# Historical journey in a linguistic archipelago

Descriptive concepts and case studies

Edited by Émilie Aussant Jean-Michel Fortis

History and Philosophy of the Language Sciences 3

### History and Philosophy of the Language Sciences

Editor: James McElvenny

In this series:


# Historical journey in a linguistic archipelago

Descriptive concepts and case studies

Edited by

Émilie Aussant

Jean-Michel Fortis

Aussant, Émilie & Jean-Michel Fortis (eds.). 2020. *Historical journey in a linguistic archipelago*: *Descriptive concepts and case studies* (History and Philosophy of the Language Sciences 3). Berlin: Language Science Press.

This title can be downloaded at: http://langsci-press.org/catalog/book/286 © 2020, the authors Published under the Creative Commons Attribution 4.0 Licence (CC BY 4.0): http://creativecommons.org/licenses/by/4.0/ ISBN: 978-3-96110-292-1 (Digital) 978-3-96110-293-8 (Hardcover)

ISSN (print): 2629-1711 ISSN (electronic): 2629-172X DOI: 10.5281/zenodo.4269397 Source code available from www.github.com/langsci/286 Collaborative reading: paperhive.org/documents/remote?type=langsci&id=286

Cover and concept of design: Ulrike Harbort Typesetting: Katharina Berking, Felix Kopecky, Sebastian Nordhoff Proofreading: Adrien Barbaresi, Agnes Kim, Amir Ghorbanpour, Annika Schiefner, Aviva Shimelman, Christopher Straughn, Felipe Neis Araujo, Francesco Screti, Jeroen van de Weijer, Jessica Brown, Lachlan Mackenzie, Linda Leembruggen, Plinio Barbosa, Sebastian Nordhoff, Selçuk Eryatmaz, Yvonne Treis Fonts: Libertinus, Arimo, DejaVu Sans Mono, Source Han Serif Typesetting software: XƎLATEX

Language Science Press xHain Grünberger Str. 16 10243 Berlin, Germany langsci-press.org

Storage and cataloguing done by FU Berlin

## **Contents**


### Contents


## **Preface**

This volume brings together 12 papers presented at the 14th International Conference on the History of the Language Sciences (ICHoLS XIV), held at Université Sorbonne Nouvelle in Paris, August 28 to September 1, 2017.

The present book is not the only publication presenting a selection of papers delivered at ICHoLS XIV. A second collection has been assembled by the same editors in a companion volume published by John Benjamins and entitled *History of linguistics 2017*. An attempt has been made to strike a balance between the two volumes, in terms of time range and thematic diversity.

In addition to these two volumes reserved for independent contributions, the proceedings of seven of the nine thematic workshops of *ICHoLS XIV* have been – or will be – published in books or in special issues of journals:


### Preface

*da Ciência da Linguagem, das Teorias Linguísticas, da construção do conhecimento sobre a(s) língua(s)*: https://periodicos.ufsm.br/fragmentum/issue/ view/1438/showToc

• "Cross-readings: Andalusian grammarians reading Oriental grammarians, Oriental grammarians reading Andalusian grammarians", published in *Histoire Épistémologie Langage* 40.2 (2018), edited by Francesco Binaghi: https: //www.hel-journal.org/fr/articles/hel/abs/2018/02/contents/contents.html

> The editors, Paris, April 2020

## **Acknowledgments**

Our deepest thanks to the team of Language Science Press for their help, patience and meticulous work. Special credit is due to James McElvenny and Sebastian Nordhoff, who have very kindly helped us through all the steps for the preparation of the manuscript.

## **Introduction**

### Émilie Aussant

Histoire des Théories Linguistiques (CNRS/Université de Paris)

### Jean-Michel Fortis

Histoire des Théories Linguistiques (CNRS/Université de Paris)

The present volume is divided thematically into two parts: (I) *Metalinguistic concepts and representations*, and (II) *Fields, authors and disciplinary commitments*. Within each part, papers are arranged chronologically. There is of course a certain artificiality in this two-part division, given the variety of the topics and perspectives illustrated in the papers. Let us say that the first part is more concerned with descriptive concepts and the second with case studies involving specific fields and authors.

The first part begins with a paper by Mazhuga, who sets out to explore the origin of the technical use of the terms *accidere/accidens* in Latin grammar. Is this origin Greek? Is it to be sought in the preceding "technicization" of what would be the Greek counterparts of *accidens*, namely *συμβεβηκός* and *παρεπόμενον* in their grammatical usage? Mazhuga thinks not, and shows that the Latin and Greek terminologies have followed their own idiosyncratic courses of development. This leads him to discuss the Greek terms in some detail, before coming to the ultimate goal of the paper, which is to clarify the Latin terms. In this respect, it is of particular interest to mention the hypothesis he puts forth: *accidentia*, he claims, was used in rhetoric to designate the qualities pertaining to a particular case (causa). It is from there that it gained its technical use in grammar.

Fortis' contribution (delivered as a plenary lecture, hence its greater length) does not bear on a linguistic category per se, but rather on a philosophically loaded perspective which has been defended throughout the history of Western linguistics. This perspective corresponds to a family of descriptions which can be conveniently brought together under the label of "localism", a name that first

became established only during the 19th century, although the basic idea behind it is much older and can be traced back to Aristotle. Localism traditionally argues for the primacy of spatial relations in the semantics of particular categories, typically prepositions and cases. Remarkably, ideas reviving this tradition surged again in the second half of the 20th century, for reasons explored in this paper. After a historical retrospective and a presentation of the recent context, two main neolocalist lines of investigation are examined: localist descriptions in lexicalgrammatical semantics, and localist theories of thematic roles.

The next paper cross-cuts and complements the preceding one. In her study, Chalozin-Dovrat discusses the status of space in linguistic discourse, scrutinizing three theories she samples in view of the important role they confer on spatializing descriptions of linguistic time. The theories in question are due to Beauzée, Guillaume and the strand she identifies as the "conceptual school of cognitive linguistics" (comprising here Lakoff, Langacker, Talmy and Traugott). She suggests that spatializing descriptions, aside from their potential in modelling linguistic facts, also have a function of legitimation: by resorting to a fundamental category of Western science, space, they contribute to providing linguistics with scientific credentials. This strategy is what is designated as "scientification" in the paper.

It is a fortunate circumstance that the next chapter by Mazziotta provides yet another way to envisage the relation of linguistics to spatialization, in the guise, this time, of diagrammatic representations of grammatical structure. These appear, as the author claims, at a juncture when linguistics increasingly turns from a word- to a syntax-centered perspective. The focus is on early diagrams, especially those proposed by Clark, Reed and Kellogg around the mid-19th century. As can be expected, diagrams reflect pre-conceived grammatical analyses. But the graphical expedients employed to capture grammatical structure have their own logic, which may interfere with that of the analysis. These expedients are therefore not theoretically neutral, as the author shows, for instance, in the treatment of subordination.

If the subject of spatial relations is of timely importance, the same can be said of the question addressed in the following chapter: the origin of the term "polysemy", a designation which sums up a host of issues actively debated in recent times as a result of the renewed prominence of lexical semantics in linguistic scholarship. In his study, Courbon shows that, contrary to common belief, the term was not coined by Bréal but by Joseph Halévy (an orientalist), who applied it to cuneiform signs. The wide acceptance of the term and the fact that Bréal was credited with its invention show two things: that the context was favorable to the newly created *Sémantique* (in part through lexicography), and that Bréal

was in a better position to prescribe its use. Here, social-scientific factors come into play: we may say that Bréal's advantage lay in his institutional position and the important role he played in the rise of modern general linguistics in France.

The relevance of polysemy for contemporary semantics and even syntax (if we think of construction grammar) is more than matched by the topic of the next paper. In his survey of the state of the art, Christy touches on the status of prefabs, idioms, phraseology etc. ("fixed expressions") in today's linguistics, with an eye toward Saussure and Bally. Again, we see this issue's rise to prominence in the past decades, partly, we may say, to fill a gap left by generative linguistics, and partly because the usage-based approach inherited from diachronic linguistics joined forces with cognitive and functional linguistics. The contemporary landscape is the focus of Christy's paper, which maps recent research and gives us an overview of the criterial features of fixed expressions, of the roles and functions assigned to them by different authors, and of the new tools available to data collection and analysis. As transpires from Christy's discussion, the importance of fixed expressions culminates in usage-based approaches. In these approaches, formulaic structures form the basis for creatively assembled sequences of any degree of regularity above non-productive, non-compositional idioms (which in this respect are a limiting case).

In the second part of this volume, we have brought together case studies which bear on a specific field or author. In all cases, these fields and authors are outlined against the backdrop of overarching views or broader concerns related to sociolinguistic, philosophical, ideological, or pedagogical issues, or to the circumscription of the disciplinary boundaries of linguistics.

In the opening chapter by Li, the overarching view in question is of a sociohistorical nature. In the perspective advocated in this paper, data from a standard language are no longer prioritized and must give way to productions obtained from situations of communicative immediacy. An approach of this kind, shows Li, is especially fitting for the subject she considers, that is, Chinese Pidgin English. Since the socio-historical angle lays much importance on concrete speech situations, Li takes due account of the speech act participants, of the variety of their social roles, and highlights the ensuing variation of their linguistic productions. In addition, she offers a descriptive analysis of the Chinese Pidgin English preposition *long*, and argues that syntactic-semantic variations in its usage are bound up with varying degrees of acculturation on the part of speakers. Of note here is an interplay with the Cantonese substrate: meanings of *long* more specific to the Cantonese counterpart tend to inherit their syntactic behavior from Cantonese too.

The following chapter is a contribution to the history of school grammar and the evolution of language pedagogy. Lhomond, who is the grammarian chosen by Piron in her case study, plays a very special part in this history. For Lhomond's *Elémens de grammaire latine*, in addition to having achieved a huge editorial success, have been held up as the first truly "modern" Latin textbook, in the sense that it was to a large extent adapted to its audience of French-speaking learners. Since Latin grammaticography had always been essential for the description of vernacular languages, and since, in return, Lhomond's French grammar was intended as a kind of propaedeutic introduction to Latin, it seemed all the more apposite to have a closer look at this reciprocal relationship. What Piron demonstrates is that, while still of course dependent on the Latin model, Lhomond decidedly takes a "delatinizing" turn. As an example, although Latin cases are treated as functionally equivalent to French *de* and *à* when government is at issue, cases are no longer employed for presenting the paradigm of French nouns. To be noted too, are such innovations as the separate treatment of adjectives as a part of speech distinct from substantives, tabular expositions of paradigms, and a French propaedeutic taking its departure from the word, that is, dealing with morphology and spelling, agreement and government as features of the word.

School grammar is of course endowed with a social role, that of codifying, standardizing and transmitting. The stakes of linguistic description are very high too when it is in addition subordinated to an enterprise of acculturation like religious conversion. This does not mean that missionary linguistics is necessarily more prejudiced or ill-intentioned than supposedly cool science. In his chapter on German missionaries and scholars in Australia, Moore shows precisely this: missionaries trained in philology were more intimately in touch with the local cultures and languages, and appeared to be more empathetic, than anthropologists and linguists whose secular religions were positivism and evolutionism. Thus, Strehlow, a missionary, professed an "idiographic" orientation, defiant of the sweeping claims of evolutionary and physical anthropology. On the other hand, the matter-of-fact approach of anthropologists and linguists could be conducive to a distant vision which, by its lack of empathy, ran the risk of being more susceptible to prejudices, value judgment and collaboration with colonial power. This was indeed the case for the strand of research which Moore characterizes as "antihumanist".

The chapter by Bergounioux is concerned not so much with the latter kind of rival epistemic orientations as with the constitution of a field, the study of regional dialects in France. However, broader epistemic issues are not absent, for the field was largely shaped by new methods that differentiated themselves both

from philology and historical linguistics, and thus favored the emergence of linguistic subdisciplines. These methods were promoted by two important figures of French linguistics, namely, Gilliéron, for linguistic geography, and Rousselot, for experimental phonetics. Both scholars felt the need for a journal that would enable them to deploy their investigation in the field of dialectology, a project which was concretized in the *Revue des Patois Gallo-Romans*. Each brought to the task his own technique and interests, and his own network of local researchers. However, because Rousselot's approach was in the direction of extremely finegrained phonetic analysis — and as such rather unwieldy — and because descriptions were narrowly focused on very local varieties, the net result tended to be an "atomization" (Bergounioux' word) of the dialectal reality. Moreover, dialectology was at variance with the linguistic unification then actively implemented by the French republic and with the dim view that was taken of so-called *patois* in a large part of French society. In the end, facing a relative indifference, the journal and other publications with a similar purview were unable to perpetuate themselves.

In the case of linguistic geography and experimental phonetics, new methods help form subdisciplines and bring into focus specific aspects of linguistic and social reality. In the chapter by Frigeni, we see that a discipline already in existence and with a well-established methodology may be partly reshaped by a new perspective. In historical semantics, Meillet, it is recalled, laid a new emphasis on social and cultural matters. But was this perspective idiosyncratic enough to be regarded as a specific way of practicing historical semantics? Frigeni's answer is positive, and she argues that Meillet's perspective lived on in Benveniste's work, even if he made no explicit reference to Meillet in this connection. Evidence for a legacy, or at least a common orientation, is furnished by her case study, in which she examines Benveniste's semantic analysis of Indo-Iranian *Mit(h)ra* in the light of Meillet's initial attempt. Both scholars, it is shown, conduct their etymological inquiry by expanding it into an inter-cultural comparison (between the Indian and the Iranian contexts) in which social reality takes on an important role. This concern for a wider cultural context leads them both to hypothesize for *Mit(h)ra* a meaning with a social import, related to the notion of social pact or contract for Meillet, and to a function protective of the community for Benveniste.

The following chapter on Trần Đức Thảo by D'Alonzo illustrates yet another way of counteracting linguistic abstraction by bringing it in closer contact with the social reality of semiotic systems. In the eyes of the phenomenologist and materialist philosopher Trần Đức Thảo, the social and cognitive determinants of signs furnish the material conditions of human semiotic life. These conditions, he claims, undermine the view of sign-systems as made up of arbitrary elements with a differential value, in other words, the view commonly attributed to Saussure. Trần Đức Thảo's insistence on the motivation of signs is associated with the consideration of their antepredicative genesis (in particular in the form of indicative gestures). This would not be highly original if Husserlian phenomenology, to which this antepredicative layer refers back, were not encompassed in the same critique as Saussurian structuralism, and both characterized as idealistic. In other words, Trần Đức Thảo's critique reflects an original perspective which enrolls Marxism and genetic psychology into a reorientation of phenomenology, and in turn marshals this theoretical complex against Saussure.

## **Part I**

## **Metalinguistic concepts and representations**

## **Chapter 1**

## **Le terme** *accidentia* **chez les grammairiens romains**

Vladimir I. Mazhuga

Saint Petersburg Institute of History, RAS

This chapter deals with the custom of Latin grammarians of introducing the attributes of the parts of speech with the verb *accident* (3rd person plural), as well as with the use of the nominalized participle plural *accidentia*. Contrary to the most widespread view, this chapter demonstrates an early development of the usage in question from Greek rhetorical doctrines of the 1st century BC. The teachings and writings of Theodore of Gadara, Cornelius Celsus and Pliny the Elder had a crucial impact on the development of this usage. One does not need to search for its origin in the hypothetical grammatical vocabulary of the Stoics or in treatises on Greek grammar. Similar customs also appeared in response to the new needs of school teaching in Greek grammatical doctrine, although somewhat later. The Greek analogy to the Latin terms were the plural noun *τὰ παρεπόμενα* and the verb *παρέπομαι*, which were borrowed from the sophistic dialectic at the time of the flourishing Second Sophistic. Apollonius Dyscolus contributed much to the initial introduction of the terms in question into school textbooks on Greek grammar.

La présente contribution pourrait prendre comme point de départ l'article de Maria Teresa Vitale "Per una terminologia grammaticale : *parepómena* – *accidentia*" (Vitale 1982), tant son auteure paraît avoir magistralement traité le sujet que nous allons aborder. Elle nous semble toutefois avoir élaboré des hypothèses sujettes à caution, plutôt que d'examiner scrupuleusement le matériau disponible. Ces hypothèses se réduisent en substance à la thèse, formulée jadis par Karl Barwick, selon laquelle le terme *παρεπόμενα* des grammairiens grecs et le terme *accidentia* des grammairiens latins rendaient compte tous deux de la notion de *συμβεβηκότα*, qui joua un rôle fondamental dans la philosophie d'Aristote et, dans une moindre

Vladimir I. Mazhuga. 2020. Le terme *accidentia* chez les grammairiens romains. In Émilie Aussant & Jean-Michel Fortis (éd.), *Historical journey in a linguistic archipelago : Descriptive concepts and case studies*, 3-14. Berlin : Language Science Press. DOI : 10.5281/zenodo.4269407

### Vladimir I. Mazhuga

mesure, dans celle des stoïciens (Barwick 1922 : 107–108 ; Vitale 1982 : 212–214). Tenant pour une vérité acquise que les grammairiens latins suivaient assez fidèlement les maîtres grecs, Vitale s'est vue obligée de démontrer d'emblée que dès l'époque d'Aristote le terme *συμβεβηκότα* pouvait désigner aussi bien les propriétés stables distinctives des êtres que les particularités occasionnelles, celles auxquelles Aristote appliquait habituellement les termes *τὸ συμβεβηκός* ou *τὰ συμβεβηκότα*.

Dans sa tentative de réinterprétation de la notion de *συμβεβηκός*, Vitale s'est concentrée avant tout sur la doctrine aristotélicienne de l'*essence* de l'*être* et des attributs de celui-ci. À l'appui de l'idée selon laquelle *συμβεβηκός*, en tant qu'attribut, pouvait exprimer chez Aristote aussi bien l'essence du sujet que ses propriétés occasionnelles, elle évoque une phrase des *Seconds analytiques* :

Ὑπόκειται δὲ ἓν καθ᾿ ἑνὸς κατηγορεῖσθαι, αὐτὰ δὲ αὐτῶν, ὅσα μὴ τί ἐστι, μὴ κατηγορεῖσθαι. συμβεβηκότα γάρ ἐστι πάντα, ἀλλὰ *τὰ μὲν καθ᾿ αὑτά, τὰ δὲ καθ᾿ ἕτερον τρόπον·* (*Arist. Post. Anal.* : 83b.19–83b.21)

We have now established that in predication one is asserted of one subject, and that predicates (except those which denote essence) are not predicated of one another. They are all attributes, *some per se and others in a different sense*. (trad. par Tredennick dans *Arist. Post. Anal.*)

Il faut entendre ici *τὰ καθ᾿ αὑτά* évidemment selon son sens direct, c'est-à-dire en tant qu'attributs comme tels, et il est inutile, en revanche, d'y chercher le sens d'une qualité distinctive essentielle. Aristote désignait celle-ci par le terme spécifique *τὸ ἴδιον*, qu'il opposait le plus souvent à *τὸ συμβεβηκός*, bien qu'il concédât parfois à ce dernier le sens d'un *propre relatif et temporaire* (*πρός τι ἴδιον* ; cf. *Arist. Soph. El.* : 102b22). On peut supposer ce même sens dans l'expression *τὰ καθ᾿ ἕτερον τρόπον*. Vitale déduit néanmoins de la phrase citée une formulation tout à fait inattendue : "I due *συμβεβηκότα*, distinti in καθ' αὑτά = essenziali e καθ' ἕτερον τρόπον – accidentali, erano già presenti, in un contesto più strettamente logico, negli *Analytica Posteriora*" (Vitale 1982 : 202).

Traitant des notions de *συμβεβηκός* et d'*accidens*, il semblait naturel de borner l'analyse à la doctrine philosophique de l'*être* individuel. Aristote ne cessait de répéter que la catégorie de *συμβεβηκός* ne s'applique qu'à l'individuel. À partir de la logique des sophistes, la notion de *τὰ παρεπόμενα* ne s'appliquait cependant qu'aux qualités propres à tout un genre. Vitale passe sous silence les idées d'Aristote sur ce sujet, exprimées de manière éloquente dans ses *Réfutations sophistiques.* En dévoilant les fausses prémisses des conclusions dont les sophistes usaient dans la réfutation de leurs adversaires, il s'attaque, entre autres,

### 1 Le terme *accidentia* chez les grammairiens romains

à la fausse déduction à partir de ce qui accompagne généralement une classe de choses (*παρεπόμενον*), où une caractéristique stable est remplacée insidieusement par une propriété occasionnelle (*συμβεβηκός*) :

(…) *τὸ μὲν συμβεβηκὸς ἔστιν ἐφ' ἑνὸς μόνου λαβεῖν, (...) τὸ δὲ παρεπόμενον ἀεὶ ἐν πλείοσιν* τὰ γὰρ ἑνὶ καὶ ταὐτῷ ταὐτὰ καὶ ἀλλήλοις ἀξιοῦμεν εἶναι ταὐτά· διὸ γίνεται παρὰ τὸ ἐπόμενον ἔλεγχος (*Arist. Soph. El.* : 168b.29–168b.35).

(…) *You may secure an admission of the accident in the case of one thing only,* (...) *whereas the consequent always involves more than one thing* ; for we claim that things that are the same as one and the same thing are also the same as one another, and this is the ground of a refutation dependent on the consequent. (trad. par Barnes 1984 : 285)

À la différence de la notion grecque de *συμβεβηκότα* la notion latine apparentée d'*accidentia* a pris progressivement chez les grammairiens romains une signification particulière qui allait bien au-delà des bornes de l'*être* individuel. Mais avant d'examiner ce sujet, essayons de poursuivre le raisonnement de Vitale en ce qui concerne les précédents qu'elle a cru repérer dans la grammaire grecque. À côté du terme grec *συμβεβηκός*, elle tente de reconstituer l'histoire du terme *τὰ παρεπόμενα* traité comme équivalent de *τὰ συμβεβηκότα*. Vitale ne doutait pas que le terme *παρεπόμενα* fût établi dans la grammaire grecque dès le temps d'Aristophane de Byzance (vers 257–180 av. J.-C.) et de son élève alexandrin Aristarque (ca 216 – ca 144 av. J.-C.), c'est-à-dire dès la fin du III<sup>e</sup> siècle av. J.-C. Mais les arguments probants lui font défaut. Elle se contente de renvoyer le lecteur aux exemples contenus dans la *Τέχνη* attribuée à Denys le Thrace, grammairien célèbre du début du Ier siècle av. J.-C., en négligeant totalement les recherches de Vincenzo Di Benedetto, qui montrait bien, dès les années cinquante, le caractère tardif du corps principal de ce traité scolaire qui semble avoir été composé dans l'Antiquité tardive (cf. Di Benedetto 1958 ; 1959). Néanmoins, Vitale se voit tenue d'expliquer l'absence ultérieure du terme *τὰ παρεπόμενα* dans les textes grammaticaux avant Apollonius Dyscole, c'est-à-dire avant le deuxième tiers du IIe siècle ap. J.-C., ce en raison de l'emploi prétendument synonymique du terme latin *accidentia* chez les grammairiens romains à partir du Ier siècle av. J.-C.

L'absence de témoignages sur l'emploi du terme *τὰ παρεπόμενα* chez les grammairiens grecs est à relier, selon Vitale, à la prépondérance du vocabulaire stoïcien dans la grammaire grecque de la période car, selon une supposition de Karl Barwick, le terme *συμβεβηκότα* y jouait le rôle qui incombait à *παρεπόμενα* chez les anciens grammairiens alexandrins (Barwick 1922 : 97, 107 ; 1957 : 48). Barwick a

### Vladimir I. Mazhuga

élaboré sa doctrine en s'appuyant sur des raisonnements que l'on pourrait qualifier d'arbitraires. C'est dans les interprétations byzantines des faits de grammaire qu'il a trouvé l'unique exemple d'assimilation du terme *παρεπόμενον* à celui de *συμβεβηκός* : *παρεπόμενον δὲ ἐστι συμβεγηκός* (*Sch. Dion. Thr.* 217.23, cf. Barwick 1957 : 47). Sans prendre en compte la différenciation fondamentale que les sophistes anciens et Aristote établissaient entre la notion de *παρεπόμενον* et celle de *συμβεβηκός*, et sans essayer de l'atténuer en quelque sorte, comme Vitale a tenté, mais sans succès, de le faire, Barwick affirmait tout simplement que la signification de quelque chose d'occasionnel convenait bien aux attributs des parties du discours, définis par lui comme *zufällige Merkmale* (Barwick 1957 : 48).

Les chercheurs qui ont travaillé après Vitale ne s'intéressaient plus au rapport sémantique entre les termes *παρεπόμενον* et *συμβεβηκός*. Cependant, comme nous le verrons, certains faits les faisaient tenir à l'idée de Barwick, selon laquelle les stoïciens avaient coutume de désigner les propriétés des parties du discours par le terme *συμβεβηκότα*. On se demandait parfois encore si l'emploi du terme *παρεπόμενον* précédait chronologiquement l'emploi de *συμβεβηκός* chez les grammairiens grecs (cf. toutefois de Jonge 2008 : 154–155), mais on ne discutait point la parenté sémantique présumée de ces deux termes, et personne n'a mis en doute jusqu'ici l'essentiel des raisonnements de Vitale.

Personne n'a mené non plus l'analyse systématique des témoignages disponibles. À y regarder de près, on ne peut parler que d'une analogie dans les emplois du terme *τὰ παρεπόμενα*, chez les grammairiens grecs, et du terme *accidentia*, chez les grammairiens latins. Chacun des deux termes a eu son histoire propre et leur signification exacte témoigne d'une origine différente.

Pour terminer notre analyse du terme *τὰ παρεπόμενα*, examinons la thèse de Vitale selon laquelle ce terme était bien établi dans la grammaire grecque à l'époque d'Apollonius Dyscole (Vitale 1982 : 212). L'œuvre d'Apollonius Dyscole atteste pourtant plutôt des tout premiers débuts d'un nouvel usage terminologique.

Le traité d'Apollonius *de Pronomine* est chronologiquement le premier dans la série de ses traités grammaticaux qui nous sont parvenus, et il accuse une forte dépendance vis-à-vis des prédécesseurs d'Apollonius, surtout du grammairien du I er siècle av. J.-C. Tryphon. Dans ce traité, Apollonius emploie habituellement le verbe *παρακολουθέω* au lieu de son synonyme *παρέπομαι*, comme l'avaient fait tant d'autres auteurs de la période précédente (19 exemples apolloniens, y compris le substantif *ἡ παρακολούθησις* de même racine, contre trois exemples seulement du verbe *παρέπομαι* et un exemple du participe *παρεπόμενον*). Dans son œuvre fondamentale *Syntaxis*, le vocabulaire d'Apollonius change d'une manière impressionnante. L'auteur montre ici une prédilection pour les participes

### 1 Le terme *accidentia* chez les grammairiens romains

substantivés *τὰ παρεπόμενα* (22 exemples) et *τὸ παρεπόμενον* (13 exemples), un peu moins pour de simples participes (21 exemples) ; nous trouvons en outre dix cas où le verbe *παρέπομαι* est employé, ce qui donne au total 66 exemples, contre 22 exemples avec le verbe *παρακολουθέω* et trois substantifs de même racine. Le substantif pluriel *τὰ παρεπόμενα* est donc ici le plus fréquent, et l'on peut le concevoir comme un vrai terme technique.

Il faut signaler toutefois l'emploi prépondérant de ce terme dans le large domaine du raisonnement logique. Apollonius l'applique aux différentes propriétés sémantiques et morphologiques des parties du discours, aussi bien qu'aux formes particulières de certains mots. Mais dans sa *Syntaxis*, il n'emploie ce verbe et les mots de la même racine que sept fois seulement appliqués aux attributs des parties du discours, en leur donnant un sens général (*Ap. Dysc. Synt.* : 320.7 *παρέπομαι* ; 118.13 ; 324.3 *παρεπόμενον* ; 75.9 ; 145.9 ; 267.1–267.2 *παρεπόμενα*). Une fois seulement, Apollonius use du verbe *παρέπομαι* pour indiquer les propriétés distinctives concrètes des parties du discours :

Καθὼς ἔφαμεν, ἔστιν γενικωτάτη ἡ τῶν ἀπαρεμφάτων ἔγκλισις, ἀναγκαίως λείπουσα τοῖς προδιαπορηθεῖσι, *〈τοῖς προσώποις καὶ〉 τῷ παρεπομένῳ ἀριθμῷ, ὃς οὐ φύσει παρέπεται τῷ ῥήματι* παρακολούθημα δὲ γίνεται προσώπων τῶν μετειληφότων τοῦ πράγματος (*Ap. Dysc. Synt.* : 324.10–325.1).

Comme nous le disions, le mode infinitif est le plus général, puisque lui font nécessairement défaut les accidents dont on a vu plus haut qu'ils faisaient problème : *〈la personne et〉 le nombre – ce dernier n'étant pas par nature un accident du verbe,* mais une dépendance des personnes qui prennent part à l'acte. (trad. par Lallot 1997 : 327)

On voit bien que *παρεπόμενον* n'est pas encore chez Apollonius un terme technique du lexique grammatical, mais il est évident qu'à partir d'un certain moment ce grammairien éminent montre une prédilection pour ce terme et ses dérivés. On peut expliquer aisément cette prédilection par l'épanouissement de la Seconde Sophistique, phénomène culturel dont est fortement empreint l'enseignement scolaire du temps d'Apollonius, et qui a fait renaître entre autres la dialectique des anciens sophistes.

C'est aussi dans l'histoire de l'école et surtout dans l'enseignement de la rhétorique, étroitement lié aux études de grammaire, qu'il faut chercher les raisons de l'emploi du terme *accidentia* et surtout du verbe *accido* (sous la forme de la 3<sup>e</sup> personne du pluriel *accidunt*) chez les grammairiens latins. Mais un autre phénomène, plus ancien, a déterminé ici le choix du terme. En laissant maintenant de côté les raisonnements de Vitale, examinons de plus près les observations

### Vladimir I. Mazhuga

des savants qui ont essayé de retracer l'histoire des termes grecs *τὸ συμβεβηκός* – *τὰ συμβεβηκότα* et leurs équivalents latins *accidens* – *accidentia.* Comme nous allons le voir, tous les chercheurs partaient des exemples que l'on trouve chez Denys d'Halicarnasse (à Rome en 30/29 jusqu'à 7 env. av. J.-C.). En voici un parmi d'autres. En parlant de la maîtrise exceptionnelle des règles de la rhétorique chez Démosthène, Denys traite en général de l'apprentissage régulier aussi bien des éléments les plus simples de l'art rhétorique que de ceux de la prosodie et des catégories grammaticales.

κρατήσαντες δὲ τούτων, τὰ τοῦ λόγου μόρια· ὀνόματα λέγω καὶ ῥήματα καὶ συνδέσμους· καὶ τὰ *συμβεβηκότα* τούτοις, συστολάς, ἐκτάσεις· ὀξύτητας, βαρύτητας· γένη, πτώσεις, ἀριθμούς, ἐγκλίσεις, τὰ ἄλλα παραπλήσια τούτοις μυρία ὄντα ("Dion. Halic.": 52, 242.20–242.23).

Quand nous possédons parfaitement cela, nous étudions les parties du langage, je veux dire les noms, les verbes, les mots de liaison, et *tous leurs accidents*, abrégement, allongement, accent aigu, accent grave, genre, cas, nombre, flexion, et mille autres choses du même genre. (trad. par Aujac 1988 : 151)

Le terme *τὰ συμβεβηκότα* est appliqué ici à la fois aux attributs des parties du discours et aux particularités de la prosodie. Il s'agit de l'usage pratique et parfois personnel de différents éléments du langage. On peut penser en outre aisément que le terme *τὰ συμβεβηκότα* était employé précisément dans le sens que lui prêtait Barwick de quelque chose d'individuel et d'occasionnel. D'autres exemples de l'emploi semblable du terme *τὰ συμβεβηκότα* chez Denys renforcent encore cette impression (*Amm*. II 1–12, 421.5–432.13, spec. 421.17 ; *Comp*. 25, 131.18–132.8 ; cf. de Jonge 2008 : 147–149). Vu l'influence massive de la théorie stoïcienne du langage sur la théorie grammaticale à ses débuts, Barwick a essayé le premier de faire remonter l'emploi du terme en question au vocabulaire des stoïciens (Barwick 1957 : 48, cf. Barwick 1922 : 107–108). Mais il s'agit chez Denys généralement d'une combinaison particulière des caractéristiques grammaticales et phonétiques, tandis que considérés en soi, les attributs des parties du discours devraient évoquer principalement une idée des propriétés permanentes d'une espèce. Denys semble avoir appliqué le terme *τὰ συμβεβηκότα* plutôt à tout un répertoire de procédés stylistiques qu'aux attributs des parties du discours comme tels.

Dans les citations des œuvres des stoïciens chez les auteurs du Ier jusqu'au début du III<sup>e</sup> siècle ap. J.-C., tels Philon d'Alexandrie, Plutarque, Galien, Clément

### 1 Le terme *accidentia* chez les grammairiens romains

d'Alexandrie, le terme *τὸ συμβεβηκός* désigne cependant soit quelque chose d'occasionnel (*FDS* : 515, 746), soit un événement, distinct d'un corps matériel en tant que sa cause (*FDS* : 762, 695) (cf. de Jonge 2008 : 153). La signification du terme est assez large chez les stoïciens, mais elle ne diffère pas essentiellement de celle que le terme a chez Aristote. Tout en leur donnant une interprétation nouvelle, les stoïciens gardaient néanmoins le système primaire des catégories aristotéliciennes, et il est difficile d'imaginer qu'ils pussent appliquer le terme *τὰ συμβεβηκότα* aux propriétés des parties du discours.

Tout en partageant dans une certaine mesure les critiques émises par des savants éminents comme Jan Pinborg et Dirk M. Schenkeveld, de Jonge adhère, bien qu'avec des précautions, à l'hypothèse de Barwick sur les origines stoïciennes de l'emploi du terme *τὰ συμβεβηκότα* chez Denys d'Halicarnasse (de Jonge 2008 : 151–154). À la suite de Richard Janko, il indique en outre l'emploi similaire du terme en question dans le traité du philosophe épicurien Philodème (vers 111– 40/35 av. J.-C.) *De poematibus.* Les deux savants mentionnent cet usage aussi bien dans le passage où est cité Pausimachus de Milet, qui développait la doctrine selon laquelle l'euphonie constituait en elle-même le trait distinctif (*τὸ ἴδιον*) de la poésie, que dans celui où Philodème réfute le postulat restrictif défendu par Pausimachus (Janko 2000 : 182, 268–269 (fr. 74.1–74.6), cf. 300–301 (fr. 94.15–94.21), *Philod. Poem. Poem.* 22–25 (fr. 18.25–19.9) ; de Jonge 2008 : 155). C. de Jonge n'a cité toutefois les exemples trouvés chez Philodème que pour montrer une fois de plus l'ancienneté de l'usage du terme *τὰ συμβεβηκότα* dans le domaine des parties du discours.

Mais en fait, le savant hollandais a relevé un phénomène particulier qui est digne d'une analyse spéciale. Le terme *τὰ συμβεβηκότα* a été, de toute évidence, ancré dans le vocabulaire des philosophes épicuriens en tant que désignation commune des propriétés aussi bien permanentes qu'occasionnelles (cf. "Epic. Herod." : 40, 6.14). Tout comme dans le cas de l'emploi synonymique du terme plus usuel *τὰ παρακολουθοῦντα*, on peut cependant observer chez Épicure une tendance à désigner par ce terme surtout les propriétés stables – mais il faut noter l'habitude qu'a cet auteur de souligner expressément le caractère de stabilité de telles propriétés, par l'ajout d'une détermination spéciale indiquant leur essence physique constante : *τὴν φύσιν ἀίδιον* (voir "Epic. Herod." : 71, 24.5–24.7 ; cf. 68– 69, 22.13–23.9).´

On sait bien qu'Épicure (vers 342–270 av. J.-C.) négligeait les doctrines dialectiques des philosophes grecs, pour proposer une épistémologie originale fondée sur l'analyse des impressions. La terminologie d'Épicure et de ses adeptes n'avait pas grande chance de trouver place dans le lexique linguistique de l'An-

### Vladimir I. Mazhuga

tiquité grecque et latine – à une exception près. Les chercheurs ont bien mis en évidence que les mêmes théories de l'euphonie intéressaient les stoïciens et les épicuriens, qu'il s'agissait là d'un domaine commun de leur réflexion (cf. Janko 2000 : 181–182, 188–189 ; Campbell 2002 : 105–109). Même si Denys d'Halicarnasse parle lui-même de ses emprunts aux théories linguistiques des stoïciens, il est fort probable que les vues de Denys sur la prosodie ainsi que celles sur les particularités des parties du discours reprennent plutôt les idées des épicuriens, telles que nous les connaissons à travers l'œuvre de Philodème.

Barwick a signalé l'emploi du terme *τὰ συμβεβηκότα* appliqué aux propriétés des parties du discours chez l'auteur qu'on peut identifier comme le grammairen Polybe de Sardes (II<sup>e</sup> siècle ap. J.-C. ; "Polib. Barb. et sol." : 286.13 ; Barwick 1922 : 97 ; cf. Jones 1996). Dans le traité sur le barbarisme et le solécisme, cet auteur développe l'idée que le solécisme naît du remplacement (*κατὰ ἐναλλαγὴν*) erroné d'une partie du discours par une autre, et surtout de l'interversion entre attributs de parties du discours : *ἢ ὅταν τὰ συμβεβηκότα τοῖς τοῦ λόγου μέρεσιν εἰς ἄλληλα ἐναλλάσσηται*. Cette pensée est si proche des idées de Denys d'Halicarnasse sur l'usage souple de différentes parties du discours et de leurs formes grammaticales et acoustiques, que leur fond commun est hors de doute.

Il faut chercher ailleurs les origines de l'emploi du verbe latin au pluriel *accidunt* et du participe *accidentia* comme termes de la grammaire latine. On doit prendre en compte aussi le cas d'un processus long des transformations sémantiques de ces vocables dans le domaine de la rhétorique et dans celui de la grammaire latine, autrement dit, d'un processus à peu près parallèle à l'élaboration des termes de la grammaire grecque *παρέπεται* et *παρεπόμενα*, que nous avons esquissé ci-dessus.

Signalons quelques faits dans l'histoire de la doctrine rhétorique au début du Principat, qui pouvaient influencer d'une manière décisive la pensée des grammairiens latins. Il faut prendre en considération surtout la doctrine de Théodore de Gadara sur la répartition de la cause civile selon les questions principales qu'on y posait. Comme l'on sait, Théodore enseignait la rhétorique à Rome au futur empereur Tibère dans les années 33–32 av. J.-C. Le manuel de rhétorique qu'il a composé a été traduit en latin, comme l'atteste Quintilien (*Inst. or.* II 15, 21). Or, suivant toujours le témoignage de Quintilien (III 6, 36), parmi les nombreuses questions qui pouvaient être posées dans la cause civile, Théodore mettait en avant deux questions générales, à savoir, « le fait existe-t-il ? » et, « le fait étant certain, quelles en sont les particularités concomitantes ? » : « *idem Theodorus, qui de eo an sit et de accidentibus ei, quod esse constat, id est περὶ οὐσίας καὶ συμβεβηκότων* **existimat quaeri**. » (*Inst. or.* III 6, 36). Autrement dit, sous

### 1 Le terme *accidentia* chez les grammairiens romains

la notion de *συμβεβηκότα* ou d'*accidentia* en traduction latine, Théodore a réuni tout ce qui était à disputer et à définir sitôt après l'établissement du fait.

D'autres maîtres grecs de la rhétorique ont développé à leur tour une répartition de la cause civile en deux parties principales. Mais le cas du célèbre érudit romain Cornelius Celsus (première moitié du Ier siècle ap. J.-C.) est particulièrement intéressant pour nous. Quintilien nous apprend que, dans la répartition de la cause civile, Celsus partait de deux questions semblables à celles que Théodore posait jadis : *Celsus Cornelius duos et ipse fecit status generales*, an sit ? quale sit ? *Priori subiecit finitionem* (*Ibid.* 38). Quant à la deuxième partie de la cause, dont parle Celsus, il la nomme tout simplement *qualitas,* si l'on croit Quintilien, et la subdivise à nouveau en deux catégories : *Qualitatem in rem et scriptum dividit.* Les sujets qui constituaient le contenu de la deuxième partie de la cause civile selon la répartition générale de Théodore de Gadara et Cornelius Celsus, occupent une place prépondérante dans les traités de rhétorique que nous connaissons. Les termes *συμβεβηκότα* ou *accidentia*, aussi bien que le verbe latin au pluriel *accidunt*, pouvaient donc être appliqués par les rhétoriciens érudits à différents sujets à traiter, y compris aux propriétés morphologiques et phonétiques des parties du discours.

Les œuvres de Denys d'Halicarnasse, contemporain de Théodore de Gadara, venu lui aussi à Rome vers 29 av. J.-C., ont donc pu également subir l'influence de cet usage. Mais ce sont les écrits de Pline l'Ancien (23/24–79 ap. J.-C.) qui présentent, semble-t-il, un cas particulièrement suggestif. On sait que, dans sa *Naturalis historia,* Pline l'Ancien a puisé son matériau maintes fois chez Aurelius Celsus (cf. Münzer 1897 : 41–45, 56–70). Il est vraisemblable que Pline connaissait aussi la partie de l'œuvre encyclopédique de Celsus consacrée à la rhétorique, et qu'il s'en est inspiré dans son traité de rhétorique *Studiosus* qui n'est pas parvenu jusqu'à nous. Il est particulièrement significatif qu'on trouve un reflet possible de la terminologie de Celsus dans les fragments du traité grammatical de Pline *De dubio sermone*.

Il s'agit de l'exposé des idées de Pline l'Ancien sur la forme personnelle des pronoms définis, exposé qu'on trouve chez Clédonius, éminent grammairien du V e siècle (*GL* : V 49.27), aussi bien que dans l'*Ars Bernensis* (*GL* : Suppl. 135.1– 135.8), dont l'auteur se réfère au grammairien Sergius, actif à la fin du V<sup>e</sup> siècle et dans la première moitié du VI<sup>e</sup> siècle. Nonobstant leur caractère tardif, ces témoignages semblent offrir un reflet assez fidèle du lexique de Pline ainsi que de sa pensée même. Suivant ces témoignages, ainsi que celui du grammairien carolingien Clément, Pline considérait la forme personnelle desdits pronoms comme inhérente à eux et non comme quelque chose d'adjacent.

### Vladimir I. Mazhuga

Plinius artigraphos dicentes **pronominibus finitis accidere personas** reprehendit. tunc enim bene diceretur, **si aliud esset pronomen finitum, aliud persona, non enim una res potest esse quae accidit et cui accidit, ergo melius ita dicendum est**, ait, *eadem esse finita pronomina, quae sunt etiam personae* (Cledonii *Ars*, *GL* : V 49.27–49.32 ; cf. *Clem. Ars* : 61.8 ; Della Casa 1969 : 301).

Sed tamen Plinius Secundus grammaticus, sicut Sergius ostendit, reprehendit eos, qui dicunt **personas accidere finitis pronominibus**, ut **ipsud accidens aliud sit atque illud, cui accidit**. Hinc Plinius ipse dixit : *Sed scire debemus huiusmodi definitores non tam in ratione errare quam in ordine verborum, ut dicerent personas pronominibus accidere, cum dicere debuissent finita pronomina non recipere quasi aliunde personas* (*Ars Bernensis*, *GL* : 135.1–135.8 ; cf. Della Casa 1969 : 301–302).

Le verbe *accido*, aux différentes formes où il est employé, conserve ici encore sa signification primitive d'une connexion occasionnelle. Il est significatif toutefois que le mot soit employé précisément à propos de l'attribut d'une partie du discours (Della Casa 1969 : 301). L'autorité de Pline l'Ancien a été assez forte pour donner une première impulsion aux innovations lexicales relatives aux attributs des parties du discours.

Flavius Caper, *Magister Augusti Caesaris* (selon le témoignage de Pompeius : *GL* : V, 153.13), grammairien de la deuxième moitié du II<sup>e</sup> siècle ap. J.-C., semble avoir récapitulé ces essais. Il a composé, entre autres, un traité intitulé *De dubio genere*, dont le thème est très proche du traité plinien. Le traité ne nous est pas parvenu, excepté quelques citations tardives, mais l'on peut aisément présumer qu'il dépend du traité de Pline. Le fait est que tous les auteurs dont les manuels de grammaire accusent une dépendance par rapport à Caper emploient déjà le verbe au pluriel *accidunt* appliqué aux attributs des parties du discours. C'est le cas surtout des grammairiens d'origine africaine, comme Palladius – connu communément sous le nom de Pseudo-Probus et actif à Rome au début du IV<sup>e</sup> siècle –, ou comme le Pseudo-Augustin, (*GL* : IV, 522.4 *et passim*; V, 517.35 ; 519.27 ; 520.33 ; cf. Mazhuga 2007 : 274 ; 2011 : 104). Il en va de même des *Artes grammaticae* de Sacerdos, qui enseignait la grammaire à Rome à la fin du III<sup>e</sup> siècle (*GL* : VI, 429.16 ; 442.16 ; 444.22).

Résumons nos observations. Contrairement à l'opinion reçue, il n'y a pas de raisons suffisantes pour mettre l'emploi des termes *accidunt* et *accidentia* dans la grammaire latine en dépendance directe de l'usage des grammairiens grecs qui référaient aux attributs des parties du discours par le verbe *παρέπομαι* et le participe *παρεπόμενα* avec ses formes substantivées. Il n'y a pas besoin, non plus, d'expliquer l'emploi desdits termes latins ni par l'influence de l'emploi présumé du terme *τὰ συμβεβηκότα* dans la philosophie stoïcienne du langage, ni par son emploi bien attesté dans l'épistémologie épicurienne. On doit seulement suivre la transformation du terme latin *accidentia*, en partant de la signification du terme grec apparenté *τὰ συμβεβηκότα*, que celui-ci a acquise dans les théories rhétoriques au début du Principat.

### **Remerciements**

L'auteur exprime sa profonde gratitude à Mme Sylvie Archaimbault (CNRS/Univ. Paris-Sorbonne) qui a pris en charge de relire et de corriger le texte français de cet article. Il tient également à remercier M. Jean Lallot (École Normale Supérieure, Paris) de ses suggestions importantes et de sa relecture du texte.

### **Références**


## **Chapter 2**

## **From localism to neolocalism**

### Jean-Michel Fortis

Histoire des Théories Linguistiques, CNRS, Université de Paris

Localism is the hypothesis that spatial relations play a fundamental role in the semantics of languages. Localism has a long history. The first instance of a localist account can be found in Aristotle's *Physics*. Later, localist ideas surface time and again for the purpose of analyzing prepositions, cases and transitivity. The first part of this paper will be devoted to a short account of past localist ideas. Remarkably, new forms of localism have reappeared in the past decades. This neolocalism involves two main lines of investigation: thematic roles and lexical semantics, especially the semantic analysis of prepositional meanings. In this paper, our next task will be to contextualize the development of these two strands by placing them in their theoretical environment. Both begin to flourish at a significant juncture marked by the rise of cognitive science and by the semantic turn observable in linguistics in the 1960s. This global context is the subject of our second part and sets the stage for a discussion of neolocalist accounts in the third part. Lastly, since this paper makes no pretense at being exhaustive, we draw attention to questions that had to be left out: the existence of more "abstract" forms of localism, the connection of localism with "grounded cognition" and, finally, diachronic studies.

