic literature discussed in Chapters 1 and 5.

<sup>5</sup> Frevert, 'Emotional Politics'.

This, however, did not mean that debates became more emotional or more deeply involved with personal emotions. To the contrary, parliamentary attitudes developed towards a more abstract, observant, and distant dealing with (other people's) emotions. The emotion-mining results of the 1980s debates displayed a change in the distribution of words in the top word lists: Emotion words were still present, but the number of different unique words used dwindled. Simply mentioning or using the word 'emotion' itself had become sufficient.The politicians involved were apparently so imbued with the importance and the emotional charge of the topic that they considered further strategic embellishment of the issue with emotions redundant. These findings corroborate the claims made by Hoetink, regarding the development that emotions became less prominent in debates because of the professionalisation and technocratisation of Dutch politics.<sup>6</sup> MPs considered the building of elaborate, autonomous, and tailored war victim legislation self-supportive and strong enough in the 1980s. The emotional scaffolding put up by their colleagues could safely be dismantled, but not without leaving behind a deep emotional imprint.

Whilst it lies outside the scope of this investigation, in future research it would be interesting to assess whether and how this development of a more abstract and distant dealing developed towards andinto the twenty-first century. It seems plausible that this development expanded further. Building on the 'sensitivities' that Gerdi Verbeet mentioned in 2016, she stated five years later that '(…) the further we get from the war, the more careful we have to be'.<sup>7</sup> This could indicate, for example, the emergence of a situation wherein not only mentioning 'the emotions', but referring to 'the war' itself has become a manifestation of emotionality – an act of expressing emotion in itself.

#### **10.4 The Value of Computer-assisted Historical Research**

The aim of this study was twofold. Next to the substantive historical questions outlined above, this investigation was also an assessment of the value of integrating computational techniques for quantitative text analysis into historical research. Digital resources, computational techniques and methods, off-the-shelf software, and the large-scale digitisation or 'datafication' of historical text collections made it feasible to analyse millions of words over several decades. Speed, scale, and comprehensiveness proved to be first and obvious advantages of the computer-

<sup>6</sup> Hoetink, *Macht der gewoonte*, 421–98.

<sup>7</sup> Coen Verbraak, 'Gerdi Verbeet: "We mogen nooit vergeten dat mensen tot massamoord in staat zijn"', *NRC*, 29 April 2021, https://www.nrc.nl/nieuws/2021/04/29/gerdi-verbeet-we-mo gen-nooit-vergeten-dat-mensen-tot-massamoord-in-staat-zijn-a4041828.

assisted methodology. The outputs of emotion mining – the predominant computational method applied – were especially informative when investigating broader diachronic trends, and in mutually comparing different themes, years, or source collections. The distant perspective of the macroscope made the outer contours of historical trends, developments, and phenomena visible – more so than would have been feasible by the individual historical researcher.The outputs of emotion mining showed in this study, for example, how the curve of the diachronic development of emotional manifestations in the parliamentary discussion of 'perpetrators and collaboration' could be compared to a high mountain plateau, whereas that of the debates on the former resistance bore a greater similarity to a bumpy road.

However, the computer-assisted research methodology also raised questions about the validity of the computational methods and techniques used. However systematic, transparent, well-evaluated, or established a lexicon or other type of (linguistic) digital resource or computational technique may seem, basic evaluation of the validity of results was still necessary. Chapter 4 stated how, rather than the standards of the fields usually concerned with NLP and text mining, it was the historiographic research context that was of primary concern in this investigation. Chapter 5 showed how, in multiple steps, a basic and rudimentary evaluation of the validity of the methods was established. This process was true to the nature of the historical discipline, as it was based on secondary literature, universal principles, a well-known historical case study, and a close reading of the sources. As the results presented in Chapter 5 demonstrated, predominantly the distinction between positive emotions (trust and joy) and negative emotions (anger, fear, sadness, and disgust) was observable in emotion scores and trends. This finding concurred with the so-called 'Pollyanna principle'. However, on the more fine-grained level of individual negative basic emotions, the similarities in the patterns of scores raised questions.

