# Evidentiality, egophoricity, and engagement

Edited by Henrik Bergqvist Seppo Kittilä

Studies in Diversity Linguistics 30

### Studies in Diversity Linguistics

### Editor: Martin Haspelmath

In this series:


ISSN: 23635568

# Evidentiality, egophoricity, and engagement

Edited by

Henrik Bergqvist Seppo Kittilä

Bergqvist, Henrik & Seppo Kittilä (eds.). 2020. *Evidentiality, egophoricity, and engagement* (Studies in Diversity Linguistics 30). Berlin: Language Science Press.

This title can be downloaded at: http://langsci-press.org/catalog/book/261 © 2020, the authors Published under the Creative Commons Attribution 4.0 Licence (CC BY 4.0): http://creativecommons.org/licenses/by/4.0/ Indexed in EBSCO ISBN: 978-3-96110-269-3 (Digital) 978-3-96110-270-9 (Hardcover)

ISSN: 2363-5568 DOI: 10.5281/zenodo.3968344 Source code available from www.github.com/langsci/261 Collaborative reading: paperhive.org/documents/remote?type=langsci&id=261

Cover and concept of design: Ulrike Harbort Typesetting: Carla Bombi, Elena Moser, Sebastian Nordhoff Proofreading: Anca Gâţă, Andreas Hölzl, Annie Zaenen, Benjamin Brosig, Eva Schultze-Berndt, Ezekiel Bolaji, Gerald Delahunty, Jeroen van de Weijer, Lachlan Mackenzie, Sean Stalley, Tom Bossuyt, Yvonne Treis Fonts: Libertinus, Arimo, DejaVu Sans Mono Typesetting software: XƎLATEX

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

Storage and cataloguing done by FU Berlin

# **Contents**


# **Chapter 1**

# **Epistemic perspectives: Evidentiality, egophoricity, and engagement**

Henrik Bergqvist Stockholm University

Seppo Kittilä University of Helsinki

### **1 Introduction**

In the last decade, there has been a surge in output on various forms of epistemic marking in language, including (epistemic) modality, evidentiality, mirativity, egophoricity, and engagement.<sup>1</sup> Some of these terms are better known than others.<sup>2</sup> To begin with, epistemic modality has a long research tradition stemming

<sup>2</sup>By using terms like "evidentiality" and "egophoricity", we refer to meaning domains that signal how knowledge about events can be qualified in different ways. Usually, such domain labels come with a definition that is found in the seminal literature dealing with a given domain (e.g. Palmer 2001, for modality), but this is not always the case. Definitional criteria for comparable systems and forms are often contested and in a relatively young field such as the present one, debates concerning what counts as defining (semantic) features of a certain domain, are especially fierce. In this volume, a term like "evidentiality" is regarded as constituting a linguistic category in some languages, but consequently also refers to an epistemic domain that expresses different "modes of access" (Plungian 2010) with respect to how knowledge about events may be acquired.

Henrik Bergqvist & Seppo Kittilä. 2020. Epistemic perspectives: Evidentiality, egophoricity, and engagement. In Henrik Bergqvist & Seppo Kittilä (eds.), *Evidentiality, egophoricity, and engagement*, 1–21. Berlin: Language Science Press. DOI: 10.5281/ zenodo.3975795

<sup>1</sup>Edited volumes that deserve mention are: Aikhenvald & Dixon (2003; 2014) on evidentiality, modality, and expressions of knowing in grammar more broadly; Gawne & Hill (2017) on evidentiality in Tibetan languages and Floyd et al. (2018) on egophoricity. The list of journal articles on epistemic marking in grammar is (very) long, but we may note Evans et al. (2017a, 2017b) on engagement, Bergqvist & Knuchel (2017) on egophoricity, and San Roque et al. (2017) on evidentials and interrogativity.

### Henrik Bergqvist & Seppo Kittilä

from philosophy and qualifies an utterance in terms of possibility and probability, ranging from speculation to high certainty. The first monograph-length treatment of epistemic modality from a cross-linguistic perspective is Palmer (1986; 2001). At about the same time, research on the related category, evidentiality, also began to gain momentum. Evidentiality signals the source of information that a speaker has for an utterance. It is often sub-divided into direct and indirect evidentials where direct evidentials target the (direct) perception of the speaker, signaling sensory access (visual, auditory) to a discourse object. Indirect evidentials express other types of cognitive access, such as inference, assumption, and hearsay, and may be differentiated by how directly accessible a given type of evidence is. For example, inference and assumption are both based on the speaker's observation of a state-of-affairs that is not directly related to the event his/her claim is based on (e.g., we may infer that someone has left if that person's coat is gone). Inference is usually based on direct sensory perception, while assumption is often based on our general knowledge of the world, thus differing in type of indirect access (see e.g. Willett 1988). Aikhenvald (2004) is the first typological treatment of evidentiality, but it is Chafe & Nichols' (1986) seminal volume that is commonly regarded as the first work to investigate evidentiality from a cross-linguistic perspective.

Mirativity is regarded as separate from evidentiality by some (e.g. DeLancey 1997), but the ultimate definition of this category (and even its existence) is still under debate.<sup>3</sup>

Miratives signal new/non-assimilated knowledge and has often been said to convey the surprise of the speaker (DeLancey 1997; cf. Aikhenvald 2014). Surprise as a defining semantic component of mirativity has increasingly been rejected, however, whereas the signaling of new/non-assimilated information appears to be more widely accepted (but see Hill 2012 for arguments against the category of mirativity). In some languages (e.g. Turkish and Finnish), mirativity is found in specific uses of inferential evidential morphemes, while in other languages (such as Hare), there is a morpheme whose primary, or even only, function, is to signal mirativity (DeLancey 1997).

Egophoricity signals the epistemic authority of a speaker or addressee (speechact participant) subject to his/her involvement in a talked-about event (Bergqvist 2018a; Bergqvist & Kittilä 2017; cf. Hargreaves 2005). This dialogical property of the egophoric marker has produced a pattern where egophoric markers occur in statements with first person subjects and in questions with second person

<sup>3</sup> See e.g. the debate in *Linguistic Typology*, involving, among others, Hill (2012), DeLancey (2012), and Hengeveld & Olbertz (2012).

### 1 Epistemic perspectives

subjects: *I am leaving*. [EGO] vs. *Are you leaving?* [EGO]. This functional overlap with person marking has led some to regard egophoric marking as a form of person marking/agreement, but it is clear from diachronic, distributional, and semantic criteria that egophoric marking is distinct from person marking (see Bergqvist & Kittilä 2017, for a discussion). Egophoricity, as a term to designate this kind of epistemic marking, was proposed by Tournadre (1996: 201), but the forms he discussed using the term had already been described by Austin Hale (1980) for Kathmandu Newar, then called "conjunct" (contrasted to "disjunct"). Subsequent research on the phenomenon called into question the usefulness of the label "conjunct/disjunct" and various proposals were put forth to replace it (e.g. "assertors involvement", Creissels 2008; "congruent/non-congruent", Dickinson 2000). Floyd et al. (2018) opts for the term "egophoricity" in providing a comprehensive overview of the phenomenon.

"Engagement", finally, is a term that has had some currency in French linguistics (e.g. Desclés 2009; Guentchéva 2011) and in discourse studies (Hyland 2005), but in the work of Evans et al. (2017a, 2017b) it is used as a label for a kind of epistemic marking separate from the categories outlined above. Engagement targets the epistemic perspectives of the speech-act participants, signaling differences in the distribution of knowledge and/or attention between the speaker and the addressee. As such, it specifies whether information is shared, or exclusive to one of the speech-act participants. This contrast is exemplified with data from Southern Nambikwara (Kroeker 2001: 63–64 [our adjusted glossing and translation]):

	- a. *wa3ko3n-a1-Ø-wa2.* work-1-PRS-EXCL-IMPF 'I am working.'
	- b. *wa3ko3n-a1-ti2.tu3-wa2* work-1-SHRD-IMPF '(You and I see that) I am working'

In (1), the semantic contrast between exclusive and shared knowledge is signaled by a zero morpheme and the -*ti2.tu3*-suffix, respectively.<sup>4</sup>

While the translation suggests a visual mode of access to go along with such knowledge asymmetry, this is not encoded in said forms, but signaled by means of separate evidential morphology (see Kroeker 2001, for details). Engagement markers are generally underspecified with respect to how knowledge about an

<sup>4</sup>Kroeker's label for what we term engagement, is "verification".

### Henrik Bergqvist & Seppo Kittilä

event is gained, but Evans et al. (2017b) note that engagement markers often combine with evidentials and modals to signal (a)symmetries in terms of accessibility to and belief of some event. An issue that relates to the topic of the present volume in terms of categorical overlap, is whether the assumed distribution of knowledge between the speech-act participants concerns their actual knowledge/non-knowledge, or their rights/non-rights to knowledge (see Grzech 2020 [this volume]). It is possible that this contrast in terms of knowledge access and rights to knowledge is subject to variation across engagement systems. While engagement has yet to be widely accepted as a grammatical category in some languages, arguments for making such a claim are put forth in §3.3, below.

The organization of this introductory chapter is as follows: In §2, we will discuss the functional and semantic overlaps between the discussed categories on a general level. In §3, we will focus on the relation between the discussed categories as resources for expressing epistemic authority, and in §4, we will summarize the main points of the paper and suggest some ideas for future research.

### **2 Functional and semantic overlap between categories**

It is a widely known fact that the abovementioned categories overlap in form, meaning, and function (see Cornillie 2009; de de Haan 1999; *inter alia*). This has been especially noted with respect to modality and evidentiality, where an English modal verb like *must* may be treated as a modal, or an evidential depending on the context of use. Such ambiguity has led some researchers to analyze evidentiality as a subtype of modality (e.g. Palmer 1986; Palmer 2001), whereas others have argued for a strict separation between the two categories (e.g. Aikhenvald 2004).

A comparable overlap has also been observed for evidentiality, mirativity, and egophoricity. With respect to evidentiality and mirativity, an inferential (evidential) form may also serve as a mirative depending on the context of use, as in Turkish (Slobin & Aksu 1982). Evidentiality and egophoricity overlap to the extent that egophoric markers may be part of evidential paradigms and thus contrast with evidential forms. Semantically, some have argued that egophoric marking cannot be regarded as a kind of evidential marking due to the fact that such markers do not signal a source of information as such (e.g. Aikhenvald 2004). Others have argued the opposite and regard egophoric markers as the strongest kind of access that speakers employ to justify a statement (Plungian 2010; San Roque & Loughnane 2012; cf. Boye 2012, evidentiality as "justification"). Egophoric markers have sometimes been analyzed as a kind of mirative marker for Tibetan and

### 1 Epistemic perspectives

Barbacoan languages (DeLancey 1997; Dickinson 2000; but see Curnow 2002, for a critique). Kittilä (forthcoming), shows that egophoric markers have features in common with general knowledge/factual evidentials and that claims of factuality may formally resemble markers of both visual evidence and volitional participation in an event.

Several papers in this volume discuss categorical overlaps of the kind sketched above. We will mention a few of them here. Liljegren notes for the Indo-Aryan language Palula that indirect evidentiality is produced by the contextualized token-use of perfect forms in addition to employing sentence final particles denoting hearsay and inference. The use of aspectual forms and particles to serve evidential functions are a commonly noted phenomenon, which has also been attested for Turkic (Slobin & Aksu 1982), Caucasian (Tatevosov 2001), and Persian languages (Lazard 1996).

Grzech accounts for epistemic markers in a variety of Quechua that previously have been described as evidentials, but which Grzech analyzes as signaling epistemic authority. Given the defining role of epistemic authority in egophoric marking, this constitutes a clear case of conceptual overlap between the two categories where a set of evidential markers (i.e. that are cognate to evidentials in related languages) have developed egophoric semantics. While assigning epistemic authority is implied by the use of direct and indirect evidentials (see §3.1, below), it has become encoded in Tena Kichwa, an Ecuadorian variety of Quechua. Conversely, languages with egophoric marking (see e.g. Tournadre 2008) may imply the cognitive access that a speaker claims for making an assertion, while the speaker's epistemic authority is encoded in the form (see §3.2, below).

For Kalapalo, Basso accounts for an enormously rich variety of epistemic markers with different grammatical status that includes modal and evidential notions, but which also feature participation and the positioning of knowledge between the speech-act participants. In a language like Kalapalo, it may not be fruitful to base categorical distinctions on formal criteria, given their distribution in different parts of the grammar. Rather than a small number of forms serving many functions, Kalapalo has many forms serving overlapping functions, thus constituting an entirely different kind of categorical overlap than the one that e.g. Liljegren reports for Palula.

Henrik Bergqvist & Seppo Kittilä

### **3 The relation between evidentiality, egophoricity, and engagement**

How may we account for the overlaps discussed above? The conceptual overlap between categories is evident from a functionalist perspective, but how this corresponds to their semanto-pragmatic properties and distinct development requires further discussion. With respect to the development of markers belonging to the categories that are the topic of the present volume, it is clear that the presence of pragmatically contingent components of meaning (e.g. inference as a tokenfeature of perfects) makes possible an eventual encoding of this implied feature to become part of the semantics of a form. This conventionalization of implicature (Levinson 2000) is a driving force in grammaticalization processes more generally and we may observe this for epistemic marking, as well (see Bergqvist 2018a). The aim of this section is to account for how implied meaning in the use of certain markers suggests the relation between the investigated categories and the reason for the frequently attested overlaps between categories.

### **3.1 Cognitive access vs. (dis)claiming epistemic authority: Evidentiality**

The notion of evidentiality in language has been equated with providing justification for how information was acquired, i.e. the source of information that a speaker has for making an assertion about an event. Willett (1988) is still a good starting point for an illustration of the basic semantic features and divisions relevant to evidentiality in a cross-linguistic sense.

Figure 1.1: Evidential categories (after Willett 1988)

### 1 Epistemic perspectives

There is an implicit expectation for the speaker to provide the highest form of evidence for an assertion, which produces a hierarchical relation in terms of strength of evidence. Direct evidentials conveying "visual" access (i.e. 'I know this from having seen it') usually constitutes a stronger form of evidence than other sensory access (i.e. 'I know this from having heard/felt/smelt it'). It is, however, important to note that states-of-affairs display variation in this regard. For example, olfactory evidence is more direct and reliable for a claim like 'There must be a gas leak here somewhere', and tactile evidence outranks visual evidence for claims such as 'The water is hot'. In general, though, visual evidence ranks highest, a suggestion that is supported by the use of visual evidential forms to express participatory/factual evidence in some languages (see §3.2, below). For indirect evidentials, it is possible to argue that making an inference based on e.g. visual evidence (i.e. 'I know this because I know/see what caused it') is a stronger form of assertion than one based on report (i.e. 'I know this because someone told me about it'). As stated above, inference and assumption are in principle very similar types of evidence, and both of them are classified as personal and indirect evidence by Plungian (2010: 37). However, these two evidence types differ from each other as regards their directness and reliability; inference is based on a more reliable (usually directly observable) evidence, while assumption is based on our general knowledge of the world (see the definitions offered by Plungian 2010: 37). We can also see that the hierarchical relation between direct and indirect evidence in terms of strength may not always hold, since e.g. "folklore", as a kind of reported evidential (i.e. 'I know this from (our) oral tradition'), may be deemed a very reliable source despite the fact that no member of the speech community has direct access to the events portrayed.

Possibly the most thoroughly investigated aspect of evidentiality is reported speech,<sup>5</sup> which also corresponds to the "simplest" kind of evidential system where reported speech is the only evidential marker. Estonian exemplifies such a system with only a hearsay marker (see Aikhenvald 2004 for details). But despite this emphasis on reported speech in the literature on evidentiality, a narrow conception of evidentiality as a verbal category tends to emphasize the role of direct perception (i.e. visual, auditory) in grammaticalized evidentiality systems. In de Haan's (2013) survey of evidential systems, however, systems that feature both direct and indirect evidentials are much less common than systems with only

<sup>5</sup>This is an impressionistic claim that would be difficult to substantiate statistically given the (by now) vast literature on evidentiality. Reportive evidentials were, however, discussed as early as Jakobson (1957) and has continued to occupy research on European languages as well as more cross-linguistically oriented research (see e.g. Boye 2012).

### Henrik Bergqvist & Seppo Kittilä

indirect evidentials. Evidential systems with only *direct* evidentials are not attested at all in the survey. From this cross-linguistic patterning, we may gather that the grammatical expression of evidentiality is primarily a means to signal indirect access to events, and that this indirect access sometimes is (explicitly) contrasted to direct access. In the default case (i.e. indirect access to an event), evidentiality is not so much a strategy to signal ownership of knowledge through some cognitive channel, but a disclaimer of epistemic authority as a consequence of restricted (i.e. indirect) sensory access (cf. Mushin 2001).

Signaling indirect access to an event in terms of inference, or assumption, does not necessarily mean that there is a restriction present on the sensory access that a speaker has to the talked-about event. In data resulting from the use of an interactive elicitation task developed by Nick Evans and colleagues (described in San Roque et al. 2012), it is clear that explicit, visual representations of people, things, and events will prompt the use of inferentials and assumptives even in cases where such representations appear unambiguous (e.g. Quartararo 2017). The simple fact that depicted events and people are outside of the speaker's domain of epistemic authority, may be sufficient to warrant a more cautious approach to asserting such events by using indirect evidentials. The speaker may mark an event as being inferred from his/her point of view, rather than claiming direct perceptual access to the contents of the picture, possibly because such contents pertain to previously unknown characters in a fictional universe. Curnow (2003) discusses the use of non-visual/indirect forms with first person subjects to produce unintentional/non-volitional readings of utterances (see Example 2b, below). Such interpretation effects are in line with the hypothesis that indirect evidentials function as disclaimers of epistemic authority. However, it is also possible to use evidentials to *claim* epistemic authority by means of direct evidentials. This may be a less prominent function of evidentials, but one that links evidentiality to egophoricity (see §3.2, below). Just like indirect evidentials may be used even in cases where the speaker has direct sensory access to a talked about event, so are direct evidentials sometimes used to signal other forms of access than their semantics may suggest (i.e. visual, auditory). One such form of access that may be signaled by the use of direct evidentials, is "(volitional) participation". Participatory meaning in the context of evidentiality may result from the distribution of direct evidential forms with subject pronouns (i.e. first vs. third person), or they can constitute distinct forms that are part of paradigms alongside other direct forms that signal visual access to the referent (see Example 3, below). Examples of how the distribution of direct evidentials may produce participatory meaning according to subject person, is discussed by Curnow (2002: 188–190, citing Ramirez 1997: 133):

1 Epistemic perspectives

(2) Tucano

	- plate break-REC.PAST.NON3.NONVISUAL 'I broke the plate accidentally (I didn't see it on the table).'

Curnow discusses the examples from Tucano as an instance of interpretation effects resulting from the distribution of forms signaling a visual/non-visual contrast with first person subjects. Such effects are also reported for other languages and depending on what evidentials are present in each language, different effects may arise (see Curnow 2002; 2003 for details).

In Central Pomo, the evidential paradigm contains a form, -*la*, which denotes "personal agency" (Mithun 1999: 181):

(3) Central Pomo *da-ché-w=la* pulling-seize-PRF=PERSONAL.AGENCY 'I caught it.' (I know because I did it)

It should be noted that performative, or participatory, evidentials assume part of the function of person agreement, given the implied agency of a first person subject in such forms. Subject identity is not an encoded feature, however, and reference to the actions of third person subjects featuring performative/participatory forms may produce a factual reading that corresponds in epistemic status to participatory meaning when referring to events and actions involving the speaker. This means that factual events involving third persons may be marked in the same way as events involving one of the speech-act participants as a participant. Bergqvist & Kittilä (2017; cf. Bergqvist 2015) explores participation/involvement as part of evidential systems in order to place this notion against egophoric marking, in which volitional participation/involvement has been suggested as a defining notion (see directly below). One reason to make such a comparison relates to the ongoing debate on whether egophoric marking is a kind of evidential marker, or if egophoric marking constitutes a separate grammatical expression altogether. If "source of information" (Aikhenvald 2004) is the preferred definition, then participation will be difficult to accommodate within such a definition (but see San Roque & Loughnane 2012 for a discussion). If viewed from the perspective of epistemic marking in language, more generally, then cognitive access

Henrik Bergqvist & Seppo Kittilä

to events must be situated against related means to signal access more broadly, including access from participation/involvement (see Bergqvist 2017; cf. Boye 2012).

### **3.2 Involvement and epistemic authority: Egophoricity**

Egophoricity is a recently proposed term for a form of epistemic marking that prototypically occurs with first and second person subjects in declarative and interrogative clauses, respectively (see Floyd et al. 2018, for a cross-linguistic overview). Egophoricity is also known as "conjunct/disjunct"-marking in the literature (e.g. Bickel & Nichols 2007), but competing terms also exist (see §1, above). Example (4) portrays the two combinations of subject person and sentence type that trigger egophoric marking in Kathmandu Newar (Hale 1980). All other combinations of subject person and sentence-type produces non-egophoric marking (i.e. 1S+interrogative/2S+declarative/3S+any sentence type):

	- a. *ji* 1S.ABS *ana* there *wanā* go.EGO 'I went there.'
	- b. *cha* 2S.ABS *ana* there *wanā* go.EGO *lā* INTERR 'Did you go there?'

In addition to this distributional pattern, there are also restrictions on what verbs may take the egophoric marker. In Kathmandu Newar, only verbs that feature (volitional) agents are permitted. The contrast between verbs that denote volitional actions and ones that do not, is illustrated in (5) (Hargreaves 2005: 12– 13):

	- a. *jĩ:* 1.ERG *jyā* work *yān-ā* do-PST.EGO 'I did the work.'
	- b. *ji* 1.ABS *mhiga* yesterday *then-a* arrive-PFV.ALLO 'I arrived yesterday.'

### 1 Epistemic perspectives

c. *jĩ:* 1.ERG *thul-a* understand-PFV.ALLO 'I understood (it).'

Bergqvist & Knuchel (2017) take a broad approach to analyzing egophoric marking by outlining the boundaries of speech-act participant involvement in such systems. Although the volitional actions of a speech-act participant purportedly are a defining semantic component of egophoric marking, it has become increasingly clear that this is not the only kind of involvement that may trigger the use of an egophoric marker. Involvement as a basis for epistemic authority may in some instances target the affectedness, or even the attitude of a speaker/addressee (Bergqvist & Knuchel 2017: 369). Given that such an encompassing formulation of involvement is applicable to some egophoric markers, Bergqvist & Knuchel propose that "epistemic authority" is actually the core semantic notion that may define egophoric marking against other forms of epistemic marking, such as evidentials and epistemic modals. Focusing on the notion of epistemic authority also bridges the gap to seemingly unrelated phenomena such as "ethical datives", which share formal and functional features with egophoric marking, as described for a language like Standard Tibetan (see Bergqvist & Knuchel 2017, for details). The involvement of a speech-act participant produces a kind of epistemic inalienability that permits the speaker to assign epistemic authority to him/herself, or the addressee without necessarily specifying what this involvement consists of.

The notion of epistemic authority may be conceptualized as a driving force in verbal interaction more generally and serves to situate information with respect to the speech-act participants. As such, it goes well beyond the use of egophoric markers in languages where this is an attested form of epistemic marking. For English (arguably a language without egophoric marking), the notion of epistemic authority may be seen in the correspondence between sentence-type and communicative function, an issue that has concerned speech-act theory since its formulation (Searle 1969). Sometimes an assertion may function as a question in a communicative sense, and vice versa. In fact, the majority of polar questions in American English have the form of an assertion (Stivers 2010; Stivers & Rossano 2010). In order to explain this apparent discrepancy, Heritage (2012) argues that the notions, "epistemic status" and "epistemic stance" are key for understanding discrepancies between grammatical form and (social) action. Epistemic status, as an index of relative epistemic authority, is formulated with reference to the notion of A and B-events (Labov & Fanshel 1977), where A-events are known only to the speaker (speaker authority) and B-events are known only to the addressee

### Henrik Bergqvist & Seppo Kittilä

(addressee authority). Typical B-events include the addressee's opinions, beliefs, and bodily states, but may also include his/her professional expertise. Heritage offers the following definition of epistemic status:

(W)e can consider relative epistemic access to a domain as stratified between actors such that they occupy different positions on an epistemic gradient (more knowledgeable [K+] or less knowledgeable [K−]), which itself may vary in slope from shallow to deep […]. We will refer to this relative positioning as epistemic status, in which persons recognize one another to be more or less knowledgeable concerning some domain of knowledge (…) (Heritage 2012: 32)

The speaker's epistemic stance can be congruent or incongruent with the speaker's epistemic status, as seen in the example, *you're married*, provided by Heritage, which may be understood as a request for confirmation despite its assertive formulation, given that it pertains to the addressee's marital status. A statement such as, *you're sad*, using an assertive form, may be deemed incongruent to the speaker's epistemic status, given that it must be regarded as K-, given the addressee's obvious epistemic authority over his/her own emotional states.

Speakers of any language continuously keep track of what others know and how their own knowledge can be related to the knowledge of others, and Heritage offers us a detailed and empirically grounded picture of how this works in everyday conversation. Such assumptions most prominently involve the addressee and explicit formulations of how the addressee's perspective is attended to have received some attention in the field of discourse studies, notably in work by Ken Hyland (e.g. 1999; 2001; 2005). More recently, cross-linguistic research has led to the formulation of "engagement" as a bona fide epistemic category in some languages (see Evans et al. 2017a) and the relation between engagement and epistemic authority deserves to be explored in the definition of both notions.

### **3.3 Shared vs. non-shared access/rights to knowledge: Engagement**

Engagement consists of a contrast between the speaker's assertion about the addressee's knowledge of/attention to an event (*e*) as either *shared* or *non-shared* with the speaker. *I know e and I assume that you know e too*, is contrasted with, *I know e and I assume that you do not* (Evans et al. 2017a; cf. "complex perspective", Evans 2005; cf. "complex epistemic perspective", Bergqvist 2015; 2016; 2017). This semantic contrast may in principle concern any aspect of epistemicity, and is as such relevant for any form of epistemic marking (see below). Engagement targets "knowing" from a socio-centric perspective, where the speaker's assertion

### 1 Epistemic perspectives

contains an embedded assertion assumed to belong to the addressee. Such assumptions may be implied in the token-use of other forms of epistemic marking, but in a language with engagement as a category, this implicature has become encoded in forms to signal asymmetries in the knowledge/attention states of the speech-act participants.

In Kogi (Arwako-Chibchan; Bergqvist 2016; cf. Ortíz Ricaurte 1994), speakers can choose one of four auxiliary prefixes that encode engagement.<sup>6</sup> These prefixes may be divided into two sets that take the speaker and the addressee as their respective starting points. A focus on the perspective of the speaker is found with *na-*/*ni-*, where *na-* means that 'the speaker knows *e* and expects the addressee to be unaware of *e*' (6a), and *ni*- means that 'the speaker knows *e* and expects the addressee to know *e* too' (6b) (Bergqvist 2016: 2):

(6) Kogi


*Na*-/*ni*- are in turn contrasted to *sha*-/*shi*-, which encode a corresponding distinction in terms of non-shared/shared knowledge from the addressee's perspective. *sha*- means that 'the speaker expects the addressee to know *e* while the speaker is unaware of *e*' (7a), and *shi*- means that 'the speaker expects the addressee to know *e*, and the speaker knows *e* too' (7b) (Bergqvist 2016: 3):

(7) Kogi


<sup>6</sup>Bergqvist (2015; 2016; 2017) discusses the Kogi system using the term "complex epistemic perspective" without arguing for a more general applicability of this term to similar systems and forms in the literature. Such applicability is, however, considered in Evans et al. (2017a, 2017b) under the term engagement.

### Henrik Bergqvist & Seppo Kittilä

*Shi*-/*sha*- are used to signal the speaker's acknowledgement of the addressee as primary knower, but at the same time encodes the speaker's assertion (without reduced certainty) of a talked about event.

Engagement in Kogi may appear to be an exotic system in a little-known language, but this notion echoes with well-known phenomena like modal particles in Germanic languages (modal particle). Descriptive accounts of cognates of the German *ja* ('as you know'/'of course', Abraham 1991), in Danish (*jo*, Davidsen-Nielsen 1996), Norwegian (*jo*, Andvik 1992) and Swedish (*ju*, Lindström 2008) agree that reflexes of this form signals "available knowledge through shared experience" (Lindström 2008: 74, for Swedish).<sup>7</sup> But although modal particles are relatively frequent in e.g. spoken Swedish (see Bergqvist 2017), they are generally not viewed as part of core grammar, given their non-obligatoriness, weakly paradigmatic organization, and function as markers of discourse. Modal particles have an established (Eurocentric) descriptive tradition, but have recently been compared to analogous particles in non-European languages such as Chinese and Japanese (e.g. Abraham & Leiss 2012). A continued exploration of modal particles by means of cross-linguistic comparison, will surely contribute to a more developed understanding of engagement as a linguistic category.

From the typological overview presented in Evans et al. (2017b), it is clear that engagement can combine with modals and evidentials and that the shared/nonshared contrast may fuse with such forms of epistemic marking. Hintz & Hintz (2017) account for evidentials in Sihuas Quechua and argue that this language has developed two distinct sets of evidential markers, which feature an -*i*/-*a* alteration encoding individual/mutual knowledge, respectively. Papuan languages like Foe (Rule 1977) and Angal (Sillitoe 2010) also display engagement semantics fused with evidential forms that produce readings like 'as I could see, but you could not'. Schultze-Berndt (2017) reports that markers of epistemic authority in Jaminjung (Mirndi, Australia) are contrasted according to whether epistemic authority is considered to be shared, or exclusive to one of the speech-act participants.

Ika, a language closely related to Kogi (above), features a version of egophoric marking that mirrors some of the semantic contrasts found in Kogi, but by way

<sup>7</sup> For Danish, Davidsen-Nielsen argues that "*jo* signals that the hearer is assumed to be aware of and accept the states of affairs described [...]" (Davidsen-Nielsen 1996: 285). He notes that the notion of something being "familiar to the receiver" is subject to some variation, but that the semantic component of including the hearer's perspective remains part of utterences featuring *jo* in Danish (Davidsen-Nielsen 1996: 293). For Norwegian, Berthelin et al. (2013) note that the meaning of *jo* encodes that "the hearer and speaker both have access to all the evidence required for entertaining *p* as true".

### 1 Epistemic perspectives

of a distinct system that signals the involvement of the speech participants in relation to their respective epistemic authorities (Bergqvist 2012; 2018b;2018a). The epistemic marking systems in Kogi and Ika shows how a functional pressure to assign epistemic authority to one, or both of the speech-act participants may produce distinct systems in closely related languages. Drawing on this comparison, Bergqvist (2018a) argues that any subjective evaluation/positioning may develop a sensitivity to whether this is shared with the addressee, or not, and that the possibility of such a development stems from the very nature of indexical reference.

An eventual typology of engagement must answer questions regarding its relevant dimensions of meaning, i.e. whether encoded (a)symmetries of knowing concern the assimilated knowledge of the addressee, or the speech-act participant's respective rights to know in terms of epistemic authority. With respect to the diachronic development of engagement, this must be accounted for in the context of engagement being part of related functional categories, but also as a distinct grammatical expression.

### **4 Concluding thoughts and a view to the future**

From the overview and discussion provided in this introductory chapter, it should be clear that evidentiality, egophoricity, and engagement are closely related notions that overlap in both form and function. The notion of epistemic authority is argued to be central to a functional analysis of said categories, either as an underlying (largely implicit) motivation for the use of evidentials, or as a defining semantic component of egophoric marking. It may also be assigned to one, or both, of the speech-act participants by forms of engagement. The general role of epistemic authority as an integral part of the "epistemic engine" that drives verbal communication (Heritage 2012) lends further support to idea that epistemic authority is a key concern for speakers engaged in conversation.

In order to explore epistemicity in language with this focus, interactive language data is required. The dialogic function of language was discussed as a defining feature of language already by Jespersen (1922), but throughout the 20th Century this function all but disappeared from linguistic analysis (but see Givón 2001; Halliday 1973). Recently, dialogicity has come back on the agenda in research by Du Bois (2007; 2014) and Evans (2012), among others. A return to the dialogical features of language use and its consequences for grammar appears crucial to account for the ongoing exploration of epistemic marking strategies in languages everywhere.

### **Abbreviations**


### **References**


1 Epistemic perspectives


Henrik Bergqvist & Seppo Kittilä


# **Chapter 2**

# **Epistemic primacy, Common Ground management, and epistemic perspective**

### Karolina Grzech

University of Stockholm

In this chapter, I discuss the epistemic discourse clitics attested in Upper Napo Kichwa, a Quechuan language spoken in the Ecuadorian Amazon. I show that contrary to how they have been described in other Quechuan dialects, in Upper Napo Kichwa the enclitic =*mi* and =*cha* should not be treated as evidentials, but as markers related to the (lack of) epistemic authority/primacy, i.e. origo's relative right to know a certain piece of information. I also demonstrate that although Quechuan evidentials have previously been analysed as focus markers, this analysis, too, cannot be sustained for Upper Napo Kichwa, where the markers in question are associated with focal constituents, but cannot be said to "mark" focus.

With examples from a corpus of Upper Napo Kichwa monolingual discourse, I show that the considerations related to ownership and distribution of knowledge play a role in the management of Common Ground in interaction. I show that by using the two enclitics, speakers can indicate whether or not at a given point in interaction the information they convey should be integrated into Common Ground. In order to generalise this analysis, I propose situating linguistic items dedicated to Common Ground management, such as =*mi* and =*cha*, within the cross-linguistic functional domain of epistemic perspective.

### **1 Introduction**

This chapter explores the meaning and functions of =*mi* and =*cha* in Upper Napo Kichwa,<sup>1</sup> an under-documented Quechuan language spoken in the Ecuadorian

<sup>1</sup>Upper Napo Kichwa is one of the varieties of Ecuadorian Amazonian Kichwa. In my previous work (e.g. Grzech 2016a, Grzech 2016b) I have been referring to the same variety with the name "Tena Kichwa".

Karolina Grzech. 2020. Epistemic primacy, Common Ground management, and epistemic perspective. In Henrik Bergqvist & Seppo Kittilä (eds.), *Evidentiality, egophoricity, and engagement*, 23–60. Berlin: Language Science Press. DOI: 10.5281/zenodo. 3975797

### Karolina Grzech

Amazon. In other Quechuan varieties described to date, these two enclitics have been analysed as direct and inferential/conjectural evidentials, respectively. As I show in this chapter, in Upper Napo Kichwa, they are more adequately analysed as markers of epistemic primacy (=*mi*) or lack thereof (=*cha*). That is, they indicate the origo's (lack of) "relative right to tell, inform, assert or assess" (Stivers et al. 2011: 13). Moreover, the markers form part of a larger paradigm of free enclitics loosely associated with focal status of the constituents they attach to.

By way of introduction, I outline the facts underpinning the research presented here, providing an overview of the research objectives ( 1.1), giving background information on the language under study and describing the data collection methods as well as the resulting corpus ( 1.2), and defining the notions used in my analysis ( 1.3).

In the ensuing sections, I discuss the Upper Napo Kichwa paradigm of discourse enclitics (§ 2) and the previous analyses of the markers =*mi* and =*cha* (§ 3). I discuss the markers' semantics and functions in discourse (§ 4). Subsequently, I propose a unified analysis of the different aspects of meaning of the two enclitics (§ 5) and integrate this proposal with a cross-linguistic framework for the description of epistemic marking systems (§ 6). Finally, I provide some conclusions and suggestions for further research (§ 7).

### **1.1 Research objectives**

The main objective of this chapter is to spell out the analysis of =*mi* and =*cha* in terms of epistemic primacy, and to explain how the markers interact with focus. Furthermore, I aim to show how through fulfilling both these functions – marking epistemic primacy, and association with focal status of referents – the enclitics in question contribute to the management of Common Ground in Upper Napo Kichwa discourse. I also show the cross-linguistic relevance of this analysis, discussing the possible place of =*mi* and =*cha* within the "domain of epistemic perspective" (Bergqvist 2017).

### **1.2 Language and research background**

Before I proceed to the description of the enclitics =*mi* and =*cha*, it is in order to provide some background on the language which I analyse here. Upper Napo Kichwa is a Quechuan language of the QII subgroup, spoken in the province of Napo, in the Ecuadorian Amazon. The estimated number of speaker varies from 20,000 (Lewis 2016) to ca. 46,000 (INEC 2010).

### 2 Epistemic primacy, Common Ground management & epistemic perspective

Upper Napo Kichwa belongs to the dialectal grouping of Amazonian Kichwa, and is but one of several Quechuan varieties spoken in Ecuador. Although Quichua<sup>2</sup> is recognised by the Ecuadorian constitution as the "official language of intercultural relations" (ANCE 2008), the legislation and resulting policies on the national level fail to recognise the existence of multiple Quechuan languages within Ecuador. Instead, the focus is on Unified Kichwa - an official, "standard" variety created in the 1980s with the aim of providing a single, official orthography for all Quechuan dialects spoken in Ecuador. Unified Kichwa is used in Spanish-Kichwa bilingual education across the country, and its knowledge is required of all teachers of Kichwa, even if they are native speakers of other varieties. It has become the prestige variety of the language, and is adopted by speakers of other varieties for official purposes, including use in administration and cultural activities (cf. e.g. Wroblewski 2014). Consequently, its official status contributes to the weakening rather than strengthening of the regional varieties (cf. Hornberger & King 1996; Grzech 2017).

The assessments of the vitality of Upper Napo Kichwa have delivered varying outcomes. While Lewis (2016) claims the language is "vigorous", Moseley (2010) evaluates it as "seriously endangered". Such discrepancy arises most likely due to the fact that Lewis (2016) does not take into account the difference between Unified Kichwa and Upper Napo Kichwa (cf. Grzech 2017). Participant observation I carried out in 2013 and 2014 suggests that while Upper Napo Kichwa is not yet seriously endangered, this might be the case within a generation. In the communities where I conducted fieldwork, the last generation who uses Upper Napo Kichwa in all social contexts are people around the age of 30. Teenagers have passive knowledge of the language, but tend to communicate with each other in Spanish, and small children are generally spoken to in Spanish, and use the language amongst themselves. Nonetheless, my main field site was relatively well connected to the provincial capital, and it is likely that the language is faring better in more secluded settlements (Arthur Cognet, p.c., December 16th, 2016 ).

In terms of its morphosyntactic characteristics, Upper Napo Kichwa exhibits a number of features typical of the Quechuan language family. It is agglutinative, exclusively suffixing, and the two main word classes are verbs and nominals, characterised by differing patterns of inflection. In Quechuan languages the dominant word order is generally SOV, but some studies also characterise them as discourse-configurational (cf. Muysken 1995). A preliminary study of

<sup>2</sup> In Ecuador, the term *Quichua* is preferred over *Quechua*. I use the name *Quichua* when referring to Ecuadorian Quechua in general, but for the name of the language with which this chapter is concerned, I use the newer orthography, hence writing it *Kichwa*. This orthography is the one most widely used by the members of the community I work with.

### Karolina Grzech

the Upper Napo Kichwa word order suggests that in this variety the orders SOV and SVO are equally permissible (cf. Grzech 2016a: ch.4). Moreover, the language is characterised by less morphological complexity than most other documented Quechuan dialects (cf. Adelaar & Muysken 2004); for instance, similarly to other Ecuadorian varieties, e.g. Imbabura Quichua (Cole 1982), Upper Napo Kichwa only exhibits residual object agreement marking on the verb.

The data on which this research is based were collected during ten months of language documentation fieldwork in 2013 and 2014. My main field site was the village of Nuevo Paraíso, situated on the bank of the Napo River, about fifty kilometres west from Tena, the capital of the Napo Province. Nuevo Paraíso is accessible by river and by a dirt road. Buses to and from Tena pass through Nuevo Paraíso several times a day, and the journey takes about two hours. As of September 2014, 53 associates (Spanish: *socios*) lived in the village, most of whom were heads of families. In Kichwa communities, families are formed by parents, unmarried children, and sometimes also widowed grandparents. They range in size from three to about ten people, the average number of children per family in this particular community being around five.

The corpus I used for this research comprises two parts: an eleven-hour corpus of naturalistic Upper Napo Kichwa discourse, and a two-hour corpus of elicited discourse. Both parts of the corpus were recorded on audio and video, transcribed in Upper Napo Kichwa and translated into Spanish. In addition, the elicited discourse corpus was also parsed and annotated with morpheme-by-morpheme glosses.

The naturalistic discourse part of the corpus consists of recordings of communicative acts characterised by different degrees of spontaneity, ranging from everyday conversation, through storytelling, to recordings of community events and political discourse. The elicited discourse corpus includes "staged communicative events" (Himmelmann 2006): discourse resulting from presenting consultants with video and/or picture stimuli, or asking them to perform specific tasks. The stimuli I used to collect this part of the corpus included e.g. the "Pear story" video (Chafe 1980), and the tasks for two consultants from the "Questionnaire on information structure" (Skopeteas et al. 2006). These type of tasks allow for obtaining naturalistic parallel data (San Roque et al. 2012: 137), and for comparing constructions used by various speakers in the same discourse situation. They also permit the researcher to control what information is, and is not, shared between discourse participants – a task unattainable in case of naturalistic discourse.

The documentation project was carried out in collaboration with a team of Kichwa researchers: Nilo Licuy, Jacobo Chimbo, Wilma Aguinda and Edwin Shi-

### 2 Epistemic primacy, Common Ground management & epistemic perspective

guango. Sofía Alvarado also contributed to the transcription and translation of corpus. The members of the research team selected the topics to be documented, as well as the participants for the interviews. They also worked as camera and sound operators, interviewers, transcribers and translators (for more details on collaborative documentation, cf. Grzech 2016a: sec. 1.3.3).

### **1.3 Definitions**

As stated previously, the main objective of this chapter is to explore the meaning and discourse functions of the Upper Napo Kichwa enclitics =*mi* and =*cha*, which I analyse as markers of the origo's epistemic primacy, or lack thereof. To accomplish this, I first define the basic notions used in the analysis sections of this chapter. I discuss evidentiality and epistemic primacy, and briefly acquaint the reader with the domain of epistemic perspective. Subsequently, I focus on the notions pertinent to information structure that are necessary for the adequate description of =*mi* and =*cha*: focus and Common Ground.

### **1.3.1 Evidentiality**

Although I do not analyse the Upper Napo Kichwa enclitics as evidential markers, the notion of evidentiality is mentioned frequently throughout the chapter, and therefore needs to be clarified. I understand evidentiality in the "narrow" sense of the term, as the linguistic coding of the source of information (cf. e.g. Willett 1988: 54; Nikolaeva 2000; Dendale & Tasmowski 2001: 342–343; Aikhenvald 2004) or "mode of access" (e.g. González Ruiz et al. 2016). Under this view, the source of information on which a proposition is based is independent of the speaker's beliefs about the veracity of that proposition; evidentiality marks the source of information on which a proposition is based, while epistemic modality evaluates the likelihood that this proposition is true (Cornillie 2009). Nonetheless, evidential and modal meanings are often hard to separate (Palmer 2001), and cross-linguistic evidence shows that evidential and epistemic modal meanings can be encoded by the same set of markers (Willett 1988: 55).

Narrowly defined evidentiality and epistemic modality have also been regarded as two sub-types of the category of EPISTEMICITY (Boye 2012). According to this approach, both evidentiality and epistemic modality provide JUSTIFICATORY SUP-PORT for propositions. Evidential expressions provide EPISTEMIC JUSTIFICATION, which can be either direct or indirect. Epistemic modal expressions, in turn, provide EPISTEMIC SUPPORT, which can be full (certainty), partial (probability) or neutral (lacking epistemic qualification) (cf. Boye 2012: 36). The term EPISTEMIC

### Karolina Grzech

MEANING is used in this chapter in a broader sense than the one adopted by Boye (2012: sec. 1.5); Following Bergqvist (2017), I see both evidentiality and epistemic modality as sub-domains within the domain of EPISTEMIC PERSPECTIVE, which I discuss in more detail in §1.3.3, after introducing other notions necessary for understanding it.