### **1 Introduction**

In this paper, "neolocalism" refers to localist accounts which have flourished since, approximately, the 1960s, in lexical semantics as well as in analyses of thematic roles. Let us first circumscribe our subject matter, *localism*, a little more closely.<sup>1</sup>

<sup>1</sup> As far as can be ascertained, the term "localism" first circulated among German-speaking authors. Unfortunately, I have been unable to identify the place of its first occurrence. Neither Bopp, Wüllner nor Hartung, who are generally regarded as the first localists by succeed-

### Jean-Michel Fortis

John Lyons defines localism as "the hypothesis that spatial expressions are more basic, grammatically and semantically, than various kinds of non-spatial expressions (…). Spatial expressions are linguistically more basic, according to the localists, in that they serve as structural templates, as it were, for other expressions; and the reason why this should be so, it is plausibly suggested by psychologists, is that spatial organization is of central importance in human cognition" (Lyons 1977: 718). A paramount example is furnished by those semantic analyses which strive to identify a spatial core meaning for at least a subset of prepositions and cases. Neolocalist descriptions of adpositions in cognitive semantics squarely fall within this tradition.

Lyons' definition captures the essence of traditional localism, which may be conveniently summed up in two basic claims: (1) a "morphogenetic" thesis: some non-spatial expressions (e.g. temporal) are derived, or extended, from spatial ones, and this transfer is expected to be one-way; (2) the transfer is unidirectional for cognitive reasons: the palpability and ubiquity of spatial relations in phenomenal experience, the range of conceptual distinctions they afford (location, motion to, via and from, and their further determinations) provide an experientially grounded framework for the construal of relations and properties of a more abstract nature. In spite of Lyons' allusion to the work of "psychologists" (perhaps an allusion to the 1976 book of Miller & Johnson-Laird), this two-pronged version of localism has first and foremost been defended by linguists and without any other support than the methods of their discipline. This has overwhelmingly been the case up to a very recent time.

The definition above, however, is not ideally suited to an important strand of neolocalism, that is, to those theories in which local relations impart structure to what is today designated as thematic roles. Thematic roles need not be transparently expressed by means of spatial forms, hence they may at least partly deviate from the morphogenetic claim; likewise, claiming that changes of state rest, for their linguistic structuring, on the kind of conceptualization that gives rise to the expression of motion events is a step that an analysis of surface forms does not warrant straightforwardly. Further, some authors may be reluctant to endorse

ing authors, describe their own theory as a brand of "localism". On the other hand, the term *Lokalisten* ('localists') is found in contexts in which the defenders of various forms of localism are lumped together and subjected to a critical examination, e.g. Michelsen (1843) and Rumpel (1845). The latter authors refer back to Grotefend (1835) as an early opponent of localist claims, but I was not able to consult Grotefend's essay and check whether he was the first to coin the terms localist/localism. At any rate, the terms seem to have been first used by critics, in order to characterize a doctrine that was either repudiated (Grotefend, Rumpel) or found to be too one-sided (Michelsen).

### 2 From localism to neolocalism

the claim of a cognitive primacy of spatial relations; we shall see examples of this stance in the course of the discussion. In a nutshell, the meaning to form orientation of some modern theories is at variance with the semasiological perspective of traditional localism, while the cognitive primacy of space may be, in theory, if not in practice, disputed by some authors.

In what follows, we shall see how these two forms of neolocalism have come to coexist through a revival, rediscovery and reelaboration of traditional localism. Our discussion of neolocalism will be divided in two parts. The first part will be devoted to the environment which favored this new surge of localist ideas. The second part will deal more specifically with neolocalist accounts in semantics and in the field of thematic roles. Finally, in the fourth part, a short discussion reviews some aspects of today's research which could not be explored in sufficient detail in the preceding sections. But first of all, since allusion has been made to the localist tradition, a brief overview of past localist ideas is in order.

### **2 A retrospect**

The best-known source on the history of localism remains Hjelmslev's doxography in his *Catégorie des Cas* (1935–1937). His purview, however, is not purely historical: some precursors were singled out not for their historical role but in proportion of their value in Hjelmslev's eyes. Further, his description, including that of theories he praised most, can be faulted on several counts (Fortis 2018a).

To my knowledge, the first localist analysis can be found in Aristotle's *Physics*, in the context of a semantic clarification of the relation of containment as expressed in the Greek preposition *ἐν* ('in'). In the *Physics*, this clarification is motivated by Aristotle's definition of space as the (minimal) place containing a body. As is the case for 'being' in *Metaphysics*, the notion of spatial containment is circumscribed and semantic anarchy staved off by distinguishing the various meanings of *ἐν* in a principled way (in particular according to the categories), and showing that they all presuppose a primary meaning, which Aristotle describes as local (namely 'being in a place'). This primacy, says Aristotle in the *De anima*, rests on the fact that our objects of thought reside in sensible forms, and place is a necessary condition of the existence of these forms (423a3–423a8). The justification, therefore, is both ontological and cognitive.

The most direct localist legacy of Aristotle's *Physics* dates from the rediscovery of the text in the Middle Ages, and is embodied in the corpus of studies assignable to *speculative (or modistic) grammar* (13th–14th centuries), a synthesis of grammar and logic which in effect promotes an autonomous syntactic anal-

### Jean-Michel Fortis

ysis aiming at universal validity, though obviously taking its data from Latin (Bursill-Hall 1971: 29). The universal scope of the analysis is ensured by moving to a level of description that is based on metaphysical and physical notions. For example, for the purpose of defining verbhood, tense is no longer an ultimate feature. Rather, the ability to carry tense is grounded in the fundamental semantic import of verbs, which is to convey motion, hence time. From this *physicalist* conception of verbs sometimes follows a redefinition of cases in *localist* terms, i.e. as expressing the origin/source or the goal of motion. We see, for example, Simon of Dacia, defining the accusative as a "casus dicens terminum motus" (Kelly 1977). After the Modists, localist statements can be found in various places, especially in authors with a philosophical inclination, such as J.C. Scaliger who, in his *De Causis*, characterizes the class of prepositions with the Aristotelian category of place. As for non-spatial uses, they are related to spatial uses, he says, by *analogy* (2018 [1540]: 152), a post-Aristotelian term which, in this context, refers to the relation to a primary sense.

In the period which approximately spans the 17th and 18th centuries, localist statements concern nearly exclusively prepositions. Although the pedagogical practice of glossing Latin cases with vernacular prepositions (and conversely) had alerted grammarians to the functional equivalence of prepositions and cases (Colombat et al. 2010: 26), they were manifestly more reluctant to analyze cases than prepositions in localist terms. Arnauld and Lancelot, for instance, while recognizing this functional equivalence (Arnauld & Lancelot 1969 [1660]: 62) provide no extensive systematic account of cases; in all likelihood, their arbitrariness was for them a challenge to the very possibility of such an account. In addition, cases were associated with grammatical relations, for example in discussions on the natural word order. This must have made them appear to be of a degree of abstraction not amenable to a reduction to a local primary sense. Not until the time of Harris (1773 [1751]), it seems, are cases treated in localist terms on a par with prepositions. Doeleke may therefore be right when he praises the British Neoplatonist for having been the first to explain the meaning of cases in terms of spatial relations (1814: 7).

An obstacle to localism was that the whole class of so-called *particles*, including prepositions, had been associated with acts of thought, not with conceptual content (Nuchelmans 1986). When Leibniz, stimulated in particular by Locke, turned his attention to particles, his localist analysis of prepositions was a way of undoing the act/content distinction by providing them with a conceptual substance, with the ultimate purpose of paraphrasing them in a universal language. In the general epistemic context of the time, linguistic analysis took on a new im-

### 2 From localism to neolocalism

portance, because of the status of language for the theory of knowledge: language revealed operations and concepts of the mind (for Leibniz, cf. Dascal 1990) but also had a potential of obfuscation (especially for Locke, Dawson 2007), aspects which both attested to its cognitive power. The nominalist proclivity of empiricism, quite perceptible in Locke's treatment of mixed modes ("it is the name that seems to preserve those essences, and give them their lasting duration", Locke 1975 [<sup>4</sup> 1700]: 434), by conferring to language a capacity to form universals, could only reinforce its cognitive relevance (see an important discussion of this point in Formigari 1988).

In this broad context, forms with a spatial meaning are particularly central for they can lay bare a fundamental aspect of cognition, the apprehension of spatial relations. In the diachronic perspective of the time, this apprehension was approached in two different, sometimes coexistent, ways: in view of the importance of this apprehension in the cognitive development of mankind, in a priori genealogies of mind and language (cf. Condillac 1775, esp. II.13 on the primary spatial meaning of French *de*), or in the first attempts of modern historical linguistics, as justifying hypotheses on the origin of forms, especially of cases. Localist hypotheses of an a priori nature could coexist with "technical" considerations on the evolution of forms: this was the case in Doeleke's study, and Bopp himself entertained localist ideas, hence his claim that some endings of Sanskrit, Latin and Greek declensions originated from prepositions and demonstratives with a primitive spatial meaning (1826).

All throughout the 19th century and well into the 20th century (cf. Kuryłowicz 1964), the primacy or importance of spatial meaning for cases is a claim many authors endorse or feel called upon to discuss or criticize, mostly in the perspective of historical linguistics, but in psychological linguistics and philosophy as well (cf. Marty 1910). The 19th century is the centerpiece of Hjelmslev's overview, which is still worth consulting for more information (but again, caution has to be exerted; Fortis 2018a). The most developed localist account of cases, with an emphasis on Latin, Greek and Sanskrit, is put forward by Wüllner (a student of Bopp) in two books, of which the first (Wüllner 1827) proposes a semantic analysis focusing on Greek and Latin cases, while the second (1831) introduces Sanskrit data and is more historical. In both studies, Wüllner defends the radical view that the fundamental meanings (*Grundbedeutungen*) of all cases (except the nominative and vocative) are local. This "pan-localism" is obviously achieved at the price of sweeping the nominative case under the rug, thus dodging the problem of analyzing it in localist terms (cf. also Hjelmslev's own convoluted localist account). Otherwise, Wüllner's basic tenet is simple: the *Grundbedeutungen* of

### Jean-Michel Fortis

the genitive, accusative and dative cases are spatial intuitions (*Anschauungen*), respectively of a starting point (*Anfangspunkt*, *woher* 'where from'), of a goal (*wohin* 'where to') and a localization (*wo* 'where') (the ablative is considered as a secondary differentiation of the dative). Other senses are derived from these intuitions, forming semantic networks akin to those found in present-day cognitive linguistics (without, however, the now usual diagrammatic representations).

The testimony of Rumpel (1845) tells us that in the domain of cases localist accounts enjoyed a supremacy, especially for pedagogical reasons. This supremacy, for Rumpel, had to be contested and overcome. Together with Curtius (1864), Rumpel (1845; 1866) championed an anti-localist reaction which betrayed a certain weariness of philosophical grammar. Localism was for him an offshoot of an outdated conception of grammar, inherited from the Enlightenment, and according to which language was the creation of the human mind reflecting on its own operations, in a bootstrapping process taking its origin from sensory features and embodied cognition. The description could fit Wüllner's perspective pretty accurately and, beyond him, might have had Herder and his *Besonnenheit* ('reflection') in mind. What Rumpel aimed at was a more formal definition of grammatical relations, redirecting grammar toward surface forms and taking as basic the most fundamental fact of human thinking, the subject-predicate structure, for which localists had no explanation. He also inveighed against the proliferation of senses entailed by the semantic network approach of localists, a proliferation, he said, which in effect transferred to cases the semantic features which had traditionally been used to classify verbs in cooccurence with the different cases.<sup>2</sup>

Authors who were less philosophically committed than Wüllner and Hartung and proceeded more matter-of-factly showed some reluctance to accept that the genitive was more than the default adnominal case, or that the accusative had a fundamental spatial value, since this use was marginal with the bare case. Further, with the evolution of historical grammar, Latin and Greek cases were more and more seen as syncretic with respect to Sanskrit, which made it more difficult to confer a unitary value on them. The consequences are best appreciated by considering Holzweissig's semilocalist account (Holzweissig 1877), as summarized in Table 2.1. In Holzweissig's system, cases with one fundamental spatial value are restricted to a subset of Sanskrit cases and distinguished from the grammatical cases (nominative, vocative, accusative, genitive). Latin and Greek have reshuffled the values of Sanskrit local cases; the Latin ablative, and the Greek genitive and dative are described as *Mischcasus*.

The shift from spatial to non-spatial values is relativized to a historical stage. Note that Sanskrit lays bare the fundamental meanings present in a more primitive stage of Indo-European.

<sup>2</sup>A strategy found e.g. in Despautère (1527 [1509]) and Lancelot (1653)

### 2 From localism to neolocalism

Table 2.1: Holzweissig's (1877) semilocalist account of the evolution of cases.


The assumption that the concepts underlying cases (*Grundbegriffe*) must be found in a primitive stage is precisely what Wundt (1912) rejects as "mythological". Being cognitively (and affectively) motivated through and through, expressions of relations must reflect, for any language at whatever stage, the linguistic decomposition of thought into attribution and predication, as well as an openended set of phenomenal properties (*Gegenstandsbegriffe*), or "external determinations", which may involve causality, comitativity, similarity…, in addition to spatial relations. On this view, localism unduly restricts "external determinations" to spatial values, a mistake justifiable by the fact that external determinations are maximally distinctive in the spatial domain (Wundt 1912: 115).

### **3 The global context of neolocalist views**

We shall now try to describe in broad strokes the environment in which neolocalist views were incubated, with the goal of understanding the conditions which favored or legitimized them. For convenience, these conditions have been sorted along disciplinary lines.

### **3.1 The global context: The rise of semantics**

Neolocalist descriptions may be conveniently categorized into two classes. One class comprises semantic analyses of lexical items.<sup>3</sup> The other includes localist accounts of thematic roles. Both kinds of description appear to emerge at a juncture which corresponds to the so-called "cognitive revolution" (*revolution* is a term we would not endorse for reasons we cannot go into here). What should

<sup>3</sup> In this paper, we will designate as "lexical semantics" the semasiological analysis of forms like *out, up, over* etc. as well as the onomasiological study of the notion of verticality in Nagy (1974). Let us observe that the onomasiological orientation of Nagy remains rather marginal (as we shall see).

### Jean-Michel Fortis

most arrest us in this "revolution" is a number of theoretical changes directly relevant to the new rise of localism. On the side of linguistics, especially in the United States, semantic concerns gain in importance, after a relative eclipse among the post-Bloomfieldians; on the other hand, thematic roles come to the fore of syntactic analysis, especially with Fillmore, as we shall see later. Both trends, as can be gathered from testimonies of the time and from the turn of events itself, notably the advent of generative semantics, were significantly encouraged by generative grammar, and in particular, the notion of deep structure invoked in the *Aspects* model. In this paper, I shall not delve further into this evolution and the semantic turn prompted by transformational grammar (TG); this has been documented and discussed elsewhere (Newmeyer 1986; Harris 1993; Huck & Goldsmith 1995; Fortis 2015). However, the important role played by TG should not mislead us into thinking that semantics would have remained a blind spot if generative grammar had not entered the scene. Hymes and Fought's (1981) observation that, over the long term, American structuralism progressively expanded into syntactic and semantic territories does not exclusively rest on the success of generativism. Their judgment can be confirmed by the fact that forays into semantic issues were accomplished by practicioners of a "late structuralist" bent, i.e. by linguists who were extending up to a semantic layer the stratal organization of forms into allophones/phonemes/morphophonemes-allomorphs and morphemes. We cannot go into the details of this evolution here, but suffice it to say that it was a short step to conceiving of the morpheme as an abstract semantic unit as soon as one had analyzed, e.g., the /u/ of *took* as an allomorph of /ed/; for /ed/ could then be glossed as /past/ (Hockett 1954, Lamb 1964; for a discussion on this history of the notion of morpheme, cf. Matthews 1993).

Chafe (1962) was an early proponent of this extension to semantics: in this paper, he proposed considering morphemes as arrangements of semantic features, on the analogy of phonetic features and, in all likelihood, on the model of componential descriptions put forward in "ethno-semantics", especially for the purpose of analyzing kinship terms (e.g. Lounsbury 1956; on this short-lived strand, cf. Murray 1982). The examples of generative grammar and generative semantics would encourage Chafe (1970) to develop his theory in the direction of a stratal model in which surface forms are generated from a considerably enriched semantic stratum. In this elaborate model, the semantic stratum was in charge of inventorying forms along semantic parameters, and of stating selectional restrictions, semantic changes effected by derivations and inflections, and even pragmatic aspects (such as intonational variation and information structure). Note that this conception would lead Chafe to abandon the notion of morpheme alto-

### 2 From localism to neolocalism

gether, hence to divorce the semantic stratum from the segmentation of linear sequences of forms.

The ambiguity of the notion of morpheme, i.e. its being a class of forms or something *represented* by phonemes, is quite explicitly one of Lamb's (1964) starting points in developing his own version of a stratal grammar. To put some order into this confusion, Lamb recommends distinguishing carefully what a unit is composed of from what it is taken to represent. For example, he insists that <sup>M</sup>/good/ and <sup>M</sup>/bett-/ (in *better*) are morphemes which cannot be put on the same level as what they represent, namely a superior unit called "lexon" and glossed as <sup>L</sup> /good/. The latter, in turn, is separated from the semantic unit it represents, its *sememe*. This separation of a semantic plane, as it was for Chafe, apart from considerations having to do with the inner logic of the system, is motivated by the wish to account for the properties which he thinks can only be stated on this level, and between this level and lower ones, for instance properties such as the synonymy and polysemy of forms, e.g. the fact that the sememe <sup>S</sup> /also/ is represented by <sup>L</sup> /also/ and <sup>L</sup> /too/ and, in a negative environment, by <sup>L</sup> /either/. In short, in Lamb's framework, a stratificational grammar (SG) should provide an account of the "tactical" pattern (= combinations) proper to each stratum and of their interrelations. To this purpose, Lamb resorts to a formalism of his own which consists of networks presented in diagrammatic fashion. Units are connected via various types of AND and OR nodes which account, respectively, for the composition and alternations of units (or classes). The first developed presentation of this formalism was submitted in his short opus of 1966.

It is all the more important to mention Lamb's version of SG since it was perceived at the time as a rival to TG, and one that could favorably compete with it insofar, e.g., as it afforded an explicit measure of the complexity of grammars (roughly, as a function of the number of links between items). However, a major drawback of Lamb's theory was that it was not fully expounded until Lockwood's textbook (Lockwood 1972). This inconvenience, together with other adverse circumstances (Nielsen 2010), ensured that SG would never win a support in any way comparable to the success enjoyed by generative grammar.

On the whole, both TG and stratal grammars contributed to the rise of semantics, and this rise manifested itself, among other things, in semantically-oriented studies of prepositions and in inchoate localist analyses of verbs. White's (1964) analysis of English prepositions, based on a corpus, offered an early example of the former kind of study. It was framed in Lamb's formalism and exploited the potential of Lamb's systemic approach. That is, White considered a *system* made up of 11 non-compounded prepositions which he analyzed into their sememes

### Jean-Michel Fortis

(senses), based on the commutation with other prepositions and on distributional evidence. For example, the commutability of *about* with *around* in certain contexts was taken to justify positing a separate sense corresponding to this use of *about*; in other cases, distribution, that is co-occurrence with a set of semantically cohesive verbs, provided evidence for a distinct sememe. A residue of uses was characterized as idiomatic. Lamb's network notation served to represent the semantic interconnections of prepositions in the system. Worthy of note was the fact that his method was conducive to a proliferation of senses, and that no localist hypothesis was formulated, and understandably so, since his tests conflated *at home* and *at noon*, and, on the other hand, distinguished subtle shades of meaning as a function of cooccurrences.

An example of an early localist attempt coming from the transformational circle is provided by Lakoff (1976 [1963]): to simplify, in this paper, some selectional restrictions on verbs were associated with a classification of these verbs by semantic features, and this classification was partly localist. There was for instance a class of verb of "directed change" which was subdivided into goalvs source-oriented items (resp. *I became insane* vs *I lost my sanity*). Talmy's dissertation (1972) could also be seen as an (unorthodox) emanation of the transformational approach, close to the generative semantics movement. His deep syntactico-semantic structure was tailored to the analysis of structures referring to motion events, and in this initial stage of his theory, this structure could be interpreted as a linguistic template transferrable to non-spatial events. This was hypothesized to follow from the cognitive centrality of the structuring of motion events (cf. Talmy 1975: 234). There is in this respect a seamless evolution leading to Talmy's more direct concerns with cognitive matters and his future affiliation with cognitive linguistics (Fortis & Vittrant 2016).

The most elaborate study adopting the SG framework was due to Bennett (1975) and chose a strategy opposite to that of White, i.e. Bennett strove for the maximal reduction of polysemy. Bennett's plea for monosemy was facilitated by his methodology (neither based on corpus nor on distribution) and the fact his scope was confined to spatial and temporal uses, which implied that no attempt was made at deriving or explaining more abstract uses. From his testimony, we learn that his inclination to monosemy came from Jakobson's description of cases as expressing a *Gesamtbedeutung*, Fillmore's (1968) attempt at identifying a list of universal deep cases (in modern parlance, thematic roles), and finally componential analyses of the kind promoted by SG. In short and to simplify somewhat, his semantic descriptions were structured sets comprised of five local cases (locative, source, goal, path, extent) combined with various specifications such

### 2 From localism to neolocalism

as, e.g., 'interior' (for *in*), 'proximity' (for *by*) or 'visibility' (for implicitly viewercentered uses, that is uses in which a viewer is a reference point). The semantic structuring of prepositions was represented in the form of tree diagrams, which thus served to express scope relations: In *The passenger fell asleep and went past his station*, 'past his station' was glossed as 'at a goal at the end of a path via the proximity of his station with respect to a reference point', with 'at —' having scope over 'goal', 'goal' over 'path' etc. As for the stratificational formalism, it was used, in Lamb's parlance, on the semotactic and semolexemic levels. To semotactics belonged the task of expounding the various kinds of locative propositions available in English, the range of choices at the disposal of speakers (such as the distinction between extent and locative expressions, and the inventories of their components) and selectional restrictions, for instance the cooccurrences between aspectual classes of verbs and prepositions. The semolexemic diagrams encoded the lexemic realization of semantic structures.

Bennett's position is not localist, insofar as he never declares that space has any kind of primacy and he does not venture into diachronic and cognitive considerations. He claims, however, that local cases have spatial and temporal uses in common; in addition, he proposes an analysis of tenses in terms of local cases. If Bennett were to be assigned to a family of theorists, he might be deemed closest to those who have always denied the primacy of spatial meanings or, in Anderson's words, who have insisted on the *neutrality* of the principles of organization with respect to the domain in which they are instantiated (be it spatial, temporal, or abstract). In this group we may include Beauzée (1786) and Pottier (1962 and later studies by the same author), who both stand for a notion of locative meaning encompassing domains other than space. For Pottier, e.g., the locative layer of cases encompasses space, time and "abstract" meanings he calls "notional" (e.g. Pottier 1974: 53–55). Whoever would summon diachronic evidence for asserting the primacy of space would be, according to Pottier, misguided: synchronically, relators have a permanent potential for expressing spatial, temporal and notional relations (Pottier 1962: 126). According to Anderson (1971; 1994), Hjelmslev also adheres to the neutrality stance, which he suggests distinguishing from bona fide localism and for which he reserves the special name of "localistic". Localistic views, therefore, either are agnostic on the cognitive grounding of local cases, like Bennett, or explicitly reject the primacy of space, like Beauzée, Hjelmslev and Pottier, yet see commonalities between spatial and non-spatial relations.

A hallmark of stratificational approaches was their full autonomization of a semantic plane. In this they differed from generativists, be they adherents of the interpretative or generative version of TG, who regarded semantics, respectively,

### Jean-Michel Fortis

as a matter of providing interpretations of syntactic structures, especially to disambiguate them, or as a very deep level in charge of the generation of surface forms. There were for sure semantic representations and specific semantic rules in analyses affiliated to TG, for example in Katz and Fodor's famous attempt at specifying what an interpretative component should be like (Katz & Fodor 1963). By contrast, some studies, including some not affiliated to SG, also undertook to state properly semantic rules not indexed to the generation of surface structures, and to this end had developed a semantic notation apt at representing what they regarded as the "quasi-deductive system" underlying semantic interpretation. "Quasi-deductive system" was the term employed by Weinreich (1972: 163), whose blueprint for a semantic theory incorporated a notation for representing semantic interactions between cooccurring terms and sketched a generative account directly mapping semantic-categorial structures (such as 'verb + circumstance') to morphosyntactic forms. In the same spirit was Leech's (1969) attempt at devising a formulaic (and rather cumbersome) notation for semantics. Leech made use of two basic relations (predication and attribution), various symbols for representing definiteness, quantifiers, negation, inchoativity etc. and, like Weinreich, analyzed lexical contents into bundles of features (or "clusters"). His field of inquiry, he declared (1969: 28–30), was properly the semantic plane viewed as an *autonomous* level, in contradistinction to generative grammar. Most significant for us was the fact that he applied his apparatus to the domains of place, time and modality, which betrayed the philosophical background of his seemingly purely linguistic essay: in effect, the fundamental concepts and domains he was working with were the Kantian forms of intuition and categories. Presumably, then, place was chosen as a domain of application in view of its being a received category of Western epistemology (for a discussion of this point, see Chalozin-Dovrat, this volume). It was also chosen because spatial markers (reduced to prepositions and names of compass points and object parts) appeared to form a *system* whose features were amenable to a complete inventory. In short, place expressions were both basic and manageable.

### **3.2 The global context: Cognitive linguistics**

It is far beyond the scope of the present paper to explore the origins of cognitive linguistics. The reader may consult other texts in which I have attempted to narrate this history (e.g. Fortis 2015). Of immediate relevance for our subject is the fact that, after the schism which caused cognitive linguists to split from generative grammar, semantics offered itself as a promising niche. These linguists (notably Lakoff, Langacker and Talmy) were all the more inclined to engage in

### 2 From localism to neolocalism

semantic issues since they had all been involved in the semantically-oriented dissident movement known as generative semantics. In this favorable environment, localist ideas surfaced again. We have seen the case of Talmy above; unfortunately, since his dissertation contains no references to previous work, the possible inspiration of his early localism remains inscrutable. Lakoff and Johnson, with their conceptual metaphor theory, had embarked on an empiricist program which defended as a corollary rather extensive localist views, thus claiming that "most of our fundamental concepts are organized in terms of one or more spatialization metaphors" (Lakoff & Johnson 1980: 17). In fact, spatial metaphors such as 'more is up' (e.g. *prices have gone up*) were cited by Johnson (1981) as a counterexample to the view that metaphors rest on similarities between a source and a target concept. They played a role, therefore, in establishing the very notion of conceptual metaphor, that is, the idea that a metaphor is not based on preconceived similarities but "serves as a device for reorganizing our perceptual and/ or conceptual structures" (Johnson 1981: 31). It should be noted that this aboutturn in Lakoff's theoretical concerns, from generative semantics to metaphor theory, did not come out of the blue. Metaphor was a much discussed subject in the 1970s, a "hot topic" as Honeck (1980) puts it, partly because of the redirection of psycholinguistics from transformational grammar to semantic issues.

Just as Leibniz (1923 [1685–1686]), among others, had claimed that the nonspatial senses of particles are connected through tropes with their primary spatial senses, figures (and especially metaphors) were employed for deriving "abstract" senses from concrete ones. Further, through the adoption of prototype theory, imported from psychology and legitimated by it, though simplified in the process, linguists had found a convenient tool for handling polysemy (Kleiber 1990; Fortis 2018b). The combination of empiricist views and of prototype theory obviously favored the reintroduction of localist semantics, especially in the traditional area of particles and adpositions, as we shall see in the next section. In the present case, speaking of a "rediscovery" might not be quite appropriate. Although cognitive linguists had been raised in the lap of generative grammar, and hence were relatively cut off from the tradition of semantics, we cannot exclude that transmission did take place. Langacker, for instance, was acquainted with Nagy's dissertation, had read Anderson's (1971) *Case grammar*, and had declared himself to be "basically sympathetic" with its localist orientation (Anderson 1973). Likewise, the work of Nunberg (1978) was known to Lakoff (as is attested in the acknowledgements), and in this text reference was made to historical linguistics, with an eye toward the application of principles of semantic change to the synchronic treatment of polysemy. In particular, Nunberg hinted at the work of Darmesteter (1887), which, if it had been consulted by Lakoff, may have inspired

### Jean-Michel Fortis

him to represent a semantic network in diagrammatic form. The junction of empiricism, prototype theory and semantic networks at least partly arose out of these historical circumstances.

### **3.3 The global context: Psychology and cognitive science**

The rise of semantic concerns in linguistics coincides with significant changes brought about in psychology by the demise of behaviorism, although, as we shall see with Osgood and Nagy, the divide between pre- and post-behaviorist psychology was not such that no continuity nor conciliation can be witnessed.

The changes hinted at here can be conveniently placed under the banner of "cognitive psychology", first so named in Neisser's book of 1967. Opposition to stimulus-response psychology had gathered momentum in the preceding years (mostly, from the 1950s on) and a series of studies had converged toward the idea that subjects actively (re)construct the stimuli they are exposed to. For example, Bousfield (1953) had observed that subjects tended to recall words in clusters corresponding to semantic categories, that is, they tended to reorganize the material presented. In a similar vein, Bransford & Franks (1971) observed that subjects recognized as old a sentence that was in fact new but semantically coherent with the sentences that had been presented during the acquisition phase. Attneave (1957) had demonstrated that subjects found familiar a shape that had not been part of the experimental items but had served to model them by systematic deviations, and which he called for that reason the *prototype* of the series. Neisser (1967) laid much emphasis on these constructive processes, which he regarded as ubiquitous and spanning all the range of human abilities, from perception (cf. his notion of iconic memory as a buffer storing items for constructive processing) to the hierarchical verbal structures posited by TG, strangely likened to superimposed *Gestalten*. On the whole, constructive processes and structuring went hand in hand with the relevance devolved to semantic factors.

In some quarters, the mind-as-computer metaphor and the comparison of cognitive processes with states of a Turing machine (the so-called functionalist view of Putnam 1960) were enthusiastically seized upon as offering a free hand to speculate on mental representations, including in linguistics (Katz 1964), without having to worry too much about their ontology. This new freedom was a favorable environment for the reintroduction of notions which had been repressed, though not entirely banned, during the behaviorist era, such as mental images, voluntary attention, or teleological behavior ("will"). For our subject, the fact that mental images were rehabilitated is of particular importance, since spatial or diagrammatic representations would later flourish in cognitive linguistics, and would be

### 2 From localism to neolocalism

identified with the meaning of spatial morphemes. This rehabilitation of imagery was progressive, anticipated in some late behaviorist work (Mowrer 1960: 281ff) and in the margins of mainstream psychology (e.g. in research on hallucinations, Holt 1964). It was legitimated with experiments which had a strong persuasive power because they exhibited striking linear relationships between processing time and variables hypothesized to be proper to visual images (Baars 1986: 161, cf. Shepard & Metzler 1971 and other studies). There was, however, some resistance to accepting that phenomenal properties of representations (such as the visual properties of images) may play a functional role in cognition. This resistance may be seen both as a prolonged aversion to anything smacking of introspection, and as a consequence of the computational view of the mind, according to which mental representations with an effective computational role must have a propositional structure (on this debate cf. Fortis 1994). This resistance, however, was largely overcome, and the endorsement of imagery opened up avenues of research for psychologists and neuropsychologists, who devoted a considerable number of publications to the role of imagery in recall, to the relations between images and words (e.g. in studies on picture naming), or to the relative share of modality-specific representations (e.g. visual) in various categories of "concepts", a topic hotly debated in neuropsychology (Fortis 1997).

The importance laid upon the visuo-spatial/linguistic interface was enhanced by the prospect of making the sciences of the mind/brain converge into a unifying approach, cognitive science. It is therefore no coincidence that George Miller, a psychologist who was a staunch advocate of this unified science of cognition (see e.g. his contribution in Arbib et al. 1978), launched into an ambitious study entitled *Language and perception* (1976, written in collaboration with Philip Johnson-Laird). While the first part of the book focused on psychological matters, the second part more directly addressed the psychological underpinnings of the linguistic representation of the perceptual world. Unsurprisingly, the expression of spatial properties (shape, location, and motion) was considered as a fundamental semantic domain, and the authors justified this privilege by appealing to the central role played by spatial relations in cognition, citing in this regard the localist declaration of the philosopher Urban: "our intellect is primarily fitted to deal with space and moves most easily in this medium. Thus language itself becomes spatialized, and in so far as reality is represented by language, reality tends to be spatialized" (Urban 1939: 186, in Miller & Johnson-Laird 1976: 375). Urban's localism echoed British empiricism, Bergson's reflection on spatializing thought, and also Cassirer's views on the spiritualization of concrete determinations effected through language as a medium of representation (Cassirer 1923). These also formed the background, it seems, of Miller and Johnson-Laird's conception,

### Jean-Michel Fortis

but most of their discussion was confined to a semantic analysis of spatial terms (paraphrased with basic concepts expressed in first order predicate calculus) and linguistic coordinate systems.

A surge of universalist ideas accompanied the quest for the unification of the sciences of the mind which was proclaimed to be the goal of cognitive science. From different quarters, universalist hypotheses were boldly put forth: Chomsky's universal grammar (from *Aspects* on), Berlin & Kay's theory of a universal path of color terms differentiation (Berlin & Kay 1969), Ekman's theory of universal emotions (Ekman 1971), Lenneberg's late work on the biological determinants of language acquisition and processing (Lenneberg 1967). In a universalist perspective, the analysis of locative expressions could be framed in terms of universal cognitive constraints on their acquisition and use. Such was for example the way the psychologist Herbert Clark described his own endeavor: demonstrate that the child's a priori knowledge of space, e.g. the "vectorization" of space determined by the asymmetry of the human perceptual apparatus, constrains the acquisition of spatial relators. From a developmental angle, a localist hypothesis implies that spatial markers are acquired before their metaphorical extensions. This was indeed how Clark dealt with temporal expressions, which he considered to be grounded in metaphors based on the experience of motion and lexicalized by primarily spatial markers (Clark 1973). This aspect of Clark's work would later be appropriated by Lakoff in support of his own localism.

### **4 Neolocalist views**

The stage is now set for the emergence of neolocalist views. What follows is an exposition of the views themselves. As was said in the introduction, they address two main issues: lexical semantics and the nature and functioning of thematic roles.

### **4.1 From cross-domain associations to figurative patterns: Osgood and Nagy**

In American lexical semantics of the post-war period, the first clearly localist endeavor can be traced back to a neo-behavioristic framework antedating the semantic turn promoted by transformational grammar. Neo-behaviorism accommodated inner responses which mediated the production of overt behavior, and it was therefore receptive to hypotheses positing unobservable reactions. In Osgood's theory, meaning was precisely such a mediational process: the meaning

### 2 From localism to neolocalism

of a sign was defined as a fractional response, i.e. as a part of the total behavior associated with the referent of the sign; by dint of repetition, this response was reduced to a kind of abbreviated replica, a "disposition" (Osgood 1952). Now, the conditioning of signs to primary dispositions, through generalization, extended to mediators elicited in proportion to their similarity to original reactions, but also thanks to spontaneous cross-domain elicitation, that is, synesthesia (Karwoski et al. 1942). But synesthesia was only the most vivid manifestation of a wider phenomenon, the correlation of properties belonging to different dimensions. Osgood and his associates set out to demonstrate that it made sense, even for subjects who were not synesthetes, to correlate scales which belonged to different modalities, for example the soft-loud scale and the large-small scale, the happy-sad scale and the bright-dark scale etc. From there, the next step was to define the semantic profile of a sign as the set of its values determined by subjects on a large number of continua. This set, the "semantic differential", was thus conceived of as an operational definition of meaning, although, as admitted by Osgood himself, meaning was in effect reduced to connotative associations (Osgood 1952: 231).

While Osgood (1952) had alluded to the cross-cultural relevance of correlations associating spatial relations, especially verticality (up-down) with positivenegative values, he had not pursued this idea along localist lines. In his dissertation, Nagy (1974) applied Osgood's notion of a bipolar organization of meaning by selecting a specific spatial scale, the up-down axis, with the goal of studying the productivity of its application to non-spatial domains across the lexicon. He called *figurative pattern* a mapping from the verticality scale to another domain, noting for instance that many predicates similar to *high* occur in the domain of prices (*prices were above guidelines, one effect of the war was to boost the price of gasoline, stocks prices climbed slowly/declined/dipped* etc.). Distinctly localist in Nagy's study was the notion of an asymmetrical dependence of non-spatial domains on spatial axes. This asymmetry, given the then current concern for generative capacity, was tentatively captured by redundancy rules stating, for example, that lexical items referring to vertical position could be used with terms in the domains of prices, pitch, opinion etc. The restrictions constraining the productivity of figurative patterns proved to be very difficult to state: metonymies had to be taken into account (*stocks went down*), and contextual effects had to be factored in (\**A low suggestion* but *His suggestion as to how much we should pay them was much lower than the legal minimum wage).* This concern for stating the limitations on the productivity of figurative patterns remains a hallmark of Nagy's study. When localist ideas got appropriated by cognitive linguistics, the

### Jean-Michel Fortis

issue of productivity and of stating rules for limiting it receded into the background, probably as a consequence of an antagonism to generative rules on the part of cognitive linguists.

### **4.2 Localist cognitive semantics: Particles and adpositions**

In the preceding sections, we have enumerated a number of circumstances which may explain why semantic analyses of particles and adpositions would gain so much prominence once linguistics, and especially American linguistics, had rediscovered the importance of semantics. To these circumstances we should add the obvious fact that Indo-European languages possess rich inventories of prepositions, preverbs, locative adverbs and particles, which fulfill functions that other languages entrust to a generic adposition or to other strategies, such as verb-serialization, applicative markers and posture expressions (Fortis & Vittrant 2016). In addition, in the special case of English, "relators" such as *at, in, out, off, up* etc. had not disappeared from the linguistic horizon for a reason that is related to the complexity of their morphosyntactic behavior and the attendant difficulty of assigning them to clear-cut categories: are they particles? prepositions? adverbs? "adpreps"? Taking a stance on these matters often meant that semantic considerations had to be brought into the discussion. Some authors, for example, correlated the fact a relator had a "literal" or "concrete" meaning with its having an adverbial function, or being susceptible of receiving contrastive stress (see Lindner 1981, for a review). In a study quite remarkable for treating together morphosyntactic phenomena, semantic aspects and pragmatic intent (such as contrastive stress), Bolinger (1971) underlined the interplay between the literalness of the particle, its ability to move to last position, and its being susceptible of emphasis. Since phrasal verbs in which particles retained their "literal" or "concrete" meaning seemed to be less cohesive, it could be argued that this was due to their making an independent semantic contribution to the phrasal verb, hence that the "literal" meaning was the original one and preceded tighter integrations of the particle with the verb. For this reason, Bolinger claimed for example that the primitive meaning of *up* was directional; he further speculated that this directional meaning had got associated with perfectivity (as in *choke up, rev up*) through physical events of completion (e.g. because filling a glass means the level of the liquid goes up), and also via the notion that a gap between the thing viewed and the viewer was thereby closed (cf. *He came up to me*). In conformity to traditional views and in anticipation of cognitive linguistics Bolinger claimed that such semantic extensions were metaphorical: "the phrasal verb", he said, "is a floodgate of metaphor" (Bolinger 1971: xii).

### 2 From localism to neolocalism

There are therefore circumstantial and more perennial reasons for the importance given to semantic analyses of particles and prepositions in cognitive linguistics: the persistence and revitalization of empiricist views, an interest for the language/perception interface, promoted by cognitive science, a long-standing interest in "particles"<sup>4</sup> on the part of linguists, the rise of semantics, of cognitive linguistics, and the avenues of research in lexical semantics opened up by prototype theory.

As far as I know, the first two studies of "particles" conducted in the spirit of cognitive linguistics were those of Lindner (1981), her PhD thesis, supervised by Langacker, on *out* and *up*, and Brugman (1981) on *over*. The latter, though less documented than the former, was to rise to celebrity especially through its being exploited in Lakoff's bestseller, *Women, fire and dangerous things* (1987),<sup>5</sup> in which the description of polysemous items, like *over*, was presented as an application of prototype theory (or rather, a simplified form of its psychological version). Lindner, Brugman and Lakoff all regarded spatial meanings as primary, a localist bent that was further shored up, in Lakoff's case, by his empiricist conception of metaphorical thinking.

These first studies and the numerous ones of the same style which followed share a number of characteristics: they are semasiological, taking as their object a single lexical item at a time, rather than a system (contrary to what was done by Bennett and Leech for instance); there is a tendency to neglect pragmatic and contextual factors, i.e. variations of use conditioned by a contrastive emphasis on a specific kind of information, given the system of expressions available in the language; as a consequence of the foregoing, polysemy tends to proliferate, and little is done to reduce it to a minimal set of features (unlike, e.g., in Pottier 1962; for a discussion, see Fortis 2009); finally, although cognitive semantics can be expected to live up to its name only if it can lay a serious claim to the psychological validity of its analyses, in the overwhelming majority of the cases, no attempt is made at devising a method other than introspective, e.g. no psychological experimentation is conducted (Sandra & Rice 1995 being a rare exception). On the whole, then, and if we abstract away from their philosophical backdrop, more or less explicitly articulated, little seems to distinguish these early studies from Aristotle (1957) on *ἐν*, Leibniz (1923 [1685–1686]) on *ad*, Harris (1773 [1751]) on *over*, and Condillac (1775) on *de*.

<sup>4</sup> The term "particles" is intended to be neutral, as far as the morphosyntactic category of these forms is concerned (adverbs, prepositions or adpreps). The morphosyntactic behavior and categorization of these forms are not a priority of the studies of cognitive semantics we are considering here.

<sup>5</sup>According to Peeters (2001), the book sold around 100,000 copies.

### Jean-Michel Fortis

### **4.3 Neolocalist accounts of thematic roles: Gruber and Jackendoff**

We now come to the last strand of neolocalism, the set of theories dealing with thematic roles. There are, among these accounts, commonalities which set them apart from the past localist tradition; this is not to say they are detached from this tradition, but their relation to past localism is quite variable and at times rather obscure. In spite of this difficulty, the observation that they all emerge in a rather narrow period, approximately spanning the years 1965–1975, cannot but incite us to have a closer look at possible lines of influence and transmission.

Chronologically, the first theory with a localist (or better, localistic) bent is the one proposed by Gruber (1965). In this study (his dissertation) Gruber sets himself the task of providing a *syntactic* (not semantic) representation that should state the kind of complements a verb can occur with, that is, whether it takes a direct object, and/or a prepositional complement, which kind of preposition is compatible with it, and the selectional restrictions that hold of its arguments. This syntactic level of representation or, as Gruber calls it, "prelexical" structure, consists in the marking of "deep" cases incorporated in verbs in the form of prepositions. For example, verbs which, like *obtain*, take a goal as subject are noted as TO V, the TO subject argument (to the left of the verb) being obligatorily incorporated. Note again that the interpretation of these prelexical structures is left to a semantic component, in conformity with the *Aspects* model; this merely attests to the reluctance of mixing syntax and semantics, in a time when generative semantics has not yet really caught on. It is in circumscribing the number of roles and identifying their semantic import that spatial relations show their usefulness. For Gruber, verbs expressing concrete motion ("positional transition") incorporate roles which, for some of them, are common to other, "abstract" fields; further, all abstract roles appear to have a concrete counterpart. These generalized roles are: *theme* (the located or moving entity), *source*, *goal*, *location, agent*. Generalized *source* and *goal* are for example instantiated in domains like "activity" (*The climate changed from being rainy to manifesting the dryness of the desert*)*,* "possession" (*John obtained a book from Mary, John gave a book to Bill*) *or* "abstract transfer" (*John reported to Mary from Bill that he would like to see her*). Importantly, such analogies do not give rise to any sweeping declaration on the cognitive primacy of space. In fact, Gruber explicity declares that on his view "there is no particular priority intended for the sense of concrete motion" (Gruber 1965: 48), that is, "motional" seems to be a substitute for "dynamic". In the terminology of Anderson, Gruber's theory should therefore be classified as belonging to the family of localistic analyses.

### 2 From localism to neolocalism

The conundrum Gruber confronts us with is the following: his text simply contains no reference to any previous study. We seem to be dealing with a rediscovery based on linguistic facts such as the cross-domain analogies cited above. Like Wüllner before him, Gruber (1967) analyzes the accusative as a goal-case, yet there is no evidence he was aware that he had predecessors. Nor was Jackendoff (1969), in all likelihood, when he first borrowed Gruber's list of "thematic relations".<sup>6</sup> The context in which this borrowing took place is rather puzzling too, since there was no clear motivation, from Jackendoff's own point of view, for adopting Gruber's thematic relations rather than Fillmore's cases. For thematic relations were introduced in Jackendoff's discussion merely to state a condition on reflexivization and the control of ∅ arguments in complement clauses, and to this purpose the hierarchy of thematic relations proposed by Fillmore would have served as well. Thus, the semantic part of the condition on reflexivization (the *thematic hierarchy condition*) stated that a reflexive should not be higher on the hierarchy AGENT > LOCATION, SOURCE, GOAL > THEME than its antecedent, which ruled out \**John was shaved by himself* (with *John* as theme, *himself* as agent) or \**I talked about Thmug to himself* (with *Thmug* as theme and *himself* as goal). The only remote allusion to a cognitive motivation can be found in the very general assertion that "to suppose a universal semantic representation is to make a strong claim about the innateness of semantic structure", and that "presumably the semantic representation is very closely integrated into the cognitive apparatus of the mind" (Jackendoff 1969: 1).

It should be noted that the thematic hierarchy condition was but one piece in a machinery designed to push back generative semantics, i.e. was aimed at supporting Chomsky's rival version of TG. The condition was regarded as an interpretative rule filtering out unacceptable interpretations, not as a semantic condition formulated at deep structure. And globally, the book (and its updated version of 1972) was intended as a refutation of the level-mixing infesting GS, and a rehabilitation of the role of surface structure in semantic interpretation.

The cognitive justification for positing local thematic relations is much more elaborated on when Jackendoff, from 1976 on, embarks on the description of autonomous semantic representations, with the aim of providing explicit procedures mapping these representations to syntactic structure. In essence, Gruber's generalized roles are now reformulated as "conceptual" predicates (first termed

<sup>6</sup> Jackendoff first acknowledged localism had a past in his 1983 book, referring back to Anderson as a source (Jackendoff 1983: 188). But even at this date, it may be doubted that Jackendoff had delved very deep into historical matters, since he declared, without further ado, that Gruber's essay offered the best demonstration for localist(ic) ideas to date.

### Jean-Michel Fortis

"semantic functions") specified with respect to ontological domains, but with cognitive primacy being granted to the "positional" domain and to "the innate conception of the physical world" (Jackendoff 1976: 149). For example, changes of state will be paraphrased by means of the predicate GO specified in the domain called "identificational" (Gruber's term), that is, concerning the inner properties of an entity, or "circumstantial", that is, referring to events. A simple illustration of the latter case is offered in the following example (Jackendoff 1976: 129), in which BILL takes on the status of theme by virtue of being the first argument of GO, while the other arguments are, respectively, a circumstantial source (unspecified) and a circumstantial goal:

	- b. CAUSE (JOHN, GOCirc (BILL, y, BILL SCREAM))

Just like in Gruber, Anderson and later Lakoff, a central argument in favor of such semantic functions was the observation that inferences valid in the spatial realm analogically carry over to non-spatial domains, with concomitant variations. For example, while GO from X to Y generally implies that the theme is no longer at X when it has reached Y, whether in space or when a change of state occurs, the inference does not hold in contexts referring to spatial extents such as *The road extended/reached from Altoona to Johnstown*, nor to abstract identificational extents as in *This theory ranges from the sublime to the ridiculous*.