Some exceptions aside, differences between patterns in the scores for the four negative basic emotions (anger, fear, sadness, and disgust) were often negligible. The 'words behind the scores' for each individual basic emotion (presented in tables in the supplement) often overlapped.The complexity and versatility of emotions and the way they are expressed, manifest, or circulated in real-life social interactions inform an initial possible explanation for the similarities between the different emotion scores. As discussed in Chapter 2, many of the emotions humans feel, express, evoke, and encounter consist of a combination of two or more basic emotions. The theoretical underpinning of the more fine-grained differentiation between the various basic emotion categories is derived from neuroscience and psychology.This study's findings raise the question whether (and if so, how) a theoretical differentiation is applicable outside the psychological, personal, or clinical realm. Another possible explanation for the similarities between the different negative basic emotions' scores could lie within the nature of the methods or resources used. The difficulty to differentiate between basic emotions in emotion mining could also be the result of the choices made in creating the NRC EmoLex (as its makers often attributed a single unique word to more than one basic emotion category). Another possibility (which would present an interesting avenue in future research) is that the differentiation between basic emotions becomes fuzzy for source-related reasons. Perhaps the basic emotion categories in lexicons do have their differentiating capabilities only when applied to small, very specific texts created by a single author. It is also possible that the issue is particular to the historical Dutch text corpora analysed. It is, for example, plausible that MPs transform an emotion such as disgust into an expression of anger, as they perceive the latter to have stronger performative or persuasive qualities in the parliamentary context.

The application of lexicon-based emotion mining in historical research also showed that emotion scores are never a final destination in and of themselves. Emotion-mining results, even when convincingly visualised in neat graphs or pretty boxplots, are not end results on which conclusions can be readily drawn, or historical claims can be instantly made. The results of computational methods in a computer-assisted research methodology should be understood in relation to the methodological and analytical decisions made, to the resources used, and not as self-evident or objective facts about the textual sources under scrutiny. These decisions can, in their turn, also be considered as interpretive acts that shape the outcomes of an investigation.<sup>8</sup> What this study did show, is how the output of text mining can be used as another source that requires further critical scrutiny. In understanding and fully grasping these emotion-mining outputs, the statistical evaluation proved to be the first important next step in this process.

The statistical evaluation of the numerical output (Chapters 5, 6, and 8) complemented the emotion-mining scores and visualisations. The statistical tests indicated, for example, that differences observed in scores were either not as significant as they initially seemed, or confirmed observed variation and change – or the lack thereof. The same goes for the value of the calculated effect magnitudes. The statistical evaluation of the results complemented the 'raw' emotion-mining scores and visualisations, and made emotion mining a better descriptor of what was possibly going on in the debates in terms of historical change and trends, and in the comparison of debates.These results informed further close reading, investigation, and interpretation. Sometimes in this study, the output of emotion mining pointed rightfully at peaks or troughs in emotionality, however, sometimes the results could only be explained by external factors that had more to do with other factors than with the manifestation of emotions in language. I argue that even when working with largescale historical text collections – particularly those covering a longer time span –

<sup>8</sup> Drucker et al., 'Distant Reading and Cultural Analytics'.

or when arriving at straightforward conclusions, diving deeper into the sources remains an indispensable part of the analysis.

This study clearly demonstrated that not every historical phenomenon or perceived historical change was equally well observable with (the outputs of) emotion mining. However, the use of computational techniques kept open the possibility to return to the sources.The historical phenomenon of 'the emancipation of emotions' was, for example, much better understood after closer scrutiny of the sources. This not only entailed taking a closer look at the quantitative output and at the emotional terms behind those scores, but also a close reading of qualitative aspects as they unfolded over the course of the discussions. Especially the outcomes of the closereading analyses in Chapters 7 and 9 pointed at reification (the fallacy of misplaced concreteness) as it made clear the boundaries of the apparent objectivity of the numbers and quantitative outputs of the methods used. An example is the sadness peak of 1955, observed in Chapter 6. After in-depth analysis in Chapter 7, this peak turned out to be,in fact,an elaborate sequence of poignant examples of sad situations.Close reading of the same debate displayed how this was more an act of showing compassion, than it was an exceptionally sad episode in parliamentary history.