### **1.3.2 Epistemic primacy**

Another notion indispensable for the accurate description of the two Upper Napo Kichwa enclitics discussed here is EPISTEMIC PRIMACY (Stivers et al. 2011). It can be conceptualised, alongside evidentiality, as related to the dimensions of knowledge in interaction (cf. Stivers et al. 2011: 13), presented in Table 2.1:


Table 2.1: Dimensions of Knowledge

The three dimensions listed above correspond to the different "levels" on which knowledge can be grounded in conversation. Evidentiality clearly falls within the dimension of epistemic access, since it relates to the type of evidence. Epistemic modality also falls within that domain, as it is related to the degree of certainty.

Epistemic primacy is more subjective than epistemic access;<sup>3</sup> while epistemic access is concerned with the relationship between the proposition and the origo, epistemic primacy has to do with the distribution of knowledge between participants of the speech event. Epistemic primacy is the asymmetry "in the depth, specificity or completeness of their [speech act participants'] knowledge" (Stivers et al. 2011: 13). Consequently, the making of epistemic primacy is grounded in the subjective assessment of the origo's knowledge state rather than in the relationship of that knowledge to the discourse-external world. In the literature,

<sup>3</sup> Subjectivity can be defined as "(…) the way in which natural languages, in their structure and their normal manner of operation, provide for the locutionary agent's expression of himself and his own attitudes and beliefs" (Lyons 1982: 102). Understood in this manner, subjective expressions index the attitudes or viewpoint of the speaker.

### 2 Epistemic primacy, Common Ground management & epistemic perspective

epistemic primacy is often used interchangeably with EPISTEMIC AUTHORITY (cf. e.g., Grzech 2016a; García-Ramón 2018). Nonetheless, an important distinction can be made between those notions: epistemic authority is gradable: one possesses it if one knows something, although it is possible to know more or less about the matter at hand. Moreover, epistemic authority can, but does not have to be relative. One can have epistemic authority on a subject independently of the knowledge state of other discourse participants. Epistemic primacy, on the other hand, is relative by default: whether or not one can claim to have epistemic primacy always depends on the knowledge states of other participants of a given interaction. It follows that epistemic primacy can only be established in a particular interactional context.<sup>4</sup> The epistemic authority that one might have over a certain matter only allows one to claim epistemic primacy with respect to interlocutors who know less than one does. While epistemic primacy/authority often arises as a result of having the best possible type of evidence for the information in question, or being certain that the proposition is true, it need not be grounded in direct evidence or certainty.

The third domain – EPISTEMIC RESPONSIBILITY – is related to the information that the speaker has an obligation or a right to know. For instance, it is expected of everyone to know their own name, etc. On the other hand, there is information about other people, their internal states and experiences, or private affairs, about which their interlocutors do not have a responsibility, or even right, to possess knowledge. I will not devote more attention here to this last domain, (see Grzech 2016a: ch.5 for a more detailed discussion), as it is only relevant to the analysis presented here inasmuch as the domains of epistemic primacy and epistemic responsibility correlate with one another. It should be expected that if the origo has an obligation/right to know a certain piece of information, she is also likely to have epistemic authority over it.

### **1.3.3 Epistemic perspective**

The dimensions of knowledge discussed above relate mainly to the origo, and to the extent to which she can know, or claim to know, a piece of information. The only reference to the interpersonal aspect of communication in Table 2.1 is the observation that epistemic primacy is "a *relative* right to know or claim". The relative nature of epistemic primacy suggests that the origo's interlocutor and

<sup>4</sup> I am grateful to Amparo García-Ramón for the discussion which helped me clarify this distinction. The authority/primacy distinction is related, but not tantamount to, the epistemic stance/epistemic status distinction (e.g. Heritage 2012).

### Karolina Grzech

her state of knowledge should also be taken into account by the speaker when she chooses to use a given marker of epistemic primacy.

In line with this observation, Bergqvist (2017) proposes to analyse evidentiality and other epistemic marking systems within the functional domain of EPISTEMIC PERSPECTIVE, 5 encompassing not only the relationship of origo to the information, but also the distribution of information between the speech act participants. Proposing the existence of the domain of epistemic perspective is based on two assumptions: (1) that evidentiality and related systems share a *functional space* in the grammar of languages in which they occur; (2) that these epistemic marking systems allow the speaker to adopt different perspectives with respect to both information and interlocutors Bergqvist (2017: 11). Consequently, the domain can be divided into different subdomains of epistemic meaning, stratified according to the level of (inter)subjectivity they express.<sup>6</sup> Figure 2.1 shows this stratification:

```
SCOPE
```
(wide)

```
SPEAKER-HEARER LINKS illocutionary modality
SPEAKER INVOLVEMENT egophoricity
INFORMATION SOURCE evidentiality
POSSIBILITY epistemic modality
NECESSITY
```
KNOWLEDGE (A)SYMMETRY *complex epistemic perspective (CEP)*

(narrow)

Figure 2.1: Dimensions of the epistemic perspective (after Bergqvist 2017: 12)

As mentioned above, the notion of *epistemic perspective* points to the fact that all the categories shown in Figure 2.1 are to some extent concerned with marking the different dimensions of epistemic access of the discourse participants. From bottom to top, the different epistemic marking systems shown on the right-hand

<sup>5</sup>Bergqvist (2017) uses the term FUNCTIONAL DOMAIN coined by Givón (1981); functional domains are conceptual structures which can overlap with one another and encompass several levels of meaning (Bergqvist 2017: 11).

<sup>6</sup> Subjectivity was defined in Footnote 3 above. Intersubjectivity relates to the speakers' "acknowledgement of and attention to the addressee" (Traugott 2010: 2). Consequently, intersubjective expressions take into account the attitudes of, or distribution of knowledge between, the speaker and the addressee.

### 2 Epistemic primacy, Common Ground management & epistemic perspective

side are organised from the least to the most intersubjective. On the left-hand side, Figure 2.1 shows the different functional domains corresponding to the epistemic marking systems shown on the right. The arrow corresponds to the levels of meaning, based on the assumptions that the categories at the bottom of the scale encode more propositional meaning, and hence have narrower scope, and the categories towards the top encode non-propositional meaning. The marking of complex epistemic perspective refers to the fully intersubjective systems, in the encoding of which the perspectives of both interlocutors are equally important. The composition of the domain presented in Figure 2.1 is based on a preliminary cross-linguistic survey, including the grammatical categories attested in a sample of languages which exhibit the exemplified epistemic marking systems (Bergqvist 2017).

The number of functional sub-domains – and corresponding grammaticalised marking systems – is bound to increase if other languages are taken into account. Examples of such systems could include the marking of engagement (cf. e.g. Landaburu 2007), information status of discourse participants (cf. San Roque 2008), or the marking of epistemic primacy, discussed at length in this chapter. In Japanese, epistemic primacy is encoded by the dedicated marker *yo* (Hayano 2011). In the following sections, I show that epistemic primacy is also morphologically marked in Upper Napo Kichwa, where the distribution of epistemic authority between speaker and hearer is encoded by =*mi* and =*cha*. Therefore, at least for certain languages including Upper Napo Kichwa, epistemic primacy should be considered one of the *functional sub-domains* within the overarching domain of epistemic perspective.

### **1.3.4 Common Ground and focus**

The remaining notions that need to be explained before I proceed to the analysis of the Upper Napo Kichwa enclitics =*mi* and =*cha* are Common Ground and focus, both pertinent to the study of information structure. I define them in turn below. The readers should keep in mind that the definitions provided here are not meant to tackle the conceptual complexity of the discussed notions in full detail. Rather, they are meant to provide basic definitions, so as to allow for a consistent, clear interpretation of the analysis presented in the following sections.

COMMON GROUND (henceforth CG) consists of information which is mutually known to be shared by the discourse participants (cf. e.g., Stalnaker 1974; Clark 1996). This information includes discourse referents that the interlocutors are familiar with, and 'a set of propositions which the participants in the conversation mutually agree to treat as true for the purpose of the exchange' (Stalnaker

### Karolina Grzech

& Cole 1978). CG constantly develops over the course of communication, and Krifka (2007) points out that there are two aspects of CG relevant to communication: CG CONTENT, which includes all the truth-conditional information within the CG, and CG MANAGEMENT, which indicates the way in which CG content should develop. Both CG content and CG management are shared between discourse participants. The aspects of information structure (henceforth IS) that have truth-conditional impact can be associated with CG content, and those relating to the pragmatic use of expressions – with CG management (Krifka 2007: 18). The notion of CG management has been applied in the analysis of e.g. German modal particles (cf. Repp 2013, Döring 2016). Over the course of this chapter, I show that Upper Napo Kichwa discourse enclitics contribute to CG management, rather than to CG content.

As mentioned above, CG develops constantly in the process of communication. For this to occur, utterances need to contain not only information already known to both interlocutors, but also information known to the speaker, but new to the hearer, so that it can be added to CG in the process of communication. However, there are cognitive constraints on how much new information can be conveyed at a time. According to the ONE NEW IDEA CONSTRAINT (Chafe 1987; 1994), for processing reasons, every clause in connected discourse can contain only one concept which falls under the scope of assertion: the focus of the clause. Clauses also contain presupposed content, and within it, expressions denoting a referent the clause is about (cf. Lambrecht 1994: 127). In IS terms, such a referent is the topic of the clause. Both topic and focus are relational notions. That is, no referent is inherently focal or topical. Rather, the topic or focus relation arises between discourse referents and propositions as a result of the speaker's strategy of CG management.

### **2 Upper Napo Kichwa discourse enclitics: the paradigm**

In this section I present the morphosyntactic paradigm of Upper Napo Kichwa discourse enclitics, of which both =mi and =cha form part. As mentioned in § 1.2, Upper Napo Kichwa is an exclusively suffixing language. Consequently, all the clitics found in the language are in fact enclitics. Moreover, all the enclitics attested in Upper Napo Kichwa are free, that is, they attach to hosts from any grammatical category, as long as their hosts function as phrasal heads.<sup>7</sup>

<sup>7</sup> In §1.2, I mention that Upper Napo Kichwa has two main grammatical categories, nouns and verbs. Nonetheless, members of minor word classes, e.g., adverbs, can also function as phrasal heads/hosts for enclitics.

### 2 Epistemic primacy, Common Ground management & epistemic perspective

The presence of free enclitics is attested in all described Quechuan varieties (cf. e.g., Parker 1969; Cole 1982; Weber 1986; Cusihuamán 1976), but most authors do not grant them a comprehensive description, focusing only on a few selected markers. In the Upper Napo Kichwa data collected to date, as many as fifteen free enclitics are attested (cf. Grzech 2016a: ch.3). Of those, nine were identified as DISCOURSE (EN)CLITICS in the sense of Spencer & Luis (2012: 37) as clitics which express discourse functions, rather than inflectional categories. The occurrence of discourse enclitics is not conditioned by grammar, and hence no contexts were identified in which their occurrence would be required for the syntactic wellformedness of the clause. Rather, they are used to enhance discourse coherence by encoding "cues for interpretation, (…) emphasis, rhetorical effects, or the attitude of the speaker" (cf. Spencer & Luis 2012: 35). Of the nine discourse enclitics in Upper Napo Kichwa, one, namely =*ga*, seems to be associated with presupposed information (cf. Grzech 2016a: ch.4). The remaining eight markers – among them the two enclitics described in this chapter – associate with focal referents. These eight enclitics, along with the analyses proposed for their cognates in different Quechuan languages, are listed in Table 2.2.

All of the enclitics listed in Table 2.2 occur on focal referents, and attach to their hosts word-finally, following all the inflectional morphology. However, apart from being associated with focus, they also play different roles in Upper Napo Kichwa discourse. For the sake of space, this chapter only concentrates on two of the markers listed above, =*mi* and =*cha*.

As mentioned above, all the Upper Napo Kichwa enclitics are syntactically non-obligatory, and the same applies to the two markers in question. In the parsed-and-glossed part of the Upper Napo Kichwa corpus, comprising 1537 turns, =*mi* occurred in 5.9% of the turns (n=96), and =*cha* was even less frequent, occurring in 2.15% of turns (n=33). This low frequency suggests that despite their association with focus, other discourse-related factors are decisive when it comes to the markers' use in discourse. In the following sections, I explore these factors, including their epistemic primacy semantics and their role in CG management.

### **3 Quechuan evidential enclitics and the Upper Napo Kichwa "evidential" paradigm**

As mentioned in the previous sections, in most other varieties of Quechua described to date, =*mi* and =*cha* are analysed as evidential markers encoding direct and conjectural/inferential evidence, respectively. In this section, I present the Quechuan evidential paradigm (3.1 ) and present the analysis of =*mi* and =*cha* in

### Karolina Grzech

Table 2.2: Focus-related discourse enclitics in Upper Napo Kichwa


### 2 Epistemic primacy, Common Ground management & epistemic perspective

other Quechuan varieties (3.2), so as to contextualise the analysis of the Upper Napo Kichwa markers presented in §4 and § 5.

### **3.1 The Quechuan evidential paradigm**

Apart from the two markers described in this chapter, the evidential paradigm found in most Quechuan varieties also contains a third marker, encoding reportative evidence. This three-choice system is illustrated in (1) below with data from Cuzco Quechua, spoken in Peru (adapted from Faller 2002: 122):

(1) Cuzco Quechua<sup>8</sup>

```
a. Direct/Best possible ground9 =mi
   Parashanmi.
   para-sha-n=mi
   rain-PROG-3=MI
   'It is raining.' [speaker sees that it's raining]
b. Inferential/conjectural =chá
   Parashanchá.
   para-sha-n=chá.
   rain-PROG-3=CHÁ
   'It is raining.' [speaker conjectures that it's raining]
c. Reportative =si
   Parashansi.
   para-sha-n=si.
   rain-PROG-3=SI
```
'It is raining.' [speaker was told that it's raining]

The paradigm shown in (1) is attested in most described Quechuan languages. However, some varieties diverge from it, offering the speakers more choices. South Conchucos Quechua (QI) is reported to have five evidential markers (Hintz 2012a; Hintz & Hintz 2017). Six markers have been described for Sihuas Quechua (QI) (Hintz 2012b; Hintz & Hintz 2017) and Huamalíes Quechua (Howard 2012), both closely related to South Conchucos. In these varieties, the speaker's choice of evidential marker is influenced not only by the type of evidence, but also by the distribution of knowledge between discourse participants. All these systems

<sup>8</sup>Language names are only included in examples from languages other than Upper Napo Kichwa.

<sup>9</sup> I explain the notion of 'Best possible ground' below.

### Karolina Grzech

have the direct, indirect and reportative markers. In addition, South Conchucos has markers asserting mutual knowledge and indicating shared conjecture (Hintz & Hintz 2017). Sihuas has a system of three contrastive pairs, indicating the distinctions between individual and mutual knowledge, individual and shared conjecture and individual knowledge from report vs generalised knowledge from report (Hintz & Hintz 2017). Humalíes has markers indicating speculation, affirmation of knowledge co-constructed by the speaker and the hearer, and negation of such knowledge<sup>10</sup> (Howard 2012). In the above systems, the markers do not correspond strictly to source of information – the use of INDIVIDUAL vs SHARED KNOWLEDGE MARKERS is also influenced by the speaker's opinion about whether she has epistemic authority over the information conveyed (Hintz & Hintz 2017: 96). Epistemic authority in particular is relevant to the description of the 'evidential' markers in Upper Napo Kichwa, as I show in § 4.

In the varieties discussed above, the evidential paradigm consists of three or more markers. However, in some varieties, including Imbabura Quechua (QII) (Cole 1982: 164–165), spoken in the Ecuadorian Highlands,<sup>11</sup> and Upper Napo Kichwa, described here, reports are marked periphrastically by means of the verb *ni*- ('say') rather than by the use of a dedicated marker. Interestingly, the reportative marker is attested in Pastaza Quichua (QII) (Nuckolls 1993; Nuckolls 2012), which is related to Upper Napo Kichwa more closely than the highland Imbabura variety.

In Upper Napo Kichwa, the reportative marker seems to have been replaced by a periphrastic construction, whereby the verb of speech *ni*- ('say') is used in all hearsay/reportative contexts as a marker of reported speech, and the speech complements are often marked by the enclitic =*mi* (see examples (5) in § 3.2 and (21) in § 6). Given that the reportative enclitic is not attested in the Upper Napo Kichwa data, I limit the discussion here to the direct and conjectural/inferential markers.

### **3.2 Analyses of =***mi* **and =***cha* **in other Quechuan varieties**

The analyses of =*mi* as a direct evidential in different dialects of Quechua point to the fact that it can mark propositions for which the origo has direct, sensory

<sup>10</sup>In Howard's (2012) terms, co-constructed knowledge results from the exchange of information between the speaker and the addressee. This is different from Hintz & Hintz's (2017) 'shared knowledge', understood as knowledge shared by participants, acquired either through a linguistic exchange, or through a shared non-linguistic experience. The latter understanding is more in line with widely accepted definitions of Common Ground.

<sup>11</sup>Cole (1982: 165) observes that while the marker -*shi* is attested in Imbabura, it seems to have undergone semantic shift from marking hearsay to marking speculation.

### 2 Epistemic primacy, Common Ground management & epistemic perspective

evidence. However, a broader definition of the marker's semantics was proposed by Faller (2002), in order to reflect the fact that the direct/sensory access analysis cannot account for all uses of =*mi* in Cuzco Quechua (QII). Faller (2002) analysed =*mi* as the marker of BEST POSSIBLE GROUND (henceforth BPG). BPG corresponds to direct evidence if the information in question belongs to the speaker's own life experience. However, in case of encyclopedic knowledge, which tends to be learnt from authority rather than through direct experience, the BPG can correspond to reportative evidence. Under Faller's view, having the best possible source of information is a necessary, but not sufficient condition for making a =*mi*-marked statement. In order to felicitously use =*mi*, the speaker needs to have the BPG for making a statement. To have the BPG, apart from having the best possible source of information, the speaker also needs to have assimilated the proposition into his network of beliefs (Faller 2002: 140–141).

As for =*cha*, the marker has previously been analysed as a "conjectural" (Weber 1986; Faller 2002), "inferential" (Floyd 1997) or "dubitative" (Muysken 1995) marker in the different Quechuan varieties. Despite the different labels, in all the varieties for which it has been described, the marker is reported to cover indirect evidence based both on reasoning and conjecture. Floyd (1997: ch.5) analyses the meaning of the Wanka Quechua -*chr(a)* as prototypically indicating that a given utterance is an inference. According to Floyd, the prototypical meaning of the marker is "attenuation in the domain of commitment", which "equates nonincorporation into reality (…) and is encoded in terms of likelihood values" (Floyd 1997: 106). He further claims that the non-prototypical uses of =*chr(a)* have to do with attenuation in other domains, including e.g., the "psychological distance between the hearer and the proposition".

Although Floyd does not explicitly refer to epistemic modality, the fact that he analyses the marker as encoding commitment to the likelihood of propositions amounts to an evidential/modal analysis, which is the one that Faller (2002; 2007) proposes in her work on Cuzco Quechua (QII). According to Faller, the conjectural marker is the only CQ evidential that is both an evidential and an epistemic modal. The evidential meaning of =*chá* is to indicate that the speaker "bases his or her statement on a mental process", be it inference, conjecture, guesswork or any other process involving reasoning (Faller 2002: 176). However, if the speaker bases her statement on partial direct evidence, -*chus hina* – a marker which occurs in complementary distribution with other evidential enclitics (Faller 2006) and is glossed below as 'result' – is preferred over -*chá*. Consider the following example from Cuzco Quechua:

### Karolina Grzech

	- a. *?Unqusqachá* unqu-sqa-chá sick-PRT-CONJ *kashanman.* ka-sha-n-man be-PROG-3-COND 'She may be sick.' [Context: Marya looks very pale.]
	- b. *Unqusqachus* unqu-sqa-chus sick-PRT-RES *hina* hina *kashanman.* ka-sha-n-man be-PROG-3-COND 'She appears to be sick.' (Faller 2007: 4)

The marker -*chus hina*/*chu shina* means roughly 'I guess'/'I think'/'apparently' (Faller 2006: 3). In Cuzco Quechua, -*chá* cannot be used if the speaker is certain that the proposition is true or false, even if she arrived at that conclusion through reasoning (Faller 2007: 5). This supports the epistemic modal analysis of -*chá*. For a -*chá*-marked proposition to be felicitous, the speaker needs to believe in the possibility that the proposition expressed is true and needs to have arrived at that belief by her own reasoning.

The analyses of =*mi* and =*cha* presented above cannot be applied to their Upper Napo Kichwa cognates. In the case of =*mi*, neither the 'direct evidential' nor the BPG analysis account for all uses of the marker. The following examples show that the marker can be used in statements based on different evidence types, including direct/visual source of information, as in (3), inference or conjecture, as in (4) and (5), or reportative evidence, as in (6):


*[Cesar]* [Cesar] [Cesar] *mingamami* minga-ma=**mi** collective.work-DAT=**MI** *rishka.* ri-shka go-ANT

'[Cesar] went to the minga' [the speaker has just arrived at C's home. C. is not there, it's the day of the minga, and the speaker knows C. always participates in community events] [el\_21092014\_01 051]

### 2 Epistemic primacy, Common Ground management & epistemic perspective

(5) Inference/conjecture *Ñuka* ñuka 1SG *yaya* yaya father *shamuwmi* sham-u=**mi** come-PROG=**MI** *yachin.* yachi-n seem-3

> 'It seems my father is coming.' [speaker hears footsteps outside, and was expecting his father to come] [el\_21092014\_01 035]

(6) Reportative evidence

*Rimawanun* rima-wa-nun say-1OBJ-3PL *Saida* Saida NAME *ungushkami* ungu-shka=**mi** fall.ill-ANT=**MI** *sirik* siri-k stay-NMLZ *nisha.* ni-sha say-COREF 'They tell me Saida is ill.' [el\_25092014\_01 113]

Example (3) can easily be reconciled with both the direct evidential and the BPG analysis of =*mi*. Example (6), where =*mi* is used within a reportative construction, could indicate that in this case =*mi* can be analysed as a marker of BPG, as the fact which the speaker reports is not accessible to him directly, and thus the best possible evidence is that of a verbal report. However, according to the analyses presented above, examples (4) and (5) could not be marked by =*mi* in any of the varieties of Quechua discussed there. Rather, they would be marked by =*cha*, as in both cases the speaker arrives at the proposition expressed by the utterance by reasoning. Moreover, as shown below, the Upper Napo Kichwa =*mi* is also felicitously used when the speaker cannot be certain of having the BPG:

(7) Guesswork / Inference based on partial evidence *Lluki* lluki left *puramami* pura-ma=**mi** side-DAT=**MI** *rin,* ri-n go-3 *llukipurama...* lluki-pura-ma left-side-DAT '[the seed] goes to the left, to the left' [el\_03102014\_01 076]

The utterance in (7) comes from an elicitation session during which two participants watched recordings of a magician performing six games of a three-shell game. For each game, they first watched it without the final part in which the location of the seed was revealed, and were asked to guess where the seed was. After they took their guesses, the same recording was re-played again, this time with the finale, so that the participants could confront their guesses with reality.<sup>12</sup> The participant who uttered (7) was guessing the location of the seed. He

<sup>12</sup>This elicitation session is accessible online in the ELAR Upper Napo Kichwa deposit, under the following link: https://elar.soas.ac.uk/Record/MPI1034554.

### Karolina Grzech

had already made several attempts at guessing where the seeds were located, but each of them failed. Thus, the use of =*mi* in (7) cannot be explained by the speaker's certainty of having the BPG for the proposition. In §4, I demonstrate that conversely to the direct evidence/BPG analysis, the epistemic primacy analysis of =*mi* can account for all the examples given above.

The Upper Napo Kichwa =*cha* also does not lend itself well to the analyses proposed for its cognates. If it was in fact a conjectural/inferential marker, we would expect to find it in examples such as (4) and (5) above, which are, however, marked by =*mi*. Moreover, =*cha*-marked statements, when presented without further context, are always interpreted as interrogative, rather than declarative utterances. Consider:

(8) a. *#Tamiashkacha.*

tamia-shka=**cha** rain-ANT=**CHA** Intended meaning: 'It rained/It must have rained' [speaker haven't seen the rain, but sees the ground is wet] [elicited]

b. *Tamiashkacha?* tamia-shka=**cha** rain-ANT=**CHA** 'Has it rained? / It has rained, hasn't it?' [elicited]

The example (8a) was presented to consultants in an elicitation context, but they rejected it, proposing (8b) instead. If =*cha* was a conjectural evidential, its use in a declarative statements would not be problematic, as suggested by evidence from other varieties of Quechua.

Although =*cha*-marked declaratives are not attested in elicitation, they do occur in naturalistic discourse, in utterances which can be interpreted as rhetorical questions, as in (9), or dubitative statements, as in (10):

(9) Rhetorical question


### 2 Epistemic primacy, Common Ground management & epistemic perspective

(10) Dubitative

*Ima* ima what *shutiracha,* shuti-ta=**cha** name-ACC=**CHA** *Shangricha* Shangri=cha NAME=CHA *nijkuna* ni-k-kuna say-NMLZ-PL *akay...* a-ka=y AUX-PST=EMPH 'What was his name, I think they called him Shangri…' [in\_26052013\_02 132]

Example (9) is an excerpt from a conversation between two farmers, each of whom owns a plot of land on the shore of the Napo river, in roughly the same area. Although the speaker has first-hand knowledge about the quality of the soil on the shore, he still chooses to use =*cha*, possibly because the plot of land under discussion belongs to the other farmer. Example (10) further shows that Upper Napo Kichwa =*cha* does not lend itself to the analyses proposed for its cognates. In this case, the speaker is an elderly woman trying to remember the name of a shaman she used to know in her youth. Her interlocutor is a young interviewer, who could not have possibly known the shaman. Here, =*cha* is used despite the fact that the speaker has direct, though partial evidence for the proposition, as she knew the shaman in person. Thus, as shown in example (2) above, in Cuzco Quechua that same example would have been marked by -*chus hina*, rather than =*cha*. As in the case of =*mi*, the occurrences of =*cha* in the examples cited in this section cannot be accounted for by the evidential analysis, but can be explained if the marker is analysed as related to epistemic primacy. I discuss this analysis in detail in the following section.

### **4 Epistemic primacy in Upper Napo Kichwa discourse**

In the previous section, I have shown why the evidential analysis does not apply to the Upper Napo Kichwa enclitics =*mi* and =*cha*. Here, I spell out their analysis as markers of epistemic primacy. Firstly, I discuss the semantic contribution that the two markers make to the clause ( 4.1). Secondly, I explain how their epistemic primacy semantics can be reconciled with their association with focus (4.2).

### **4.1 Epistemic primacy semantics of =***mi* **and =***cha*

As mentioned in § 1.3.2, epistemic primacy can be defined as a "relative right to know or claim" (Stivers et al. 2011). As such, the notion operates on a different level of discourse than that of evidence. Evidence or access to information by default relates to the text-external world. The same does not hold for epistemic primacy, which is more social or interpersonal. Conviction of having superior

### Karolina Grzech

knowledge or right to claim something can of course also be grounded in the discourse-external reality, but that need not always be the case. Claims to superior knowledge can arise on the basis of subjective mental processes unrelated to external discourse contexts. Consequently, the notion of epistemic primacy is also broader. All the instances of the speaker having direct, sensory access to events can be explained as cases of evoking both direct evidence and epistemic authority/primacy, which arises by virtue of direct access to an event. However, cases such as the =*mi*-marked guess in example (7) can be explained by the speaker invoking epistemic primacy, but not by the direct evidence or BEST POSSIBLE GROUND analysis (see §3.2).<sup>13</sup>

As mentioned in §1.3.2, an important component of the epistemic primacy meaning is its relative nature; the "relative right to know or claim" certain pieces of information arises as an interpersonal aspect of the context of the speech situation. Let us imagine a situation in which I explain the meaning of the Upper Napo Kichwa word *ayllu* ('family'/'relatives') to a fellow linguist who works on another language family. In this encounter, I could safely assume epistemic primacy, since I am more qualified to talk about Quechuan than my colleague. The knowledge about Quechuan languages falls within my TERRITORY OF INFORMA-TION (cf. Kamio 1997). However, if I were discussing the meaning of the same lexeme with any native speaker of Upper Napo Kichwa, it would be them, not me, who would hold epistemic primacy, by virtue of their native knowledge of the Kichwa language and culture. It follows that to claim epistemic primacy we do not need to be extremely highly qualified, or even certain about the veracity of the proposition; we do need to assume however that we are more qualified to voice an opinion than our interlocutor is. If we assume that Upper Napo Kichwa =*mi* encodes epistemic primacy, examples of its use in utterances (3) to (7) as presented in § 1.3.2 can all be accounted for.

Another aspect of the meaning of =*mi* which was not mentioned so far is that it undergoes ORIGO SHIFT. That is, it is anchored to the speaker in declarative clauses, and to the hearer in interrogative clauses - hence the use of the term *origo* rather than *speaker* in reference to the source of epistemic authority indexed by =*mi*. Consider:

<sup>13</sup>In this particular case, "guesswork" could be regarded as inference based on partial evidence, rather than no evidence at all – despite the earlier erroneous guesses, the speaker might be conviced that this time their observation of the seed's location is correct (I am grateful to Kasper Boye for pointing this out). This does not invalidate the main point of the argument – that previous analyses do not hold for Upper Napo Kichwa =*mi*. As discussed earlier, the Cuzco Quechua =*mi* cannot be used in case of only partial direct evidence, where -*chus hina* is preferred.

### 2 Epistemic primacy, Common Ground management & epistemic perspective


In (11), the speaker introduces herself, and it is clear that the epistemic authority to give this type of information lies exclusively with her. In case of (12), however, it is not the speaker but the hearer who is the source of epistemic authority. Origo shift is not a singular feature of the Upper Napo Kichwa epistemic marking - it occurs in a great many evidential/epistemic marking systems described to date (cf. e.g., Hargreaves 2005).

I turn now to the semantics of the Upper Napo Kichwa =*cha*. In the previous section, I have shown that it is not satisfactorily analysed as a conjectural/inferential evidential. Instead, I propose to analyse the marker as indicating the speaker's lack of epistemic primacy. This means that by using =*cha* the speaker indicates she doesn't have the "relative right to know or claim". This analysis is compatible with the examples (9) and (10) above, since having no epistemic primacy is compatible with doubt. However, the use of the marker in example (8), which demonstrates that in the absence of context, =*cha*-marked utterances are interpreted as interrogatives, requires additional explanation, which I provide in § 5.

The examples of =*cha* given so far do not provide sufficient evidence for claiming that the marker indicates lack of epistemic primacy, as they would also be permissible if the marker simply had a dubitative meaning. What differentiates the dubitative from the "lack of epistemic primacy" expression is that, by using the first, the speaker indicates that he considers the proposition in question as a possibility. In the latter case, the speaker indicates that she is not in a position to evaluate the probability of the proposition. The Upper Napo Kichwa data suggest that this in fact the case. In §3.2 I have shown that Upper Napo Kichwa dubitative statements are marked with the epistemic modal *yachin* 'seem-3', potentially accompanied with =*mi*, rather than with =*cha* (see example 5). Consider:

### Karolina Grzech

	- b. *Yaya* yaya father *yachin,* yachi-n, seem-3 *paywa* pay-pa 3SG-GEN *yayacha…[?]* yaya=**cha** father=**CHA** 'It seems [he is the] father, isn't he his father?/is he his father?' [elicited]

The difference between (13a) and (13b) lies in the fact that in example ( 13a) the speaker presents his own point of view, supposing that the farmer from the "Pear story" video (Chafe 1980) is the father of the boy who comes over to steal the fruit. In (13b), the point of the utterance is different - the speaker first expresses a supposition that the man in the film is the boy's father using the epistemic modal, and then acknowledges the lack of "right to claim", turning to the addressee for the evaluation of the proposition. Examples (14) and (15) further support the claim that =*cha* encodes lack of epistemic authority, rather than doubt:


In (14) and (15) =*cha* occurs on interrogative pronouns in rhetorical content questions, rather than marking propositions the veracity of which is not evident to the speaker, as a dubitative marker would. This is clear in (15), where the first clause of the utterance is marked with =*cha* to show the lack of epistemic authority, and it is only in the second clause that the speaker makes a supposition, marking it with *yachin* 'seem-3'. The above examples support the analysis of the marker as indicating that by using it the speaker assumes the position of lacking the knowledge necessary to evaluate its veracity.

There are two other important aspects of the meaning of =*cha* which should be mentioned here. Firstly, unlike =*mi*, Upper Napo Kichwa =*cha* does not undergo

### 2 Epistemic primacy, Common Ground management & epistemic perspective

origo shift; the "lack of epistemic primacy" is always anchored to the speaker, irrespective of whether the utterance is interrogative or declarative. The research to date does not suffice to explain why this is the case.<sup>14</sup> Further research is also needed to clarify the motivations of speakers when they choose to use =*mi* and =*cha* in questions. It also remains to be established whether such choices are motivated by the content of the utterance, the epistemic context of discourse, or considerations related to politeness. Initial observations indicate that speakers resist using =*mi* when asking questions concerning the addressee. In second person interrogatives =*cha* or the negative/interrogative marker =*chu* are strongly preferred. When I tried uttering declarative clauses with second person subject, my interlocutors corrected me, suggesting that what I wanted to say was actually a =*chu* or =*cha*-marked question. As shown in (12) above, the use of =*mi* is not an issue in third person questions.

Lastly, it is important to note that the "lack of epistemic primacy" meaning does not always entail that the speaker's epistemic access is inferior to that of the interlocutor. The examples (13) through (15) above suggest that =*cha* is used in discourse in such contexts, and this is most often the case. However, this is not the case in (10), repeated below for the sake of clarity:

(10) Dubitative

*Ima* ima what *shutiracha,* shuti-**ta**=**cha** name-ACC=**CHA** *Shangricha* Shangri=**cha** NAME=**CHA** *nijkuna* ni-k-kuna say-NMLZ-PL *akay...* a-ka=y AUX-PST=EMPH 'What was his name, I think they called him Shangri…' [in\_26052013\_02 132]

As mentioned above, in this case the speaker is trying to remember an event from her youth, and marks the propositions she is not sure about with =*cha*. However, her interlocutor is a young person who could not have known the shaman whose name the speaker is trying to remember. Therefore, a more plausible analysis is that the speaker uses =*cha* to indicate that she does not have epistemic primacy in this case, although as the narrator and witness of the events she is recounting, she could be expected by the interlocutors to have it. Although examples like this one need further analysis, they are still in line with the main point

<sup>14</sup>The currently available data suggest that unlike =*mi*, the meaning of =*cha* does not assign epistemic primacy to a particular participant of the interaction. Rather, it serves to indicate that the speaker does not hold it. This aspect of analysis needs to be developed in more detail, but it would explain why =*cha* does not partake in origo shift (I am grateful to Henrik Bergqvist for this observation).

### Karolina Grzech

presented here – that Upper Napo Kichwa =*cha* can be analysed as a marker of "lack of epistemic primacy" on the part of the speaker.

In this section, I have shown that the two Upper Napo Kichwa markers described in this chapter, namely =*mi* and =*cha*, can be analysed as encoding "epistemic primacy of the origo" and "lack of epistemic primacy of the speaker", respectively. In the following section, I discuss the relationship between the markers' semantics and their association with focus.

### **4.2 Epistemic primacy and focus**

In Table 2.1 (§2), I show that the cognates of the markers =*mi* and =*cha* in other Quechuan varieties were often analysed as focus marker (cf. e.g., Muysken 1995; Sánchez 2010; 2015). In the same section, I mention that in Upper Napo Kichwa, =*mi* and =*cha* seem to be associated with focus, but are too infrequent to be plausibly analysed as focus markers. I have discussed the interaction of focus with =*mi* and =*cha* in detail elsewhere (Grzech 2016a: ch.4). Here, for the sake of space, I limit the discussion to those aspects of the markers' association with focus that are relevant to describing the interdependence between their focusrelated function and their epistemic primacy-related semantics.

The marker =*mi* can occur in different types of focus structures, but it seems to only be obligatory for the felicity of utterances which contain contrastive focus. Consider:


Contrastive focus, as understood here, exhibits two main properties: (1) having a set of identifiable alternatives (cf. Kiss 1998; Repp 2010) and (2) implying the rejection of those alternatives (Repp 2010: 1336). Both apply to (16) above and (17) below, which the speakers also resisted accepting without =*mi*:


The contrastive constructions exemplified in (16) and (17) are also corrective, that is, the alternatives to the focused constituent are not only identifiable, but

### 2 Epistemic primacy, Common Ground management & epistemic perspective

also overt (cf. Repp 2010). Correction involves rejection of an alternative proposition that is currently under discussion, or which the hearer assumes to form part of CG. The =*mi*-marked constituents are introduced to the CG in the aftermath of the rejection of the =*chu*-marked alternatives. Consequently, in both cases, propositions which one of the speakers considered true before the exchange are rejected, and replaced with the new, =*mi*-marked content. The fact that =*mi* is required for the felicity of such constructions is fully compatible with the marker's 'epistemic primacy' semantics. By uttering (16) and (17), the speaker contradicts a proposition the hearer held as true, and introduces a new one, the factuality of which was outside the interlocutor's awareness. Hence, in both cases, it is the speaker who holds the epistemic primacy. This would explain the fact that =*mi* is required for the felicity of this type of utterances. The marker can also occur in information focus structures, but consultants deem those equally permissible with and without =*mi*. The non-obligatoriness of the marker in such contexts plausibly stems from the fact that in the case of information focus it is not always clear - as it is in cases of corrective, contrastive foci - that the interlocutor was neither aware of, nor expecting, <sup>15</sup> the focal content of the utterance.

Let us compare the above to the correlation of focus with =*cha*, which occurs only in information focus contexts. Consider the following exchanges:

	- b. *Ayajtá* ayaj=**tá** bitter=**TÁ** '[it IS] bitter' [in\_05092014\_01 033-34]

<sup>15</sup>Expectation is another factor which seems to be relevant to the occurrence of =*mi* in Upper Napo Kichwa. For reasons of space, I do not discuss this in detail here, but a discussion of its role in the distribution of =*mi* is provided elsewhere (cf. Grzech 2016a: ch.6).

Karolina Grzech

> b. *Shindzimi* shindzi=**mi** strong=**MI** '[It is] strong' [in\_20092013\_01 186-87]

In the question-answer pairs in (18) and (19), =*cha* is used to indicate the focus of the interrogative clauses. In the corresponding answers, the focus is marked with the verum focus marker =*tá* and with =*mi*, respectively. The enclitic =*cha* is not syntactically obligatory in the contexts shown above, but (quasi- )interrogatives are its most frequent context of use. The fact that it tends to indicate information focus in rhetorical and confirmation questions is compatible with its "lack of epistemic primacy" semantics; marking the focus of interrogative utterances is tantamount to highlighting the portion of the utterance about which the speaker does not have sufficient knowledge. Hence, she is requesting more information from the addressee. Therefore, highlighting focus in dubitative/interrogative contexts and marking the speaker's "lack of epistemic primacy" can be regarded as complementary aspects of the discourse function of =*cha*.

### **5 Epistemic primacy and Common Ground management**

The discussion so far suggests that both =*mi* and =*cha* are polyfunctional markers, encoding meanings related to epistemic primacy and associated with the IS category of focus. Consequently, a question that emerges is whether these two aspects of their semantics could be considered jointly, so as to arrive at a unified analysis of the markers' function in discourse. In the paragraphs that follow, I propose that such an analysis is possible if we take the notion of Common Ground into account.

In § 1.3.4, I defined the notions of COMMON GROUND (CG) and CG MANAGEMENT. I mentioned that CG develops constantly over the course of communication. Its content expands as the interlocutors put forward new propositions, which they mutually agree to treat as true for the purpose of the exchange (cf. e.g., Stalnaker 1974). However, CG consists not only of its truth-conditional content, but also of CG management, that is, linguistic devices indicating how CG should develop – what propositions should be added to it, and at which point in the conversation. The discussion in the following paragraphs is underpinned by another important assumption: that any update in the CG results from solving a qUES-TION UNDER DISCUSSION (henceforth QUD) (cf. Asher 1993; Matić 2015). QUDs are introduced into discourse to advance communication, often by means of interrogative clauses which require an answer. QUDs influence the development

### 2 Epistemic primacy, Common Ground management & epistemic perspective

of the CG content by directing the communicative exchange towards saturating certain propositions, so that they can then be added to CG.

As mentioned above, the development of CG is also informed by expressions which contribute to CG management, rather than content. These type of expressions are typically non-truth-conditional; they do not add to the proposition expressed, but rather constitute processing clues for the interlocutors, indicating whether and how certain propositions could be integrated into the CG content. I propose that the Upper Napo Kichwa enclitics discussed in this chapter belong to this class of expressions, with the epistemic primacy and information-structural aspects of their meaning contributing to their overarching function of CG management.

Although Quechuan epistemic/evidential markers have not been analysed in this way before, the idea that epistemic meanings and IS meanings are related to each other is not a new one. Kamio (1997: 3) pointed out a conceptual relation between the TERRITORIES OF INFORMATION and IS, as both relate to "the character of information expressed in natural language". In Quechuan literature, these relations also have not gone without notice. Hintz & Hintz (2017) observe that "the inter-subjective nature of certain evidential systems" invites discussion of how evidentiality relates to CG. In her analysis of Cuzco Quechua evidentials, Faller (2002) also mentions in passing that evidential marking is related to IS. In some previous studies (e.g., Muysken 1995), the focus-related and evidential functions of the enclitics were acknowledged, although the relation between the two was not explained.

Upper Napo Kichwa =*mi* and =*cha* can be analysed as markers of CG management not only because of their meaning, but also because of the type of contribution they make to the proposition expressed. Both markers are non-truthconditional, which is confirmed by the fact that their contribution to the propositions expressed by the utterances in which they occur cannot be negated (for discussion, see Grzech 2016a: ch.5; 2016b). It follows that they cannot contribute to the CG content, which is truth-conditional by definition.

Let us discuss the CG management function of =*mi* on the basis of its occurrence in declarative clauses, where the origo is tantamount to the speaker. As mentioned previously, the speakers use =*mi* not only to indicate focus, but also to assert their epistemic primacy. In doing so, they highlight information as congruent with their own world knowledge, but not necessarily with that of the hearer. Consequently, the use of =*mi* indicates that the speaker has information that is sufficient to resolve a certain QUD, and that this information is not shared with the interlocutor. The association of =*mi* with contrastive focus is congruent with this analysis; in examples (16) and (17), the contrastive focus associated with =*mi*

### Karolina Grzech

was also corrective. A correction implies that the speaker assumes being the only participant of the discourse who knows how to answer the current QUD. Consequently, the procedural meaning of =*mi*, integrating both epistemic primacy and focus association, is to encourage the hearer to accept the =*mi*-marked information as part of CG, despite the misgivings they might have about it. Consequently, the marker is a CG-management device, the function of which is to accelerate the integration of a certain proposition into CG. This is illustrated in (20) with an exchange that took place between myself (A) and a good friend of mine (B), when we talked about whether I would be able to prepare *chicha*, a traditional drink made from manioc:

	- b. *Ushanguimi!* usha-ngui=**mi** can-2=**MI** 'Yes you can' [attested]

In the example above, my friend corrected me, using a =*mi*-marked utterance to dispel my doubts about my capability of preparing the drink. In this specific context, the use of =*mi* with second person subject is not aimed to undermine my authority over my own actions or skills – rather, my friend wanted to emphasise the fact that she knows I am skilled, and she also knows how to make *chicha*. She used the =*mi*-marked assertion as a form of encouragement, urging me to re-assess my skills.