In subsequent texts, this account of "conceptual structure" does not undergo any substantial change. In *Semantics and cognition* (Jackendoff 1983: 188), it is claimed to rest on a fundamental hypothesis of basic and presumably universal "semantic functions": "In any semantic field of [EVENTS] and [STATES]", says Jackendoff, "the principal event-, state-, path-, and place-functions are a subset of those used for the analysis of spatial location and motion." However, as localist as this statement may sound, Jackendoff's final position may be more adequately described as localistic. On a speculative note, he declares himself in favor of the idea that thematic structure is a generalized abstract organization which is not grounded in spatial metaphors, at least in synchrony and during ontogeny (phylogeny being another matter; Jackendoff 1983: 210).

### **4.4 Neolocalist accounts of thematic roles: Anderson**

Anderson's case grammar may be regarded as the most sophisticated localist account of thematic roles; it is remarkable in another respect: Anderson has always taken care to refer to past localist ideas (notably the doxography contained in Hjemslev's *Catégorie des Cas*) and he has often presented his theory as an

### 2 From localism to neolocalism

ongoing debate with rival models, two aspects which confer to his texts a historiographical dimension. The complexity of Anderson's theory prevents us from presenting but a rough sketch.<sup>7</sup>

His initial motivations (Anderson 1968) for introducing thematic roles (or, in his terms, *case roles/relations*) into the base component of grammar are close to those voiced by Fillmore (1966; 1968) around the same time. Like Fillmore, Anderson is unsatisfied with the configurational definition of grammatical relations proposed in generative grammar, a definition which he deems insufficient to capture generalities which hold within English itself as well as in the cross-linguistic comparison of accusative and ergative languages. In a nutshell, grammar should first and foremost be based on subcategorization, and subcategorization is expressed in terms of "deep" cases. At this early stage of Anderson's thinking, these cases are the Nominative (i.e. 'absolutive') and the Ergative. At a deep level, ergative and accusative languages are non-distinct. For English, various rules take care of the accusative realignment of arguments, in other words, and this is a leitmotiv of Anderson's theory, grammatical relations are a superficial phenomenon and should not be expressed at deep structure (*pace* generative grammar).

Although Anderson's case grammar will undergo some changes with the years, its essential characteristics are presented in his 1971 book, *The grammar of case*, and will remain largely unaltered (Fortis 2018a). His syntactic model consists of a dependency grammar in which predicates govern case roles taking as their dependents nominal expressions. Unlike Fillmore, and like Hjelmslev, he insists on the fact that case roles should exhibit systematicity, i.e. should form a system whose members are differentiated by features belonging to a semantic field (included their ∅ marked counterparts). This is where his "localist hypothesis" enters the stage: syntactic representations are to be "constructed out of predications that are locational or directional or non-locative non-directional" (Anderson 1973: 10).

In his *Grammar of case*, Anderson settles on a list of four fundamental subcategorial features of the functional category "case role": Nominative, Ergative, Locative and Ablative (capitalized here, in order to distinguish them from surface cases). This list is of course reminiscent of semi-localist accounts of the past, and in this Anderson may have been inspired by Hjelmslev and by timely remarks made by Lyons. Shortly before Anderson put forth his "localist hypothe-

<sup>7</sup>Andor (2018) provides an excellent and updated overview of Anderson's theory. The historical background and the evolution of Anderson's case grammar is the subject of my paper in the same volume (Fortis 2018a).

### Jean-Michel Fortis

sis", Lyons (1967) had drawn attention to cross-linguistic parallels between locative, existential and possessive constructions, and more to the point, his linguistic textbook of 1968 included a discussion which laid some importance on the distinction between grammatical and local cases and alluded to a minimal inventory of three essential local cases (AT, FROM, TO). Now, Anderson's own list of fundamental categorial features is unorthodox and perplexity may arise: what about the goal-case? Or the experiencer? And the instrumental case, which can obviously cooccur with an ergative, and which Fillmore had for this reason separated from it? Can a goal-case be conflated with Anderson's "Nominative"? But is it plausible to treat the Nominative and what is usually assimilated to "dative" uses as instances of one and the same case? On the other hand, since we are not dealing here with surface cases but with a localist description of semantic roles, it would be tempting to reduce agentivity (i.e. the Ergative) to the notion of source (i.e. the Ablative): is the Ergative not superfluous from a localist point of view?

We cannot go here into all the subtleties involved in the discussion of these issues and the solutions proposed by Anderson. Suffice it to say that one solution for capturing fine-grained semantic distinctions among roles is to posit complex subcategorial features: experiencers, for example, will be glossed as both ergative and locative, [ERG, LOC], and distinguished from simple datives, [LOC]. A further solution will be to introduce contextual rules: a goal will be coded as a Locative in the context of a predicate which subcategorizes for a source, i.e. an Ablative. An Instrumental in a circumstantial phrase will then be glossed as a variety of path, i.e. as [LOC, ABL]. Finally, as regards the Ergative and the Ablative, their semantic kinship prevails and in his 1977 book, where Anderson explicitly marks their commonality through the feature "source", restricting the name *Ablative* to locational sources. This modification leads to the following, penultimate version of his system of case roles (note the "nominative" has been renamed *Absolutive*).

Table 2.2: Features of cases in Anderson (1977)


This abstracts away from contextual effects, for instance that a Locative can be a goal in the presence of a locational source. No mention will be made either of the reanalyses prompted by this featural reformulation. Again, expounding the theory in its details, its posterior evolution and its final stage would take us too

### 2 From localism to neolocalism

far afield. Our first purpose was to convey to the reader a sense of Anderson's approach. However, a few concluding words need to be said about the localist(ic) spirit in which this theory was elaborated.

In Anderson's discussions, cognitive motivations have never loomed large, although he does occasionally hint at the importance of spatial cognition (e.g. Anderson 1994). This relative agnosticism is the reason why he initially described his theory as "localistic". He is well-informed about past localist accounts, and his initial semi-localist grammar evinces this familiarity, as does his foray into a localist description of aspect and tense, for which he refers back to Darrigol (1827) on Basque (Anderson 1973). His quest for systematicity, with a minimal list of basic features and a ∅-marked, or neutral, case (the absolutive) reflects a modern, structuralist perspective. Modern too is the fact that his main purpose is not to deliver a semasiological analysis of surface cases; like Fillmore, he envisages cases as roles abstracted from their surface realizations. This separation of levels and the formal, generative-like style of his analyses entail, as for Fillmore and Jackendoff, that rules mapping his deep structures to their surface realizations have to be explicitly formulated. Or in other words, his theory bears the mark of the generative period and has to be contextualized in the debates surrounding the relation of deep structure to semantics and grammatical relations.

### **5 Other perspectives**

In the foregoing, I have made no pretense at presenting all the facets of modern localism. I see at least three aspects that could have been examined more closely and that I will briefly mention here. First, we have dealt here with the clearest instances of localism, or what Desclés (1991) calls "pure localism", in which a direct mapping is posited between spatial relations and linguistic meanings. Some other varieties of localism were therefore left out. Desclés' own view would be more aptly characterized as a *partial localism*, in which spatial reference points (*repères*, a term also found in Culioli) are but a primitive stratum giving rise to a set of more abstract cognitive operations; in particular, states of affairs and events are situated with respect to enunciative reference points which cannot be regarded as purely spatial. Like Wundt, he also insists on the role of intuitive but non-spatial primitive notions, such as intentional control (Desclés 1991; 1993). In a way similar to Desclés' higher reference points, it might be questioned whether the visuo-spatial diagrams used by Langacker for the sake of representing semantic focus, headhood, tenses, aspects and modalities, the meanings of various relational expressions etc. are to be affiliated to localism. His first ver-

### Jean-Michel Fortis

sion of cognitive grammar went by the name of *Space grammar*, but space was invoked first and foremost because he had adopted a diagrammatic representation of the linguistic strata making up a clause. On the other hand, and from an early date on, he has proposed semantic analyses one would be tempted to describe as localist: in a 1975 paper, for example, possessive structures which can be glossed as 'x is by/at y' are declared to issue from a locative metaphor (Langacker 1975: 384–385). However, he denies that his diagrams are to be assimilated to the kind of visual imagery studied by psychologists (Langacker 1986: 6). The brand of diagrammatic localism advocated by Langacker is not the first of its kind. We feel that examining this family of theories would necessarily take us to territories unexplored here: the import of visual representations in linguistics and science at large; the relation of linguistics to place and space as themes scrutinized in physics, mathematics, philosophy and psychology, in other words, the role of the global epistemic context in the importance attributed to place and space, and the possible impact of changing conceptions of space on linguistic theories with a philosophical background. These issues overlap with those addressed in this volume by Chalozin-Dovrat, who claims that diagrammatic localism can be viewed as a way of furnishing scientific credentials by connecting linguistics to a category of Western science.

A second point we have not mentioned relates to what is designated today as "grounded cognition", a perspective according to which meanings are based on multimodal representations which are reenacted when these meanings are activated (Barsalou 2008). This trend partly takes its origin in cognitive linguistics and the conceptual metaphor theory of Lakoff and Johnson, but it has now developed into a body of research which extends beyond the boundaries of linguistics. Of relevance for us here is e.g. Boroditsky's psychological work on facilitation (or interference) caused by visually perceived motion on the processing of temporal expressions (e.g. Boroditsky 2001). Views of this sort do not only argue for what we designated as morphogenetic localism; they hypothesize that spatial conceptualization is, as it were, constantly active when spatial *and* metaphorically spatial forms are being processed.

Our third omission involves the junction between diachronic linguistics and the more or less explicit neo-empiricist views we have reviewed above. Our retrospective has shown there is no novelty in this diachronic twist. For historians, it may be significant that the later strand was revived by people who were connected with the tradition of historical linguistics, and were in a position to reinstate the notion of grammaticalization.<sup>8</sup> Bybee and Traugott illustrate this

<sup>8</sup>On cognitive linguistics as reactivating themes and perspectives of historical linguistics, see

### 2 From localism to neolocalism

revival (for a localist application, see Traugott 1975), which got further reinforcement from typological research in the spirit of Givón: we are alluding here to the work of what may be identified as the Cologne school, brought together in the 1970s on the occasion of a project on language universals. Among the linguists involved, Bernd Heine achieved perhaps the greatest fame. He would later join forces with cognitive linguistics and instill localist elements in his research on grammaticalization, showing for instance that possessive constructions result from different metaphors, some of them spatial or partly spatial: 'Y is located at X', 'Y is (intended) for XGOAL', 'Y exists from X', 'X grasps Y' etc. (Heine 1997). Of course, although the localist hypothesis fares well with respect to the semantic evolution of adpositions, it is by no means implied that spatial expressions are the only starting points. Counter-localist evolution may even be observed, but this appears to be exceptional (for a discussion on Romance, cf. Fagard 2010).

### **6 Conclusion**

In the sections above, we have seen that localist ideas were deeply entrenched in the history of linguistics and recurred in different contexts. However, in spite of this historical depth, some neolocalist views cannot be straightforwardly traced back to their antecedents. With the exception of Lyons' scattered reflections and Anderson's case grammar, the link which binds neolocalist theories of thematic roles with their past seems to be rather thin. Indeed, the legacy of past localist accounts is at times so dimly visible that one may be tempted to speak of reinvention, as we saw in the case of Talmy, Gruber and, in his wake, Jackendoff. In semasiological studies, especially those devoted to prepositions, the connection is somewhat more apparent. On the one hand, conceptual metaphor theory and prototype theory have clear links to a tradition we may describe as empiricist (from Aristotle to Locke and beyond). Empiricist tenets were in the intellectual horizon of linguists and were congenial to some before they could really be supported by extensive research of their own. For example, although it is not clear if empiricist philosophy was on Langacker's mind, it is rather striking to see him rediscover the Lockian problem of the general triangle, or use the Humean archetype of a ball hitting another to illustrate prototypical causation (Langacker 1991: 13). As for Lakoff and Johnson, philosophically oriented reflection was the backdrop, among other influences, to their conceptual metaphor theory (Johnson 1981); however, they seem to have both underestimated the extent to which

Geeraerts (2010).

### Jean-Michel Fortis

their views had been anticipated and exaggerated the novelty of their brand of empiricism, sometimes by omitting sources (Fortis 2018b). On the other hand, and beyond this revival of philosophical grammar, localist ideas substantiated by linguistic or psychological evidence collected in the pre-cognitive era had, as it were, percolated into semantics. We can refer the reader to what has been said above about Osgood, Nagy and Lindner.

Whether the history recounted here concerns very general ideas whose paths of transmission are for this reason difficult to map, or more specific claims with clear antecedents, we may at least hope that our depiction of their intellectual environment goes some way toward understanding why they could be put forward and received favorably from the 1960s on.

### **References**


### 2 From localism to neolocalism


### Jean-Michel Fortis


2 From localism to neolocalism


### Jean-Michel Fortis


### 2 From localism to neolocalism


### Jean-Michel Fortis


### 2 From localism to neolocalism


## **Chapter 3**

## **Spatialization of time as a scientification strategy: Beauzée, Guillaume and the conceptual school of cognitive linguistics**

### Lin Chalozin-Dovrat

Cohn Institute for the History and Philosophy of Science and Ideas, Tel Aviv University

The common thread interconnecting the work of Enlightenment grammarian Nicolas Beauzée (1717–1789), the typically modernist "psychomechanics" of Gustave Guillaume (1883–1960), and the conceptual school of cognitive linguistics emerging from the tumultuous 1970s American scene (e.g. George Lakoff, Leonard Talmy, Elizabeth Traugott, Ronald Langacker), is far from obvious. Yet, as I demonstrate in this essay, despite their dissimilarities these three moments in the history of linguistics exemplify a common theoretical gesture: construing grammatical time in terms of spatial concepts, which, I argue, functions in all three cases as a robust scientification strategy, meant to reinforce grammar's claim to scientificity

### **1 Introduction**

Various topics may attract the gaze of historians of linguistics, typically specific works or theories, or the entire œuvre of a particular linguist or school of thought. In this essay I wish to focus on a different kind of historical object: a specific strategy of scientification – a technique of transmuting knowledge into scientific knowledge. Scientification strategies may differ from one another in their distinctive epistemological practices, styles of reasoning, methodologies, models or operative concepts. With the rise of modern science and the celebration

Lin Chalozin-Dovrat. 2020. Spatialization of time as a scientification strategy: Beauzée, Guillaume and the conceptual school of cognitive linguistics. In Émilie Aussant & Jean-Michel Fortis (eds.), *Historical journey in a linguistic archipelago: Descriptive concepts and case studies*, 51–66. Berlin: Language Science Press. DOI: 10.5281 / zenodo.4269411

### Lin Chalozin-Dovrat

of classical mechanics as science's ultimate exemplar, the most diverse domains have experienced pressures to develop forms of knowledge akin to those of the natural sciences (Bod et al. 2014). Hence, scientification strategies are characteristic of the humanities and the social sciences in the post-Newton era. Grammar, no exception, has employed a gamut of strategies to devise knowledge in ways generally considered scientific. These strategies of scientification, I will argue, merit their own genealogies.

Time, or what we today call TAM (Tense–Aspect–Mood or Modality), is a salient domain in grammar, particularly dominant in the study of Indo-European languages, where it is visible through elaborate systems of verbal tense. Time is also a traditional object of metaphysics, natural philosophy, mechanics, and modern physics. The history of TAM theories shows that tense, particularly the systematic ways in which tenses co-relate, enabling speakers to establish order relations between events and the moment of speech, associated more readily with physical and mathematical theories than did mood or aspect. Tense relations construed as order relations are easy to represent using graphic representations emulating Euclidian geometry. This is one manifestation of the spatialization of time: the depiction of grammatical tense (or other expressions of temporal cognition) via graphical representations, reimagining temporal notions as relations in space. The theorization of time may also engage with spatial concepts via terminology that we recognize as "spatial." Whether in verbal or graphical forms, the spatialization of time invariably relates to a more or less explicit claim (depending on the theory) that linguistic time is in some manner spatial, or that we perceive time through the spatial categories of the mind. Such claims, most common in cognitive linguistics, often motivate diachronic and synchronic explanations highlighting spatial terms and metaphors or using geometry-like illustrations.<sup>1</sup>

The spatialization of time relies on the prevailing conception of space as a dimension conditioning metric relations among objects. This presently common idea of space, "the unlimited expanse in which everything is located," entered the

<sup>1</sup>Building on previous research in linguistics and psychology, John Lyons designated as "Localism" "the hypothesis that spatial expressions are more basic, grammatically and semantically, than various kinds of non-spatial expressions," and "that they serve as structural templates […] for other expressions" (1977: 718). In fact, Lyons' work helped popularize the "localist hypothesis" to which he fully subscribed (for a short history of localism, see Fortis 2018 and the chapter by the same author in this volume). Thus, we should not confound "Localism" and "Spatialization." While the term 'localism' refers to a semantic phenomenon (i.e. the alleged primacy of spatial cognition in language and thought), 'spatialization' refers to a phenomenon in the field of science: The tendency of certain scientific theories, notably in the language sciences, to represent time via spatial concepts. To the extent that localist trends in the late twentieth century seem to co-occur with specimens of spatialization of time (which are often used to promote scientification), localism is what a theory of spatialization may study.

### 3 Spatialization of time as a scientification strategy

general lexicon following the propagation of Isaac Newton's work in the eighteenth century (Chalozin-Dovrat 2019b). This was also when the modern notion of space began shaping grammatical thought. The term probably first appeared in the context of grammatical theory in James Harris's *Hermes* (1751) but did not evolve into a grammatical category until the twentieth century (e.g. "spatial prepositions"). Hence, understood as a meta-scientific object of study – pertaining to scientific theories – the spatialization of time is a modern phenomenon originating in modern conceptions of space first inculcated by classical mechanics and later by the theory of relativity.<sup>2</sup>

This essay discusses in tandem three seemingly unrelated moments in the history of linguistics: the work of the Enlightenment grammarian Nicolas Beauzée (1717–1789), the typically modernist "psychomechanics" of Gustave Guillaume (1883–1960), and the scientific project of the conceptual school of cognitive linguistics emerging in 1970s America through the works of Leonard Talmy (b. 1942), Ronald Langacker (b. 1942) George Lakoff (b. 1941), Elizabeth C. Traugott (b. 1939), and others. While Guillaume was most probably well familiar with Beauzée's work (Fournier 2013), and Traugott (1975) designated Guillaume a prime reference on the spatialized nature of grammatical time, these three theoretical moments do not derive one from another in any significant historical sense. However, as I shall argue, despite their obvious differences these three theoretical gestures share similar scientific motivations and deploy similar tactics in realizing them. Thus, Beauzée, inspired by both Cartesian and Newtonian perceptions of natural philosophy, enlisted the metaphysics of time and space to reproduce the successes of physical science in the field of grammar. Guillaume's theory of the spatialization of time emerged in the heyday of relativity theory, when physical theory aroused considerable public interest. Applying the interchangeability of time and space to grammatical phenomena, Guillaume hoped to devise a theory of the psychological mechanisms behind linguistic temporality. Finally, relying on its close relations with other theories of human cognition (such as language acquisition or cognitive psychology), the conceptual school of cognitive linguistics strove to construct its theories on general cognitive principles. Grounded in research on spatial cognition, the spatialization of time bolstered the scientificity of cognitive grammar while enhancing its affinity with the general project of cognitive science. In the conclusion, I shall suggest that this retrospective genealogy

<sup>2</sup>Twentieth-century cases of time spatialization often blend different physical theories of space, theories of spatial perception, and various modern and premodern conceptions of space, place, location, distance, and extension. Such fusions, combining the most diverse theories and concepts, may reflect attempts to reach an effective scientification strategy.

### Lin Chalozin-Dovrat

of the spatialization of time uncovers a distinct form of continuity in the history of linguistics (Auroux 1980; Colombat et al. 2010) reproducing and regenerating not only past ideas, theories, and concepts, but also successful scientification strategies.

### **2 Nicolas Beauzée (1717–1789)**

In 1756 Beauzée, a grammar professor in the *École royale militaire*, joined the cohort of writers for Diderot and D'Alembert's *Encyclopédie* (Le Guern 2009). Over the following decade he contributed more than 140 entries to the *Encyclopédie* on various grammatical concepts, including a substantial article presenting his renowned theory of tense (Beauzée 1765). As was customary, the *Encyclopédie* treated grammar as a branch of the communicative arts (D'Alembert & Diderot 1751), while Beauzée sought to elevate grammar to the rank of science. Enthused by his era's scientific achievements and inspired by natural philosophy, he aspired to salvage grammar from its status as a premodern form of knowledge primarily associated with rhetoric and pedagogy. Building from the Port-Royal *Grammaire générale et raisonnée* (Arnauld & Lancelot 1660), Beauzée both revived the notion of general grammar and strove to upgrade it to "grammatical metaphysics:" a sure foundation for a science of grammar modeled upon Cartesian precepts (Chalozin-Dovrat 2019a).

His quest for "grammatical metaphysics" represented Beauzée's hope to attain in the study of grammar the scientific rigor and clarity achieved by natural philosophy. Fully adopting the Cartesian model establishing metaphysics as the necessary foundation for any science, Beauzée asserted that "only metaphysics, that is, the most thoughtful and analytic examination of abstract ideas," can discover the true principles of general grammar (Beauzée 1767: vol.1: xxxiij–xxxv). Hence, some objects of inquiry, such as tense, could not fit within grammarians' "confused notions," but required, Beauzée advocated, the light of metaphysics (Beauzée 1767: 96). Time was a traditional object of metaphysics and natural philosophy and understandably necessitated a scientific mode of inquiry. While Beauzée did not overtly claim that grammatical time and natural time were one and the same, his tense theory implicitly depended on the presumed identity between tense and time. "Let me resort here to the blazing torch of metaphysics," he implored in opening his entry on grammatical tense, "the only one that can indicate all the ideas comprehended in the nature of *tenses*" (ibid.).

Beauzée obtained the metaphysical basis for his theory from the definition of time by the Cartesian academician Étienne Simon de Gamaches. "Time is the

### 3 Spatialization of time as a scientification strategy

very succession attached to the created being's existence," postulated Gamaches in his work of astronomical physics (1740: 28), a formulation Beauzée converted into a geometry-like theoretical apparatus. "[T]he successive existence of beings," he argued, "is the only measure of time that is within our reach" (Beauzée 1765: 96). Measuring this successive motion required breaking the free flow of existence using fixed points of reference, which Beauzée termed "epochs" (*époques*) – from the Greek *ἐπέχειν* 'to stop'. The portion of time demarcated between two such "stops" – between beginning and concluding epochs – he termed a "period," bounded, he asserted, on all sides "just like a space around which one can turn" (ibid.). This graphic depiction led Beauzée to a general definition of tense as a system of reference in which "tenses are verb forms expressing different existential relations to the various epochs that one can imagine in time" (ibid.).

Beauzée's analysis relied on three basic distinctions or "divisions" characterizing the different tenses:


### Lin Chalozin-Dovrat

These basic distinctions allowed Beauzée to analyze the entire French tense system as a complex reference system *à la* geometry, establishing order relations between the period, or its two demarcating epochs, and the moment of speech. This three-layered reference system (Auroux 1991; Fournier 2013) determined the full tableau of time relations characterizing each verb form. Ultimately, Beauzée's notion of tense equals a "constellation" of time relations: a typical arrangement of the possible reference points and their respective positions. Conceptualizing each tense as if it were an astronomical constellation, representing a specific configuration of temporal relations (distinct from other tenses' configurations) necessitated a complex procedure of spatialization to transform the complete tense system into an array of such constellations. While Beauzée's precise concept of space remains tacit, this strategy of spatialization, reimagining tenses as configurations of positions and relations, is the key to his tense theory.

Beauzée's mastery of time spatialization demonstrates his theoretical agility, transposing one type of knowledge into a different epistemological setting in a meaningful way. The great originality of Beauzée's tense theory resides in his selecting the epoch of comparison – the point of reference that enables us to break the free flow of existence and situate events and actions in time – as the main point of reference calibrating the system. This choice may seem counter-intuitive, as traditional theories of grammar construe tense in relation to the moment of speech – that is, in relation to the speaker rather than to metaphysical notions such as time or existence. "I believed," admitted Beauzée, that "I should treat the principles of language as we treat those of physics, geometry, those of all sciences" (1767: xvi, vol. 1). Abandoning the privileged point-of-view of the speaker while relying on the spatial example of astronomical physics allowed Beauzée to analyze the tense system as an objective natural phenomenon, entitled to its own genuine science.

### **3 Gustave Guillaume (1883–1960)**

Guillaume's unconventional career path – transforming him from a young bank clerk into a groundbreaking linguist (Valin 1982) – gives a human face to his original but somewhat eccentric work.<sup>3</sup> Guillaume's theory, "The psychomechanics of language," aspires to uncover the "systematic totality" (Guillaume 1965 [1945]: 15) behind language – the psychological machine underlying the linguistic sys-

<sup>3</sup>Often considered opaque, Guillaume's work has nonetheless acquired many disciples. The resources list at the *Fonds Gustave Guillaume*'s site testifies to the rich on-going research on Guillaume's theory. See: http://www.fondsgustaveguillaume.ulaval.ca/

### 3 Spatialization of time as a scientification strategy

tem. Guillaume's major theoretical challenge was thus providing a glimpse into that which cannot be seen: the mechanisms conditioning language without leaving direct evidence in discourse. While Guillaume was Antoine Meillet's protégé and heir to a structuralist lineage, his ambition to expose the systematic apparatus behind observable linguistic facts links his theory to cognitive trends in linguistic research (Puech & Savatovsky 1997). As contemporary historians of linguistics have commented, it seems "Guillaume merely lacked the term *cognition*" (Bottineau 2006: 41).

The production of linguistic temporality – the way linguistic systems construct and express time – stood at the center of Guillaume's work (Joly & Lerouge 1980). Indo-European linguistics has traditionally considered verb and time inseparable, thus facilitating immediate association between the notions of verb, time, and action. Building on this conceptual nexus, Guillaume aspired to ascertain the psycholinguistic mechanisms transforming abstract notions of time into verbal images. "Time is so abstract," maintained Guillaume in his first book on psychomechanics, *Temps et verbe* (1929), "that its simple rendering as a clear image already suggests powerful concretization" (Guillaume 1965 [1929]: 7). Through his notion of "time-image" (*l'image-temps*) Guillaume conceptualized verbal time as a dynamic tension between time's abstract nature and the visual concretization of temporality in language. This psycho-mechanic procedure, consisting in visualizing time, is accomplished, according to Guillaume, by means of spatialization:

The human mind is thus made that it has the experience of time but does not have its representation. It must seek it through constructive and descriptive means which are spatial in character. The linear representation of time that flees is one of these means: it is already in its primary and […] fundamental simplicity a certain *spatialization of time*. (Guillaume 1965 [1945]: 17)<sup>4</sup>

Guillaume's "spatialization of time" is a psycholinguistic mechanism converting the abstract experience of time into concrete spatial representations – geometric relations encoded in language. Guillaume's use of the terms 'concrete' and 'abstract' requires elucidation, as it marks a break from a tradition comprehending these notions as semiotic modalities. Accordingly, considering an object and

<sup>4</sup> "L'esprit humain est ainsi fait qu'il a l'expérience du temps, mais n'en a point la représentation. Il lui faut la demander à des moyens constructifs et descriptifs qui sont de l'ordre de l'espace. La représentation linéaire du temps qui fuit fait partie de ces moyens : elle est déjà dans sa simplicité première et […] fondamentale […] *une certaine spatialisation du temps*." (Guillaume 1965 [1945]: 17; emphasis is mine, LCD)

### Lin Chalozin-Dovrat

its mode united, as the senses experience them (e.g. *the white paper*, *the round box*), would yield the "concrete sense," while abstracting the mode from the object or the object from the mode (e.g. *whiteness, paper*) would produce the "abstract sense."<sup>5</sup> Guillaume, however, employs the terms 'abstract' and 'concrete' as properties: time is thus inherently "abstract" and space "concrete," which ultimately signifies "visual" – perhaps because classical mechanics often uses graphical illustrations to represent mathematical relations in space.

Figure 3.1: The Latin spatialization of time on three dimensions. From Gustave Guillaume (1965 [1945]: 37).

Guillaume translated his ideas about the relations between time and space into vigorous geometry-like instruments, purportedly reconstructing the mental apparatus of psycholinguistic temporality. According to Guillaume's diagrams, pre-verbal time comprises *n* dimensions, just like physical space. In this diagram, for example (see Figure 3.1), Guillaume presented his thesis about the spatial relationships between tenses and moods in Latin. The diagram's top section shows the infinitive forms: the present infinitive *amāre* 'to love' and below it the perfect

<sup>5</sup>Eighteenth-century cognitive theory did not use these terms as properties (e.g. "time is abstract;" "space is concrete"), but as modalities. We can conceive a body without its form, or whiteness without a body (*res absque modo* or *modus absque re*) explained French linguist César Chesneau Dumarsais in his cognitive theory of tropes (1730). Employing the terms 'concrete sense' and 'abstract sense' as semiotic modalities allowed Dumarsais to understand abstraction as "a sort of separation made by thought" (1730: 260).

### 3 Spatialization of time as a scientification strategy

infinitive *amāvisse*. The central section represents the subjunctive, a binary relationship between two tenses and two aspects: the present and perfect subjunctive (*amem* and *amāverim*), and on top, the same relationship between the anterior actions: the imperfect *amārem* 'I would love to', and the pluperfect *amāvissem* 'I had loved to'. The lower section of the diagram represents the indicative, the basic form of the verb, with the simple (*amo*) and perfect (*amavi*) forms. The same order relations appear in the past (to the left: the imperfect *amabam* and the pluperfect *amaveram*) and the future (to the right: the future *amabo* and the future perfect *amavero*). The diagram as a whole represents *chronogénèse*: the mental operation of generating time-images. Each section stands for a *chronothèse* (initial, medial, and final): the static aspects of the chronogenetical dynamics, fixing the mental time-imagery (Boone & Joly 1996).

Guillaume's inventive terminology, highlighting the dynamic dimension of language, borrowed extensively from physics, geometry, biology and physiology (Bottineau 2006). As Valette (2003) has shown, starting in the 1920s, around the time he expressed his first ideas about psychomechanics, Guillaume gradually adopted more and more scientific terms at the expense of his formerly more richly philosophical vocabulary. The spatialization of time, with its architectonic illustrations and mechanical allure, was the focus of much of Guillaume's effort to devise an expressly scientific theory of the systematic apparatus underlying language. It allowed Guillaume not only to illustrate the structure of tense systems as he understood it, but also to theorize the cognitive mechanisms motivating grammatical time. In that sense, the spatialization of time, construed here not as a cognitive mechanism but as a strategy of scientification, enabled Guillaume to formulate a proto-science of psycholinguistics.

### **4 The conceptual school of cognitive linguistics (1970s to the present)**

By 1975, as similar ideas about the spatial nature of grammatical time were gaining currency among various linguistic schools, the "spatialization of time" was no longer an oddity. "[T]he spatial nature of temporal expressions in many languages is widely recognized," stated the American linguist Elizabeth C. Traugott in a widely cited article (1975: 207). Inspired by the pioneering works of Eve V. Clark (b. 1942, in 1971) and Herbert H. Clark (b. 1940, in 1973) on children's language acquisition – establishing that the same principles that guide general perception determine the acquisition of temporal and spatial vocabulary – Traugott brought together time-honored diachronic methodologies and novel cogni-

### Lin Chalozin-Dovrat

tive sensibilities. Well versed in the writings of Continental structuralists, she selected a quote by Guillaume for her article's epigraph, the same quote I cited above.

Building on existing diachronic datasets, Traugott argued that the grand majority of English prepositions derive from locatives, thus demonstrating "[t]he locative character of temporal relations" (1975: 209). Thence, she embarked on ascertaining "whether there are any constraints on the selection of spatial terms for temporal relationships" (1975: 207). This undertaking, attempting to determine the cognitive logic behind the time/space nexus in language, relied on a definition of linguistic time ("the expression of our experience of time") that Traugott distinguished from both physical and calendrical time (1975). Equipped with a distinct scientific object and a well-defined scientific task, Traugott's adaptation of the "spatialization of time" theme set an example for future research in the functionalist and cognitive trends.

Prepositions also attracted the attention of other linguists, such as Leonard Talmy, for whom spatial cognition is a "fine structure", a fundamental conceptual subdivision of language (1983: 225). Talmy used topological schemata to extrapolate the cognitive principles accounting for the constraints on the distribution of prepositions. Emphasizing the role of spatial cognition in semantics enabled Talmy and other theoretical linguists to connect their work with general cognitive principles, thus linking it to other domains of cognitive science and particularly to research on visual perception and motion.

For George Lakoff (1993) the relations between our notions of time and space are determined by conceptual metaphor. Together with Mark Johnson, he established metaphor as an all-pervasive cognitive mechanism producing concepts and governing semantic change (Lakoff & Johnson 1980). According to their theory of conceptual metaphor (CMT), we conceptualize common abstract concepts, such as TIME, ACTION, or CAUSATION via metaphor, generating non-figurative ideas according to concrete experiences. "Abstract reasoning," argued Lakoff, "is a special case of image-based reasoning," mapping mental images from concrete experiential domains onto abstract ones (Lakoff 1993: 229). Conceptual metaphor builds on visual principles, projecting the structure of one domain onto another, reproducing concepts in new domains, and eventually generating new knowledge. According to Lakoff, the same cognitive mechanisms participate in conceptualizing time: Lacking a biological "detector" of time, we perceive it in terms of objects, locations, and motion in space (Lakoff 1993: 218).

I employ the term "conceptual school of cognitive linguistics" to refer to those linguists who place concepts and conceptualization at the center of their theoretical work. Cognitive linguists generally portray space not as an abstraction or as

### 3 Spatialization of time as a scientification strategy

a set of cognitive faculties but as a natural concept. Ronald Langacker's theory, titled *Space grammar* (1982; 1987), is a case in point, treating space as a general conceptual apparatus governing grammatical rationality. For Langacker the priority of space is all-encompassing, applied – but not limited – to grammatical temporality. Langacker proposed a particularly elegant specimen of time spatialization when he attempted to provide a semantic construal of lexico-syntactic differences. Hence, the following pair of sentences:


exemplifies, according to Langacker, two modes of temporal perception governed by a differential visual mechanism (1987: 146, example 19). The verbal form in (1) expresses *sequential scanning*: a series of event images whose continuous progression replaces one temporal configuration with another. The nominal form in (2) stands for a procedure of *summary scanning*: an additive progression of the event image which is accessible simultaneously. While the cognitive modality of sequential scanning simulates a consecutive series of snapshots, during summary scanning snapshots instantly add up to form one visual synthesis. The conceptual difference between conjugated verbs and nominalization, proposes Langacker, is essentially visual: the verbal form expresses the perception of a series of images as a moving sequence (like a motion picture composed of still frames), whereas the noun sums up the complete spatial information in one image. In fact, the semantic difference between the two lexico-syntactic forms reflects the distinction between two general principles underlying spatial cognition and event processing.

As Langacker has explained, the power of space – both as an epistemic, explanatory principle and as a comprehensive cognitive mechanism governing grammar – stems from its special status in perception:

[I]t hardly seems appropriate or feasible to consider three-dimensional space as a concept definable relative to some other, more fundamental conception. It would appear more promising to regard the conception of space (either two- or three-dimensional) as a basic field of representation grounded in genetically determined physical properties of the human organism and constituting an intrinsic part of our inborn cognitive apparatus. That is, our ability to conceive of spatial relationships presupposes some kind of representational space creating the potential for such relationships, but it is doubtful that conceptual analysis can go beyond positing this representational space and elucidating its properties. (Langacker 1987: 148)

### Lin Chalozin-Dovrat

Hence, according to Langacker, space is a primitive concept that no other more fundamental idea can ascertain or define. Genetically determined, the conceptual apparatus of space is itself a "representational space," enabling us to represent spatial relationships. Mirroring its representational competence in a sort of *mise-en-abyme*, this "representational space" is an innate feature of our cognitive mechanisms grounded in our physiology. This convoluted theory of representation, apparently self-reflecting to infinity, has one clear boundary: the theory's concept of space begins and ends with the biological necessities constraining the species' physical and cognitive properties, permitting no further conceptual analysis.

Cognitive semantics has evolved significantly since the late 1980s, popularizing the idea that linguistic temporality is based on spatial perception. Attempts to advance the thesis of the spatial nature of temporality in language and cognition have often proclaimed subsidiary theoretical goals – such as ascertaining the universality of temporal experience (e.g. Alverson 1994); advancing a refined research program for linguistic typology (e.g. Haspelmath 1997); or demonstrating the affinities between language, thought, and cognitive functions such as motor action (Casasanto & Boroditsky 2008). In its various forms, spatialization of time has also served as a more general strategy to enhance grammar's authority as a scientific form of knowledge.

### **5 Conclusion**

Beauzée's metaphysical tense grammar; Guillaume's "psychomechanics" of verbal time; late twentieth-century theories about spatial cognition's role in linguistic temporality – these three theoretical gestures emerged in very different historical and epistemological contexts. Within these multifaceted contexts the spatialization of grammatical time changed in its foci, forms, and functions. Yet one purpose remained unchanged throughout the centuries: to promote and enhance grammar's scientification.

Beauzée was an activist for the cause of the scientification of grammar at a time when the sole established science, in the modern sense, was classical mechanics. Manipulating the similitude between natural and grammatical time, Beauzée conceived his theory of tense with the ambition of changing the course of general grammar. With the example of astronomical physics in mind, Beauzée reconstructed the tense system as a multilayered system of reference, theorizing each tense and the entire tense system as if they were objective constellations, natural phenomena disengaged from discursive considerations. Thus, spatializing

### 3 Spatialization of time as a scientification strategy

time was for Beauzée a strategy for transforming "an old problem," such as the coherence of the tense system, into "a new science."

Guillaume's scientific project developed amid a functioning modern disciplinary environment. However, his interest in the dynamics of psycholinguistic mechanisms ill-fitted the existing framework of normal science, necessitating, in Guillaume's view, a new scientific language. The idea of time as the fourth dimension of spacetime was then attracting considerable public attention, especially following Einstein's visit to Paris in 1922 (Biezunski 1987). The growing interest in modern physics most probably also inspired Guillaume's increasingly science-based terminology. Thus, Guillaume's "spatialization of time" tied up many loose ends: his interest in the dynamics of verbs and the linguistic expression of time; the need for a psycholinguistic mechanism to associate linguistic temporality with mental images; and the desire to articulate the general laws of language production in a scientific way.

By the 1970s the stakes were very different. The main theoretical concerns of cognitive linguists in the conceptual trend arose within the context of the "Linguistics Wars." For those hoping to shift linguistic theory's focus from syntax to semantics, conceptual paradigms like that proposed by the time/space nexus in language came in handy. Space had another significant advantage in the competition between linguistic theories: It linked theoretical linguistics with other fields of cognitive science, enhancing both its empirical and its theoretical standing. Spatializing time allowed cognitive semantics to demonstrate how general principles of cognition propose semantic explanations while reshuffling traditional distinctions like that between lexicon and grammar.

There is wide agreement among historians of linguistics as to the high degree of continuity in the history of linguistic ideas and the discipline's low "rewriting rate" (*taux de réinscription*) – a concept Sylvain Auroux (1980) introduced to emphasize that linguistics progresses without eradicating past theories via paradigmatic change. The continuity traced in this study is not historical per se: it is difficult to measure the degree of influence Beauzée's work had on Guillaume's "spatialization of time," and Traugott's reference to Guillaume was clearly superficial. Yet, I would like to argue that this genealogical study of the spatialization of grammatical time suggests that the history of scientificity (Auroux 1990) and scientification cannot solely rely on isolated evidence documenting the transmission of knowledge by individuals. It must also be attentive to scientific motivations and to the concrete webs of interests wherein these motivations are situated. Such a theoretical framework could enrich our understanding of the modes of continuity and progression typical of linguistics and the way it functions as a

### Lin Chalozin-Dovrat

scientific discipline. This study of time spatialization shows that, whenever the modern concept of space was available to them, linguists employed the novel practices and opportunities it opened up to promote the scientification of grammar. Understood as a strategy, spatialization of time may illuminate the ways by which scientific memory persists and prevails.

### **References**


### 3 Spatialization of time as a scientification strategy


### Lin Chalozin-Dovrat


## **Chapter 4**

## **Grammar and graphical semiotics in early syntactic diagrams: Clark (1847) and Reed-Kellogg (1876)**

Nicolas Mazziotta

Université de Liège

This paper shows that the authors who first used diagrams to represent syntactic structures chose graphical conventions that constrained the way they could represent their analyses. Some graphical conventions may look similar despite following different rationales. Conversely, they may also look different and yet be grounded in the same logic. This second possibility becomes obvious when comparing the diagramming systems proposed by Clark (who uses aggregated bubbles) and Reed & Kellogg (who use strokes). Nevertheless, two-dimensional objects such as bubbles offer more varied layout possibilities for drawing diagrams than one-dimensional objects such as strokes do. Consequently, authors have to add the representation for abstract concepts such as inclusion by means that are compatible with the basic objects they use.

### **1 Introduction**

This paper focuses on the graphical depiction of syntactic analysis. I will compare two early diagramming systems in order to show how graphical formalisms put constraints on what can be expressed, because of the spatial properties of the graphical units. I will focus on two aspects: how graphical devices represent linguistic units and how these graphical devices interact on the plane. The diagrams under study were created in the United States, and date from the 19th century (Brittain 1973), thus preceding the advent of modern syntactic trees that

Nicolas Mazziotta. 2020. Grammar and graphical semiotics in early syntactic diagrams: Clark (1847) and Reed-Kellogg (1876). In Émilie Aussant & Jean-Michel Fortis (eds.), *Historical journey in a linguistic archipelago: Descriptive concepts and case studies*, 67–81. Berlin: Language Science Press. DOI: 10.5281/zenodo.4269413

### Nicolas Mazziotta

are in use today. The introduction of syntactic diagrams occurred at a time when grammatical description in America had undergone a paradigm shift from wordcentric to syntactic analysis.<sup>1</sup> Scholars began to focus on the "deep structure" of the sentence. Pioneers made extensive use of diagrams to encode the structural order of syntactic relations and discarded the linear succession of words.

In section §2, I will introduce the formal and semiotic concepts that I use to describe the diagrams. In section §3, I will present early diagramming systems, especially the ones produced by Stephen W. Clark (1810–1901) on the one hand, and Alonzo Reed (d. 1899) and Brainerd Kellogg (1834–1920) on the other hand. Section §4 compares the graphical representation of three specific structures in these two systems.

### **2 Visual depictions of syntax**

Visual depictions of syntactic analyses are diagrams, i.e., icons that represent relations and allow for creative thinking (Peirce 1994 [1931–1935], cf. Chauviré 2008: 36–42). Syntactic diagrams represent words and grammatical concepts, and depict the way they interact. Since words and grammatical concepts are just *thoughts* (albeit highly organized ones), they cannot be expressed without a way of making them perceptible. Syntactic diagrams achieve this goal by graphical means. In section §2.1, I introduce the concepts of *reification*, *graphical entities* and *spatial configurations*, which I rely on to describe the way linguistic analyses are graphically represented. Section §2.2 illustrates the variety of diagrams.

### **2.1** *Semiotic concepts***: Reification, graphical entities, spatial configurations**

To begin with, let us consider Figure 4.1, depicting a classical immediate constituent analysis ("ICA") of (1).<sup>2</sup>

(1) The fur warms a bear.

Each word is represented by itself (by self-reference) in the form of concatenated letters. Conceptual units such as constituents are represented in the same

<sup>1</sup>Aarts & McMahon (2006: 76). Arguably, such a paradigm-shift is already achieved in the works of several French scholars at the beginning of the 18th century (see Mazziotta 2020).

<sup>2</sup>The sentence is taken from Clark (1870: 44) for comparison purposes, but the analysis is mine. It should be noted that the diagram depicts a simple ICA in order to illustrate the grounding rationales of such a representation for comparison purposes. Figure 4.1 is not seen here as a state of the art representation.

### 4 Grammar and graphical semiotics in early syntactic diagrams

Figure 4.1: Sample ICA tree

way. On the other hand, strokes represent part-whole relations. I call *reification* such a discrete representation of a concept (*to reify* 'to turn into an object'; Kahane & Mazziotta 2015). When a concept is reified in a graphical environment, it becomes a *graphical entity* (Groupe µ 1992), i.e. a sign that we are able to perceive, to delimit, and to describe from a visual point of view. Graphical entities are thus bound to conceptual units.

Graphical entities interact with one another by forming *spatial configurations*: in Figure 4.1, one can understand that *warms* and the second NP are immediate constituents of VP because they are both placed below VP, and they are each connected to VP by a stroke.<sup>3</sup> Spatial configurations express conceptual relations in the diagram.

### **2.2 Variety in the graphical representation of analyses**

It is well known that there are various ways to represent the relations between words. The same conceptual units/relations can be reified in different ways. Moreover, graphical entities and spatial configurations are competing means of expressing analyses of the same kind. To illustrate this, I will compare the two diagrams in Figure 4.2: (a) is the very first syntactic diagram, proposed by Gustav Billroth (1808–1836) in 1832 (the analysed sentence can be literally translated as: 'Miltiades, leader of the Athenians, restored nearly suppressed freedom to the entire Greece, in the battle at Marathon'); (b) is a diagram from *Graded lessons in English* (1879 [1876]) by Alonzo Reed and Brainerd Kellogg.

Despite the fact that both diagrams represent words and the way they interact in the sentence, and despite the fact that both diagrams make use of concatenated

<sup>3</sup> For the sake of simplicity, unless otherwise stated, I will use shortcuts such as "NP" for "the graphical entity that reifies NP".

Nicolas Mazziotta

(b) Reed & Kellogg (1879 [1876]: 43)

Figure 4.2: Various values of strokes

characters and strokes, their rationales are very different. In Figure 4.2a, words are reified by concatenated characters, whereas in Figure 4.2b, words are reified by a combination of concatenated characters and strokes below them, i.e. *labeled strokes*. Both diagrams express how words are grouped into larger structures, such as phrases and sentences. In Figure 4.2a, strokes reify the grouping of words in a way that we are very accustomed to: many diagrams of all sorts use similar conventions. On the contrary, unless we learned to read Figure 4.2b at school, it takes a while to get accustomed to the fact that strokes in Figure 4.2b do not reify syntactic relations as they do in Figure 4.2a. Actually, in Figure 4.2b, relations are mostly represented by the relative positions of the words, i.e., by spatial configurations.

This comparison demonstrates two important features of graphical entities. Firstly, they are complete signs with a specific form of expression and a value: using the same form of expression does not automatically entail that the same value is implied, and *vice versa*. The configurational properties of the entities differ accordingly: in Figure 4.2a, strokes are drawn *between* words, whereas in Figure 4.2b, they are drawn *under* the words. Secondly, it is up to the creator of

### 4 Grammar and graphical semiotics in early syntactic diagrams

the diagram to choose between reifying his analysis or representing it by means of configurations. Comparing Figure 4.2 to Figure 4.1 delivers another piece of evidence for the demonstration: in Figure 4.1, part-whole relations are reified independently by strokes. Consequently, some diagrams may look similar but are conceptually very different indeed, and conversely, some diagrams that look different are actually similar. Formal and semiotic description helps identifying units inside a given system and comparing units across systems.

In the next sections, I will focus on diagrams that share similar rationales with the ones in Figure 4.2b. In section §3, I will show how similar rationales are shared by diagramming systems that make different representation choices. In section §4, I will show that graphical choices constrain what part of the analysis can be expressed, and further force the reification of abstract concepts.

### **3 Early syntactic diagrams in the United States: Basic rationales**

Most diagramming systems from the 19th century reify words, but hardly any relation. Section §3.1 describes the fundamental rationales of the first of these diagramming systems, proposed by Clark in 1847, and compares it with the Reed & Kellogg system. Section §3.2 briefly illustrates the continuity between the two systems with intermediary diagrams that were proposed by other authors.

### **3.1 Clark's seminal** *Practical grammar* **and the successful Reed-Kellogg system**

In Clark's system (1847; see Mazziotta 2016), words are reified by means of labeled bubbles that aggregate with one another (Figure 4.3). The syntactic relations between words are expressed by the relative positions of the bubbles.