Despite the drawbacks, the most important advantages of the lexicon-based approach remain. The indispensable formalisation of research criteria in emotion mining made explicit the criteria that define what is considered emotional, and what is not. This helped, first and foremost, in making the analytical process replicable and transparent, and pushed the boundaries of (tolerance for) fuzziness in the right direction. In addition, it forced a critical rethinking and evaluation of (personal) preconceptions, commonplace views, and received wisdoms in the process. Emotion mining makes researchers less reliant on their own personal preconceptions, ideas, and interpretations, or commonplace views. Lexicon-based emotion mining offered this study an outside perspective when investigating commonplace views on the opposition between, for example, 'silence' and 'emotionalisation'. More than with a traditional historiographical approach, and more than would ever be feasible for an individual researcher, emotion mining offered a comprehensive, distant, and outside perspective that was true to the (semi-) quantitative nature of the historiographic claims regarding broad and diachronic developments in publicly expressing emotions. The – now empirically supported – farewell to the received wisdoms of the 'silence of the 1950s' or the 'emotionalisation of the 1970s' is a particularly valid example of the merits of the computer-assisted research methodology.

This brings this book to a conclusion regarding the computer-assisted methodologies used. Lexicon-based emotion mining as a computational method in a historical research process remains rather crude in its nature, as the more subtle human judgement is inevitably lost in the process. Outputs gave only partial, sometimes distorted insights into historical developments. However, the fact that a measuring instrument is somewhat fallible makes it neither worthless nor redundant. In many research contexts, incomplete, fallible measuring instruments are not only convenient and valuable, but also necessary. Take body temperature; not a very precise instrument for detailed medical diagnosis, but if the measurement is over 40 degrees Celsius, you know that something is wrong. The same goes for radar: Whilst it may not show you precisely what the particular object rapidly approaching you is, at least you know there is something to steer away from (or to drop the bombs).The position of emotion-mining results in historical analysis, interpretation, and claims is similar. Here considered as fundamental part of 'the historian's macroscope', emotion mining offered an auxiliary perspective that allowed for the drawing of rough contours of historical phenomena and trends. Instead of making only inferences about these contours, emotion mining created a truly empirical perspective on them. The emotion-mining outputs drew the contour lines,and with additional statistical evaluation and visualisation, the results showed if and where 'something was going on'. Emotion mining offered a first glimpse of 'the great unread'. This study has shown how the computationalmethods offer an expansion of the historian's research scope. Plotted in graphs, boxplots, or tables, the outputs made it possible to uncover diachronic trends, and to set a baseline to which the more particular case studies of Chapters 5, 6, 7, 8, and 9 could be related. The computational approach to emotions offered a contextual view of the proportionality, size, and course of broader trends. In other words, the outputs of computational methods provided a considerable degree of additional and auxiliary context.

This investigation also demonstrated that computational methods are not a substitute for traditional historiographical practicesin the field of academic history.The outputs of emotion mining still had to be filled in with the results of sustained, close reading – a practice common to the historical field. This entailed more than simply reading, as the hermeneutic nature of academic history is about clarification, interpretation, understanding, and finding ways to grasp and explain the meanings of developments, or the continuities or changes observed in the sources. Text mining, its outputs, statistics, and graphs proved useful, offering an alternative, complementary, and auxiliary perspective on the historical sources.This perspective has the important advantage of being less reliant on personal preconceptions and interpretations, as it allows for the comparison of results of close- and distant-reading analysis. In this investigation, the output of emotion mining offered complementary, systematic, and diachronic evidence, and a comprehensive contextual framework of the role of emotions in parliamentary discussions over time. Relying solely on traditional historiographic practices, such a framework was hitherto impossible.