The function of =*cha* as a CG management device is a mirror image of the function of =*mi*. As discussed above, Upper Napo Kichwa =*cha* is associated with information focus and indicates that the speaker renounces epistemic primacy. Consequently, in terms of CG development, the use of =*cha* indicates that the speaker does not have the information sufficient to answer the current QUD, and that input is needed from the hearer to integrate the =*cha*-marked proposition into CG. This explains why, as shown in example (8), =*cha*-marked propositions tend to be interpreted as interrogatives in the absence of further context. It is also congruent with the association of =*cha* with information focus, as the enclitic attaches to that portion of the utterance that requires saturation in subsequent discourse. Consequently, the procedural meaning of =*cha* in terms of CG management is to indicate that the current QUD is not yet solved, preventing the

### 2 Epistemic primacy, Common Ground management & epistemic perspective

hearer from accepting the proposition as part of CG without further elaboration. This was shown in examples (18) and (19) above.

In this section, I have shown that the two aspects of the semantics of =*mi* and =*cha* discussed in the previous parts of this chapter can be considered jointly, if we analyse the enclitics as markers of CG management. This suggests that rather than referring to them as markers of epistemic primacy associated with focus, they can be analysed as (part of) a morphosyntactic paradigm the main function of which is to manage the CG. In the following section, I show how this proposal can be integrated into a broader, cross-linguistic functional domain of EPISTEMIC PERSPECTIVE.

### **6 Common Ground management and the epistemic perspective domain**

In §1.3.3, I introduced the notion of the domain of EPISTEMIC PERSPECTIVE (Bergqvist 2017), which brings together epistemic marking systems attested in various languages, proposing that they share a "functional space" in the grammar. The structure of the domain as proposed by Bergqvist was shown in Figure 2.1 in §1.3.3 , and is repeated below with minor modifications, incorporating the subdomain of CG management:

$$\text{score}$$

$$\text{(wide)}$$

KNOWLEDGE (A)SYMMETRY *complex epistemic perspective (CEP) CG management* SPEAKER-HEARER LINKS *illocutionary modality* SPEAKER INVOLVEMENT *egophoricity* INFORMATION SOURCE *evidentiality* POSSIBILITY *epistemic modality* NECESSITY

(narrow)

Figure 2.2: Dimensions of the epistemic perspective domain (based on Bergqvist 2017: 12)

The placement of CG management shown in Figure 2.2 is based on the semantics of the Upper Napo Kichwa CG management markers. As shown in § 4, the Upper Napo Kichwa CG-managing enclitics indicate knowledge (a)symmetry

### Karolina Grzech

between discourse participants, and hence are related to the category of KNOWL-EDGE ASYMMETRY.

Bergqvist (2017) based the bottom-to-top ordering of the different sub-domains of epistemic perspective on the assumption that the more intersubjective meanings correspond to a wider scope of linguistic expressions. This has been confirmed in the literature on grammaticalisation (Traugott 1995; 2003; 2010; 2012; Traugott & Dasher 2002), including grammaticalisation of discourse markers (Traugott 1995) and deictic elements, such as demonstratives (e.g., Kratochvíl 2011). Traugott (e.g., 2010) observes that subjectified elements tend to appear at the periphery of the constituent or clause. This observation applies both to Upper Napo Kichwa CG management markers and to evidentials in other Quechuan varieties (e.g., Muysken 1995; Sánchez 2015).

Although I grouped the markers of CG management together with the marking of COMPLEX EPISTEMIC PERSPECTIVE (CEP) on the basis of both sub-domains relating to knowledge (a)symmetries, it remains unclear how CG management markers relate to CEP. Bergqvist characterises CEP as a subtype of multiple perspective (cf. Evans 2005), in which "one perspective [is] embedded in another" (Bergqvist 2017: 6). In Upper Napo Kichwa, =*mi* – but not =*cha* – was attested to occur in the scope of verbs of speech and thinking, and thus participate in constructions which could be defined as encoding multiple perspectives. Such constructions are used to convey reports about other speakers' utterances, as well as in self-corrections. Consider:

(21) *Muyura* muyu-ta fruit-ACC *pitiwn…* piti-w-n… cut-PROG-3 *ima…* ima… what *Coco…* coco... coconut *Mana,* mana NEG *cocomi* coco=**mi** coconut=**MI** *[an]* [a-n] [COP-3] *nini,* ni-ni... say-1 *cocochá…* coco=**cha**... coconut=**CHA**

'He is cutting [harvesting] fruit….what…[It's a] coconut…No, I said "coconut"… [but is it a] coconut?' [el\_24092014\_03 003-5]

In (21), the speaker's previous opinion about the type of fruit harvested in the "Pear story" video is embedded in the utterance conveying her current opinion. As such, the example could be analysed as an instance of CEP. Note that =*cha* is also used in (21), but not within the complex perspective construction – the speaker marks her new opinion with that enclitic, having realised she is not able to name the fruit she is seeing.

In the light of the above discussion, the tentative positioning of the category of 'CG management' in the vicinity of CEP seems to be accurate. However, it

### 2 Epistemic primacy, Common Ground management & epistemic perspective

requires further research into the scope of the Upper Napo Kichwa CG management markers. As mentioned above, the bottom-to-top ordering of the different sub-domains in Figure 2.2 is based on the assumption that the scope of the linguistic expression of these domains is narrow at the bottom and gets wider towards the top. Consequently, the expressions of CEP are supposed to scope over all the other types of markers. The scope properties of =*mi* and =*cha* need to be thoroughly investigated before a definite decision on the placement of the sub-domain of CG management can be made.<sup>16</sup>

Consequently, the discourse-semantic considerations presented above need to be investigated in more detail before definite conclusions can be reached about where CG management fits into the epistemic perspective domain. The relationships between the different sub-domains of epistemic perspective are also likely to be revised in the light of ongoing documentation of the world's languages.

In sum, the Upper Napo Kichwa markers of CG management functionally and semantically fit within the functional domain of epistemic perspective postulated by Bergqvist (2017). The data from Upper Napo Kichwa also show that the domain is not complete in the shape in which it was proposed originally.<sup>17</sup> Determining the place of the sub-domain of CG management within the general domain of epistemic perspective, as well as its relationship with other sub-domains, requires further, cross-linguistic research. In this context, the study of the Upper Napo Kichwa CG management enclitics brings us one step closer to describing and analysing the cross-linguistically valid functions and properties of epistemic marking systems across languages.

### **7 Summary and conclusions**

Over the course of this chapter, the reader has become acquainted with the system of marking epistemic primacy in Upper Napo Kichwa, exemplified by the enclitics =*mi* and =*cha*. I have shown that the two Upper Napo Kichwa enclitics

<sup>16</sup>The scope of the markers is not discussed in this chapter, but I do tackle it to some extent in my previous work (Grzech 2016a: ch.5). Preliminary investigation shows that =*mi* can be embedded under verbs of speech and thinking, and under certain modals, but not under negation. The marker =*cha*, on the other hand, was not attested in embedded contexts. Both markers are ungrammatical with overtly marked imperatives.

<sup>17</sup>In the time between when this paper was first written and its going into print, a category of 'engagement' was put forward, which encompasses CEP and a broader range of phenomena related to epistemicity (Evans et al. 2018). While the analysis presented here does not take that development into account, it will have to be addressed in future studies of Upper Napo Kichwa epistemic marking.

### Karolina Grzech

belong to a larger paradigm, and that they can be analysed as marking epistemic primacy and lack thereof, respectively. I have also shown that although the enclitics are associated with focus, they cannot be considered as focus markers. Subsequently, I have demonstrated that the epistemic primacy semantics of the markers and their relation to focus can be brought together under one overarching function: CG management. Finally, I have discussed the place of CG management within the cross-linguistically valid epistemic perspective domain.

Numerous aspects of the Upper Napo Kichwa CG management markers remain to be investigated. First and foremost, although the considerations presented in this chapter help account for the occurrence of both enclitics in discourse, they are not sufficient to account for their non-occurrence. More factors, including (counter)expectation (cf. Matić 2015), need to be taken into account in the future research into determining why the Upper Napo Kichwa enclitics are only attested in a very small percentage of the utterances in which they could grammatically occur. As for =*cha*, it is still unclear what role politeness plays in motivating its occurrence in discourse. Furthermore, it remains to be explored how the epistemic markers attested in undescribed languages such as Upper Napo Kichwa behave in discourse. Data from such languages have yet to be integrated into the discourse-analytical research concerned with how epistemic meaning is expressed through discourse strategies and discourse structure, which has been on the rise since the publication of the seminal paper of Heritage & Raymond (2005).

Nonetheless, the analysis of =*mi* and =*cha* presented here contributes to our current knowledge of epistemic marking systems in a variety of ways. The chapter adds to our current knowledge of systems encoding epistemic meanings in Quechuan languages. As mentioned above, in most described Quechuan varieties, the cognates of the enclitics described in this chapter are analysed as evidential markers. Consequently, the analysis provided here adds to our crosslinguistic knowledge of evidential and related systems. The chapter also contributes to the description of epistemic primacy as a semantic category. Relatively little work has been done cross-linguistically on morphosyntactic systems and linguistic expressions encoding epistemic primacy (cf. e.g., Stivers et al. 2011; Hayano 2011). As a result, the description of the Upper Napo Kichwa system contributes to what we know about the category's cross-linguistic properties. The preliminary discussion of the relation of the Upper Napo Kichwa epistemic primacy markers to CG management in discourse, and to the epistemic perspective domain (Bergqvist 2017) make the analysis presented here relevant to the study of linguistic devices encoding epistemic meaning in a cross-linguistic perspective.

### 2 Epistemic primacy, Common Ground management & epistemic perspective

### **Acknowledgments**

I would like to acknowledge the generous support granted to this research by ELDP (Endangered Languages Documentation Programme, grant no. IGS0166). The data collected during from the ELDP-financed documentation project were deposited in the ELAR archive and can be accessed under: https://elar.soas.ac.uk/ Collection/MPI849403.

I would like to thank Lydia Chimbo and her family for giving me a home away from home in Ecuador, and all the inhabitants of Nuevo Paraíso for supporting the documentation project. In particular, I would like to acknowledge the work and support of Nilo Andy, Jacobo Chimbo and Wilma Aguinda. Lastly, I am grateful to two anonymous reviewers and to the editors of this volume for their comments and suggestions. All shortcomings of this work remain my own.

### **Abbreviations**


### **References**


### Karolina Grzech


2 Epistemic primacy, Common Ground management & epistemic perspective

Givón, Talmy. 1981. Typology and functional domains. *Studies in Language* 5(2). 163–193.


### Karolina Grzech


2 Epistemic primacy, Common Ground management & epistemic perspective


# **Chapter 3**

# **Egophoricity, engagement, and the centring of subjectivity**

### Alan Rumsey

Australian National University

In egophoric systems formal patterns that are associated with first person subjects in declarative sentences are associated with second person subjects in questions. This difference in formal patterning is associated with a difference in the centring of subjectivity, whereby, for example, epistemic authority regarding the state of affairs that is described in a declarative sentence is vested in the speaker, whereas in a question it is vested in the addressee. Such egophoric patterning is but one instance of a wide range of phenomena that involve more-or-less regular shifts in the usual centring of subjectivity as between speaker and addressee. Here I examine three other such phenomena: 1) interactions between person marking and intentional modality; 2) shifts between speaker-centred and addressee-centred kinship terms when used in direct address; and 3) the prompting of children with utterances that are voiced as if from the child's perspective. Evidence is drawn from the Papuan language Ku Waru and from comparison with other languages. I present an extended example of "engagement" in Ku Waru and compare it with 1-3. I show that, while grounded in the same basic aspects of human interaction, it differs from 1-3 in treating the centring of subjectivity as potentially variable, emergent, and distributed across the interaction rather than as prototypically related to the speech roles of speaker and addressee and their alternation across speech-act types.

### **1 Introduction**

In egophoric systems<sup>1</sup> formal patterns that are associated with first person subjects in declarative sentences are associated with second person subjects in ques-

<sup>1</sup> In recent literature on egophoricity including the discussions of it in this volume and the detailed cross-linguistic survey in San Roque & Schieffelin (2018) that term is taken to be synony-

Alan Rumsey. 2020. Egophoricity, engagement, and the centring of subjectivity. In Henrik Bergqvist & Seppo Kittilä (eds.), *Evidentiality, egophoricity, and engagement*, 61–93. Berlin: Language Science Press. DOI: 10.5281/zenodo.3975799

### Alan Rumsey

tions. From a functional viewpoint this difference in formal patterning is associated with a difference in the centring of subjectivity, whereby, for example, epistemic authority regarding the state of affairs that is described in a declarative sentence is vested with the speaker, whereas in a question it is vested in the addressee. Egophoric patterning of this kind is but one instance of a wide range of phenomena that involve more-or-less regular shifts in the usual centring of subjectivity as between speaker and addressee. In this chapter, I will examine three other such phenomena. I will compare and contrast them with egophoric patterning and ask what, if anything, is special about the latter. Evidence for my argument will be drawn from the Ku Waru language of Highland Papua New Guinea — with particular emphasis on interactions between young children and adults — and from comparison with other languages. The phenomena to be considered include: 1) interactions between person marking and intentional modality; 2) shifts between speaker-centred and addressee-centred kinship terms when used in direct address; and 3) the prompting of children with utterances that are voiced as if from the child's perspective. Finally I will present an example of the nascent grammatical category of "engagement" in Ku Waru and discuss what it has in common with 1) - 3) and how it differs from them.

### **2 Egophoricity**

As a typical case of egophoricity let us consider that of Northern Akhvakh as described by Creissels (2008). Akhvakh is a Nakh-Daghestanian language spoken in western Dagestan. It has both nominal case marking and gender-number agreement marking on the verb, both of which show ergative-absolutive alignment. Within the perfect tense/aspect Akhvakh has a formally marked distinction between the presence *vs* absence of what Creissels calls "assertor's involvement". He defines the "assertor" as "the speaker in statements and the addressee in questions" (2). The relevant criterion of "involvement" in Northern Akhvakh is a fully grammaticalized one: the "involved" participant is the A argument of a transitive verb or the S argument of a lexically specified subclass of intransitive verbs the subjects of which are typically volitional. This is illustrated in (1) and (2) by the distribution of the verb suffixes -*ari* and -*ada* for +/- "Assertor's Involvement" (ASSINV) (where "non-involvement" is signalled as the residual alternative by the Perfective suffix (PFV) alone).

mous with what has in the past more often been called 'conjunct-disjunct patterning'. Here I follow the above-cited literature in treating the terms as synonyms, and in using 'egophoricity' in preference to 'conjunct-disjunct patterning', noting, as do San Roque & Schieffelin, that this is a distinct usage of 'egophoric' from that of Hagège (1974) and Dahl (2001; 2008).

3 Egophoricity, engagement, and the centring of subjectivity

	- b. *me-de* 2SG-ERG */* / *hu-sw-e ̱* DEM-OM-ERG */* / *hu-λ̱-e* DEM-O<sup>F</sup> -ERG *kaʁa* paper *q̵war-ari* write-PFV 'You / he / she wrote a letter.'
	- c. \*de-de kaʁa q̵war-**ari**.
	- d. \*me-de / \*hu-s̱w-e / \*hu-λ̱-e kaʁa q̵war-**ada**. (Creissels 2008: 1)
	- b. *de-de* 1SG-ERG */* / *hu-sw-e ̱* DEM-OM-ERG */* / *hu-λ̱-e* DEM-O<sup>F</sup> -ERG *čũda* when *kaʁa* paper *q̵war-ari?* write-PFV 'When did I / he / she write a letter?'
	- c. \*me-de čũda kaʁa q̵war-**ari**?
	- d. \*de-de / \*hu-s̱w-e / \*hu-λ̱-e čũda kaʁa q̵war-**ada**? (Creissels 2008: 1)

In a wide-ranging survey of grammatical patterning of this kind San Roque & Schieffelin (2018: 4) refer to it as *egophoric distribution*. This is shown in Table 3.1, which they have adapted from Hale & Watters (1973), who originally interpreted egophoricity (or what they called conjunct/disjunct marking) as a kind of person marking.

Table 3.1: Typical distribution of egophoric and non-egophoric markers with respect to person and sentence type (after San Roque & Schieffelin 2018: 5)


A complicating factor in languages that have this kind of system is that in reported speech, including indirect discourse, the egophoric marker is typically

### Alan Rumsey

used as if grounded in the "reported" speech situation rather than the "reporting" one. An example from Akhvakh is shown in (3).

(3) a. *ha* DEM *ĩgora* bread *de-de* 1SG-ERG *magazi-gune* shop-EL *b-eχ-e* N-buy-CVB *j-eq'-ada.* F-come-PFVASSINV '**I brought** this bread from the shop.' b. *ilo-de* mother -ERG *eƛ̱'-iri* tell-IRR *waša-su-ga, ̱* boy-OM-LAT *ha* DEM *ĩgora* bread *ĩ-λ̱-e* ANA-O<sup>F</sup> -ERG *magazi-gune* shop-EL *b-eχ-e* N-buy-CVB *j-eq'-ada.* F-come-PFVASSINV 'The mother told the boy that **she had brought** this bread from the shop.'

(Creissels 2008: 3)

The common denominator between the reported speech cases as in (3b) and the simpler, non-embedded ones as in (1), (2) and (3a) is that egophoric marking, or what Creissels calls "assertor's involvement" marking is associated with the participant who ostensibly has what Hargreaves (1991) refers to as "epistemic authority" concerning the event that is being described. In (1), (2) and (3a) that is the speaker in the primary speech event. In (3b) it is the speaker in the reported speech event, the mother.

In most languages with egophoric marking its association with epistemic authority is overridden in certain contexts, but languages vary both in the extent to which this happens and in the contexts where it happens. One such context for some languages is rhetorical questions. This is illustrated from Akhvakh in (4),<sup>2</sup> where it can be seen that in rhetorical questions the same pattern as in true questions applies, despite the fact that in the case of the rhetorical questions the addressee is not really being treated as the epistemic authority.

(4) *de-de čũda eƛ̱'-ari ha-be?*

1SG-ERG when say-PFV DEM-N

1. 'When did I say that?' – I don't remember, perhaps you do (true question)

2. 'When did I say that?' – You should know that I never did (rhetorical question) (Creissels 2008: 11)

<sup>2</sup> In the original source (Creissels 2008: 8) the interlinear gloss of *de-de* shows 2SG in initial position rather than 1SG. Creissels (personal communication 6 October 2016) has confirmed that this is a mistake and has corrected it in an updated version of the original Word doc file that he has kindly provided me with.

### 3 Egophoricity, engagement, and the centring of subjectivity

In other languages with egophoric marking, including Newar (Hale & Watters 1973: 249) and Awa Pit (Curnow 2002: 614–615), questions with first and third person subjects routinely appear with egophoric marking when they are rhetorical questions and otherwise with non-egophoric marking, which more accurately tracks the epistemic authority.

Another kind of variation within languages with egophoric marking is in how, and to what extent, the semantics of volitionality are taken into account as a potentially overriding factor.

This is illustrated from Newar by (5).

	- b. *ji* 1.ABS *pyāt-a* be.wet-PST.DJ
	- c. *jī:* 1.ERG *(tha:-yāta)* (self-DAT) *pyā-k-ā* be.wet-CAUS-PST.CJ 'I got (myself) wet.' (Hargreaves 2005: 29)

By contrast, within Northern Akhvakh, for any given verb there is no choice of marking its subject as either egophoric or non-egophoric in accord with its volitionality or any other contextual factor. As Creissels puts it, "Northern Akhvakh is a nearly perfect example of a fully syntacticized system of assertor's involvement marking, in the sense that, when the assertor of a clause in the perfective positive coincides with the A/S argument of a verb encoding a controllable event, the omission of assertor's involvement marking is very exceptional" (Creissels 2008: 12–13).

### **2.1 Egophoric distribution and evidentiality**

As will be evident from the above discussion, the kind of involvement in a predicated event or state of affairs that is deemed to be relevant for egophoric marking is typically treated as a matter of epistemic authority – the presentation of oneself as *knowing about* that event or state of affairs. As was also exemplified above, a related kind of involvement that also figures in many egophoric systems is volitionality. This pertains to the difference between actions or states of affairs that have putatively been intentionally performed or brought about by the referent of the A or S argument and ones that were not under his/her control. But it is important to note that there are other linguistic phenomena which may not qualify as egophoricity *per se*, but which nonetheless show the same kind of alternation

### Alan Rumsey

in the centring of subjectivity as the one between questions and statements that is shown in Table 3.1. One such area is evidentiality, as demonstrated in detail by Aikhenvald (2004) and by San Roque et al. (San Roque et al. 2015, San Roque & Schieffelin 2018), who show how it overlaps and interacts with egophoricity in that respect. This is exemplified in (6) from Duna, a Trans-New Guinea language spoken in the Southern Highlands of Papua New Guinea.

	- b. *Keni* PSN *hutia=pe?* come.PFV.VIS.P=Q 'Has Kenny come? (according to your visual evidence?)'
	- c. *Rodni* PSN *kho* 3SG *ayu* now/today *hutia* come.PFV.VIS.P *ri-tia* say-PFV.VIS.P '[Someone] said Rodney came today (according to their visual evidence).' (San Roque & Schieffelin 2018: 56, from San Roque 2008 and field notes)

As can be readily seen from this example, the patterning of the "previous visual evidence" (VIS.P) evidential category shows the same shift in the imputed knower of the evidence across statements, questions and reported speech as did the Akhvakh egophoric or "asserter's involvement" marker in examples (1), (2), and (3) respectively. This correspondence seems expectable (once it has been discovered!) in view of the fact that both evidentiality and egophoricity have to do with *knowledge*, and the speech-act participants' relation to it – evidentiality having to do with their *sources* of knowledge and egophoricity with their *presumed authority* over it or lack thereof. But interestingly, the same shift is also found with respect to other linguistic categories that pertain to another kind of "assertor involvement" besides knowledge, namely *intention*. In order to illustrate this I now turn to a discussion of intentional modality, beginning with an example from Ku Waru.<sup>3</sup>

<sup>3</sup>As explained in §2.2, by 'intentional' modality I refer to the modal categories that express an intention or desire on the part of the speaker or other relevant 'modal subject' (optative, imperative, hortative, etc.). I take this to be synonymous with what is sometimes also called 'volitive' modality, but not with 'volitionality' in the sense that that term is most often used by linguists, for reasons discussed below.

### 3 Egophoricity, engagement, and the centring of subjectivity

### **2.2 Egophoric distribution and intentional modality**

Ku Waru<sup>4</sup> has eight TAM categories that are indicated by verb suffixes along with seven person-number categories. One of the TAM categories is what we call the "Optative" mood. The suffixes that mark it are shown below with rough English glosses.


Some examples of optative 1SG -*ab* are:


'Now I want to hear your (PL) final determination' [lit: …hear the thing that you have said finally].

(9) *ab* woman *ilyi* this *ly-ab* get-OPT:1SG 'I want to marry this woman.'

<sup>4</sup>Ku Waru is spoken in the Western Highlands Province of Papua New Guinea and belongs to a dialect continuum that includes what Ethnologue classifies as four distinct languages: Melpa, Mbo-Ung, Imbonggu, and Umbu-Ungu. In those terms, Ku Waru belongs to the Mbo-Ung language (ISO code *mux*). My collaborator Francesca Merlan and I use 'Ku Waru' in preference to 'Mbo-Ung' because it has more salience for the local people as a name for a regional speech variety. Mbo-ung, by contrast, just means 'Indigenous language' and does not correspond to a territorially bounded speech variety at the level of *a* language.

### Alan Rumsey

When a 1SG optative verb is used in a question, the usual modal origo or relevant centre of intention shifts from speaker to addressee.<sup>5</sup> Examples are:

(10) *na* I *lku* house *suku* inside *w-ab-i* come-OPT:1SG-Q 'Can I come into the house?', 'Is it o.k. with you for me to come into the house?'

(11) *s-ab* give-OPT:1SG *mola* or *mol* no 'Shall I give [it to you] or not?'

(12) *na-nga* 1SG-GEN *mai* ground *aprali* seize *te-k* do-NF:2/3PL *lyi-ng-lum* get-PRF:2/3PL-COND *na* I *tena* where *p-ab* go-OPT:1SG 'If they take my land, where I am supposed to go? [i.e. where do you propose that I go?].'

(13) *ab* woman *obi-nga* penis-GEN *aki* that *te-pa* do-NF:3SG *suku* inside *pe-lym* be/lie-HAB:3SG *na* I *to-p* hit/do-NF:1SG *pilyi-lyo* know-HAB:1SG *enebu* tiredness *to-kum* hit/be-PPR:3SG *pilyi-kir-ayl* know-PPR:1SG-DEF *mel-ayl* thing-DEF *ti te-ab mel nar*

another do-OPT:1SG thing what

'When a woman has sex with a man that's how it is, I know; I'm an expert at it. I know it's tiring, but what else can I do?'<sup>6</sup>

<sup>5</sup> For a similar shift of origo in questions involving 2SG hortatives in another Trans-New Guinea Papuan language of the Papua New Guinea Highlands, Duna, see San Roque 2008: 448–450. <sup>6</sup>This example comes from a paternity dispute, a full transcript and analysis of which is presented in Merlan & Rumsey (1986). In order to understand the point of this remark, it is important to know that, at least as of 1983, when the dispute took place, Ku Waru people firmly believed that is impossible for a woman to become pregnant from having sex with a man only once, multiple applications of semen being necessary to stem the flow of menstrual blood and build up the foetus. The woman who was accused of adultery in this case became pregnant at a time when she was living mainly apart from her husband, during which, as they both agreed, they had had sex between three and six times. The speaker of (13) is humorously offering himself as an expert witness, attesting that in his experience it takes many more acts of intercourse to impregnate a woman than that – so many that one becomes exhausted. For the full context, see Merlan & Rumsey (1986: 100).

3 Egophoricity, engagement, and the centring of subjectivity

(14) *ab ilyi ly-ab-i*

> woman this get-OPT:1SG-Q

'Do you want me to marry this woman?', 'Are you expecting me to marry this woman?'

Now let us consider a partly comparable case from Mwotlap, an Oceanic Austronesian language of Northern Vanuatu. Mwotlap does not mark person, number or tense on the verb. It has what François (2003, 2004) calls a Prospective marker *so* that occurs with a wide range of modal and other meanings. In an analysis that is consistent with other theorists of modality such as Halliday (1970), Verstraete (2005) and Lehmann (2012) (but different in its terminology), François (2003, 2004) distinguishes between the subject of the *modus* (i.e., of the modal projection encoded by the Prospective marker) and the subject of the *dictum* (of the sentence itself). While the latter is explicitly encoded within the sentence, the subject of the *modus* is left unspecified, and must be retrieved from the context. There are certain default patterns to this inference, one of which relates to the difference between declarative sentences and questions. Let us first consider a declarative example, as provided to me by François (personal communication March 12, 2016) based on one of the examples in François (2003: 221).

(15) *Nok* 1SG *so* PROSP *leg* marry *mi* with *kē.* 3SG 'I <Prosp> marry her.'

Here are possible readings given to me by François for (15) as a statement:


### Alan Rumsey

e) **I'm going to** marry her

non-modal readings e.g. typically in a dependent clause: 'She will move to my village, that's because I'll be marrying her…'

f) **If/When** I marry her… suspended modal origo (yielding a conditional protasis)

Here are possible readings given by François for that same sentence with question intonation:

(16) *Nok* 1SG *so* PROSP *leg* marry *mi* with *kē?* 3SG

'I <Prosp> marry her?'


modal subject = an authority, whether specific (father+) or non-specific (custom…) François adds:

"For example, if I confide to my uncle who knows well how custom or kinship systems or marriage rules work, and assuming I want to do well to please my extended family and marry the 'righ' person, then the modal origo would not be my addressee, but rather, a diluted, perhaps impersonal subject, whatever we understand as 'custom' (as it is often called in Melanesia) or 'law' (as among indigenous Australians)…" [cf. François 2003: 229]

d. **Am I going to** marry her?

non-modal readings (assuming there's a context where one could ask this question with no particular feelings or modal contents, as in: 'What's the plan exactly? Will I marry her?")

These uses of Prospective marking in Mwotlap compare interestingly with those of the optative mood in Ku Waru as discussed above. The main difference is that the uses of the latter are much more wide-ranging than the former. To see this, compare Ku Waru example (9) with Mwotlap example (15). As can be seen, (9) has the same sense as the one given for Mwotlap in (15a). But it cannot

### 3 Egophoricity, engagement, and the centring of subjectivity

express non-modal meanings (as in 15e) or any of the other modal ones in (15b)– (15d), all of which are expressed in other ways in Ku Waru. But notwithstanding this difference, when Mwotlap *so* does express a modal meaning, it shows the same crossover between speaker- and addressee-centred modality in questions as opposed to statements. This is shown in Table 3.2, which is reproduced from François (2003) with the title shown in English. (Note that "Cén" in the top row stands for "centre énonciative", or "modal subject" as François renders it in the examples above.)


Table 3.2: The crossover between volitive and deontic values of the Prospective in Mwotlap (after François 2003: 228)

The correspondence between the crossover in Table 3.2 and that in Table 3.1 is striking, especially in view of the fact that San Roque & Schieffelin's table is presented as pertaining to egophoricity and evidentiality with no reference to modality, and François' table as pertaining to modality with – so he tells me – no thought of its possible relevance for any other grammatical domain at the time when he produced it (personal communication March 25, 2016).

While this kind of crossover has become almost definitional of egophoricity or "conjunct/disjunct" marking, and also has been recognized with regard to evidentiality as discussed above, it has seldom been noted with respect to modality, which has generally been treated as if it were exclusively speaker-centred. Besides François (2003, 2004), the main exception that I am aware of is Lehmann (2012), who develops a systematic typological comparison among six languages

### Alan Rumsey

in this respect, and on that basis argues for an overarching category of "subjective" modality which includes, for all six of the languages, modal categories within which such interrogative-declarative crossover is found.

I have developed the Ku Waru - Mwotlap comparison in particular here for two reasons. The first is that, although the examples I presented from those languages are in full accord with the "egophoric distribution" shown in Table 3.1, they fall outside the scope of most of the existing literature on egophoric systems, almost all of which is exclusively concerned with the role in grammar of *epistemic* authority or involvement. In other words, the "ego" in "egophoric" is taken to be a *knowing* ego and the relevant asymmetries among speech act participants are taken to be ones of knowledge. By contrast, the relevant semantic considerations here are ones of deontic or intentional modality – what the entailed ego *wants*, or what others want of him/her.

The other reason for my comparison between Ku Waru and Mwotlap is that I think it shows with particular clarity how patterns that are fully grammaticalized in one language may be evident in the discourse patterning of another. In the following section, I build on that by moving beyond the semantics and pragmatics of egophoricity, evidentiality and modality to other aspects of language use that do not show the same pattern of crossover between speaker and addressee, but which are similar to the above examples insofar as they involve more-or-less regular shifts in the usual centring of subjectivity as between speaker and addressee. Arising as they do from my current work on child language socialization, a common factor among these remaining examples is that they all involve speech that is addressed to young children.

### **3 Altercentric kin term usage**

There are at least two ways in which Ku Waru adults' use of kin terms to children differs from their usage to other adults, and from children's use of them to anybody. The first of these is what Merlan (1982) in an Australian context dubbed "altercentric usage". To understand what that involves, it is helpful to introduce the term "anchor" as used by Dahl & Koptjevskaja-Tamm (2001). For any given kin term within a given context, the anchor is the person or persons from whom the relevant relation is reckoned. So for example the anchors of "Fred's uncle", "our aunt", and "Daddy", when used by a child to her father, are "Fred", 1PL and 1*sg* respectively. Usually when kin terms are used without explicit reference to an anchor, the implicit anchor is the speaker as in the "daddy" example above. But alternatively it may be the addressee, as for example when a mother says

### 3 Egophoricity, engagement, and the centring of subjectivity

to a child "Give it to Mommy" or "Where's Daddy?". What Merlan (1982) drew attention to was: 1) that this phenomenon, which she called "altercentric" usage, is very common around the world; 2) that it is especially common in speech addressed to young children by adults and older children; and 3) it is typically non-reciprocal, i.e., it is used when children are being addressed by adults but not vice versa.<sup>7</sup>

Altercentric usage of kin terms is common in Ku Waru, and is in accord with Merlan's generalizations. Examples are:


Note that in all these examples (as in the English examples of altercentric usage above) there is no explicit indication of the anchor. Just as in the English examples, it would have been possible to include one, with a possessive pronoun 'your' (*nunga*). But in both cases, as in many other attested ones from around the world, that indication of the anchor tends to be left out in altercentric kin-term usage by adults to children.

Another thing to note is that in both the English case and the Ku Waru one (again, as is common around the world), the kin terms that are used in this way are ones that in other contexts are characteristically used by children: "daddy" and "mommy" rather than "(my) father", "(my) mother", etc. Likewise in Ku Waru the terms *ma*, *tata* and *ana* are ones that are in other contexts used more by children than adults.

I suggest that what is going on here bears a family resemblance to what we saw above regarding the shift of typical modal origo as between statements and

<sup>7</sup> For further elaboration of this typology including other kinds of anchor-shift and examples of them see Agha (2006).

### Alan Rumsey

questions in egophoric, evidential and modal contexts, in two ways. First, in both cases the shift is between Speaker and Addressee as the relevant implicit ground in relation to which the meaning is figured. Second, in both cases this happens within a specific context: questions *vs* statements in the case of the modal shift and speech by adults to young children *vs* other adults and older children in the case of kin term usage. Another similarity is that the shifts in question are very widely attested in the languages and speech communities of the world. Below I will address the question of why this should be the case. First I will introduce another kind of shift that is widely found in the use of kin terms to children.

### **4 "Address inversion"**

In examining transcripts of interactions involving Ku Waru children one thing I have been struck by is a regular pattern whereby an adult when speaking to a child addresses him/her with the kin term that the child uses when addressing that adult. So for example a man addresses his ("actual" or classificatory) son or daughter by a term that ordinarily means 'daddy', and addresses his nephew or niece by a term that ordinarily means 'uncle'. Examples are:


<sup>8</sup>The 'daughter' who is being addressed here is in European terms not the speaker's own daughter, but the daughter of one of his 'classificatory' brothers, which in Ku Waru (as in many languages) places her in the same kin category as his own daughter (Kroeber 1909). Similar considerations apply to the kin relations in examples (21)–(23).

3 Egophoricity, engagement, and the centring of subjectivity

(23) a. John, to his daughter (age 3): *ep* now *na-nga* 1SG-GEN *oma* fish *mari* some *tuju-ni?* hit:BEN-FUT:2SG 'Now will you catch some fish for me?'

> b. Saina (the girl's mother): *toju-ba-yl* hit:BEN-FUT:3SG-DEF

'Indeed she will.' c. John: *e?* really? *a* oh *na-nga* 1SG-GEN *papa-n…* daddy-ERG

'Really? Oh, my 'daddy' [i.e., daughter] [will catch some]

In all these examples, in place of the junior term within a given dyad (fatherdaughter, uncle-niece, father-son) the senior term is used in address and reference to the junior member of the pair, regardless of his or her gender. Although such usage may seem unusual from an Anglophone perspective, it is actually quite widely attested from elsewhere in the world, including Albanian, Arabic, Bengali, Bulgarian, Greek, Hungarian, Italian, Norwegian, Persian, Romanian, Russian, Tok Pisin, Turkish and many other languages where it has been documented under the rubrics of *allocuzione inversa* (Renzi 1968), *umgekehrte Anrede* (Beyer & Kostov 1978), and "address inversion" (Braun 1988). This is actually a misnomer, for two reasons. The one that is especially relevant here is that, as shown by Braun, in the great majority of attested practices of this kind, there is actually not a full inversion. That is, while the senior party uses the term for the junior one that the latter uses for him or her, the junior party does not change his or her usage in the opposite direction. For example the man's classificatory daughter in example (20) does not reciprocate by calling him by a term for "daughter". Nor does the man's son in (22) call his father "son". Rather the children continue to call him by a father term.

What are we to make of these usages? As different as they are from the altercentric ones discussed above, there are, I suggest, three important similarities:


### Alan Rumsey

That 3 is true in the case of altercentric kin term usage should be obvious. It is also true of "address inversion" insofar as the terms that are used, even as their reference is reversed, are ones that are normally used by children to adults, just as in the altercentric case. But why the apparent reversal of reference? Friederike Braun (1988) has suggested that "the senior [kin term] used for inversion could be regarded as an archilexeme referring to the dyad as a whole, though explicitly naming only one of its partners"<sup>9</sup> (Braun 1988: 285). I'm not sure how well that proposal could be supported on formal grounds in the Ku Waru case, but in functional terms there does seem to me to be something right about the idea that the "inverse" uses of kin terms highlight the relationship *per se*. And in Ku Waru and elsewhere it seems to me no accident that the terms which do that, and which shift the anchor from speaker to addressee, should be particularly associated with contexts in which adults are speaking to children. For the highlighting of relationships is one of the main ways in which children are socialized and culture is transmitted.

### **5 Prompting routines**

I now turn to a much more general phenomenon involving the same kind of transposition of perspectives as in the use of altercentric kin terms. This is the phenomenon of prompting, i.e., of adults presenting utterances to the children for them to repeat to others.<sup>10</sup> An example is shown in (24). This bit of conversation took place at Kailge, a settlement which lies at the end of a long and winding unpaved road about one hour's ride from Mount Hagen, the capital of Western

<sup>9</sup> In that respect 'address inversion' is like 'dyadic' kin terms such as discussed by Merlan & Heath (1982) and Evans (2006). Dyadic kin terms are ones which refer to a pair of people who stand in a given relationship to each other. They mean things like 'pair of brothers', 'mother and child', etc. There is a widespread pattern found among those terms, first pointed out by Merlan & Heath (1982), such that whenever they are based on the word for one member of an intergenerational pair such as father-son, uncle-nephew, etc., it is the term for the senior member that becomes the basis for the dyadic term. For example in Kayardild, an Australian Aboriginal language spoken on the South Wellesley Islands of Northern Queensland, such terms are formed with a dyadic suffix –*ngarrba*; the expression for "mother and child", *ngamathungarrba*, is formed by combining this suffix with the word for 'mother', *ngamathu*. Likewise in Icelandic the dyadic terms *feðgin* "father and daughter" and *mæðgin* 'mother and son' are derived from the words for 'father' and 'mother' respectively (Evans 2006).

<sup>10</sup>For extensive discussion and exemplification of prompting in Kaluli, a Papuan language spoken in the Bosavi region about 120 km to the southwest of Ku Waru see Schieffelin (1990). For crossculturally comparative accounts of this very widespread phenomenon see Demuth (1986) and Moore (2012).

### 3 Egophoricity, engagement, and the centring of subjectivity

Highlands province. The speakers were a man named Taka and his eighteenmonth old daughter Laplin. Both of them were looking at an older boy named Mawa who was passing by.

(24) a. Father:

*mawa* (boy's name) *wi* call.out.to *to* do:IMP:SG 'Call out to Mawa.'

b. Daughter [shouts]: mawai! 'Hey Mawa!'

c. Father: *kar-na*

*p-abiyl* go-OPT:1DU *wa!* come:IMP:SG *kar!* truck

'Come, let's you and me go in the car! The car!'

d. Daughter: *wa* come:IMP:SG 'Come!'

truck-LOC

The father's utterance in line c is a typical instance of prompted speech in that it is voiced from his daughter's point of view, for her to repeat. Sometimes such prompts are explicitly framed with the imperative form of the word for 'say' (*nya*), but more often they are not, as in this case. The daughter nonetheless understands this as a prompt, and responds appropriately. She is at the one-word stage of language acquisition, so her father's utterance with its concatenated first person dual optative verb and locative-marked accompanying NP is far too complex for her to understand and repeat in its entirety, but she nonetheless understands that it is a prompt, picking out its most central element, the imperative verb *wa* 'come', and repeating it as if from her own viewpoint.

Another, more complex example of prompting is shown in 25. This is from a conversation between a Ku Waru mother (Wapi) and her two-year-old son (Jesi), in the presence of his five-year-old brother Alex.<sup>11</sup>

<sup>11</sup>For some of the son's utterances in this transcript there are two parallel lines of Ku Waru. The first line shows what he actually said and the line below it shows what our Ku Waruspeaking transcription assistant has offered as an adult equivalent of that utterance. For further discussion of this interaction see Rumsey (2014).

(25) a. Mother: *ana* same.sex.sibling *kola* cry *naa* not *ti=o* do:JUS=VOC *nya* say:IMP:SG 'Tell your brother not to cry [lit: 'Say "brother, don't cry"']' b. Son: *ana* same.sex.sibling *kola* cry *naa* not *ti=o* do:JUS=VOC 'Brother don't cry.' c. Mother: *kali* kalyke *p-abiyl=o* go-OPT:2DU=VOC 'Let's go to Kailge.' d. Son: *teka* kalyke Kailge *pabi=o* p-abiyl=o go-OPT:2DU=VOC 'Let's go to Kailge.' e. Mother: *sispop* cheesepop *ly-abiyl* get-OPT:2DU *p-abiyl=o* go-OPT:2DU=VOC 'Let's go get some cheesepops.' f. Son: *titopa-ti* sispop-ti cheesepop-IDF *nabi* n-abiyl eat-OPT:2DU 'Let's eat a cheesepop.' g. Mother: *p-abiyl* go-OPT:2DU 'Let's go.' h. Son: *pebil=o* p-abiyl=o go-OPT:2DU=VOC

'Let's go.'

3 Egophoricity, engagement, and the centring of subjectivity

i. Mother: *kola* cry *naa* not *ti=o* do:JUS=VOC 'Don't cry.' j. Son: *pike naa ti*

bighead not do:JUS 'Don't be a bighead (disobedient).'

All of the mother's utterances in this example are prompts offered to her son for him to repeat. Only the first one (25a) is explicitly framed as such by the imperative verb *nya* 'say'. But again the son understands them all as prompts and responds appropriately. In lines (25b), (25d), (25f) and (25h) he does this by repeating the utterances he has been prompted to say (sometimes in his own toddler-talk variants). In line (25j) he responds more creatively by saying something that does not repeat the prompt verbatim but is fully within the spirit of it, chastising his brother for being a "bighead". The framing relationship between the mother's prompt in line (25a) and the prompted response that she projects for her son in it is shown in Figure 3.1. The sons's response in line (25b) and all of his subsequent ones are interactionally complex in that they are positioned both as responses to his mother and as injunctions directed to his brother Alex. The same is true of the daughter's responses to her father in 24. In interactional terms, the utterances by the parents that prompt these responses are similar to what happens with altercentric kin term usage as exemplified above, in that the adult speaker is in effect aligning his or her perspective with that of the child, while at the same time retaining his/her position as the controlling party from whose perspective the child's utterance is projected.<sup>12</sup> Accordingly, in common with altercentric kin term usage, the projection of perspectives within prompting routines is asymmetrical in that it is always done by the adult in relation to the child, never the other way around.

<sup>12</sup>A reviewer comments regarding this form of prompting that it 'seems to be very different from the perspective shift in altercentric kinship terms. Here the "perspective shift" is a natural consequence of the fact that the prompt is given verbatim, as direct (pre-)reported discourse, which is the simplest way to instruct a child what to say.' I agree that prompting is different in kind from altercentric kinterm usage in that it is in some instances explicitly marked as what the referee calls 'direct (pre-)reported discourse'. But in those many instances where it is *not* explicitly framed in that way by a verb of saying, it is more like altercentric kinterm usage in that the shift of perspective (or, in Bühler's 1990 terms, transposed origo) must be indirectly inferred by the addressee, in this case the child. I suggest that the children's ability to do that so readily from a very early age results in part from the regularity of the direction in which that shift takes place, as identified in the following sentence.