(2) The king of shadows loves a shining mark.

In Figure 4.3, which depicts the analysis of (2), horizontally arranged bubbles express that the sentence is a combination of a subject, a predicate and, optionally,

Figure 4.3: Sample bubble diagram (Clark 1847: 23)

### Nicolas Mazziotta

an object. These elements can be complemented by adjuncts, which are represented by vertically connected bubbles. Note that adjuncts can also have adjuncts of their own: the representational convention is recursive. As it can be observed at the bottom left of the diagram, the representation of the terms of prepositional phrases combines two ways of aggregating bubbles. Prepositional phrases are a combination of a preposition (called the *leader* in Clark's terminology) and a noun (the *subsequent*). The leader is represented by a bubble that connects the phrase complements vertically to the word, and is horizontally arranged with its subsequent.

Similar ideas were adopted in the manuals written by A. Reed and B. Kellogg, albeit with alternate graphical means of expressing the analyses. As already highlighted in section §2.2, Figure 4.2b is a diagram containing words reified by labeled strokes that aggregate with one another. The sentence is a combination of a subject, a predicate and, optionally, an object. It is represented by horizontally arranged strokes. These elements can be complemented by modifiers, which are reified by vertically connected strokes. Prepositional phrases are a combination of a preposition and a noun: a vertically connected stroke for the preposition, which in turn is horizontally arranged with the noun.

### **3.2 Continuity between Clark (1847) and Reed & Kellogg (1879 [1876])**

The theoretical assumptions in Clark (1847) and Reed & Kellogg (1879 [1876]) are roughly the same: both systems acknowledge a distinction between principal parts and adjuncts/modifiers, and posit a hybrid status for prepositional phrases. From a graphical perspective, both systems reify words by means of labeled geometric shapes that aggregate with one another either horizontally or vertically. Such similarities demonstrate some continuity between the diagramming systems. Additionally, there was a huge variety of diagramming systems that were proposed by different authors between Clark's grammar and the *Graded lessons of English* of Reed and Kellogg. Many of them relied on variations of the same principles, demonstrating the popularity of Clark's ideas. Figure 4.4 illustrates this with diagrams collected by Brittain (1973: 37, 56, 75).

It is striking that each author chooses a different geometric shape to reify words, but the way the entities are arranged remains mostly identical as far as the basic structures are concerned (horizontal arrangement of the principal parts and vertical arrangement of the adjuncts). The core descriptive stance of these diagrams is *sparsity*: whenever it is possible, only the words are reified, and any other piece of information is expressed by configurational conventions.

Bubbles, open boxes, closed boxes, and strokes are theoretically equivalent. However, strokes are one-dimensional entities, whereas bubbles and boxes are

### 4 Grammar and graphical semiotics in early syntactic diagrams

Figure 4.4: Various diagrams

two-dimensional ones. The initial choice of basic entities has consequences for the depiction of more complex constructions. This will be demonstrated in section §4 by comparing Clark's system to the one developed by Reed and Kellogg.

### **4 The logic of space: Case studies**

In the following subsections, I will study the graphical means at use to represent three kinds of syntactic relations in Clark's system and in Reed & Kellogg's diagrams, namely: the relation between the subject and the predicate (§4.1), coordination (§4.2), and subordination of the subject clause (§4.3). All three relations somewhat break the sparsity of Reed & Kellogg's model by introducing additional entities.

### **4.1 Subject-predicate relation**

A fundamental relation that every syntactic diagramming system has to depict is the one that holds between the subject and the predicate. This relation has already been illustrated in §§2.2 and 3.1.

Nicolas Mazziotta

Figure 4.5: Visual models for the subject-predicate relation.

Figure 4.5(a) is a *visual model*: it represents the configurational rules in Clark's grammar: Clark simply connects two bubbles from left to right. Bubbles clearly delimit their perimeters without any possibility of confusion, and their relative positions are sufficient to express the role of the words they contain. Therefore, only words are reified and only two entities are necessary.

In Reed & Kellogg's approach, the stroke corresponding to the subject and the one corresponding to the predicate in the graphical model (Figure 4.5(b)) could not have been joined by a mere concatenation. Because of the one-dimensionality of the strokes, the resulting configuration would have been a single line, with no means of distinguishing between syntactic roles. Reed and Kellogg understood this shortcoming only too well:

I will draw on the board a heavy, or shaded, line, and divide it into two parts [. . . ]. I will consider the first part as a sign of the subject of a sentence, and the second part as a sign of the predicate of a sentence. (Reed & Kellogg 1879 [1876]: 17)

They use a vertical dash to part the subject from the predicate, which yields the model in Figure 4.5(b). From a visual point of view, this vertical dash is an additional entity that does not correspond to a word. Paradoxically, using a graphical device to *separate* two units creates an entity that corresponds to the grammatical relation that *unites* them. For this reason, three entities are necessary: one for the subject (a horizontal stroke), one for the predicate (a second horizontal stroke), and a third one for the relation between the subject and the predicate (a vertical dash).

The hierarchy between the subject-predicate relation and the predicate-object relation is not obvious in Clark's diagrams, nor in the text that accompanies the diagrams.<sup>4</sup> Some diagrams that contain a subject and a predicate as well as an object were interpreted by some contemporary authors (Jewell 1867: 30) in a way that conceptualizes equivalent structural positions. By contrast, diagrams such as Figure 4.6 suggest a visual hierarchy. Reed & Kellogg state several times that the *object complements* "complete" the predicate: "You will see that the line

<sup>4</sup> See Mazziotta (2020: §5.3.2).

4 Grammar and graphical semiotics in early syntactic diagrams

Figure 4.6: Objects in Reed & Kellogg's system

standing for the *object complement* is a continuation of the predicate line, and that the little vertical line only touches this without cutting it" (Reed & Kellogg 1879 [1876]: 54).

(3) Fulton invented the first steamboat.

Again, the use of a stroke to separate the verb from the object complement is actually an entity that reifies the relation that unites them. This graphical convention corresponds to a hierarchical analysis whereby the object is part of the predicate. This ICA-like analysis is related to the entities used to render the words and their relations: aggregated bubbles alone would allow for such a reification of part-whole relations only by inclusion.

### **4.2 Coordination**

Traditional grammatical theory posits that coordination adds complexity to syntactic structures by allowing for several units to share the same grammatical role.<sup>5</sup>

In Clark's system, words that are coordinated aggregate with the surrounding bubbles in the same way as they would if they were not. In Figure 4.7(a), *the heaven* and *the earth* are both objects of the predicate *created*, and they are connected to it horizontally. The coordinative conjunction *and* unites them by aggregating vertically with both. Should the conjunction be absent, Clark would use the symbol × to express that it is implied. The model in Figure 4.7(b) provides a synthesis of the entities at use: there are four of them, one for each word. The two-dimensionality of the bubbles makes them extensible vertically and horizontally. Therefore, the size of the bubbles may be altered to facilitate the marking of connections with no change in their values.<sup>6</sup>

<sup>5</sup> Some modern approaches prefer to assume that the conjuncts have asymmetric syntactic roles. This difference is expressed by the way the conjuncts are encoded in the syntactic structure – see Mouret (2007) for such examples in constituent trees and Polguère & Melʹčuk (2009: 50–51) for an example of asymmetric dependency-based description.

<sup>6</sup> There are several other examples of such alteration in Clark's work and most of them are coordinations.

Nicolas Mazziotta

Figure 4.7: Coordination in Clark's system

The one-dimensional strokes of the diagrams by Reed and Kellogg do not have the same properties: the only ways they can be manipulated is by changing their thickness (which does not really add another dimension), their slope, or their continuity (using dots instead of a continuous line). They cannot be extended vertically. Therefore, the authors had to create a device to represent coordinations. They describe the first example given in Figure 4.8(a) as follows:

The short line following the subject line represents the entire predicate, and is supposed to be continued in the three horizontal lines that follow, each of which represents one of the parts of the compound predicate. These lines are united by dotted lines, which stand for the connecting words. The × denotes that an *and* is understood. (Reed & Kellogg 1879 [1876]: 47–48)

Figure 4.8: Coordination in Reed & Kellogg's system

The graphical model of coordination in Figure 4.8(b) shows that the number of strokes exceeds the number of nodes. The slanted lines, which are thinner than the horizontal ones, represent the part-whole relation between a syntactic unit taken as a whole and its component (the conjuncts). The notation integrates three additional entities that aggregate with the words: two of them are used to

### 4 Grammar and graphical semiotics in early syntactic diagrams

represent the part-whole relation, and the third one represents the larger syntactic unit. In the case of coordination, using strokes with a specific slope offers no other solution but to introduce additional entities: one for each part-whole relation between the conjuncts and the group they form, and, of course, one for this very group.

### **4.3 Subordinate clauses as subjects**

Subordinate clauses are not handled in a consistent way, either in Clark's grammar (see Mazziotta 2016: 319–322 and 328–329) or in the diagrams by Reed and Kellogg. Adjunct clauses are not represented with the same graphical conventions as the ones introduced by complementizers as objects or as subjects.<sup>7</sup> I will focus on the latter, which Clark names "auxiliary sentences". Auxiliary clauses assert "dependent proposition[s]" (1870: 187).

(4) That good men sometimes commit faults cannot be denied.

Figure 4.9: Subordinate subject in Clark's system.

In Figure 4.9(a), the subject clause, *that good men sometimes commit faults* is wrapped in half an ellipsis that represents an abstract subject – I will not comment on the position of the complementizer. The spatial configuration that expresses the relation between the subject and the predicate (cf. §4.1) is not sufficient to represent complex structures that are used as subjects: the subordinate predicate *commit* is already aggregated to the right of the subject *men*, and to the left of the object *faults*. There is no room left to its right for another predicate, for the structure would then concatenate five bubbles without any means of distinguishing the predications from one another. The introduction of an additional entity establishes a hierarchy between the predications. It does so by a specific configuration: an inclusion layout that is made possible by the two-dimensionality

<sup>7</sup>This may be due to the heterogeneous treatment of dependents of the verb, since neither Clark nor Reed & Kellogg overtly acknowledge verb-centrality (§4.1).

### Nicolas Mazziotta

of the entity. There are two important consequences to this choice of representation: (i) the part-whole relations between the subject and its constituents are iconic; (ii) the combination of this wrapping bubble with the predicate *can be denied* operates exactly in the same way as if it were a single word.

Due to the one-dimensionality of the lines, the corresponding representation by Reed & Kellogg does not use inclusion.

(5) That stars are suns is taught by astronomers.

Figure 4.10: Subordinate subject in Reed & Kellogg's system

The description of the diagram in Figure 4.10(a) clearly identifies a special device used as a "support" for the subordinate clause: "As this [sentence] subject cannot, in its proper form, be written on the subject line, it is placed above, and, by means of a support, the [sentence] diagram is made to rest on the subject line" (Reed & Kellogg 1879 [1876]: 107).

Since the configurational rationales of the combination between the subject and the predicate are the same as in Clark's system (horizontal connection), it follows that lines offer no other possibility but to introduce an additional entity in such cases. This entity could have been a wrapping device much as Clark's or a connecting entity. Reed & Kellogg chose the latter. The one-dimensionality of strokes does not allow the depicting of inclusion in an iconic way. As a result, what is expressed by this entity is a relation of equivalence between the abstract subject and its realization as a clause – interestingly, the subordinate clause is connected via its predicate to the abstract subject.

### **5 Conclusions**

Diagrams are graphical means of representing syntactic analyses. There are several ways to visually represent words and analytic concepts. The first distinction

### 4 Grammar and graphical semiotics in early syntactic diagrams

is between *entities* and *spatial configuration*: as shown in section §2.2, grammarians may choose to represent relations between words with entities that reify them (Figure 4.2(a)) or by their layout (Figure 4.2(b)). When different grammarians choose to reify the same units, they may use different entities to express similar analyses (cf. §3). However, the material choice of different shapes (bubbles in the case of Clark, and strokes for Reed & Kellogg) as diagrammatic entities has consequences for the spatial configuration of the entities, and on the amount of entities needed to express the analyses. In the case of the relation between the subject and the predicate, Clark could use a simple configurational convention at his disposal to arrange the bubbles representing a subject and a predicate, whereas Reed & Kellogg had to introduce an additional entity. In an indirect way, this additional entity can be interpreted as a representation of the relation between the subject and the predicate. Similar means can be used to identify the verb-object relation, thus leading to a model that decomposes the sentence into two strata of relations (cf. §4.1). A related issue concerns the representation of coordinate constructions: since conjuncts share the same positions in the syntactic structure, diagrams should represent this feature. With two-dimensional entities, it is possible to do so in an iconic way, whereas the use of strokes implies the use of additional entities that reify abstract units representing the combination of the conjuncts (cf. §4.2). The representation of subordinate subjects (or objects) also implies the introduction of abstract entities in all kinds of diagrams, be they a wrapping entity or a connecting one (cf. §4.3).

From these observations, it appears that diagrams making use of one-dimensional objects are less minimalistic. The use of one-dimensional strokes forces the reification of abstract units that are not words, but rather represent part-whole relations or the integrity of complex structures considered as wholes. Indeed, current constituent trees (Figure 4.1) constantly and consistently represent this kind of syntactic arrangement. A similar tendency was already developing at the time when Alonzo Reed and Brainerd Kellogg were drawing their first diagrams.

Graphical entities chosen to express the analyses always constrain what kind of relations can be represented. More often than not, they imply the creation of *artefacts*, i.e. graphical objects that are not part of the knowledge transmitted by the analysis, but entirely pertain to the graphical formalism that encodes it.

### **References**

Aarts, Bas & April McMahon (eds.). 2006. *The handbook of English linguistics* (Blackwell handbooks in linguistics). Malden, MA: Blackwell.


## **Chapter 5**

## **Aux origines de la notion de polysémie en français : la formation du concept**

### Bruno Courbon

Université Laval

For at least 80 years, the name of Michel Bréal has been both associated with the development of the field of semantics within French linguistics and with the invention of the term and concept of *polysémie* (polysemy) in the 1880s. The present contribution deals with the origin of the concept of polysemy, when Bréal used it for the first time. Although Bréal contributed to define polysemy thoroughly, neither the idea, nor the term were new. In the first place, the term was coined by Joseph Halévy ; secondly, dictionary makers and 19th century linguists often conceived of lexical meaning in terms of polysemy.

> […] la pratique métalinguistique des lexicographes pose le problème de l'émergence des notions avant leur dénomination. Le fait de nommer des concepts permet de circonscrire des portions de réalité, de reconnaître des séquences de problématisation et de questionner des phénomènes que l'acte de dénomination vise à cerner.

> > (Bisconti 2016 : 23)

Bruno Courbon. 2020. Aux origines de la notion de polysémie en français : la formation du concept. In Émilie Aussant & Jean-Michel Fortis (éd.), *Historical journey in a linguistic archipelago : Descriptive concepts and case studies*, 83-96. Berlin : Language Science Press. DOI : 10.5281/zenodo.4269415

### Bruno Courbon

### **1 Introduction**

La polysémie peut être considérée comme un cas emblématique dans l'histoire des idées linguistiques.<sup>1</sup> Il s'agit d'abord d'un phénomène intuitif, dont les sujets parlants font consciemment l'expérience lorsqu'ils remarquent qu'une même forme lexicale sert à parler de réalités différentes que relie une impression commune.<sup>2</sup> Dans un second temps, cette expérience donne lieu à des propos qui indiquent la conscience d'un « fait sémantique ». Les lexicographes usent ainsi depuis longtemps d'une terminologie discrétisante pour parler de la pluralité sémantique des formes lexicales (« ce mot a plusieurs sens », « le mot *X* reçoit plusieurs acceptions »…). Cette vision analytique du sens lexical est clairement exprimée depuis le XVII<sup>e</sup> siècle dans la tradition métalinguistique de langue française.<sup>3</sup> Il faut pourtant attendre la fin du XIX<sup>e</sup> siècle pour que l'idée intuitive qu'à une même forme lexicale puissent correspondre plusieurs significations prenne la forme abstraite du concept de *polysémie*. Ce concept est forgé à une époque où les questions relatives au sens linguistique font l'objet d'un examen attentif, qui s'affranchit de l'habitude consistant à observer d'abord les effets de sens spécifiques à des discours particuliers.<sup>4</sup>

<sup>1</sup>Ce texte s'inscrit dans le prolongement d'un précédent travail (Courbon 2015). On y trouvera moins de développements relatifs à ce qui fait la spécificité du point de vue de Joseph Halévy (1827-1917) et de Michel Bréal (1832-1915) sur la polysémie ; cependant, l'accent est mis sur deux aspects : 1) d'une part, la continuité entre a) la conception générale de ce qui est intuitivement perçu comme de la polysémie et b) la conceptualisation nouvelle que rend possible l'importation du terme *polysémie* en sémantique, et, d'autre part – et dans une moindre mesure –, la distribution entre l'usage du terme chez Halévy et l'utilisation conceptuelle spécifique qu'en fait Bréal ; 2) l'appartenance du concept de polysémie à une mythologie des fondations (celles de « la sémantique » comme discipline), qui n'aurait sans doute pas déplu à Bréal, mythologue et découvreur de talent.

<sup>2</sup>Ces éléments appartiennent toujours à la définition du concept de polysémie : « un mot polysémique (un polysème) est un mot qui rassemble plusieurs sens entre lesquels les usagers peuvent reconnaître un lien » (Nyckees 1998 : 94).

<sup>3</sup>Nous ne voulons pas dire par là que l'idée de ce qui est désormais appelé « polysémie » n'existerait pas avant le XVII<sup>e</sup> siècle, mais seulement souligner le fait qu'au moment où le terme *polysémie* est forgé, cette conception était chose relativement banale dans les pratiques de description du français, notamment en lexicographie. En fait, cette conception remonte sans doute bien avant les premières attestations qu'on en trouve dans l'Antiquité : qu'une forme lexicale à valeur référentielle spécifique puisse servir à parler de choses relativement différentes bien qu'en partie apparentées est en effet un phénomène suffisamment fréquent et saillant pour que les sujets parlants en aient conscience assez tôt dans leur expérience linguistique.

<sup>4</sup>Cette contribution ne présentera pas l'évolution conceptuelle qu'a connue l'étude des questions de sens au cours du XIX<sup>e</sup> siècle. Il est toutefois évident que la focale change, et que l'on se préoccupe davantage, dans la deuxième moitié du siècle, du sujet sémantiseur, qu'il soit « cognitif », « psychique » ou social… Explicitement présenté comme intersubjectif, le sens linguistique est de moins en moins conçu comme un pur effet discursif (sens sémiotextuel).

### 5 Aux origines de la notion de polysémie en français

Le fait d'étudier la « cristallisation » du concept de polysémie dans la période charnière durant laquelle la sémantique reçoit son nom moderne présente un double intérêt : d'une part, en ce qui concerne l'histoire des idées relatives aux faits de signification linguistique ; d'autre part, en ce qui concerne la sociologie des sciences linguistiques. Pour ce qui est de l'histoire des idées sur la langue, resituer la genèse du concept de polysémie dans le contexte intellectuel de l'époque contribue à éclairer quelques-uns des enjeux liés à son utilisation. De plus, du point de vue de la sociologie des sciences du langage, l'histoire de ce concept participe d'un mythe fondateur associé, dans le monde francophone, à la constitution de « la sémantique » en discipline. Dans ce mythe originel, l'aura symbolique du père fondateur se surimpose à la complexité des formes de contribution. Bien que ce soit à Michel Bréal (1832–1915), alors ténor de la linguistique institutionnelle, que l'on attribue couramment la paternité du terme et de la notion de « polysémie », nous verrons que la situation dans laquelle le concept s'est développé est loin d'être aussi simple, et l'affaire sans doute plus délicate qu'elle semble à première vue.

Même si elle n'est pas formellement nommée, l'idée que recouvre le concept de polysémie est souvent énoncée dans la linguistique francophone du dernier tiers du XIX<sup>e</sup> siècle. Nous verrons aussi que c'est à l'orientaliste Joseph Halévy (1827–1917), et non à Michel Bréal, que l'on doit les premières utilisations du terme *polysémie*. Étant donné que Bréal et Halévy se côtoyaient régulièrement, on peut faire l'hypothèse que le contexte théorique général en linguistique, comme, en particulier, la conception halévienne, ont pu exercer une influence sur l'interprétation que Bréal donna à la notion de polysémie. Nous reviendrons dans la présente contribution sur cette « erreur » d'attribution de la paternité du terme et de la notion de polysémie à Michel Bréal, en tâchant de les resituer de façon sommaire dans le contexte de l'époque (i.e. à partir des années 1860–1870).<sup>5</sup>

<sup>5</sup>L'un des relecteurs d'une version antérieure de ce texte résume, à juste titre, qu'il y est fait état d'une « querelle de paternité ». À notre avis, cette « erreur » d'attribution est moins grave qu'on pourrait le croire. En revanche, il nous semble qu'elle devrait intéresser l'histoire et la sociologie de la linguistique à plus d'un titre : 1) au plan de la circulation des termes « consacrés » : son origine oubliée, le terme *polysémie* est devenu, une fois repris en sémantique, un puissant instrument conceptuel ; 2) au plan de la genèse des concepts métalinguistiques, ici clairement située entre conception commune et intuitive, tradition descriptive et proposition dénominative réussie ; 3) au plan de l'effet sociologique que peuvent exercer les sphères d'appartenance et les réseaux d'influence sur la fortune de certaines conceptions en linguistique (un autre relecteur anonyme note que les projets respectifs de Bréal et d'Halévy qui sous-tendent l'utilisation qu'ils font du terme *polysémie* s'opposent nettement : Bréal, dans une optique laïque, utilise ce concept pour illustrer les progrès de l'esprit humain ; Halévy, au contraire, utilise le concept de polysémie pour défendre une thèse d'ordre religieux – outre l'influence personnelle et institutionnelle de Bréal à l'époque, sa position est sans doute plus adaptée au contexte intellectuel français de la fin du XIX<sup>e</sup> siècle).

### Bruno Courbon

### **2 En amont de la polysémie comme concept dénommé**

Les sujets parlants ont l'intuition de ce que l'on a pris coutume de nommer « polysémie (d'un mot, d'une expression) ».<sup>6</sup> Assez tôt dans l'histoire des descriptions de la langue française des sujets experts ont exprimé cette conception plurisémantique. Parmi les façons de traduire l'intuition ordinaire qu'une forme particulière peut « comporter » plusieurs significations, deux d'entre elles semblent se dégager au fil de l'histoire :


Dans la tradition française, l'idée que l'on puisse donner à un mot une pluralité d'« acceptions » est clairement formulée au moins depuis le XVII<sup>e</sup> siècle. La sous-entrée ACCEPTION de la réédition du *Dictionaire* [sic] *universel* de Furetière (1619–1688) (Furetière 1701) en témoigne : « Sens auquel un mot se prend. Ce mot a plusieurs *acceptions*. Dans sa première & plus naturelle *acception*, il signifie & c. » (ibid.). À cette même époque, divers auteurs parlent d'un mot ou d'une proposition en disant qu'ils « ont » ou « reçoivent » « plusieurs sens » ou « plusieurs significations ».<sup>7</sup> Voici quelques exemples de ces propos :

<sup>6</sup>L'expression *concept dénommé* est ici redondante. La position adoptée dans cet article est la suivante : tant que le terme *polysémie* n'est pas utilisé, il existe certes une idée intuitive de ce que l'on nomme aujourd'hui « polysémie », ses manifestations peuvent être décrites par les sujets parlants, la réalité notionnelle correspondante peut même être exprimée de diverses autres façons (cf. *pluralité de significations*) ; pour autant, le *concept* de polysémie n'existe pas encore à proprement parler. Il ne commence à circuler qu'à partir du moment où la dénomination spécifique de *polysémie* est utilisée. Parler de concept de polysémie avant la création du terme est donc une forme d'anachronisme.

<sup>7</sup>Un examen systématique montrerait certainement que la question de la pluralité des sens glisse en partie, entre le XVII<sup>e</sup> siècle et le XIX<sup>e</sup> siècle, du texte vers le mot. Alors que les domaines

### 5 Aux origines de la notion de polysémie en français


Avant l'emploi du terme *polysémie* en linguistique, la conception qui consiste à voir dans l'unité lexicale une pluralité de significations est explicitement formulée, notamment dans des contextes didactiques :

(6) « La recherche des sens divers d'un même mot, complète […] les études nécessaires pour arriver à la connaissance de la signification exacte des mots et de la propriété de l'expression. » (Michel 1858 : 252).

Ces quelques exemples montrent que des francophones expriment l'intuition qu'ils ont de ce que l'on appelle aujourd'hui « polysémie », et ce bien avant la création du terme.<sup>8</sup> Non seulement ces discours révèlent la conscience qu'ont certains auteurs de cette intuition, mais ils indiquent aussi la représentation que des sujets experts se faisaient de la signification linguistique, représentation qui se teinte des idées du temps : avec le développement, au XIX<sup>e</sup> siècle, des études sur l'histoire de la langue française, l'intérêt pour la diversité sémantique tend à

originellement associés à la question des « sens divers » relevaient de l'herméneutique et de l'exégèse (droit, théologie, philosophie), le phénomène est progressivement mis en valeur dans les arts du langage (rhétorique, grammaire, lexicographie, puis histoire de la langue).

<sup>8</sup> Il est certain que ce type d'intuition n'est pas propre aux francophones. Une recherche approfondie devrait couvrir une période plus large et inclure d'autres langues (dont des langues sans tradition écrite). En restant sur le plan des constituants morphologiques du terme, on remarque que l'adjectif *polysemous* est employé en anglais dès 1853 dans un sens proche de celui qu'on lui connaît encore : [à propos de la gestuelle en Italie] « […] he [un homme napolitain, avec un ami] shakes his head and hands, uttering […] the monosyllabic but polysemous exclamation "Eh!" which, like a Chinese word, receives its meaning from its varying accent. » (Wiseman 1853 : 534).

### Bruno Courbon

passer de l'analyse de textes (effets de sens et usages) à une conception abstraite de l'unité sémiosémantique. Émerge une conscience aiguë du développement de nouvelles significations (ce qu'on appelle aujourd'hui « néologie sémantique »).

La préoccupation pour la « multiplication des significations » est présente dans les premières utilisations que Michel Bréal fait du terme *polysémie*. Mais elle figure aussi dans les textes de quelques-uns de ses prédécesseurs immédiats, comme Émile Littré (1801–1881) et Arsène Darmesteter (1846–1888).<sup>9</sup> Darmesteter (1874 : 11–12) parle de « transformation » ou de « succession des sens dans les mots » (cf. Darmesteter 1876 ; voir aussi Bailly 1874), Littré (1880 : 1) de « mutations de signification ». On relève, dans l'*Histoire de la langue française* de ce dernier (Littré 1863), les germes de la vision « polysémiste » de la signification lexicale. Bréal connaissait suffisamment l'œuvre de Littré pour qu'elle pût inspirer sa réflexion. Dans le premier texte où il présente le concept de polysémie (Bréal 1887), il fait référence au travail du lexicographe. Bréal est également à l'origine de la réédition posthume de l'opuscule de Littré « Pathologie verbale ou lésions de certains mots dans le cours de l'usage » (Littré 1880), auquel il préfère donner le titre de *Comment les mots changent de sens* (Littré 1888).

Les liens entre les deux « sémantistes » que sont Darmesteter et Bréal sont plus marqués encore. Arsène Darmesteter engage dès le début des années 1870 une réflexion sur la signification lexicale. On peut souvent lire, en filigrane des analyses de Bréal, la présence de Darmesteter. On remarquera aussi que les premières occurrences du terme *polysémie* chez Bréal (au nombre de 3) sont publiées dans un texte intitulé « L'histoire des mots », paru en 1887 en réaction aux éléments de substrat organiciste que Bréal relève dans l'ouvrage « éminemment suggestif » de Darmesteter (Meyer 1888 : 469), *La vie des mots étudiée dans leurs significations* (sur ce point, voir Delesalle 1987). Dès le titre s'affrontent deux conceptions. L'une historiciste (« histoire des mots »), l'autre naturaliste (« vie des mots »). Le point de vue de Darmesteter, cependant, est plus nuancé – et sensé – que Bréal ne le laisse entendre. Dans sa thèse sur la néologie, parue en 1877 (soit dix ans avant le texte de Bréal), Darmesteter s'interrogeait déjà, dans les termes suivants, sur les mécanismes à l'œuvre dans l'extension de la signification :

Ce mot [*carré*], compris de tous, a des significations multiples ; pour en faire le nom de l'objet nouveau, le peuple [parlant de l'objet du jardin anglais] sera obligé de faire un travail intellectuel qui, par une extension dans la signification, approprie le mot à la chose […]. (Darmesteter 1877 : 33)

<sup>9</sup>Auteurs que Nerlich (2001a : 1605) considère comme les prédécesseurs immédiats de Bréal en « sémantique » (le terme est alors anachronique).

### 5 Aux origines de la notion de polysémie en français

Écrit dix ans avant que Bréal n'emploie le terme *polysémie* pour la première fois, ce texte montre l'existence d'un questionnement sur la multiplicité des significations. L'idée de polysémie est en germe.

### **3 Autour de l'invention du terme** *polysémie* **dans les années 1870–1880**

Même les théories plus ou moins bien connues et souvent discutées ne sont pas connues dans leurs connexions historiques. Ainsi, par exemple, on attribue presque toujours à Saussure les distinctions entre *langue* et *parole*, entre *signifiant* et *signifié*, entre *synchronie* et *diachronie*, toutes distinctions que Saussure a retrouvées dans la tradition, qu'il a, sans doute, reformulées et auxquelles il a donné en partie une interprétation nouvelle, dans le cadre d'un système cohérent, mais qu'il n'a pas été le premier à formuler. (Coseriu 1967 : 74)

Le propos de Coseriu (1921–2002) au sujet de distinctions systématisées par Saussure peut s'appliquer à nombre de notions devenues des concepts-clés en linguistique, et dont la création est attribuée à des « fondateurs » (d'une discipline, d'un domaine, d'un courant…). Ainsi en va-t-il de la notion de polysémie : Bréal (1897), fondateur officiel de la sémantique en France, reprend des idées anciennes à propos de la multiplicité des sens, idées qu'il développe et systématise d'une façon singulière, et qu'il nomme en (ré)utilisant un terme (*polysémie*) employé depuis plus d'une décennie par un confrère qu'il côtoie, écoute, et dont il commente les travaux présentés dans divers cercles savants du Paris des années 1870–1880.<sup>10</sup>

### **3.1 Joseph Halévy : l'inventeur du terme** *polysémie*

Contrairement à une croyance répandue, ce n'est pas à Michel Bréal, mais à l'orientaliste Joseph Halévy que l'on doit l'invention du terme *polysémie* en français.<sup>11</sup> C'est dans le but de démontrer l'origine sémitique de la langue transcrite au moyen de l'écriture cunéiforme que J. Halévy a l'habitude d'employer ce terme,

<sup>10</sup>École pratique, Société de linguistique, Académie des inscriptions, etc. À ce sujet, voir Courbon (2015).

<sup>11</sup>Pour une présentation plus détaillée du sens que donnent Halévy puis Bréal à la notion de polysémie, voir Courbon (2015 : 5–7).

### Bruno Courbon

qui correspond à un concept central de son œuvre. Halévy présente la polysémie comme l'une des « particularités les plus saillantes du système graphique assyro-babylonien » (Halévy 1876 : 298).

Ce terme renvoie alors à la multiplicité de valeurs sémantiques que revêtent les signes cunéiformes. Celle-ci peut être exprimée par l'intermédiaire de signes linguistiques différents. Le concept de polysémie se rapproche ainsi du concept classique de synonymie,<sup>12</sup> mais comporte déjà, chez Halévy, les principaux traits que nous lui connaissons encore : pluralité de significations associées à un signe (ici, un idéogramme) et proximité relative des différentes « lectures » de ce signe. Halévy décrit sa conception de la polysémie dans une communication donnée en 1878 :

[C]haque signe envisagé comme idéogramme est en général rendu par plusieurs mots […], ce qui revient à dire que le signe comporte à la fois plusieurs sens, qu'il est polysème. (Halévy 1883 : 275)

La polysémie, qu'Halévy décrit comme la multiplication de sens équivoques, est consubstantielle au système d'écriture cunéiforme :

[L]'accumulation infinie d'équivoques dans l'accado-sumérien, caractérise celui-ci comme un système idéographique, ou [sic] la polysémie des signes est un principe fondamental et inéluctable. (Halévy 1883 : 276)

L'idée de la pluralité des significations d'un signe idéographique se trouvait déjà cinquante ans auparavant chez Champollion (1790–1832) (1828 : 311–312), mais elle ne recevait alors pas d'autre nom que celui, classique, de « signes synonymes » (ibid.).

### **3.2 Une « erreur » d'attribution : « Michel Bréal, inventeur du terme** *polysémie* **»**

L'attribution à la personne de Bréal de la paternité du terme de *polysémie* – ou, quelquefois, du concept correspondant – est présentée dans différents textes : Firth (1957 [1935] : 15), Ullmann (1952 : 199), Ricoeur (1975 : 147), Delesalle & Chevalier (1986 : 286), Delesalle (1986 : 89 ; 1987 : 300–305), Nerlich (1993 : 22 ; 2001b : 1625), Peeters & Eiszele (1993 : 169–170), Peeters (1994), Desmet & Swiggers (1995 :

<sup>12</sup>Sur la proximité conceptuelle entre synonymie et polysémie à cette époque, voir Delesalle (1986).

### 5 Aux origines de la notion de polysémie en français

27), Branca-Rosoff (1996 : 118), Victorri & Fuchs (1996 : 11), Surcin (1999 : 16), Auroux & Delesalle (2000 : 215), Siblot (2000 : 156), Nerlich & Clarke (2003 : 4), Girardin (2004 : 131), Piron (2006 : 51), Cusimano (2008 : 55), Larrivée (2008 : 22), Thibault (2009 : 116), Mazaleyrat (2010 : 10), Pauly (2010 : 23), Jakimovska (2012 : n. p.), Fr. Rainer, dans Lieber & Štekauer (2014), Derradji (2014 : 59), Sorba & Cusimano (2014 : 6), Bisconti (2016 : 23), Bruns (2016 : 4), De Palo (2016 : 103).

Les dictionnaires contemporains perpétuent cette croyance. Ainsi, la date de première attestation du terme présentée dans le *Petit Robert 2018* et dans le *Trésor de la langue française* est 1897, année de parution de l'*Essai de sémantique* de Bréal (1897), plus de vingt ans après son utilisation courante par Halévy. Cette erreur d'attribution (donc de datation) s'explique par le succès rapide qu'a connu l'*Essai* de Bréal. Halévy, quant à lui, n'est pas associé à l'histoire de la sémantique, au contraire de Bréal, dont le nom est associé à la constitution de la sémantique en champ disciplinaire, ainsi qu'à l'usage du terme *polysémie* (le terme, lui, est associé dès 1887 : Baale 1887).

La valeur générale du terme *polysémie* envisagée par Halévy est proche de celle que Bréal définit en 1887. Toutefois, l'association de ce terme à un domaine de la linguistique en émergence, et à celui que l'on considère comme son fondateur, prévaut sur la chronologie des faits et sur la paternité réelle de la dénomination.

### **3.3 La polysémie : un concept unificateur dès les débuts de la sémantique**

La place (deux chapitres de l'*Essai*) et la valeur particulière que Bréal accorde au concept de polysémie indiquent son importance.<sup>13</sup> En utilisant une dénomination simple pour parler d'un ensemble de phénomènes sémantiques relativement complexes, Bréal souligne la puissance significative des formes linguistiques, qui résultent parfois d'une réduction (cf. le rapport qu'établit Bréal entre polysémie et « ellipse »). En outre, par ce geste d'objectivation de phénomènes sémantiques intuitifs, Bréal assoit la légitimité de prendre le sens lexical comme objet d'étude à part entière, contribuant ainsi à consolider le projet de fonder la « sémantique » en un champ de réflexion et de connaissances spécifique.<sup>14</sup>

<sup>13</sup>Selon Nerlich, le concept de polysémie constitue le « point focal de [l']œuvre » de Bréal (Nerlich 1993 : 22).

<sup>14</sup>« [L]'avènement de la Sémantique et le surgissement de la polysémie » vont de pair, écrit Delesalle (1987 : 83). Sur les relations étroites, pendant cette période, entre la lexicographie et la sémantique « naissante », voir Bisconti (2016).

### **4 Conclusion**

Créé en français par J. Halévy dans les années 1870, le terme *polysémie* est réemployé par M. Bréal une dizaine d'années plus tard. L'audace terminologique de J. Halévy, en dehors des préoccupations propres à la linguistique historique, permet d'appréhender de façon plus stable le problème de la pluralité sémantique. La solution dénominative qu'offre Halévy à la communauté des linguistes ne comporte qu'une faible valeur conceptuelle. Cette proposition ne s'insère pas directement dans le champ de réflexion contemporain sur le sens linguistique. Mais, la création du terme simple et relativement transparent de *polysémie* contribue à attirer le regard sur une série de problèmes qui, tout au long du XX<sup>e</sup> siècle, n'ont eu de cesse d'être (re)posés. Il faut reconnaître à Bréal d'avoir insufflé à ce terme, en le plaçant parmi les éléments cardinaux de sa sémantique, une valeur particulière.

L'histoire primitive de la « polysémie » – entre idée générale, dénomination simple et stabilisation conceptuelle – illustre l'interdépendance constitutive entre histoire des idées linguistiques (plan conceptuel) et histoire des réseaux de circulation des idées linguistiques (dimension sociale). Se dessine, à travers cette histoire singulière, l'effet, sur le long terme, du sociosymbolique sur le cognitif, à travers notamment la forte influence d'un sémantiste de renom, Michel Bréal, considéré dans le monde francophone comme le « père fondateur » de la sémantique. L'histoire primitive du concept de polysémie, telle qu'on la rencontre depuis près d'un siècle, appartient au mythe originel de la discipline.

Bréal avait pour ambition de faire de la sémantique une « science nouvelle ». Il a insisté sur les nouveautés de sa démarche (fonction promotionnelle caractéristique des discours de fondation). Le concept de polysémie fait partie des fondements conceptuels de ce mouvement épistémologique. Dans l'entreprise fondationnelle, la référence au néologue Halévy, créateur du terme *polysémie*, aurait sans doute compliqué le tableau. On l'a oublié. Dans les faits, la nouveauté qu'apporte Bréal tient plus à l'articulation d'ensemble qu'aux phénomènes discutés ou aux termes employés et parfois présentés comme neufs. Contrairement à Halévy, dont les préoccupations intellectuelles sont différentes, Bréal présente dans son *Essai* une somme théorique, dont la force réside dans la cohérence des développements et dans la consistance des exemples. Bien qu'il pût avoir l'intuition de l'effet que produirait l'aboutissement de ses réflexions de sémantiste, Bréal ne pouvait, dans les années 1880–1890, connaître l'avenir de cette discipline dont il jetait explicitement les bases en lui donnant un nom en français. Il n'imaginait sans doute pas l'effet, sur les générations suivantes, de ses propos relatifs à la

### 5 Aux origines de la notion de polysémie en français

nouveauté du terme *polysémie*. Une autre histoire reste à écrire, qui consisterait à suivre les voies par lesquelles le mythe fondateur d'un Bréal inventeur du terme et/ou de la notion de polysémie s'est installé durablement dans le paysage de la linguistique française.<sup>15</sup>

### **Références**


<sup>15</sup>Je tiens ici à remercier les deux relecteurs anonymes pour leurs commentaires constructifs sur une précédente version de ce texte.


## **Chapter 6**

## **The unfixed status of fixed expressions: Past and present approaches to a pervasive linguistic feature**

### T. Craig Christy

University of North Alabama

Fixed expressions figure pervasively in speech, and have attracted much attention in language learning, translation and stylistics, yet linguistic analyses have been sporadic and inconclusive. Even Saussure offered little by way of accounting for how or why these "locutions toutes faites" suffuse language. In this study I contrast fixed expressions with other varieties of idiomaticity and discuss their importance in usage-based accounts and second-language learning. I also relate the traditionversus-freedom dialectic to Saussure's anagram studies.

> You shall know a word by the company it keeps. (Firth 1957: 11)

### **1 Terminology**

It would be hard to conjure a better candidate for illustrating the fuzziness of linguistic theory than what I am here exploring under the term "fixed expressions". That the phenomenon remains ill-defined and poorly understood is attested by the many rubrics under which it is researched, each of which, to greater or lesser extent, reveals a weighted emphasis in this or that direction of analysis. These rubrics include, alongside "fixed expressions", "lexical bundles", "multi-word expressions", "formulaic language", "set expressions", "formulaic sequences", "mul-

T. Craig Christy. 2020. The unfixed status of fixed expressions: Past and present approaches to a pervasive linguistic feature. In Émilie Aussant & Jean-Michel Fortis (eds.), *Historical journey in a linguistic archipelago: Descriptive concepts and case studies*, 97–111. Berlin: Language Science Press. DOI: 10.5281/zenodo.4269417

### T. Craig Christy

ti-word chunks", "memorized lexical phrases", "institutionalized utterances", "ritualized phrases", "prefabricated patterns/phrases", "ready-made constructions", "entrenched collocations", "phraseologisms", "phrasemes", "complex lexical units", "-grams", "idiomatic sequences", as well as, in French, "expressions toutes faites", "locutions figées", Bally's "locutions phraséologiques", and, perhaps most widely known, Saussure's term, in the *Cours*, "locutions toutes faites", to which I will return. For simplicity's sake, "fixed expressions" will be the default generic term in this paper.

### **2 Notable approaches**

Bally's (1865–1947) emphasis, for example in his *Le langage et la vie*, was on the role these fixed expressions play in stylistics, specifically the way they enable the telegraphing of emotive nuances in the communication of ideas (Bally 1952 [1913]: 141).<sup>1</sup> That recurrent experiential moments should become thus ossified as fixed expressions is both economical, from the point of view of memory, for instance, and unsurprising.<sup>2</sup> Fixed expressions facilitate the spanning and linking of informational units within pragmatic, discourse analytic, and speech act-specific contexts, as Bally observes early on in his emphasis of the influence emotive elements exert on the crystallization of fixed expressions:

Un grand nombre de tours syntaxiques sont nés de l'action du sentiment. …Il s'ensuit que plus l'expression est affective, plus elle tend vers une forme ou synthétique ou disloquée. (Bally 1952 [1913]: 68)

Saussure (1857-1913), in his brief discussion of syntagmatic relations in the *Cours*, acknowledges the classificatory challenges which are posed by invariable ready-made phrases ("locutions toutes faites"), and concludes that

Where syntagmas are concerned …one must recognize the fact that there is *no clear boundary* separating the language, as confirmed by communal usage, from speech, marked by *freedom of the individual*. In many cases it is

<sup>1</sup>As Bally (1952 [1913]: 68) argues, "…les éléments émotifs de la pensée tendent à immobiliser les articulations de la phrase logique, c'est-à-dire analytique. …Il s'ensuit que plus l'expression est affective, plus elle tend vers une forme ou synthétique ou disloquée. …l'association avec une forme ayant même sens que le groupe entier finit par faire oublier l'analyse des parties du groupe."

<sup>2</sup>Richards & Schmidt (1983: 191) also underscore the pervasiveness of this process. The Appendix offers a few representative examples taken from colloquial English.

### 6 The unfixed status of fixed expressions

difficult to assign a combination of units to one or the other. Many combinations are the product of both, in proportions which cannot be accurately measured. (Saussure 1986 [1916]: 123, emphasis mine)

In their functioning as units of meaning, fixed expressions appear to belong to the lexicon, yet, for the most part, they also show both regular (e.g., *hold your horses*, *scratch the surface*, *when the dust has settled*) and irregular (e.g., *all of a sudden*, *believe you me*, *once upon a time*) syntactic patterning.<sup>3</sup> For Saussure, the whole issue seems weighted toward determining to what extent the speech of an individual exhibits freedom of choice versus conformity to automated, situationally-triggered prefabricated constructions. It is precisely this tension between the gravitational harness of tradition, versus the freedom to intentionally mold and steer the elements of language, that suffuses Saussure's vexingly haunting "anagram" studies.<sup>4</sup>

More recent approaches to the analysis of fixed expressions have mainly focused on theorizing and typologizing idiomaticity, the many varieties of which are reflected in the multiplicity of terms used to refer to fixed expressions. Building on the work of Bolinger (1907–1992, in Bolinger 1976), Fillmore (1929–2014) and coauthors (1988), Makkai (1972), Richards & Schmidt (1983), Sinclair (1933– 2007, in Sinclair 1991) and others, and taking inspiration from Construction Grammar, Warren (2005) proposes "A Model of Idiomaticity" in which fixed expressions are seen as learned combinations that are just as integral to language competency as the lexical items and rules that dominate traditional accounts. These fixed combinations, acquired experientially through immersion in a language, figure prominently in usage-based accounts. In Warren's model fixed expressions figure as one type along a continuum of word-phrase-sentence idiomaticity. Gries (2008), speaking specifically of "phraseologisms", provides another useful survey of recent developments along with a proposed six-parameter rubric for differentiating among the various types and degrees of idiomaticity which fixed expressions exhibit, thus making it

…possible to define a variety of interrelated concepts from different frameworks including, but not limited to, idioms, word-clusters, n-grams, collocations, collostructions, constructions, patterns, fixed expressions and phraseologisms. (Gries 2008: 21)

<sup>3</sup>Cf. Wood (2002: 2–3), who cites Richards & Schmidt (1983: 36) on this point.

<sup>4</sup>On Saussure's anagrams, see Christy (2007), Christy (1999b), Christy (1999a) and Carr et al. (1999). On the relation of the homophonic principle to proper names in the anagrams, see Christy (1995).

### T. Craig Christy

The sheer fact that fixed expressions have commanded the attention of scholars in a broad spectrum of fields, ranging from theoretical linguistics, lexicology, lexicography, discourse analysis, psycholinguistics, and speech act theory, to applied linguistics, pragmatics, computational linguistics, corpus linguistics, natural language processing, machine translation, and second language learning underscores the pervasiveness, and importance, of the phenomenon. Given the predominant emphasis, in linguistic theory and practice, on the operation of rules on well-defined units – typically single, grammatically pigeonholed words – it's not altogether surprising that the status of, by comparison, messy and often intractable "fixed expressions" should have remained wobbly at best.

### **3 Fixed expressions in second language learning**

While the jury's still out as regards how best to situate and interpret fixed expressions within linguistic theory, their mastery has long been recognized as indispensable for achieving fluency in second language learning. In her article "Formulaic language in learners and native speakers" Alison Wray settles on the term "formulaic sequence" to signal any

…sequence, continuous or discontinuous, of words or other meaning elements, which is, or appears to be, prefabricated: that is, stored and retrieved whole from memory at the time of use, rather than being subject to generation or analysis by the language grammar. (Wray 1999: 214)

Estimates of the number of such formulaic sequences or fixed expressions all trend towards the high end: Jackendoff (1997: 156) observes that "There are vast numbers of such memorized fixed expressions", and, using a list of over 500 multiword expressions his daughter culled from the popular Wheel of Fortune game show, where contestants guess words and phrases on the basis of minimal clues, he concludes that his own "…extremely crude estimates [of candidate expressions for the show] suggest that their number is of about the same order of magnitude as the single words of the vocabulary." "Multi-word expression", or MWE, is the term most used in computational linguistics and work on natural language processing, where it refers, specifically, to a combination of two or more words that acts as a single unit that can exhibit syntactic and/or semantic idiosyncrasies. Multi-word expressions cross word boundaries, and lie at the interface of grammar and lexicon. They fall on a continuum that extends from transparently compositional combinations to opaque, non-compositional frozen constructions including idioms proper. Though some might think Jackendoff has cast the net too

### 6 The unfixed status of fixed expressions

widely in including all manner of compounds ranging from adjective-noun, nounnoun, participial, verbal, idioms, proper names, clichés, titles, quotations, foreign phrases, and more, the plain truth is that nothing stands out as more awkward in L2 speech than collocational combinations that stray from the beaten path. \*"I'm not large on sweets" and \*"He had a pain in the little of his back", though grammatically correct, are clearly not felicitous stand-ins for "I'm not big on sweets" and "He had a pain in the small of his back", to cite some examples from the transparent, compositional end of the continuum. Similarly, we speak of *fast food* and *strong coffee*, not \**quick food* and \**powerful coffee*. In these cases it is the asymmetrical distribution of synonyms, whose values (*valeurs*) diverge, that throws the issue of collocational fixity into sharp relief. Again, the sheer number of words appearing in set combinations makes this both a pervasive feature of language, and a prerequisite to fluency.