Especially the two detailed case studies have shown how the benefit of the historian's macroscope lies in its ability to zoom both in and out on the sources. In ideal circumstances, distant and close reading, qualitative and quantitative methods, are not considered as disconnected phases in the analysis. Rather than the opposition between the two, it is their combination, in a back-and-forth approach that allows for a greater diversity of historical evidence and research perspectives. The benefit here lies in the combination of close reading with the results of computational methods. Combining these two approaches brings the best of both worlds to the fore.This investigation's analytical workflow can therefore serve as an example of how these approaches are complementary and strengthening each other in historical research practice.

To conclude, this study revealed the imprints of historical emotions, as they were empirically observable in the sources. This led, not to a history *of* emotions per se, but to an investigation in which the study of emotional imprints provided an important additional perspective in the re-evaluation of a more mainstream historical phenomenon. Computational methods in a computer-assisted methodology made possible an assessment of the question whether emotions played a role in shaping and debating war-related legislation.The results providedinsights into how this role developed in the Dutch parliament between 1945 and 1989. Rather than emotions *making* history themselves, parliamentarians *gave* emotions a role in the language of their speeches and discussions. This investigation, however, never transcended the universe of the spoken and the recorded. Although the computer-assisted approach offered valuable insights in something as volatile, complex, and versatile as emotions in history, they are on, and perhaps even slightly over, the edge of what is knowable after all.

## **Supplements**

In this supplement, to which is referenced throughout the book, graphs, statistics, tables, and background information can be found.

#### **Supplement 1 | Chapter 3 | Materials and Data**

#### 1.1 Controlled Vocabulary and the *Handelingen* Metadata

The table below contains all documents from the Political Mashup *Handelingen der Staten-Generaal* dataset retrieved and selected as being relevant to the case studies of Chapters 5, 6, 7, 8, and 9. The date and document metadata are derived from the Political Mashup dataset, the subset, thematic cluster, and topic tags are added during the retrieval and selection procedures described in Chapter 3. For layout- and space-related reasons, the complete transcripts of the discussions are here not included in the table.














#### 1.2 Additional Search Terms

These search terms were used in the additional manual retrieval and selection steps described in Chapter 3. The keywords are derived from the relevant literature and the NIOD World War II-thesaurus.<sup>1</sup> They were used in additional keyword search in the Political Mashup *Handelingen der Staten-Generaal* dataset.The keywords served

<sup>1</sup> NIOD Institute for War, Holocaust, and Genocide Studies, 'WO2-thesaurus'; Van Nispen and Jongma, 'Holocaust and World War Two Linked Open Data Developments in the Netherlands'.

the retrieval of a few additional debates that were not (yet) retrieved by the retrieval and selection procedure described in Chapter 3.

#### Additional search terms


#### 1.3 Blocklisted Lexicon Words

Very frequently occurring words in the *Handelingen* dataset (1945–1989) without an emotional charge in the particular context of parliamentary discussion are put on a block list. This blocklist is based on the 100 most frequent NRC Word-Emotion Association Lexicon words (per year) in the *Handelingen* dataset and the following criteria:


The blocklisted lexicon words that are removed from EmoLex are:

#### Blocklisted EmoLex words


### **Supplement 2 | Chapter 5 | Peering Through the Macroscope**

#### 2.1 The 'Emotionalisation' of Dutch Society

#### 2.1.1 The Queen's Speeches

*Graph 17: Negative emotions in the Dutch Queen's speeches (1945–1989).*

#### *Graph 18: Positive emotions in the Dutch Queen's speeches (1945–1989).*

*Table 18: Output linear regression analysis of negative emotion scores in the Dutch Queen's speeches (1945–1989).*


*Table 19: Output linear regression analysis of positive emotion scores in the Dutch Queen's speeches (1945–1989).*


#### 2.1.2 The Newspapers

*Graph 19: Negative emotion scores in Dutch newspaper De Telegraaf (1949–1989).*

*Graph 20: Negative emotion scores in Dutch newspaper De Volkskrant (1945–1989).*