Alan Rumsey

Figure 3.1: Interactional framing in conversation between Wapi and Jesi

### **6 Discussion**

In the first part of this chapter I have discussed the regular shifts in the centring of subjectivity as between speaker and addressee that are entailed in the use of the grammatical categories of egophoricity, evidentiality and intentional modality. I then discussed other such shifts that are typically found in the use of kin terms to young children, and in the prompting of children by adults and older children. In all of these kinds of shift there was a high degree of regularity but there were considerable differences between the former, category-based alternations and the latter, situationally based ones, namely:


### 3 Egophoricity, engagement, and the centring of subjectivity

Although these two kinds of shifts may seem very different, in some contexts there is interaction between the two. This happens on at least two different levels. First, in keeping with what San Roque & Schieffelin (2018: 451) have suggested regarding speech to small children in the Kaluli language (about 200 km west-southwest of Ku Waru), the practice of "speaking for" the child in prompted utterances involves at least a partial displacement of epistemic and intentional authority from the child to the adult, or a virtual merging of the two. For example in line (24c), the parent uses an optative form, the general modal meaning of which involves the attribution of an intention by the speaker to him/herself – a self-attribution. Here the intention is presented to the addressee as if it were (also) her own, to be voiced by her in address to a third party, which she then does. The same is true of line c in example (25).

Additionally there is a shift in what Nuyts (2005, 2006) would call the "controlling participant", or "first-argument participant" (and Lehmann 2012 the "executor"), i.e., the person who would carry out the intention. That is, it is not the speaker of the prompting utterance nor her addressee Jesi, but Jesi's addressee in the projected utterance that he is prompted with.

In other cases, there is an interesting transposition of perspective within adult speech to children that is the inverse of the one discussed above regarding modality. An example is (26), the likes of which I have often heard in adult's speech to small children.

(26) *lku* house *suku* inside *w-ab-i* come-OPT:1SG-Q 'Do you want to come into the house?'

Note that this example is the same as the last three words of example (10) above, the gloss of which was 'Can I come into the house'. But here the understood controlling participant (the person whose coming into the house is at issue) is the addressee rather than the speaker. So here there is a double displacement, whereby it is not only the modal subject that shifts from speaker to addressee but also the controlling participant. This is in accord with the fact that such examples meet both of the contextual criteria for perspective shift that I have discussed above, namely, occurrence of modal verbs in questions as opposed to declaratives and occurrence in the speech of adults to children.

The same thing sometimes also happens with the use of the *future*, which in Ku Waru, as in most or all other languages, is not a purely temporal category, but also partly modal in value. An example is (27).

### Alan Rumsey

(27) a. Adult:

*nu* 2SG *ku* stone/money *mare* some *pe-nsi* be/lie-CAUS *molt-i* be/stay:HAB:2SG-Q 'Do you have any money?'

b. Child:

*al* that:ENDO:DEF *na-n* 1SG-ERG *sib-ayl* give:FUT:1SG-DEF 'I'll give it [to you].'

c. Adult: *naa* not *lyi-nsi-bu-e* take-BEN-FUT:1SG-Q 'Oh, so you're not going to take it [for yourself].'

In line (27c) as in (26) there is a shift of perspective by the adult to that of the child, both with respect to the modal subject of *lyinsibu* (the person to whom the intention as attributed) and the controlling participant. But in (27) there is a further complexity in that the utterance in line (27c) as a whole is not voiced entirely from the projected viewpoint of the child. Rather it is a hybrid or "double perspective" construction (Evans 2006) in that the question marker -*e* is voiced from the father's perspective, as the person who is questioning the child's intention.

Below I will introduce another grammatical category, the nascent one of "engagement", and discuss its relation to all of the phenomena discussed above. Before doing so I will first turn to a question pertaining to the three categories that I have discussed so far. The question arises from the fact that in previous treatments of egophoricity (a.k.a. conjunct-disjunct patterning) it has been so closely associated with the crossover pattern shown in Table 3.1 as to make that seem almost definitional of it. But as has long been appreciated, that pattern is also found with respect to evidentiality, and as shown above, it is also associated with some kinds of modality. So the question is, among all the phenomena discussed above, what if anything is distinctive about egophoricity? If anything, I think it is this: While all the phenomena discussed above involve shifts in the locus of relevant subjectivity, egophoric marking is the only one that explicitly tracks that by making use of morphemes whose main function is to indicate whether or not a given argument or other nominal referent is the primary locus of relevant subjectivity. By contrast in example sentences (8) – (14) from Ku Waru and most of the readings of (15) and (16) from Mwotlap the main function of the morphemes in question was to express intentional modality, and the main function of Ku Waru kin terms in the following examples was to refer to particular people, albeit in

### 3 Egophoricity, engagement, and the centring of subjectivity

ways that involve particular shifts of perspective that correspond to the ones involved in egophoric marking. But I think we must add that the contrast with egophoric marking is only a partial one, because not all of it does actually indicate the locus of subjectivity. This is shown to be the case in Akhvakh by example (4), where it can be seen that the presence or absence of egophoric marking does not take account of the difference in the centring of subjectivity as between actual questions and rhetorical ones. Rather, it is determined in a semantically bleached, fully "syntacticized" way (Creissels 2008: 11) by its occurrence in a question with a first person subject. Furthermore, as is also clearly brought out by Creissels, the incidence of egophoric marking is not directly conditioned by the presence or absence of imputed "assertor involvement" at the level of actual situated utterances, but rather is grammatically and lexically determined. It occurs only in constructions with perfective verbs. When they are transitive, it is obligatory on A and when they are intransitive it is obligatory on S for a subclass of intransitive verbs whose subjects are typically in control of the action, and proscribed in construction with verbs whose subjects are typically not in control. The operative word here is "typical": egophoric marking in Northern Akhvakh is not sensitive to semantically differing uses of a single verb in different contexts, but instead operates as what Creissels calls:

a particular type of agreement, for which the term *assertive agreement* can be proposed. The difference with person agreement is that, in person agreement, verb morphology reflects the coincidence between particular argument roles and the speech act roles*speaker* vs. *addressee* vs. *non-SAP*, whereas in assertive agreement, verb morphology (or the morphology of a subclass of verbs) reflects the coincidence between a particular argument role and the speech act role of *assertor* (Creissels 2008: 11).

While the egophoric system of Northern Akhvakh is perhaps unusual in the extent to which categorial factors override discourse-contextual ones in this way, it seems from San Roque & Schieffelin's (2018) wide-ranging survey of such systems that all of them show this tendency to a certain extent. In other words, the marking of what San Roque & Schieffelin call "epistemic authority" (Creissels' "assertor involvement") is probably never realized in a completely exceptionless way by any single formal device. Conversely, many or all of the other formal categories discussed above probably also serve that function to some extent in certain discourse contexts. This must be taken as a qualification to the answer I have offered above to the question of what is distinctive about egophoric marking. But in my view it still leaves the category of egophoricity intact as a valid

### Alan Rumsey

cross-linguistic one, which is recognizable within particular languages to the extent that they have markers that signal epistemic authority or assertor involvement as their *primary* function if not the only one. Having argued as much I now turn to a consideration of the fledgling linguistic category of "engagement" and its relation to the other linguistic phenomena discussed above.

### **7 Engagement**

Evans et al. (2018a: 112) characterize engagement as "grammatical means ... for intersubjective coordination". Their focus on those means has no doubt been stimulated in part by the exciting work that has been done in recent years by psychologists, primatologists and cognitive scientists on what is known as "joint attention" (Elian et al. 2005), "shared intentionality" (Tomasello 1999, Tomasello et al. 2005) or "triangulation" (Hobson 2004), our capacity to focus jointly with others on objects of attention and share and exchange intentions and perspectives with respect to them. It is one of our most distinctively human capacities and propensities, not shared to any great extent even by our close primate cousins, but emerging in typically developing humans before speech – from the age of around nine months. Although in its basic form this capacity does not depend on language, the evolution of language has almost certainly depended on *it*, and in turn has greatly enriched it. Building on the work of Landaburu (2005, 2007) the recent proposals by Evans et al. (2018b) for a cross-linguistic (meta-)category of engagement are an attempt to elucidate the grammatical means by which that capacity is exercised in language use.

Evans et al. note that, although our understanding of "the full panoply" of those means "remains basic" (2018a: 112), one of them that has been widely studied at least in western European languages is "the definiteness/indefiniteness contrasts expressed in article systems" (117). Ku Waru has a somewhat similar system, but with a wider distributional scope. Rather than articles, it has suffixes, two of which my colleague Francesca Merlan and I (Merlan & Rumsey 1991: 336–337) call "definite" and "indefinite". The "definite" marker is -*ayl*/-*iyl*/-*yl*. It means roughly "… which you and I know about". It contrasts most directly with the indefinite marker -*ti*, which means roughly "… which I know about but you may not". In some contexts it contrasts with another marker -*ja*, which means roughly "perhaps" or "you and I are not sure about this". All three of these suffixes can occur not only on nouns, but also on inflected verbs, in which case their scope typically includes the entire clause.

Uses of the latter two markers are illustrated in (28), which is an excerpt from an audio and video recording of an interaction between a 3 ½ year old Ku Waru

### 3 Egophoricity, engagement, and the centring of subjectivity

boy Ken, his father Lep and his uncle, John Onga. A video of this stretch of interaction is available online at https://vimeo.com/257625252 Password: Kailge. As can be seen there, Ken is sitting in his father's lap, facing away from his father toward the video camera, which is being operated by John. In the lead-up to this stretch of interaction, John suggests that he is going to give Ken some money. Instead of responding directly to John, Ken puts his hand into Lep's pocket and starts feeling around for coins there. Instead of money he finds a bit of dried tobacco leaf, which he pulls out and hands to Lep. He then puts his hand back into Lep's pocket and starts searching for money again. That is the point at which the video and transcript start, as shown in (28).<sup>13</sup>

(28) a. Child (Ken)

*ti* one *lyi-bu* take-FUT:1SG 'I'll take one [coin].'

b. Child (Ken)

*ti* one *pe-lym-ja* be/lie-HAB:3SG-MAYBE 'Maybe one is there.'


*ti* one *pe-lym-ja* be/lie-HAB:3SG-MAYBE *kan-ui* see/look-JUS 'Maybe one is there. Look.'

e. Child (Ken) *ti* one *pe-lym-iyl* be/lie-HAB:3SG-DEF 'One is indeed there.'

<sup>13</sup>The root *pe*- (in example 28b) is one of five existential verbs in Ku Waru which differentially characterize their subjects, either with respect to their inherent properties, or with respect to their state-of-being within particular contexts of discourse (Rumsey 2002). The specific value of *pe*- when used in the latter way, as in this case, is to indicate the referent of its S (the coin that may be in Lep's pocket) is latent or concealed (loc. cit. pp. 188–197).

### Alan Rumsey

f. Adult (John) *kan-kun-i* see/look-PPR:2SG-Q 'You see?'

This short stretch of interaction is a classic case of joint attention, in which the child Ken and his uncle John are focusing on Lep's pocket and Ken's attempt to find money there. The intersubjective coordination within it takes place along several different dimensions at once. These include at least the following:


### 3 Egophoricity, engagement, and the centring of subjectivity

am referring to the meaningful interplay of repetition and variation as theorized by Roman Jakobson (1960), Michael Silverstein (2004), James Fox (2014) and others. A prime example of this in (28) is the relation between -*ja* in line lines (28b)–(28d) and -*iyl* in line e as discussed above.<sup>14</sup>

### **8 Conclusions**

In this chapter I focused initially on egophoricity and how it has been identified as a system for indicating "epistemic authority" or "assertor involvement" in relation to predicated actions or states of affairs, with related reversals of marking as between speaker and addressee in statements *vs* questions. I have shown that those reversals, which have been called "egophoric patterning", are not specific to egophoricity *per se*, but (as has long been recognised) are also found with respect to evidentiality and (as less often recognised) also with respect to some kinds of modality. In all three cases there are regular shifts in what I call the "centring of subjectivity" as between speaker and addressee. Drawing from my ongoing study of Ku Waru children's language socialization, I then turned to a consideration of shifts which are similar in that respect, but which differ from the former in being associated with particular contexts of interaction (i.e. speech to young children by adults and older children) rather than with particular grammatical categories. I showed that, notwithstanding this difference, there are, in certain contexts, interactions between the two kinds of shift, resulting in a kind of "double displacement" in the centring of subjectivity. Finally, turning to a Ku Waru example of the nascent category of "engagement" as "grammatical means ... for intersubjective coordination" (Evans et al. 2018a: 112), I showed how the grammatical dimension of such coordination is thoroughly intertwined both with other dimensions of language and discourse patterning including intonation, prosody and parallelism, and with non-linguistic dimensions including gaze direction and facial expression.

An issue that I left out of that discussion of engagement but will treat here in order to link it to the rest of the chapter is: what is the status of subjectivity with respect to examples such as (28)? Here, in relation to "engagement", the notion of the "centring of subjectivity" seems to me less straightforwardly applicable than it was to the other phenomena discussed in the chapter — at least if we take that expression to refer to a shifting centre of epistemic authority, involvement

<sup>14</sup>For a more detailed analysis of parallelism in (28) that draws on Du Bois concept of 'dialogic syntax' (2014, Du Bois & Giora), and fuller treatment of 'engagement' in Ku Waru in general, see Rumsey 2019.

### Alan Rumsey

or intention that is at any given moment always or mainly located in the individual mind of one of the interacting parties or another. Rather, the intersubjectivity that is in play in the processes of engagement as "intersubjective coordination" is not something that is entirely shaped by the individual, preexisting subjectivities of the interacting parties, but rather is a process through which those subjectivities themselves are partly shaped. With respect to (28) for example, consider the fact that Ken's utterance in line a is not entirely innovative within this interaction, but partly repeats something his father Lep has said a few turns (14 seconds) before in the conversation while reaching into his own pocket, namely:

(29) *mare* some *pe-lym-ja* be/lie-HAB:3SG-MAYBE *kan-abil* see-OPT:1DU Maybe some is there; let's (you and I) see.

Lep in this utterance was in turn responding in part to John's remark that I referred to earlier, that he was going to give Ken some money. John's remark had provided a context in which it could be taken as a matter of shared understanding that the referent of Ken's *mare* "some" is money, and as a matter of *presumed* understanding that if Lep finds money in his pocket he himself will give at least some of it to Ken (the modal and grammatical subject of his *kanabil* 'Let's see' being first person dual, referring to Ken and himself). Lep's utterance in (29) then provides a context for Ken's lines (28a) and (28b), in two ways:


In line (28d) John picks up on Ken's use of *ti* 'one' in place of Lep's *mare* 'some', and Ken's repetition of Lep's *pelym-ja* 'maybe is there'. What this amounts to in terms of "epistemic authority" is that it has in effect been distributed among three people: Lep, Ken and John. The associated *intention* – for Ken to get money – has also been distributed among the three of them, in that it has started out with John, been taken up in a modified form by Lep (with him as the donor

### 3 Egophoricity, engagement, and the centring of subjectivity

rather than John), and then in a further modified form by Ken (with him as the active taker rather than the passive recipient). During lines (28b)-(28d) there is a thoroughgoing intersubjective coordination among the three of them in that all of them are focusing with that shared intention and uncertainty on Ken's probing hand in Lep's pocket, which Ken emphasises with his prosody and intonation in line (28c), and John with his use of a jussive form of the verb *kan*- in line (28d). Then there is coordinated marking of their transition from uncertainty that is initiated by Ken's change from -*ja* in line (28e) to the definite marker -*iyl* in line (28e), a suffix that explicitly flags the presence of the coin in Lep's pocket (and its being concealed from sight there) as a matter of shared knowledge among them. This is then further reinforced by John's switch from jussive marking on *kan*- in line (28d) to Present Progressive marking in line (28f) (Present Progressive being a verbal category which signals indicative mood in addition to its temporalaspectual meaning).

In short, in keeping with its definition as an aspect of intersubjective coordination, engagement shares with egophoricity, evidentiality and some modal categories its grounding in basic aspects of human interaction, but differs from all of them in the extent to which it treats the centring of subjectivity as potentially variable, emergent, and distributed across the interaction rather than as prototypically related to the speech roles of speaker and addressee and their alternation across speech-act types. It is important to note that I have expressed this difference as a matter of degree ("the extent to which…") rather than as a completely categorical one. For the complexities of language in use always exceed our ability to pin them down analytically, and as interestingly exemplified by other chapters in this volume, the formal devices that we identify as egophoric and evidential in particular languages may also have "engagement" functions and vice versa.

### **Abbreviations**


### Alan Rumsey

### **Acknowledgements**

I wish to thank Henrik Bergqvist for inviting me to participate in the Stockholm Symposium on evidentiality, egophoricity, and engagement for which the original version of this article was written, and the Swedish Research Council for funding my travel costs to it. Many thanks to all the workshop participants who commented on my presentation there, to Lila San Roque, Francesca Merlan, Naomi Peck and the anonymous Open Linguistics referees for helpful feedback on intermediate versions of the article, and to John and Wapi Onga for their invaluable advice regarding aspects of Ku Waru grammar while on a visit to Canberra at the time I was revising it. Thanks also to Australian Research Council for funding the field research on which it is based.

### **References**


3 Egophoricity, engagement, and the centring of subjectivity


### Alan Rumsey

*guages of Latin American II, University of Texas at Austin, 27th-29th October 2005*. http://www.ailla.utexas.org/site/cilla2\_toc.html.


3 Egophoricity, engagement, and the centring of subjectivity


# **Chapter 4**

# **An egophoric analysis of Dhivehi verbal morphology**

### Jonathon Lum

University of Melbourne

Egophoricity is a typologically rare category in which first-person statements and second-person questions share the same marking ("egophoric"), while other contexts generally lack this marking. This chapter presents evidence that certain verbal suffixes in Dhivehi (Indo-Aryan, Maldives) mostly show an egophoric distribution, contrary to previous analyses of these suffixes as first/second person markers. Dhivehi is thus the first Indo-European language reported to show an egophoric pattern. Aside from the distribution of the relevant suffixes in first-person statements and second-person questions, further evidence for an egophoric analysis may be found in contexts where a third-person nominal subject refers to the speaker or addressee ‒ in these contexts the appearance of the relevant suffixes is more consistent with egophoricity than person marking. However, egophoricity in Dhivehi is restricted to finite, volitional stems in certain tenses/aspects/moods, and some relics of an older person-marking system survive. Finally, it is proposed that egophoricity in Dhivehi may have developed from a reanalysis of person markers in reported speech, in a process similar to that described by Widmer & Zemp (2017) for certain Tibeto-Burman languages.

### **1 Introduction**

### **1.1 Background and outline**

In Dhivehi, the Indo-Aryan language of the Maldives, finite volitional-stem verbs in some tenses/aspects/moods are seemingly marked for person, though different linguists have listed different forms for certain parts of the verbal paradigm. This is true especially for the second person, which has been grouped both with

Jonathon Lum. 2020. An egophoric analysis of Dhivehi verbal morphology. In Henrik Bergqvist & Seppo Kittilä (eds.), *Evidentiality, egophoricity, and engagement*, 95–139. Berlin: Language Science Press. DOI: 10.5281/zenodo.3975801

### Jonathon Lum

the first person (Cain & Gair 2000: 23–27) and partly with the third person (Fritz 2002: 166–169). This chapter offers a new analysis of the situation, presenting data from recent fieldwork as well as from written material to argue that Dhivehi in fact displays an egophoric or "conjunct-disjunct" pattern of verbal marking. Egophoricity is a typologically rare pattern in which verbs in second-person questions are marked like verbs in first-person statements, while verbs in secondperson statements are marked like verbs in third-person questions/statements (e.g., Hale 1980; Creissels 2008; San Roque et al. 2018). The presence of this pattern in Dhivehi may partly explain the inconsistencies between some existing descriptions of the language, given that verbs with second-person subjects may resemble verbs with either first- or third-person subjects, depending on the sentence type.

More importantly, the existence of egophoricity in Dhivehi is of considerable typological significance, as grammatical systems of this kind are highly unusual cross-linguistically, and have never before been reported for an Indo-European language. The origins of egophoricity in Dhivehi are therefore of some interest, especially given the isolation of the Maldives from regions (such as the Tibeto-Burman area) where egophoricity has previously been reported. The details of egophoric systems vary greatly (Floyd et al. 2018), and the Dhivehi system shows some features that have not been (widely) reported on before, and which may shed light on egophoricity as a cross-linguistic phenomenon. These include the use of egophoric marking in first-person questions, and an interaction with a tendency (motivated by politeness) towards third-person nominal reference in place of first- and second-person pronouns in many contexts.

Like many other languages with egophoricity, Dhivehi also has a reported speech construction in which egophoric markers may be seen as showing coreference between the subject of the matrix clause and the subject of the embedded clause, while egophoric marking is absent when these subjects are not co-referential. And like in many other egophoric systems, Dhivehi egophoric markers are restricted to volitional contexts, though the existence of separate "involitive" morphology means that the alternation between egophoric and alterphoric marking of the same stem is never deployed to show contrasts in volition, unlike in some other languages (e.g., the Tibeto-Burman language Newar, Hale 1980).<sup>1</sup> It is therefore hoped that this chapter will be of interest both to scholars of Dhivehi and to scholars working on egophoricity and related areas such as volitionality, person marking, and reported speech.

<sup>1</sup> See § 1.3 for definitions of the terms *egophoric* and *alterphoric*.

### 4 An egophoric analysis of Dhivehi verbal morphology

This chapter is divided into six main sections. The current section introduces the prospect of an egophoric analysis for Dhivehi and the significance of such an analysis, provides a brief profile of the language, its speakers and dialects. The chapter also makes some brief notes on terminology and data collection. § 2 provides a summary of some previous accounts of person marking in Dhivehi, highlighting their inconsistencies. § 3 introduces some issues with those accounts and shows how an egophoric analysis is a better fit for the data, presenting evidence from second-person statements as well as from contexts where speakers use third-person nominal reference in place of first- or second-person pronouns. Data from reported speech are also shown to be consistent with this analysis. However, elements of an older person-marking system appear to be present in the context of first-person questions and in the distribution of the archaic/literary suffix ‑*mu*. §4 discusses person marking in some of Dhivehi's conservative southern dialects, and suggests a possible way in which egophoricity could have developed in the northern dialects (including the standard Malé dialect). In particular, I propose that (northern) Dhivehi may have undergone a similar process to that described by Widmer & Zemp (2017), in which a person-marking system is gradually reanalyzed and reshaped into an egophoric system, via a semi-direct speech construction. Finally, § 5 summarizes the chapter and suggests some areas for future research.

### **1.2 Dhivehi**

Dhivehi (or "Maldivian") is an Indo-Aryan language spoken throughout the Maldives, where it is the national language. Dhivehi has more than 340,500 speakers (Lewis et al. 2014), mostly in the Maldives but also in smaller numbers abroad. A dialect of the language (sometimes known as Mahl) is spoken on Minicoy, an island belonging to the Indian union territory of Lakshadweep, to the north of the Maldives. Despite the increasing encroachment of English, Dhivehi continues to enjoy a dominant status in the Maldives, where it is the main language of communication in mass media, government, and home life.

There are two main dialect groups (see Figure 4.1, below): a northern group spanning from Minicoy all the way to Laamu Atoll, and a southern group comprising the dialects of Huvadhu, Fuvahmulah, and Addu. At the heart of the northern dialect group is the standard dialect based on the language of Malé. This standard variety is used throughout the country in mass media and in official contexts; it is understood all across the archipelago. The atolls nearest to Malé use this dialect with only some slight variants, while more far-flung islands

### Jonathon Lum

or atolls (such as Minicoy and Laamu) have dialects that are related but more clearly distinct.<sup>2</sup>

The dialects in the southern group are the most conservative and show similarities with older varieties of Sinhala, the language to which Dhivehi is most closely related.<sup>3</sup> The Addu and Fuvahmulah dialects are described by Fritz (2002), though the Huvadhu dialect is practically undocumented (except for a small amount of information in Maumoon 2002 and Wijesundera et al. 1988). These dialects are not understood by Malé speakers (though as indicated earlier, southerners can understand the Malé dialect due to exposure), and are apparently not mutually intelligible either.

Even the standard/Malé dialect has attracted little scholarly attention, in part because the Maldives has been relatively inaccessible to outsiders until the late 20ℎ century, and in part because the language has sometimes been assumed to be very similar to Sinhala. Thus, although the first word lists date back to the 17ℎ century (Pyrard 1619) and the first grammatical sketches to the early 20ℎ century (Geiger 1919), comprehensive dictionaries and grammatical descriptions were not made until the last few decades. At present, the best Dhivehi-English dictionary is Reynolds (2003), and the most detailed grammars are Wijesundera et al. (1988), Cain (2000) (also re-worked into a sketch grammar, Cain & Gair 2000), Fritz (2002), and especially Gnanadesikan (2017). There are also a number of Dhivehi-medium works on the language, housed in various educational and research institutions in the Maldives. Most of these are prescriptive (e.g., Ahmad 1970; Saudiq 2012).

As is the case for Sinhala, there is evidence of substantial historical contact between Dhivehi and Dravidian languages (Cain 2000); and in later periods Dhivehi has also come into contact with Arabic, Portuguese and English. Typologically, Dhivehi has much in common with Sinhala (spoken in Sri Lanka) and other languages of the South Asia region. Dhivehi has a predominant SOV word order, and noun phrases are consistently head-final. The language makes considerable use of clause chaining, with sentences made up of only one finite clause

<sup>2</sup>The northern dialect group is probably not as homogeneous as has sometimes been claimed (e.g., Fritz 2002: 13). According to the consultants from Malé and Laamu with whom I worked during field trips in 2013‒2015, as well as some expatriates in Australia, Malé speakers can barely understand the Laamu dialect, if at all. They also report that some northern islands such as Naifaru (in Lhaviyani Atoll), have very distinct dialects too. More work is needed on this subject. Nonetheless, it does appear to be true that the northern dialect group is more homogeneous than the southern dialect group.

<sup>3</sup>According to Cain (2000), Dhivehi and Sinhala may have begun to diverge as early as the 3‒1 centuries BCE, but were in contact for many centuries after that.

Figure 4.1: Dialect Map of Maldives

### Jonathon Lum

preceded by any number of non-finite medial clauses. Pro-drop is typical, especially in spoken language. Clauses tend to omit as many arguments as are retrievable from contexts. For reasons that are not entirely clear (but probably due in part to language contact), Dhivehi has lost much of the morphological complexity found in Sanskrit and the Prakrits, from which it descends. Gender is not a category in the standard language (even in pronouns), and number is not obligatorily marked except on nouns denoting humans. There is no morphological distinction between the nominative and accusative (both falling under an unmarked "direct" case), though the standard language does also have separate dative, locative, genitive, instrumental/ablative, comitative and vocative cases. In terms of alignment, Dhivehi therefore shows "neutral" alignment in its nominal and pronominal morphology (i.e., core arguments are never distinguished from each other morphologically).

Verb paradigms vary according to stem types, with slightly different patterns for monosyllabic stems (e.g., *ka-nī* 'eating'), polysyllabic *a*-stems (e.g., *jaha-nī* 'striking'), polysyllabic *nn*-stems (e.g., *ganna-nī* 'buying'), *e*-stems including inactive/intransitive/involitive verbs (e.g., *e ge-nī* 'knowing'), and irregular verbs (Cain & Gair 2000: 22–25; Gnanadesikan 2017: 136–146). For all stem types, a number of tenses, aspects and moods may be distinguished morphologically. Verbs in the present, simple past, simple future, perfect, potential/optative, and irrealis/conditional are generally marked for person (e.g., Cain & Gair 2000: 23–27; Gnanadesikan 2017: Chapter 8), though in this chapter I will argue that this is in fact egophoric marking. In contrast, verbs with progressive aspect simply take the suffix ‑*(n)ī* regardless of person or number. The same suffix also function as a focus marker, attaching to verbs that appear in non-canonical (i.e., non-final) position regardless of person, number, or aspect (this "focus marking", along with the unusual word order and various prosodic cues, puts pragmatic focus on whatever constituent follows the verb). Progressive/focus verb forms are very common in the language and appear in many contexts where English or other languages would use a conjugated verb.

Like Sinhala, Dhivehi also has an alternation between active, volitional verbal morphology and inactive/intransitive/involitive morphology (which Cain & Gair 2000 usefully refer to as IN-morphology), with IN-verbs never carrying person marking. This morphological category interacts with various syntactic constructions and plays an important role in the grammar of the language. IN-verbs are used in many intransitive clauses, passive (or at least, "inactive") clauses, as well as in clauses where the agent acts accidentally or against his own will. In addition, IN-verbs may be used to show politeness, abilities, or counter-expectations

### 4 An egophoric analysis of Dhivehi verbal morphology

of the speaker, among other functions (see Cain 1995; Cain & Gair 2000: 56–60; Gnanadesikan 2017: 248–257).

Along with progressive/focus verb forms, Dhivehi makes much use of various other verb forms and verbal nouns that do not show person marking. In particular, converbs are used in clause chains, and participles in adverbial clauses, complement clauses, and relative clauses. In fact, as Gnanadesikan (2017: 289) notes, in general Dhivehi allows at most one finite verb per sentence. It should also be noted that there are no copular verbs in the language ‒ statements of equivalence are made by attaching a copular marker directly to the subject noun phrase (see Gnanadesikan 2017: 236–237). The prevalence of progressive/focus verb forms, IN-verb forms, converbs and participles (and other non-finite verbs), in conjunction with the lack of copular verbs, means that there are relatively few verbs in ordinary Dhivehi discourse that show any kind of person marking, a point to which I return in §1.4.

### **1.3 Terminological note**

In this chapter I use the terms *egophoric* and *alterphoric*, following Post (2013). The term *egophoric* (from Tournadre 1992; 1994) has become fairly well established in recent years (e.g., Floyd et al. 2018), displacing earlier terms such as *conjunct* (Hale 1980). Egophoric markers are used in first-person statements and second-person questions, though they sometimes appear elsewhere in some languages. Verbal marking that appears only in other contexts, especially thirdperson statements/questions and second-person statements, was originally labelled *disjunct* marking by Hale, though other terms such as *non-egophoric* have more recently become popular. Instead of *non-egophoric*, I use the term *alterphoric* as a convenient way of referring specifically to non-egophoric marking on finite, volitional-stem verbs in the relevant tenses/aspects/moods, i.e., the same stems which may carry egophoric marking. The term *non-egophoric* is less suited to this purpose because it tends to imply 'not egophoric', yet in Dhivehi, nonfinite verbs (including converbs, participles, and infinitives), IN-verbs, and verbs in the progressive/focus form are unable to carry either egophoric or alterphoric marking, but in a general sense are non-egophoric (i.e., not egophoric) too. This will become clearer throughout this chapter. For recent overviews of terminology used in the egophoricity literature, see San Roque et al. (2018: 6–9) and Widmer & Zemp (2017: 35).

### Jonathon Lum

### **1.4 Data sources**

The data presented in this chapter mostly come from elicitation sessions conducted with a 34-year-old native Dhivehi speaker in Fonadhoo, Laamu Atoll during fieldwork in 2014‒2015. This speaker is fluent in both the Laamu and Malé dialects, and provided judgements and example sentences for both dialects. These judgements and sentences were subsequently verified separately with more than twenty other native speakers in Laamu, Malé, and in Australia (where some native speakers reside for work or study). To supplement the elicited data, a number of online searches for written language examples were conducted with the assistance of another native speaker consultant. Literacy rates are extremely high in the Maldives, and recent years have seen a rapid growth in Dhivehi language material online, including news, social media, blogs, and collections of short stories. Much of this material is written in Thaana, a right-to-left script unique to Dhivehi. Using a search engine, searches were made using combinations of common subjects (e.g., pronouns and kin terms) with common verbs (e.g., *kuranī* 'doing') in their volitional form and in the appropriate tenses/aspects/moods for the grammatical contexts at issue. The examples obtained by this process all come from short stories published online, on various websites and by various authors. Where necessary, more context is provided with each example.

Finally, a few written language examples from her corpus of Dhivehi online news stories were kindly shared by Amalia Gnanadesikan (personal communication), and some other examples are sourced from existing descriptions of the language such as Cain & Gair (2000) and Gnanadesikan (2017).

The use of these data sources was necessitated by the nature of the research question, which relates to verbal marking in a range of grammatical contexts. Some of these contexts, such as second-person statements and first-person questions, are relatively rare and may not necessarily appear in a corpus of spoken language texts.<sup>4</sup> This problem is common in studies on egophoricity, and unsurprisingly much of the literature going back to Hale (1980) relies at least in part on elicited data. It should also be noted that for Dhivehi in particular, it is not easy to find in naturally occurring texts the relevant data points to decide between a person-marking analysis and an egophoric analysis. Aside from the fact that the right contexts (e.g., second-person statements) are somewhat rare anyway, egophoric/alterphoric marking (or person marking according to previous

<sup>4</sup>The corpus of spoken texts I collected during my fieldwork in the Maldives, for example, did not contain clear examples of finite, volitional-stem verbs in the grammatical contexts at issue. This corpus, which was compiled for a separate project on spatial language and cognition (see Lum 2018; Palmer, Gaby, et al. 2017; Palmer, Lum, et al. 2017), mostly included recordings of speakers engaged in description tasks and instructional texts, with some narratives.

### 4 An egophoric analysis of Dhivehi verbal morphology

analyses) in Dhivehi occurs only on finite, volitional-stem verbs that are not in the progressive/focus form. This means that the appearance of IN-verbs (which are required in inactive/intransitive/involitive sentences, as mentioned in §1.2) does not help to decide between analyses, nor does the appearance of non-finite or progressive/focus forms. As discussed in §1.2, these verb forms are extremely common in Dhivehi. The use of direct elicitation and the World Wide Web therefore facilitated easier access to verb forms and grammatical contexts that are relatively uncommon in ordinary discourse.

### **2 Previous accounts of person marking in Dhivehi**

Descriptions of person marking in (standard) Dhivehi are provided by Geiger (1919), Wijesundera et al. (1988), Cain (2000), Fritz (2002), and most recently Gnanadesikan (2017). A few other works (Cain & Gair 2000; Maumoon 2002) contain descriptions of person marking based on Cain (2000). In addition, there are some prescriptive works (in Dhivehi language) that offer guidelines on person marking (e.g., Ahmad 1970; Saudiq 2012). However, many of these various accounts disagree on aspects of the Dhivehi person-marking system. In particular, they mostly agree on third-person forms, but not on first- and second-person forms. Geiger's (1919) description contains some clues as to how the system worked in the early 20ℎ century (see § 4.3), but unfortunately appears to confuse different tenses and aspects throughout paradigms. Wijesundera et al. (1988) describe first-person and third-person forms, but not the second person. I will therefore concentrate on the more recent descriptions provided by Cain & Gair (2000) and Fritz (2002), with some mention also of the prescriptive literature. The most recent description of the language, Gnanadesikan (2017), analyses the relevant markers as person markers, but like Wijesundera et al., describes them as firstperson and third-person markers. Referencing an earlier version of the current chapter, Gnanadesikan (2017: 138) acknowledges that second-person forms are variable, and accepts that at least some speakers show a split, with second-person subjects in statements triggering the same verbal agreement as third-person subjects, but in questions the same verbal agreement as first-person subjects.

### **2.1 Fritz's (2002) account**

According to Fritz (2002: 166), "[t]he finite verb is characterized by a three person system distinguishing singular and plural". For the simple present tense of

### Jonathon Lum

polysyllabic a-stem verbs in the Malé dialect, she gives the paradigm outlined in Table 4.1 below.<sup>5</sup>


Table 4.1: Person marking on simple present tense polysyllabic a-stem verbs according to Fritz (2002: 168–169)

Fritz thus presents a simple first- versus non-first-person distinction in the singular, but a more complicated picture in the plural. She states that -*mu* is used as a first/second-person plural marker, and lists this in her own tables. However, she notes in her prose that -*n* is an alternative to -*mu* in the first-person plural, and -*V*: may be used in the second-person plural, both of which she interprets as analogical formations based on the equivalent singular forms. The third person is simply -*V*: in both singular and plural.

For the simple past (in her terms "finite preterite"), which has a different stem, Fritz (2002: 174–176) describes the same basic pattern for person agreement: a first- versus second/third-person dichotomy in the singular, but a first/secondversus third-person dichotomy in the plural, with the third person having the same form (-*Ø*) in both singular and plural. She does not specify whether the first- and second-person plural also have alternative forms identical to their singular equivalents as they do in the present tense. Additionally, she regards the perfect as a compound of the "absolutive" (converb) form of the main verb followed by the simple past of the now obsolete verb \**fianī* 'put' (Fritz 2002: 225– 226). Thus, the perfect follows the same person-marking template as the simple past. However, for the simple future (which also has its own stem), Fritz (2002: 176‒178) presents a first- versus second/third-person distinction that is not sensitive to number. She also gives an alternative, archaic form -*ū* for the first/secondperson plural, which she reports is mostly confined to literary usage. These various paradigms are summarized in Table 4.2 below.

<sup>5</sup> In this table and elsewhere, -*V*: indicates lengthening of the final vowel in the stem.

### 4 An egophoric analysis of Dhivehi verbal morphology


Table 4.2: Person suffixes according to Fritz (2002)

### **2.2 Cain & Gair's (2000) account**

Cain & Gair (2000: 23–27) present a simpler picture (based on Cain 2000: 54– 63) of person marking in the contemporary standard language of Malé. According to their analysis, many tenses/aspects/moods show a distinction between an unmarked third-person form on the one hand and a "non-third" (henceforth, "first/second") person form on the other, with no distinction between singular and plural. Tenses/aspects/moods with this marking include the simple present, simple future, simple past, perfect, irrealis, and optative. According to Cain & Gair, the third person is unmarked. They report that the first/second-person marker for the simple present and simple past is -*n*, but they state that the underlying form is -*m*/-*mu*, which appears in certain dialects and sometimes also in literary Dhivehi. Their paradigms for the simple present, simple past, perfect, and simple future are shown in Table 4.3 below.

Table 4.3: Person suffixes according to Cain & Gair (2000)


### Jonathon Lum

### **2.3 Comparison and discussion**

In a review of Fritz (2002), Cain (2004) criticizes Fritz's account of the personmarking system, which he argues is based partly on an incorrect phonological analysis of the nasals *n* and *m*. According to Cain, both nasals neutralize to [ŋ] word-finally (rendered as <n> in writing), and the verbal endings ‑*n*, -*m*, and -*mu* are simply allomorphs of the same suffix. The first of these appears wordfinally, whereas -*m* is used before a vowel, such as when followed by the sentencefinal marker ‑*eve* (Cain 2004: 355). Unfortunately, Cain does not mention what phonetic environment attracts the -*mu* form. He does, however, observe that *mu* can be used with singular subjects, which is contrary to Fritz's account. Cain also points out that the data in Fritz's text materials (Fritz 2002: Vol. 2) sometimes differ significantly from Fritz's own description, with examples like *nikumejjaim-eve* '(I) went out' (2002: Vol. 2, 136) and *ahālaifī-m-eve* 'I asked' (2002: Vol. 2, 141), in which -*m* is used for the first-person singular (rather than -*n*), and also examples like *duśi-n ta?* 'Have (you) seen…?' (2002: (Vol. 2) 154) in which -*n* is used for the second-person singular (rather than -*Ø*).

Tables 4.4 and 4.5<sup>6</sup> below summarize the different accounts of person marking given by Fritz (2002) and Cain & Gair (2000) (whose analysis is also followed by Maumoon 2002) for four tenses/aspects.<sup>7</sup> Suffixes involving -*n* (or -*an* in the future) or ‑*mu* (a possible allomorph or variant) are in bold.


Table 4.4: Person suffixes according to Fritz (2002)

<sup>6</sup>There are also some small but inconsequential differences according to stem type (e.g., Cain & Gair 2000: 23‒27). To simplify matters, this table is intended to represent the suffixes for polysyllabic *a*-stem verbs in particular, though in most respects it is also accurate for verbs of other stem types.

<sup>7</sup>The more marginal irrealis/conditional and potential/optative are not compared here as Fritz (2002) does not explicitly list their forms according to person and number. No other tenses/aspects/moods are considered to have person marking.

### 4 An egophoric analysis of Dhivehi verbal morphology


Table 4.5: Person suffixes according to Cain & Gair (2000)

If we accept the argument that -*mu* is an allomorph or variant of -*n*, then the two accounts come to look more similar than first meets the eye. This would basically reconcile the first-person plural (barring the archaic form -*ū* for the future) and most of the second-person plural across the two accounts. Both accounts already agree entirely on the first-person singular and third-person singular and plural. However, the second-person singular and in some tenses the second-person plural remain problematic. Cain & Gair (2000) consistently group the second person with the first person. On the other hand, Fritz (2002) groups the second-person singular with the third person (though as noted earlier, her data sometimes contradict this), and groups the second-person plural sometimes with the third person and sometimes with the first person, depending on the tense/aspect, as shown in Tables 4.4 and 4.5 above. Evidently, the status of the second person still requires some clarification.

More recently, Gnanadesikan (2017: 138–140) has also analyzed the verbal endings -*n* [ŋ] and ‑*m* as allomorphs of the same suffix, which she writes as -*m̊* to better reflect the underlying form. Based in part on an earlier version of the current chapter, she describes ‑*m̊* as a suffix marking the first person in questions and statements and the second person in questions only, with second-person statements showing the same verbal agreement as the third person. §3 of the current chapter will provide evidence for this distribution, though unlike Gnanadesikan, I argue that the relevant forms are egophoric/alterphoric markers rather than person markers. As for the form ‑*mu*, Gnanadesikan (2017: 138–140) describes it as a "fancy" literary suffix for the first and second person. She notes that the native grammatical tradition (e.g., Ahmad 1970) prescribes -*m*̊ for the first-person singular and -*mu* for the second-person singular and the first- and second-person plural. However, she observes that ‑*mu* is also attested for the first-person singular. The suffix ‑*mu* and its relationship with ‑*m̊* will be discussed further in § 3.4 and §4.

### Jonathon Lum

### **3 Towards an egophoric analysis**

Data collected during my own fieldwork in 2014‒2015 as well as from online sources (see § 1.4) show that several clause types fail to exhibit the person markers in the expected distribution (according to either of the two main accounts discussed in the previous section). Although this data does not contain any previously unattested person markers (with the exception of some distinct future tense forms in the Laamu dialect, for which see §4.1), it shows a distribution of these markers that is sensitive to sentence type. Unexpected marking (or lack of marking) emerges in certain sentences with first-, second-, and even thirdperson subjects. In this section I will argue that the data form a pattern that is mostly consistent with egophoricity rather than person marking, though to an extent some features of person marking are also present. I first present evidence from the distribution of markers across statements and questions (§ 3.1) and then from contexts where third-person nominal reference is used in place of first- or second-person pronouns (§3.2 ). I then show that verbal marking in reported speech is consistent with egophoricity even if it can be explained in other ways (§ 3.3). In §3.4 I revisit the archaic/literary person marker -*mu* (first described in §2 ) and show how on the available evidence, it appears to be a genuine first/second-person marker rather than an egophoric one. Finally, §3.5 sums up the section and compares the Dhivehi system to some other egophoric systems described in the literature, concluding that Dhivehi has elements of both person marking and egophoricity.

### **3.1 Statements and questions**

According to my data, verbs with second-person subjects may pattern either with the first person or with the third person, depending on whether the utterance is a question (first-person pattern is used) or a statement (third-person pattern is used). There is no singular versus plural distinction. This is shown in the examples below:<sup>8</sup>

(1) *miadu* today *ma/aharemen̊* 1SG/1PL *kai-fim̊=ta?* eat.CNV-**PRF.EGO**=Q 'Have I/we eaten today?' (elicited)

<sup>8</sup>Glossing abbreviations used in this chapter mostly follow the Leipzig Glossing Rules, and are listed at the end of this chapter.

In these and subsequent examples, I follow Gnanadesikan (2017) in transcribing /m/ as -*m̊* rather than -*n* word-finally to reflect the underlying form, as described in § 2.3.