### **4 The turn to data**

The technology-driven rise of extensive corpora of language usage has been a major boon to the analysis and quantification of fixed expressions. In his overview of the current state of the relatively new field of computational phraseology, Ulrich Heid (2008: 348) points out that "In WordNet [a dictionary-thesaurus of English used mainly in artificial intelligence applications], roughly half of the entries are MWEs. Current research," he goes on, "suggests that these figures are rather low estimates, and that the number of MWEs in our languages is much greater than that of single words." In their widely discussed research into nativelike fluency, Pawley & Syder (1983: 191–192) argue

…that fluent and idiomatic control of a language rests to a considerable extent on knowledge of a body of "sentence stems" which are "institutionalized" or "lexicalized" …[and] whose grammatical form and lexical content is wholly or largely fixed; its fixed elements form a standard label for a culturally recognized concept. …The stock of lexicalized sentence stems known to the ordinary mature speaker of English amounts to hundreds of thousands.

Referring to the pioneering corpus linguist John Sinclair's (1933–2007)<sup>5</sup> "idiom principle" – that a large number of prefabricated or semi-preconstructed phrases

<sup>5</sup> Sinclair's efforts, dating back to the 1960s, to track statistically relationships between lexical units and meaning in spoken English prefigured corpus linguistics in the big-data sense in which it is now understood (cf. Stubbs 2013).

### T. Craig Christy

are available to speakers – other researchers have suggested that fully half of fluent native text is constructed according to this principle (Erman & Warren 2000: 29). The methodology of corpus linguistics brings statistical precision to the task of identifying co-occurrences that happen with a greater-than-chance frequency. The distribution of these co-occurrences across time can also be tracked, which at once gives some indication of the "life span", so to speak, of specific expressions. Notable corpora for English include, for example, the American National Corpus, the Collins Corpus, Bank of English, the Oxford Hector Pilot Corpus, and the British National Corpus. For ease of use, the Google Books -gram viewer is hard to beat as a means of getting a quick snapshot of an expression's usage data over time. Even so, its potential limitations may yet render it less reliable than others. These limitations include:

	- (1) Their houseplants *died* while they were away on vacation.
	- (2) Their houseplants *bit the dust* while they were away on vacation.

While (1) is matter of fact, (2) is marked by a colloquial casualness typical of a register used to signal solidarity with an interlocutor: (2) furthermore conveys that the speaker is rather glib about this turn of events, and not particularly saddened. In other words, the fixed expression evokes not only a meaning, but also a mood;

4. the data on usage of fixed expressions that are frequently cited in research literature (e.g., *spill the beans*, *skating on thin ice*, *kick the bucket*) may accordingly be skewed. Thus an expression that might otherwise be considered dated might rise in the frequency ratings just by dint of having been recently studied.

### 6 The unfixed status of fixed expressions

Despite its limitations, the Google Books -gram viewer still, I think, gives an overall sense of usage patterns and frequencies associated with fixed expressions. Looking at a few fixed-expression -gram charts (see Figures 6.1–6.3) can offer a glimpse of insights that could be buttressed or rendered more accurate using other available corpora.

Figure 6.3: -gram of *in a jiffy*

### T. Craig Christy

It should probably come as no surprise that the majority of high-frequency fixed expressions tend to show usage patterns that link the present to the past two or three generations: that is, our speech most closely tracks that of our parents, grandparents, and, perhaps, great-grandparents. The semantic unity of fixed expressions – which is to say their lexicalization – furthermore offers a staging ground for grammaticalization: think of *manage to*, *going to*, *couldn't help* + present participle, *bound to*, and so on. In other words, fixed expressions have an enhanced probability of being the nexus for nascent language change, which makes sense given that they evolve and function in the formative interface between grammar and lexicon, or, in Saussure's terms, between *langue* and *parole*. In their functioning as units of meaning, they appear to belong to the lexicon, yet, for the most part, they also show both regular (e.g., *hold your horses*, *scratch the surface*, *when the dust has settled*) and irregular (e.g., *all of a sudden*, *believe you me*, *once upon a time*) syntactic patterning.<sup>6</sup>

### **5 The rise of usage-based accounts**

In contrast to the prevailing linguistic view that the rules of language are acquired and then applied to the lexicon to generate grammatically correct novel utterances, according to the usage-based view, which aligns well with grammaticalization theory,<sup>7</sup>

…most language is [in fact first] acquired lexically [in the form of prefabricated lexical chunks] and then broken down and reassembled in new combinations. It is formulaic speech that provides the basis for the creative ruleforming processes by which the syntactic system is mastered. (MacKenzie 2000a: 174)

In second language teaching this concept of language acquisition has given rise to the "lexical approach", which sees fixed expressions, not words, as the real units of language.<sup>8</sup> In this view, "…language consists of grammaticalized lexis, and not lexicalized grammar" (MacKenzie 2000b: 63). Because fixed expressions are entrenched in long-term memory, ready for deployment when prompted by situational cues, they are closely associated with actual, or at least seeming,

<sup>6</sup>Cf. Wood (2002: 2–3), who cites Richards & Schmidt (1983: 36) on this point.

<sup>7</sup> See Christy (2010) for a detailed comparison of the usage-based grammaticalization theory with that of generative grammar.

<sup>8</sup>Cf. Lewis 1993.

### 6 The unfixed status of fixed expressions

fluency in a language, and are frequently even referred to as "fluency devices" (MacKenzie 2000a: 174; Wood 2002: 3): thus *you know*, *and so on* and similar frequent fixed expressions routinely crop up in the target-language sentences of English-speaking language learners.<sup>9</sup> Fixed expressions seem furthermore to survive, and even surface with increased frequency, in impaired speech, yet further evidence of their entrenchment and quasi automation (Granger & Meunier 2008: xxiv). That fixed expressions figure importantly in both second language learning and impaired speech underscores their primacy in language. In Sinclair's (2008: 409) view, "…the normal primary carrier of meaning is the phrase and not the word; the word is the limiting case of the phrase, and has no other status in the description of meaning." Sinclair sees words as comparable to phonemes: both are basic units involved in encoding meaning but, by themselves, are of insufficient magnitude to transact meaning, a function he assigns to the phrase. This view of the primacy of the phrase has given rise to the claim that grammatical competency emerges from a database of fixed expressions, a claim that has not gone unchallenged, with linguists in the generative tradition arguing just the opposite, namely that "…formulas do not appear to constitute the relevant input for the acquisition of syntax, and do not constitute a bootstrapping mechanism into grammar" (Bardovi-Harlig & Stringer 2017: 85).

"Memorized fixed expressions", to use Jackendoff's term, are stored in longterm memory and are processed more quickly as chunks. Precisely because they efficiently encapsulate specific high-frequency meanings, situations and attitudes, their use allows the speaker to compose longer stretches of speech than could otherwise be supported by working memory.<sup>10</sup> Examples of fixed expressions include, in English, *by the way*, *all of a sudden*, *be that as it may*, *in the long run*, *for the most part*, *for the record*, *before long*, *by all means*, *at any rate*, *so to speak*, *first and foremost*, and many more: in French we have, for instance, *à quoi bon?* ('what's the use?'), *tout de suite* ('immediately'), *tout à fait* ('totally'), *quand même* ('still'), *à peu près* ('approximately'), and so on. These serve as handy sentence connectives and stream-of-thought facilitators: for all intents and purposes, they behave as individual words, though, at least in orthographical perspective, the spaces separating their components disqualify them for official "word" clas-

<sup>9</sup>As Wood (2002: 7) observes, "A great proportion of the most familiar concepts and speech acts can be expressed formulaically, and if a speaker can pull these readily from memory as wholes, fluency is enhanced."

<sup>10</sup>Bolinger (1976) also underscores the value of prefabricated phrases in relation to memory constraints.

### T. Craig Christy

sification.<sup>11</sup> In Sinclair's (2008: 410) formulation, "…the meaningful units do not coincide regularly with the units of the orthography …". In the language of natural language processing and machine translation, fixed expressions present a "tokenization" challenge: that is, the challenge is how to represent and process them. The typical solution is to represent them as single items, "words", within processing algorithms, but that alone does not solve the problem of recognizing the semantic uniqueness of fixed expressions in the first place, of detecting the shift from compositional to non-compositional formulation. One promising approach is based on identifying topic transitions within a text, abrupt departures from the dominant theme as would be the case when, say, ice-breaking would pop up in a description of a meeting (*break the ice*) or bucket-kicking in a narrative about a person or animal (*kick the bucket*). So the sheer fact that fixed expressions – ranging from relatively transparent collocations to opaque idioms – stand out like a sore thumb itself becomes the basis for their identification in artificial intelligence systems.

Despite the indisputable importance of such fixed expressions to fluent speech and writing, and their acknowledged importance in language learning, "…the concept of phraseology is still notably absent from studies on translation theory and practice" (Granger & Meunier 2008: xxiv) – and, unfortunately, from most language teaching materials as well. Surely no foreign language teacher has been spared the consequences of students not being aware that what they are translating bit by bit is in fact an idiomatic, and idiosyncratic, fixed expression: thus, to cite a few German examples, \**bei dem Weg* is frequently used, instead of the correct *übrigens*, to render English 'by the way', just as \**für diesen Grund* ('for this reason') is used to translate what should be *aus diesem Grund*: these infelicitous translations are made "…because L2 learners make faulty assumptions about the transferability of restricted collocations …" (Kim & Kim 2012: 833).<sup>12</sup> In other words they are simply not sufficiently aware of what is and is not idiomatic and fixed since, in their native language, there is no need for a speaker to be particularly attentive to these categories.

<sup>11</sup>As Moon (1998: 43) points out, hyphenated chains of words similarly behave like single units: "Hyphenation is an indicator of the process of institutionalization and lexicalization …The catenation of strings into quasi-single words signals the writer's intention to consider a string as a unit …".

<sup>12</sup>Howarth (1998: 34) suggests that "…second language learners have trouble figuring out which of a range of collocational options are restricted and which are free" (cited by Kim & Kim 2012: 833).

### **6 Concluding remarks**

In this brief study I have tried to highlight some of the major issues associated with fixed expressions, and challenges that accommodating these issues to linguistic theory presents. These include:


While I have only scratched the surface, clearly these are all complex questions whose resolution will require close collaboration across disciplinary lines. I think you get the picture: there remains an indisputable need to bring clarity to the current unfixed status of fixed expressions.

### **Acknowledgements**

This research was supported by funding from Dr. and Mrs. William R. Welborn, and by the University of North Alabama.

## **Appendix**


### **References**


### T. Craig Christy


6 The unfixed status of fixed expressions


## **Part II**

## **Fields, authors and disciplinary commitments**

## **Chapter 7**

## **Language history from below: Pidgins and Creoles as examples**

### Michelle Li

Caritas Institute of Higher Education, Hong Kong

This paper argues for pidgins and creoles as examples of a relatively new perspective to historical sociolinguistics called "language history from below," which focuses on the language use of ordinary people. While recent research on major European languages adopting this perspective has produced some significant results, a "from below" perspective of pidgins and creoles has not yet been explored. Focusing on Chinese Pidgin English data from various historical sources, this paper shows how its evolution, groups of users, and the linguistic variation of the preposition *long* exemplify the idea of "language history from below."

### **1 Introduction**

One of the greatest challenges to the development of historical linguistics is the problem of bad data which "may be fragmentary, corrupted, or many times removed from the actual productions of native speakers" (Labov 1972: 100). Given this incompleteness of information presented by historical data, Labov (1972: 101) argues that linguistic research should adopt the uniformitarian principle, which states that "the linguistic processes taking place around us are the same as those that have operated to produce the historical record." This principle proposes that linguistic changes that occurred in the past can be explained by observing changes in the present (on the antecedents of this principle, see Christy 1983). A consequence of this approach is an emphasis on the use of synchronic data and quantitative methods for interpreting linguistic and social variations. In recent decades, historical linguistics has made significant progress due to technolo-

Michelle Li. 2020. Language history from below: Pidgins and Creoles as examples. In Émilie Aussant & Jean-Michel Fortis (eds.), *Historical journey in a linguistic archipelago: Descriptive concepts and case studies*, 115–128. Berlin: Language Science Press. DOI: 10.5281/zenodo.4269419

### Michelle Li

gies such as digitalization, which allows for the construction of large historical corpora. The fast-developing field of historical sociolinguistics has opened new avenues for discussion of theoretical and methodological concerns in the field, the problems and challenges of using historical data, and the value of written historical sources (Hernández-Campoy & Schilling 2012; Russi 2016).

In addition to the major languages, pidgin and creole languages also feature significantly in historical research because, as Romaine (2005: 1696) states, "language is both a historical and social product, and must therefore be explained with reference to the historical and social forces which have shaped its use." As the evolution and linguistic structure of pidgins and creoles are products of special sociohistorical contexts, these languages are prime examples for the study of language historiography and historical sociolinguistics.

This paper is organized as follows. §2 introduces recent developments in historical sociolinguistics and the concept of "language history from below." §3 argues for the relevance of pidgins and creoles within the "from below" approach in language history. Focusing on Chinese Pidgin English (CPE), spoken on the China coast, §4 presents the major groups of CPE users. In §5, linguistic variation in the use of the preposition *long* in different historical sources will be examined. §6 concludes the paper.

### **2 The "from below" approach to language history**

The linguistic data used in historical linguistics often come from written texts. As mentioned above, there are various problems with the value and quality of such written data. Hernández-Campoy & Schilling (2012) summarize the seven major problems with data in historical sociolinguistic research as representativeness, empirical validity, invariation, authenticity, authorship, social and historical validity, and standard ideology. These problems are all related to the quality of written texts and the data derived from them. However, one should not discard the value of this medium of communication simply because of the problems and limitations it presents. Compared with speech, writing is often viewed as asynchronous and more formal. Romaine (1988: 1461), however, argues that, from the point of view of a stylistic continuum, writing and reading do not necessarily represent the formal end of the continuum. Moreover, like speech, written data can also show sociolinguistic variation and can be used to reconstruct diachronic variation, such as in her study of the linguistic variation of Middle Scots (Romaine 1982). The impression that written language represents the educated elite, scholarly literature and standard language is partly due to historical linguists'

### 7 Language history from below: Pidgins and Creoles as examples

traditional bias towards the use of data from the most formal end of writing styles. Milroy (1999), for example, shows that, from approximately the 16th century onward, research on the history of English has almost always referred to the standard variety.

The notion of "from below" was first applied in the discipline of history (Thompson 1966), focusing on points of view or accounts offered by ordinary or common people rather than scholars or the higher social classes. A parallel approach called "language history from below" is developing in historical linguistics. This new interest in the language use of the lower social strata is partly due to the historical reliance on the standard variety of a language for linguistic data. Oesterreicher (1997: 198) points out that the traditional view of language history has favored the elaborate, socially exclusive and distance-orientated variety, usually the standard, which has resulted in only partial descriptions of the history of a language.

The "from below" approach highlights two sources of linguistic data: (i) language use by ordinary people and (ii) selection of texts showing speech or vernacular features (Elspaß 2007). Although who or what is considered "below" may vary from study to study, the term in general refers to ordinary people (Hailwood 2013), who are important to study because while constituting the majority of any population of language users, their language use is often neglected (Elspaß 2007). The types of materials selected for examination are mainly texts with a resemblance to speech, including ego-documents such as personal letters, diaries, autobiographies, and catechisms, rather than literary works or formal documents. In discussing the features of different types of written texts, Elspaß (2012) devises a continuum and argues that at one end lie texts of "distance" such as legal documents, whose language is mainly literate, formal, and planned, and at the other end are texts of "immediacy," which have an oral, informal, and unplanned nature similar to speech. Other types of ego-documents showing varying degrees of vernacular features include trial proceedings, diaries, travelogues, drama texts, sermons, and proclamations (Van der Wal & Rutten 2013). Ego-documents have been employed as sources in the study of the grammar and spelling of private emigrant letters written by Germans (Elspaß 2007; Elspaß et al. 2007) and letters written in seventeenth-century Dutch (Nobels 2013).

### **3 Pidgins and creoles as examples of "language history from below"**

The formation of pidgins and creoles was mainly a solution to the immediate communication barrier between indigenous people and Europeans under differ-

### Michelle Li

ent contact circumstances. While many languages undergo standardization at different levels of language use, the formation of pidgins and creoles is the result of negotiations and compromises on the spot among the groups of users concerned. Therefore, standardized usage was not the principal concern of the users. So far, works within the "language history from below" approach have mainly focused on major languages like English, German, and Dutch. In the following, I will argue that pidgins and creoles are ideal examples of languages of the common people.

From the 16th century onward, the European trading and political conquests in Africa and Asia created an environment where intercultural communication was commonplace. These languages of mixed sources are often depicted as marginal, makeshift, baby talk, or corrupted and broken versions of their European superstrates. Their use was mainly restricted to informal domains, and the speakers were mainly those of a lower social class. As such, few have cared about standardizing pidgins and creoles, but it is exactly this lack of attention that gives them room for innovation. The "creole continuum" (Stewart et al. 1965, Bickerton 1975), where we find different varieties of a creole, ranging from the least to the most similar to a standard, is a manifestation of the richness of linguistic and social variation of contact languages. In sum, we can see that the concept of "language history from below" can be applied both socially and linguistically to understand the history of pidgins and creoles in language contact settings.

Pidgins are used for restricted communication, such as trade, and thus have limited vocabulary and employ simpler grammatical structure (Holm 2000: 5). One of the best-known examples of pidgin is Chinese Pidgin English (CPE), which was initially used as a lingua franca for trading first in Canton and Macau in the 18th century and later in other treaty ports in China after the opening of foreign trade in the 19th century; it gradually declined in the first half of the 20th century (Matthews & Li 2013). The vocabulary of CPE consists mainly of English words with some borrowings from Portuguese, Malay, and Chinese; grammatically, it shows substantial influence from Cantonese, the local Chinese dialect spoken in Canton (Ansaldo et al. 2010). Social and political circumstances also played a part in the development of CPE. Officially, foreigners were prohibited from learning the Chinese language, and there was mutual hostility between the Chinese and foreigners (Baker 1990). These factors provided a space for CPE to develop as a lingua franca for interethnic communication. While negative views on pidgins and creoles abound, praises of their usefulness such as the one on CPE given by the old China hand William Hunter are harder to find.

### 7 Language history from below: Pidgins and Creoles as examples

In the Canton book-shops near the Factories was sold a small pamphlet, called *Devil's Talk*…I have often wondered who the man was who first reduced the "outlandish tongue" to a current language. Red candles should be burnt on altars erected to his memory and oblations of tea poured out before his image, placed among the wooden gods which in temples surround the shrine of a deified man of letters. (Hunter 1882: 63)

Carl Crow, a long-term resident in China, also recognized the effectiveness of CPE as a means of communication:

In no language could the conversation have been any more definite and in none could fewer words have been used. In spite of its barbarous nature pidgin was quite sufficient for all commercial transactions and in default of any other means of communication could cover a wide range of subjects. (Crow 2011 [1940]: 31)

As with studying other historical languages, data on CPE are small and fragmentary, though they come from a wide variety of sources. CPE recorded in the English language appears in diverse sources, including travelogues, journals, memoirs, newspaper and magazine articles, guide books, etc.

In Chinese language sources, which are mainly instructional materials, CPE is transcribed in Chinese characters. Various versions of anonymously written socalled redhaired phrasebooks containing a vocabulary of about several hundred essential words or phrases in pidgin English, like the one mentioned in Hunter (1882), were sold in Canton (Bolton 2003). It is clear that these coarsely made, cheap phrasebooks were prepared by and for ordinary people. A notable instructional manual source is called *The Chinese and English Instructor* 英語集全 (Tong 1862). Note that the "Chinese" in the title refers to the local dialect, Cantonese instead of Mandarin. Apart from Cantonese and English, what makes the book a valuable source is the inclusion of pidgin equivalents of Cantonese and English dialogues in some parts of the book. Due to the comprehensiveness of its content and the systematicity of the transcription, this source is of great importance to the grammatical analysis of CPE. The compiler of the *Instructor* was a prominent Cantonese merchant called Tong Ting-kü 唐廷樞 [1832–1892]. In traditional Chinese society, the merchant class was considered the lowest in the Chinese social hierarchy, and many of them had no or limited formal education. Tong was an exception, having attended the Morrison Mission School in Hong Kong, and was highly praised for his proficiency in English. Most of the CPE examples from the Chinese-language sources are taken from this book.

Michelle Li

### **4 The users of CPE**

Accounts from historical sources indicate that the users of CPE, Chinese and foreigners alike, came from the non-elite, lower social class, thus were people "from below." This section will look at the major groups of users constituting the speech community of CPE.

### **4.1 European traders and other foreigners**

The foreign community in the old Canton city was multinational, so the immediate challenge was to find a common language for communication with the Chinese. Therefore, economic, social, and political factors contributed to the rise of CPE as the lingua franca. The need to be familiar with CPE was evident in John Robert Morrison's [17 April 1814–29 August 1843] book *A Chinese commercial guide consisting of a collection of details respecting foreign trade in China* (1834) where a "glossary of words and phrases peculiar to the jargon spoken at Canton" was found. The jargon mentioned here referred to the Canton pidgin. Some examples in the glossary are as follows.

(1) *Can do?* 'Will it do?'

*Chop-chop* 'quick, fast', as *too muchy chop-chop* for 'very quick'

In the 19th century, the function of CPE expanded as families of foreign traders were allowed to reside in China and there was need to hire Chinese servants to help them in their housekeeping. The dialogue in (2) is a communication between a European master and a Chinese boy.

(2) Boy: *You makee ling?* 'Did you ring, sir?' Master: *Yes, sendee catchee one piecee tailor man.* 'Yes, send for a tailor.' Boy: *Just now have got bottom side.* 'He is below at present.' Master: *Show he come top side.* 'Tell him to come up.' (Anonymous 1860: 43)

### **4.2 Chinese**

### **4.2.1 Merchants and compradors**

As Chinese government officials did not trade directly with foreign traders, merchants and compradors functioned as their intermediaries. In order to perform their duties, "[t]he comprador[s] … speak a broken English mixed up with Portuguese, some Dutch, and French, the same as most of the Chinamen who come

### 7 Language history from below: Pidgins and Creoles as examples

about the ship. It is rather difficult to understand them at first, but one soon gets used to hearing them" (Tyng 1999: 29).

Though of the lowest class, a merchant could become rich and influential through his familiarity with Westerners and Western cultures; "through his expertise in pidgin English and his knowledge of the West, he became a middleman between East and West, not only economically but also socially, politically, and culturally" (Hao 1970: 180). A good example is the author of the *Chinese and English Instructor*, Tong Ting-kü, who was highly respected by the Chinese as well as Westerners.

### **4.2.2 Interpreters**

Another group of Chinese who had close connections with the foreign community were interpreters, also referred to as linguists in the 18th and 19th centuries (Hunter 1882; Van Dyke 2005: 50). Like merchants, interpreters were indispensable partners of foreign traders. Linguists were appointed by the Hoppo (administrator of Canton Customs) to act as interpreters. Apart from providing translation services for the Chinese and Westerners, their duties and functions were various, including managing the daily activities of the foreigners (Chin 2016). In the early 18th century, as many interpreters came from Macao, they tended to also speak (pidgin) Portuguese. The example in (3) is a linguist's translation of a speech in Chinese to "China Anglish."

(3) *Yeckhing, chin chin the gentleman alla proper. Yeckhing very much oblige to soupcarg, who have wantchee buy him cargo pigeon. He chin chin gentlemen good voyagee, hopee go home his country No.1 good, and catchee many per cent.*

'Yeckhing thanks the gentlemen assembled, for their polite attention, and is very much obliged to the supercargo who bought his chop; he wishes him a pleasant return voyage, and hopes that he will derive a handsome profit on the purchase.' (original translation, Tiffany 1849: 126)

### **4.2.3 Domestic servants**

As mentioned above, the growth of the foreign community in China resulted in more Chinese being employed in foreign households. Zhang (2009: 309) shows that, in early Hong Kong, typical European families employed anywhere from three to more than ten Chinese servants. This means that not only was there an Michelle Li

increase in the number of users of CPE, but more importantly, that the expansion in domains of use triggered an elaboration of the pidgin's vocabulary and grammar.

### **4.2.4 Other users**

The foreign community's activities were not restricted to the office and home; they were also in contact with people like tailors, barbers, shopkeepers, prostitutes, etc. who provided various services and sources of entertainment to them. Shops could be found along the Old China Street and Hog Lane adjoining the Thirteen Factories in Canton. Example (4) shows a typical exchange between a shopkeeper and a European.

(4) European: *How you do, Hipqua?* Chinese: *Welly wen, tankee; how you do?* European: *I well. What have got?* Chinese: *Anyting have got. What ting wantyee?* European: *I no sabee; lettee my see something. How muchee this cigar-boxee?* Chinese: *Oh! dat cigar-boxee! dat tree quart dollar.* European: *Too muchee. More cheap have got?* Chinese: *No; more cheap no got.* (Duer 1860: 301)

### **5 Linguistic variation in the preposition** *long*

As speakers of Chinese Pidgin English come from different cultural and linguistic backgrounds, linguistic variation is expected. The preposition *long* in CPE demonstrates linguistic variation at different levels. Li (2011) shows that the functions and meanings of *long* are a conflation of its source languages: *along (with)* in English and *tung4* in Cantonese. The double etymology of *long* is evident in the following linguistic variation:


### 7 Language history from below: Pidgins and Creoles as examples

In Cantonese, the preposition *tung4* is placed before the verb. To illustrate how the word order of *long* varies in the sources, take the most productive meaning 'with' as an example. In the CPE data, *long-*prepositional phrases can be found after the verb phrase (5), as in English, or before it (6), as in Cantonese *tung4*. Compare (5 ′ ) and (6 ′ ), which are the Cantonese translations of (5) and (6), respectively. It is clear that the postverbal placement of *long* does not come from Cantonese, but is modelled on English.

	- 3 with guest talk ASP thing

The meanings of *long* also recombine the meanings in the source languages. In addition to the comitative usage shown above, other meanings of *long* are used to indicate source, benefactive, and malefactive. The occurrences of each meaning in Chinese- and English-language sources are summarized in Table 7.1.

je5

The comitative use of *long* conflates English *along* (*with*) and Cantonese *tung4*. As both share this meaning, this may explain the high frequency of this use. Other uses of *long* contributed by Cantonese *tung4* include source and benefactive, as shown in examples (7) and (8) respectively. The malefactive meaning in (9) seems to come from one use of *for* in English (Li 2011).

	- my [long you] takee some (Tong 1862: VI.12) 'I will buy some from you'

my too much fear some war ship mans want for make bobbily [long china mans] (Tilden 1834-36: 968)

'I fear very much that the sailors want to make troubles for the Chinese'

### Michelle Li

Table 7.1: Distribution of the meaning and syntax of the preposition *long* in Chinese Pidgin English (Li 2011).


A closer look at Table 7.1 suggests a close relationship between the semantics, syntax, and source of attestation of *long*. For example, though both Cantonese and English contribute to the comitative meaning, there is an overwhelming preference (90%) for the postverbal word order in the English language sources, which suggests that the writers' language might have come into play. The two instances of a preverbal *long* in English-language sources and the more or less balanced distribution of different word orders of *long* in Chinese-language sources may be due to varying degrees of acculturation. The source and benefactive meanings are exclusively attested in Chinese-language sources and syntactically in the preverbal position, demonstrating a clear case of the influence of the substrate language, Cantonese.

In sum, the preposition *long* in CPE shows the following features: first, the form and function of *long* show different levels of recombination of features from English and Cantonese grammar (see Li 2011 for a detailed account of this conflation of features); second, some uses of *long* show a close relationship between semantics and syntax; and third, the presence or absence of certain uses of *long* in certain types of sources may be related to the linguistic backgrounds or acculturation of the writers.

### **6 Conclusion**

Recent studies on the historiography of major languages have laid an important foundation for the development of theories and methods for investigating language history. The study of pidgins and creoles offers language historians additional perspectives for understanding languages evolution in intense and rapid contact situations. Using Chinese Pidgin English as an example, this paper shows that contact languages are good examples of "language history from below" because, unlike standard varieties, pidgins and creoles developed without much human intervention and were subject to little pressure of standardization. As a result, language variation instead of homogenization seems to be the norm. An example of such variation in language use has been demonstrated with the preposition *long* in CPE. Recent developments in historical linguistics, especially historical sociolinguistics, have given rise to significant advances in theories and methodologies. A shift of focus from the study of higher register and standard varieties to that of the language use of ordinary people, such as speakers of pidgins and creoles, can provide new insights for the "from below" approach to language history.

### **References**

Anonymous. 1860. *The Englishman in China*. London: Sauders, Otley & Co.


### Michelle Li

Downing, C. Toogood. 1838. *The Fan-Qui in China in 1836-7*. Vol. 1. London: Henry, Colburn Publisher.

Duer, John K. 1860. Pigeon-English. *Knickerbocker* 55. 300–303.


## **Chapter 8**

## **Analyse comparative des** *Élémens de la grammaire françoise* **de Lhomond et de ses** *Élémens de la grammaire latine*

### Sophie Piron

Université du Québec à Montréal

This paper compares two grammars published by Lhomond at the end of the 18th century. One is dedicated to Latin, the other to French. The paper shows how similar both grammars are. They are constructed in a way that the French grammar can be used as an introduction to the Latin grammar. Most importantly, the French grammar has its specific theoretical organization, based on the parts of speech. Such an organization makes concordance syntax fall into an orthographic approach.

### **1 Introduction**

Charles-François Lhomond publie ses *Élémens de la grammaire latine* (EGL) en 1779. Il existe assez peu d'informations sur les éditions parues du vivant de l'auteur. Il semble y en avoir eu neuf, mais la liste complète reste lacunaire, et toutes les éditions ne sont pas disponibles : <sup>1</sup> 1779, <sup>2</sup> 1780, <sup>3</sup> 1781, <sup>4</sup> 1784, <sup>9</sup> 1793. La troisième édition, qui remonte à 1781 [1779], est la plus récente qui soit consultable à la Bibliothèque nationale de France. Elle servira par conséquent de référence dans le cadre de notre étude.

Lhomond publie les *Élémens de la grammaire françoise* (EGF) un an après sa grammaire latine. La liste des éditions connues et parues du vivant de l'auteur est, elle aussi, plutôt réduite : <sup>1</sup> 1780, <sup>5</sup> 1786 et <sup>7</sup> 1790. Les différences entre ces éditions sont mineures. Elles relèvent de la mise en page et du style. L'édition de 1790 [1780], la dernière du vivant de l'auteur, servira ici de référence.

Sophie Piron. 2020. Analyse comparative des *Élémens de la grammaire françoise* de Lhomond et de ses *Élémens de la grammaire latine*. In Émilie Aussant & Jean-Michel Fortis (éd.), *Historical journey in a linguistic archipelago : Descriptive concepts and case studies*, 129-141. Berlin : Language Science Press. DOI : 10.5281/zenodo.4269421

### Sophie Piron

Ces deux publications ont joué un rôle capital dans l'histoire de la grammaire scolaire, tant latine que française. L'importance se jauge à la fois sur le plan éditorial et sur le plan grammatical. Les nombreuses rééditions du vivant de l'auteur, mais surtout tout au long du XIX<sup>e</sup> siècle (et même encore au début du XX<sup>e</sup> ), ainsi que les adaptations, confirment le succès de librairie sur le marché grammatical. Le catalogue en ligne de la BnF propose ainsi environ 320 ouvrages de Lhomond et de ses adaptateurs pour la grammaire latine et à peu près 480 pour la grammaire française. Pour cette dernière, Chervel a comptabilisé que "La Bibliothèque nationale en possède 760 exemplaires, d'éditions toutes différentes" (Chervel 1977 : 63). Quel que soit le chiffre exact des rééditions et adaptations, le succès de ces deux ouvrages est incontestable.

Aucune grammaire française, sans doute, n'a connu un succès aussi durable que celle de Lhomond. (Chervel 1977 : 63)

Les *Élémens de la grammaire latine* de LHOMOND restent, et de loin, le manuel le plus utilisé au XIX<sup>e</sup> siècle […]. (Chervel 1979 : 16)

Avec le succès commercial désormais attaché au nom de Lhomond croît sa valeur sur le marché symbolique de la grammaire, que ce soit pour les *Élémens* français (Chervel 1977 ; 2006 ; Colombat et al. 2010) ou les *Élémens* latins (Chervel 1977 ; Colombat 1999). Si aucune innovation théorique ne porte son empreinte, l'apport de Lhomond reste incontestable, mais il se situe ailleurs. Il relève, pour les EGF au moins, d'un tour de force pédagogique opéré sur le matériau grammatical et l'ordre de présentation au sein de l'ouvrage (Piron 2019).

La date clé dans l'histoire de la production grammaticale à usage scolaire au XVIII<sup>e</sup> siècle n'est pas le Restaut de 1732, qui distingue les deux orthographes, ni le Wailly de 1754 qui abandonne la déclinaison, mais le petit Lhomond de 1780. (Chervel 2006 : 220)

En France, ils [les EGL] ont constitué la grammaire latine de référence pendant près d'un siècle […]. Bien que fondamentalement différents, la méthode de Du Marsais et la grammaire de Lhomond sont sans doute les travaux qui ont le plus marqué l'enseignement du latin au XVIII<sup>e</sup> siècle. […] la première cédera assez rapidement la place à la seconde qui offrira à l'enseignement du latin un fondement d'autant plus incontesté que son caractère traditionnel ne risquait pas de choquer […]. (Colombat 1999 : 106)

L'objectif du présent article est de comparer les grammaires latine et française de Lhomond, telles qu'elles ont été publiées du vivant de l'auteur. La comparaison se trouve justifiée par quatre arguments. Tout d'abord, la grammaire latine

### 8 Analyse comparative des Élémens… de Lhomond

constitue le modèle descriptif en usage depuis la naissance de la grammaire française, mais sous une forme évoluée, *la grammaire latine étendue* (Auroux 1994). Par ailleurs, à partir du XVII<sup>e</sup> siècle (avec la *Nouvelle méthode latine* de Port-Royal, en particulier depuis la réédition de 1650), "la langue latine est désormais décrite du point de vue du francophone" (Colombat 1995 : 11), ce qui transforme encore davantage le modèle latin, mais resserre en même temps les liens entre grammaire latine et grammaire française. À cet égard, les EGL de Lhomond inaugurent une certaine forme de modernité en proposant "une méthode d'initiation à la langue latine à l'usage des francophones" (Colombat 1999 : 166).

Ensuite, du côté du français cette fois, le XVIII<sup>e</sup> siècle confère désormais à la grammaire française le rôle de porte d'entrée pour l'étude du latin, et donc de sa grammaire (Chevalier 2006 [1968], Chervel 1977). Ce rôle est d'ailleurs reconnu aux *Élémens* français de Lhomond : "Par bien des aspects, l'ouvrage [EGF] ne constitue qu'une propédeutique à l'étude du latin." (Colombat 1999 : 165). Lhomond lui-même le précise dans sa préface aux EGF : "C'est par la langue maternelle que doivent commencer les Etudes […] & cette connoissance leur [aux enfants] sert comme d'introduction aux Langues anciennes qu'on veut leur enseigner." (Lhomond 1790 [1780] : 3). Enfin, chose on ne peut plus évidente, la similitude des titres – *Élémens de la grammaire françoise, Élémens de la grammaire latine* – invite à un parcours comparatif. Nous prendrons les EGF comme point de référence, ce qui ne correspond pas à la chronologie de parution des ouvrages, mais bien à leur articulation.

### **2 Structure**

Malgré une structure peu explicite, les EGF se divisent en deux parties. La première est consacrée aux espèces de mots<sup>1</sup> et occupe presque la totalité de l'ouvrage (76 pages). La seconde traite de l'orthographe (10 pages). Les EGL présentent, quant à eux, trois parties : les espèces de mots (130 pages), la syntaxe latine (59 pages) et la méthode latine pour traduire les gallicismes en latin (88 pages). Les plans rédactionnels des deux ouvrages semblent donc assez différents au premier abord et ne suivent pas les mêmes proportions pour le traitement des

<sup>1</sup>Lhomond n'utilise pas l'expression *partie du discours*, mais *espèce de mot.* Cet usage se coupe volontairement de la tradition, on le trouve cependant chez d'Olivet (1767), à qui Lhomond se réfère par ailleurs (1790 [1780] : 63). Lhomond ne justifie pas ce choix terminologique, par contre il justifie celui d'autres termes comme *nom* à la place de *substantif* . Comme il l'explique dans la préface des EGL et dans celle des EGF également, il recherche une terminologie plus claire, une terminologie que les enfants puissent entendre.

### Sophie Piron

parties du discours. La présence, dans les EGL, d'un chapitre consacré à la syntaxe est pratique courante dans les grammaires latines et, plus spécifiquement, dans les ouvrages de type *rudiments* (Colombat 1999) : on relèvera surtout le texte phare de Lancelot (1644), mais aussi par exemple Chompré (1751). Il existe également des ouvrages de type *méthode*, qui s'inscrivent dans la pratique du thème. Ces techniques de traduction peuvent suivre des notions de syntaxe, ce que propose déjà Lancelot (1644).

Il est normal de ne pas trouver de méthode de traduction dans les EGF. Par contre, l'absence d'une section explicitement consacrée à la syntaxe dans cette petite grammaire française est ce qui frappe le plus quand on compare les deux grammaires de Lhomond. Cela étant dit, si la syntaxe constitue un incontournable dans une grammaire latine, elle se définit moins comme un passage obligé pour les grammaires élémentaires du français. En effet, sur neuf publications similaires aux EGF et parues au XVIII<sup>e</sup> siècle, seules quatre proposent une partie consacrée à la syntaxe : Gaullyer (1722), Wailly (1759), Bertera (1773) et Royon (1777) 2 . La présence d'une partie explicitement consacrée à l'orthographe dans les EGF relève, en revanche, d'une pratique bien implantée dans les petites grammaires du français. De fait, dans les grammaires que nous avons consultées à ce sujet, seuls Wailly (1759) et Viard (1763) ne proposent pas de chapitre intitulé *orthographe.* Cela ne signifie pas pour autant que ces notions n'y soient pas abordées.

Au-delà de ces différences entre EGL et EGF, le traitement des parties du discours y est rédigé selon le même principe : d'abord, une série de chapitres proposant des définitions et observations générales pour chaque partie du discours ; ensuite, un chapitre (assez long) apportant des précisions sur chaque espèce de mot. Ce chapitre s'intitule *suppléments* dans les EGL, *remarques particulières* dans les EGF. Un tel découpage de la matière grammaticale en une section plus générale et une autre plus complexe est assez fréquent, à l'époque de Lhomond, dans les grammaires consacrées au latin (Colombat 1999). Cette pratique de dédoublement semble également exploitée dans les grammaires françaises, bien qu'elle soit parfois difficile à déceler ou qu'elle consiste à présenter de nouveau la matière sous un angle un peu différent : on pensera notamment à Wailly (1759), chez qui des remarques sur les parties du discours prennent place après une partie consacrée aux différentes classes de mots et à la syntaxe ; on pensera aussi à Restaut (1732a), qui propose des observations sur les parties du discours après les avoir toutes passées en revue au sein de différents chapitres.

<sup>2</sup> Sans syntaxe : Restaut (1732a), Panckoucke (1749), Viard (1763), Chompré et al. (1778) et Domergue (1778).

### 8 Analyse comparative des Élémens… de Lhomond

La liste des parties du discours est identique, sauf pour l'article (qui n'est présent que dans les EGF) et pour le participe (qui est, en latin, mis en équivalence avec le gérondif et le supin). L'ordre est, lui aussi, le même si ce n'est que la préposition est vue avant l'adverbe dans les EGF et après celui-ci dans les EGL. Le point fréquemment relevé (notamment Chervel 1977, Colombat 1999) concerne l'adjectif : il est hissé au rang de partie du discours, et ce dès 1779 dans les EGL. Comme le signale Colombat (1999), cette position a déjà été adoptée par la grammaire latine de Goulier (1773). Pour le français, les prédécesseurs de Lhomond sont Girard (1747) et Beauzée (1767), mais aussi les grammaires élémentaires de Royon (1777) et de Domergue (1778).

### **3 Traitement des parties du discours**

Les EGL et les EGF adoptent un traitement identique des parties du discours. Le développement se déploie en cinq temps, plus ou moins perceptibles dans la rédaction de ces deux grammaires et dans leur mise en page : la définition de la partie du discours traitée ; les catégories qui affectent la classe (traditionnellement nommées *accidents*, mais ce terme n'est pas utilisé), à savoir le genre, le nombre, la personne, etc. ; la morphologie du mot lorsqu'il varie (féminin, pluriel, conjugaison) ; la syntaxe d'accord et la syntaxe de régime.

### **3.1 Définition**

La comparaison entre les EGL et les EGF pour ce qui est de la définition des espèces de mots fait clairement apparaître une volonté rédactionnelle d'uniformiser les deux publications. Il s'agit de proposer deux ouvrages qui puissent être utilisés en séquence, sans coupure théorique.

Le nom, l'adjectif, le pronom et le verbe sont définis de manière pratiquement identique, voire identique entre les deux textes, Lhomond ayant seulement supprimé des EGF tout ce qui avait rapport au latin. Ainsi, aucune mention n'est faite des cas et des déclinaisons dans les définitions du nom et de l'adjectif.

L'Adjectif est un mot que l'on ajoute au nom pour marquer la qualité d'une personne ou d'une chose, comme *bon* pere, *bonne* mere ; *beau* livre, *belle* image : ces mots, *bon*, *bonne*, *beau*, *belle*, sont des adjectifs joints au nom *pere*, *mere*, & c. (Lhomond 1790 [1780] : 12)

*L'Adjectif* est un mot que l'on ajoute au nom pour marquer la qualité d'une personne ou d'une chose, comme *bon* pere, *bonne* mere ; *beau* livre, *belle*

### Sophie Piron

image. *Bon*, *bonne*, *beau*, *belle*, sont des adjectifs : ils se déclinent en latin, & ils ont les trois genres, masculin, féminin & neutre. (Lhomond 1781 [1779] : 113 )

Le participe est la partie du discours qui reçoit le traitement le plus différencié entre les deux grammaires. En latin, il est défini au premier chef comme un adjectif, même s'il tient à la fois du verbe et de l'adjectif. Il est surtout défini par le fait qu'il s'accorde, comme un adjectif. Or la définition de l'adjectif n'avait pas présenté cette catégorie de mot sous cet angle. Dans la version française de la grammaire, les caractéristiques du participe ressortissant à l'adjectif en reproduisent la définition, entièrement fondée, cette fois, sur la qualification.

Le *Participe* est un mot qui tient du verbe & de l'adjectif, comme *amant, aimé* : il tient du verbe, en ce qu'il en a la signification & le régime : *aimant Dieu, aimé de Dieu* : il tient aussi de l'adjectif, en ce qu'il qualifie une personne ou une chose, c'est-à-dire, qu'il en marque la qualité. (Lhomond 1790 [1780] : 60)

Les *Participes* sont des Adjectifs qui viennent des verbes ; ils s'accordent en genre, en nombre & en cas avec le nom auquel ils sont joints, & de plus ils gouvernent le même cas que le Verbe d'où ils viennent ; c'est pour cela qu'on les nomme *Participes*, parce qu'ils tiennent de l'adjectif et du verbe. (Lhomond 1781 [1779] : 90)

Les autres parties du discours – adverbe, préposition, conjonction et interjection – sont définies avant tout comme des mots indéclinables dans les *Élémens* latins.

L'*Adverbe* est un mot indéclinable, qui se joint le plus souvent à un Verbe, & en détermine la signification. (Lhomond 1781 [1779] : 92)

La *Préposition* est un mot indéclinable, qui joint à un Nom, ou à un Pronom, veut ce Nom ou Pronom à l'Accusatif ou à l'Ablatif. (Lhomond 1781 [1779] : 94)

La *Conjonction* est un mot indéclinable qui sert à lier les parties du discours. (Lhomond 1781 [1779] : 96)

<sup>3</sup>Les exemples suivants d'adjectifs seront en latin, seuls ceux qui permettent d'introduire la notion d'adjectif sont en français.

### 8 Analyse comparative des Élémens… de Lhomond

L'*Interjection* est un mot indéclinable qui sert à marquer les différens mouvemens de l'ame. (Lhomond 1781 [1779] : 97)

Lhomond a évacué cet angle morphologique de ses *Élémens* français et les inscrit ainsi dans un projet délatinisant de la grammaire. Ce n'est que plus tard que les adaptateurs de Lhomond adopteront un angle orthographique et introduiront le concept de *mot invariable* en remplacement de *mot indéclinable* (par exemple, Le Tellier 1811).

On décèle quelques différences d'approche dans les définitions des mots invariables, excepté pour l'interjection. Si la préposition sert à joindre dans les deux ouvrages, un éclairage différent prend place dans cette partie du discours. En latin, la préposition s'inscrit dans une relation à deux termes : la préposition et le mot dont elle régit le cas. En français, elle s'inscrit dans une relation à trois termes : les deux mots reliés par la préposition, et la préposition elle-même. Cette fois, la préposition ne régit pas de cas (cependant elle régit bien le mot qui suit), mais marque un rapport d'ordre sémantique (le lieu, l'ordre, l'union, etc.).

L'adverbe, dans les EGL, se joint habituellement à un verbe, mais rien n'est dit sur les autres mots qu'il pourrait déterminer, tandis qu'il est précisé dans les EGF qu'il peut également se joindre à un adjectif. À l'inverse, la conjonction présente un comportement plus restreint dans les EGF : au lieu de pouvoir lier n'importe quelles parties du discours, elle ne joint que des phrases ou propositions.

### **3.2 Accidents**

Après chaque définition d'une espèce de mots variables, Lhomond propose habituellement les accidents<sup>4</sup> qui l'affectent, en tout cas ceux qui ont une incidence sur l'apprentissage élémentaire de la langue : genre, nombre, personne, mode et temps. Ces attributs sont identiques en latin et en français, mais ne prennent pas toujours les mêmes valeurs. Ainsi, en latin, outre le masculin et le féminin, la valeur du genre peut être le neutre. Autre exemple, le conditionnel est un mode en français, alors qu'il est absent de la liste des modes en latin.

La présentation des accidents manque quelquefois de systématicité. Ainsi, l'adjectif possède un genre, mais rien n'est dit sur le nombre alors même que le chapitre propose une règle pour la formation du pluriel des adjectifs. Relevons, par ailleurs, que le nom ne possède pas la catégorie de la personne, alors que celle-ci est mise en œuvre dans la règle d'accord du verbe avec son sujet : "Tout verbe doit être du même nombre & de la même personne que son nominatif ou sujet"

<sup>4</sup>Lhomond n'utilise pas le terme *accident* et n'en propose aucun autre en remplacement.