*Graph 21: Positive emotion scores in Dutch newspaper De Telegraaf (1949–1989).*

*Graph 22: Positive emotion scores in Dutch newspaper De Volkskrant (1945–1989).*

*Table 20: Output linear regression analysis of negative emotion scores in newspaper De Telegraaf (1949–1989)*


*Table 21: Output linear regression analysis of negative emotion scores in newspaper De Volkskrant (1945–1989).*


*Table 22: Output linear regression analysis of positive emotion scores in newspaper De Telegraaf (1949–1989)*


*Table 23: Output linear regression analysis of positive emotion scores in newspaper De Volkskrant (1945–1989).*


#### 2.1.3 The Parliamentary Debates

*Graph 23: Negative emotion scores in Dutch parliamentary debates (1945–1989).*

*Graph 24: Positive emotion scores in Dutch parliamentary debates (1945–1989).*

*Table 24: Output linear regression analysis on negative emotion scores in parliamentary debates (1945–1989).*


*Table 25: Output linear regression analysis on positive emotion scores in parliamentary debates (1945–1989).*


#### 2.2 Pepetrators, Collaboration, and Emotion Peaks

#### 2.2.1 Boxplots and Summary Statistics

*Boxplot 5: Anger scores for each 250-word chunk in the debate subset.*

*Boxplot 7: Disgust scores for each 250-word chunk in the debate subset.*

*Boxplot 8: Sadness scores for each 250-word chunk in the debate subset.*

*Boxplot 9: Joy scores for each 250-word chunk in the debate subset.*

#### *Boxplot 10: Trust scores for each 250-word chunk in the debate subset.*

*Table 26: Summary statistics of emotion scores of all 'perpetrators and collaboration' debates (1945–1989).*


#### 2.2.2 Graphs over Time

*Graph 25: Anger scores of the 'perpetrators and collaboration' subset.*

*Graph 26: Fear scores of the 'perpetrators and collaboration' subset.*

*Graph 27: Disgust scores of the 'perpetrators and collaboration' subset.*

*Graph 28: Sadness scores of the 'perpetrators and collaboration' subset.*

*Graph 29: Joy scores of the 'perpetrators and collaboration' subset.*

*Graph 30: Trust scores of the 'perpetrators and collaboration' subset.*

#### 2.2.3 Output Tables Linear Regression Analysis

*Table 27: Output linear regression (anger scores) in the 'perpetrators and collaboration' subset.*


*Table 28: Output linear regression (fear scores) in the 'perpetrators and collaboration' subset.*


*Table 29: Output linear regression (disgust scores) in the 'perpetrators and collaboration' subset.*


*Table 30: Output linear regression (sadness scores) in the 'perpetrators and collaboration' subset.*


*Table 31: Output linear regression (joy scores) in the 'perpetrators and collaboration' subset.*


*Table 32: Output linear regression (trust scores) in the 'perpetrators and collaboration' subset.*



#### **Supplement 3 | Chapters 6 and 8 | Case Studies**

#### 3.1 Chapter 6 | 'The Resistance'

#### 3.1.1 Boxplots and Summary Statistics

This section contains the boxplots displaying the emotion scores for all 250-word chunks for each debate on WBP and WBPZ. Grey dots represent the individual observations (per chunk), the black line is the median and the symbols show the mean score (per debate).