4 An egophoric analysis of Dhivehi verbal morphology


This distributional pattern suggests that the relevant verbal marking is not (only) motivated by the grammatical category of person, but by some other organizing principle. What organizing principle could this be? I propose that instead of simply marking person agreement, the suffixes in (1) 9 ‒(6) above (and their equivalents in other tenses/aspects/moods ‒ see §2) behave more like epistemic markers, except possibly in first-person questions (which I return to in later in this section). More precisely, they mark whether or not the subject is also the source of information for the proposition. In most first-person contexts, the subject (i.e., the speaker) is the source of information, since she has the "epistemic

<sup>9</sup>A reviewer asks when such a question would ever be uttered, and whether it would include a pronoun. It is true that some of the elicited examples in (1)‒(6) are somewhat artificial in certain ways, and that pronouns are typically dropped when the referent is obvious from the context (though the verb form does not change), as mentioned in §1.2. It is also true that the necessary contexts for some of the combinations of persons and sentence types are generally unlikely to arise. However, as I discuss later in this chapter, that is part of the point, and helps to explain both the inconsistencies in existing descriptions of the language (§2) and the probable diachronic origins of egophoricity in Dhivehi (§4). Examples (1)‒(6) are simply intended to show the full range of permutations (of persons and sentence types) using minimal pairs and with pronouns included for maximal clarity. While some of these sentences are unlikely to be uttered, some possible contexts are discussed later in §3.1.

### Jonathon Lum

authority" (following Hargreaves 1991; 2005) to report on her own actions, experiences, desires, etc. Second-person questions are similar in that the subject (i.e., the addressee) again has epistemic authority, the question being about the addressee's actions (or experiences, etc.). In second-person statements and thirdperson questions/statements, however, there is a mismatch between the subject and the epistemic source: in second-person statements, the speaker *tells* the addressee about the addressee's actions, and so the speaker is the epistemic source for a statement about another person's actions; in third-person sentences, the epistemic source is the speaker (in statements) or the addressee (in questions), but in either case the subject of the sentence is a third party. This is what motivates the shared marking of first-person statements and second-person questions on the one hand, and the shared marking of second-person statements and thirdperson statements/questions on the other.

As discussed in §1.3, this type of verbal marking is often referred to as a "conjunct-disjunct" system (following Hale 1980) or more recently as "egophoricity" (e.g., Tournadre 1992; 1994; Post 2013; Floyd et al. 2018). In most egophoric systems, the egophoric (or conjunct) form appears where the subject is the epistemic source, and the alterphoric (or non-egophoric, disjunct, etc.) form appears elsewhere. Egophoricity has been documented in a number of Tibeto-Burman languages including Newar (Hale 1980), Lhasa Tibetan (DeLancey 1992; 2001) and Sherpa (Schöttelndreyer 1980; Kelly 2004), other languages of the Himalayas including the Mongolic languages Mongghul (Åkerman 2012), Mangghuer (Slater 2003) and Bonan (Fried 2010) and the Sinitic language Wutun (Sandman 2016), as well as certain languages of the Caucasus (Creissels 2008), South America (Dickinson 2000; Curnow 2002; Bergqvist 2012), and New Guinea (Loughnane 2009; San Roque & Schieffelin 2018). However, as far as I am aware, this type of verbal marking has not been reported for any other Indo-European languages, nor has it been documented in any Dravidian or other contact languages through which the system may have entered into Dhivehi (languages in the region use various kinds of person agreement systems, or have no person agreement at all ‒ see Hock 2016 for an overview).

Under this analysis, what has previously been described as first-person or first/ second-person marking in Dhivehi is in fact egophoric marking (and is glossed as such in examples (1)‒(6)). This includes the suffix -*m̊* for the simple present and simple past, the suffix -*fim̊* for the perfect, and the suffix -*am̊* for the simple future.<sup>10</sup> These markers generally indicate that the epistemic source is the

<sup>10</sup>For reasons of space, I only provide examples of the perfect in (1)‒(6) above, though the same distributional pattern applies in the other tenses/aspects mentioned here. Some examples of these other tenses/aspects will be provided in the remainder of the chapter.

### 4 An egophoric analysis of Dhivehi verbal morphology

subject of the verb. Meanwhile, forms previously described as third person or as second/third person are in fact alterphoric markers: final vowel lengthening for the simple present, -*Ø* for the simple past, ‑*fi* for the perfect, and -*e* for the simple future. These markers are generally used where the epistemic source is not the subject of the verb.

This explanation may partly account for the different paradigms offered by Cain & Gair (2000) and Fritz (2002), if we suppose that Cain & Gair based their analysis of the second person on the interrogative form, while Fritz took the declarative form to be representative at least in the singular (see §2.3). Both descriptions correctly identify marking that is used in second-person contexts, but neither description tells the full story. This is perhaps not altogether surprising, as second-person declaratives are relatively rare in Dhivehi. Of the secondperson statements that do occur, some are nonverbal copular sentences, and many others involve non-finite verbs, IN-verbs, or verbs with progressive/focus marking ‒ verbs in these forms do not carry the suffixes at issue, as mentioned in §1.2. However, example (4) above demonstrates that when the right tense/aspect coincides with a volitional stem in a second-person statement, the verbal marking is the same as for the third person, which can be analyzed as alterphoric marking. Example (7) below (from a website of Dhivehi stories) contains two further instances of alterphoric marking in second-person statements, this time in the future tense:<sup>11</sup>

(7) *kalē* 2SG *bēnum̊* want *nu=vi=yas* NEG=be.PST.PTCP=CNCS *kalē-ge* 2SG-GEN *zamīru* conscience *kuran̊* do.INF *bēnun̊\_ve=gen̊* want\_be.CVB=SUCC *kalē* **2SG** *ti=kam̊* **DEM2=action** *kurāne.* **do.FUT.ALTER** *kale=akī* 2SG=COP *vakīl-ek̊.* lawyer-INDF *ēnā-ge* **3SG-GEN** *furāna* **life** *salāmat̊\_kuran̊* **safety\_do.INF** *kalē* **2SG** *masakkat̊* **work** *kurāne* **do.FUT.ALTER**

'Even if you don't want to, because your conscience wants to do [that], you will do that. You are a lawyer. You will work to save his life.' (from www.esfiya.com/1849/)

While the different verbal marking in second-person statements and questions points to an egophoric system, a potential problem for this analysis is the be-

<sup>11</sup>Note that the first clause in (7) has a second-person subject but as a non-finite clause does not show egophoric/alterphoric marking. Also note that the second sentence in (7) (translating to 'You are a lawyer') has a second-person subject, but as a copular sentence in Dhivehi it lacks a verb.

### Jonathon Lum

haviour of first-person questions. In typical egophoric systems, first-person questions pattern like second-person statements and third-person statements/questions, and show different marking to first-person statements and second-person questions (e.g., Hale 1980). In first-person questions, *I* ask *you* about myself, and the addressee (temporarily) has epistemic authority over the speaker's actions, experiences, etc., which are usually in the speaker's own epistemic territory. Since the subject (in this case the speaker) is not the epistemic source in firstperson questions, alterphoric rather than egophoric marking is expected in this context, and indeed has been reported in typical egophoric systems (San Roque et al. 2018: 4–5). Dhivehi, however, uses egophoric marking (or first/second-person marking under previous analyses) in first-person questions, as shown in (1) earlier. This situation is summarized in Table 4.6 and Table 4.7 below.

Table 4.6: Typical distribution of egophoric and alterphoric markers in egophoric systems


Table 4.7: Distribution of egophoric and alterphoric markers in Dhivehi


In Dhivehi, verbal marking in first-person contexts therefore looks like genuine person agreement (despite being glossed here as EGO), while verbal marking in second-person contexts looks like egophoricity, and verbal marking in thirdperson contexts is consistent with both systems. There are at least three ways to interpret this type of distribution: (i) as a person-marking system with a quirk in the second person; (ii) as a hybrid of person marking and egophoricity (perhaps representing a transitional phase in the diachronic development of one system into the other); or (iii) as an egophoric system with a non-canonical distribution

### 4 An egophoric analysis of Dhivehi verbal morphology

of markers. These three analyses lie on a cline ‒ for example, a non-canonical egophoric system may display elements of person marking, and a hybrid system may be closer to the egophoric end or to the person-marking end. On the basis of the evidence introduced thus far, it is somewhat difficult to decide, and a conservative approach would probably be to advocate (i), considering that egophoricity is rare cross-linguistically (and is in fact unattested in the Indo-European family and southern South Asia region). However, as I will show, there are good reasons to think that (ii) and/or (iii) may be correct, and that egophoricity is thus a part of the Dhivehi verbal system.

Firstly, as others have also noted (e.g., San Roque et al. 2018: 26–27), firstperson questions are usually pragmatically marked, as it is uncommon for the answer to be genuinely unknown to the speaker. For example, a first-person question may be posed to test the addressee's knowledge of the speaker, or it may be a rhetorical one. Example (8) below contains two first-person questions, both of which appear to be rhetorical:

(8) *ekamaku* but *balā\_bala…* look.CVB\_IMP *aharen̊* 1SG *moya* crazy *kam-ek̊* action-INDF *kura-m̊=ta?* do.PRS-EGO=Q *nūnī* or *duvah-aku=ves* day-UNSP=EMPH *kuri-m̊=ta?* **do.PST-EGO=Q** 'But look…do I do anything crazy? Or did [I] ever do [anything crazy]?' (from www.vaguthu.mv/evaguthu/story/210155/)

Rhetorical questions are problematic because the speaker believes ‒ and in fact advertises ‒ that she already knows the answer to the question. As such, rhetorical questions do not truly bestow epistemic authority upon the addressee, and unsurprisingly in some egophoric systems rhetorical first-person questions may attract egophoric marking (see Hale & Watters 1973 for Newar). It is difficult to find first-person questions in Dhivehi that are unambiguously "genuine" as opposed to rhetorical. The example in (9) below is a good candidate, though the question has a permission reading and so is not a real request for information:

(9) *aharen̊* 1SG *ja*[]*burōl-ek̊* rose.apple-INDF *naga-m̊=ta?* **take.PRS-EGO=Q** 'Can I take a rose apple?' (from www.dhiggaru.com/946)

Since first-person questions in Dhivehi are rarely genuine requests for information, their use of egophoric marking is still in keeping with a system that is at least partly egophoric in nature. While the use of egophoric marking in

### Jonathon Lum

first-person questions is still unusual cross-linguistically, it is not completely unattested. The same distributional pattern is found in the future tense of Kaluli (Trans New Guinea), and has been analyzed as being partly egophoric (San Roque & Schieffelin 2018). Moreover, the exact distribution of egophoric markers varies considerably across egophoric systems anyway (see San Roque et al. 2018 for an overview), and the use of egophoric marking in first-person questions is arguably only a relatively small departure from the canonical system described earlier.

Secondly, some Dhivehi speakers accept alterphoric marking in at least some first-person questions. This was the case for one of my consultants (a 34-year-old man from Fonadhoo, Laamu Atoll), who accepted alterphoric marking in firstperson questions directed at others, but in self-directed first-person questions only accepted egophoric marking, as shown in (10) vs. (11) below:


According to this consultant, (10) might be used by an old man who has forgotten if he has already eaten that day and is asking somebody to remind him, while (11) might be used by the same old man talking to himself. This distinction is interesting because it relates to epistemic authority: in self-directed questions, epistemic authority remains with the speaker, but in (non-rhetorical) addresseedirected questions, epistemic authority is with the addressee. The use of egophoric marking in (11) and alterphoric marking in (10) is therefore entirely consistent with an egophoric system, but is inconsistent with a person-marking analysis. However, other consultants in Laamu and Malé rejected such a distinction, accepting only egophoric marking in both contexts. This may partly reflect the difficulties of eliciting such an unusual (and pragmatically marked) sentence type, but probably does nonetheless point to a general preference for egophoric marking in all first-person questions. Still, this general preference is not absolute, and it is possible that speakers' differing intuitions reflect a change in progress (a point I return to in §4).

### **3.2 Pronoun avoidance and the use of third-person nominal reference**

Aside from the distribution of forms across sentence types, there is another piece of evidence for egophoricity in Dhivehi: the use of egophoric markers in sen-

### 4 An egophoric analysis of Dhivehi verbal morphology

tences where speakers refer to themselves or their addressees with third-person nominal reference, such as a kin term, name or title. Because such references are strictly speaking third-person forms, when they are the subject of a verb they would be expected to trigger third-person agreement if the language uses a canonical person-marking system. But in Dhivehi, this context triggers egophoric marking (or first/second-person marking under previous analyses). For example, in (12) below, the speaker asks her mother if she (the mother) has eaten:

(12) *mamma* mother *kai-fim̊=ta?* eat.CVB-**PRF.EGO**=Q *kobā* where *Shihānā=āi* Shihaanaa=CONJ *donta?* sister *emīhun̊* 3PL *kai-fi=ta?* eat.CVB-PRF.ALTER=Q 'Has mother [=addressee] eaten? Where are Shihaanaa and sister? Have they eaten?'(from http://vnews.mv/517)

In (12), the kin term *mamma* 'mother' is used in lieu of the second-person pronoun *kalē* 'you'.<sup>12</sup> Despite the use of this third-person form as subject, the perfect egophoric suffix -*fim̊* (normally associated with first/second person) is used. This is very different to the expected marking in a person-agreement system (cf. the English question *Is sir ready to order?* which shows third-person agreement ‒ \**Are sir ready to order?* is ungrammatical). The use of egophoric marking in (12) therefore appears to be motivated by the fact that the mother is both the subject and the epistemic source, regardless of whether the speaker refers to her in the second or third person. Note that although in some egophoric systems egophoric marking can be triggered when the subject is a close relative of the speaker (San Roque et al. 2018: 33), the egophoric marking in (12) only relates to the fact that the subject is the epistemic source, regardless of the relationship with the speaker. The second question in (12) is about some other close relatives/associates of the speaker, but uses alterphoric marking because the question is not actually posed to them. And if the question about the mother were instead directed at somebody else, the alterphoric perfect suffix -*fi* would be used, as in (13) below:

(13) *mamma kai-fi=ta?*

> mother eat.CVB-**PRF.ALTER**=Q

'Has mother eaten?' (not directed at mother) (elicited)

<sup>12</sup>That *mamma* is the subject of the verb *kaifim̊* and not a free-standing vocative expression is supported by the fact that both words would belong to the same intonation unit if the sentence were used in speech.

### Jonathon Lum

Referring to one's addressee by a name, kin term or title is extremely common in Dhivehi, largely because the second-person pronoun *kalē* 'you' is now generally regarded as impolite (see Gnanadesikan 2017: 70). Somewhat less often, speakers also refer to themselves in the third person. This occurs especially in child-directed speech. An example is (14) below, where a mother is telling her child where she (the mother) went the other day:

(14) *kurin̊* earlier *duvah-aku=ves* day-UNSP=EMPH *mamma* mother *e=ge-aṣ̊* DEM3=house-DAT *diya-im̊* go.PST-**EGO** 'The other day as well mother [=speaker] went to that house.' (from www.dhivehivaahaka.com/read/601)

Again, the use of egophoric marking with a formally third-person subject would be anomalous in a person-marking system (though see (26) in § 3.4), but is entirely consistent with an egophoric system that is sensitive to epistemic roles. In this case, the speaker is the epistemic source for the proposition, and so the use of egophoric marking is well motivated even though she refers to herself in the third person.

Sentences in which speakers use third-person pronominal reference in place of first- or second-person pronouns as subjects provide a useful window on the underlying nature of markers which in many other contexts may look equally like person markers or epistemic (i.e., egophoric/alterphoric) markers. This grammatical context has hardly been explored in the egophoricity literature, though the general prediction would be for true egophoric markers to follow the pattern illustrated for Dhivehi in (12) and (14), and for person markers (and perhaps some "hybrid" markers) to be sensitive to the way in which the subject of the verb is formally expressed.

### **3.3 Reported speech**

Like many other languages with egophoric systems (e.g., Newar, Hale 1980), Dhivehi makes use of an egophoric/alterphoric opposition in reported speech. Egophoric marking appears where the reported subject (i.e., the subject of the reported speech) is the same as the subject of the matrix clause, and an alterphoric form appears when there is a mismatch between subjects. This is presumably because egophoric markers are used when the epistemic source is also the subject of the clause (cf. §3.1) ‒ in reported speech clauses, the epistemic source is the person reporting on what was said (i.e., the subject of the matrix clause), and so egophoric marking appears on the reported verb only when the reported subject

### 4 An egophoric analysis of Dhivehi verbal morphology

and matrix subject are co-referential. Thus, where the subject of the matrix clause is the speaker, egophoric marking appears only if the speaker is the subject of the embedded clause, such as in (15) as opposed to (16) and (17):<sup>13</sup>


Where the subject of the matrix clause is the addressee, egophoric marking appears only if the reported subject is also the addressee:


And where the subject of the matrix clause is a third party, egophoric marking appears only if that third party is also the reported subject:

(21) *Ali* Ali *bunī* say.PST.FOC *ma* 1SG *Māle* Malé *diya=yē* go.PST-**ALTER**=qUOT 'Ali said that I went to Malé.' (elicited)

<sup>13</sup>Note that the egophoric marker (or first/second-person marker under previous analyses) in these examples is ‑*im̊* rather than ‑*m̊* (the form presented in §2 for the simple past) because the verb *diya* 'go.PST' is a monosyllabic-stem verb rather than a polysyllabic *a*-stem verb (see Gnanadesikan 2017: 145–146).

### Jonathon Lum


However, even though the reported speech data is perfectly consistent with egophoricity, there is another possible explanation. Existing descriptions of Dhivehi analyze the marker =*ē* (allomorph *yē*) simply as a marker of direct quotations (Cain & Gair 2000: 47; Gnanadesikan 2017: 302), even if they sometimes note that the pronoun identity in the original utterance is not always the same in the quotation (Gnanadesikan 2017: 302–303). In practice, pronouns and other noun phrases are often omitted when they are obvious from the context, and so in many cases one cannot tell for sure whether the omitted pronoun would have been faithful to the original utterance or whether it would have been deployed from the current speaker's perspective. For example, if *ēnā* '3SG' in (23) had been omitted (as is both possible and idiomatic in Dhivehi), the sentence could perhaps be analyzed as containing a direct quotation with a first-person subject (i.e., 'Ali said, "[I] went to Malé"'). However, the examples in (15)‒(24) show that when a pronoun is included, it is deployed from the perspective of the current speaker rather than the original speaker. This may be because the inclusion of a pronoun, being unusual, is pragmatically marked, and is more likely to occur in emphatic contexts where the speaker feels a need to draw attention to the identity of the reported subject. This is most easily done from the speaker's perspective in the current speech context. The data in (15)‒(24) therefore show elements of both direct and indirect speech: the pronoun in the reported quote is deployed from the perspective of the current speaker, while the marking on the reported verb is calculated from the perspective of the original speaker. This pattern, known as "semi-(in)direct speech" (e.g., Aikhenvald 2008; 2011), "hybrid reported speech" (Tournadre & Dorje 2003), or "deictically mixed speech" (Widmer & Zemp 2017) among other terms (see Evans 2012), is not inconsistent with

### 4 An egophoric analysis of Dhivehi verbal morphology

a person-marking analysis because in semi-direct speech, person marking on the reported verb does not have to agree with the reported subject.<sup>14</sup>

Dhivehi also has a logophoric pronoun *timannā* (plural *timannāmen̊*) that is sometimes used in reported speech to refer to an embedded/reported subject that is co-referential with the matrix subject (see Gnanadesikan 2017: 96–97). This pronoun occurs with egophoric (or "first-person") marking on the verb. An example is (25) below:

(25) *Maumūn̊* Maumoon *amilla-aṣ̊* self-ADVZ *bunī* say.PST.FOC *timannā* **LOG** *30* 30 *aharu* year *verikam̊* rulership *koṣ̊-fīm=ē* do.CVB-PRF.EGO=qUOT 'Maumoon himself said that he had ruled for 30 years.' (adapted from Gnanadesikan 2017: 96)

However, there is no distinct logophoric marking on verbs in Dhivehi; instead, the same egophoric/alterphoric markers (or first/second-person vs. third-person markers under previous analyses) are available, as illustrated in (25) and in the examples earlier in this section. It is therefore not entirely clear how or whether the logophoric pronoun *timannā* relates to egophoricity in Dhivehi. It appears to simply be a special pronoun used in some cases where the reported subject and matrix subject are co-referential, and has no bearing on the marking of the reported verb (which would still attract egophoric/first-person marking even if the pronoun were deployed from the perspective of the current speaker or omitted entirely).

### **3.4 The suffix -***mu*

§ 3.1 and § 3.2 presented evidence for egophoricity in Dhivehi, with the caveat that verbal marking in the unusual context of first-person questions may point to the Dhivehi system being a hybrid of person-marking and egophoricity. I now turn to another possible piece of evidence that person-marking is present in Dhivehi: the distribution of the archaic/literary person marker -*mu* (first introduced in § 2).

<sup>14</sup>Various kinds of semi-direct reported speech constructions, or fuzzy boundaries between direct and indirect speech, are attested in South Asia (e.g., Tamil: Lehmann 1989: 373–375; Malayalam: Asher & Kumari 1997: 2–7). Masica (1991: 403) observes that reported speech constructions in Sinhala and many other Indo-Aryan languages may be Dravidian calques, and that in some Indo-Aryan languages there is "no clear distinction between indirect and direct quotation".

### Jonathon Lum

Recall that Fritz (2002) identifies -*mu* as a first/second-person plural suffix, while Cain & Gair (2000) and Gnanadesikan (2017) analyze it as an archaic/literary suffix for the first and second person (regardless of number), though it was traditionally prescribed for the second-person singular and first/second-person plural (e.g., Ahmad 1970). Fritz (2002: 169) and Cain & Gair (2000: 27) assume a historical connection between -*mu* and -*m̊* in at least some parts of the paradigm. This raises the question of whether -*mu* has (or had) much the same distribution as -*m̊*, i.e., first-person statements/questions (though possibly restricted to firstperson plural) and second-person questions, but not elsewhere. The evidence appears to be mixed. On the one hand, -*mu* is sometimes found in second-person statements (as well as the expected contexts of first-person statements/questions and second-person questions).<sup>15</sup> On the other hand, speakers tend to reject the use of -*mu* in second-person statements, accepting it only in first-person statements/questions and second-person questions. In any case, ‑*mu* is now seldom used in spoken language, being mostly restricted to literary contexts.

The example in (26) below shows ‑*mu* in a statement with a second-person singular subject:<sup>16</sup>

(26) *manikufānu=eve!* excellency=END *qānūnu\_asāsī* law\_basis *galu\_aḷā=fai\_ot̊* stone\_put.down.CVB=SUCC\_lie.PST.PTCP *duvas\_varu* day\_amount *manikufānu* excellency *vidāḷu\_vī-mu=eve.* say.HHON\_be.PST-**1**/**2**=END 'Your excellency! In the days when the [preparation of the] constitution had stalled [lit. 'had been hooked on a rock'], your excellency said [it].' (A. Gnanadesikan, pers. comm.; originally from *Haama Daily* online newspaper, 2010)

In (26), -*mu* cannot straightforwardly be analyzed as an egophoric marker because the subject of the verb *vidāḷu\_vī-mu* is the addressee rather than the

<sup>15</sup>Thanks to Amalia Gnanadesikan for bringing this point to my attention.

<sup>16</sup>The subject of this sentence is the title *manikufānu* '(your) excellency', which Fritz (2002: 136) analyzes as a second-person pronoun used in reference to members of the highest level of society (such as the president), though it could alternatively be regarded as a (formally thirdperson) noun ‒ the boundary here is unclear as some other Dhivehi pronouns derive historically from nouns, such as the deferential first-person pronoun *aḷuga ḍu* (lit. 'slave piece'). If *manikufānu* is analyzed as a third-person reference, the use of second-person agreement in (26) would be odd for a person-marking system, following the discussion in § 3.2. The same issue applies to *tiyabaimīhun̊* (lit. 'that group of people near you') which is used in (27) as a second-person plural pronoun.

### 4 An egophoric analysis of Dhivehi verbal morphology

speaker, and the sentence is a statement. It therefore appears to be a secondperson marker in this context. Further, (27) below shows -*mu* (in the perfect form ‑*fīmu*) with a second-person plural subject:

(27) *tiyabaimīhun̊* 2PL *timannāmen̊-ge* LOG.PL-GEN *gedor-aṣ̊* house-DAT *vade\_gane* enter.CVB\_take.CVB *h̤amalā\_dī* attack\_give.CVB *e=tan̊*∼*tan̊* DEM3=place∼REDUP *halāku* damage *kos-fīmu=eve. ̊* do.CVB-**PRF**.**1**/**2**=END 'You people have come to our homes, attacked, and damaged them.' (A. Gnanadesikan, pers. comm.; originally from *Haama Daily* online newspaper, 2010)

In (27) too, the use of -*fīmu* cannot be straightforwardly analyzed as egophoric, because the subject is second person and the sentence is a statement.<sup>17</sup> Nonetheless, the suffix ‑*(fī)mu* is accepted by most of my consultants only in secondperson questions as well as first-person contexts, though some who are familiar with prescriptive grammar books claim that it is also "correct" in second-person statements, even if they would not personally use it in that context. Additionally, ‑*mu* can apparently be used in first-person singular contexts, contrary to Fritz (2002) and the prescriptive guides (e.g., Ahmad 1970) which list it only for the first-person plural and the second person. This is shown in (28) below:

(28) *aharen̊* 1SG *e=ge-aṣ̊* DEM3=house-DAT *diyai-mu* go.PST-**1**/**2** 'I went to that house.' (elicited)

Thus, for many speakers at least, -*mu* and -*m̊* are basically the same, though ‑*mu* is generally regarded as an archaic, literary, or "fancy" form. Still, because ‑*mu* does sometimes appear in second-person statements (as shown in (26) and (27) above), I analyze it conservatively as a first/second-person marker rather than as an egophoric marker. However, more work needs to be done to explain the fact that traditional prescription, actual usage, and speaker judgements each paint somewhat different pictures of ‑*mu*. The diachronic account that will be proposed in §4.3 goes some way towards addressing this.

<sup>17</sup>Note that (27) also includes the (plural) logophoric pronoun *timannāmen* because it is taken from a larger quotation in which the speaker is cross-referenced.

### Jonathon Lum

### **3.5 Discussion**

The data in some of the previous sections are problematic for a simple personagreement analysis of Dhivehi verbs. Not only does "first/second-person" marking fail to appear in some second-person contexts (§ 3.1), it actually appears in some "third-person" contexts (§ 3.2). This data is, however, consistent with egophoricity. On the other hand, the literary ‑*mu* (§ 3.4) does appear to be a genuine person marker (though it is falling out of use even in literary contexts, and current speaker judgements often do not match traditional prescription with regard to its distribution), and the marking of verbs in first-person questions (§3.1) could also be regarded as evidence of a person-agreement system (though as discussed, the data here are not entirely incompatible with egophoricity either). The data from reported speech (§ 3.3) are equally consistent with egophoricity and personmarking, assuming a semi-direct speech construction in the case of the latter. On the whole, the evidence therefore points to a mixture of egophoricity and personmarking in Dhivehi, with the language seemingly moving closer to egophoricity with the decline of the archaic/literary person marker ‑*mu* (see §4.3 for more on this).

While egophoricity appears to be a good explanation for (much of) the data, such a grammatical system is typologically unusual, and Dhivehi would be the first Indo-European language reported to have such a system. This raises questions as to whether there might be any other ways to account for the data. Gawne (2017: 83–84) points out that individual features within egophoric "systems" may or may not overlap in different languages. According toGawne, the co-occurrence of certain constituent features (such as certain evidential markers and a "rule of anticipation" in which questions pre-empt the person marking of the anticipated response) may result in an epiphenomenally egophoric pattern.

For Dhivehi, some relevant constituent features are: (i) second-person statements marked like third person, but second-person questions marked like first person (perhaps under a "rule of anticipation" which also extends to first-person questions for some speakers); (ii) egophoric markers (or under some previous analyses, first/second-person markers) used for co-reference in reported speech (or alternatively, a pattern of semi-direct speech); and (iii) marking on verbs sensitive to the discourse context, rather than the formal expression of the subject ‒ e.g., a speaker using a name or noun phrase to refer to herself uses egophoric (or first/second-person) marking on the verb, rather than alterphoric (or thirdperson) marking. It is possible that these three features are independent phenomena in Dhivehi, and that they just happen to co-occur in such a way that gives the appearance of an underlying epistemic system. This kind of explanation might

### 4 An egophoric analysis of Dhivehi verbal morphology

be advantageous in that it avoids appealing to a typologically rare grammatical "system" that is completely unexpected in the region and language family. On the other hand, it is obviously more parsimonious to appeal to a single phenomenon that can explain the various constituent features, some of which would be unusual for the region and language family anyway. Throughout this chapter I adopt the more parsimonious analysis, but it is not possible to completely discount the notion that egophoricity in Dhivehi may be epiphenomenal. At the very least though, I hope to have demonstrated that Dhivehi shares a number of interesting features with other "egophoric" languages in the literature, and that an egophoric analysis may be just as appropriate for Dhivehi as it is for many of those languages.

How then does the Dhivehi pattern compare to other examples of egophoricity in the literature? I have already mentioned that the Dhivehi distribution of egophoric markers resembles the distribution of future tense forms in Kaluli (Trans New Guinea; San Roque & Schieffelin 2018), which is slightly different to the distribution found in canonical egophoric systems. I have also discussed similarities between reported speech in Dhivehi (which involves a particular distribution of egophoric markers) and reported speech in other egophoric languages. But aside from the grammatical distribution of egophoric markers, what about possible connections with systems of evidentiality, mirativity, or volitionality, which often interact with egophoricity (e.g., Creissels 2008; San Roque et al. 2018)? In some languages (e.g., Newar: Hale 1980), for example, speakers can use alterphoric markers on verbs with first- or second-person subjects to show that the subject acted without volition, while egophoric marking is restricted to verbs that describe intentional acts. This is not the case for Dhivehi, in which accidental or involuntary events are encoded by separate verbal morphology known as the "inactive/intransitive/involitive" or "IN"-form, introduced in § 1.2. <sup>18</sup> IN-verbs do not take any kind of person marking or egophoric/alterphoric marking, as shown in (29) below:

	- 'I/(s)he closed the door (accidentally).' (adapted from Cain & Gair 2000: 58)

Dhivehi's egophoric/alterphoric opposition therefore applies only to volitionalstem verbs, and alterphoric forms within this opposition are never deployed to show a lack of volition. For example, (30) below is ungrammatical:

<sup>18</sup>The IN-form of a typical Dhivehi verb is derived via an umlaut process and/or the addition of a dedicated suffix, depending on the stem type of the verb (Cain & Gair 2000: 57–61).

### Jonathon Lum

(30) *\*aharen̊* 1SG *doru* door *leppi* close.PST.ALTER 'I closed the door (accidentally).'

Hence, although Dhivehi has a regular system for marking volitional vs. nonvolitional distinctions on verbs, it is separate from the egophoric/alterphoric system which only comes into play for active, volitional verbs (and even then only for non-progressive aspects). Nonetheless, it is curious that many other languages with egophoricity also attend to volitionality, and that the egophoric form in these languages is often restricted to verbs describing intentional, controllable actions (San Roque et al. 2018: 14–15, 29–30). It therefore seems plausible that there may be an underlying relationship between the egophoric system and the volitional system in Dhivehi, though for now the nature of that relationship is unclear.

Thus far, I have outlined the issues with previous accounts of Dhivehi verbal morphology which revolve around typical notions of person marking, and have shown that an egophoric analysis appears to be a better fit for the data in many grammatical contexts. There are of course some limitations and counterexamples to an egophoric analysis too. However, these are in keeping with the widespread variation found in egophoric systems, and/or reflect a combination of egophoricity and person marking in Dhivehi. While it is conceivable that the Dhivehi data can be explained in terms of the co-occurrence of a number of separate, possibly unrelated grammatical and pragmatic phenomena, this is also true of other egophoric systems reported in the literature. It is therefore valid to discuss Dhivehi in terms of egophoricity and to consider its potential contribution to our understanding of egophoricity cross-linguistically.

### **4 Dialectal variation and historical development**

As mentioned in earlier parts of this chapter, egophoric marking has not been reported for the Indo-European or Dravidian language families nor for any other languages with which Dhivehi has had historical contact. In order to shed light on how egophoricity came to emerge in the standard Malé dialect from which the data in the previous sections were drawn, it may be instructive to consider the southern dialects of Huvadhu, Fuvahmulah and Addu (briefly introduced in §1.2), which are in most respects more conservative than Malé Dhivehi (Fritz 2002: 13). In addition, the Laamu and Minicoy dialects are spoken at the extreme ends of the northern dialect group (see Figure 4.1 in §1.2) and are distinct from the

### 4 An egophoric analysis of Dhivehi verbal morphology

Malé dialect in many ways. In the following sections I briefly summarize what is known about "person marking" in the non-Malé dialects for which information is available: §4.1 on the Laamu dialect and §4.2 on the dialects of Fuvahmulah and Addu. In § 4.3 I then propose that the northern dialects may have undergone a similar process to that outlined in Widmer (2015) and Widmer & Zemp (2017) for some Tibeto-Burman languages, in which the distribution of person markers in semi-direct speech fosters a reanalysis of those markers as egophoric/alterphoric markers.

### **4.1 Laamu**

In the dialect of Laamu Atoll (traditionally known as *Haddummatī bas*), verbal marking is mostly identical to that in the standard dialect, according to data collected during my own recent fieldwork. One salient but inconsequential difference is that the progressive/focus suffix is ‑*(n)ū* rather than ‑*(n)ī* (e.g., *danū* 'go.PRS.PROG'). The suffixes for the simple present, perfect, and simple past are the same as those in Malé. Future-tense forms in Laamu are different, however: egophoric forms end in ‑*m̊* instead of the Malé -*nam̊* and alterphoric forms end in -*ḷa* instead of the Malé -*ne*. As with the corresponding Malé forms, the main evidence that these are egophoric and alterphoric markers respectively is their distribution in second-person clauses (‑*m̊* for questions, ‑*ḷa* for statements) and in clauses where a speaker/addressee subject is expressed with third-person nominal reference (-*m̊* used where the subject is the epistemic source, regardless of the formal expression of the subject). For example, (31) and (32) below demonstrate that -*m̊* behaves as an egophoric marker in second-person clauses while -*ḷa* acts as an alterphoric marker:


### **4.2 Fuvahmulah and Addu**

According to Fritz (2002: 164–184), the southern dialects of Fuvahmulah and Addu have comparatively richer systems of person marking. Fuvahmulah distinguishes between all six person and number combinations, like Literary Sinhala.

### Jonathon Lum

Addu distinguishes between first-person singular, second/third-person singular, first-person plural, and second/third-person plural.<sup>19</sup> These patterns are demonstrated along with the Malé paradigm in Table 4.8 below for the present tense of the verb *balanī* 'looking'.<sup>20</sup>


Table 4.8: The simple present tense of *balanī* 'looking' in three dialects (adapted from Fritz 2002: 169)

Recall that Fritz's account of the Malé dialect does not mention the split between second-person statements and questions, and this raises the question of whether there might be a similar split in the southern dialects too. However, on the basis of Fritz's examples, as well as some reports I have received from native speakers of the Addu dialect, it appears that Addu does have a genuine personagreement system that contrasts first-person with second/third-person and singular with plural, without any egophoric-like distribution across sentence types. Fuvahmulah has a fully-fledged person-agreement system according to Fritz's description, contrasting first-, second-, and third-person in both the singular and plural. Thus on the available evidence, these southern dialects use personmarking systems rather than egophoric ones. It is relevant to note that these person-marking systems are relatively conservative ‒ the Fuvahmulah system in particular closely resembles the six-way system in Literary Sinhala (Fritz 2002: 168–175), the language most closely related to Dhivehi.<sup>21</sup>

<sup>19</sup>In addition, Addu has some special interrogative forms for certain persons and numbers depending on the stem type and tense, resulting in slightly richer interrogative "paradigms" (Fritz 2002: 244–247). However, as these are not in egophoric distribution and Fritz has a plausible phonological explanation for them, I will not discuss them further here.

<sup>20</sup>Note that the forms listed here for Malé are based on the data in §3 rather than Fritz's description, and that for simplicity I omit forms involving the archaic/literary suffix ‑*mu*.

<sup>21</sup>In contrast, Colloquial Sinhala does not have any person or number agreement on verbs (Gair 1990: 15).

### 4 An egophoric analysis of Dhivehi verbal morphology

### **4.3 Possible origins of egophoricity in Dhivehi**

The previous sections showed that Dhivehi's conservative southern dialects retain person-agreement systems, while the northern dialects of Laamu and Malé instead use egophoric systems with some elements of person agreement. But how did egophoricity come to exist in northern Dhivehi? Given that the egophoric and alterphoric markers in the northern dialects are highly similar (and in some cases identical) to the person markers in the southern dialects and in Literary Sinhala (see Table 4.8 in the previous section as well as Fritz 2002: 168–175), it is highly likely that egophoricity in northern Dhivehi is a recent development from a purely person-marking system. This is also supported by a number of other observations: languages of the region have person agreement rather than egophoricity (Hock 2016), prescriptive grammar books in Dhivehi list personagreement suffixes (e.g., Ahmad 1970), an early description of Dhivehi suggests a person-marking system (Geiger 1919), and even more recent descriptions also report person marking, as discussed in § 2. Thus, all indications point to a personmarking system existing in the language until quite recently, when the person markers must have changed in their distribution across sentence types, resulting in the egophoric distribution described in §3. Although the full diachronic development is difficult to reconstruct precisely, some aspects are reasonably clear and others may be inferred. In this section I will first sketch the likely development of person marking in Dhivehi until it reached the system described by Ahmad (1970), and then I will address the likely development of that person-marking system into the egophoric system used in the modern standard variety (and other northern dialects).

It is highly probable that Dhivehi once had a six-way distinction along the lines of Literary Sinhala, but at some point this system began to collapse except on Fuvahmulah, where the original system survived mostly unchanged (see Table 4.8 in the previous section). In Addu the distinction between the old secondand third-person markers was lost, and it appears that northern Dhivehi also experienced this change, given that it currently lacks dedicated second-person markers. The extant forms (in statements) are based on the old third-person ones, though northern Dhivehi has also lost its third-person plural marker (which was still attested at the time of Geiger 1919), extending the singular form to the plural as well.

The suffix ‑*mu* (§ 3.4) in northern Dhivehi must be related to the Literary Sinhala first-person plural suffix of the same form (Fritz 2002: 168), and also to the first-person plural suffix ‑*mā* in the southern dialects. However, as discussed in §3.4, the northern Dhivehi suffix ‑*mu* is used also for the second person and

### Jonathon Lum

sometimes for the first-person singular. At some point, ‑*mu* must have spread to second-person contexts, though the spread to the first-person singular may be much more recent ‒ according to Geiger (1919) and the native grammatical tradition (e.g., Ahmad 1970), ‑*mu* is used only for the first-person plural and second person (singular and plural). More recently (i.e., in the later part of the 20ℎ century), ‑*mu* started to fall out of use, and it is possible that those speakers (or writers) who use it at all in first-person singular contexts have reanalyzed it simply as a 'fancy' equivalent of ‑*m̊* (which is also used for the first and second person).

As for ‑*m̊* itself, there is a clear relationship with the Literary Sinhala firstperson singular suffix -*m* (Fritz 2002: 168) as well as with -*m̊* in the southern dialects, where it also marks the first-person singular. The appearance of -*m̊* (in northern Dhivehi) in second-person questions and in the first-person plural is difficult to date precisely, though the evidence suggests this unfolded in the late 20ℎ century. Geiger (1919: 83–88) presents some examples of ‑*m* in these contexts (e.g., *aharamen̊ kakkāfīm* 'we cooked'),<sup>22</sup> probably a reduction of ‑*mu*, and not far off the present-day ‑*m̊* (pronounced [ŋ] word-finally). Prescriptivists writing in the second half of the century (e.g., Ahmad 1970) do not comment on ‑*m* or ‑*m̊* in these contexts, however. This suggests at the very least that ‑*m̊* in these contexts was not standard practice in writing by the 1960s, and probably also indicates that it was not yet widely used in speech or writing (else it would have been remarked upon, even if only to proscribe its use). However, descriptions of the language around the turn of the century (Cain & Gair 2000; Fritz 2002) have reported at least some aspects of the new distribution, as discussed in §2. Quite possibly, the spread of -*m̊* is directly related to the decline of ‑*mu*, which would have occurred during much the same period ‒ I return to this point later.

Although ‑*mu* could have reduced to -*m*̊ through loss of the word-final ‑*u* as Fritz (2002: 169) suggests, this alone does not explain why -*m̊* is not also used in second-person statements (recall that -*mu* was formerly used with both firstand second-person subjects). The correct explanation must account for why -*mu* was completely lost in second-person statements and replaced by third-person forms in that grammatical environment. The diachronic process I wish to propose here accounts for this: Dhivehi speakers reanalyzed person markers as epistemic markers because of their distribution in semi-direct speech, and when this epistemic reanalysis was overgeneralized to basic clauses, former second-person marking ‒ now reanalyzed as egophoric marking ‒ disappeared from secondperson statements but not from second-person questions. This process is very

<sup>22</sup>Transliteration adapted.

### 4 An egophoric analysis of Dhivehi verbal morphology

similar to the one recently described by Widmer (2015) and Widmer & Zemp (2017) for the Tibeto-Burman language Bunan, but with some differences that will be addressed later in this section.

In Dhivehi, the -*u* of -*mu* is deleted before the quotative marker =*ē* (see §3.3) in semi-direct speech, resulting in [m]. This [m] could have been reanalyzed as /m/ underlyingly (and is written in this chapter as *m̊* word-finally or *m* elsewhere, following Gnanadesikan 2017). This is also identical to the pre-existing first-person singular suffix, and so even before any reanalysis took place, [m] would have been used in all quotations where the original speaker was first or second person, as semi-direct speech in Dhivehi preserves the person marking of the original utterance (see §3.3).

But a functional reanalysis of this marker in semi-direct speech must have also taken place. Semi-direct speech is the only grammatical construction in Dhivehi that allows a mismatch between a subject pronoun (or noun phrase) and the marking on a finite verb, the former being calculated from the current speaker's perspective and the latter from the original speaker's perspective. This makes it the most likely grammatical construction to lend itself to a reanalysis of person markers (cf. Widmer & Zemp 2017: 54–56). For example, the first-person marking on the reported verb in a sentence like 'Ali said that he went to Malé' could be reanalyzed as marking co-reference of the matrix subject with the reported subject, or even as marking that the reported subject is the epistemic source (i.e., egophoric marking), especially since the otherwise similar sentence 'Ali said that he went to Malé' displays different marking (third-person) on the verb, which may naturally be reanalyzed as marking a lack of co-reference or a lack of epistemic authority on the part of the subject (i.e., alterphoric marking).

An epistemic interpretation in particular (as opposed to a purely syntactic one to do with co-reference) would have been plausible for speakers because it would have fit with the way the marker was already being used in most basic clauses, where co-reference is (probably) not in play.<sup>23</sup> The suffix -*m̊* was already being used in first-person clauses, and almost all first-person clauses are statements in which the first-person subject has full epistemic authority over the proposition. Further, the phonologically similar suffix ‑*mu*, which was probably already in the process of reducing to ‑*m̊* in speech, was being used in first-person plural and second-person contexts, where the subject is also generally the epistemic source (bearing in mind that first-person questions and second-person statements are rare compared to first-person statements and second-person questions). Thus,

<sup>23</sup>Though see Hale (1980) for a co-referential analysis of egophoric/alterphoric markers in basic clauses, and San Roque et al. (2018: 51–54) for discussion and criticism.

### Jonathon Lum

once an epistemic reanalysis had been made in reported speech, the generalization to basic clauses would have been a very natural one.