### Sophie Piron

(Lhomond 1790 [1780] : 47) ; "Tout Verbe s'accorde en nombre & en personne avec son nominatif" (Lhomond 1781 [1779] : 31). Le chapitre consacré au verbe ne présente pas non plus très clairement la personne comme un accident de cette espèce de mot, mais plutôt comme une information qui provient de l'emploi des pronoms *je, tu,* etc. Quant à la classe des pronoms, elle n'est pas mise en relation avec la catégorie de la personne. Les pronoms personnels sont toutefois présentés sous cet angle<sup>5</sup> . Il ne faut pas imputer cette incohérence à Lhomond lui-même, qui ne fait ici que s'inscrire dans la tradition. À l'époque, la notion de personne, centrale dans l'accord du verbe, n'est pas attachée au substantif puisque celui-ci ne présente pas de marque morphologique de personne (Colombat 1999).

### **3.3 Morphologie**

Les aspects morphologiques nécessitent de nombreux tableaux dans les EGL puisqu'il faut présenter les différents modèles de déclinaison du nom<sup>6</sup> et de l'adjectif, et également les déclinaisons de chaque type de pronom. Il faut aussi présenter les modèles de conjugaison<sup>7</sup> . À part pour les pronoms, les EGF ne requièrent de tableaux que dans le traitement du verbe<sup>8</sup> . En effet, Lhomond a évacué de la grammaire française la déclinaison du nom, choix déjà répandu à l'époque, mais pas encore entériné par tous les grammairiens. Plusieurs d'entre eux qui publient justement des grammaires élémentaires à la même époque que Lhomond proposent encore des tableaux de déclinaison associés au nom : notamment Viard (1763), Bertera (1773), Royon (1777)<sup>9</sup> et Chompré et al. (1778).

Lhomond procède à une délatinisation de la grammaire française en excluant ainsi la déclinaison de ses EGF. C'est un choix qui mérite encore d'être souligné à la fin du XVIII<sup>e</sup> siècle, d'autant que cette option théorique établit une différence non négligeable entre les EGL et les EGF. En effet, soulignons-le une fois de plus, malgré certaines apparences, les deux ouvrages sont construits de manière similaire et de façon à pouvoir être utilisés en séquence. Se soustraire à la tradition de la déclinaison en français, c'est se couper d'une passerelle pédagogique ai-

<sup>5</sup>L'attribut y est exploité comme un élément de désignation d'une véritable personne (celle qui parle, à qui on parle ou de qui on parle) plutôt que comme un trait grammatical.

<sup>6</sup>Le nom est la classe de mots la plus complexe de ce point de vue avec cinq déclinaisons et dix modèles puisque certaines déclinaisons présentent plusieurs comportements.

<sup>7</sup>Douze tableaux : quatre conjugaisons pour chacune des trois sortes de verbes (actif, passif, déponent).

<sup>8</sup>Douze tableaux : *avoir* ; *être* ; verbes en -*er*, en -*ir*, en -*oir*, en -*re* ; temps primitifs réguliers, puis irréguliers ; verbes passifs, neutres, réfléchis et impersonnels.

<sup>9</sup>Royon (1777) propose des tableaux de déclinaison, très complets, mais lors du traitement de l'article.

### 8 Analyse comparative des Élémens… de Lhomond

sée vers le latin. Néanmoins, cette délatinisation laisse place, en filigrane, à une conception très latine de la morphologie nominale et adjectivale. Dans les EGL, la déclinaison est rapidement introduite : "En latin le Nom change sa derniere syllabe" (Lhomond 1781 [1779] : 2). Dans les EGF, les règles morphologiques sont présentées sous le même angle, comme une modification de la finale du mot : "Pour former le pluriel, ajoutez *s* à la fin du nom" (Lhomond 1790 [1780] : 9) ; "Quand un adjectif ne finit point par un *e*, on y ajoute un *e* muet pour former le féminin" (Lhomond 1790 [1780] : 12). Bien sûr, le lien entre la déclinaison latine et les modifications morphologiques du français reste ténu, mais il doit être souligné, nous semble-t-il. Enfin, tout en établissant en français une conception morphologique prête à être réinvestie dans l'étude du latin, Lhomond propose un traitement orthographique de la formation du pluriel et du féminin. Il en expose les principales règles : -*x* (*caillou, cailloux*), consonne double (*cruel, cruelle*), -*que* (*public, publique*), etc. Fait non négligeable, il positionne cette information très tôt lorsqu'il aborde une partie du discours. Ce choix montre une volonté d'associer l'orthographe à la présentation des parties du discours. Celles-ci sont donc conceptualisées sous cet angle. Il s'agit, plus précisément, de l'orthographe de principe de Restaut, celle des "différentes terminaisons des noms par rapport aux genres ou aux nombres, & des verbes par rapport aux tems & aux personnes" (Restaut 1732b : 48), et non de l'orthographe d'usage.

### **4 Syntaxe d'accord**

Lhomond établit clairement la différence entre *syntaxe d'accord* et *syntaxe de régime*. Il le fait au commencement de la partie consacrée à la syntaxe dans les EGL, soit à peu près à la moitié de l'ouvrage, mais beaucoup plus tôt dans les EGF, plus précisément dans une note de bas de page au sein du chapitre consacré aux adjectifs. Il définit la syntaxe d'accord comme la syntaxe "par laquelle on fait accorder deux mots en genre, en nombre, & c." (Lhomond 1790 [1780] : 15 ; Lhomond 1781 [1779] : 131). Ce n'est autre que la syntaxe de concordance ou de convenance, généralisée au XVI<sup>e</sup> siècle et bien en place à l'âge classique, dans les grammaires latines en particulier (Colombat 1999).

Fait majeur : les aspects syntaxiques que Lhomond aborde dans la partie *syntaxe* des EGL sont classés par espèce de mots (syntaxe des noms, des adjectifs, des verbes, des pronoms, des participes, des prépositions, des adverbes et des conjonctions) et ensuite par type de syntaxe. Ils dédoublent à l'occasion la règle principale qui a déjà été présentée dans les chapitres consacrés aux différentes espèces de mots, mais c'est véritablement dans cette 2<sup>e</sup> partie du livre que l'exposé

### Sophie Piron

des règles est fait rigoureusement. La terminologie utilisée mérite d'être relevée. Dans la 1re partie de l'ouvrage (les espèces de mots), Lhomond fait mention de règles (*règle des adjectifs, des pronoms adjectifs, du* qui *et du* que *relatifs* et *de la règle générale pour les verbes*). Dans la 2<sup>e</sup> partie (la syntaxe), il utilise cette fois le terme *accord* (*accord de deux noms, de l'adjectif avec le nom, du verbe avec le nominatif ou sujet, du pronom avec l'antécédent*), excepté pour le participe, dont la règle est exposée en tant que *syntaxe des participes.* Lhomond a ainsi ajouté d'autres règles à celles sur lesquelles s'entendaient les grammaires latines humanistes : le verbe avec son nominatif, l'adjectif avec le substantif et le relatif avec son antécédent (Colombat 1999). Ne consacrant aucun chapitre explicite à la syntaxe dans ses EGF, Lhomond a intégré les règles d'accord dans la première partie de l'ouvrage, plus précisément dans chaque partie du discours concernée. Des titres détachent clairement les sections consacrées à la syntaxe d'accord et c'est bien ce terme qui est employé : *accord des adjectifs avec les noms, accord des verbes avec leur nominatif, accord du participe passé.* Il arrive toutefois que le titre n'utilise pas le terme *accord*, mais use simplement du mot *règle* : *règle des pronoms, règle du* qui *ou* que *relatif.*

### **5 Syntaxe de régime**

Lhomond définit la syntaxe de régime comme la syntaxe, en français, "par laquelle un mot régit *de* ou *à* devant un autre mot" (Lhomond 1790 [1780] : 15), en latin, "par laquelle un mot régit un autre mot à tel cas, à tel mode, & c." (Lhomond 1781 [1779] : 131). Les aspects syntaxiques présentés dans les EGF couvrent toutefois plus que ce que ne prévoit la définition étroite qui y est donnée. La ventilation de la matière reproduit le choix rédactionnel de ce qui a été fait pour la syntaxe d'accord, tant dans les EGL que dans les EGF. De nouveau, le plan rédactionnel des EGF intègre la syntaxe – de régime, cette fois – dans le traitement de chaque espèce de mot, à la suite de la syntaxe d'accord lorsque les deux syntaxes sont à l'œuvre. Si le terme *régime* n'est pas systématique, la présence de la syntaxe de régime est indéniable : pour joindre un nom à un mot précédent ; régime des adjectifs, régime des verbes (actifs, passifs, neutres, etc.) et des conjonctions.

### **6 Conclusion**

La comparaison des deux grammaires élémentaires de Lhomond a permis de constater que seule la partie consacrée aux espèces de mots se retrouvait dans la structure des deux publications. Au-delà de cette similitude très sommaire, plusieurs parallèles ont été relevés. Les définitions des parties du discours sont très proches, voire parfois identiques. Le traitement en deux niveaux de difficulté est reproduit dans les EGF. Le choix de considérer l'adjectif comme une espèce de mot à part entière prend place dès 1779 dans les EGL. Chaque partie du discours est abordée selon un mouvement en cinq temps, lorsque cela est possible : définition, accidents, formes, syntaxe d'accord, syntaxe de régime.

Le véritable point commun entre les deux grammaires de Lhomond est la charpente grammaticale. Celle-ci est constituée des espèces de mots : elle soutient les suppléments, la syntaxe, l'orthographe. Lhomond opère toutefois un changement majeur lors de la rédaction des *Élémens* français. En effet, il renforce le rôle de cette structure sous-jacente en intégrant complètement la syntaxe dans les espèces de mots. Le plan rédactionnel était en germe dans les EGL, où la syntaxe s'immisçait dans les espèces de mots.

Lhomond procède à la fois à une délatinisation et à une latinisation en filigrane de la grammaire française. Tout en supprimant les déclinaisons et le concept de mot indéclinable, il parvient à traiter la variation sous l'angle de la morphologie flexionnelle, mais par le fait même il l'envisage sur le plan de l'orthographe. Les *Élémens* français proposent ainsi une évolution majeure sur deux points. La grammaire est désormais une grammaire des mots, qui déploie tout ce qui s'y rapporte : orthographe de principe, syntaxe d'accord, syntaxe de régime, mais aussi orthographe d'usage. La nature d'un mot est désormais la porte d'entrée dans la grammaire, ce qui la rend beaucoup plus aisée à consulter et à comprendre. Autre point d'évolution : le positionnement de la syntaxe d'accord fait basculer cette partie de la syntaxe dans le domaine de l'orthographe. Ainsi transformés, les *Élémens de la grammaire françoise* constituent le point de départ de la grammaire scolaire.

### **Références**


Olivet, Pierre-Joseph d'. 1767. *Remarques sur la langue françoise*. Paris : Barbou.


## **Chapter 9**

## **Crosscurrents in linguistic research: Humanism and positivism in Central Australia 1890–1910**

### David Moore

University of Western Australia

Trained in nineteenth century humanist traditions of philology, German Lutheran missionaries conducted linguistic fieldwork in the Dieri (Diyari) language near Lake Eyre in South Australia and in the Aranda (Arrarnta, Arrernte) and Loritja (Luritja) languages at Hermannsburg in the Northern Territory. As the discipline became increasingly positivist in the late nineteenth century, anthropologists and linguists with this very different orientation also took an interest in the languages of Central Australia. In this paper I contrast humanist and positivist researchers of Central Australian languages arguing that common metascientific orientations are more significant factors than nationality for understanding their research.

### **1 Introduction**

The period 1890–1910 saw the publication of the first comprehensive grammars and dictionaries of Central Australian Aboriginal languages. Lutheran missionaries from the Hermannsburg Missionary Institute in present-day Lower Saxony in Germany conducted linguistic fieldwork in the Dieri (Diyari) language near Lake Eyre in South Australia from 1866. After the Hermannsburg Mission in the Northern Territory was established in 1877 Hermann Kempe (1844–1928) researched the Aranda (Arrarnta, Arrernte) language. The Hermannsburg missionaries left Central Australia in 1891. Carl Strehlow (1871–1922) arrived in 1894 after training at the Neuendettelsau Mission Institute in Franconia, Germany in which humanist philology played a greater role than it had at Hermannsburg.

David Moore. 2020. Crosscurrents in linguistic research: Humanism and positivism in Central Australia 1890–1910. In Émilie Aussant & Jean-Michel Fortis (eds.), *Historical journey in a linguistic archipelago: Descriptive concepts and case studies*, 143–158. Berlin: Language Science Press. DOI: 10.5281/zenodo.4269423

### David Moore

The missionaries were not alone. During 1907 and 1908 five descriptions of Central Australian languages were published. In tracing the history of German anthropology Kenny (2013: 51) claims that "Strehlow had little contact with his British-Australian contemporaries". Discussing the antagonism between the English biologist and anthropologist Walter Baldwin Spencer and Strehlow, Kenny does not discuss like-minded English-speaking researchers such as R.H. Mathews (Mathews 1907; Thomas 2007) and N.W. Thomas who collaborated with Strehlow. The "German *fin de siècle* anthropological tradition that was language based" (Kenny 2013: 99) was not monolithic and there is a need to take account of the discontinuities in German Ethnology which are so evident on reading the German sources. Citing Monteath (2013) and acknowledging that Antihumanists were also "well represented among the Germans", Kenny (2013: 228) also fails to acknowledge the antagonism of German Antihumanists towards Strehlow.

Previously I argued that Strehlow was engaged in philology and cultural translation rather than a form of "ethnography" which was a transitional stage to modern anthropology (Moore & Ríos Castaño 2018: 336). In dual roles of missionary and researcher, he translated for distinct purposes and audiences. His investigation of indigenous religion was limited by the attitudes of the church authorities which viewed it as "heathen" (Moore & Ríos Castaño 2018: 338). For a brief time (1906–1910) he was engaged in disinterested scholarship. He kept his research separate from his evangelical work (Brock 2017: 232). Further, "he stopped his investigations on completion of his manuscript and the death of his patron" (Brock 2017: 236), returning to Bible translation from 1913 to 1919 (Moore & Ríos Castaño 2018: 336).

In this paper I am seeking to explain the differences between Strehlow and some of his contemporaries in terms of their attitudes to linguistic research. I characterize two contrasting kinds of research as "humanist" and "positivist" according to the influences of nationality, education and training (Errington 2008: 94), arguing that metascientific orientations and language ideologies (Moore 2020) more than nationality determined their approaches to linguistic research. An understanding of these factors is necessary for interpreting their linguistic descriptions and also for understanding the collaborations between some researchers in the field and the antagonisms between others.

9 Crosscurrents in linguistic research

### **2 Humanist and positivist paradigms in the early twentieth century**

The labels "humanism" and "positivism" refer to philosophies or epistemologies or means to finding "a method by which humans could be classified and known" (Zimmerman 2001: 186), reflecting a division that existed in Germany since the *Aufklärung* ('Enlightenment'), into *Naturwissenschaften* ('natural sciences') and *Geisteswissenschaften* ('human sciences').<sup>1</sup> Humanism arose in the Renaissance as the study of the classical world (Giustiniani 1985: 172).<sup>2</sup> Although this term refers to diverse branches of scholarship (Adams 1998: 258), I focus upon the adoption of Humanism by the sixteenth-century Lutheran Reformers, the rise of the German philosophy of language and its language ideologies and the extension of philology to indigenous languages (Moore & Ríos Castaño 2018: 328; Moore 2020). Humanist scholars privileged the study of language for understanding other societies (Zimmerman 2001: 53), developing the methods of textual criticism, hermeneutics and translation to understand texts.

Early anthropologists such as James George Frazer (1854–1941) were trained in Classics as the study of the ancient civilizations of Greece and Rome. A division of labour developed as the two disciplines diverged (Marett 1908; Kluckhohn 1961). The subject material of classical studies was "civilized peoples", that is, those with a written literature while the subject material of anthropology was the "natural peoples", those without written literatures.<sup>3</sup> The division between humanism and positivism is a cline rather than rigid, reflecting the situation in German academic life in the first decade of the twentieth century. Within Germany the move from the humanist philological sciences to positivist sciences was underway about 1850 (Smith 1991: 26), reflecting wider changes in society in the "age of positivism" (Massin 1996: 120). There are strong similarities between general linguists, missionary linguists and moderate positivists. That they corresponded about the study of Australian languages is evidence of this affinity, reflecting the persistance of language ideologies from the German philosophy of language.

<sup>1</sup>The *Geisteswissenschaften* included both the humanities and social sciences, for which there was no clear division in the nineteenth century (Adams 1998: 282)

<sup>2</sup>Another term which contrasts with "positivist" is "idealist" (Vossler 1904), which applies to developments in German philosophy later than the Humanist origins in the sixteenth century.

<sup>3</sup>Anthropology is the only discipline which is allied to the humanities, social sciences and natural sciences.

### David Moore

### **3 Central Australian missionary linguistics as Humanist research**

### **3.1 Kempe and Strehlow**

Missionaries compiled grammars and wordlists for the purposes of biblical translation from German into Aboriginal languages (Moore & Ríos Castaño 2018). Initially, they sought words in Indigenous languages as translations of key theological terms in order to translate the Catechism and later, the Bible into Aboriginal languages. Kempe published a grammar and wordlist of "the language of the Macdonnell Ranges" (Kempe 1891). His treatment of the language was tentative: "the following pages, therefore, do not profess to contain *a* complete vocabulary, nor one which would satisfy the learned philologist" (Kempe 1891: 1). The Neuendettelsau curriculum was based upon philology and Lutheran theology with the purpose of enabling the mission candidates to translate and interpret biblical texts. Strehlow's training replicated the "classical orientation" (Kenny 2013: 83) in which proficiency in reading ancient languages enabled scholars to understand the biblical and classical Greek and Roman worlds. Language training included instruction in Latin, Greek, Hebrew, English and German syntax and word formation, and prepared missionaries for translating languages (Völker 2001: 8; Nobbs 2005: 26). Strehlow revised Kempe's earlier grammar, dictionary and hymnbook. Aranda became a language of interest to European scholarship and the need to obtain reliable data from the field prompted Freiherr Moritz von Leonhardi (1856–1910) to request information from Strehlow who extended his research to the collection of texts and translations of *Die Aranda- und Loritja-Stämme in Zentral-Australien* (Strehlow 1907–1920).

For German Lutheran missionaries the first step in understanding was to acquire Aboriginal languages. They learned the languages rapidly through social interaction. The Neuendettelsau-trained Lutheran missionary in North Queensland Wilhelm Poland said that "[t]here was in fact only one way of learning the language, and that was to mix with the older generation in their camps" (Poland 1988: 103). The Aranda elder and Evangelist Moses Tjalkabota was "surprised at the rapid rate of progress which Carl Strehlow made with the language" through reading Kempe's grammar and hymnbook (Latz 2014: 65–66). The Aboriginal elders Loatjira, Pmala, Tjalkabota and Talku worked with Carl Strehlow on the compilation of *Die Aranda* (Kenny 2013: 29). Talku (c.1867–1941) told him the Loritja myths for the 1908 and 1911 volumes of *Die Aranda.* Strehlow's collections of Loritja texts, grammars and dictionary were the first comprehensive record of a Western Desert language.

### 9 Crosscurrents in linguistic research

Monolingual speakers of Aboriginal languages were often not able to understand the researcher's questions. In the preface to his grammar, Kempe (1891: 1) describes the problem for linguistic research as follows:

The result of an attempt to analyse a language of which the people speaking it have only a colloquial knowledge, and who are consequently incapable of answering or even understanding grammatical questions, must be in many respects imperfect. The difficulty is increased by the wandering habits of the people.

Kempe was aware that Europeans would be told what they wanted to hear because of the gratuitous concurrence which occurs when an informant agrees with the researcher from a desire to please the questioner (Liberman 1985: 198). Kempe (1891: 1) emphasized the need to check language statements thoroughly: "Concerning the vocabulary, it may be mentioned that it has been carefully compiled and revised several times with different natives, so that the words may be relied upon as correct."

Leonhardi discussed the difficulty of eliciting information:

One should never develop his own view and then put the question, "Is this how it goes?" The question must be rather expressed, "What have your elders taught about the matter?" Then some blacks are smart enough, to find the answer. In this way one can go back and check, whether it is correct. (Leonhardi 1907: 286)

In his time at Bethesda from 1892 to 1894, Carl Strehlow, with J.G. Reuther, evaluated Dieri (Diyari) terms which would be useful for the Dieri New Testament translation and gained experience in translation, building on the work of Hermannsburg trained missionaries who had preceded Johann Flierl (Kneebone 2001). Strehlow became aware that Aboriginal languages are very different from European languages. For example, *neji* in Diyari cannot be directly translated as 'brother' (J. Strehlow 2011: 83), as Aboriginal languages have separate words for 'younger brother' and 'older brother'. He researched kinship with the ethnographic researcher Francis Gillen (Mulvaney et al. 1997). Leonhardi's questions reflected such contemporary interests of European scholars as totemism, initiation rites and kinship, views of conception and ceremonial objects or *Tjurunga* (Schmidt 1908).

It was the systematizing and generalizing by Spencer and Gillen that led Strehlow to record the particular and local to find out what Aranda and Loritja said

### David Moore

in their own words. Strehlow recorded texts in order to understand Aboriginal culture. In criticising Spencer and Gillen, Leonhardi wrote to Strehlow:

The big mistake of the books by these two researchers, it seems to me, is the fact that they systematise too much, that they try too hard to show universal views in a large area, while there should be no more than individual legends, local views and customs etc. and not a closed well-ordered system of custom. Only by providing individual stories and customs is it possible to bring out, through comparison, general aspects. (VL 1904-1-2, 28/8/04).

This "emic" approach was later identified by Kenneth Lee Pike as "studying behavior from inside the system" (Pike 1967: 37; Bolinger 1975: 542), rather than taking an external perspective. Leonhardi's *Linguistische Feststellungen* (linguistic findings) are the interlinear texts and the free translations of the myths and the songs of *Die Aranda*. Copious footnotes included translations and explanations of words which appear in the texts. The texts which were recorded only in German translation were apparently regarded as of less evidentiary value. Strehlow was also working on a grammar and comparative Aranda-German-Loritja dictionary which would help the reader to understand the *Urtext.*

The importance of the *Urtext* can be understood from the comment of Leonhardi's editorial successor Hagen (1991 [1911]: 285): "It is of some importance to know that the most important matter, the focal point so to speak, viz. the intellectual culture of the Aranda and Loritja, are in the main secured." The uncompleted sections of *Die Aranda* deal mainly with material culture. Most critical for humanist research was to record what "the Other" said in their own words.

### **3.2 Collaboration with philologists in linguistic research**

Some German scholars were interested in language classification and typology, particularly the "general linguists" of the Humboldtian school. However, contemporary comparative philology in Germany was narrowly focused upon the Indo-European languages:

von der aufblühenden historisch-vergleichenden Sprachforschung wurde die typologische Sprachwissenschaft im Sinne Humboldts ziemlich in den Hintergrund gedrängt.

the flourishing historical-comparative language research pushed typological linguistics of the Humboldtian school somewhat into the background. (Deeters 1937: 216)

### 9 Crosscurrents in linguistic research

Although Kempe (1891: 1) claimed that his grammar and wordlist were submitted "in the hope that they would prove interesting to the philologist", he doesn't appear to have consulted with philologists. Contact between Strehlow and a general linguist was facilitated by Leonhardi, who could see the benefit of making Strehlow's research available to European scholars. Franz Nikolaus Finck, a professor of linguistics at the University of Berlin, provided comments on Carl Strehlow's texts which Leonhardi sent him. In a letter to Strehlow Leonhardi writes:

I would like you to know that Prof. Finck in Berlin, to whom I had sent the 'Aranda Legends' has in the last few days expressed high praise for the Aranda texts in a letter to me. Since Prof. Finck is a first-rate authority on Austr. Oceanic languages and I had sent him your essay as well, as you know, I am very pleased about this recognition. (VS 1908-1-1)<sup>4</sup>

### **4 Positivism**

Positivists based their research on the natural sciences. Their favorable valuation of the natural sciences followed contemporary trends and, by the late nineteenth century, led to a "sense that scientific discourse was more correct than others" (Crick 1976: 154). For some positivists language was typically one category of human behavior among many behaviors that could be described. The view that visual observation provided the only reliable evidence about the "Other" meant that "fieldwork" became an essential practice within anthropology and began to replace "armchair" scholarship around the turn of the century. Prominent in the development of fieldwork was Adolf Bastian (1826–1905), co-founder of the *Gesellschaft für Anthropologie, Ethnologie und Urgeschichte* (Society for Anthropology, Ethnology and Prehistory) with Rudolf Virchow (1821–1902) and the first director of the Museum für Völkerkunde in Berlin in 1873 (Köpping 1983). *Völkerkunde* "classified and generalized the results of a strictly descriptive ethnography" (Buchheit & Köpping 2001: 19–25, cited in Gingrich 2005: 87).<sup>5</sup> Bastian typified the positivist view, attacking interpretation, history and literature as unreliable ways of understanding the "Other" (Zimmerman 2001: 61). However, he retained a respect for the philological tradition (Gingrich 2005: 89). After his death in 1905, he was succeeded by his younger associates, whom Gingrich (2005:

<sup>4</sup>The correspondence from Moritz von Leonhardi to Carl Strehlow is held in the archives of the Strehlow Research Centre, Alice Springs.

<sup>5</sup>*Völkerkunde* can be translated as "cultural anthropology". *Anthropologie* is translated as "physical anthropology" (Massin 1996: 82) and is therefore a false friend with current English "anthropology".

### David Moore

91) characterizes as "moderate positivists". Positivists were not necessarily evolutionists and Bastian and others were opposed to evolutionism. In the following sections, the "moderate positivists" (Gingrich 2005: 99) are contrasted with radical positivists or Antihumanists (Zimmerman 2001; Monteath 2013).

### **4.1 Moderate positivists**

The moderate positivists were overshadowed by the diffusionists in the first decade of the twentieth century in Germany but "remained as systematic fieldworkers and museum documentarists" (Gingrich 2005: 92), closer to the international mainstream of anthropology and particularly close to the German-influenced linguistic anthropology that was emerging in the USA. Among the moderate positivists were Konrad Theodor Preuss (1869–1938) and Karl von den Steinen (1855–1929). Preuss (1908; 1909) reviewed Carl Strehlow's *Die Aranda* positively.

### **4.2 Antihumanists**

Antihumanists followed social evolutionary anthropology, which was the dominant paradigm in British anthropology by the turn of the twentieth century. Walter Baldwin Spencer (1860–1929) and Francis J. Gillen (1855–1912) may be characterized as Antihumanist. As Spencer admitted, "my anthropological reading was practically confined to Sir Edward Tylor's 'Culture' and Sir James Frazer's 'Totemism' " (Spencer 1928: 184). Spencer and Gillen followed Frazer's lead and their monographs clearly show the influence of Frazer's *Golden Bough* and the list of priorities for the collection of significant ethnographic "facts" outlined in his short questionnaire (Urry 1993: 45). Frazer separated particular facts from their cultural contexts and arranged them within a continuous discourse of evolutionary development. When Virchow's "inductive positivism" (Massin 1996: 138) was rejected, social evolutionism became more influential in Germany. I examine German Antihumanist researchers in the following sections.

### **4.2.1 Basedow**

Herbert Basedow (1881–1933) was a medical practitioner whose family had migrated from Berlin to South Australia in the 1850s. He trained in Breslau under Hermann Klaatsch (1863–1916), an anatomist and physical anthropologist who founded an institute of physical anthropology in Breslau in 1907 (Massin 1996: 84) and invited Basedow to study there in the same year (Basedow 1925: ix). Klaatsch

### 9 Crosscurrents in linguistic research

was one of the first German physical anthropologists to adopt social evolutionary theory. In 1904 he travelled to Australia, visiting Melville Island, Tasmania and northwestern Australia (Oetteking 1916: 423). He claimed that the Aborigines were "a relic of the oldest types of mankind" (McGregor 1997: 42), based upon the anatomical comparison of Aboriginal people with the Neanderthals and other earlier humans. Australian languages were also primitive: "The Australian dialects seem in many respects to be fragments of the primitive speech of man" (Klaatsch 1923: 38). Adopting the evolutionary view of his mentor, Basedow (1908: 208) compared sounds made by speakers of Aboriginal languages with those made by apes:

Es ist von Interesse, dass Garner in seinem bekannten Werk über die Affensprache gefunden zu haben angibt, dass die von ihm beobachteten Affen denselben Laut "ng" besitzen und zwar im Zusammenhang mit dem Ausdruck der Zufriedenheit "ngkw-a".

It is of interest that Garner in his well known work on ape language, found that the apes observed by him use the same sound 'ng' in the context of an expression of satisfaction 'ngkw-a'. (Basedow 1908: 208)

Basedow was a member of the South Australian Government North-West Prospecting Expedition, led by L.A. Wells. He expected his officers to learn Aboriginal languages and "to treat the natives in a friendly and considerate, yet firm and masterly way" (Zogbaum 2010: 49). Basedow collected "a vocabulary of about 1500 words of the Aluridja (Western Desert) and Aranda languages" (Harmstorf 2004 [1935]: v). He admitted that he did not consult other sources and that his "article on language is not intended to be at all comprehensive" (Basedow 1925: xii). The short-term nature of Basedow's trips was useful for compiling wordlists but not for learning to speak languages fluently. His wordlist (Basedow 1908) is rich in names for physical objects but not mental and religious aspects of culture.

Although his evolutionary views are in strong contrast with Strehlow's, Basedow appears to have been sympathetic to the Lutheran missionaries and appreciative of their linguistic research (Harmstorf 2004 [1935]: vi). He was a Lutheran and had strong connections to the Barossa Valley and South Australian Lutherans who supported the Hermannsburg Mission. He visited the mission station in 1919 and later wrote that with Strehlow's death "Science has lost an indefatigable and conscientious worker" (Basedow 1925: ix).

### David Moore

### **4.2.2 Eylmann**

Erhard Eylmann (1860–1926) included two chapters about language in his study of Aboriginal people in Australia (Eylmann 1908). Monteath (2013: 34) characterizes him as an "Antihumanist" as he focused upon material culture rather than pursuing a humanist interest in language. He admitted the limitations of his understanding of Aboriginal languages:

Über den Bau der Sprachen vermag ich keine nennenswerte Auskunft zu geben. Ich habe mich in Südaustralien nirgends solange aufgehalten, daß ich nach Erledigung der notwendigsten Arbeiten noch Sprachstudien treiben konnte.

Concerning the structure of the language, I cannot provide any great amount of information. I have not stayed anywhere in South Australia for a length of time that would have permitted me to pursue language studies after I had completed the work of the highest priority. (Eylmann 1908: 81)

Eylmann (1908: 81) admitted that he had difficulty eliciting a word which was equivalent to English 'and' from a speaker of the Awarai language. He became tired in a "surprisingly short time".

### **5 Positivism in linguistic research: Planert**

August Schleicher (1983 [1850]) first suggested that linguistics was a natural science, casting linguistics in terms of biological metaphors and creating a disciplinary matrix for a linguistics founded upon the natural sciences (see also McElvenny 2018). Linguistics increasingly came under the influence of positivism in the late nineteenth century.

Wilhelm Planert (b. 1882) claimed to be "scientific". In his inaugural dissertation at the University of Leipzig, Planert (1907b) claimed: "In this treatise, for the first time, an attempt is made to correspond to the intentions of modern linguistics". Planert was a student of Carl Meinhof (1857–1944), professor at the School of Oriental Studies in Berlin from 1905. Meinhof was involved in developing the Language Institutes (Seminars) as "Hypermetropolitan spaces", laboratories where phonetic and linguistic information could be easily extracted from informants (Pugach 2012: 138). Languages were recorded with phonographs, played back and "observed". As Pugach (2012: 93) points out, "the new discipline of phonetics recast linguistics as a natural science, distancing it from humanistic philology by

### 9 Crosscurrents in linguistic research

refocusing attention on bodies and the sounds they produced instead of written texts".

Planert's usual method of operation at the Oriental Institutes was to interview language speakers who were visitors to Germany. He was limited to working in the metropole and the laboratory away from the context of language use. Planert (1908) acknowledged in a response to Carl Strehlow's criticisms of his Aranda Grammar that errors were made because of a lack of reliable informants and he was disparaging about the training of Missionary Nicol Wettengel who was his informant for the Aranda Grammar (Planert 1907a). Wettengel had worked at Hermannsburg in the Northern Territory of Australia from 1901 to 1906 (Strehlow 2011: 1154) and gained some familiarity with the Aranda language. Planert worked with an informant who was not a native speaker of Aranda and who had a less than adequate grasp of the language.

Languages were manipulated to serve colonial goals (Errington 2008: 88). The nation required a "school-mediated, academy-supervised idiom codified for the requirements of reasonably precise bureaucratic and technological communication" (Gellner 1983: 57). Planert's PhD dissertation was published as *Syntactic relationships in Swahili* (Planert 1907b). Swahili was a language of administration which the German colonists elaborated into a language of "civilization and progress" to rule East Africa. Planert also described "Bushman" and "Hottentot" (1905b), Nama (1905a) and Jaunde (Nekes 1911).<sup>6</sup> While the Germans were conducting a genocidal war against the Nama and Herero peoples in South West Africa (Hull 2005), Planert was writing his grammar of the Nama language. He collaborated with the colonial authorities in language engineering and control, in contrast with the moderate positivists who had very little to do with colonialism or were even opposed to it.

### **6 Conclusions**

Researchers from a wide variety of metascientific orientations attempted to understand the "Other" through their languages.

Kenny's assertion that the "humanism of German anthropology with its pluralistic outlook and anti-evolutionistic position lasted nearly to the eve of World War I" fails to explain Klaatsch's evolutionary anthropology. The views of Basedow, Eylmann and Planert reveal the degree to which Antihumanism had, in fact, become established and dominant in German Ethnology. There was more

<sup>6</sup>A language used in the German colony of Kamerun (Cameroon) and now written as "Ewondo".

in common between researchers of different nationalities who shared a similar orientation than those of the same nationality who had different philosophical orientations. It would be most accurate to say that Strehlow had little sympathy for those contemporaries who saw research in a very different way, anthropologists whose primary metascientific orientation was Antihumanist, including German Antihumanists.

The critical difference between the missionaries and the Antihumanists in the Central Australian field was that the missionaries could understand the "Other" through the strong focus upon language of their humanist training. Antihumanist interpretations were often hampered by literalism and misunderstandings. Although claiming to be "scientific" and objective, they were often biased through their support for pre-existing theories and their affiliations to colonial forces. Significantly, missionary research filled in gaps in the knowledge of Central Australian languages at a time when neither anthropology nor comparative philology took an interest in the languages of Australia. Further research on these rich sources is needed to understand missionary research, language ideologies and experiences of fieldwork.

### **References**


Basedow, Herbert. 1925. *The Australian Aboriginal*. Adelaide: F.W. Preece & Sons.


*Sammeln, Vernetzen, Auswerten: Missionare und ihr Beitrag zum Wandel europäischer Weltsicht*, 173–218. Tübingen: Gunter Narr Verlag.


## **Chapter 10**

## **La** *Revue des Patois Gallo-Romans* **(1887–1892) et la représentation de l'oral**

Gabriel Bergounioux Université d'Orléans/LLL – UMR 7270

> The *Revue des Patois Gallo-Romans* appeared five years after the Jules Ferry laws, which regulated the teaching of French and established a modern schooling system, were adopted by the French Republic. The *Revue* was soon confronted with issues raised by the description of oral varieties in a country shaped by a written culture and actively enforcing linguistic standardisation. The problems scholars had to face involved the choice of more or less fine-grained phonetic transcription, determining the competence required by investigators, and the choice of reliable speakers. A survey of the contributions to the *Revue* shows how dialectologists coped with the exigencies of the standardisation program and the new importance laid on dialects by comparative linguistics.

### **1 Introduction**

En France, les revues ont constitué, de 1859 à 1970 environ, un observatoire privilégié pour comprendre la façon dont s'est effectué le déploiement scientifique de la linguistique (Bergounioux 1984 ; Chevalier & Encrevé 2006). Dans la deuxième moitié du XIX<sup>e</sup> siècle, elles ont constitué une préfiguration (Bergounioux 1997) ou une alternative, souvent conflictuelle, à l'organisation de l'enseignement supérieur dans ce domaine (Desmet 1996, Bergounioux 2002).

La *Revue des Patois Gallo-Romans* (désormais *RPGR*) constitue cependant un cas particulier car la dialectologie, dans son état contemporain, ne s'était pas implantée dans les universités. Les cours ouverts dans les facultés des lettres de province étaient consacrés aux états anciens (Bergounioux 1984). La seule re-

Gabriel Bergounioux. 2020. La *Revue des Patois Gallo-Romans* (1887–1892) et la représentation de l'oral. In Émilie Aussant & Jean-Michel Fortis (éd.), *Historical journey in a linguistic archipelago : Descriptive concepts and case studies*, 159-173. Berlin : Language Science Press. DOI : 10.5281/zenodo.4269425

### Gabriel Bergounioux

connaissance institutionnelle, au sein de la IV<sup>e</sup> section de l'École Pratique des Hautes Études, était la direction d'études confiée à Jules Gilliéron en 1887. Après avoir rappelé le contexte politique et l'historique des travaux sur les « patois », on considérera la *RPGR* en tant qu'elle a fourni des éléments de réponse à une question centrale pour la recherche : quelle doit être la représentation de l'oral et comment doit être comprise la solution proposée par celui qui en a conçu le système de transcription, l'abbé P.-J. Rousselot ?

### **2 Une situation particulière**

Si l'on observe la situation il y a cent cinquante ans, la France, dans les frontières qui étaient les siennes et qui n'ont guère changé, présente un cas de figure unique en Europe. Pas de revendication d'indépendance de ses territoires comme, à la même époque, en Norvège (par rapport à la Suède) ou en Irlande (face au Royaume-Uni) ; pas d'objectif de rattachement des francophones de Belgique ou de Suisse alors que l'Allemagne, l'Italie ou la Grèce affirmaient leur volonté de fédérer en une seule entité toutes les contrées où une majorité de la population se référait, pour la représentation officielle de son parler, à la même langue écrite. Comme en Autriche-Hongrie, dans l'Empire Ottoman et en Russie, le pays comptait un très grand nombre de locuteurs dont la langue maternelle n'était pas celle qui avait le monopole de l'enseignement et de l'administration mais, à la différence de ces trois états, la France est demeurée unitaire et monolingue.

Une explication s'impose : à défaut d'être francophone, la France a été francisée. Le processus a duré des siècles (Brunot 1905-1937 ; Boyer & Gardy 2001; Kremnitz & Broudic 2013) sans soulever de résistances significatives fondées sur l'identité linguistique. Les lois scolaires de Jules Ferry en 1881–1882 ont précipité un processus qui avait commencé dès le Moyen-Age dans un contexte que la guerre de 1870 avait définitivement transformé.

Opérant un renversement symbolique, le romantisme avait conféré une dignité aux « arts et traditions populaires ». La reconnaissance par les sciences historiques de strates de peuplement et de langues littéraires différenciées supplantait la condescendance pour des parlers frustes et désuets. Ceux-ci représentaient l'expression authentique des populations rurales, le réceptacle des traditions, du *folklore* (un mot emprunté à l'anglais en 1885). Témoignages de la permanence d'une population d'origine celtique sur le même territoire, ils appartenaient de plein droit au patrimoine de la nation. Pourtant, ils restaient associés à l'obscurantisme et apparaissaient comme de possibles ferments de fédéralisme voire de séparatisme après la signature du traité de Francfort le 10 mai 1871.

### 10 La *Revue des Patois Gallo-Romans* (1887–1892) et la représentation de l'oral

La réponse des romanistes parisiens passait par la confusion de l'ensemble des parlers métropolitains issus du latin en une seule et même langue dont la forme littéraire proviendrait de la transposition écrite du « francien » (Auroux et al. 1996). Les variétés écartées de l'usage officiel se trouvaient par suite ravalées au rang de « patois » – un terme en concurrence avec *jargon* et autres vocables dépréciatifs –, une dénomination étendue au breton, au basque et au flamand, assimilés par leur statut aux dialectes romans dont ils partageaient le décri.

### **3 La dialectologie : rétrospective et méthodologie**

Dans le prolongement d'un engouement plus empathique que savant qui recoupait le travail entrepris à l'École des Chartes (créée en 1821) à partir des archives et des manuscrits, les premières études avaient abouti à la confection de dictionnaires comme le *Lexique roman* 1838-1844 de F. Raynouard et le *Glossaire* de H. Jaubert (1856-1858) qui préfiguraient un intérêt renouvelé pour les langues régionales et les formes vernaculaires. Après l'inventaire des sources scripturales sous l'autorité de la commission des travaux historiques de l'Académie des Inscriptions et Belles-Lettres en 1834, les missions de terrain du ministère de l'instruction publique avaient assuré la collecte de données ethnographiques (chansons, proverbes…) dans les années 1850. Parallèlement, en Suisse et en Belgique, des monographies affirmaient la spécificité de la Wallonie et du pays romand. Les enquêtes d'O. Bringuier et de Ch. de Tourtoulon (1876) et la contre-enquête d'A. Thomas (1879) ont relancé la discussion sur les frontières linguistiques internes de la France romane (Brun-Trigaud 1990).

La dialectologie s'est constituée méthodologiquement au point de rencontre entre trois techniques : l'enquête de terrain, la projection cartographique et la transcription. La constitution des données au moyen d'enquêtes est empruntée aux sciences sociales et, du fait des populations étudiées, la recherche s'est avérée plus proche de l'ethnologie que de la sociologie, une orientation déterminée par le peu de sources écrites disponibles sur le monde rural. La cartographie, reprise à la géographie (Gilliéron présentera une partie de ses travaux sous l'intitulé de « géographie linguistique »), accompagne le développement de cette discipline et figure en synchronie des variations que la grammaire comparée traitait dans leur profondeur diachronique.

La représentation des matériaux sonores pose la question de la notation des langues. Il y a un lien entre l'étude sur le terrain et la phonétique de laboratoire, deux disciplines antinomiques par leurs pratiques mais qui partagent une même défiance à l'encontre de la philologie, de la lettre et des textes, même si le proGabriel Bergounioux

gramme de la *RPGR* aboutissait à restituer des usages parlés sous la forme de documents écrits.

### **4 Les auteurs de la RPGR**

La *RPGR* paraît la même année que la *Revue des Patois* de Léon Clédat. Elle a pour directeurs Rousselot et Gilliéron. Elle comporte cinq tomes, publiés de 1887 à 1892 (pas de parution en 1889) et totalise 1666 pages. Dans le titre, l'idée d'un substrat gaulois – « gallo-roman » –, est secondaire, seule comptant la réunion en un seul ensemble linguistique des patois « (…) appartenant comme elle [notre langue littéraire] au latin vulgaire qui est parlé dans les Gaules depuis la conquête romaine » (Rousselot 1887 : 1).

Pierre-Jean Rousselot est prêtre. Il a commencé par enseigner la phonétique à l'École des Carmes (l'Institut Catholique de Paris) où il a installé un laboratoire transféré au Collège de France en 1897 auprès de Michel Bréal. « Préparateur » trente années durant, il est élu à soixante-dix-sept ans dans une chaire de phonétique. Il a soutenu son doctorat en 1891 et l'édition de sa thèse occupe une part importante des livraisons de la *RPGR*.

Jules Gilliéron a soutenu son doctorat en 1880 sur le *Patois de la commune de Vionnaz (Bas-Valais)* complété l'année suivante par un atlas (Gilliéron 1881). Il est nommé chargé de conférences à l'EPHE en 1883 et, trente ans après sa naturalisation prononcée en 1886, directeur d'études en 1916. Ses contributions à la *RPGR*, une soixantaine de pages, sont consacrées pour l'essentiel à l'exploitation d'informations tirées d'enquêtes de terrain menées en Savoie.

Au Collège de France, les soutiennent Henri d'Arbois de Jubainville, premier titulaire de la chaire de « langue et littérature celtiques » qui confie à la revue son étude sur « La langue latine en Gaule » (1887 : 161–171) et Gaston Paris à qui la *RPGR* est dédicacée : « A M. Gaston Paris, hommage respectueux et reconnaissant de ses élèves (Gilliéron et Rousselot) ». Paris apparaît une seule fois au sommaire, à l'occasion de son allocution solennelle devant le congrès des sociétés savantes en 1888 sur « Les parlers de France » (1888 : 161–175). Sa présence ne fait que mieux ressortir l'absence de Paul Meyer, le titulaire de la chaire de « Langues et littératures de l'Europe méridionale » (comprenant l'occitan) de 1876 à 1906.

Interviennent aussi sur des questions de portée générale Jean Psichari (à propos de la dialectalisation du grec moderne), Louis Gauchat (pour une appréciation critique du livre de D. Schindler sur le vocalisme du patois de Sornetan) et Eduard Koschwitz (sur l'interprétation phonétique de graphies médiévales), trois personnalités étrangères (ou d'origine étrangère) pour des questions qui se posent en dehors des frontières nationales et de l'état actuel des langues.

### 10 La *Revue des Patois Gallo-Romans* (1887–1892) et la représentation de l'oral

Au nombre des enquêteurs, le premier, Edmond Edmont, est le futur collaborateur de Gilliéron pour l'*Atlas Linguistique de la France*. Son *Lexique saint-polois* (1887 ; 1897) est publié en fascicules qui représentent près d'un quart de la collection complète de la revue (385 p.). A côté de cette monumentale étude sur une localité du Pas-de-Calais, on relève les interventions de :


Seul Jean Passy traite d'un parler qui n'est pas celui de la région où il a grandi. D'autres participants ont rédigé entre une et dix pages, tels G. Dottin, A. Jeanroy, P. Lejay, F. Nougaret, P. Passy et une douzaine d'abbés. Les autres auteurs étrangers sont, pour la Belgique, A. Horning et M. Wilmotte et, en Suisse, H. Morf (sur les Grisons).

### **5 Terrains et objets**

Cent huit contributions correspondent à des études territoriales sur l'aire galloromane. La Wallonie a fourni une soixantaine de pages (dix articles de Wilmotte, Marchot, Doutrepont) et la Suisse romande, la patrie de Gilliéron, seulement trois pages rédigées par l'enfant du pays. L'aire de la langue d'oc correspond à une trentaine de départements dont seuls une douzaine sont représentés. Aucun ne se trouve à l'est du Rhône, le berceau des provençalistes. En comparaison, une trentaine de départements d'oïl ou franco-provençaux sur une cinquantaine sont mentionnés avec des témoins en grande partie recrutés par Rousselot parmi ses élèves. L'image de la France est anamorphosée. On ne saurait considérer que le Boulonnais (étude d'Edmont sur Saint-Pol) et le Confolentais comptent pour la moitié du territoire parce qu'ils occupent la moitié de la pagination.

Sur les vingt-et-un articles consacrés aux pays occitans, quatorze sont des éditions de textes, des transcriptions (chansons, contes…) accompagnées d'un commentaire strictement philologique et deux des études phonétiques (Nougaret et Jean Passy). A part Rousselot, seul Camélat a entrepris la description systématique d'un parler dans l'Hérault. Le peu de réalisations de monographies locales exhaustives recoupe l'absence d'études sur des isoglosses en dépit d'un intérêt

### Gabriel Bergounioux

affirmé pour les frontières du traitement des palatales ou du /a/. Le postulat d'une variation progressive, sans véritable frontière, induisait une attention concentrée sur le niveau communal.

En terre d'oïl, trois auteurs ont entrepris de rédiger un lexique ou des études étymologiques (Devaux, Edmont, Marchot) et deux ont conçu la description complète d'un parler : Fourgeaud en Charente et Rabiet en Côte d'Or. En tout, 25 articles font une part à la phonétique.