*Boxplot 11: Anger scores for each 250-word chunk in the 'extraordinary government employees' debates subset.*

*Boxplot 12: Fear scores for each 250-word chunk in the 'extraordinary government employees' debates subset.*

*Boxplot 13: Disgust scores for each 250-word chunk in the 'extraordinary government employees' debates subset.*

*Boxplot 14: Sadness scores for each 250-word chunk in the 'extraordinary government employees' debates subset.*

*Boxplot 15: Joy scores for each 250-word chunk in the 'extraordinary government employees' debates subset.*

*Boxplot 16: Trust scores for each 250-word chunk in the 'extraordinary government employees' debates subset.*

*Table 33: Summary statistics of emotion scores of all 'extraordinary government employees' debates (1945–1989).*


#### 3.1.2 Output Kruskal-Wallis Test

*Table 34: Output Kruskal-Wallis test on emotion scores of all 'extraordinary government employees' debates (1945–1989).*


#### 3.1.3 Output Wilcoxon Rank Sum Test

As a measure of statistical significance of differences between the emotion scores of the multiple debates, a pairwise comparison using aWilcoxon rank sum test is used. This test calculates the statistical significance of all emotion scores of every possible pair of debates on the 'extraordinary government employees' in the dataset.<sup>2</sup> The output tables show the two debates compared, the effect size, the number of 250 word chunks for each debate, the magnitude of the effect, and a p-value. Because of

<sup>2</sup> Unless otherwise specified a confidence interval of 95% (α = 0.05) is used. A difference is statistically significant when (p =< α).

the size of the Wilcoxon rank sum test output tables, these are not incorporated in this supplement. The complete test results can be found on Github.


#### 3.1.4 Graphs over Time

*Graph 31: Anger scores of the 'extraordinary government employees' debates subset.*

*Graph 32: Fear scores of the 'extraordinary government employees' debates subset.*

*Graph 33: Disgust scores of the 'extraordinary government employees' debates subset.*

*Graph 34: Sadness scores of the 'extraordinary government employees' debates subset.*

*Graph 35: Joy scores of the 'extraordinary government employees' debates subset.*

*Graph 36: Trust scores of the 'extraordinary government employees' debates subset.*

#### 3.1.5 Output Tables Linear Regression Analysis

*Table 35: Output linear regression (anger scores) in the 'extraordinary government employees' debates subset.*


*Table 36: Output linear regression (fear scores) in the 'extraordinary government employees' debates subset.*


*Table 37: Output linear regression (disgust scores) in the 'extraordinary government employees' debates subset.*


*Table 38: Output linear regression (sadness scores) in the 'extraordinary government employees' debates subset.*


*Table 39: Output linear regression (joy scores) in the 'extraordinary government employees' debates subset.*


*Table 40: Output linear regression (trust scores) in the 'extraordinary government employees' debates subset.*


#### 3.1.6 Top Emotion Word Lists

The top 20 emotion words (based on TF-IDF weights) are plotted for each emotion category and each parliamentary debate (aggregated per day) in the 'extraordinary government employees' subset. As can be seen in the graphs, not every debate contained (more than) 20 emotion lexicon words.

*Graph 37: Top anger words in the 'extraordinary government employees' subset.*

*Graph 38: Top fear words in the 'extraordinary government employees' subset.*

*Graph 39: Top disgust words in the 'extraordinary government employees' subset.*

*Graph 40: Top sadness words in the 'extraordinary government employees' subset.*

*Graph 41: Top joy words in the 'extraordinary government employees' subset.*

*Graph 42: Top trust words in the 'extraordinary government employees' subset.*

#### 3.2 Chapter 8 | 'War Victims'

#### 3.2.1 Boxplots and Summary Statistics

This section contains the boxplots displaying the emotion scores for all 250-word chunks for each thematic debate cluster. Grey dots represent the individual observations (per chunk), the black line is the median and the symbols show the mean score (per debate).

*Boxplot 17: Anger scores for each 250-word chunk in the 'war victims' debates subset.*

*Boxplot 18: Fear scores for each 250-word chunk in the 'war victims' debates subset.*

*Boxplot 19: Disgust scores for each 250-word chunk in the 'war victims' debates subset.*

*Boxplot 20: Sadness scores for each 250-word chunk in the 'war victims' debates subset.*

*Boxplot 21: Joy scores for each 250-word chunk in the 'war victims' debates subset.*

*Boxplot 22: Trust scores for each 250-word chunk in the 'war victims' debates subset.*

*Table 41: Summary statistics of emotion scores of all war victim debates (1945–1989).*