As this reanalysis is based on verbal marking in reported speech, where both the (former) first-person singular marker and first-person plural/second-person marker are realized as [m] (as mentioned earlier), the actual form of the newly reanalyzed egophoric marker to spread to basic clauses would have been ‑*m̊* rather than ‑*mu*, though we cannot rule out the possibility that some speakers were already beginning to reduce ‑*mu* to -*m̊* in basic clauses prior to this point, perhaps on analogy with the existing first-person singular form. The relative order of these changes is of little consequence, however. If ‑*mu* had mostly been replaced by ‑*m̊* in basic clauses already, then the new egophoric reanalysis of ‑*m̊* in reported speech would simply have led to a functional reanalysis of ‑*m̊* in basic clauses. If on the other hand ‑*mu* (or perhaps ‑*m* with apocope of the final vowel) was still in popular use, then as the new egophoric system spread from reported speech ‑*m̊* would have begun to displace ‑*mu* in basic clauses. In either scenario ‑*mu* would begin to fall out of usage, and an egophoric distribution of ‑*m̊* would start to take shape across basic clauses. This is exactly the situation that (northern) Dhivehi is currently in ‒ the former person-agreement suffix ‑*mu* is still known but is now largely restricted to literary contexts, and a mostly egophoric distribution has taken hold in basic clauses.

The actual shift from person marking to egophoricity in basic clauses would have been fairly subtle in that most of the pieces were already in the right places. The 'new' egophoric marker ‑*m̊* was already being used in first-person statements, and the 'new' alterphoric markers (former third-person markers) were already used in third-person sentences. In the emerging epistemic system, however, speakers must have started to use alterphoric/third-person marking in second-person statements (and as described above, would have increasingly come to replace ‑*mu* with the phonologically similar ‑*m̊* in first-person contexts and second-person questions, assuming they had not started this process already). In addition, names, kin terms, and other formally third-person references would have started to trigger egophoric marking in the right contexts as the former person-sensitive system was overridden by an epistemically-sensitive one. However, the marking of first-person questions survives as a relic of the former personagreement system (though as discussed in §3.1, one consultant in Laamu did accept alterphoric marking in some first-person questions). This might partly be related to the fact that the first person was already marked with ‑*m̊* (at least in the singular) before the egophoric system emerged, unlike the second person which was marked with the disappearing ‑*mu*. But it is more likely to be because first-person questions are unusual and pragmatically marked (see §3.1),

### 4 An egophoric analysis of Dhivehi verbal morphology

and because interrogatives in general tend to preserve former person-agreement markers for longer than declaratives do when a new epistemic system emerges (Widmer & Zemp 2017). Eventually, however, first-person questions ‒ or at least, non-rhetorical first-person questions that are not self-directed ‒ might also be predicted to take alterphoric marking instead of first-person/egophoric marking, resulting in a fully-fledged egophoric system. The full chain of development outlined here is represented in Figure 4.2 below (lighter shading indicates diachronic changes in the person-marking system up to at least the middle of the 20ℎ century; darker shading indicates the recent shift from person-marking to egophoricity):

Figure 4.2: Possible development of egophoricity in Dhivehi

Although the transformation of person marking into epistemic marking may seem an unusual diachronic development, it is not without precedent. Widmer

### Jonathon Lum

& Zemp (2017) propose a similar diachronic process in three Tibeto-Burman languages: Sunwar, Dolakha Newar, and Bunan (see also Widmer 2015 for Dolakha Newar and Bunan). According to Widmer & Zemp, these three languages represent different stages of a process by which person agreement transforms into epistemic marking. Sunwar is at an early stage in which there is no evidence of an egophoric/alterphoric opposition in basic clauses, but the language has a semi-direct reported speech construction with a binary opposition between firstand third-person marking in reported clauses. Dolakha Newar is at an intermediate stage in which person markers in semi-direct speech have been reanalyzed as epistemic markers, and are occasionally used in declarative clauses to mark an egophoric opposition. Finally, in Bunan the innovative egophoric system is widely used in both declarative and interrogative clauses, but some remnants of the old person-marking system remain, specifically in interrogative contexts.

The diachronic development described by Widmer & Zemp has some differences to the one outlined for Dhivehi in Figure 4.2 though the basic process is the same. In Widmer & Zemp's account of Bunan, second-person endings gradually become obsolete as first- and third-person markers are reanalyzed as egophoric and alterphoric markers respectively. With the loss of second-person markers among younger speakers of the language, Bunan appears to be moving towards a four-way system that marks egophoric versus alterphoric in both the singular and plural (though number distinctions are also beginning to disappear in the Bunan verbal system ‒ see Widmer 2014: 575–576). In contrast, the available evidence suggests that the person-marking system of (northern) Dhivehi had already simplified (including the complete loss of dedicated second-person markers) prior to its reanalysis as an egophoric system, and the new system does not involve a number distinction at all. Still, both languages developed an egophoric system through a functional reanalysis of person markers in reported speech. In both languages, egophoric markers developed from former first-person markers, and alterphoric markers from former third-person markers. And both languages also show vestiges of their former person-marking systems, especially in interrogative contexts (e.g., first-person questions in Dhivehi) and in the dwindling use of certain agreement markers (e.g., ‑*mu* in Dhivehi) that are outside the egophoric/alterphoric opposition that is now operational. The similarities with Bunan add weight to the diachronic development proposed here for Dhivehi, and suggest that person marking may be a plausible diachronic source of egophoric marking in other egophoric languages too. Data from other egophoric languages for which historical records or descriptions are available may shed further light on the nature of this process.

### 4 An egophoric analysis of Dhivehi verbal morphology

### **5 Conclusions**

In this chapter I have presented a new analysis of verbal marking in finite tenses/aspects/moods in Dhivehi. The analysis is that in the simple present, simple past, perfect and (finite) future tenses/aspects of active, volitional-stem verbs, Dhivehi employs an egophoric/alterphoric opposition that indexes whether or not the subject is the epistemic source for the proposition. While this may be a significant departure from previous accounts of person marking in the language, it appears to be the best explanation for the data presented in §3. In particular, the conclusion that Dhivehi has recently developed an egophoric/alterphoric distinction is able to account for the following:


In addition, this analysis partly explains the differing accounts of second-per son marking in some existing descriptions of the language (Cain & Gair 2000; Fritz 2002), since the egophoric/alterphoric distinction splits the second person by sentence type. However, the analysis presented here stops short of claiming that Dhivehi shows canonical egophoricity, since first-person questions are marked like first-person statements and second-person questions, instead of being marked like second-person statements and third-person statements/questions. Although there are good reasons for thinking that first-person questions might be aberrant because they represent a pragmatically-marked context (see § 3.1), the marking of first-person questions might also suggest that the Dhivehi system is a hybrid of egophoricity and person agreement. The continued (though declining) use of the first-person plural/second-person suffix ‑*mu* is further evidence for a hybrid system, or perhaps more accurately, the (temporary) coexistence of two systems.

Dhivehi's egophoric system is typologically unusual, and may even be unique among Indo-European languages and the languages of the (southern) South Asia

### Jonathon Lum

region. Egophoric marking is found in certain Tibeto-Burman languages as well as some languages of the Caucasus, South America, and New Guinea, though unlike in many of these languages, egophoric marking in Dhivehi is not obviously connected to evidentiality, mirativity, or volitionality. However, the egophoric/alterphoric opposition is restricted to volitional-stem verbs, and so we should not discount a subtler connection of this type, especially given that volitionality is an important grammatical category in Dhivehi.

An interesting question is how egophoricity emerged in Dhivehi and how it relates to person marking. The modern system in the northern dialects may well have developed from an older, six-way person-marking system that simplified somewhat over time, gradually reducing to a binary distinction as speakers began to reanalyze person agreement as egophoric/alterphoric marking. This reanalysis most likely had its genesis in semi-direct speech, the only grammatical construction in the language where 'mixing' of subject identities and person agreement would have been possible.

Further research is required to better understand the historical development of egophoric marking in Dhivehi, its semantics and grammatical behaviour in the contemporary language, and its typological significance. Future studies might look for evidence of egophoricity in the southern dialects (in particular the undocumented Huvadhu dialect) and in certain northern dialects too, such as the dialect spoken on the remote island of Minicoy. A study tracing the development of person/epistemic markers in written Dhivehi would help to pinpoint the timing of the shift, and may also shed further light on how the shift unfolded. Finally, the collection of additional data from unusual grammatical contexts such as firstperson questions and second-person statements would help to check the analysis presented here, and may uncover valuable details concerning the operation of the egophoric system in the contemporary language.

### **Acknowlegements**

I am grateful to Ismail Ibrahim, Ahmed Hassaan, Aniya Rasheed, Nazim Sattar, Ahmed Rasheed, Mohamed Anas Riza and Amalia Gnanadesikan for providing data and for some valuable conversations about many of the points in this chapter. I am also grateful to Alice Gaby, Lila San Roque, Anna Margetts, Manuel Widmer, and Lauren Gawne for their encouragement and for providing comments on earlier versions of the chapter, to Peter Johnson for creating the map, and to two anonymous reviewers for their helpful feedback. Any errors are of course my own, and the various claims and analyses presented do not necessarily reflect the views of the people acknowledged here.

### **Abbreviations**


### **References**


Asher, Ronald E. & T. C. Kumari. 1997. *Malayalam*. New York: Routledge.


### Jonathon Lum


### 4 An egophoric analysis of Dhivehi verbal morphology


### Jonathon Lum


Wijesundera, Stanley, G. D. Wijayawardhana, J. B. Disanayaka, Hassan Ahmed Maniku & Mohamed Luthufee. 1988. *Historical and linguistic survey of Dhivehi: Final report*. Colombo: University of Colombo.

# **Chapter 5**

# **Emerging epistemic marking in Indo-Aryan Palula**

### Henrik Liljegren

University of Stockholm

While evidentiality is neither systematically nor obligatorily signaled in Indo-Aryan Palula [phl; phal1254] (Pakistan), it can be observed in so-called scattered coding. It is most obviously reflected in three sub-systems of the language: a) as a secondary effect of tense—aspect differentiation, mostly clearly seen in the use of the perfect for indirect evidence vis-à-vis the use of the simple past for direct evidence; b) by a set of utterance-final mood markers, involving an emerging threeway paradigmatic contrast: *thaní* as quotative, *maní* as hearsay and *ɡa* as inferred knowledge; and c) by (at least) one member of a set of second-position discourse particles, *xu*, marking surprise. Although evidentiality contrasts akin to the perfect vs. simple past were indeed part of the ancestral Indo-Aryan tense system, there are plenty of parallels in adjacent languages to the epistemic contrasts noted for Palula, suggesting that more recent language contact must have contributed to, or largely facilitated, the emergence of epistemic marking in the language.

### **1 Introduction**

While evidentiality, mirativity and related notions have been discussed at length for Sino-Tibetan (DeLancey 1986; 2001) and for Turkic languages (Johanson 2000; 2003), relatively little attention has been given to similar phenomena in the multilingual Hindukush-Karakoram (Afghanistan, Pakistan, Kashmir), a mountainous region home to approximately 50 distinct – mostly Indo-Iranian – languages, lying at the crossroads of South Asia and the Tibetan and the Turkic worlds. Only some preliminary suggestions regarding evidentiality and its origin in the region have been offered, and a few language-specific studies focussing on epistemic

Henrik Liljegren. 2020. Emerging epistemic marking in Indo-Aryan Palula. In Henrik Bergqvist & Seppo Kittilä (eds.), *Evidentiality, egophoricity, and engagement*, 141–163. Berlin: Language Science Press. DOI: 10.5281/zenodo.3975803

### Henrik Liljegren

aspects have been carried out (see §2). While in a few languages, contrasts in the realm of evidentiality are actually part of verbal morphology, in others it is mainly indicated with particles, or else such distinctions are present in "scattered coding", i.e. not as part of a single sub-system but instead encoded in various parts of the grammar. In the present study, texts belonging to a corpus resulting from recent fieldwork on Indo-Aryan Palula have been analysed, and some preliminary conclusions as well as open-ended questions are offered regarding various types of epistemic marking found in the language, their possible semantic scope and their relationship to other language-particular grammatical distinctions. It has been found that epistemic marking is entailed in some tense-aspect distinctions (§3); in the use of mood-markers (§4); and in the use of discourse particles (§5). In as much detail as possible, derivational paths will be discussed (see §6), and also language-contact effects, e.g. the possible influence of grammaticalized inferentiality in neighbouring languages or in languages of wider communication on Palula. The findings of the study are summarized in §7.

### **2 Background**

Palula [iso 639-3: phl; glottocode: phal1254] is an Indo-Aryan language belonging to the Shina group. It is spoken by approximately 10,000 people in the southern part of Chitral district in northern Pakistan (35.38, 71.78; see Figure 5.1). Speakers are to a varying degree bilingual in Khowar, another Indo-Aryan language widely spoken in Chitral, and/or in Pashto, an Iranian language that is one of the most important lingua francas of northwestern Pakistan. Educated speakers also know Urdu, the nation-wide lingua franca of Pakistan, and to a lesser extent English. The author conducted linguistic fieldwork in this language, primarily in the period 1998 to 2006, with the compilation of a corpus of Palula narratives and other texts as one of its aims.<sup>1</sup> In the present study, that corpus has been consulted, along with various field notes and other types of language data, e.g. obtained experimentally or by means of direct elicitation.

The district as well as the surrounding region where Palula is spoken is linguistically highly diverse and multilingual. The mountainous north of Pakistan counts nearly 30 distinct ethnolinguistic communities, and another 20 or so can be added if we also include the adjacent, and equally diverse areas of northeastern Afghanistan and Indian-administered Kashmir. Those languages represent

<sup>1</sup>The interested reader is encouraged to consult the following works dedicated to the description and documentation of Palula: phonetics and phonology (Liljegren & Haider 2009); morphology and syntax (Liljegren 2010; 2016); vocabulary and semantics (Liljegren & Haider 2011; 2015a); glossed and translated texts (Liljegren & Haider 2015b).

### 5 Emerging epistemic marking in Indo-Aryan Palula

six genera: Indo-Aryan<sup>2</sup> (to which Palula and above-mentioned Khowar belong), the dominant one as far as the number of languages is concerned; Iranian (apart from Pashto and Dari, the Afghan form of Persian, these are relatively small language communities in remote areas); Nuristani (concentrated in an area of Afghanistan just across the border from Chitral); Sino-Tibetan (represented by Balti, spoken in the eastern-most part of this region); Turkic (mainly in the borderlands between Afghanistan and the former Soviet republics of Central Asia); and the language isolate Burushaski.

Figure 5.1: Chitral district and the surrounding region. Language names (only those that feature in the article) in italics.

There are only a few studies discussing evidentiality in this region, if we do not consider the observations made regarding the general pervasiveness of it in the grammars of Turkic and Sino-Tibetan, two of the genera represented here – although in a rather peripheral way, as we saw above. In a master's thesis, Jones (2009) analyses evidentiality and mirativity in Balti, and concludes that the language has a grammatical category of mirativity, reflected by present and past mirative markers, as well as a reportative verb suffix and some newly developed strategies for marking inference and supposition.

Although evidentiality seems a less pervasive or significant feature in Iranian languages in general than is the case in Turkic and Sino-Tibetan, there are nevertheless indications that certain verb forms in the Persian varieties spoken in or

<sup>2</sup> Indo-Aryan, Iranian and Nuristani are usually regarded as subgroups on the same level, subsumed under the Indo-Iranian branch of Indo-European.

### Henrik Liljegren

near this region have epistemic uses. While the scope of epistemic uses is limited to past-time reference in Persian of Iran, Perry (2000) presents evidence of much wider uses in Dari, i.e. the variety spoken in Afghanistan, encompassing quotative, inference, presumptive and speculative functions, in the present and future as well as in the past. Bashir also notes the use of a so-called "distant perfect" in Iranian Wakhi as a marker of inferentiality or mirativity (2006; 2007b: 840) and a second-position clitic in Pashto carrying out certain evidentiality-related functions (2006).

Evidentiality distinctions in Indo-Aryan are in general not particularly significant or easily identifiable (Masica 1991: 279–291; de de Haan 2013), but there are individual or areally significant exceptions, a fact noted by Bashir (2006). Khowar, the Indo-Aryan language mentioned above, and its closest relative Kalasha, are particularly interesting in this regard, as they may be part of an areal configuration, also including non-Indo-Aryan languages in the Hindukush-Karakoram region, where the semantic parameter of evidentiality to a varying extent has been grammaticalized (Bashir 1988; 1996; 2013; 2003: 823; 2010). In another degree work, Lubberger (2014) analyses a set of metarepresentation markers in Indus Kohistani — an Indo-Aryan language spoken in the central parts of the Hindukush-Karakoram region — of which at least two have a definite bearing on evidentiality.

Based on previously, but severely limited, published research, Bashir also presents evidence for what she refers to as "robust inferential/indirective systems" (2006) in several of the lesser-described Nuristani languages, as well as a special past-tense form only found in one of the dialects of Burushaski, which imparts an inferential-mirative meaning (2010: 14).

### **3 Evidentiality and tense–aspect differentiation**

The first evidential sub-system in Palula to be discussed, is that of verbal categories. Seven main TMA categories can be identified in the language (Liljegren 2016: 247–263): four (Future, Present, Simple Past, Imperative) that are purely morphological categories, and another three that are periphrastically formed by the addition of auxiliaries (Past Imperfective, Perfect, Pluperfect). Palula is, like most other languages of the surrounding region, verb-final, and most frequently SOV. Verb morphology is suffixal, whether related to tense-aspect or agreement marking, and auxiliaries occur subsequent to the main verb.

Evidentiality is not a primary category in this system. Instead, the use of Simple Past vis-à-vis Perfect in Palula narratives often entails a contrast between

### 5 Emerging epistemic marking in Indo-Aryan Palula

direct and indirect evidence with events in the past. This is for instance reflected in the choice of verbal category when applying Dahl's (1985) TMA questionnaire. The event described in ( 1) is the speaker's eye-witness account, whereas in (2), it is what the speaker's brother experienced and told the speaker prior to the moment of speaking that is being described. In the first case, the Simple Past (morphologically expressed with the perfective and ergative person/number agreement) is used. In the second case, the Perfect category (morphologically expressed with the perfective and the present tense form of a 'be'-auxiliary, both agreeing in person/number with the O or S) is instead applied.

(1) Palula – Simple Past

*tíi* 3SG.REM.OBL *áa* INDF *báaṭ* stone *uc̣h-i* lift-CVB *ba* TOP *ǰhandra-í* snake-OBL *the* to *uṛíit-u* let.go.PFV-MSG *so* 3.SG.REM.NOM *múṛ-u* die.PFV-MSG [This happened to my brother, I saw it:]'He took a stone and threw it at the snake. It died.' (PHL-TMAQ-NH:174–175)

(2) Palula – Perfect

*tíi* 3SG.REM.OBL *áa* INDF *báaṭ* stone *uc̣h-í* lift-CVB *ba* TOP *ǰhandra-í* snake-OBL *the* to *uṛíit-u* let.go.PFV-MSG *hín-u* be.PRS-MSG *ta* SUB *so* 3SG.REM.NOM *múṛ-u* die.PFV-MSG *hín-u* be.PRS-MSG [This happened to my brother, and he told me:]'He took a stone and

threw it at the snake. It died.' (PHL-TMAQ-NH:179–180)

The example in (3) is from an interview, where the narrator tells how, a long time ago, he went to his father-in-law who told him a story. Here, the tense is the Simple Past. This can be compared with the corresponding Perfect forms used in the lines of (4), belonging to a story about a boy named Katamosh who was told by his mother to go to his grandmother up in the high pastures. The latter is obviously part of a non-witnessed event with numerous components of fiction, such as animals acting and talking.

(3) Palula – Simple Past

*áa* one *deés* day *táa* there.REM *ɡúum* go.PFV.MSG *ta,* SUB *máathe* 1SG.DAT *qisá* story *th-íil-u* do-PFV-MSG 'One day I went there, and [he] told me a story.' (PHL-Hunter:009)

### Henrik Liljegren

(4) Palula – Perfect

*tasíi* 3SG.GEN *yéei* mother *taste* 3SG.DAT *áak* INDF *ṭíki* bread.cake *th-íil-i* do-PFV-F *hín-i* be.PRS-F *kaṭamúš* Katamosh *ṭíki* bread.cake *ḍóok-a* back-OBL *pharé* along *ɡhaṇḍ-í* tie-CVB *sóon-a* pasture-OBL *dúši* toward *ɡúum* go.PFV.MSG *hín-u* be.PRS-MSG

'His mother made a cake of bread for him. Katamosh tied the bread to his back and set out to the high pastures.' (PHL-Katamosh:009-010)

However, it should be noted that this "extended" use of the Perfect is only a tendency, far from any obligatory marking of indirect evidence. It also seems that some authors or narrators are more prone than others to use it. In another, equally fantastic, story, an unnamed person in a distant past goes hunting, and while sitting down to eat the cooked meat of a markhor, a *ṭhaaṭáaku*, a hairy and frightening creature suddenly appears. Here, however, as can be seen in (5), the line of the narrative uses the Simple Past.

(5) Palula – Simple Past

*anɡóor* fire *ǰeel-í* light-CVB *táa* there.REM *pačaá* cook.CVB *kh-ainií* eat-INF *široó* starting *th-íil-u* do-PFV-MSG *široó* starting *th-íil-u* do-PFV-MSG *ta* SUB *tíi* 3SG.REM.OBL *maǰí* in *áa* INDF *ǰhaṭíl-u* hairy-MSG *ṭhaaṭáaku* ogre *yh-óol-u* come-PFV-MSG 'When he had made a fire, he cooked the meat and started eating. Meanwhile, a *ṭhaaṭáaku* suddenly appeared.' (PHL-Thaataaku:004-005)

An even less predictable contrast is that between Present and Future, as in (6) and (7), respectively. Both can be used with future-time reference. However, the choice in this case does not necessarily have a bearing on evidentiality per se.

(6) Palula – Future

*ma* 1SG.NOM *nis* 3SG.PROX.ACC *aáǰ* today *kh-úum* eat-1SG *ta* SUB *rhootašíi-a* morning-OBL *ba* TOP *kanáa* what *bh-úum* become-1SG 'If I eat it today, what should I then do tomorrow?' (PHL-HunterMonkey:005)

### 5 Emerging epistemic marking in Indo-Aryan Palula

(7) Palula – Present

*uth-í* stand.up-CVB *maníit-u* say.PFV-MSG *hín-u* be.PRS-MSG *ki* COMP *eé* o *lhéṇḍu* bald(M) *ɡóoi* where.from *yh-óol-u* come-PFV-MSG *ma* 1SG.NOM *tu* 2SG.NOM *kha-áan-u,* eat-PRS-MSG *muṣṭú* first *ma* 1SG.NOM *thíi* 2SG.GEN *ráat* blood *pil-áan-u* drink-PRS-MSG *théeba* then *ma* 1SG.NOM *thíi* 2SG.GEN *lhéṇḍ-i* bald-F *kakaríi* scalp *čap-áan-u* gnaw-PRS-MSG 'I will eat you. First I will drink your blood, and then I will gnaw on your bald scalp.' (PHL-Katamosh:030–032)

When on the other hand two present-time referring utterances are being contrasted as in (8) and (9), there is a clearer correspondence between the use of Present and direct evidence, on the one hand, and the use of Future and indirect evidence, on the other.


### **4 Evidentiality and utterance-final mood markers**

Palula has a closed set of markers that one way or another specify the relationship between an utterance as a whole and the speaker and/or hearer, e.g. signaling a polar question or a request. Almost exclusively, such mood markers occur utterance-finally, i.e. in most cases following the finite verb. At least three of those markers have functions related to, or partly related to, evidentiality.

The most frequent (in my corpus) of the three, *thaní*, is a quotative (Liljegren 2016: 267, 377–387). In Palula narratives, it usually marks — or closes — directly quoted speech, as in (10).

### Henrik Liljegren

(10) Palula – quotative

*so* 3MSG.NOM *ba* TOP *maidóon-a=wée* field-OBL=into *be* go.CVB *ba* TOP *áa* one *khur* foot *raál* High *the* do.CVB *ba* TOP *kar* around *kar* around *kar* around *gír-a* turn-3SG *koó* who *hín-ee* be.PRS-MPL.Q *yh-óoi* come-IMP.PL *thaní* qUOT 'He was spinning around and around in the field, holding up one leg, [and he was calling out, *saying*,]"Is there anyone here [brave enough]? Come on!"' (PHL-JangibazKhan:037–038)

However, it may occasionally extend into (non-uttered) reported thought, as shown in (11), here used along with a preposed *ki*-complementizer.

(11) Palula – quotative

*ɡhrast-á* wolf-OBL *karáaṛ-a* leopard-OBL *asíi* 1PL.GEN *xiaál* opinion *ki* COMP *ɡóo* maybe *mheer-íl-u* kill-PFV-MSG *heentá* CONDL *kh-óol-u* eat-PFV-MSG *thaní* qUOT 'We thought that perhaps a leopard or a wolf had killed and eaten him.' (PHL-GhaziSamad:011)

If an explicit PCU (perception-cognition-utterance) predicate precedes the complement, the use of the quotative is more variable. In (12), where the utterance predicate *ṭeekílu* 'called (out)' is used, there is no closing *thaní*, while in (13), the quotative *thaní* is co-occurring with the preceding predicate of knowledge acquisition, *búda hína* '(have) understood'.

(12) Palula – utterance without quotative

*áak* INDF *šúma* parrot.OBL *ṭeek-íl-u* call-PFV-MSG *ée* o *kúṛi* woman *thíi* 2SG.GEN *míiš-i* man-GEN *paaṇṭí* clothes *ṣ-éel-i* dress-PFV-F *hín-i* be.PRS-F *tu* 2SG.NOM *míiš* man *ba* TOP *na* NEG 'A parrot called out: "O woman, you have dressed like a man, but you are not a man.' (PHL-WiseMinister:012-013)

(13) Palula – utterance with PCU predicate and quotative *búd-a* understand.PFV-MPL *hín-a* be.PRS-MPL *ki* COMP *phaí* girl *wíi-a* water-OBL *ɡíi* go.PFV.FSG *thaní* qUOT 'They understood that the girl must have thrown herself into the water.' (PHL-ShepherdBoy:060)

### 5 Emerging epistemic marking in Indo-Aryan Palula

Another, highly grammaticalized, function of *thaní* is when it occurs postposed to a proper noun and carries the meaning 'called, thus named', referring to the immediately preceding noun. An example is provided in (14).

(14) Palula – 'name' *miír* Mir *thaní* qUOT *áak* INDF *míiš* man *heens-íl-u* exist-PFV-MSG *de* PST 'There was a man *called* Mir.' (PHL-GhaziSamad:051)

Hearsay can be (but is not necessarily) marked with an utterance-final *maní*. The reported content, preceding it, is in such cases often mythical or unexpected, as in (15).

(15) Palula – hearsay

*dac̣h-áan-u* look-PRS-MSG *ta* SUB *eeteeṇ-ú=ee* such-MSG=CONJ *áak* INDF *šay* thing *yh-óol-u* come-PFV-MSG *maní* HSAY *maaxustán* evening *de* be.PST *maní* HSAY *áa* INDF *šay* thing *yh-óol-u* come-PFV-MSG *babár* furry.thing *búd-u* understand.PFV-MSG *ki* or *na* NEG 'Then a creature appeared, it is said, in the evening, it is said, a furry one,

you know, don't you.' (PHL-AyanMir1:065)

However, the use of *maní* is not restricted to narrative discourse. It is also used in everyday conversation, as in (16), an excerpt from an online chat conversation. It should be noted that it is only the first of the two clauses that is hearsay-marked.

(16) Palula – hearsay

*asíi* 1PL.GEN *atshareet-á* Ashret-OBL *bíiḍ-u* much-MSG *kir* snow *dít-u* fall.PFV-MSG *hín-u* be.PRS-MSG *maní* HSAY *hiimeel-í* glacier-PL *bi* SEP *whéet-im* come.down.PFV-FPL *hín-i* be.PRS-F '[I have been told that] a lot of snow has fallen in our [village] Ashret, and there have been avalanches as well.' (PHL-CHN070320)

Finally, a third, and in the present corpus much more infrequently occurring marker, *ɡa*, signals inferred, presumed or assumed knowledge, as in (17).

### Henrik Liljegren

(17) Palula – assumption *anú* PROX.NOM.MSG *dhút-a* mouth-OBL *de* give.CVB *baaǰá* harmonium *bhanǰa-áan-a* beat-PRS-MPL *eetáai* from.there *aawaáz* sound *yh-áand-u* come-PRS-MSG *eh* oh *rueeleé* government.official.PL *hín-a* be.PRS-MPL *ɡa* ASS *rueeleé* government.official.PL *ba* TOP *aní* PROX *sarkaarí* official *sipaahi-aán* soldier-PL *hóons-an* exist-3PL *de* PST 'I heard a sound like a harmonium being played. "Oh," I thought, "These must be government officials, such professional soldiers they are."' (PHL-Hunter:067-068)

### **5 Evidentiality and second-position discourse particles**

Palula discourse markers are second-position clitics that specify the discourse role of a preceding unit (often a noun phrase) in relation to adjacent units. A secondary effect of some discourse markers is that they indicate how larger units (e.g. clauses) are interrelated, especially when used in pairs, or when the same marker is used repeatedly in two adjacent clauses. The latter use overlaps with the function of conjunctions.

The particle *xu* is such a second-position clitic. It signals surprise, as in (18), or emphasis, occurring postposed to the first-position constituent that is thus being focused. While it is often difficult to find good translation equivalents in English, it is strikingly similar in meaning to the Swedish modal particles *ju* or *visst*.


Other members of the set of second-position discourse particles are: *bi* (separation marker, see 16 for an example), *ba* (switch-topic marker), *ta* (contrast marker), *ee* (amplification marker). The extremely frequently occurring switchtopic marker *ba*, is for instance used to make a contrast with an immediately preceding subject explicit, as in (19), but has a number of other functions, some of them challenging to define exactly (Liljegren 2016: 419–425). (See 6 for another example of the contrastive function.)

5 Emerging epistemic marking in Indo-Aryan Palula

(19) Palula – contrast *míi* 1SG.GEN *ɡhoóṣṭ* house *lookúṛi* Lokuri [place] *hín-u,* be.PRS-MSG *iskuúl* school *ba* TOP *asíi* 1PL.GEN *kaṇeeɡhaá* Kanegha [place] *hín-i* be.PRS-F 'My house is in Lokuri, while our school is in Kanegha.' (PHL-Our school:004)

### **6 Origins and grammaticalization paths**

As for the contrastive use of tense-aspect categories to signal indirect vs. direct evidence, there is evidence of a similar-functioning differentiation in Old Indo-Aryan, i.e. in the ancestral language (or a closely related language to that) of Palula and other modern-day Indo-Aryan languages. In the system described by the Indian grammarian Pāṇini, there were three categories with past-time reference: Aorist, Imperfect and Perfect. Of these three, the Perfect was used with special reference to reported, less recent, events, that is excluding the speaker's direct witnessing the event reported, while the Imperfective had the same time reference as the Perfect but implied that the speaker was indeed a direct witness. The Aorist, which functioned as a more general past tense, was the only one of the three that could refer to recent events (Cardona 2002: 235). Deshpande (1981: 62) summarizes, in the same vein, this three-way contrast as: a) [+past +recent +/-seen] (Aorist), b) [+past -recent +seen] (Imperfective), and c) [+past -recent -seen] (Perfect), thus making the presence or absence of a [seen] feature the minimal contrast between the latter two, a distinction that Bashir (2006) argues has been passed on to some of the descendant languages, the two Chitral languages Kalasha and Khowar in particular.

In Khowar, which is locally influential in the district where Palula is spoken, and also is the second language of many Palula speakers, the distinction between indirect evidence [+past -seen] and direct evidence [+past +seen] is upheld in the tense-aspect system (which has retained forms that were lost in most other Indo-Aryan languages), as can be seen in comparing (20) with (21). Here, the [-seen] value of the Past Inferential is further specified with a *birai* (a past participle of 'become' whose epistemic function has developed later) which adds a mirative meaning to the reported event. The encoding of such evidentiality-related differentiation, is non-optional in Khowar (Bashir 2007a: 221–222).

### Henrik Liljegren

(20) Khowar – Past Actual *hase* 3SG.REM.NOM *boht-o* stone-OBL *ɡan-i* take-CVB *ayi-o* snake-OBL *ṭek-o* TOP-OBL *lak-it-ai* let.go-PST-3SG *hase* 3SG.REM.NOM *o-br-it-ai* PST-die-PST-3SG [This happened to my brother, I saw it:] 'He took a stone and threw it at the snake. It died.' (KHW-TMAQ-AA:174-175) (21) Khowar – Past Perfective Inferential *hase* 3SG.REM.NOM *boht-o* stone-OBL *ɡan-i* take-CVB *ayi-o* snake-OBL *ṭek-o* TOP-OBL *lak-iru* let.go-PPTC *bir-ai* become.PST.INFER-3SG *hase* 3SG.REM.NOM *birdu* die.PPTC *bir-ai* become. PST.INFER-3SG

[This happened to my brother, and he told me:] 'He took a stone and threw it at the snake. It died.' (KHW-TMAQ-AA:179-180)

Very similar distinctions are being made in the Kalasha tense-aspect system. For instance, two distinct past-time referring forms of 'were' are used. In (22), the past actual *asini* is used as the speaker points to the domestic animals as witnessed entities in the real world. In (23), the past inferential *asta* instead is used, as a means for the narrator to portray the participants in the story as the creations of fiction.

(22) Kalasha – Past Actual

*tara* there.REM.SPC *as-ini* be.PST.ACT-3PL *ɡak* cow *tara* there.REM.SPC *ɡordok* donkey *hãš* horse 'There were cows, donkeys and horses there.' (Heegård Petersen 2015: 136)

(23) Kalasha – Past Perfective Inferential *ek* one *ɬawak* fox *ek* one *šara* markhor *malɡiri* friend *asta* be.PST.INFER.3 'Once a fox and a markhor were friends.' (Heegård Petersen 2015: 182)

Kalasha is not a common second language of Palula speakers. However, considerable interaction apparently took place between the communities in the past, and there is reason to believe that Kalasha, at least in part of what is now Palulaspeaking territory, is exerting substratal influence, as conversions from the traditional Kalasha religion often resulted in a gradual language shift from Kalasha to one of the surrounding languages spoken by a Muslim majority population,

### 5 Emerging epistemic marking in Indo-Aryan Palula

Palula in those days being one of the main candidates (Cacopardo & Cacopardo 2001: 117–118; Decker 1992: 55–60).

Although the distinction in Palula also could have been inherited, it is equally probable to have been areally influenced and/or reinforced. Apart from evidentiality-related distinctions in the verbal systems of Khowar and Kalasha, a few other languages in the immediate region reflect similar contrasts in their TMA systems. There are for instance epistemic verb forms in regionally influential Persian varieties (Perry 2000; Windfuhr & Perry 2009: 461). Bashir (2010: 14–15; 2007b: 839) also mentions the perfect in the Iranian language Wakhi as well as in Tajik Persian as specifically correlated to indirect evidence. Another possible parallel is the contrastive use of a proximate vs. a distal perfect in Pashai, another Indo-Aryan language (or perhaps more correctly, group of related language varieties), spoken across the border in northeastern Afghanistan (Lehr 2014: 295– 297). Perhaps this, in turn, is part of a considerably larger areal configuration in Western and Central Asia, something that Dahl (1985: 152) is hinting at, when he describes the extension of perfects into the realm of quotatives as an areal phenomenon with an approximate geographical correlation with the former Ottoman Empire (including e.g. Kurdish and Turkish), but it goes without saying that the secondary use of perfects for such functions has indeed been verified cross-linguistically much farther afield (Aikhenvald 2004: 112).

As for the mood markers described in Section 4, they are to a varying extent grammaticalized in Palula, which is reflected in the textual occurrence of these markers (Table 5.1). It should be noted, however, that identical forms may occur in other uses in the material.


Table 5.1: Text occurrences of *thaní*, *maní* and *ɡa*, and of forms related to them. (In a text corpus consisting of 76 transcribed and annotated Palula texts, mainly narrative.)

The quotative marker *thaní* is the most frequent of the three. The relatively frequent occurrence of the same form but in other uses is largely accounted for by its post-nominal "naming" function, as shown in example (14) above. It is highly grammaticalized as a quotative, but its co-occurrence with a preposed (and "borrowed") *ki*-complementizer, on the one hand, and the alternative construction

### Henrik Liljegren

with *ki* altogether lacking an quote-final *thaní*, as in (24), on the other hand, points to an ongoing competition with the "new" Persian-derived *ki*-strategy (further reinforced by the corresponding construction in Urdu), in which the *thaní*-less construction most likely is winning out in the long run: …*thaní* (as in 10) > *ki*...*thaní* (as in 11) > *ki*... A similar development has been observed in neighbouring Kalasha with regard to the indigenous utterance-final *ɡhõi* and the "imported" utterance-initial *ki* (Bashir 1988: 266–324).

(24) Palula – quotative

*dun-áaṭ-u* think-AG-MSG *bh-íl-u* become-PFV-MSG *hín-u* be.PRS-MSG *ki* COMP *aní* 3FSG.PROX.NOM *ba* TOP *kateeeṇ-í* what.kind-F *ǰuánd* life 'He started thinking, "What a life!"'(PHL-Katamosh:057)

Formally, *thaní* is a converb form (or conjunctive participle) of the verb *thané*- 'call, say'. In contemporary Palula, few other forms of that verb are in fact in use, once more confirming the level of grammaticalization that this converb has reached. Bashir (1996) found in a study of SAY-quotatives, and similarly derived markers, that they are present in many of the region's languages (and beyond), and argues for areal convergence. Examples of such markers include: Indus Kohistani (Indo-Aryan) *karee* (Lubberger 2014: 67–69); Kalasha (Indo-Aryan) *ɡhõi* (Heegård Petersen 2015: 67); Khowar (Indo-Aryan) *reé* (Bashir 1996: 225– 235); Gawri (Indo-Aryan) *är(o)* (Baart 1999: 147–149); Dameli (Indo-Aryan) *ɡani* (Perder 2013: 176–177); Gilgiti Shina (Indo-Aryan) *theé* (Radloff & Shakil 1998: 28); Balti (Sino-Tibetan) *zer*/*zere* (Bashir 1996: 270; Jones 2009: 64); and possibly Burushaski (isolate) *nusé(n)* (Bashir 1996: 262), although a pre-posed *ke* is the preferred marker of direct speech (Berger 1998: 193). While many of them are indeed grammaticalized forms of a verb with the meaning 'say', a number of them are instead ultimately related to a 'do'-verb; that is likely the case with Palula *thaní* (cf. *the*- 'do'), the corresponding markers in other Shina varieties, as well as Indus Kohistani *karee* (< *kar*- 'do').

Dameli is another geographically close neighbour of Palula, spoken in the next valley to the south. Here, too, the quotative, which is derived from a converb of 'say', is extended to predicates of cognition, as in (25), and additionally is postpositioned to a noun phrase with the meaning 'called, thus named', as in (26), just like in Palula.

5 Emerging epistemic marking in Indo-Aryan Palula


The Palula hearsay marker *maní*, is of lower frequency in the text corpus, and is possibly grammaticalized to a lesser extent than *thaní*. When it is used it is often in order to emphasize the non-witnessed, and often questionable (in the mind of the speaker), character of the particular event or situation thus marked, rather than to signal just any instance of reported speech or hearsay. Note that *maní* in (27) only occurs after the second finite verb of this utterance, not after the first.

(27) Palula – hearsay

*eesé* REM *baačaá-ii* king-GEN *bóoš* twelve *zára* thousand.PL *kuṛíina* women.PL *heens-íl-im* live-PFV-FPL *de* PST *tasíi* 3SG.GEN *áaṣṭ* eight *zára* thousand.PL *kuṇaak-á* child-PL *heens-íl-a* live-PFV-MPL *de* PST *maní* HSAY 'This king had twelve thousand wives, and is said to have had eight thousand children.' (PHL-AboutAKing:007-008)

The marker *maní* is like *thaní* derived from a converb, in this case of the verb *mané*- 'speak, recite, say'. In the corpus, quite a number of other uses of this verb occur, including a few instances of it as a non-grammaticalized converb, entirely homophonous with the hearsay marker. An example of the latter is seen in (28), where *maní* heads a subordinate clause with a subsequence reading, thus repeating the preceding lexical finite verb 'said' in its converb form 'having said', simply corresponding to 'then' in English.

(28) Palula – maní as converb


### Henrik Liljegren

*wháat-u* come.down.PFV-MSG *heentá* CONDL *ma* 1SG.NOM *tu* 2SG.NOM *kh-úum* eat-1SG *eendáa* like.that *man-í* say-CVB *ba* TOP *iṇc̣* bear *áak* INDF *keeṇ-í* cave-OBL *šíiṭi* inside *the* to *ɡúum* go.PFV.MSG *hín-u* be.PRS-MSG 'The bear said: "Good, if you come down fattened I will eat you." Then [lit. Having said that,] the bear went into a cave.' (PHL-Katamosh:025-026)

Reportative markers with similar functions (and in many cases with a similar history) have been noted for other languages in the region, e.g. Indus Kohistani *lee* (Lubberger 2014: 22–23); Kati (Nuristani) *mem* (Strand 2016); Waigali (Nuristani) -*le* (Degener 1998: 173–182); and Balti/Purik (Sino-Tibetan) -*lo* (Jones 2009: 57–62; Zemp 2013: 776–792).

In the example from Indus Kohistani in (29), the speaker quotes one of her sons speaking to her at the time when he had come home after the big earthquake in 2005. The reportative *lee* cliticizes to the finite verb in each of the two clauses.

(29) Indus Kohistani – reportative

*iskul-ãĩ* school-GEN.F *kùṛ-muṛ* wall.FPL-ECHO *bazíthe=lee* go.PRS.PFV.MPL=HSAY *hãã* and *maasmá* child.PL *búṭ* all *báč* saved *hu-úthe=lee* become-PRS.PFV.MPL=HSAY 'The walls of the school and stuff went down, but the children all escaped.' (Lubberger 2014: 26–27)

There are only a few, and in some cases rather dubious, instances of *ɡa* as an utterance-final marker of inferentiality in the Palula corpus. It is most likely related to the homophonous indefinite-interrogative pronoun *ɡa* 'what, any, what kind of, any kind of'. Probably, it was first used in the utterance-final position as a tag. The assumption here is that it is even less grammaticalized or established than *maní* and remains somewhat varying in its application: as a request for confirmation, marking suspense or surprise, etc. The particle *bo* in Kohistani Shina, see example in (30), has similar semantics and distribution (Schmidt & Kohistani 2008: 204).

(30) Kohistani Shina – presumption

*mi* my *qǽæs* guess *ǰo* some *kudí* where *múṭho* tree *khári* under *níiz-iǰ-aan-o* sleep-ABSP-is-3MSG *bo* ASS 'I guess he must be asleep under a tree somewhere.' (Schmidt & Kohistani 2008: 204)

### 5 Emerging epistemic marking in Indo-Aryan Palula

Careful elicitation (in 31), in which imaginary situations were described, confirms the emerging paradigmatic contrast ZERO vs. *maní* vs. *ɡa* as corresponding to a) an event directly observed by the speaker, b) an event heard about but not seen by the speaker c) an inference/assumption on the part of the speaker.

	- a. *kir* snow *dít-u* put.PFV-MSG *hín-u* be.PRS-MSG 'It has snowed.' [directly observed] (PHL-20157027-elic:001)
	- b. *kir* snow *dít-u* put.PFV-MSG *hín-u* be.PRS-MSG *maní* HSAY 'It has snowed.' [not seen but heard from sb else] (PHL-20157027-elic:002)
	- c. *kir* snow *dít-u* put.PFV-MSG *hín-u* be.PRS-MSG *ɡa* INFER 'It has snowed.' [not directly observed but inferred from other evidence, e.g. snow on somebody else's boots] (PHL-20157027-elic:003)

Interestingly, the same elicitation task resulted in a two-way differentiation in neighbouring Khowar (32).