### **6 Phonétique et dialectologie : la question de l'alphabet**

Le lien consubstantiel de la dialectologie à la phonétique s'explique par le caractère oral de ces parlers et la subtilité des réalisations qui les différencient. À l'opposé de l'étude du français (ou des grandes langues littéraires), les analyses se centrent sur l'enquête, le recueil des paroles. Les textes qui ne sont pas transcrits en alphabet phonétique sont donnés comme des figurations approchées dont l'intérêt premier est d'offrir une représentation des réalisations sonores. Très peu de documents avaient été consignés à l'écrit avant d'être reproduits dans la *RPGR*, surtout des chants traditionnels, moins encore avaient été publiés (un article de journal en patois). Sur 48 contributeurs, 17 ont consacré tout ou partie de leur production à des questions phonétiques mais aucun n'a redoublé sa description par l'archivage d'un enregistrement. Il faudra attendre l'initiative des Archives de la parole de Ferdinand Brunot en 1911 (Cordereix 2001). A côté d'études à portée générale (Rousselot, Psichari, Koschwitz), on relève une analyse sur la phonotaxe de Nougaret et deux sur la morphophonologie par Marchot et par d'Arbois de Jubainville. Les autres travaux concernent :


Dans tous les cas, une même question se posait concernant la façon dont doivent être restituées les productions orales dans une notation scripturale dérivée des caractères latins. D'un côté, il y avait les limites imposées par les polices et les fontes des imprimeurs. De l'autre, des solutions avaient déjà été mises

### 10 La *Revue des Patois Gallo-Romans* (1887–1892) et la représentation de l'oral

en pratique dans l'adaptation aux différentes langues romanes (digraphes, diacritiques, ponctuation…). Les romanistes ont divergé dans la correspondance entre ressources typographiques et variations phonétiques, les deux principaux alphabets en concurrence à la fin du XIX<sup>e</sup> siècle étant celui de Böhmer (1871) révisé par Ascoli (1873) et celui de Rousselot.

La principale innovation de Rousselot pour décider d'une écriture en 1887 est de partir du signal et non des notations écrites afin de choisir celles qui seraient le mieux ajustées aux productions sonores. Il n'a fondé son système ni sur les oppositions, ni sur les perceptions mais sur les qualités articulatoires que ses instruments lui permettaient d'observer. La description ne commence pas par la séparation en consonnes et voyelles mais par celle des « résonnances » (pharyngale ou nasale) et des « sons ». L'approche est fondée sur des unités que définit leur contenu et non leur fonction. Le « Système graphique » est présenté juste après l'introduction générale, dès la page 3 du tome I. Il n'y a pas de tableau récapitulatif et la description des signes préconisés est exposée de façon strictement linéaire.

Après les résonances (consignées sous forme d'indices ou de suscriptions) sont énumérés les « sons fondamentaux » subdivisés en « consonnes », « résonnantes » (= sonantes), semi-voyelles et voyelles. La liste s'établit à trente unités, soit toutes celles présentes en français auxquelles sont ajoutées la glottale /h/, les deux fricatives dentales /θ/ et /ð/, une dorsale palatale /ç/, vélaire /γ/ et uvulaire /χ/, une nasale /ŋ/, une lambdaïque /λ/ et quatre rhotiques notées [R ɾ̪ r̥ ɾ]. Afin d'augmenter le nombre de variantes sont prévus des diacritiques correspondant à la mouillure, à la fricativisation, à une réalisation gutturale et aux deux positions *Advanced Tongue Root* ATR et *Retracted Tongue Root* RTR. Pour les sonantes, le passage d'une valeur consonantique à une valeur syllabique est indiqué par un point souscrit. A chacune des trois semi-voyelles du français est affecté un symbole distinct, comme dans l'API.

Concernant les voyelles, Rousselot part de sept timbres qui correspondent à /a/ /e/ /i/ /o/ /u/ /œ/ /y/ auxquels est ajouté le schwa /ə/. Les timbres sont dédoublés par l'indication de leur quantité (un accent circonflexe à l'envers ou à l'endroit), leur degré d'aperture (accent grave et accent aigu) et la nasalisation (tilde) ou la demi-nasalisation (un tilde aplati). Comme les degrés d'aperture peuvent excéder les quatre pertinents en français, une même voyelle moyenne peut recevoir un indice qui permet d'en démultiplier les réalisations. De la combinaison des diacritiques résultent d'inévitables difficultés d'impression qui ne seront pas résolues.

Les « sons intermédiaires » sont marqués par la suscription de l'un des deux. Un son entre sourde et sonore, laissant dans l'indécision le choix d'un /t/ ou d'un

### Gabriel Bergounioux

/d/ par exemple, sera écrit soit avec un t surmonté d'un d plus petit, soit l'inverse, sans qu'à aucun moment il ne soit fait de distinction entre ces deux notations. Le système est compliqué par la taille des caractères : une gradation est établie en notant les sons en cours d'apparition ou de disparition par une police de moindre taille.

Rousselot conclut sa présentation par le rappel de quelques principes :

Tel est l'ensemble du système graphique que nous proposons. On voit :

1° Que nous empruntons à l'alphabet et aux usages typographiques français la plupart de nos signes.

2° Que nous conservons à ces signes la valeur qu'ils ont en français, et que nous modifions la forme de ceux dont nous sommes obligés de modifier la valeur.

3° que chaque signe a toujours la même valeur et que chaque son est toujours représenté par le même signe. (…)

4° que chaque son est figuré par un seul caractère ; *ch* est devenu pour nous *c*; *ou*, *u* ; *eu*, *œ* ; *gu*, *g* ; *ss*, *s*, etc.

5° enfin que nous n'employons aucun signe qui ne serve à figurer la prononciation. Nous ne faisons donc usage ni de l'apostrophe, ni du trait d'union. (Rousselot 1887 : 6–7)

Suivent dix pages consacrées aux difficultés d'application, en particulier pour la reconnaissance de certaines distinctions (mouillure, /h/ buccal ou pharyngal, l'opposition /c/ *vs* /k/, les diverses réalisations des rhotiques, les affriquées…). Entre /a/ et /i/, Rousselot découpe huit degrés en comprenant les extrêmes, sept entre /a/ et /u/. Comme à aucun moment il n'a récapitulé la liste de tous les sons que son système permettrait de représenter, on peut en livrer une approximation en intégrant l'ensemble des traits retenus pour distinguer les voyelles entre elles :


### 10 La *Revue des Patois Gallo-Romans* (1887–1892) et la représentation de l'oral

• 36 x 2 (nasalisation) + 36 x 2 (semi-nasalisation) = 144 possibilités

Enfin, une voyelle peut toujours être accentuée (rien n'est dit de l'accent secondaire), soit

• 144 x 2 (accentué vs non accentué) = 288 combinaisons possibles

à quoi il convient d'ajouter le schwa et, avec la taille des polices, l'indice d'un amuïssement ou d'une épenthèse en cours. La puissance générative des traits dépasse l'inventaire de n'importe quelle langue. Comme l'analyse part de la phonétique articulatoire et que toutes ces voyelles sont effectivement réalisables, Rousselot s'arrête à ce constat. Au contraire, confronté aux données de terrain, Gilliéron formule des réserves :

La voyelle finale qui est notée *a* sans distinction de timbre ni de quantité est un son qui, dans un seul et même patois, peut avoir une existence de la même plénitude qu'une voyelle ordinaire non accentuée, ou être un de ces sons que nous rendons par des caractères plus petits, ou même totalement disparaître dans certaines conditions de son existence. Il en est de même de *è* et de *é*. Il ne faut rien conclure de l'absence de ce son dans certains de nos patois. (Gilliéron 1888 : 33–34n)

Les limites d'une approche instrumentale sont patentes dans l'absence de notation spécifique pour les diphtongues, si fréquentes dans les parlers d'oïl et d'oc. Identifiées par leur timbre, elles ne sont pas considérées comme des phonèmes spécifiques. D'autres difficultés affleurent constamment. E. Rabiet fait part de ses scrupules dans l'établissement d'une graphie phonétique normalisée du patois de Bourberain en Bourgogne (1887 : 243) et F. Nougaret, étudiant le parler de Bédarieux, hésite à attribuer aux segments des caractères qui sont décidés *in fine* par la phonotaxe (1890 : 216).

C'est par sa dimension critique que la contribution de Rousselot a marqué son époque. A défaut d'élaborer une méthode qui, au-delà de la description du matériel sonore, établirait les principes définissant la liste des symboles pertinents pour la transcription d'une langue donnée, il a mis en évidence en quoi la tradition scripturale déformait la représentation des langues, qu'il s'agisse de l'orthographe félibréenne (Rousselot 1887 : II-158), de l'orthographe du français (Rousselot 1887 : III-239) et même de la transcription en API proposée quelques années auparavant par Paul Passy (Rousselot 1887 : III-238).

### Gabriel Bergounioux

### **7 Une tâche impossible ?**

La *RPGR* a construit son projet suivant une démarche contradictoire : scientifiquement, les parlers « gallo-romans » sont appréhendés comme des langues dignes d'être étudiées pour elles-mêmes, rompant avec le préjugé d'idiomes barbares ou d'un français déformé. Conçus comme la transmission populaire du latin parlé en Gaule sur un territoire rural donné, pour reprendre la définition de l'époque, ils ne le cèdent en rien par leur antiquité et leur authenticité au français malgré la disparité des destins. Politiquement, à la différence de la *Revue des Langues Romanes*, proche des Félibres, ou des travaux conduits en Allemagne et en Italie, la *RPGR* ne s'assignait pas pour fin de relever le statut des « patois » dont le refoulement était entériné par la scolarisation obligatoire mise en place quelques années auparavant.

Le compromis entre les principes du monde savant et les exigences de l'Etat s'est résolu par une atomisation des parlers. Au lieu d'une forme écrite de reconnaissance transdialectale qui unifierait de vastes ensembles, les directeurs sollicitaient des enquêtes de terrain restreintes à l'échelle d'une commune, voire en deçà (Rousselot à Cellefrouin, Edmont à Saint-Pol). La moindre nuance phonétique relevée d'une localité à l'autre était consignée, requérant un alphabet phonétique d'autant plus complexe qu'il ne restitue pas des différences à l'intérieur d'un système mais des variations entre villages, voire entre locuteurs.

La méthodologie de la revue excluait les contributions directes des patoisants, sinon à titre de témoins. Les auteurs ne pouvaient être que des savants informés des avancées de la dialectologie ou des lettrés qui, par leur trajectoire, maîtrisaient le patois comme langue maternelle et le français (et le latin) par une fréquentation prolongée du système scolaire. A leurs côtés, des amateurs auraient pu proposer à la revue des productions littéraires régionales ou des études de folklore. Ceux-ci étaient plus attirés par l'exposé des mœurs et des objets que par la description phonétique ou grammaticale – à l'exception notable d'E. Edmont. Quant aux écrivains, ils étaient d'avance rebutés par Rousselot qui stipulait dans une « chronique » :

Des correspondants nous proposent des compositions littéraires en patois. (…) Mais ces compositions ne sont à leur place dans notre *Revue* que si elles réalisent les deux conditions suivantes : 1° Représenter exactement le patois d'un lieu déterminé (…). 2° Etre transcrits suivant le système graphique de la *Revue*. (Gilliéron & Rousselot 1887-1892 : III-159)

Les enfants issus des campagnes dont l'ascension sociale résultait d'une prise en charge des frais de scolarité par leur futur employeur étaient soit des prêtres,

### 10 La *Revue des Patois Gallo-Romans* (1887–1892) et la représentation de l'oral

soit des instituteurs. La mission de l'école laïque – substituer le français aux patois dans les départements allophones – s'avérait peu compatible avec un intérêt pour la description des parlers régionaux. Quant aux prêtres, sollicités directement par Rousselot à l'École des Carmes, sauf Fourgeaud et Rabiet, ils ne sont guère allés au-delà de leur collaboration à l'établissement d'une transcription.

### **8 Rousselot et Gilliéron : deux approches discordantes**

Si, en couverture, les directeurs de la *RPGR* sont présentés par ordre alphabétique, Rousselot signe seul l'introduction programmatique de la revue. L'implication rédactionnelle des deux directeurs a évolué au fil des livraisons (voir Tableau 10.1).

Tableau 10.1 : Nombre de pages rédigées par les directeurs de la RPGR (entre parenthèses les documents établis avec les prêtres de l'École des Carmes)


E. Edmont est le seul collaborateur attitré de Gilliéron à qui a été confié un nombre significatif de cahiers pour qu'il puisse publier une partie de son lexique. A l'inverse, l'entourage de Rousselot bruit de nombreuses soutanes. L'accroissement spectaculaire des interventions de Rousselot dans les tomes IV et V, alors que Gilliéron s'est retiré, s'explique par l'impression de sa thèse. Le tome V édité pour solde de tout compte est partagé entre une liste des « mots français usités en saint-polois » (p. 7 à 144) et la deuxième partie de la thèse de Rousselot : « Modifications historiques de l'ancien fond du patois » qui, afin de préparer la publication en volume sans retouche des placards, fait sauter la numérotation de la page 144 à la page 208.

Gilliéron s'est orienté vers la géographie linguistique qui l'a conduit à développer une étude fondée sur l'onomasiologie (1905, 1918) tandis que Rousselot, abandonnant la dialectologie, s'est consacré de façon exclusive à la phonétique

### Gabriel Bergounioux

instrumentale (1897-1908). En dépit de leurs divergences, ils partageaient un point commun : une indifférence à une considération anthropologique qui a contribué à creuser le fossé entre l'école française et les écoles allemande, américaine ou russe. Le Tableau (10.2) relève les points de divergence entre les deux directeurs.


Tableau 10.2 : Orientations des directeurs de la RPGR

Si les raisons de l'échec de la *RPGR* sont à chercher avant tout dans la situation faite aux patois en France et dans l'absence d'un public suffisant de producteurs

<sup>1</sup>Par opposition à une étude aréale qui compare des échantillons de lexique à l'intérieur d'une zone donnée, la recherche monographique concentre l'enquête sur une seule localité en essayant de donner une représentation exhaustive du parler et de la culture populaire.

<sup>2</sup>D'origine suisse, Gilliéron n'a pas suivi les formations classiques en France (classes littéraires, ENS, agrégation) et ne s'est intégré ni aux réseaux politiques de ce temps, ni aux salons intellectuels, ni aux revues à forte visibilité (*Revue des Deux Mondes*, *Revue de Paris*, *Revue Politique et Littéraire)*. L'E.P.H.E. n'était pas au centre d'un réseau de relations comparables à celui que pouvait ouvrir une chaire en Sorbonne ou un accès aux ministères.

et de lecteurs, l'antagonisme des directeurs, qui ne s'accordaient pas sur la finalité du projet, expliquerait aussi l'interruption rapide d'une publication qui dès la fin de sa deuxième année avait dû suspendre temporairement sa parution. Les difficultés de la *Revue des Patois* de Clédat, rebaptisée en 1889 *Revue de Philologie française et provençale* (en distinguant les deux langues) et le projet inabouti d'une Société des Parlers de France dont le premier numéro du *Bulletin* en 1893 n'a pas eu de suite, confirment les résistances collectives qui contrevenaient à la reconnaissance des dialectes.

### **9 Conclusion**

La dialectologie représentait, en France, un point de rupture avec la priorité que les romanistes accordaient aux documents écrits. Ce que les écoles russe et américaine effectuaient au contact des peuples premiers présents sur leur territoire, les linguistes français le réalisaient dans un cadre radicalement différent : les populations rurales auprès de qui sont conduites les enquêtes sont de même origine, linguistiquement et anthropologiquement, que les savants qui viennent recueillir leur témoignage. La distinction, en termes de géographie, de niveau social et de culture est affaire de degré, non d'extériorité. Dans le même temps, pour les linguistes, reste omniprésente en arrière-plan la référence au latin et au français.

La *RPGR* est bien l'un des lieux où s'est accompli le partage entre la linguistique et la philologie d'une part, l'anthropologie d'autre part, sans que jamais ne parvienne à s'établir en France une revue pérenne consacrée à un domaine et à des langues qui s'imposaient pourtant par leur importance numérique et politique. Le refus de leur reconnaissance symbolique a entravé la recherche et la contradiction entre le soutien scientifique dont a bénéficié la *RPGR* et l'absence de réponse de l'Etat comme de la société est emblématique d'une contradiction jamais résolue.

### **Références**


### Gabriel Bergounioux


Gilliéron, Jules. 1888. Mélanges savoyards. *RPGR* 2. 31-37.


Rabiet, Eugène. 1887. Le patois de Bourberain (Côte d'Or). *RPGR* I. 241-255.


## **Chapter 11**

## **"Mithra aux vastes pâturages": L'antropologia di Émile Benveniste**

### Silvia Frigeni

Sapienza Università di Roma, Université Sorbonne Nouvelle – Paris 3

In the first part of this article we present two different ways of approaching ethnological concerns: the so-called "anthropology" of Émile Benveniste, the meaning of which will be further explained, and Antoine Meillet's sociolinguistic point of view. In the second part, these approaches are illustrated by two comparative studies by Meillet and by Benveniste respectively. Both studies happen to bear on the same subject (the Indo-Iranian god Mitra) and both are intended to provide a perspective on the culture and life of the people whose language they scrutinize.

### **1 Due diverse prospettive: antropologia e sociolinguistica**

### **1.1 L'***antropologia* **di Émile Benveniste: qualche precisazione**

La necessità di presentare un articolo tramite un titolo che riassuma e illustri il contenuto, magari riuscendo a interessare il lettore, si scontra a volte con la necessità di utilizzare termini il più possibile onnicomprensivi e vaghi, magari suggestivi ma bisognosi di una precisazione. Nel caso di questo articolo il termine "antropologia", accostato a un linguista e filologo quale fu Émile Benveniste (1902–1976), richiede senz'altro qualche spiegazione in più, che illustri tanto il senso in cui lo si intende quanto il ruolo che avrà nella trattazione.

A partire dalla pubblicazione nel 1966 del primo volume dei *Problèmes de linguistique générale*, la cui quinta sezione porta il suggestivo titolo "L'homme dans la langue", Benveniste passò da una notorietà circoscritta agli studiosi di lingue iraniche e ai filologi indoeuropeisti a un riconoscimento da parte di un pubbli-

### Silvia Frigeni

co ben più vasto, che comprendeva filosofi, psicanalisti, teorici del linguaggio, letterati e antropologi. La ragione era da cercare nell'interesse interdisciplinare mostrato da Benveniste negli articoli che compongono i due volumi dei *Problèmes* (il secondo sarebbe uscito nel 1974, a cura di collaboratori di Benveniste che mantennero la struttura voluta dall'autore per il primo volume). Da un'altra prospettiva, legata agli studi di linguistica storica, la pubblicazione del *Vocabulaire des institutions indo-européennes* nel 1969 fu accolta con favore dagli antropologi, interessati dall'uso fatto della ricostruzione etimologica come mezzo per ricostruire i rapporti sociali, economici, religiosi e politici espressi dai termini delle istituzioni presenti nelle lingue indoeuropee.<sup>1</sup>

Si intenderà quindi con "antropologia" di Benveniste la definizione che ne ha dato di recente Charles Malamoud, riferendosi al *Vocabulaire* come luogo in cui i confini tra indagine linguistica e ricerca antropologica diventano più sfumati. Il riferimento è alla distinzione tra "signification" e "désignation" che Benveniste pone nell' "avant-propos" dell'opera, con cui l'autore vorrebbe delimitare il suo campo di ricerca rispetto a quello degli studiosi di storia e di sociologia:

[...] étudiant des vocables ou des structures linguistiques plus complexes, Benveniste est amené à parler de ce que *désignent* ces vocables et de ce que révèlent ces structures. C'est-à-dire qu'il y a ce problème qu'il a lui-même maintes fois traité et qui est un des chapitres des études benvenistiennes – des études de Benveniste et des études sur Benveniste –, à savoir le rapport entre signification et désignation. En ce qui concerne la désignation, il y a des données, des faits, des institutions, des manières d'être, des gestes qui caractérisent à tel ou tel moment, dans tel ou tel domaine de la civilisation humaine.

Quand il s'agit de la désignation, c'est-à-dire ce que désignent les termes dont le linguiste s'efforce d'élucider la signification, nous sommes, me semble-til, dans l'anthropologie dès lors que le corpus des textes considérés parle de ce dont parlent les anthropologues: des termes de parenté, relatifs à la vie sociale, à la vie économique, à la vie religieuse. (Malamoud 2016: 246, in corsivo nel testo)

Assumendo questa propensione di Benveniste a interessarsi di ciò che effettivamente viene denotato dai termini di cui egli studia la significazione, che, se-

<sup>1</sup> Si veda il giudizio dato da Claude Lévi-Strauss in occasione del necrologio scritto per Benveniste e pubblicato sulla rivista d'antropologia *L'Homme*, che lui e Benveniste avevano fondato insieme: "il n'est pas excessif de dire que son ouvrage en deux volumes sur le *Vocabulaire des institutions indo-européennes* apporte à l'anthropologie sociale une contribution d'importance majeure." (Lévi-Strauss 1976: 5).

### 11 "Mithra aux vastes pâturages": L'antropologia di Émile Benveniste

guendo Malamoud, si indicherà qui con il nome di "antropologia", il presente articolo vorrebbe mostrare come questa prospettiva fosse già presente in altri lavori precedenti di linguistica storica.

Ci si occuperà in particolare di un articolo in cui Benveniste offre la sua interpretazione di un epiteto riservato al dio avestico Mithra: la scelta di questo lavoro in particolare è dovuta al trattarsi di un tema parzialmente sovrapponibile a quello affrontato dal suo maestro Antoine Meillet (1866–1936), che in un saggio rimasto famoso si occupò del significato del nome del dio indoiranico Mitra. L'analisi di entrambi gli articoli cercherà di mostrare i punti di contatto e le differenze tra le due prospettive.

### **1.2 Antoine Meillet e la sociologia linguistica**

L'interesse antropologico che compare nei lavori di Benveniste, così come è stato definito nel paragrafo precedente, si discosta in molti aspetti dalla prospettiva sociolinguistica assunta da Meillet. Si è già sostenuto che questo interesse sociologico ha dei tratti in comune, quando addirittura non ha influenzato, i lavori di Benveniste e di un altro celebre allievo di Meillet, Georges Dumézil (1898–1986), che pure seguiranno strade metodologicamente diverse dal maestro (Monod-Becquelin 1988; Lincoln 2012, per citarne alcuni): l'impostazione e gli scopi sono tuttavia divergenti se non opposti. La sociolinguistica di Meillet si rifà alla sociologia di Émile Durkheim (1858–1917), da cui riprende la nozione di *fatto sociale*. Nel suo celebre articolo "Comment les mots changent de sens" (Meillet 1906a), pubblicato nella rivista *L'Année Sociologique* fondata da Durkheim, afferma:

le langage est donc éminemment un fait social. En effet, il entre exactement dans la définition qu'a proposée Durkheim ; une langue existe indépendamment de chacun des individus qui la parlent, et, bien qu'elle n'ait aucune réalité en dehors de la somme de ces individus, elle est cependant, de par sa généralité, extérieure à chacun d'eux ; ce qui le montre, c'est qu'il ne dépend d'aucun d'entre eux de la changer et que toute déviation individuelle de l'usage provoque une réaction ; [...] Les caractères d'extériorité à l'individu et de coercition par lesquels Durkheim définit le fait social apparaissent donc dans le langage avec la dernière évidence. (Meillet 1906a: 230)

Sempre al 1906 risale anche un'altra menzione della lingua come fatto sociale: si tratta della lezione inaugurale del corso di Grammatica comparata del Collège de France, poi riportata nell'articolo "L'état actuel des études de linguistique

### Silvia Frigeni

générale" (Meillet 1906b). Va osservato (come fa Koerner 1988: 68) che Meillet non distingue tra *langage* e *langue* al modo di Ferdinand de Saussure, e usa i due termini in maniera pressoché equivalente. Soprattutto, la sua concezione sociale della lingua (o del linguaggio) si differenzia da quella del suo maestro Saussure: in Meillet, ma non in Saussure, è presente l'idea che i cambiamenti linguistici siano condizionati dai mutamenti sociali, e che quindi chi voglia studiare questi cambiamenti debba rifarsi al rapporto tra la grammatica della lingua in esame e lo stato di civilizzazione della società che utilizza quella lingua (Koerner 1988).

Benveniste citerà esplicitamente Meillet quando criticherà questo progetto, ormai considerato irrealizzabile (Benveniste 1966: 14–15). Tuttavia riprende lì dove le ricerche di Meillet avevano dovuto interrompersi: già alla fine degli anni Sessanta, un suo seminario tenuto al Convegno internazionale Olivetti a Milano sarà dedicato a questo tema (Benveniste 1974). Se non c'è isomorfismo tra lingua e società, argomenta Benveniste, non si può nemmeno negare il ruolo privilegiato che ha la lingua nell'indicare i cambiamenti sociali. Se "il n'y a de correspondance ni de nature ni de structure entre les éléments constitutifs de la langue et les éléments constitutifs de la société" (Benveniste 1974: 93), questo vuol dire che il rapporto che si cerca tra lingua e società non può essere una correlazione strutturale, ma sarà piuttosto di natura trascendentale:

la langue représente une permanence au sein de la société qui change, une constance qui relie les activités toujours diversifiées [...] de là procède la double nature profondément paradoxale de la langue, à la fois immanente à l'individu et transcendante à la société. (Benveniste 1974: 65).

Il rapporto che questa dualità ha con la società è sincronico, di tipo semiologico: si tratta del rapporto dell'interpretante con l'interpretato, la lingua interpreta e contiene la società. La scoperta della base comune a lingua e società, che Benveniste aveva posto come unica condizione possibile per realizzare il piano di studi prospettato da Meillet, si risolve nella primazia della prima sulla seconda tramite una via semiotica che non era quella del maestro.<sup>2</sup>

Due lavori pubblicati dai due studiosi a mezzo secolo di distanza sullo stesso argomento, l'interpretazione della figura del dio indoiranico Mitra, possono

<sup>2</sup>Non si entrerà qui nel merito della distinzione, che pure bisognerebbe fare, tra le elaborazioni teoriche a cui giunge Benveniste tra gli anni Trenta e Cinquanta e quelle degli anni Sessanta, quando diventa centrale la questione della semiologia della lingua. Per quanto riguarda le possibili eredità etnolinguistiche di Meillet in Benveniste, ad es. un punto di vista semantico sulle forme grammaticali viste come traduzioni simboliche di risposte ai problemi propri di ciascuna lingua, e che permetterebbe la comparazione di lingue isolate, cf. Monod-Becquelin (1988) e Benveniste (1966: 117).

### 11 "Mithra aux vastes pâturages": L'antropologia di Émile Benveniste

aiutare a illustrare questa continuità discorde. Come si vedrà, entrambi pongono dei problemi di natura sociologica e religiosa alla materia trattata, in cui lo studio delle etimologie, da cui pure partono, ha un valore marginale. Ma se pure Benveniste si pone un obiettivo che per certi aspetti va a completare il lavoro intrapreso da Meillet (l'interpretazione di una forma grammaticale il cui senso era rimasto oscuro), il metodo e il punto di vista sono necessariamente divergenti.

In questo senso andrebbe approfondito il ruolo che riveste la fraseologia formulare in entrambi i lavori. Si tratta di un compito che non si può portare a termine qui per mancanza di spazio, ma che sarebbe di grande interesse per future ricerche. Permettendo di non fermarsi all'etimologia della singola parola, ma di comparare le frasi formulari tra diverse lingue all'interno di una stessa famiglia, la fraseologia è stata uno strumento importante per gli indoeuropeisti, che però hanno spesso trascurato l'aspetto semantico delle formule considerate, preferendo concentrarsi sul solo aspetto etimologico. Nel lamentarsi di questa mancanza, Calvert Watkins (1933–2013) ha sottolineato come Benveniste sia stato una delle rare eccezioni, dato che sarebbe riuscito a riconoscere "the function of these formulas as expressions of an underlying semiotic system" (Watkins 1995: 43). L'importanza della semantica negli studi di Meillet, e la differente prospettiva semiotica e interpretativa adottata da Benveniste, potrebbero fornire uno spunto di riflessione da cui esaminare l'impiego della fraseologia così come si presenta nei due autori, e le diverse questioni teoriche e di metodo che questo comporta.

### **2 Gli studi sul dio Mitra**

### **2.1 "***Le dieu indo-iranien Mitra***"**

L'importanza accordata alla semantica, l'attenzione rivolta all'aspetto sociale delle lingue storiche, la scrupolosità dell'analisi linguistica: sono tutti aspetti che avvicinano Benveniste all'insegnamento di Meillet.

Suo predecessore all'École pratique des Hautes Études e al Collège de France, fu Meillet a indirizzare il giovane Benveniste allo studio delle lingue iraniche: un dominio che sarebbe rimasto il principale tra i molti trattati da Benveniste nel corso della sua carriera. Lo stesso Meillet pubblicò alcuni lavori nel settore, sia pure occupandosi quasi esclusivamente di avestico e antico persiano: vanno ricordate la *Grammaire du vieux-perse*, pubblicata nel 1915 e ristampata con sostanziali modifiche a opera di Benveniste nel 1931, e le *Trois conférences sur les Gâthâ de l'Avesta* del 1925. Ma già nel 1907 era stato pubblicato "Le dieu indo-iranien Mitra", che sarebbe rimasto un punto di riferimento per gli studiosi successivi, e

### Silvia Frigeni

che è stato considerato l'opera in cui "the sociological and philological approaches to the study of religion earlier associated with Fustel and Max Müller first powerfully came together" (Lincoln 2012: 13).

In questo breve saggio, Meillet respinge l'ipotesi condivisa da diversi autori, tra cui il pioniere degli studi sulle lingue iraniche Christian Bartholomae (1855– 1925), che vedeva nel dio vedico *Mitra-* e nella sua controparte avestica *Mithra*una divinità solare, invocata insieme al cielo. Pur riconoscendo il significato del nome comune *mitrá*, codificato come 'contratto' nel suo *Altiranisches Wörterbuch* (1904, col. 1183), Bartholomae non aveva chiarito il nesso tra questo sostantivo e il nome proprio del dio indoiranico. Le interpretazioni successive avrebbero attribuito una connotazione morale al nome del dio, che con la sua luce protegge la verità e combatte la menzogna: il nome comune sarebbe perciò derivato da una funzione esercitata dalla divinità.<sup>3</sup>

Rettificando questa visione, Meillet osserva che non c'è differenza tra il nome proprio del dio e i nomi comuni presenti in sanscrito (*mitrá-* 'amico') e in avestico (*mithra* 'contratto'). Entrambi sono derivati da un comune termine indoiranico \**mitrá-*, di cui Meillet rintraccia la radice i.e. *\*mei-* 'scambiare', presente in diverse forme sia nominali che verbali di altre lingue indoeuropee.<sup>4</sup> Fanno parte di questa discendenza comune ad es. il verbo sanscrito *máyate* 'egli scambia', ma anche termini nominali come l'antico slavo *měna*, che vuol dire 'scambio' ma soprattutto 'contratto': un altro sostantivo slavo *miră* 'pace, ordine' ha portato al russo *mir* 'comunità', poi 'comunità di paesani' e quindi 'villaggio'.

Il significato di 'amico, amicizia' presente nel termine comune sanscrito non è quello attribuibile al nome proprio della divinità, afferma Meillet, perché "[il] ne se concilie pas avec le caractère général du dieu" (Meillet 1907: 145): inoltre non è confermato dal corrispondente termine iranico. *Mitra-* è "la personnification du contrat", non diversamente dalle dee greche Temi e Dike per la giustizia o della dea romana Venere per la grazia femminile. Qui Meillet si rifà al suo maestro Michel Bréal (1832–1915) che aveva spiegato il ruolo delle Erinni comparandole alle αρ(Ϝ)άι, le maledizioni personificate, sopperendo così alla scarsa

<sup>3</sup>Cf. Schmidt 2006

<sup>4</sup>È d'obbligo segnalare (anche se per motivi di spazio non si potrà entrare nel dettaglio) che le posizioni di Meillet sono state successivamente criticate da altri studiosi. La ricostruzione etimologica da lui fornita è solo una delle possibili e, quanto all'interpretazione del nome, sono state proposte connotazioni che si accordassero meglio ad alcuni dei contesti dell'utilizzo, come 'alleanza, obbligo morale' (cf. ad es. Herzferd 1947, Brereton 1981), o al carattere compassionevole del dio (come in Lentz 1964 e in Gonda 1972): vedi il già citato Schmidt (2006). Per una panoramica sul dibattito filologico generato dal lavoro di Meillet vedi Manfred Mayrhofer, "mitrá-", in Mayrhofer 1996: 354–355.

### 11 "Mithra aux vastes pâturages": L'antropologia di Émile Benveniste

chiarezza etimologica del loro nome. L'etimologia non è perciò il metodo principale né il fine dell'indagine: quando rintracciabile, funge da controprova della giustezza dell'interpretazione proposta. "Les personnalités divines dont le nom est étymologiquement clair dans les langues indo-européennes sont toutes ainsi des personnifications de noms communs" (Meillet 1907: 145–146).

Riprendendo il caso del dio del sole Helios così come interpretato nella *Griechische Mythologie* del mitografo tedesco Otto Gruppe, pubblicata l'anno prima, Meillet chiarisce la tripartizione in cui vede definirsi il nome del dio Mitra:

Hélios est d'abord ce que son nom indique, le soleil ; en second lieu, la puissance naturelle mystérieuse qui agit dans le soleil ; en troisième lieu, la personne dont on rapproche cette puissance naturelle, et les trois notions sont interchangeables." De même l'indo-iranien *Mitra-* est le contrat, la puissance mystique du contrat, et une personne ; et les trois notions s'interchangent constamment. (Meillet 1907: 146)

Una tale definizione richiede un contesto in cui essere valida. Meillet passa qui alla seconda parte della sua analisi, quella in cui deve verificare "si cette doctrine rend compte de ce qu'indiquent les plus anciens documents connus, en l'espèce, les Védas et l'Avesta, sur le caractère du dieu indo-iranien *Mitra-*" (Meillet 1907: 146).

Per quanto riguarda i testi vedici, Mitra ha un solo inno del Rigveda (III, 59) a lui consacrato. Pur nella sua brevità, questo testo è sufficiente per Meillet a mostrare due tratti fondamentali del dio: "d'une part, il ne présente aucun trait qui indique un caractère naturaliste quelconque du dieu, et de l'autre, il est clair que *Mitrá-* surveille sans sommeil les tribus humaines, et qu'on doit demeurer dans le contrat formé avec lui" (Meillet 1907: 147). Assieme a Varuna, la divinità con cui compare spesso accoppiato nel pantheon vedico, Mitra è un Aditya, un guardiano dell'ordine universale: si tratta di personificazioni puramente morali, prive di natura fisica (Meillet 1907: 147).

Escluso dalle grandi religioni ufficiali, il Mithra iranico sembra appartenere a un culto antico e importante poi introdotto nel sistema del mazdeismo zoroastriano: questo culto doveva essere di origine indoiranica e non attribuibile a un prestito indiano, vista la sua scarsa importanza nei Rigveda. Fin dall'inizio dell'inno X dello Yasht a lui dedicato, Mithra viene presentato come colui con il quale non è possibile rompere un contratto, e che non può essere ingannato (Yasht X, 2). Qui Meillet rintraccia la sua parentela col sole: considerato in molte tradizioni indoeuropee simile a un occhio che vede tutto, il sole diventa qui l'occhio della divinità incaricata di punire la menzogna (analogamente a quanto avviene per

### Silvia Frigeni

gli Adityas nella tradizione indiana) e deve sorvegliare le azioni degli uomini. Lo Yasht X, 7, in cui si invoca Mithra "aux mille oreilles, bien fait, aux dix mille yeux, haut, à la connaissance étendue, fort, sans sommeil, éveillé", è paragonabile al passo del Rigveda (VII, 34, 10) in cui Varuna viene definito "fort, aux mille yeux" (Meillet 1907: 150).

Nel caso dell'Avesta la relazione tra il dio e il sole è meno netta: introdotto solo successivamente nel sistema zoroastriano, Mithra sorveglia in prima persona le infrazioni ai contratti. Luce che penetra ovunque e illumina ogni trasgressione, Mithra non è quindi una divinità solare ma è stata a poco a poco identificata con il sole: "étant le contrat, [il] a tous les moyens de punir les violations du contrat" (Meillet 1907: 153), fino ad arrivare al carattere guerriero che contraddistingue il dio iranico rispetto alla sua controparte indiana.

Chiarito questo aspetto del dio e quindi accantonato il suo carattere naturalista, Meillet dedica la terza parte della sua trattazione a giustificare una simile conclusione.

On ne doit pas être surpris de voir diviniser le contrat ; car le contrat était dès le principe un acte religieux, entouré des cérémonies définies, fait avec certains rites ; et les paroles qui l'accompagnaient n'étaient pas de simples promesses individuelles ; c'étaient des formules, douées d'une force propre, et qui se retournaient, en vertu de cette force interne, contre le transgresseur éventuel. Le *Mitra-* indo-iranien est à la fois le « Contrat » et la puissance immanente du contrat. (Meillet 1907: 156)

Il legame con la religione e con le formule rivela più di ogni altra cosa l'importanza sociale del dio, la sua valenza antropologica, soprattutto per la cultura iranica. Mentre il Mitra vedico è rimasto poco sviluppato, la divinità iranica è diventata talmente importante e potente da imporsi alla tradizione ortodossa mazdeana, e a diffondersi presso le popolazioni che hanno subìto l'influenza iranica, come gli armeni, fino a venire stravolta nel culto mitraico dei Romani. In tutte rimane però la sua caratteristica principale, comune al tipo religioso di epoca indoeuropea: "ce n'est pas un phénomène naturel, c'est un phénomène social divinisé" (Meillet 1907: 159).

### **2.2** *"Mithra aux vastes pâturages"*

A differenza di Meillet, Benveniste si occuperà a più riprese del dio Mithra, privilegiandone la versione iranica. In un articolo pubblicato nel 1960 e dedicato esclusivamente a questa divinità, Benveniste non cita nemmeno il lavoro del maestro

### 11 "Mithra aux vastes pâturages": L'antropologia di Émile Benveniste

di più di cinquant'anni prima. Eppure la sua influenza è evidente fin dal principio della trattazione. Lo scopo dell'articolo è quello di chiarire il significato dell'epiteto *vouru.gaoyaoitiš*, che ricorre costantemente associato a Mithra nell'Avesta e in particolare nello Yasht X a lui dedicato. Nonostante la sua frequenza debba denotare qualche carattere specifico del dio, i commentatori l'hanno tradotto ovunque come "Mithra aux vastes pâturages", che una vecchia tradizione voleva collegato al suo ruolo del dio del sole che fa fruttificare le campagne. "Mais on ne croit plus à cette image naturiste du dieu. Cependant, le même type d'interprétation persiste dans l'exégèse moderne" è l'unico commento che faccia pensare a un riferimento di Benveniste a Meillet, sia pure velato (Benveniste 2015 [1960]: 277).

D'altra parte lo stesso epiteto compariva già nell'articolo del 1907. Lì Meillet si era arreso all'impossibilità di definirne la seconda parte, etimologicamente oscura, ma aveva tentato una soluzione:

il est sans doute impossible de pénétrer entièrement le sens d'un mot fixé par la tradition et qu'un long usage rituel a usé et obscurci ; mais le rapprochement des passages védiques montre que la *gávyūtih* qui répond à la *gaoyaoitiš* iranienne est un espace où le fidèle demande au dieu, et notamment à Mitra, de le protéger. L'épithète *vourugaoyaoitiš* atteste donc le caractère indo-iranien du dieu et concorde avec le rôle qui lui est attribué ici. (Meillet 1907: 156)

Benveniste non cita questa conclusione di Meillet, che pure va nella sua stessa direzione non naturalista. Si limita a osservare che anche gli studi più recenti persistono nell'interpretare Mithra come colui che fa scendere la pioggia, assicurando così l'acqua ai campi: lo stesso ruolo che doveva avere il dio vedico Soma, il cui titolo è il corrispettivo indiano *urúgávyūti-*. Ma questo ruolo non può essere proprio di Mithra, visto che altri dèi mazdeani ben più legati al dominio delle acque non ricevono un tale epiteto. La sola strada da intraprendere è perciò per Benveniste lo studio del termine nella sua forma avestica e in quella vedica "selon sa forme étymologique d'abord, puis dans ses emplois textuels" (Benveniste 2015 [1960]: 278), la stessa percorsa da Meillet.

La parte etimologica è liquidata in fretta da Benveniste come non problematica: "tout l'essentiel du problème est hors de l'étymologie" (Benveniste 2015 [1960]: 278). Si tratta quindi di esaminare i testi vedici e avestici per ottenere il contesto da cui trarre il senso del termine.

Per quanto riguarda i testi avestici, la sola occorrenza di *gaoyaoiti-* non in composizione si trova nello Yasht X, consacrato a Mithra. "On peut dégager de cette

### Silvia Frigeni

strophe – l'unique exemple, rappelons-le, de *gaoyaoiti* en contexte non formulaire – une conception assez précise de la notion" (Benveniste 2015 [1960]: 281). A chiarirne il senso è l'epiteto presente subito dopo, e occorrente anche in un altro inno, che esprime il sentimento di indipendenza che provano bestie e uomini quando Mithra dona loro strade larghe, profonde per la *gaoyaoiti.*

La *gaoyaoiti* est cette zone de sécurité collective dont le dieu trace les accès dans les pays où il reçoit les égards dus. Qu'hommes et bétail y trouvent subsistance peut confirmer que le terme désignait d'abord un « pâturage », mais bien plus importante est cette connotation de la *gaoyaoiti* comme lieu d'asile sous la protection de Mithra. (Benveniste 2015 [1960]: 281)

L'epiteto *vouru.gaoyaoitiš* esprimerebbe quindi il vasto spazio di sicurezza che Mithra accorda a coloro che gli si affidano. Il suggerimento di Meillet viene quindi sostanzialmente confermato da Benveniste: non però tramite l'utilizzo di analisi etimologiche più avanzate o di nuove conoscenze sopraggiunte, ma grazie a un utilizzo più sistematico (e forse più spericolato) del raffronto degli impieghi testuali, fuori dall'etimologia.

Tale significato viene ribadito da Benveniste in un altro composto, l'epiteto *vasō.gaoyaoiti-*, che attribuirebbe a Mithra la facoltà di dispensare la *gaoyaoiti* a suo piacimento. Questo però per Benveniste non può essere tradotto come è stato fatto, attribuendo a Mithra la capacità di dispensare pascoli. Qui l'analisi di Benveniste sconfina in una breve trattazione antropologica del popolo iranico, non certo la più significativa né la più approfondita fra quelle da lui fatte, ma che mostra il possibile sconfinamento dell'analisi nel territorio della *désignation* (per cui cf. Benveniste 1969: 10):

un fidèle mazdéen n'a jamais demandé à un dieu – ni surtout à Mithra – des pâturages ; l'espace ne manquait pas à ces tribus iraniennes des premiers âges. L'objet de leurs prières, leur plus constant souci, était double : l'eau et la sécurité. L'eau avait ses dieux, ses mythes, ses rituels. Mais pour la sécurité, on comptait d'abord sur Mithra. (Benveniste 2015 [1960]: 282)

L'analisi dei testi vedici conferma sostanzialmente quanto Benveniste ha già affermato. Ma la controparte del Mithra iranico non è il solo Mitra, piuttosto l'entità Mitra-Varuna. Dal punto di vista del corrispettivo formale *urúgávyūti-* ricorre una volta sola, ma è sostituito nella sostanza dall'espressione *urvī gávyūti-*, la vasta *gávyūti*, che nella tradizione vedica può essere concessa anche da altri dèi, come il già citato Soma.

### 11 "Mithra aux vastes pâturages": L'antropologia di Émile Benveniste

Dalla ricerca del significato di un termine tramite il metodo comparativo si è arrivati così alla definizione di un ente concreto, spazialmente e temporalmente localizzabile. "La *gávyūti* est, adjacent à la localité habitée, un territoire tribal, protegé par un dieu souverain, où les hommes et le bétail sont à l'abri des incursions et des calamités" (Benveniste 2015 [1960]: 284).

Questa sicurezza può essere concessa solo da un dio: la differenza tra il mondo vedico e quello iranico fa supporre che la sua relazione con Mithra si sia rinsaldata nella protostoria iranica. L'antichità del culto di Mithra in Iran e la sua speciale appartenenza al mondo iranico è la conclusione che accomuna Benveniste e Meillet, sia pure per vie diverse.

Anche la scarsa trasparenza etimologica del termine, ormai non riconducibile alla sua radice, viene spiegata da Benveniste in maniera non diversa da quanto affermato da Meillet. Il continuo uso rituale che ne è stato fatto ha legato definitivamente questa parola alla nozione di potere divino, staccandola dalla sua origine, ma allo stesso tempo rendendola rivelatrice di antiche credenze e di aspetti poco noti della figura divina di Mithra.

### **3 Conclusioni**

Era stato Meillet a negare un'interpretazione naturalistica in favore di una sociale. Tuttavia il lavoro di Meillet non viene mai menzionato, in un testo in cui pure se ne menzionano altri, anche lavori con cui Benveniste non si trova d'accordo.

Una prima spiegazione può essere che, in effetti, per Meillet l'espressione oggetto dell'articolo rimane oscura, anche se l'intuizione da lui esposta (lasciata volutamente incerta perché mancante di prove sufficienti) va nella stessa direzione dell'interpretazione data da Benveniste.

Un'altra possibile spiegazione riguarda il fatto che il lavoro di Meillet sia considerato la base implicita da cui parte l'analisi di Benveniste. Il suo articolo ha segnato una svolta nel modo di concepire la figura di Mithra tale per cui non è necessario menzionarlo, basta far riferimento al fatto che non si considera più il dio da un punto di vista naturalistico. Ma soprattutto, Benveniste non ha bisogno di menzionarlo perché nei fatti ne è erede e continuatore.

Pur nella diversità del metodo, la lingua storica rimane, nell'uno come nell'altro autore, lo strumento attraverso il quale comprendere la società che si esprime attraverso di essa. Semantica e fraseologia sono ciò che spinge Benveniste a trarre le conclusioni finali riguardanti il vero significato di *gavyuti*: il significato di insieme della frase, la considerazione delle altre parole in essa presenti, la valenza sacrale delle espressioni.

Il culto del dio come dio di giustizia, la stipulazione di un contratto, il valore della mancanza di menzogna che sola permette di stipulare patti, può far comprendere perché fosse a esso demandata la creazione della sicurezza.

Dio della repressione ma anche della ricompensa verso chi lo teme, il Mithra avestico visto da Meillet e Benveniste è il dio che sostiene le fondamenta della società, intesa come luogo che si regge sull'accordo tra persone basato sull'integrità e il rispetto degli accordi. Si può perciò comprendere perché questo dio, nella sua valenza tutta particolare a lui riservata nel pantheon iranico e differente rispetto alla tradizione vedica, attirasse l'attenzione di Meillet e Benveniste: quest'ultimo ci tornerà lungo tutta la sua carriera.

La presenza in Benveniste di un pensiero linguistico in cui la semantica ha un posto sempre più rilevante, dove si dà valenza al concetto antropologico di lingua intesa come espressione di un mondo, e di vita insieme, viene qui espressa in modo stringato e circoscritto a un problema linguistico, come del resto gli è proprio. Allo stesso tempo, è proprio questa coerenza metodologica di Benveniste a permettere di notare la vicinanza di questo lavoro alle sue coeve riflessioni di carattere generale sull'uomo, la lingua e la società. La vicinanza delle sue riflessioni alla lezione di Meillet, qui più che mai evidente, ci consentirà forse allora di comprenderne meglio lo sviluppo e di rintracciare nella grammatica comparata una delle sue possibili origini.