#### 3.2.2 Output Kruskal-Wallis Test

*Table 42: Output of Kruskal-Wallis test on individual debates on war victims (1945–1989).*


*Table 43: Output of Kruskal-Wallis test on thematic clusters of war victim debates (1945–1989).*


#### 3.2.3 Output Wilcoxon Rank Sum Test

As a measure of statistical significance of differences between the emotion scores of the multiple debates, a pairwise comparison using a Wilcoxon rank sum test is used. This test calculates pairwisely the statistical significance of all emotion scores of every possible pair of debates on the 'extraordinary government employees' in the dataset.<sup>3</sup> The output tables show the two debates compared, the effect size, the number of 250-word chunks for each thematic cluster of debates, the magnitude of the effect, and a p-value. Because of the size of the Wilcoxon rank sum test output ta-

<sup>3</sup> Unless otherwise specified a confidence interval of 95% (α = 0.05) is used. A difference is statistically significant when (p =< α).

bles, these are not incorporated in this supplement.The complete test results can be found on Github.


#### 3.2.4 Graphs over Time

*Graph 43: Anger scores of the 'war victims' debates subset.*

#### *Graph 44: Fear scores of the 'war victims' debates subset.*

*Graph 45: Disgust scores of the 'war victims' debates subset.*

*Graph 46: Sadness scores of the 'war victims' debates subset.*

*Graph 47: Joy scores of the 'war victims' debates subset.*

*Graph 48: Trust scores of the 'war victims' debates subset.*

#### 3.2.5 Output Tables Linear Regression Analysis

*Table 44: Output linear regression (anger scores) in the 'war victims' debates subset.*


*Table 45: Output linear regression (fear scores) in the 'war victims' debates subset.*


*Table 46: Output linear regression (disgust scores) in the 'war victims' debates subset.*


*Table 47: Output linear regression (sadness scores) in the 'war victims' debates subset.*


*Table 48: Output linear regression (joy scores) in the 'war victims' debates subset.*


*Table 49: Output linear regression (trust scores) in the 'war victims' debates subset.*


#### 3.2.6 Top Emotion Word Lists

The top 20 emotion words (based on TF-IDF weights) are plotted for each emotion category and each thematic cluster of debates in the 'war victims' debates subset.

*Graph 49: Top anger words in the 'war victims' debates subset.*

*Graph 50: Top fear words in the 'war victims' debates subset.*

*Graph 51: Top disgust words in the 'war victims' debates subset.*

*Graph 52: Top sadness words in the 'war victims' debates subset.*

*Graph 53: Top joy words in the 'war victims' debates subset.*

*Graph 54: Top trust words in the 'war victims' debates subset.*

## **Bibliography**


Boddice, Rob. *Pain and Emotion in Modern History*. Palgrave Studies in the History of Emotions. Houndmills, Basingstoke, Hampshire: Palgrave Macmillan, 2014.


Bosscher, Doeko. 'De jaren vijftig epischer geduid'. *BMGN − Low Countries Historical Review* 112, no. 2 (1997): 209–26. https://doi.org/10.18352/bmgn-lchr.4453.

Bossenbroek, Martin. *De Meelstreep*. Amsterdam: Bakker, 2001.


Brezina, Vaclav.*Statisticsin Corpus Linguistics: A PracticalGuide*.Cambridge University Press, 2018.


*ropean Countries*, edited by Jolande Withuis and Annet Mooij, 241–68. Amsterdam: Amsterdam University Press, 2010.


Goldstone, Andrew. 'Producing a Segmented Corpus of Novels', 19 April 2015, 4.


Leenes, Jelle.*Hollandse luchten: ruiken aanNederland*. Amsterdam: Atlas Contact, 2011.


Mooij, Annette Jacoba. *De Strijd Om de Februaristaking*. Amsterdam: Balans, 2006.

Moretti, Franco. '"Operationalizing": Or, the Function of Measurement in Modern Literary Theory', Pamphlets of the StanfordLiteraryLab, 60, no. 1 (March 2014): 3–19. https://doi.org/10.15794/JELL.2014.60.1.001.