(32) Khowar minimal contrastː direct (a) vs. hearsay (b)/assumption (c)


### Henrik Liljegren

The discourse particle *xu* is most likely a loan from Pashto. In Pashto, *xo* is used as an emphatic particle (a second position clitic), with the approximate meaning 'in fact'. In Pashto as well as in Palula *xo*/*xu* is homophonous, or alternatively polysemous, with a clause-initial adversative conjunction 'but'. The examples in (33) and (34) illustrate the strikingly similar uses and clause positions of the particle in the two languages.

(33) Palula – use of the particle *xu*

*lo* 3MSG.DIST.NOM *tu* 2SG.NOM *keé* why *kh-óo* eat-3SG *lo* 3MSG.DIST.NOM *xu* EMPH *thíi* 2SG.GEN *bhróo* brother *atsharíit-u* Ashreti-MSG *thíi* 2SG.GEN *qóom* tribe 'Why would he eat you? He is your Ashreti brother and of your own tribe.' (PHL-GhaziSamad:019-020)

(34) Pashto – use of the particle *xo*

*dā* this.DIR *xo* EMPH *zmā* 1SG.GEN *wror* brother *day* be.PRS.3MSG *də* of *bel* other *saṛ-i* man-OBL *wror* brother *na* NEG

'He is in fact my brother, not some other man's brother.' (David 2013: 375)

A parallel pattern of use has also been reported for *xu* in Dameli, another neighbouring Indo-Aryan language (Perder 2013: 168), and similar sets of particles in many languages in the region (especially to the west) have been described, for instance in Nuristani Waigali by Degener (1998: 166–188) and Indo-Aryan Gawri (Baart 1999: 159–166). Degener describes one of the particles in Waigali, *be*, as for instance signalling empathy or some pre-understanding for a particular situation that has arisen. In the example in (35), there is also a component of reproach, again difficult to find a good equivalent for in English but somewhat easier in German (which like Swedish has its own set of modal particles); *be* frequently corresponds to German *doch*, but in individual examples Degener also offers the translations *bloß*, *aber*, *bestimmt*, *also*, *nun*. Some of those uses seem to overlap with Palula *xu*, others with *ba*.

(35) Waigali – use of the particle *be yi* this *manaṣa* man *be* EMPT *ämeba* 1PL.GEN *pũt* way *süräy* blocked 'This man has apparently blocked our way. [Dieser Mann hat *doch* unseren Weg versperrt!]' (Degener 1998: 167)<sup>3</sup>

<sup>3</sup>No morpheme glossing is provided in the source, only a free translation of the sentence as a whole. The glossing and the English translation are my own.

### 5 Emerging epistemic marking in Indo-Aryan Palula

There is evidence for a relatively high degree of "borrowability" across languages (Svärd 2014; Liljegren & Svärd 2017; Perder 2013: 183–184) with these types of markers, pointing e.g. to a possible diachronic link between Waigali *be* and Dameli/Palula *ba*. Probably, the use of such modal or discourse particles is a relatively new strategy of signalling evidentiality and inferentiality as far as Palula is concerned.

### **7 Summary**

Evidential marking and evidentiality-related distinctions in Palula are, like in many Indo-Aryan languages, observable in so-called scattered coding. It is not forming an independent system, and it is often optional or comes about as a secondary effect of the coding of other main functions. It is most obviously reflected in three separate parts of the language system, namely a) as a component, or byproduct, of tense—aspect differentiation, b) by a subset of utterance-final mood markers, and c) by (at least) one member of a set of second-position discourse particles. In the case of tense—aspect, the choice between a simple past and a perfect, when referring to a past-time event, is at least partly correlated with direct (witnessed) vs. indirect (non-witnessed) evidence. A less stable correlation for present-time reference can be observed between the use of present tense and direct evidence, and between the use of future tense and indirect evidence. Three mood markers have epistemic functions: *thaní* is a quotative; *maní* is a hearsay (or possibly mirativity) marker; and *ɡa* indicates inferred or assumed knowledge, although the latter seems only marginally grammaticalized. The second-position discourse marker *xu* signals surprise or emphasis.

As for tense-aspect and its correlation with direct vs indirect evidence, a similar contrast was already part of the ancestral Indo-Aryan tense system. There are also plenty of modern-day parallels in the immediately surrounding region, in which the use of a perfect serves as a marker of indirect evidence. That has for instance been noted in regional varieties of Persian as well as in Iranian Wakhi and Indo-Aryan Pashai. A highly grammaticalized evidentiality differentiation that is an integral part of the tense-aspect system, has been described for neighbouring Khowar and Kalasha, two languages that even in other respects have exerted influence over Palula, in the first case mostly as a superstrate, and in the latter as a substrate.

The mood markers—involving an emerging three-way paradigmatic contrast are to varying degree grammaticalized in Palula. Two of them are derived from converbs meaning 'to call, to say' and 'to speak, to say', respectively, another

tendency with numerous parallels in the wider region, and in several other Indo-Aryan languages. The third, and possibly less grammaticalized, marker is probably derived from an indefinite-interrogative pronoun 'what, any'.

The discourse particle *xu* is most likely a (relatively recent) loan from Pashto, a language of wider communication, but has become part of a set of discourse particles that seem to be particularly important and extensive in a subareally defined group of languages in the Pakistan-Afghanistan borderland to which Palula belongs.

### **Acknowledgements**

This work is part of the project Language contact and relatedness in the Hindukush region, supported by the Swedish Research Council (421-2014-631).

### **Abbreviations**


### **References**

Aikhenvald, Alexandra Y. 2004. *Evidentiality*. Oxford: Oxford University Press.

Baart, Joan L. G. 1999. *A sketch of Kalam Kohistani grammar*. Islamabad: National Institute of Pakistan Studies & Summer Institute of Linguistics.


### 5 Emerging epistemic marking in Indo-Aryan Palula


### Henrik Liljegren


5 Emerging epistemic marking in Indo-Aryan Palula


# **Chapter 6**

# **On the existence of egophoricity across clause types in Totoró Namtrik**

### Geny Gonzales Castaño

Laboratoire Dynamique du Langage, Université Lumière Lyon

The Barbacoan languages are known for having egophoricity systems (Dickinson 2000, Curnow 2002, Curnow & Liddicoat 1998), which exhibit a verbal marking pattern in which "speaker subjects in statements are marked the same way as addressee subjects in questions" (Curnow 2002: 614). Nevertheless, the existence of such a pattern in Namtrik had not been recognized. A recent paper by Norcliffe (2018) claims that in the Guambianovariety of Namtrik "the verb marking diverges from what might be considered canonical egophoricity marking, since it does not occur when the subject is second person in questions" (Norcliffe 2018). The current paper presents novel data from the highly endangered variety of Totoró Namtrik and argues that this variety possesses a set of verbal suffixes exhibiting the crosslinguistically recurrent pattern of an egophoricity distribution. The goal of this paper is to show that although Namtrik's egophoricity system is similar to the systems in other languages, it was not analyzed as a fully fleged egophoricity system in the past because the egophoricity suffixes are not always easily recognizable in interrogative clauses due to morphosyntactic and phonological processes. Additionally, this paper shows that Namtrik has an "undergoer" egophoric marker which exhibits a pattern of egophoricity distribution, shifting from speaker to addressee perspective in interrogatives.

### **1 Introduction**

Namtrik, also known as Guambiano, is a Barbacoan language (Curnow & Liddicoat 1998) spoken in the Colombian Andes. Traditionally, Namtrik speakers live in four so-called *resguardos* (settlements recognized by the Colombian State, with

Geny Gonzales Castaño. 2020. On the existence of egophoricity across clause types in Totoró Namtrik. In Henrik Bergqvist & Seppo Kittilä (eds.), *Evidentiality, egophoricity, and engagement*, 165–196. Berlin: Language Science Press. DOI: 10.5281/zenodo. 3975805

### Geny Gonzales Castaño

territorial autonomy and ruled by traditional authorities named *Cabildo*): Guambia, Ambaló, Quizgó and Totoró. These settlements are situated in two towns, Silvia and Totoró, located in the department of Cauca, in the Southwest of the country. Currently most researchers consider Totoró and Guambiano to be two dialects of the same language (Curnow & Liddicoat 1998), which the speakers refer to as *Namtrik*, or *Namuy Wam* 'our sound' (Gonzales Castaño 2013).

This paper will focus on a highly endangered variety of Namtrik, spoken in the community of Totoró. According to a census conducted by the authorities of the community in 2013, there are 76 native speakers (approximately 1% of a total population of 7023 people), who are all over 50 years in age (Gonzales Castaño 2013: 11). In Totoró the use of Spanish has displaced Namtrik in all daily interactions and there are now two generations of monolingual Spanish speakers. There is no intergenerational transmission of the language. Namtrik in Totoró is clearly more endangered than the Guambiano dialect in Guambia, which numbers about 8000–9000 speakers (Adelaar 1991: 66).

The data presented in this paper, beside a few examples constructed in elicitation, consists mostly of naturally-occurring examples. This data comes from two sources. The first is a lexicon collected between 2006 and 2008 within the frame of workshops on morphosyntax of Namtrik for community teachers and Namtrik speakers, which were directed by Tulio Rojas Curieux and Beatriz Vasquez. One of the results of this work was a lexicon (382 words and 448 sentences) edited and published in 2009 as *Léxico de la lengua Namtrik de Totoró* (Rojas Curieux et al. 2009). The other source of this data is a video and audio corpus, consisting of ten hours of natural speech data and additional elicitation data, collected between 2015 and 2017, in the framework of the project *Documentation and Description of Namtrik, an endangered language of the Colombian Andes* founded by HRELP-SOAS.<sup>1</sup>

The Barbacoan languages are known for their egophoricity systems (Dickinson 2000; Curnow 2002; Curnow & Liddicoat 1998; Floyd 2018), which "employ special verb forms […] where the subject (and/or another privileged argument role) is first person in declaratives or second person in interrogatives" (San Roque et al. 2018). Available grammatical sketches based on the Namtrik varieties of Guambiano and Totoró postulate the existence of a subject agreement marking system, which distinguishes between first person 'locutor' singular and plural and second and third person 'non locutor' (Pabón 1989; Triviño Garzón 1989; Vásquez de Ruiz 1987; Vásquez de Ruiz 1988).

<sup>1</sup>The data presented in this paper is available at the Endangered Languages Archive at SOAS University of London: https://elar.soas.ac.uk/Collection/MPI1012401.

### 6 On the existence of egophoricity across clause types in Totoró Namtrik

According to Vásquez de Ruiz (Vásquez de Ruiz 1987; Vásquez de Ruiz 1988) and Triviño Garzón (1989) the Guambiano variety of Namtrik presents an agreement verbal marking system distinguishing between a speaker singular subject marker -*(a)r*, a speaker plural subject marker -*er* and a non-speaker marker, singular and plural -*(a)n*. In the case of the Totoró variety, besides a different speaker singular subject marker -*(o)r*, Pabón (1989) proposes the same forms for the other egophoricity markers. In these previous descriptions of Namtrik (Pabón 1989; Triviño Garzón 1989; Vásquez de Ruiz 1987; Vásquez de Ruiz 1988), there is no information available about the behavior of egophoricity markers in questions; nevertheless the data presented show clearly that declaratives in Namtrik follow the expected pattern of an egophoricity system.

The existence of an egophoricity system in Namtrik was not argued for until a recent paper by Norcliffe (2018), which is based on the Namtrik variety from Guambia. Norcliffe (2018) notes, however, that while declaratives in Guambiano present the expected egophoricity pattern, the egophoricity markers do not appear in interrogatives. Nevertheless "layers of egophoric or egophoric-like marking are visible in Guambiano's grammar" (Norcliffe 2018), in Guambiano the verb marking pattern "does not exhibit the canonic egophoric pattern, since the egophoric markers do not appear in questions" (Norcliffe 2018).

The first goal of this paper is to show that the Totoró variety of Namtrik (TTK) possesses an egophoricity verbal marking pattern across clause types which "distinguishes speakers from non-speakers in declaratives, and addressees from nonaddressees in interrogatives" (Creissels 2008: 1); however, some morphophonological processes make it occasionally difficult to recognize this pattern in interrogative clauses. A second goal is to demonstrate that Namtrik in Totoró has an "experiencer" marker, which exhibits an egophoricity distribution, shifting from speaker to addressee perspective in interrogatives. Although I will show some usage patterns of the egophoricity markers, the semantic analysis falls outside the scope of this paper.

This paper gives a first account of the egophoricity system in TTK, beginning with a discussion on the morphophonological processes which make it difficult to recognize egophoricity markers in interrogatives. It then focuses on the description of the general usage pattern of egophoricity markers across clause types. And it concludes with the description of the use of the experiencer egophoric marker *-t* and the distribution of this marker in question predicate constructions. Geny Gonzales Castaño

### **2 Morphophonological processes affecting egophoricity markers in Totoró Namtrik**

Namtrik has a variety of morphophonological processes that occur at morpheme boundaries, which have not been fully argued for in previous descriptions. Understanding these processes is crucial for not only the description of the egophoricity system but also the description of Namtrik's grammar in general.

In TTK, there are two different morphophonological processes which make the presence of egophoricity markers in interrogative clauses unrecognizable: vowel deletion and consonant deletion.

Writing about the egophoricity system in Guambiano, Norcliffe (2018) proposes the existence of a set of verbs that present an alternation correlated with the existence of a vestigial egophoricity contrast in Guambiano, which show palatalized consonant forms in ego environments. In this section it will be argued that this alternation is not directly correlated with Namtrik's egophoricity system but with morphophonological processes.

### **2.1 Vowel deletion**

The Totoró variety of Namtrik (TTK) has two egophoric suffixes: singular -*or*<sup>2</sup> and plural -*er*, and one non-egophoric suffix: -*an*, as is shown in Table 6.1.


Table 6.1: Egophoricity markers in Totoró Namtrik

<sup>2</sup>The egophoric marker -*ar* which has been reported in Guambiano Namtrik is also attested in the Totoró variety in very few examples in the data. This egophoric marker has very restricted uses, occurring only in constructions with auxiliary verbs in declarative clauses, as is shown in example (i). The egophoric marker -*ar* is not attested in interrogative clauses. For these reasons it falls outside the scope of this paper.

(i) Totoró Namtrik (LEX1/118) *ye* potato *ma-ap* eat-DUR *wa-ar* sit.SG-EGO.SG 'I'm eating potatoes.'

### 6 On the existence of egophoricity across clause types in Totoró Namtrik


Table 6.2: Egophoricity markers on verb stems ending in consonants

The egophoricity markers are fully visible when they follow a verb stem ending in a consonant, as shown in the Table 6.2, which presents verb stems ending in a consonant, hosting the egophoric markers singular -*or* and plural -*er* and the non-egophoric marker -*an*.

In previous descriptions of Namtrik's verbal morphology (Pabón 1989; Triviño Garzón 1989; Vásquez de Ruiz 1987, 1988; Norcliffe 2018), the existence of allomorphs of the egophoricity markers has been postulated. The egophoric singular marker has two realizations -*ar* and -*r* (-*or* and -*r* in the case of Totoró) and the non-egophoric marker has the realizations -*an* and -*n*. However, the phonological context determining these realizations has not been clearly identified, with the vowel of the suffix described as sometimes being absent "depending on the value of the final vowel of the verb stem" (Norcliffe 2018).

Table 6.3: Egophoricity markers on verb stems ending in vowels


### Geny Gonzales Castaño

Table 6.3 shows examples of the TTK egophoric singular marker -*or*, the egophoric plural marker -*er* and the non-egophoric marker -*an*, following verb stems ending in vowels. Examples in Table 6.3 show that in all the cases either the final vowel of the verb stems or the initial vowel of the egophoricity markers is absent. The purpose of this section is to clarify the phonological conditioning of this process of deletion.

Although in TTK it is possible to find sequences of two vowels within lexical stems or suffixes, the language puts constraints on vowel sequences at morphological boundaries. When a stem or a suffix ends in a vowel and it is followed by a suffix beginning in a vowel, one of the two vowels in this sequence is elided. The identity of the elided vowel does not depend upon the position of the vowels in the sequence, but on the particular combination of vowels. Table 6.4 shows the combinations attested in the corpus and the results of vowel elision, as well as an example for each pattern.

These patterns of vowel elision may be summarized in the rule illustrated in Figure 1, which also shows that there exists a vowel hierarchy in TTK with respect to this morphophonological process. In this hierarchy the vowels /i/, /e/, /o/ and/u/ are stronger than the high central vowel, which is stronger than the vowel /a/.

	- i e o > ɨ > a u

The process of vowel elision was identified in TTK by Pabón (1989). Pabón proposes that TTK puts constraints on the combination of high vowels and high or middle high vowels (1989: 16). Nevertheless, the constraints on the vowel sequences at morphological boundaries do not concern only these combinations of vowels, as shown in the Table 6.4. This is a regular and widespread phonological process in the language and does not solely affect the egophoricity markers.

As noted above, the process of vowel elision is found with other suffixes of the form -V(C) or -C(V) as well as the egophoric and non-egophoric markers. Every morpheme ending or beginning with a vowel can lose its last phoneme, as is the case of the vowel /ɨ/ in the morpheme -*mɨ* 'NEG' in coalition with the morpheme -*elɨ* 'NMZ.PL' in example (1), or its first phoneme, as example (2) shows in the case of the vowel /e/ in the morpheme -*elɨ* 'NMZ.PL' affixed to the noun *nosrka*- 'brother'.


### Table 6.4: Possible combinations of phonemic vowels

### Geny Gonzales Castaño


This process makes some morphemes unrecognizable in certain phonological contexts. This is the case of the imperative second person singular marker -*ɨ* which is visible when attached to a verb stem ending in consonant, as shown in examples (3) and (4), but fully deleted following a vowel ending verb stem as shown in (5).


The only sequence which has the vowel /a/ as outcome of the phonological process of vowel deletion is the sequence /ae/. The vowel /a/ is the "weakest" one in the hierarchy which determines this process in TTK. This is important to keep in mind for the description of egophoricity system in TTK; since the nonegophoric suffix -*an* begins with /a/ and is often affected by the process of vowel

### 6 On the existence of egophoricity across clause types in Totoró Namtrik

elision, which makes it often difficult to recognize, especially in interrogative clauses. As we will see later, in Namtrik, the egophoricity markers lose their final consonant in interrogative clauses, so the non-egophoric suffix -*an* suffers a total deletion in questions following a vowel-final verb stem. This is similar to the deletion of the imperative second person singular marker -*ɨ* in this phonological context, as shown in example (5).

### **2.2 Consonant deletion of the egophoricity markers in interrogative clauses**

Questions in TTK are marked by a rising intonation, a high pitch on the last syllable of the clause and the deletion of the final consonant of the egophoric and non-egophoric suffixes or, less frequently, the replacement of the final consonant by a glottal stop /ʔ/. The following set of examples shows two sentences using -*an* — one declarative and one interrogative – in similar kinds of predicate structure and with the same verb stem.

In the second example of each set, it can be seen that the final consonant in the non-egophoric marker -*an*, following verb stems ending in consonant — *pen*- 'fall' in (6b) and *pasr*- 'stand.SG' in (7b) — is retained in the surface form of the interrogative clauses.

	- b. Totoró Namtrik (namtrik\_483/057) *chu* chu where *pen-a* pen-an fall-NON.EGO 'Where did it fall?'

### Geny Gonzales Castaño

b. Totoró Namtrik (namtrik\_058/112) *chine* chi-ne Q-3 *marɨp* ma-rɨp do-DUR *pasr-a* pasr-an stand.SG-NON.EGO 'What is she doing?'

Although it is possible to recognize the non-egophoric marker -*an* in interrogative clauses when it follows a verb ending in a consonant, when it follows a vowel-final verb it is completely omitted, as is illustrated in the following set of examples. The examples (8) and (9) (a declarative and an interrogative clause), show the non-egophoric marker -*an* hosted by the copula *kɨ*-. In example (9) the surface form of the verb seems to have a bare stem form, without any verbal marker. However, the absence of the non-egophoric marker is explained by the phonological processes of deletion of vowels in coalescence and the deletion of the final consonant of the egophoricity markers in interrogative clauses.


### **2.3 Deletion of the non-egophoricity marker in interrogative clauses**

As is claimed by San Roque et al. (2018), writing about interrogativity in evidential and egophoricity markers, a formal asymmetry with distinct interrogative morphemes or constructions is attested in several unrelated languages. Similarly in TTK there are two different question constructions, one for interrogatives involving third person subjects and another for interrogatives involving first and second person subjects.

Aside from the deletion of the final consonant of the non-egophoric marker and the rising intonation, the predicate question construction involving third person subjects has the same morphological and syntactic structure as the declar-

### 6 On the existence of egophoricity across clause types in Totoró Namtrik

ative predicate construction, as is shown in the following set of examples. Examples (10) and (11) show an interrogative and a declarative clause with similar kinds of predicate structure and the same verb stem.<sup>3</sup>

	- b. Totoró Namtrik (namtrik\_083/91, elic) *pi=pe* water=PD *mani* now *[kuakl-ap* boil-DUR *pasr-an]* stand.SG-NON.EGO 'The water is boiling.'
	- b. Totoró Namtrik (LEX1/104) *wasri* wasri sparrowhawk *atrun* atru-an come.SG-NON.EGO 'The sparrowhawk comes.'

The question predicate construction involving first and second person subjects is a complex predicate construction consisting of a main verb stem and an obligatory copula auxiliary *kɨ*-, as is shown in example (12)–(14) (see also 34 and 41). First and second person questions are marked additionally by the deletion of the final consonant of the non-egophoric marker and rising intonation (cf. example 14).

(12) Totoró Namtrik (namtrik\_031/94) *wasrchikyupe* wasr-chik-yu=pe bag-DIM-LOC=PD *chinɨ* chi-nɨ Q-3 *wai* wai put *ko* kɨ-or COP-EGO.SG 'What did you drop in the bag?'

<sup>3</sup> In the examples below, square brackets denote a complex predicate construction.

### Geny Gonzales Castaño


In questions involving second person subjects and first person plural subjects, the egophoric suffixes -*or* and -*er* are still recognizable in the surface form of questions, since the vowels /o/ and /e/ are not affected by the process of vowel deletion as illustrated in (12) and (13). However the non-egophoric marker -*an* is not visible in the surface form of the interrogative clauses involving a first person singular subject. Nevertheless, as we already mentioned, the omission of the non-egophoric marker in interrogatives is not related directly to the grammar of the egophoricity system but to the morphophonological processes of vowel deletion and the deletion of the final consonant of the egophoricity markers in interrogative clauses.

### **3 Verb alternation in verb stems ending in palatal, alveolar and dental consonants**

Before describing the usage pattern of egophoricity markers across clause types in TTK, there is a final point to explore regarding morphophonological processes and their relationship with the egophoricity system. In the description of the morphophonology of the Guambiano variety of Namtrik, Vásquez de Ruiz (1987: 58) claims the existence of a set of verbs which alternate between two verb stem forms, one ending in a palatal consonant and the other in a dental-alveolar consonant. Recently Norcliffe (2018) proposed that this alternation is correlated with the existence of a vestigial egophoricity contrast in Guambiano. This set of verbs would show palatalized consonant forms in ego environments.

Table 6.5 shows examples of verb stems ending in palato-alveolar consonants hosting suffixes beginning in a vowel: the egophoric markers singular -*or* and

### 6 On the existence of egophoricity across clause types in Totoró Namtrik

plural -*er*, the non-egophoric marker -*an*, the durative morpheme (-*ip*/-*ɨp*) and the verb nominalizer -*ap*. 4


Table 6.5: Verb alternant in verb stems ending in palatal and alveolar consonants

As is shown in the examples in Table 6.5, when the verbs end in /t/, /ɳ/ and /ʃ/, the final consonant of the verb stem is realized more alveolar: V (C palatoalveolar or alveolar#) \_ a/ +alveolar. If the verb ends in a palato-alveolar affricate /ʧ/ or an alveolar affricate /ʦ/ the verb alternation is realized only when it is hosting the nominalizer -*ap*.

Furthermore, when the verbs are followed by a suffix beginning with a consonant, the alternation is no longer observed, even in ego environments, since it is caused by the first vowel of the suffix following the verb stem, as is shown in the following sets of examples. Examples (15a)-(15c) and (16a)-(16c) show the verbs *pen*- 'fall' and *kish*- 'cry' followed by the negation marker -*mɨ* and the egophoricity markers. The above analysis suggests that this alternation is not correlated with the egophoricity system but with a morphophonological process.

(15) *pen*- 'fall'

a. *penmor* pen-mɨ-or fall-NEG-EGO.SG 'I don't fall.'

<sup>4</sup>When the verb stem ends in the consonants, *l*, *ll*, *n*, *ñ*, *ch*, and *r* , the difference between the durative (-*ip*/-*ɨp*) and the verbal nominalizer -*ap* is neutralized, and the morpheme -*ɨp* fulfills both functions.

### Geny Gonzales Castaño

	- a. *kishmor* kish-mɨ-or cry-NEG-EGO.SG 'I don't cry.'
	- b. *kishmer* kish-mɨ-er cry-NEG-EGO.PL 'We don't cry.'
	- c. *kishmɨn* kish-mɨ-an cry-NEG-NON.EGO.PL 'S/he doesn't cry.'

### **4 The general usage pattern of egophoricity markers in Totoró Namtrik**

Up to this point I have discussed the morphosyntactic and phonological processes which often make the presence of egophoricity markers in interrogative

Table 6.6: The general usage pattern of egophoricity markers in Awa Pit (Curnow 2002)


### 6 On the existence of egophoricity across clause types in Totoró Namtrik

clauses unrecognizable. These processes are mainly of two types: elision of vowels through vowel coalescence and consonant deletion in egophoricity suffixes in interrogative clauses. This section is dedicated to the discussion of the usage of egophoricity systems across clause types in TTK.

As noted above, egophoricity is canonically defined as a morphological category which marks speaker subjects in statements and addressee subjects in questions in the same way. Awa Pit, another Barbacoan language, shows a prototypical egophoricity system (Curnow 2002: 613), as is shown in Table 6.6 and in the following sets of examples of declarative clauses (17a)-(17c) and interrogative clauses (18a)-(18c).Comparing the two sets of examples, it is observed that the declarative clause involving a first person subject (17a) is marked in the same way, with the egophoric marker -*s*, as the second person subject interrogative clause (18b). These examples also show that in Awa Pit interrogatives involving first person subjects (18a), are marked in the same way, with the non-egophoric marker -*y*, as the declarative clauses involving second person subjects (17b).

	- a. *(na=na)* (1S.(NOM)=TOP) *pala* plantain *ku-mtu-s* eat-IMPF-EGO 'I am eating plantains.'
	- b. *(nu=na)* (2S.(NOM)=TOP) *pala* plantain *ku-mtu-y* eat-IMPF-N.EGO 'You are eating plantains.'
	- c. *(us=na)* (3S.(NOM)=TOP) *atal* chicken *ayna-mtu-y* cook-IMPF-N.EGO He/she is cooking chicken.'
	- a. *min=ta=ma* who=ACC=INT *ashap-tu-y?* annoy-IMPF-N.EGO 'Whom am I annoying?'
	- b. *shi=ma* what=INT *ki-mtu-s?* do-IMPF-EGO 'What are you doing?'
	- c. *min=ta-s* where=LOC-ABL *a-mtu-y?* come-IMPF-N.EGO 'Where is he coming from?'

### Geny Gonzales Castaño

Table 6.7 show the general usage pattern of egophoricity markers in TTK. Although it is similar to the Awa Pit system it also presents some differences. The TTK egophoric system has a contrast in the egophoric suffixes between singular -*or* and plural -*er* which does not exist in the non-egophoric suffix -*an* and which is not observed in the other Barbacoan languages.

Furthermore, TTK's system is not a fully symmetrical egophoricity system, since, as I already mentioned, there is a formal asymmetry with distinct interrogative constructions. TTK has two different predicate interrogative constructions, one for first and second person and another one for third person. In interrogatives involving first and second person subjects there is an obligatory copula auxiliary *kɨ* which does not appear in interrogatives involving third person subjects.


Table 6.7: The general usage pattern of egophoricity markers in Totoró Namtrik

### **4.1 Usage pattern of egophoricity markers in declaratives**

TTK has two main kinds of predicate constructions: simple and complex predicate constructions. Simple predicate constructions, whose morphological structure is shown in Figure 2, are formed by a single verbal stem suffixed with the polarity marker, egophoricity markers and/or modal epistemic markers, as shown in examples (19) and (20).


6 On the existence of egophoricity across clause types in Totoró Namtrik

(20) Namtrik Totoró (Rojas Curieux et al. 2009: 29) *nai* na-wai 1-GEN *ishukpe* ishuk=pe women=PD *kumarmɨn* kumar-mɨ-an sing-NEG-NON.EGO 'My wife didn't sing.'

In TTK, simple predicate constructions do not carry aspectual morphology and the aspectual interpretation depends on the Aktionsart of the verb. While activities are interpreted as perfective past (21), states are interpreted as simple present (22).


Namtrik also has a wide range of complex predicate constructions which vary widely in their morphosyntactic structure and the semantic relationships between the verbs comprising the complex predicates. Figure 2 shows the minimal morphological structure of a complex predicate in TTK. Although V<sup>2</sup> and V<sup>1</sup> can be inflected with verbal morphology markers in complex predicates, there is a clear split between the kinds of markers that can be hosted by each one of the verbs in the complex predicate.

As is shown in the examples (23)–(25), the verb in V<sup>1</sup> position can host the polarity marker, egophoricity markers and/or modal epistemic markers, while the verb in position V<sup>2</sup> presents a non-finite, which can be inflected with the morpheme of prospective aspect (-*tr*/-*ch*/-*ntr*) PROS, the verbal nominalizers -*ik* 'NMZ.SG' and -*elɨ* 'NMZ.PL', and the morpheme of durative aspect (-*ap*/-*ip*/-*ɨp*).


*nai* na-wai 1-GEN *kekpe* kek=pe husband=PD *[kuallip* kuall-ip work-DUR *pasran]* pasr-an stand.SG-NON.EGO 'My husband went to work.'

Geny Gonzales Castaño


Although any verb of the language can fill the position in complex predicates V2 , the position V<sup>1</sup> can only be filled by a restricted set of verbs, which are shown in Table 6.8. In complex predicates the morphological structure of V<sup>2</sup> and the kind of auxiliary determine the modal and aspectual interpretation of the predicates. The complexity of the predicate constructions in TTK falls out of the scope of this paper, nevertheless it is important to keep in mind for the description of the egophoricity system that the verb in position V<sup>1</sup> hosts the egophoricity markers in complex predicates.


Table 6.8: Auxiliary verbs in Totoró Namtrik

The canonical use of the egophoricity markers in declarative clauses is illustrated in examples (26)-(33). The canonical use of the egophoric singular marker

### 6 On the existence of egophoricity across clause types in Totoró Namtrik



The canonical use of the non-egophoric -*an* for second and third person, is illustrated in examples (30)-(33). The non-egophoric suffix -*an* is shown in a sentence with a second person singular subject in example (30), in (31) with a second person plural subject, in (32) with a third personal singular subject, and in (33) with a third plural person subject.

The TTK egophoricity system has a contrast in the egophoric suffixes between singular -*or* and plural -*er* which does not exist in the non-egophoric suffix –*an*. Nevertheless some verbal stems of auxiliary verbs, as is shown in Table 6.8, have a plural and a singular form, which allows us to identify the grammatical number of the subject. This is the case of the verbal stems in the examples (32) *pasr-an* 'stand.SG-NON.EGO' and (33) *putr-an* 'sit-PL-NON.EGO'.

### Geny Gonzales Castaño


### **4.2 Usage pattern of egophoricity markers in interrogatives**

Table 6.9 shows the interrogatives predicate constructions in TTK. As we already mentioned TTK has two different interrogative constructions: one for first and second person and one for third person. The main difference between the two constructions is the obligatory use of a copula auxiliary *kɨ*- in the interrogative clauses involving first and second person subjects which is not present in the interrogative clauses involving third person subjects.

Table 6.9: Interrogative predicate constructions in Totoró Namtrik


### 6 On the existence of egophoricity across clause types in Totoró Namtrik

The canonical use in interrogative clauses of the egophoricity markers is illustrated in examples (34)-(48). Examples (34) and (35) show the use of the nonegophoric marker -*an* in interrogatives involving a first person subject, singular in (34) and plural in (35). Since asking about one's self is usually perceived as an odd situation for Namtrik speakers, data on first person subject interrogatives is very marginal in the corpus used. (37) is a natural speech example, and (35) was elicited based on the structure of natural speech examples of first person questions.

As Floyd (2018: 13–14) notes regarding interrogatives in Chapalaa, first person interrogatives usually present a special marking which distinguishes them from second and third person interrogatives. As is shown in Table 6.9 and in the examples (34) and (35), first person interrogatives additionally have an irrealis marker -*am* which is not present in the interrogative clauses involving second and third person subjects.

Given the small amount of data concerning first person questions, it is not possible yet to argue if TTK presents a single construction for polar and content first person questions. However it is possible to recognize the use of the non-egophoric marker -*an* in interrogative clauses involving first person subjects, keeping in mind the morphophonological processes affecting this marker in questions is already described in the first part of this paper.


The use of the singular egophoric marker -*or* in interrogatives involving second singular person subjects is illustrated in examples (36)-(39). Examples (36)

### Geny Gonzales Castaño

and (37) show the use of the egophoric marker -*or* in content questions and examples (38) and (39) in polar questions. TTK has the same predicate construction for second person content and polar questions.


The use of the plural egophoric marker -*er* in content questions involving plural second person subjects is illustrated in examples (40) and (41). The use of the plural egophoric marker in polar questions is shown in examples (42) and (43)

(40) Totoró Namtrik (namtrik\_003/28) *chi* chi Q *pɨrɨ* pɨrɨ why *kishke* kish-kɨ-er cry-COP-EGO.PL 'Why are you (kids) crying?'

### 6 On the existence of egophoricity across clause types in Totoró Namtrik


Regarding interrogatives in the variety of Namtrik spoken in Guambiano, Norcliffe (2018) proposes that "in interrogatives involving second person subjects, the verb in its bare stem form is obligatorily followed by a particle which takes the form of either *ku*/*ke* or *titru*. *Ku* is used for singular second person addressees, while *ke* is used for plural second person addressees", as is shown in examples (44), (45) and (46).


### Geny Gonzales Castaño

In the case of TTK, the egophoricity system exhibit a verbal marking pattern, which is consistent with the description of egophoricity systems in other languages in which "speaker subjects in statements are marked the same way as addressee subjects in questions" (Curnow 2002). As is shown in examples (36)- (43), the forms *ko* and *ke* observed in questions involving second person singular and plural in Guambiano Namtrik, correspond to the copula *kɨ*- followed by the egophoric markers -*or* and -*er* in Totoró Namtrik.

The canonical use of the non-egophoric marker -*an* in questions involving a third person is illustrated in the examples (47) and (48). TTK also has the same construction for content and polar questions involving third person subjects.


### **5 Egophoricity markers and the category of person**

As San Roque and colleagues note regarding egophoricity marking and subject agreement marking, "under a person-marking account, egophoricity markers flag the coincidence of the epistemic authority and an argument role" (San Roque et al. 2018: 33). Since subjects usually "have good grounds to declarative direct personal knowledge of an event, it naturally follows that declaratives with a first person subject usually take egophoric marking, as do interrogatives with a second person subject" (San Roque et al. 2018: 33).

This approach may describe much of the patterns in TTK, but has trouble accounting for the data which show that under certain semantic and pragmatic factors, the non-egophoric suffix can be used with first person subjects in declaratives as well and that the egophoric plural marker -*er* can be used in question clauses with a third person subject.

It was attested in egophoricity marking in different languages (San Roque et al. 2018) that egophoric markers usually do not target non-volitional or non-control

### 6 On the existence of egophoricity across clause types in Totoró Namtrik

first person subjects. As is shown in examples (49), (50) and (51), in TTK egophoric marking is not used in declarative clauses which describe a non-volitional or non-controlled event performed by a first person subject.

Additionally, examples (49) and (50) also show the non-control marker -*ra*, which has also been reported in Guambiano by Vásquez de Ruiz (2007: 98–99) and serves to express the speaker's lack of control or evidence with respect to the described event. This marker takes part in a specific complex predicate construction whose structure is shown in Figure 4, and which however is a construction restricted to first person subjects, and is never marked with the egophoric singular marker -*or* but with the non-egophoric marker -*an*.

4. Non-control event complex predicate construction V2 -DUR-NON.CONTROL-NMZ COP-NON.EGO(-DUB)


(50) Totoró Namtrik (namtrik\_078/75, elic) *kiprain* ki-ap-ra-in sleep-DUR-NON.CONTR-NMZ *parin* parin a.lot *kiprain* ki-ap-ra-in sleep-NFV-NON.CONTR-NMZ *kɨn* kɨ-an COP-NON.EGO *nape* na=pe 1=PD *mɨintɨ* mɨin-tɨ now-REST *kasrar* kasra-or wake.up-EGO.SG ' I was sleeping a lot and I barely woke up.'

(51) Totoró Namtrik (namtrik\_070/1) *nape* na=pe 1=PD *kuap* kua-ap die-NFV *sreik* sre-ik escape-NMZ.SG *kɨpɨk* kɨ-ap-ik COP-DUR-NMZ.SG *kɨn* kɨ-an COP-NON.EGO *nape* na=pe 1=PD *srɨ* srɨ LOC.DIST *kasareik* kasar-ap-ik get.married-NFV-NMZ.SG 'I almost died when I was just married.'

Although the general usage pattern of the Namtrik egophoricity system seems to be very consistent, the system has some flexibility and there are examples

### Geny Gonzales Castaño

where the markers are not used according to the expected pattern. In examples (52)–(53) there is an egophoric plural marker -*er* in question clauses with a third person subject. It is not clear what motivates the use of the egophoric plural marker -*er* in these sentences.


### **6 Undergoer and egophoricity**

The experiencer egophoric marker -*t* is involved in different kinds of constructions in TTK. In complex predicate constructions, this marker is always suffixed to the auxiliary verb. This marker has been reported also in previous descriptions of Guambiano by Vásquez de Ruiz (2007: 101) and Triviño Garzón (1994: 615). Vásquez de Ruiz (2007) described the marker -*t* as a first person applicative, which has epistemic-evidential functions, indicating that there is a first person which is perceiving some information described in the clause (Vásquez de Ruiz 2007: 101). More recently Norcliffe (2018), writing about evidential and egophoricity markers in Guambiano, describes this marker as "a subtype of ego evidentiality ... used in contexts in which the speaker experiences an event as an undergoer, and thus has knowledge of the event as an affected participant" (2018: 25).

In TTK the main function of this marker is similar to the function that has been described by Vásquez de Ruiz (2007), Triviño Garzón (1994) and Norcliffe (2018). The main function of the -*t* morpheme is to mark a speaker which is in some way affected by the situation described in the clause, as illustrated in the examples below, which show the undergoer marker -*t* in intransitive declarative clauses.

### 6 On the existence of egophoricity across clause types in Totoró Namtrik

	- a. *nakish* nakish smoke *pesrik* pesrik angry *kɨtan* kɨ-t-an COP-EXP-NON.EGO *kashi* kashi a.bit 'The smoke is bothering me a bit.' (lit. The smoke is angry, which affects me a bit.)
	- b. *kana* kana one *sruk* sruk stone *tɨwei* tɨwei only *wepitan* we-ap-i-t-an come.out-DUR-NMZ.E-EXP-NON.EGO '[My father] just found a stone, (which the duende had hidden to make my father him sick, my father said).'
	- c. *ashampe* ashan=pe now=PD *si* si yes *kek* kek husband *palɨtan* pal-ɨ-t-an lack-E-EXP-NON.EGO 'Now I need a husband.'

The undergoer marker -*t* may be suffixed to both intransitive verbs and transitive verbs. In declarative sentences with transitive verbs, the suffix -*t* marks an 'affected' speaker, which in these cases is expressed as an object.

### (55) Totoró Namtrik (Rojas Curieux et al. 2009: 16, 25)


As is the case in other Barbacoan languages (Floyd 2018: 25), TTK has specific predicate constructions for experience verbs, including internal states, endophatic states, emotions and desires. In these cases, the non-egophoric suffix -*an* is used with first person subjects in combination with the undergoer egophoric marker -*t*.

### Geny Gonzales Castaño

	- a. *nante* na-wan-te 1-DAT-REST *kiri* kiri get.scared *kitan* kɨ-t-an COP-EXP-NON.EGO *chilliyu* chilli-yu mud-LOC *puraintrap* pura-i-ntr-ap go.through-go.SG-PROS-NFV 'I got scared going through the mud.' (lit. There was fear in me going through the mud.)
	- b. *nante* na-wan-te 1-DAT-REST *kitra* kitra be.cold *kitan* kɨ-t-an COP-EXP-NON.EGO 'I am cold.' (lit. Cold is to me.)
	- c. *nante* na-wan-te 1-DAT-REST *intsa* intsa laugh *kɨtan* kɨ-t-an COP-EXP-NON.EGO 'I laugh.' (lit. There is laughing for me.)

The data from Totoró Namtrik shows that the undergoer egophoric marker -*t* seems to have an egophoricity distribution. That is to say, the suffix -*t* refers to the speaker being affected in declarative clauses and to the hearer being affected in interrogative clauses.

The predicate construction involving the undergoer egophoric marker, which is shown in Figure 5, consists of a main verb followed by the suffix -*t* and the nonegophoric suffix -*an*. In interrogative clauses, the consonant in the non-egophoric suffix -*an* follows the same pattern of consonant deletion found in the predicate constructions without the suffix -*t*.

5. Undergoer egophoric marker predicate construction V-EXP-NON.EGO

The following set of examples shows the pattern of usage of the undergoer egophoric marker in declarative clauses and interrogative clauses.

	- a. *mɨimpe* mɨim=pe now=PD *kana* kana one *paso* paso glass *palɨtan* pal-ɨ-t-an lack-NMZ.SG?-EXP-NON.EGO 'Now I need a glass ' (lit. Now a glass lacks, which affects me.)
	- a. *chillo* chillo knife *palɨta* pal-ɨ-t-an lack-NMZ.SG?-EXP-NON.EGO 'Do you need a knife? (lit. Does it lack a knife for you?)
	- b. *maik* maik delicious *itrɨta* itrɨ-t-an smell-EXP-NON.EGO 'Does it smell delicious (according to you)?' (lit. Does it smell good to you?) c. *tapɨ kɨta*
	- tapɨ well kɨ-t-an COP-EXP-NON.EGO 'Is it okay for you (according to you)?'
	- d. *mai* maik well *mutsikɨta* muts-i-kɨ-t-an drink-NMZ.SG?-COP-EXP-NON.EGO *ñipe* ñi=pe 2=PD 'Do you want to drink?'

### Geny Gonzales Castaño

### **7 Conclusions**

Totoró Namtrik data shows a regular egophoricity pattern, exhibiting a very similar distribution as the one found in other Barbacoan languages. However, its first recognition is obscured by two morphophonological processes: vowel coalescence at a morpheme boundary and deletion of the final consonant of the egophoric and non-egophoric suffixes in questions. The fact that TTK possesses an egophoricity system is a good argument for claiming that egophoricity is a genetic feature in Barbacoan languages, as argued by Curnow & Liddicoat (1998).

The system that I illustrate in this paper is very similar to egophoricity systems in other languages. In declarative clauses the egophoric singular -*or* and plural -*er* are used for first person and the non-egophoric -*an* for second and third person. In interrogative clauses, the egophoric singular suffix -*or* is generally used for singular second person, the egophoric plural suffix -*er* for plural second person and the non-egophoric -*an* for first and third person. TTK shows a single egophoricity system both in declarative and interrogative clauses. In this respect the TTK data differ from the Guambiano data, as presented by Norcliffe (2018), which show two separate systems for declarative and interrogative clauses.