### **Riferimenti bibliografici**


11 "Mithra aux vastes pâturages": L'antropologia di Émile Benveniste


## **Chapter 12**

## **Per una semiologia materialista e dialettica: Trần Đức Thảo critico di Saussure**

### Jacopo D'Alonzo

Sapienza Università di Roma

Trần Đức Thảo was a specialist of phenomenology familiar with the French existentialists, a Marxist and an anti-colonial activist. He devoted much of his effort to describing the ontogenetic and phylogenetic origins of consciousness and language. In this vein, he proposed a general semiology that could enable him to describe all the stages of the development of human symbolic abilities. In this paper, we study the theoretical issues involved in Thảo's criticism of the semiotic model proposed in Saussure's *Cours de Linguistique Générale* and more generally of the structuralist readings of the *Cours*. In the last part, we introduce Thảo's notion of a "language of the real life".

### **1 Introduzione**

Negli ultimi anni, l'attenzione della comunità scientifica si è rivolta sempre più frequentemente ai rapporti tra fenomenologia, strutturalismo e saussurismo (De Palo 2016; Aurora 2017). A questo proposito l'opera del filosofo vietnamita Trần Đức Thảo (1917–1993), specialista della fenomenologia husserliana, merita una menzione speciale. Da un lato la (relativamente) celebre riflessione di Thảo sulla filosofia di Edmund Husserl (1859–1938) non può essere isolata dalle ricerche più generali che Thảo ha dedicato al linguaggio. D'altro canto, la teoria dell'origine del linguaggio proposta da Thảo – che ha solo di rado attirato l'attenzione degli studiosi – si inscrive nel contesto di una polemica nei confronti, sia dello strut-

Jacopo D'Alonzo. 2020. Per una semiologia materialista e dialettica: Trần Đức Thảo critico di Saussure. In Émilie Aussant & Jean-Michel Fortis (cur.), *Historical journey in a linguistic archipelago: Descriptive concepts and case studies*, 189–202. Berlin: Language Science Press. DOI: 10.5281/zenodo.4269429

### Jacopo D'Alonzo

turalismo generalizzato degli anni Sessanta, sia della fenomenologia husserliana. In simile quadro, Thảo ha condotto una critica poco nota a una certa lettura del modello semiologico introdotto nel *Cours de linguistique générale* (CLG) di Ferdinand de Saussure (1857–1913). Nel lavoro che segue, proveremo a tracciare le linee direttrici della critica di Thảo al CLG per metterne in evidenza le numerose implicazioni teoriche.

### **2 La vita e l'opera di Trần Đức Thảo**

Trần Đức Thảo può essere considerato uno dei più importanti intellettuali di lingua francese – tra i quali bisogna annoverare Jean-Paul Sartre (1905–1980) e Maurice Merleau-Ponty (1908–1961) – che, negli anni Quaranta, hanno discusso la possibilità di conciliare marxismo e fenomenologia (cf. Thảo 1946,1949). Presto, però, si produce una rottura profonda tra Thảo e l'ambiente parigino (Thảo 1950, 1951).<sup>1</sup> Dopo aver condotto i suoi studi superiori presso l'ENS di Parigi tra il 1939 e il 1942 ed essere diventato una voce autorevole nel *milieu* filosofico parigino, malgrado la sua origine straniera, nel 1951 Thảo decide di tornare nel suo paese natale, il Vietnam, per prendere parte alla lotta di liberazione nazionale. Nel 1951, poco prima della sua partenza per il Vietnam, dove resterà sino al 1991, Thảo raccoglie e pubblica i risultati delle sue ricerche sulla fenomenologia condotte durante il decennio precedente.

*Phénoménologie et matérialisme dialectique* (PMD) si apre con un'introduzione alla fenomenologia husserliana e una descrizione del metodo fenomenologico "d'un point de vue purement historique" (Thảo 1951: 5). Thảo (1951: 7) considera la fenomenologia di Husserl una forma di idealismo. La riduzione trascendentale alla sfera del vissuto come polo costituente dei fenomeni avrebbe dovuto essere molto più radicale: in tal caso, la fenomenologia avrebbe mostrato che: (1) l'ego trascendentale non è che l'ego storico e concreto; (2) la soggettività trascendentale è essa stessa costituita dal movimento della storia naturale e sociale che la precede; (3) il vissuto cosciente è sempre preceduto dall'organismo e dalla sua attività; (4) l'esperienza ante-predicativa si situa a livello della vita animale e non a quello della vita umana; (5) il vissuto è l'aspetto astratto della vita reale.

Nella seconda parte di PMD appare per la prima volta un progetto che impegnerà Thảo per il resto della sua vita, cioè una descrizione delle dinamiche naturali e storiche che hanno favorito lo sviluppo della coscienza dagli organi-

<sup>1</sup>Per maggiori ragguagli bibliografici su Thảo si veda Thảo (1991: 1–11), Thảo (1993), Giao (1988), Hémery (2013), Thao (2004), Thao (2013), Feron (2014).

### 12 Per una semiologia materialista e dialettica

smi unicellulari all'umanità; una descrizione condotta dal punto di vista di una metafisica materialista e secondo una logica dialettica.

In Vietnam, Thảo è presto escluso dalla vita politica e accademica del paese in seguito a un conflitto con il Partito comunista a proposito delle libertà civili, a cui fa seguito un periodo di reclusione. Tuttavia, negli anni Sessanta, si impegna in un vasto progetto di ricerca sulle origini della coscienza e del linguaggio, come testimonia una serie di articoli (Thảo 1966, 1969a, 1969b, 1970) riuniti poi nelle *Recherches sur l'origine du langage et de la conscience* (RLC) nel 1973 e in cui Thảo cerca di rendere ragione della cognizione umana – seguendo le indicazioni di alcuni classici del marxismo – a partire dalla vita pratica e collettiva dei nostri predecessori ominidi (Federici 1970; Caveing 1974; Haudricourt 1974; Frédéric 1974).

La teoria di Thảo s'impernia su tre ipotesi: (1) la coscienza emerge nel e grazie al linguaggio considerato nella sua materialità e nella sua funzione pratica e operativa; (2) il linguaggio non è un oggetto, ma la mediazione tra l'uomo e la realtà, tra uomo e uomo, e tra l'individuo e se stesso; (3) il linguaggio non può essere studiato come una realtà autonoma, ma bisogna osservarlo all'interno della vita sociale e pratica.

A fondamento del linguaggio umano, sia dal punto di vista diacronico che sincronico, ci sarebbero, secondo Thảo, alcuni segni naturali fondamentali (gesti, vocalizzazioni, espressioni fisiognomiche, ecc.) in cui la relazione tra il significante e il significato non è né arbitraria né convenzionale. Tali segni hanno, infatti, uno spiccato carattere corporeo e fanno parte della vita pratica e collettiva. La loro interiorizzazione psichica segue necessariamente il loro uso, nella misura in cui essi sono l'espressione immediata della vita del gruppo umano che li utilizza prima di divenire strumento d'espressione a disposizione degli individui.

### **3 Alcuni aspetti della critica di Thảo al CLG**

Thảo pubblica tra il 1974 e il 1975 un articolo dal titolo *De la phénoménologie à la dialectique matérialiste de la conscience* che serviva da introduzione, sia biografica che teorica, alle RLC. Thảo prende esplicitamente di mira la teoria del segno proposta nel CLG. Thảo reputava la sua ipotesi sull'origine del linguaggio, così come la semiologia che ne era il supporto teorico, in disaccordo radicale con alcune ipotesi del CLG, riprese poi anche dallo strutturalismo.

Preliminarmente va detto che Thảo ignorava la storia editoriale che ha condotto alla pubblicazione del CLG. Al di là della questione concernente le conoscenze che Thảo aveva a disposizione riguardo la redazione del CLG, quello che inte-

### Jacopo D'Alonzo

ressa mettere in luce è il fatto che la critica di Thảo prendeva di mira una certa lettura del CLG e in particolare quella che difendeva un modello autonomista del linguaggio. Tuttavia, la lettura che egli offre del CLG investe più livelli teorici e va ben al di là della polemica nei confronti dello strutturalismo generalizzato (Chiss et al. 2015 e Léon 2013), benché questo sia il suo punto di partenza.

a) Negli anni Sessanta e Settanta non mancava un certo disaccordo con le tesi dello strutturalismo generalizzato e con una certa recezione, circolazione e interpretazione del CLG e la cui portata trascendeva gli orizzonti della linguistica (Dosse 1991; 1992; Puech 2013a,b; Lepschy 1966).

Tra le critiche dell'estensione all'insieme degli oggetti delle scienze umane e sociali del modello semiologico incentrato sulla *langue* come forma sopra-individuale composta da elementi negativi e differenziali, si devono menzionare quelle provenienti dal fronte marxista. Nel loro insieme, le critiche di certi marxisti manifestano un posizionamento abbastanza omogeneo dal punto di vista strategico, ideologico e teorico. Per citare solo due esempi, si devono ricordare Sève (1984) e Lefebvre (1971). Secondo questi autori, l'autonomia delle strutture conduce a pensare le stesse come entità metafisiche e a identificare l'ideale con il reale, i prodotti della scienza con la realtà effettiva (si tratta della stessa denuncia dello strutturalismo ontologico che si trova anche in Eco 1968).

Thảo condivideva le stesse preoccupazioni dei suoi colleghi francesi di orientamento marxista, ma aveva un obiettivo differente: si trattava per lui di far emergere i principi fondamentali di una semiologia generale che gli avrebbe permesso di descrivere lo sviluppo filogenetico del linguaggio e della coscienza. E per questo motivo presenta il suo progetto semiologico come un antidoto per ridurre il campo d'applicazione dei due concetti chiave dello strutturalismo, quello di arbitrarietà e quello di valore. Thảo decide dunque di volgersi alla fonte di queste due nozioni:

En 1964, je reçus les premiers échos des succès retentissants du structuralisme dans les pays occidentaux. L'étude du *Cours de linguistique générale* de Ferdinand de Saussure s'imposait comme une nécessité urgente. (Thảo 1974: 39)

Da questo punto di vista, il CLG perde il suo status di oggetto storico e viene collocato nell'attualità del dibattito teorico.

b) Per Thảo il CLG è allo stesso tempo sia l'obiettivo di una critica rivolta a una prospettiva che vede nella *langue* un'entità psichica, interiore, separata

### 12 Per una semiologia materialista e dialettica

dalle pratiche linguistiche (che Thảo stesso riconosce confermata solo in parte dal testo del CLG), sia il luogo per una ricerca di un'altra semiologia possibile: "On peut lire ainsi à travers le texte du *Cours de linguistique générale,* en transparence et pour ainsi dire en pointillé, la possibilité et la nécessité d'une autre sémiologie […]" (Thảo 1974: 40). Il CLG è così il luogo di un progetto solamente abbozzato che bisogna ristabilire e sviluppare, quello, cioè, di una semiologia che prenda in conto un insieme di sistemi di segni più ampio di quello dei segni arbitrari:

Cependant l'auteur [Saussure] avait lui-même reconnu au début de la première partie du livre l'existence de toute une classe de signes présentés comme "signes naturels", soit entièrement comme la pantomime, soit partiellement comme les signes de politesse, les symboles, etc. (Thảo 1974: 39)

Thảo (1974: 42) denomina il suo progetto "*sémiologie dialectique*", il cui oggetto sarebbe il "*système général des signes intrinsèques, ou esthétiques*" (Thảo 1974: 40), cioè il sistema di segni motivati che mostrano direttamente all'intuizione sensibile il loro significato.

c) Il CLG e in particolare le nozioni saussuriane dell'arbitrarietà e del valore sono per Thảo l'oggetto di una critica, nel senso kantiano del termine, vale a dire di una delimitazione della loro legittimità teorica. Si tratta di una questione ben nota tra i lettori del CLG (De Mauro 2011: 413–416; Sofia 2013) e che, ancora oggi, è al centro del dibattito sull'eredità saussuriana (Rastier 2002; Paolucci 2012; Laks 2012; Coursil 2015). Anzitutto, si tratta per Thảo di stabilire i limiti della nozione di arbitrarietà rifiutando di prenderla come unico criterio per giudicare ogni tipo di segno e riabilitando così i segni motivati.

Thảo prende, poi, anche di mira la riduzione della significazione al valore dato che, secondo lui, la significazione concerne: i) il valore intrinseco di quei segni che non sono totalmente arbitrari; ii) il valore differenziale dei segni arbitrari; iii) la relazione tra i segni e la realtà trascendente (realtà materiale, esperienza pre-linguistica, ecc.).

Per delimitare il campo di validità dell'arbitrarietà e del valore, Thảo si impegna così in un'analisi della natura dei segni non totalmente arbitrari così come nello studio della motivazione intrinseca agli atti concreti di linguaggio:

### Jacopo D'Alonzo

Déjà le langage ordinaire cherche à obtenir par l'intonation, le choix des mots et des tournures, la disposition de la phrase, une certaine *qualité expressive* non réglée en tant que telle par les conventions du code, et qui contribue, parfois de manière décisive, à la signification. (Thảo 1974: 39–40)

Thảo (1974: 39) pensa dunque che il sistema dei segni intrinseci sia la condizione di possibilità dei sistemi di segni arbitrari di cui il miglior esempio sarebbe la lingua convenzionale delle discipline scientifiche, "qui vise essentiellement à exprimer distinctement des idées distinctes et, pour ce but, utilise autant que possible une langue conventionnelle". Per Thảo la significazione dei segni intrinseci non dipende né dalla nozione di valore né da quella di arbitrarietà, ma dalla produzione dei segni come si presenta nella vita pratica, nel "mouvement sémiotique matériel" (Thảo 1974: 39).

d) Il CLG è anche l'obiettivo di una critica che mira alle fondamenta empiriche della semiologia saussuriana. Saussure aveva spiegato che "le signe linguistique unit non une chose et un nom, mais un concept et une image acoustique" (Saussure 1995: 98). E l'immagine acustica (signifiant) "n'est pas le son matériel, chose purement physique, mais l'empreinte psychique de ce son, la représentation que nous en donne le témoignage de nos sens". Benché Saussure abbia sostenuto che la natura mentale del segno dipende dalle esperienze percettive anteriori, secondo Thảo, la sua linguistica della *langue* si rivolge solamente all'aspetto psichico dei fenomeni linguistici. Così Thảo fa notare che:

> une telle théorie s'inspirait manifestement d'une psychologie qui n'est plus acceptable de nos jours [...]. En réalité on ne peut pas séparer le langage intérieur, à titre de pure opération idéale, des mouvements réels plus ou moins esquissés, de la voix et du geste. (Thảo 1975: 25– 26)

Di conseguenza, la semiologia saussuriana tradisce un mentalismo che non può più essere difeso.

e) Allo stesso tempo, la lettura di Thảo si rivolge anche ai presupposti filosofici del CLG. Per Saussure il pensiero, prima del linguaggio, sarebbe una "masse amorphe et indistincte" (Saussure 1995: 155) e la *langue* ne ordinerebbe il flusso articolando quella massa in segmenti psichici (significanti e significati). In questo passaggio argomentativo, secondo Thảo, la semiologia di Saussure cade in errore poiché mostrerebbe l'ipotesi idealista che

### 12 Per una semiologia materialista e dialettica

la sostiene. Quella di Saussure, più che una "théorie linguistique de la signification verbale" (cioè una semantica), è una "théorie gnoséologique du concept [cioè una teoria della conoscenza]" (Thảo 1974: 41). I segni linguistici sarebbero la condizione di possibilità dell'articolazione del pensiero e lo articolerebbero in maniera esclusiva. Ne risulta che il pensiero articolato linguisticamente non necessiterebbe di alcun aggancio al mondo sensibile, corporeo e materiale esteriore al soggetto conoscente. In altre parole, Saussure cadrebbe nel medesimo errore dell'idealismo soggettivo che faceva corrispondere l'oggetto del pensiero con l'oggetto reale.

f) Thảo critica la nozione di valore secondo lo schema della critica all'economia politica volgare sviluppata da Marx nel *Capitale* (Marx 1867): "II est clair que la conception de la valeur économique exposée ici par l'auteur [Saussure], est précisément celle de *l'économie politique vulgaire*" (Thảo 1974: 42). Tra i lettori del CLG, ci sono stati alcuni che hanno provato a individuare le fonti economiche della nozione di valore (Koerner 1973: 68; cf. anche Sljusareva 1980: 541; Ponzio 2005: 2; Ponzio 2015; Joseph 2014). De Mauro (2011: n. 165) aveva già sottolineato che Saussure era a conoscenza dei dibattiti in economia politica. E come Ponzio (2005) ha ricordato, Saussure condivideva con i teorici del marginalismo e dell'economia neoclassica numerosi principi metodologici. Ma ci sono stati anche studiosi che hanno contestato questo approccio (Godel 1957: 235) ed altri che invece si sono interessati alle fonti eminentemente linguistiche della nozione (Auroux 1985: 295; cf. anche Swiggers 1982: 329; Haßler 2007).

Thảo è tra quanti riconducono la nozione saussuriana di valore alla sua origine economica. E, come altri autori marxisti, era affascinato dalla comparazione tra linguaggio e economia (Lefebvre 1966; Goux 1968; cf. anche Schaff 1968: 207; Baudrillard 1972; Latouche 1973; Bourdieu 1977; Rossi-Landi 2016: 180–181; Rossi-Landi 2003 [1968]; 1977 [1974]). L'obiettivo di Thảo è, tuttavia, diverso da quello di altri marxisti; Thảo vuole mostrare l'identità del valore e della significazione nel CLG (159–160). Anche altri autori (Malmkjær 1991: 437; Bright 1992: III, 406; Bouquet 1992: 91; Bouquet 1997: 317; Rastier 2002) difendono l'idea che ci sarebbe una coincidenza dei due concetti, di significazione e valore. Va, inoltre, sottolineato che il dibattito a questo proposito è stato recentemente riaperto a seguito della pubblicazione degli *Ecrits de linguistique générale* (ELG) nel 2002.

Assimilato al valore economico, il valore linguistico è al punto di convergenza della relazione significato-significante (nei termini del CLG: parola-

### Jacopo D'Alonzo

idea) e della relazione tra i segni (parole). Su questa base Saussure introdusse la distinzione tra il valore e la significazione: il valore di una parola dipende dalla relazione di comparazione con parole differenti, mentre la significazione verte sulla relazione di scambio di una parola con un concetto. Ma Thảo (1974: 42) aggiunge che non sarebbe possibile spiegare come i segni differiscano tra di loro senza fare appello a una significazione che sarebbe già presente.

Tuttavia, per rendere pienamente ragione della natura della significazione, si dovrebbe tenere conto anche del valore di quel segno all'interno del sistema. Thảo descrive dunque una concezione *dizionariale* del valore: per definire il valore di un segno, lo si deve definire per mezzo di altri segni del sistema. Il fatto che Saussure abbia preso come esempio la serie *craindre*, *redouter* e *avoir peur* aumenta in maniera esponenziale le difficoltà connesse alla definizione di ciò che è la significazione. A partire da tale concezione, come ha mostrato Eco (1984: 74), ci si ritrova immediatamente in un circolo vizioso: il valore presuppone la significazione e la significazione il valore. Anche i lettori più recenti di Saussure hanno messo in evidenza il medesimo problema (Joseph 2004: 67).<sup>2</sup> Come si vedrà nella conclusione, la soluzione offerta da Thảo sarà quella di introdurre dei segni fondamentali che rendano conto tanto della significazione che del valore dei segni arbitrari.

### **4 Conclusioni**

Ciò intorno a cui ruotano le analisi di Thảo è l'idea che la significazione che caratterizza i segni di cui si occupa la semiologia dialettica è la condizione *sine qua non* della significazione arbitraria:

il [le système des signes motivés] présente directement dans l'intuition sensible le contenu de signification, auquel le second [le système des signes arbitraires] donne une expression conventionnelle, formellement plus distincte, pour le développer sur le plan discursif (Thảo 1974: 40).

Certamente Thảo sembra pensare l'arbitrarietà più in termini di relazione convenzionale tra il significato e il significante che in termini di non-motivazione. La

<sup>2</sup>Tuttavia, non bisogna dimenticare che un limite al principio dell'arbitrarietà assoluta, da cui dipende la nozione di valore, era già stato indicato da Saussure stesso attraverso il fenomeno dell'arbitrarietà relativa (cf. Saussure 1995: 181–182).

### 12 Per una semiologia materialista e dialettica

convenzione deve, infatti, presupporre uno strato comunicazionale, cognitivo e sociale preesistente. Una tale osservazione potrà sembrare banale, tenuto conto della riflessione filosofica sul linguaggio dal *Cratilo* di Platone in poi. Tuttavia, l'importanza della tesi di Thảo risiede altrove: egli vuole chiarire la relazione genetica che giustifica e fonda i sistemi di segni convenzionali (e le terminologie disciplinari). In altre parole, il valore saussuriano è traccia di un meccanismo cognitivo che supporta pratiche linguistiche molto sofisticate come la capacità di distinguere e definire mutualmente i termini impiegati.

Esiste dunque un *continuum* di sistemi di segni che si sono sedimentati in seno a una comunità e si sviluppano uno dall'altro. Seguendo in questo Karl Marx (1818–1883) e Friedrich Engels (1820–1895) (cf. Marx & Engels 1958: 26, 30–31, 331), Thảo chiama l'insieme delle pratiche linguistiche che fanno parte delle attività pratiche e sociali *langage de la vie réelle* (Sprache des wirklichen Lebens). Thảo rifiuta, o riformula completamente, la nozione husserliana di esperienza antepredicativa nella misura in cui il linguaggio della vita reale precede e fonda la coscienza individuale: esso è un sistema semiologico che serve a organizzare la produzione e altri aspetti della vita pratica anziché un insieme di espressioni linguistiche arbitrarie.

Allo stesso tempo, il riferimento al mondo reale eviterà di trovarsi rinchiusi nel mondo dei segni – come avverrebbe in una semiologia che non teorizzi la referenza.<sup>3</sup> All'interno della vita pratica, è necessario che ogni discorso sia essenzialmente un dire qualche cosa a proposito di qualche cosa. A tal riguardo, Thảo afferma che ogni discorso concernente il mondo necessita dell'esecuzione di un gesto di indicazione più o meno abbozzato. Da qui la conclusione secondo la quale i sistemi di segni arbitrari poggino, in ultima istanza, sul gesto d'indicazione.

Thảo riconosce la debolezza semantica di questo genere di segni, che Eco (1984: 48) aveva già sottolineato. Il gesto di indicazione deve essere rinforzato da altre espressioni aventi funzione di metasemiotica, che aggiungano informazione all'indicazione. Nondimeno Thảo giustifica la necessità del gesto di indicazione perché esso radicherebbe il discorso nello spazio-tempo così da determinare la referenza e far funzionare un minimo di differenziazione tra i segni:

[…] tous les autres signes qu'on pourra ajouter, fonctionneront non pas comme "interprétants" du signe de l'indication, mais comme moyens de présenter les propriétés particulières de l'objet indiqué, ce qui est tout à fait différent. Le signe de l'indication signifie simplement qu'il s'agit de cet objet

<sup>3</sup>Bisogna ricordare, tuttavia, che non è mancato chi ha ammesso l'esistenza di una cripto-teoria della referenza in Saussure (Arrivé 2007; Bouquet 1992).

### Jacopo D'Alonzo

même: le "*ceci*" comme réalité objective donnée à l'intuition sensible, et *rien de plus*. (Thảo 1973: 62)

Si capisce allora cosa conduca Thảo a limitare la validità della nozione saussuriana del valore. La significazione non può essere ridotta alla relazione differenziale tra segni, poiché il valore saussuriano è qualche cosa che si aggiunge a una significazione preesistente nata sul terreno delle interazioni reali con l'ambiente sociale e fisico.

Un bilancio storico-epistemologico della lettura che Thảo offre del CLG non può ignorare un paradosso: benché Thảo dichiari di essere ormai lontano dalla fenomenologia, egli sembra applicare alla lettura strutturalista del CLG quell'approccio fenomenologico-genetico che era stato al centro del dibattito filosofico francese degli anni Quaranta e Cinquanta (specialmente in Sartre e Merleau-Ponty). Thảo non è certo interessato a mettere in campo un'indagine filologica del CLG, bensì a opporre allo strutturalismo generalizzato degli anni Sessanta e al primato della sincronia un'indagine che mostri il primato filogenetico, diacronico e assiologico della genesi sulle strutture. Questo obiettivo incontra le preoccupazioni di una certa tradizione marxista attenta a ricondurre le costruzioni ideali alla loro origine materiale – concezione non lontana dalla *Lebenswelt* husserliana orientata ora però verso le condizioni di produzione e riproduzione sociale. Allo strutturalismo marxista di Louis Althusser (1918–1990), Thảo contrappone dunque un marxismo attento alle esigenze metodologiche che erano emerse nel dibattito filosofico sulla fenomenologia nell'immediato dopoguerra.

### **Riferimenti bibliografici**


12 Per una semiologia materialista e dialettica


### Jacopo D'Alonzo


12 Per una semiologia materialista e dialettica

mo accesso 5 gennaio 2018). Paris: Société d'Histoire et d'Epistémologie des Sciences du Langage. https://hal.archives-ouvertes.fr/hal-01311898.


Aarts, Bas, 68 Adams, William Yewdale, 145 Alverson, Hoyt, 62 Anderson, John M., 25, 27, 37–39 Andor, József, 37 Anonymous, 120 Ansaldo, Umberto, 118 Arbib, Michael, 29 Aristotle, 33 Arnauld, Antoine, 18, 54 Arrivé, Michel, 197 Ascoli, Grazadio Isaia, 165 Attneave, Fred, 28 Aujac, Germaine, 8 Aurora, Simone, 189 Auroux, Sylvain, 54, 56, 63, 91, 131, 161, 195 Baale, L. M., 91 Baars, Bernard J., 29 Bailly, Anatole, 88 Baker, Philip, 118 Bally, Charles, 98 Bardovi-Harlig, Kathleen, 105 Barnes, Jonathan, 5 Barsalou, Lawrence W., 40 Barwick, Karl, 4–6, 8, 10 Basedow, Herbert, 150, 151 Baudrillard, Jean, 195 Beauzée, Nicolas, 25, 54, 55 Benveniste, Émile, 178, 183–185 Bergounioux, Gabriel, 159

Berlin, Brent, 30 Bertera, Bartolommeo Antonio, 136 Bickerton, Derek, 118 Biezunski, Michel, 63 Billroth, Johann G. F., 70 Bisconti, Valentina, 83, 91 Bod, Rens, 52 Böhmer, Eduard, 165 Bolinger, Dwight, 32, 99, 105, 148 Bolton, Kingsley, 119 Boone, Annie, 59 Boroditsky, Lera, 40, 62 Bottineau, Didier, 57, 59 Bouquet, Simon, 195, 197 Bourdieu, Pierre, 195 Bousfield, Weshton A., 28 Boyer, Henri, 160 Branca-Rosoff, Sonia, 91 Bransford, John D., 28 Bréal, Michel, 88, 89, 91 Brereton, Joel P., 180 Bright, William, 195 Brittain, Richard Coulter, 67, 72 Brock, Peggy, 144 Broudic, Fañch, 160 Brugman, Claudia, 33 Brun-Trigaud, Guylaine, 161 Brunot, Ferdinand, 160 Bruns, Beate, 91 Buchheit, Peter, 149 Bursill-Hall, Geoffrey L., 18

Burtt, Andrew, 73 Campbell, Gordon, 10 Carr, Gerald F., 99 Casasanto, Daniel, 62 Cassirer, Ernst, 29 Caveing, Maurice, 191 Chafe, Wallace L., 22 Chalozin-Dovrat, Lin, 53, 54 Chandler, Z. M., 73 Chauviré, Christiane, 68 Chervel, André, 130, 131, 133 Chevalier, Jean-Claude, 90, 131, 159 Chin, James K., 121 Chiss, Jean-Louis, 192 Chompré, Étienne-Maurice, 132, 136 Christy, T. Craig, 99, 104, 115 Clark, Herbert H., 30 Clark, Stephen Watkins, 68, 71, 72, 76, 77 Clarke, David D., 91 Colombat, Bernard, 18, 54, 130–133, 136–138 Condillac, Etienne Bonnot de, 19, 33 Cordereix, Pascal, 164 Coseriu, Eugenio, 89 Courbon, Bruno, 89 Coursil, Jacques, 193 Crick, Malcolm, 149 Crow, Carl, 119 Curtius, Georg, 20 Cusimano, Christophe, 91 D'Alembert, Jean, 54 Darmesteter, Arsène, 27, 88 Darrigol, Jean-Pierre, 39 Dascal, Marcelo, 19

De Mauro, Tullio, 193, 195 De Palo, Marina, 91, 189 Deeters, Gerhard, 148 Delesalle, Simone, 88, 90, 91 Della Casa, Adriana, 12 Derradji, Arezeki, 91 Desclés, Jean-Pierre, 39 Desmet, Piet, 90, 159 Despautère, Jean, 20 Di Benedetto, Vincenzo, 5 Diderot, Denis, 54 Domergue, Urbain, 133 Dosse, François, 192 Downing, C. Toogood, 123 Duer, John K., 122 Eco, Umberto, 192, 196, 197 Eiszele, Aileen, 90 Ekman, Paul, 30 Elspaß, Stephan, 117 Encrevé, Pierre, 159 Engels, Friedrich, 197 Erman, Britt, 102 Errington, Joseph, 144, 153 Eylmann, Erhard, 152 Fagard, Benjamin, 41 Federici, Silvia, 191 Feron, Alexandre, 190 Fillmore, Charles J., 24, 37 Firth, John R., 90 Firth, John Rupert, 97 Fodor, Jerry A., 26 Formigari, Lia, 19 Fortis, Jean-Michel, 17, 19, 22, 24, 26, 27, 29, 32, 33, 37, 42, 52 Fournier, Jean-Marie, 53, 56 Franks, Jeffrey, 28 Frédéric, François, 191

Dawson, Hannah, 19 de Jonge, Casper, 6, 8, 9

Fuchs, Catherine, 91 Furetière, Antoine, 86 Gardy, Philippe, 160 Geeraerts, Dirk, 41 Gellner, Ernest, 153 Giao, Nguyên Ngọc, 190 Gilliéron, Jules, 162, 167, 168 Gingrich, Andre, 149, 150 Girardin, Chantal, 91 Giustiniani, Vito R., 145 Godel, Robert, 195 Goldsmith, John A., 22 Gonda, Jan, 180 Goulier, 133 Goux, Jean-Joseph, 195 Granger, Sylviane, 105, 106 Gries, Stefan Th., 99 Grotefend, Friedrich August, 16 Gruber, Jeffrey S., 34, 35 Guillaume, Gustave, 56–58 Hagen, Bernard, 148 Hailwood, Mark, 117 Halévy, Joseph, 90 Hao, Yen-p'ing, 121 Harmstorf, Ian, 151 Harris, James, 18, 33 Harris, Randy Allen, 22 Haspelmath, Martin, 62 Haßler, Gerda, 195 Haudricourt, André, 191 Heid, Ulrich, 101 Heine, Bernd, 41 Hémery, Daniel, 190 Hernández-Campoy, Juan Manuel, 116 Herzferd, Ernst, 180 Hockett, Charles F., 22

Holm, John, 118 Holt, Robert R., 29 Holzweissig, Friedrich W., 20 Honeck, Richard P., 27 Howarth, Peter, 106 Huck, Geoffrey J., 22 Hull, Isabel V., 153 Hunter, William C., 119, 121 Jackendoff, Ray S., 35, 36, 100 Jakimovska, Svetlana, 91 Janko, Richard, 9, 10 Jaubert, Hippolyte, 161 Jewell, Frederick S., 74 Johnson, Mark, 27, 41, 60 Johnson-Laird, Philip N., 16, 29 Joly, André, 57, 59 Jones, Christopher P., 10 Joseph, John, 195, 196 Kahane, Sylvain, 69 Karwoski, Theodore F., 31 Katz, Jerrold J., 26, 28 Kay, Paul, 30 Kellogg, Brainerd, 70, 72, 74–76, 78 Kelly, Louis G., 18 Kempe, Hermann, 146, 147, 149 Kenny, Anna, 144, 146 Kim, Ji Hyon, 106 Kim, Soo Hyon, 106 Klaatsch, Hermann, 151 Kleiber, Georges, 27 Kluckhohn, Clyde, 145 Kneebone, Heidi, 147 Koerner, E. F. K., 178, 195 Köpping, Klaus-Peter, 149 Kremnitz, Georg, 160 Kuryłowicz, Jerzy, 19

Labov, William, 115 Lakoff, George, 24, 27, 60 Laks, Bernard, 193 Lallot, Jean, 7 Lamb, Sydney M., 22 Lancelot, Claude, 18, 20, 54, 132 Langacker, Ronald W., 40, 41, 61 Larrivée, Pierre, 91 Latouche, Serge, 195 Latz, Peter, 146 Le Guern, Michel, 54 Le Roux, Philibert-Joseph, 87 Leclerc, Jean, 87 Lefebvre, Henri, 192, 195 Leibniz, Gottfried Wilhelm, 27, 33 Lenneberg, Eric H., 30 Lentz, Wolfgang, 180 Léon, Jaqueline, 192 Leonhardi, Moritz Freiherr von, 147 Lepschy, Guilio, 192 Lerouge, Marie-José, 57 Lévi-Strauss, Claude, 176 Lewis, Michael, 104 Lhomond, Charles-François, 131, 133–138 Li, Michelle, 118, 122–124 Liberman, Kenneth, 147 Lieber, Rochelle, 91 Lighthall, G. E., 73 Lincoln, Bruce, 177, 180 Lindner, Susan J., 32, 33 Littré, Émile, 88 Locke, John, 19 Lockwood, David G., 23 Lounsbury, Floyd G., 22 Lyons, John, 16, 38 MacKenzie, Ian, 104, 105 Makkai, Adam, 99

Malamoud, Charles, 176 Malmkjær, Kirsten, 195 Marett, Robert R., 145 Marty, Anton, 19 Marx, Karl, 195, 197 Massin, Benoit, 145, 149, 150 Mathews, R. H., 144 Matthews, Peter Hugoe, 22 Matthews, Stephen, 118 Mayrhofer, Manfred, 180 Mazaleyrat, Hélène, 91 Mazhuga, Vladimir, 12 Mazziotta, Nicolas, 68, 69, 71, 74, 77 McElvenny, James, 152 McGregor, Russell, 151 McMahon, April, 68 Meillet, Antoine, 177, 178, 180–183 Melʹčuk, Igor, 75 Metzler, Jaqueline, 29 Meunier, Fanny, 105, 106 Meyer, Paul, 88 Michel, Louis C., 87 Michelsen, Conrad, 16 Miller, George A., 16, 29 Milroy, James, 117 Monod-Becquelin, Aurore, 177, 178 Monteath, Peter, 144, 150, 152 Moon, Rosamund, 106 Moore, David, 144–146 Mouret, François, 75 Mowrer, O. Hobart, 29 Mulvaney, John, 147 Münzer, Friedrich, 11 Murray, Stephen O., 22 Nagy, William Elmer, 21, 31 Neisser, Ulric, 28 Nekes, P. Hermann, 153 Nerlich, Brigitte, 88, 90, 91

Newmeyer, Frederick, 22 Nielsen, Janet, 23 Nobbs, Chris, 146 Nobels, Judith Maria Petrus, 117 Nuchelmans, Gabriel, 18 Nunberg, Geoffrey D., 27 Nyckees, Vincent, 84 Oesterreicher, Wulf, 117 Oetteking, Bruno, 151 Osgood, Charles E., 31 Paolucci, Claudio, 193 Pauly, Émilie, 91 Pawley, Andrew, 101 Peeters, Bert, 33, 90 Peirce, Charles Sanders, 68 Pike, Kenneth L., 148 Piron, Sophie, 91, 130 Planert, Wilhelm, 152, 153 Poland, Wilhelm Georg Friedrich, 146 Polguère, Alain, 75 Ponzio, Augusto, 195 Pottier, Bernard, 25, 33 Preuss, Konrad Theodor, 150 Puech, Christian, 57, 192 Pugach, Sara, 152 Putnam, Hilary, 28 Rastier, François, 193, 195 Reed, Alonzo, 70, 72, 74–76, 78 Régis, Pierre Silvain, 87 Restaut, Pierre, 132, 137 Rice, Sally, 33 Richards, Jack C., 98, 99, 104 Richelet, César-Pierre, 87 Ricoeur, Paul, 90 Ríos Castaño, Victoria, 144–146

Romaine, Suzanne, 116 Rossi-Landi, Ferruccio, 195 Rousselot, Pierre-Jean, 162, 166–168 Royon, M., 133, 136 Rumpel, Theodor, 16, 20 Russi, Cinzia, 116 Rutten, Gijsbert, 117 Sandra, Dominiek, 33 Saussure, Ferdinand de, 99, 194, 196 Savatovsky, Dan, 57 Schaff, Adam, 195 Schilling, Natalie, 116 Schleicher, August, 152 Schmidt, Hanns-Peter, 180 Schmidt, P. Wilhelm, 147 Schmidt, Richard W., 98, 99, 104 Sève, Lucien, 192 Shepard, Roger N., 29 Siblot, Paul, 91 Sinclair, John, 99 Sljusareva, Natalija A., 195 Smith, Woodruff D., 145 Sofia, Estanislao, 193 Sorba, Julie, 91 Spencer, Baldwin, 150 Štekauer, Pavol, 91 Stewart, William, 118 Strehlow, Carl, 146 Strehlow, John, 147, 153 Stringer, David, 105 Stubbs, Michael, 101 Surcin, Sylvain, 91 Swiggers, Pierre, 90, 195 Syder, Frances Hodgetts, 101 Talmy, Leonard, 24 Thảo, Trần Đức, 190–196, 198

Thao, Trinh Van, 190

Thibault, André, 91 Thomas, Antoine, 161 Thomas, Martin, 144 Thomassin, P. L., 87 Thompson, E. P., 117 Tiffany, Osmond, 121 Tilden, Bryant P., 123 Tong, Ting-kü, 119, 123 Traugott, Elizabeth Closs, 41, 53 Tyng, Charles, 121 Ullmann, Stephen, 90 Urban, Wilbur M., 29 Urry, James, 150 Valette, Mathieu, 59 Valin, Roch, 56 Van der Wal, Marijke J., 117 Van Dyke, Paul Arthur, 121 Viard, Nicolas-Antoine, 132, 136 Victorri, Bernard, 91 Vitale, Maria Teresa, 3, 4, 6 Vittrant, Alice, 24, 32 Völker, Harriet, 146 Vossler, Karl, 145 Wailly, François de, 132 Warren, Beatrice, 99, 102 Watkins, Calvert, 179 Weinreich, Uriel, 26 Wiseman, Cardinal, 87 Wood, David, 99, 104, 105 Wray, Alison, 100 Wüllner, Franz, 19 Wundt, Wilhelm, 21

Zhang, Zhenjiang, 121 Zimmerman, Andrew, 145, 149, 150 Zogbaum, Heidi, 151

Accidens / accidentia, accidents, vii, 3–13, 133, 135, 136, 139 adjectives, 133–139 adjuncts, 72 Aluridja, 151 American linguistics, 30, 32, 51, 59 American structuralism, 22 anagram, 99, 107 analysis, linguistic, 179 analytic expressions, 98<sup>1</sup> anthropology, x, 145, 149, 151, 153, 154, 170, 171, 175–177, 182, 184, 186 applied linguistics, 100 Aranda (Arrarnta, Arrernte), 143, 146–151, 153 arbitrariness, 192, 194 artificial intelligence, 106 aspect, 25, 39, 52, 59 Avesta, 179, 181–183 Avestic, 177, 179, 180, 183, 186 "Bushman", 153 Cantonese, ix, 118, 119, 122–124 case (grammatical case, case / thematic role), viii, x, 8, 16–21, 24, 25, 27, 34–39, 41, 133– 135, 138 Chinese, 87, 118–121, 124 Chinese Pidgin English, 116, 120, 122, 124, 125

cognitive linguistics, viii, ix, 16, 20, 24, 26–28, 31, 33, 40, 41, 52, 53, 57, 59, 60, 62, 63 cognitive science, 21, 27, 29, 30, 33, 53, 60, 63 collocations, collostructions, 98, 99 compounds, 101 computational linguistics, 100 concepts, 83, 84, 84<sup>2</sup> , 85, 85<sup>5</sup> , 89–91, 91<sup>13</sup> conceptual metaphor, 27, 40, 41, 60 conceptualization, 16, 40, 56, 57, 60, 74, 84, 137 construction (morphosyntax), ix, 38, 41, 73, 79, 98–100, 107 coordination, 75 corpus, corpus linguistics, 23, 24, 100–102, 116 delatinization, 136, 137, 139 designation, 4, 6, 9, 136, 176, 184 diachrony, ix,19, 25, 40, 52, 59, 60, 89, 116, 161, 170 *see also* synchrony diagram, viii, 20, 23, 25, 28, 39, 40, 58, 59, 67–79 dialects, 161, 171 dictionary, 83, 86, 87, 91, 101, 146, 148, 161 discourse analysis, 98, 100 Diyari (Dieri), 143, 147

Dutch, 117, 118, 120 English, 23, 25, 32, 37, 60, 98, 101, 102, 105, 106, 117–124, 146, 152 English grammar, 68–72 Epicureans, Epicurean philosophy, 9, 10, 13 ethno-semantics, 22 ethnography, ethnology, 144, 147, 149, 150, 153, 161, 170, 175 fixed expressions, 97–108 formulaic language, 100, 179 French grammar, 129–139 French linguistics, 19, 56, 58, 68, 98 *Geisteswissenschaften*, 145 general grammar (Grammaire générale), 54, 62 German linguistics and philology,15, 16, 19–21, 143–146, 148, 151– 153 German philosophy of language, 145 grammar comparative grammar and philology, 148, 154, 161, 175, 185 construction grammar, ix, 99 generative grammar, ix, 22–27, 32, 35, 37, 39, 104, 105 Space grammar, 40, 61 speculative grammar, Modists, 17, 18 stratificational grammar, 23, 25 transformational grammar, 22, 24, 27, 30 *see also* general grammar, cognitive linguistics grammaticalization, 40, 41, 104, 107

Greek (Ancient), vii, 3–13, 17, 19–21, 55, 146, 180 Greek (Modern), 162 historical linguistics, 177 "Hottentot", 153 humanism, 145 ideographs, 90 idioms, 97–100 Jaunde (Ewondo), 153 language acquisition, 59 language ideology, 144, 145, 154 Latin, vii, x, 3, 13, 18–21, 58, 129–139, 146, 161, 162, 164, 168, 171 lexical meaning, 83, 84, 91 lexicalization, 104 lexicography, 84, 84<sup>3</sup> , 86, 87<sup>7</sup> , 88, 100 lexicology, 100 lexicon, 7, 9, 11, 31, 63, 99, 100, 104, 107, 161, 163, 164, 169, 170 linguistic variation, 116, 118, 125 localism, 15–42, 52, 52<sup>1</sup> locutions, 98 Loritja (Luritja), 146–148 Lutheran missionaries, 143, 146, 151 machine translation, 100 Malay, 118 Marxism, 190 meaning, 84, 86–90 mechanics, 52 metaphysics, 53 metascientific orientations, 143, 144 modifiers, 72 mood, 52, 58 n-grams, 98, 99

Nama, 153 natural language processing, 100 *Naturwissenschaften*, 145 object (grammatical), 72, 75, 77 Old Persian, 179 oral language, 160, 164 orthography, 132, 137, 139 parepómenon / parepómena (παρεπόμενον, παρεπόμενα), 3– 7, 10, 13 *see also* accidens / accidentia participle, 133, 134, 137, 138 pedagogy, school, 7 pedagogy, teaching, ix, x, 18, 20, 54, 104, 106, 130, 131, 136 phenomenology, 189, 190 phraseology, 98 pidgins, 118 polysemy, 83, 84, 84<sup>2</sup> , 84<sup>3</sup> , 85, 85<sup>5</sup> , 86, 87, 87<sup>8</sup> , 88–90, 90<sup>12</sup> , 91, 91<sup>13</sup> , 92, 93 Portuguese, 118, 120, 121 positivism, 145, 153 pragmatics, 100 predicate, 4, 20, 21, 26, 31, 35–38, 71– 79 prefabricated phrases, 98, 99 prepositions, 60, 72, 116, 122, 124, 125 psycholinguistics, 100 psychomechanics, 53 real life, 190 rhetoric, 7, 8, 10, 11 Sanskrit, 19–21, 180 scientification, 51

second language, 97, 100 semantics, viii, xi, 15–17, 21–27, 30, 32–36, 39, 42, 60, 62, 63, 83– 85, 87–92, 100, 122, 124, 170, 179, 185, 186, 195 semiology, 178, 189–194, 196, 197 sentence, 68 signification, 176 signifier, signified, 191 sociolinguistics, 177 sophists, 3, 4, 6, 7 space, viii, 17, 25, 29, 30, 34, 36, 40, 52, 53, 55–58, 60–64, 73,183, 184, 197 spatialization, spatial cognition, 52, 53 speech act, 100 standardization, standard varieties, 117, 118, 125 Stoics, Stoic philosophy, 4–6, 8–10, 13 structuralism, xii, 22, 39, 57, 60, 189– 192, 198 subject (grammatical), 71–74, 77 subordination, 77 sumbebēkós / sumbebēkóta (συμβεβηκός, συμβεβηκότα), 3–6, 8–11, 13 *see also* accidens / accidentia Swahili, 153 synchrony, 25, 27, 36, 52, 89, 115, 161, 178, 179, 191, 198 syntax, viii, ix, 6, 17, 22, 26, 34, 37, 63, 68, 71, 78, 98, 99,104,105, 107, 122, 124, 129, 131–133, 137–139, 146, 153 synthetic expressions, 98<sup>1</sup>

tense, 18, 25, 39, 52, 54–56, 58, 59, 62, 63, 135, 136 terminology, vii, 3, 6, 9, 11, 34, 52, 59, 63, 72, 84, 92, 97, 131, 138, 197 time, viii, 18, 25, 26, 51–64, 197 transcription, 160, 161, 163, 167, 169 usage, ix, 8, 10, 86, 88, 97–99, 101, 103, 104, 118, 123, 137, 139, 161, 162, 177, 183 value (differential, semantic), 90, 91, 101, 135, 192–198 value (economic), 195 value (phonetic), 165, 166 Vedas, 181, 182 vocabulary (lexical store), 59, 100, 118, 119, 122, 146, 147, 151, 176 vocabulary (theoretical), 3, 5, 6, 8, 9 *see also* terminology

WordNet, 101

## Historical journey in a linguistic archipelago

This volume offers a selection of papers presented during the 14th international conference on the history of the language sciences (ICHoLS XIV, Paris, 2017).

Part I brings together studies dealing with descriptive concepts. First examined is the notion of "accidens" in Latin grammar and its Greek counterparts. Other papers address questions with a strong echo in today's linguistics: localism and its revival in recent semantics and syntax, the origin of the term "polysemy" and its adoption through Bréal, and the difficulties attending the description of prefabs, idioms and other "fixed expressions". This first part also includes studies dealing with representations of linguistic phenomena, whether these concern the treatment of local varieties (so-called *patois*) in French research, or the import and epistemological function of spatial representations in descriptions of linguistic time. Or again, now taking the word "representation" literally, the visual display of grammatical relations, in the form of the first syntactic diagrams.

Part II presents case studies which involve wider concerns, of a social nature: the "from below" approach to the history of Chinese Pidgin English underlines the social roles of speakers and the diversity of speech situations, while the scrutiny of Lhomond's Latin and French textbooks demonstrates the interplay of pedagogical practice, crosslinguistic comparison and descriptive innovation. An overview of early descriptions of Central Australian languages reveals a whole spectrum of humanist to positivist and antihumanist stances during the colonial age. An overarching framework is also at play in the anthropological perspective championed by Meillet, whose socially and culturally oriented semantics is shown to live on in Benveniste. The volume ends with a paper on Trần Đức Thảo, whose work is an original synthesis between phenomenology and Marxist semiology, wielded against the "idealistic" doctrine of Saussure.