Rockwell,Geoffrey, and Stéfan Sinclair.*Hermeneutica: Computer-Assisted Interpretation in the Humanities*. Cambridge, Massachusetts and London: The MIT Press, 2016.

Romein, C. Annemieke, Max Kemman, Julie M. Birkholz, James Baker, Michel De Gruijter, Albert Meroño-Peñuela,Thorsten Ries, Ruben Ros, and Stefania Scagliola. 'State of the Field: Digital History'. *History* 105, no. 365 (14 May 2020): 291–312. https://doi.org/10.1111/1468-229X.12969.


#### **Sources**

De Telegraaf, 1949–1989.

De Volkskrant, 1945–1989.

Handelingen Eerste Kamer der Staten-Generaal (HEK), 1945–1990.

Handelingen Tweede Kamer der Staten-Generaal (HTK), 1945–1990.

Troonredes, 1945–1989.

#### **Software and packages**


## **Author Information**

Milan Mikolaj van Lange's research interests are at the intersection of war studies, contemporary history, the digitisation of historical archives, and innovation in research methodology, including the integration of computational research techniques, text mining, data analysis, statistics, and programming in the historical research practice.

Milan was born in Enschede, the Netherlands in 1992. After completing his secondary education at Het Stedelijk Lyceum Kottenpark (2010) in Enschede, he studied History at the Radboud University in Nijmegen.Milan received his BA in History in 2013 and his MA (cum laude) in History in 2014. In 2015 and 2016, he worked as a research assistant for Prof. dr. Ismee Tames on the integration of computational methods in historical research at the NIOD Institute for War, Holocaust, and Genocide Studies in Amsterdam.

In September 2016, Milan started his PhD research project 'War and Emotions' at the NIOD Institute for War, Holocaust, and Genocide Studies and Utrecht University under supervision of prof. dr. Ismee Tames and prof. dr. Ralf Futselaar. He also worked together with prof. dr. Yasuto Nakano at Kwansei Gakuin University in Nishinomiya, Japan and participated in summer schools at the Institute for Advanced Studies in Humanities and Social Sciences in Nanjing, China (2017) and at the Sogang University in Seoul, South Korea (2019) on war, mass violence, and global post-war and post-colonial memories. In 2019, he was granted a Research Fellowship at the Luxembourg Centre for Contemporary and Digital History (C2DH) located at the University of Luxembourg campus in Belval, Luxembourg.

Between 2016 and 2021, Milan published on the post-war political engagement with war criminals, parliamentary discussions about war victims, nuclear weapons, emotions in post-war society, and the application of text- and data mining in the context of historical research. Milan is one of the initiators of the digitisation project 'First-Hand Accounts of War: War letters (1935–1950) from NIOD digitised'. He works currently as a researcher at the NIOD Institute for War, Holocaust, and Genocide Studies in Amsterdam.

# Bielefeld University Press

Alexandra Scheele, Julia Roth, Heidemarie Winkel (eds.)

#### Global Contestations of Gender Rights

 2022, 354 p., pb., col. ill. 40,00 € (DE), 978-3-8376-6069-2 E-Book: available as free open access publication PDF: ISBN 978-3-8394-6069-6

#### Silke Schwandt (ed.) Digital Methods in the Humanities Challenges, Ideas, Perspectives

 2020, 312 p., pb., col. ill. 38,00 € (DE), 978-3-8376-5419-6 E-Book: available as free open access publication PDF: ISBN 978-3-8394-5419-0

 Lise Jaillant (ed.) Archives, Access and Artificial Intelligence Working with Born-Digital and Digitized Archival Collections

2022, 224 p., pb., col. ill. 45,00 € (DE), 978-3-8376-5584-1 E-Book: available as free open access publication PDF: ISBN 978-3-8394-5584-5

All print, e-book and open access versions of the titles in our list are available in the online shop www.bielefeld-university-press.de