Further the TTK data shows an undergoer egophoric suffix -*t*, which seems to function as an egophoric undergoer suffix, signifying that the speaker is in some way affected in declarative sentences. The undergoer egophoric suffix -*t* appears to be used in questions to refer to the hearer being affected. This would mean that this suffix shows an egophoric distribution as well. However, more data is needed to confirm this.

The Totoró Namtrik egophoricity system does not always behave according to the general usage pattern. These non-canonical uses of the suffixes in the egophoricity system will play an important role in the semantic analysis of this system.

### **Abbreviations**


### 6 On the existence of egophoricity across clause types in Totoró Namtrik


### **Acknowledgements**

I would like to thank Martine Bruil for all her support, ideas and stimulating discussions which fed our presentation in the "Symposium on evidentiality, egophoricity, and engagement: descriptive and typological perspectives" (Stockholm University 17–18 March 2016) and this paper. I would also like to thank my advisers Tulio Rojas Curieux, Antoine Guillaume and Spike Gildea as well as Esteban Díaz for all their suggestions and ideas for my analysis and their support. I would also like to thank the Cabildo authorities, teachers and the Namtrik speakers from the community of Pueblo Totoroez: Erminia Sánchez, Gertrudis Benachí, Encarnación Sánchez, Gerardina Sánchez, Carmen Tulia Sánchez, Carolina Lúligo, Micaela Lúligo, Juanita Sánchez, Carlina Conejo, Ismenia Sánchez, Aristides Sánchez, Marco Antonio Ulcué, Transito Sánchez and José María Sánchez. I would also like to express my deep gratitude to Lucy Elena Tunubalá Tombé for her invaluable work and help in the transcription and translation of the data presented in this paper. Finally, I would like to thank the Ecole doctorale 3LA, Université Lumière Lyon 2, Laboratoire Dynamique Du Langage, the Grupo de Estudios Lingüísticos Pedagógicos y Socioculturales del Suroccidente Colombiano, Universidad del Cauca and the Hans Rausing Endangered Language Documentation Programme (HRELP-SOAS) for supporting my research.

### **References**


# **Chapter 7**

# **Interpersonal alignments and epistemic marking in Kalapalo (Southern Cariban, Brazil)**

### Ellen B. Basso

University of Arizona

Kalapalo epistemic features are central to the pragmatic functions of interpersonal alignments. There is a set of seven hearsay evidentials mainly used in narrative and in quoted discourse, eight initial position epistemic expressives (EXP), five evidential suffixes (EV) and hearsay quotative evidential strategies (qUOT), together with a large set of (28) second position epistemic clitics/particles (EM). The latter confirm that epistemic judgment is essentially a triadic stance procedure involving the social actors engaged in conversational interaction (an initial speaker, an interlocutor or listener-responder) and the epistemic object/proposition. I sort EM into six sets, two of which (A, B) mark the speaker's "internal" and "external" degrees of epistemic judgement. (C) mark the speaker's contraspective wishing or hoping, (D )mark the speaker's re-evaluation, correctives or denial of a proposition with a focus on new or reconsidered information. The final two sets (E, F) concern the speaker's attempts at (or conflicts with) epistemic alignment with a listener or 3rd person, based on understanding of the epistemic object. While epistemic modality in the sense of a "scale" or "grade" occurs, in Kalapalo there are other non-modal meanings including participation (or not) of the speaker and listener and a third person in an epistemic context, mirativity, incredulity (an extreme skepticism or unwillingness to believe), and conflict or denial, affinal civility and affection.

### **1 Introduction**

The idea of "interpersonal alignments" is used in this paper to refer to a languagefocused developmental process involving emergent patterns of communication

Ellen B. Basso. 2020. Interpersonal alignments and epistemic marking in Kalapalo (Southern Cariban, Brazil). In Henrik Bergqvist & Seppo Kittilä (eds.), *Evidentiality, egophoricity, and engagement*, 197–237. Berlin: Language Science Press. DOI: 10.5281/ zenodo.3975807

### Ellen B. Basso

and understanding in the interaction of distinctively different people within a community. Using this concept (with the words 'interpersonal relations') the modernist anthropologist Edward Sapir repeatedly noted in his lectures at Yale that the interaction of people of different personalities and their influence on the thought and action of a community has consequences with many historical implications (Sapir 1993: 204). When people respond to one another during conversational interaction as they receive information about the worlds of others, Sapir noted, they participate in contexts wherein judgments or evaluations of propositions are being made, often while speakers are trying to make decisions and plans for future action. More recently, John Du Bois' technical use of "alignment" refers to such ongoing activity "in which two participants in dialogical interaction ...converse in varying degrees" in taking a stance (2004: 22–23). Evidential and epistemic markers contribute to a speaker's evaluation, self-positioning, and alignments with other subjects concerning the sharing of knowledge and evaluation of the epistemic object or proposition. (Du Bois 2007: 24). Such stance enactments are central to the pragmatic functioning of sociality and everyday power (Ameka 2004: 5–6).

In this chapter, I discuss stance enactments as they occur among speakers of Kalapalo, a Southern Cariban agglutinative language currently spoken by about 630 people in the Alto Xingu region of central Brazil (Guerreiro 2015). In Kalapalo, a large number of epistemic markers are central to the pragmatic functioning of social life. My interest in this epistemic marking has grown out of my earlier anthropological work on Kalapalo narrative discourse (*akiñatunda*), a dialogical practice structured in large measure by the interaction between a narrator (*akiña otoi*) and a "listener-responder" (*etuitsofo*) (Basso 1985; 1986; 1987; 1995). Analyses of recorded narrated speech events assisted my understanding of the use of the epistemic features. In this material, dialogicality exists in the many narrated quoted conversations in which occur emergent processes of challenge, resistance, debate, deception, and the negotiation of meaning. Emotional events such as respect and endearment, as well as anger, shame, lust, greed, and envy are important for descriptions of protagonists and it is through their quoted speech that we can see how these emerge from social interaction (Basso 2007). A narrator's authority to speak of imaginary or historical subjects is an adventure in language, a kind of critical social commentary that may not be tolerated in other public situations. Presentation of sociohistorical and mythological realities include fantasies regarding human bodily processes and tricksters' appetite for subversion, as well as historical memories of women who have been abducted and who learned to live in a different kind of society, and the personal adventures of warriors and shamans. The materials presented here include quoted conversations (often, in-

### 7 Interpersonal alignments and epistemic marking in Kalapalo

terspersed with narrators' commentary), as well as from my own conversations with Kalapalo, and from the ritual speech of hereditary leaders (Basso 2009; see also Franchetto 1983, 2000).

In Kalapalo there is a set of six hearsay evidential markers (EV) mainly used in narrative and in quoted discourse (Table 7.1); eight initial position epistemic expressives (EXP) which directly mark first person judgment and sometimes 1st person evidence (Table 7.2) 1 and the large set of (28) second position epistemic clitics or free particles (EM) that I have ordered into six pragmatic sets; EM numbers correspond to the full set listed in the Appendix. 2

Kalapalo epistemic markers appear at the end of an adverbial phrase, a nominal phrase or clause, or a converbal or lexical verbal clause. There is rarely any vowel reduction; except for the morpheme *aka*, and no epistemic clitics are vowel initial. The morpheme *muke* may show elision when followed by a vowel initial phonological word. Nonetheless, most epistemic morphemes function as clitics insofar as they are phonologically bound through (second syllable) stress patterning to their host constructions, thereby playing a key role in phonological word construction. Epistemic markers have the ability to be compounded with one another, and with other types of clitics (Basso 2014), as seen in examples (6), (33b), (35), (40a), (51), (62c).<sup>3</sup>

A number of the Kalapalo parameters seem to closely match those specified by B. J. Hoff (1986) for Surinam Carib (Kari'na) and also Bruna Franchetto' s (n.d.) discussion of (so-called modal) "marker's of true speech" in the closely related Alto Xingu Kuikuro language. However, with one exception, I use different language to describe these features in keeping with my emphasis on person-person deixis and dialogicality in stance processes. What B. J. Hoff calls "grade" (strong-weaklacking), "speaker's attitude", "appeals on speaker", "change of grade" and "coexistence with supporting or conflicting evidence" are clearly semantic parameters found in Kalapalo epistemic marking. Hoff's specification of an underlying

<sup>1</sup>Another set of expressives are affective (marking fear, pain, sensory pleasure, disgust, grief, frustration). Their use by Kalapalo speakers tends to create a synergy of sentiment with listeners. The epistemic expressives are both interpersonal as well as cognitive and introspective in semantic function as shown in Table 7.2. Both subsets of the expressive word class are in the main used by speakers to introduce further descriptive commentary.

<sup>2</sup>Due to the large number of EM markers and complex semantic features, some of which (e.g., assertion, inferential, negation) may be repeated in the use of several forms, I have used numbers rather than semantically informative glosses. Full descriptions occur in the examples and the Appendix.

<sup>3</sup> Similar morphemes have been described in the pioneering work on Northern Cariban languages by B. J. Hoff (1986), Eithne Carlin (2004); and Sergio Meira (1999). Northern Cariban epistemic clitics usually appear after the first element in a clause (Wackernagel's position); the scope of the clitic including the entire clause (Hoff 1986).

### Ellen B. Basso

Table 7.1: Evidential suffixes (EV) and evidential strategies (quotatives)


contrast in the Surinam Carib particle set between the origins of "extraspective" and "introspective" evidence is also an evidential feature in Kalapalo, and this can be understood with reference to Ferdinand de Haan's (2001) approach that allows for a pragmatic description subsuming evidential and epistemic features. de Haan distinguishes between two deictic categories relevant to evidentiality: in the first, the speaker is separated (or separates themselves) from the action being described; in the second, the speaker includes themself in the description of the action. In fact, these make sense for Kalapalo epistemic marking when the stance object is foregrounded in the discourse segment (particularly in Sets A and B). However there is also the third person marked both as a logophoric interlocutor and as an epistemic object. There are four epistemic markers which can be used to describe logophoricity of several kinds: *kafa* EM17, weak but positive; *kato* EM18, troublesome ; *nafa* EM20, an animate but non-human response; *kalaka* EM23, reflexive desire. Alignment (or "engagement") is yet another deictic field, involving a speaker referencing another subject's inclusion in the sphere of epistemic stance (as in Sets E and F). These involve participant frameworks

### 7 Interpersonal alignments and epistemic marking in Kalapalo


Table 7.2: Epistemic Expressives (EXP)

relevant to Kalapalo epistemic markers in which the intersubjective relation is foregrounded and the object-subject relation recedes (though it is not entirely absent from discursive commentary).

While epistemic modality in the sense of a "scale" or "grade" is a feature of Kalapalo epistemology, there are important non-modal features that emerge as speakers evaluate information, including a) participation (or not) of the speaker and listener and a third person in a cooperative epistemic context; b) changes

### Ellen B. Basso

or correctives due to receipt of new information; c) mirativity (due to new and surprising information) and incredulity (an extreme skepticism or unwillingness to believe); d) counter-factive or contra-spective wishing or hoping; and e) acceptance or rejection of a conjoint project. I show in examples how the four logophoric markers mark responses to different kinds of experiences, speech, or desires and plans of someone other than the interlocutors: Although I follow earlier writers (Aikhenvald 2004, de Haan 1999, 2001; Nuckolls & Michael 2012), in making a distinction between evidentiality (EV) and epistemic markers (EM) as shown in the preceding tables, Kalapalo EV, EXP and EM are often combined in discourse as many of the examples will show. Additionally some EM also mark types of evidentiality (first hand experience; inference – or lack of direct speech evidence – regarding a second or third person's thoughts and plans). These are not seen in the EVset. Past tense semantics occurs with the use of *wãke* EM1 and *nipa* EM24, of interest as there is no past tense marking on the verb. In addition, four EM (*kiŋi*, plural *kiŋi-ni* EM11, *pile* EM15, tiki EM19, *tima* EM28) have negation meanings. What follows is a brief listing of the six epistemic sets sorted according to pragmatic and semantic features. These sets are discussed further with examples in Sections 2–4, followed by my final observations in section 5. Names of original narrators and location of my research recordings are given following the examples.

Outline of Epistemic Sets

	- **–** Set A: knowledge has been received internally; the speaker is a participant in the propositional context.
	- **–** Set B: knowledge is received externally; the speaker is not a participant in the propositional context but must infer or deduce from this external evidence.
	- **–** Set C: self-correctives.
	- **–** Set D: counter-factive or contra-spective acknowledgment, wishing or hoping'.

### **2 Internal versus extenal contrasts in Sets A and B**

Sets A and B are similar to Berend Hoff's (1986) specification of an underlying contrast in the Northern Cariban Kari'na (or Surinam Carib) particle set between the origins of two types of evidence. The first is "introspective", "inner world of the speaker", evidence in the mind of the speaker and private knowledge. The second is "extraspective" evidence from the "outside world", that is, evidence external to the speaker's experiences or public knowledge.

### **2.1 Set A**

Knowledge is "internal", part of the speaker's experience. There are six members of this set. These markers seem to reference an interloctor's thoughts or ideas, rather than direct speech and thus contrast with the hearsay EVgroup.

### **2.1.1** *wãke* **EM1**

With this frequently used marker, the speaker asserts a (usually distant) past experience in which knowledge or evidence has been acquired. The strong assertion is often seen in the speaker's repetition of the marker after each clause. In the example, upon hearing that his friend is the lone survivor of a massacre, the speaker declares that was why he didn't join them earlier, expecting to have to avenge them all.

(1) *ege-tomi=dye-fa* PDEM-PURP=SS-TOP *wãke* EM1 *u-te-lï* 1-go.away-PNCT *wãke,* EM1 *o-pi-ñï-ko-i* 2-avenge-AN-PL-COP *u-its-ani,* 1-EX-FUT.1 *u-ki-lï* 1-utter-PNCT *wãke* EM1

'For this purpose **I remember some time ago** I said, 'I will be the one to go away as your avenger.' (said by Kudyu at Aifa, 1980)

### Ellen B. Basso

### **2.1.2** *tifa* **EM2**

This marks the affirmation of the interlocutor's current understanding of (or thoughts about) the speaker's recently spoken ideas. This marker may be based on the hearsay EV*ti***-**. See example (40a) where EM 2 contributes to the interlocutors belief that what he is saying is true. Most examples apear to be referencing a 2nd or 3rd person's thoughts rather than speech. The following example marks an exclusive plural person's acceptance.

(2) *tis-eti-dyi-pïgï-iña=tifa* 1+3-come.out-ITR-VPE-BEN=**EM2** *ti-ŋifa-nïgï.* 1+3-teach-PS

> '**You are right** to think that on behalf of our offspring (i.e., 'those who come out of us'), is what we teach.' (said by Kudyu at Aifa, 1979)

Here, the speaker uses both EM2 in the nominalization part of the construction and *taka* EM22 agreement with someone else's description of their experience in the bracketed adverbial clause; EM22 references his mother's just-stated dismay at seeing his reddened eyes. The EM2 *tifa* references the speaker's acknowledgment of what she thought of the events being described.

(3) *(afïtï=taka* denial=EM22 *igei-ufuŋu)* IDEM-unlike *ta-ŋo-fïŋï-tifa...* DIS-NLOC-resemble-EM2 *tisuge,* 1+3 *afïtï* denial 'No, (=**because of what you saw**) you must know there's got to be another kind of place where we can live, no.' (said by Kudyu at Aifa, 1979)

Here the speaker confirms a nth2 person's understanding of a past event. *wãke* EM1 is used to assert the speaker's own participation. Because of the group context, I have used a first person plural translation.

(4) *ule-tse-ngugi=tifa=wãke,* AFR-be-interrupted=EM2=EM1 *ngikogo* fierce people *e-nïgï* come-PERF *wãke.* EM1 *aifu-pe-fa.* end-ESS-PTP

'**We saw** he was interrupted **just as you thought he would be** when he started things, **we saw** the fierce people come. It was all over for him.'

### **2.1.3** *laka* **EM3**

This marks the speaker's very weak inference, puzzlement, or a strong inability to understand the participatory context. This example includes both EM3 and the EXP*uum*, referencing thought of an imaginative or inferential nature. In de Haan's deictic model (2001), the speaker has put himself into the sphere of action. 7 Interpersonal alignments and epistemic marking in Kalapalo

(5) *u:um* **EXP** *aŋi=laka* result=**EM3** *ukuge* human being *ele-i* DEM-COP *ukuge* human being 'I suppose that could have been human, a human being.' (said by Tufule at Aifa, 1982)

In a trickster story, the speaker is surprised that he has already been told not to eat some fish parts, although he has already done so. The marker *seku* following the verb *te*-,' eat' is the mirativity marker compounded with EM3.

(6) *te-ŋe=seku=laka* eat.flesh-TRNS=MIR=EM3 *a-nïgï* EQS-PFV *u-feke=lefa* 1-ERG=SEQ 'I don't think I was already told not to eat flesh!' (said by Tufule at Aifa, 1982)

### **2.1.4** *ma* **EM4**

Used as a clitic, this dubitative form references the speaker's uncertainty due to lack of knowledge. It is used mainly, but not exclusively, with interrogative forms. It may also serve as an epistemic neutral marker (see also conversational examples 19–21, 22–24, 44c, 48c, 54c). In the example, a narrator questions how someone in his story could put a large fish inside a small flute:

(7) *ta-me=ma* CONT-FACS=EM4 *kuluta* flute *atati* inside *tu-i-ŋalï* REF-put-REV *i-feke* 3-ERG 'How could he have put it back inside something like that flute?' (said by Kambe at Aifa, 1979)

### **2.1.5** *maŋa* **EM5**

The speaker, while enunciating a proposition, at the same time denies any possibility of its occurrence; this form is used to express incredulity, often with the rhetorical (y/n) question prefix *tï* ('I can't believe you're asking/saying (X) as you and I both well know the answer'.)

The envious Trickster declares he will make the same kinds of rare and beautiful things given to his younger brother.

(8) *u:um* EXP *tï-kïtsï=maŋa* RQ-ugly=EM5 *t-iŋuG-isi* REF-make-ADV *Taugi* NAME *ki-lï* utter-PNCT '"Who thinks this is hard for me to make?/ that I can only make it badly?" (="This **is** easy for me to make"), Taugi spoke.' (said by Tufule at Aifa, 1979)

### Ellen B. Basso

A leader's ritual communication is often filled with this kind of ironic restraint, a kind of respectful devaluation of the work of his own messengers. There is a feeling of the speaker's modesty enacted by his disclaimer.

(9) *tï-kaiŋa-fïŋï=maŋa* RQ-DEST-resemble=EM5 *Ø-atsa-ki-lï* 3-run-TRNS-PNCT *atehe=gele-fa* PER=still-TOP *wãke* EM1 'Who thinks they still don't run up to some place as they had done in the past?' (Leader's talk, spoken by Ageu at Aifa, 1998)

### **2.1.6 kaɳa EM6**

The speaker expresses the unlikeliness of an event. Here, an event has taken place, and the speaker had mistakenly counted on a third person (his father-inlaw) to act in a certain way. Line (10b) has the EM11 from Set C which marks polite regret to the listener for his agreeing to let his father-in-law do so.

	- b. *ukw-oto-feke=kiŋi* dual-relative-ERG=EM11 *a-tïfïgï-ko* EQS-IMP-PL *ige-tomi.* take.away-PURP 'regrettably I **let** our relative (parent) be our escort/ leader.' (said by Kambe at Aifa, 1980)

In example (11), the speaker uses EM6 to taunt an enemy to begin shooting at him. He asserts that it's not likely he can rely on the interlocutor to bring him some arrows. The speaker uses the *ki*-applicative to change the verb 'go' from an intransitive to a transitive 'go to do X' with the complement verb' bring' .

(11) *uege=kaŋa* you=EM6 *te-ta-tiga* go.to-CONT-HAB *u-fïgi* 1-arrow *ï-ki.* bring-INST 'I can **hardly count on you** to always go to bring my arrows to me (=i.e., 'shoot me').' (said by Kudyu at Aifa, 1980)

### **2.2 Set B**

This set includes information marked as "external" to the speaker's experience or not foregrounded. There are four members of this set. As with A, there is an assumption about the interlocutor's thoughts, wishes, or plans, but not any direct hearsay evidence.

7 Interpersonal alignments and epistemic marking in Kalapalo

### **2.2.1** *nika* **EM7**

The marker is a kind of strong supposition of the speaker's regarding the interlocutor's experience or wish. It is often a strong marker of mourning (see example 16).

A husband is anxiously looking for his wives and comes to a place where he expects to find them. He politely suggests to some other women that they have seen his wives there:

(12) *aŋi=nika* result=EM7 *inde* here *u-oku-ŋi-ta* 1-liquid.food-TRNS-CONT *i-ŋi-lï* 3-see-PNCT *e-feke-ne* 2-ERG-PL 'You **might have** seen those who make my food around here.' (said by Tufule at Aifa, 1978)

As in examples (13) and (14), there are also compounded clitics.


### **2.2.2** *tata* **EM8**

There is possibility but some uncertainty as the epistemic object involved a 3rd person, or a distant past source of information, and therefore the speaker could not know for certain that the event took place. However, some probability from inferential or customary experience exists.

### Ellen B. Basso

(17) *igifagafïtï* settlement.name *anetu-gu* leader-POSS *uŋu* house *a-nïgï=tata* EQS-PFVS=EM8 *i-feke* 1-ERG 'Possibly the Igifagafiti leader has a house.' (said by Kudyu at Aifa, 1979)

A woman recounts another's actions based on common female experience, but since the event took place in the far distant past, the speaker isn't entirely certain of the accuracy of what she's saying.

(18) *lepene* then *ø-atu-ndi-lï=tata* 3-tap-TRNS-PNCT=EM8 *i-feke* 3-ERG 'Then I think she might have tapped on it' (said by Tufule at Aifa, 1979)

### **2.2.3 fïna EM9**

With this marker the speaker expresses the idea that there is some possibility but there is no direct evidence. The following is a typical remark after the Trickster seem to have acted in his usual covert manner.

(19) *Taugi=fïna=mbe* Taugi=EM9=SE 'It might have been Taugi who did that' (said by Tufule at Aifa, 1979)

Grandmother Quail sees someone who has has pulled up all her peanuts. *fïna* appears in this example with the mirative =*seku*, which in this example has scope over both utterances. *fïna* has scope over the line (20b) utterance only.

(20) a. *tï-seku=ma* RQ-MIR=EM5 *egei* ADEM *u-etigite-gï-ki* 1-peanuts-POSS-INST *ga-tiga.* make-HAB '**What is this here?** Someone's messing with my peanuts.' b. *eŋï* reason *oto-ni* food-non-existent *i-nïgï-ko=fïna=seku-fa.* EX-PFV-PL=EM9=MIR-TOP '**Could it be that's** because they don't have any food of their own?' (said by Kambe at Aifa 1979)

The forest monster is frightened by a man who disguises himself as an *adjafi* owl in order to scare him away (this owl is a bad sign to the observer).

(21) *adyafi=fïna its-a*

> owl=EM9 EX-CONT

'That could be an adafi I'm seeing.' (said by Tufule at Aifa, 1979)

### 7 Interpersonal alignments and epistemic marking in Kalapalo

### **2.2.4 koh EM10**

The speaker has no knowledge of what is being described. This morpheme is most often a particle. and is also used to begin a sentence as an expressive (example 23). The scope of *koh* covers the utterance which it follows.

A husband returns and begins to burn a pile of brush in which, unknowingly, his wife's lover is hiding. When the man runs away to escape the fire, the husband wonders why his wife never told him.

(22) *ukuge* person *wende* over.there *e-ki-nu* 2-utter-non-existent *koh-i* EM10-COP *u-feke* 1-PRSP '**Why** didn't you tell me there was someone over there?' (said by Olafu at Aifa, 1979)

Several brothers, frustrated in not having found their sister in a nearby settlement, say the following to one another:


### **3 Discussion of Sets C (Contraspective), and D (Counter-expectation)**

### **3.1 Set C**

Contraspective: in the sense there is an imaginative wishing or hoping (sometimes, in vain) for a difference in what has been told to, or what is being observed by, the speaker.

### **3.1.1** *kiŋi***;** *kiŋi-ni* **(plural) EM11**

This is used to mark regret for what has been told to the speaker. This contraspective marker is a negative nominalization of the verb *ki* 'utter'. The following example also appears in (10b); the speaker regrets having participated in a joint venture.

### Ellen B. Basso

(25) *uk-oto-feke=kiŋi* dual-relative-ERG=EM *a-tïfïgï-ko* EQS-IMP-PL *ige-tomi.* take.away-PURP 'I regret I let our relative (parent) be our escort.' (said by Kambe at Aifa, 1980)

As in this example EM11 often appears hosted by the agreement expression *eh he* whereby a speaker acknowledges affirmatively what the interlocutor has said but is expressing regret (other examples include 49d; 50b; 62b, 63c).


### **3.1.2** *muke* **EM12**

This marks the restrained or suppressed character of the utterance, an indication of the speaker politely expressing a positive wish or expectation. There can also be a realization on the speaker's part that the proposition may be in vain. This is a frequent marker in narratives; see also the conversational examples (39b), (47b), (62a), (62c), (69b).

(29) *u-ño* 1-husband *its-iŋa=muke* EX-SN=EM12 *ukuge* human 'If only he were human he might be my husband.' (said by Ugaki at Aifa, 1979)

This example shows the speaker's use of *muke* to mark his restraint and modesty as a son-in-law proposing a work party to help the family (The full conversational context appears as examples 39–40).

(30) *eŋï=muke-tsï-fa* do=EM12-EV-TOP *fesoko* fish.name *apuGi-tsofo=muke-tsï-fa* flavor-USIN=EM12-EV-TOP *ku-pehe-ne* 1+2-ERG-PL 'This way hopefully we'll make some flavoring for our fesoko fish.' 7 Interpersonal alignments and epistemic marking in Kalapalo

### **3.2 Set D**

Counter-expectation or reevaluation of evidence with regard to the receipt of new information. The speaker corrects or denies their own proposition. There are three markers.

### **3.2.1** *maki* **EM13**

The speaker reevaluates what s(he) knows. New (introspective) information results in reevaluation of speaker's own earlier proposition. In (31): the adverb *ande* is a deictic feature. Knowing there is a prior proposition comes from the context of the narrative in which the speaker claims to need to travel in another direction which the interlocutor says won't take him to the river.

(31) *ande=maki* here.now=EM13 *fanguiŋga=lefa,* river=SEQ *nïgifeke.* qUOT '"Ok, now I see that the river is here after all", he said (to him).' (said by Kambe at Aifa, 1982)

In (32), the speaker, who had thought to be a member of the Trumai group, now realizes he has been abducted as a child and says:

(32) *afïtï=maki* denial=EM13 *Tugumai-fuŋu* NAME-resemble *ku-kuge* 1+2-people 'not I realize we people resemble Trumai', i.e. 'I realize we people are are not Trumai.' (said by Ausuki at Aifa 1982)

### **3.2.2** *makina* **EM14**

New (introspective ) information leads to acceptance of the interlocutor's earlier proposition which was rejected:

A man has killed his mother-in-law but has told his wife that the woman has died from a fish on which she choked to death. The narrator begins this part of his story with a description of what the wife sees (her first-hand evidence (33a– 33b)), which is followed by the woman's surprised reaction to this evidence: her use of the *ah* expressive to assert her understanding of the truth of the event, followed by her acceptance of the husband-listener's deception, using EM17 and the mirative *ki*.

(33) a. *ege-te=gele* PDEM-DIS=still *is-isi-ŋa=gele* 3-throat-inside=still 'Still there inside her throat'


A man uses the polite reference to his parent-in-law, *uk-oto-fo-ko*.

(34) *uk-oto=fo-ko=makina* dual-parent-PL=EM.28 *ege,* he *uk-oto-fo-ko* dual-parent-PL 'Now I realize that's not our parent, our parent.'

A woman tells her brothers her abductor in the past was not one of their kind:

(35) *ah,* EXP *ukuge-fïŋï=makina=wãke* 1-person-resemble=EM14=EM1 *igei=wãke* IDEM=EM1 *u-iki-dyu=lefa* 1-abduct-PNCT=SEQ *u-feke* 1-ERG *tsa=l=ifeke,* tell-always=ERG *i-ño-pe-feke.* 3-husband-ESS-ERG

'"The fact is, the one who abducted me before wasn't a person like us as I now realize", she kept telling them about this other husband of hers.' (said by Ugaki at Aifa, 1980)

### **3.2.3** *pile* **EM15**

The speaker moves from acceptance to rejection of a proposition.

A young man has agreed throughout the narrative to go to dangerous places to get things for his brother (whose wife is his lover), but finally realizes the brother is trying to kill him.


7 Interpersonal alignments and epistemic marking in Kalapalo

### **4 Discussion of Sets E and F**

**Sets E** (five examples) and **F** (eight examples) concern the speaker's marking of epistemic alignment, that is, confirmation of a shared (or in two cases, disputed) proposition. These sets are probably large because they are used when there is a speaker's need to shift between a focus upon the epistemic object, the marking of a unique or shared perspective, and the marking of the interpersonal alignment, that is, participation with the interlocutor in an epistemic context involving a joint venture. As is often the case, such contexts themselves change throughout a conversation and there is often 'disclaimed' responsibility in which EM play a significant role. My examples include a number of examples of dialogic contexts that illustrate the co-construction of epistemic and evidential meaning. Examples of markers from sets A-D appear in these longer examples. Speakers in these examples discuss issues of responsibility, denial of conjoint participation in a pragmatic context, and engagement in deception while using the affinal civility register (Basso 2007).

### **4.1 Set E**

Markers in this set concern shared information and reference the interlocutors participating in a joint venture; one negative marks refusal to share information. Conjunct/disjunct contrasts occur and degrees of knowledge are also marked.

### **4.1.1** *tafa* **EM16**

The speaker asserts their own (or a 1st person plural) alignment with the listener. The example shows both the speaker and interloctor's use of the "distant future" FUT.2, as a reciprocal pragmatic politeness marker (see Mendoza 2016 on use of future as a politeness strategy).

Cuckoo's mother tells him to come home right away from his uncle's settlement if the man's daughter he wishes to marry decides she doesn't want him after all. Cuckoo's answer is in Line (38b).

	- b. Cuckoo answers his mother: *s-agage-dye=tafa* 3-same-SS=EM.16 *u-e-n-ïm-iŋo* 1-return-INTR-FUR.2 *ama* Mother.VOC 'I will come back, Mother, if she is the same (as you've said).' (said by Kudyu at Aifa, 1979)

### Ellen B. Basso

### **4.1.2** *kafa* **(EM17)**

A logophoric marker, *kafa* indicates the speaker's presumption of a 3rd person's wants, feeling, or experience. In this narrative section from the same story as (34), four epistemic markers appear. The speaker affirms the 3rd person has accepted their proposition and asks his wife to accept a trip to get salt, which he apparently wants to do to benefit the family. The noun 'our parent' used for 'parent-in-law' and *muke* EM12 are especially typical of the affinal civility register he is using, and the markers of politeness clearly enable what turns out to be his terribly deceptive speech.

(39) a. Use of 'our parent' as politeness marker:

*aŋi=kafa* result=EM17 *uk-oto-i* dual-parent *ŋukugu=mbo-lï* stay.behind=HYP-PNCT *ta* qUOT *i-feke,* 3-PRSP *ta-ki* qUOT-MIR *i-feke* 3-ERG

'"This parent of ours could have stayed behind **herself** after all", he said to her, surprising her by saying that.'


In line (40a), the speaker uses compounded markers EM22 *taka* for suggestion of a 2nd person prior agreement and *nafa* logophorically for marking the third person; the speaker also uses the EM2 *tifa* to affirm her need or wish. Of interest is the narrator's final comment about how even though he is lying to his wife, the speaker is able to convince her by using these epistemic markers that he should take his mother-in-law to the salt plants.

(40) a. *uge=taka=nafa* me=EM22=EM20 *ukoto–iña=tifa* dual-parent-BEN=EM2 *ikï-ponde-fï* drying.rack-arrange-ADV *u-feke* 1-ERG *nïgï-fi-feke.* qUOT-EV-ERG

> '"**I recall you may have told me (about her)** that our parent wanted me to be responsible for setting up the drying rack for her benefit ", he must have said something like that.'

7 Interpersonal alignments and epistemic marking in Kalapalo

b. *eŋgu-Gi-ta* deceive-TRNS-CONT *i-feke* 3-PRSP 'He was deceiving her.' (said by Kudyu at Aifa 1979)

### **4.1.3** *kato* **(EM18)**

This is another marker with logophoric uses, a strategy for gossipy speech as the speaker is sharing worrisome information regarding a 3rd person with an interlocutor, and denies any responsibility for what is being said. The speaker confirms what the 3rd person has said, but does not always accept it as legitimate. As the speaker aligns in this manner with the 3rd person outside the present speech context the shared proposition is puzzling, or even worrisome.

From the Kwambï, a ritual song during which gossip about the singer is "thrown back" at the original speakers:

(41) *kuGife-mbe=kato* witches.dart-SE=EM18 *ti-ka-gi-ti* REF-make-TRNS-DES *i-feke* 3-ERG 'I'm worried he wants to make witches' darts.' (sung by Kudyu at Aifa, 1967)

People are trying to kill jaguars. They come to a community to see if some youths can be prepared as warriors. The boys are tested by killing a tapir. In Line (42a) the speaker uses the third person logophiric *kafa* because the childen's actions confirms their proposal of a joint venture. In Line (42b), there is use of *kato*. Here the proposal that the children might be able to kill the jaguars is considered possible by the speaker but still somewhat weak and worrisome, which is given further context by the narrator's explanatory material in Line (42c).

(42) a. *agetsi-tsi=mbembege=dya=kafa* one-M=PE=DS=EM17 *ule-tsaŋe* AFR-DEO

> '**They seem** to have wanted to do that together to it (the tapir), what we had them do.'


'They were still frightened by what those others (jaguars) kept doing.' (said by Saŋafa at Aifa)

### Ellen B. Basso

### **4.1.4** *tiki* **EM19**

This marks the speaker's refusal to accept a role in an endeavor described by the interlocutor or by third persons. *tiki* is also a mirative form that emphasizes the speaker's sudden surprise concerning the contrastive understanding or disagreement. The form may be derived from the negative mirative suffix –*ki* (see also example 48f–48g with *tifa*).

A woman has been abducted by powerful beings and left in her hammock tied to a tree in the wilderness. When she wakes up, she says:

(43) *una=tiki egei u-e-tïfïgï*

> Q=EM19 ADEM 1-come.to-IMP

'How in the world did I (or ? could I have wanted to') come here like this?' (said by Tufule at Aifa, 1979)

A warrior from a cannibal group cannot understand why his future wife says others fear he has come to kill them all, because he's only come to marry her. (see earlier part of this conversation in (70).

	- b. *ñ-eŋe-tu-nda-ko* DE-fear-TRNS-CONT-PL *ts-ale* EX-always *igei* many *e-feke* IDEM 2-ERG 'They're frightened of what you're always like.'
	- c. *tuelï-ko* kill-PNCT-PL *faŋa-mi-ta* worry-adversative-CONT *igei* IDEM *efeke.* 2-ERG 'They're worried that you will kill them all like this.'
	- d. *ta-tiki* RQ-EM19 *Ø-ẽ-lï-ko-iña* 3-kill-PNCT-PL-DAT *u-e-na-lï* 1-come.to-INTR-PNCT 'Why should anyone say I've come here to kill you all?'
	- e. *awu-nda-fïŋï=mbe=dye* like-CONT-resemble=SE=SS *tsa=lefa.* EX=SEQ 'What they're saying is like a lie.'
	- f. *e-lï-ko-iña-la* kill-PNCT-PL-DAT-QN *tale* NEG.always *igei* IDEM *u-e-tïfïgï,* 1-come-IMP *nïgifeke* qUOT 'I never came like this to kill you all'

### 7 Interpersonal alignments and epistemic marking in Kalapalo


### **4.1.5** *nafa* **(EM20)**

With this marker, a 1st person responds to the interlocutor regarding a 3rd person animate non-human agent or subject participant. *nafa* is used in regard to the speaker assuming the interloctor has customary knowledge of the consequences of the context.

(45) *u-ifi-fo-lu=nafa* 1-touch-HYP-PNCT=EM20 *e-feke* 2-ERG *sike-feke-fa* tocandira.ant-ERG-TOP *e-fife-po-lï* 2-bite-HYP-PNCT 'You must know if you were to touch me the tocandira ant would sting you.' (said by Tufule at Aifa, 1979)

While introducing the character Cuckoo, the storyteller confirms that I already know what he's talking about, as he was keenly aware I had already worked with others identifying Kalapalo bird names.

(46) *fitsagu,* cuckoo *tu-fu-ti-sï=nafa* REF-know-TRNS-PNCT=EM20 *e-feke* 2-ERG 'Cuckoo, you must already know yourself what that is.' (said by Kudyu at Aifa, 1982)

The following example describes people escaping a massacre who are giving out dead hummingbirds as food to their fellow travellers. What occurs as a consequence is a case of oracular interpretation. Responsibility for a problematic decision is displaced deictically away from the discourse of the 2nd person human locutors and onto the 3rd person non-human epistemic object. In Line (47b), there is a use of *muke* (EM12)as 'hoping in vain'. In Line (47c), there is a use of *maki* (EM13) marking change of opinion after receipt of new information (see also line iv). Line (47d) shows EM13 suffixed by the negator –*la* (QN) as are final elements of the other clauses.

### Ellen B. Basso


reason.' (said by Ausuki at Aifa 1982)

In this example, an entire conversation is presented to show how EM19 emerges at the end of presentation of knowledge by the interlocutors. This is when the speaker expresses surprise and also questions knowledge. This utterance is constructed with a rhetorical question that emphasizes the speaker's sudden and surprising realization of the truth of what is being told him: Note in line (48d). there is a compounded hearsay EVand EM20.

(48) a. *apiči,* Grandfather *nïgifeke* qUOT 'Grandfather, they said to him.' 7 Interpersonal alignments and epistemic marking in Kalapalo


'That's what **we know as we've been told** that about you.'

e. *e-figï=nafa* 2-grandson=EM20 *egei* ADEM *itau-kuegï* woman-AUG *muku-gu-pe,* son-POSS-ESS

'**You now know** Monstrous Woman had a son, that's him over here.'

f. *uge=tifa* me=EM2 *egei* ADEM *Ñafïgï* NAME *muku-gu-pe.* son-POSS-ESS

'**I, myself, have learned** I am one of Ñafïgï's sons.'

g. *eh he!* agreement *nïgifeke* qUOT *ah* EXP *ande-ŋu=tiki* here/now-DIM=EM19 *u-fi-dyau* 1-grandchild-PL *a-nïgï* EQS-PFV '"All right!" he answered. "The fact is, **I had no reason to think** you little ones here were my grandchildren!"' (said by Kambe at Aifa 1980)

### **4.2 Set F**

This set marks differences between the speaker and listener regarding a speaker's agreement to participate in the proposal of a shared context for information. The speaker appeals to a listener with a marked grade of confidence, referencing the conjoint situation. There is a contrast between these and EM19, which marks the speaker's denial of any conjoint agreement.

### **4.2.1** *aka* **EM21**

The speaker substantiates the interlocutor's proposition. Information is shared by the speaker with the listener. The speaker declares there is existing, positive alignment between herself (often as a member of a non-inclusive plural group) and the listener. In example (49a), the speaker uses *nika* (EM7 Set B) to align with the interloctor's experience. In Line (49d), he uses *eh he kiŋi*, regretful agreement.

Based on his own experience, Cuckoo judges his wife's family's eyes are constantly bothered by the smoke in their house.

### Ellen B. Basso

	- b. *eh.* yes *ŋi-ke-fa* see-I-TOP *tisuge...* 1+3 'Yes, look at us.'
	- c. *igei-fuŋu=aka* IDEM-resemble=EM21 *tisuge,* 1+3 *igei-fuŋu* IDEM-resemble 'As you see this is what we are like, like this'.
	- d. *eh he* assent *kiŋi* EM11 'Yes, but if only it weren't so.' (said by Kudyu at Aifa, 1982)

The Dead are invited to return to the living but decline. In Line (50a), the use of *nifa*, EM25, marks that the (quoted) speaker shares a conjoint decision with the listener.

(50) a. *tis-ogo-pi-tsa=nifa* 1+3-go.back-INTR-CONT=EM25 *aña* dead *ki-lï* utter-PNCT '"All of us have agreed to go back", the Dead said.' b. *tits-e-lï=aka* 1+3-go-PNCT=EM21 *ake-ts-igei=lefa* SD-EX-IDEM=SEQ 'And so, as you see, we people must take leave of you.' (said by Tufule at Aifa, 1979)

### **4.2.2** *taka* **EM22**

This is a weak or uncertain conjunct marker. With this marker, the speaker proposes to the listener (s)he will probably agree with the proposition but speaks somewhat uncertainly. In my examples, =*taka* is seen with somewhat hesitant, rather than assertive speech, as there is no direct evidence the listener can use to assert knowledge or accept the proposition.

(51) *aŋi* exist *muke=taka* EM12=EM22 *itau=go-pïŋï* woman-PAU-DEF *fogi-tsïgï* find-IPE *u-feke* 1-ERG 'As I hoped there were a few deficient women I found by chance.' (said by Kambe at Aifa 1979)

### 7 Interpersonal alignments and epistemic marking in Kalapalo

The following example shows *taka* hosted by an "I told you so" quotative (bracketed). The quotation, which precedes the quotative, is included in the scope of the EM =*taka*.

(52) *uk-aŋi-fofo* dual-exist-IM *[u-ki-ta=taka]* 1-utter-CONT=EM22 *egei.* IDEM

> '"Let's wait a while", I suppose you **might remember** I said that.' (said by Nikumalu at Aifa 1979)

Use of *taka* with a Q word. The speaker asks a listener to help him remember someone's name:

(53) *uwa=taka* Q=EM22 *i-ñandsu* 3-sister *ititï,* name *Kamisu?* NAME

'What did I say **about** the sister's name? Kamisu?' (said by Ugaki at Aifa, 1979)

What follows is a conversational example regarding responsibility, with examples of several different logophoric markers. After travelling from place to place without finding her, the relatives searching for a woman have arrived in a community where some warriors (during the time of her abduction) had seen people on the river. They have evidence but never knew how to interpret it until they are told several years later about this woman who has been abducted. Their interlocutor uses *muke* as a politeness strategy. In Line (54b), *kafa* marks the speaker's weak presumption of the third person's experience. In Line (54e), *nika* is the second person logophoric marker emphasizing the speech of the visitor and *wãke* marks the speaker's first hand, distant past experience used to assert his knowledge to the listener. What has happened is the warriors realize they didn't have any way of knowing the people they saw on the river were the woman and her abductor. If so, they might have been able to release her. The narrator describes the following conversation:

	- b. *aŋi* result *fogi=kafa* search.for-EM17 *inde* here *iñandsu* sister *e-tïfïgï* come-IMP *inde=mukwe* here=EM12 *u-limo,* 1-children *nïgifeke* qUOT

'"Is there a chance someone who I'm looking for may be here, I hope the sister came here my sons ", he said about her.'


'How could we have known if that was the daughter we saw before (about whom you spoke) we saw before?' (said by Ugaki (Aifa, 1982)

### **4.2.3 kalaka EM23**

EM23 marks the speaker's impressions of a vague 3rd person reflexivity. The impressons are positive although the speaker may have no direct evidence regarding the 3rd person's thoughts or even the identity of the person. This contrasts with *kato* (EM18), which marks worrisome impressions of a hearsay or gossipy nature on the part of the speaker. Irrealis features in the examples contribute in several ways to the idea that *kalaka* marks an uncertainty ( because it introduces the clause marked with *koh* EM10), and even a polite proposition (because the clause is used with the -*iŋo* FUT.2 distant future).

A woman suggests to her sisters that a new fruit might be used as a drink if the Trickster invents it. The use of the irrealis distant future FUT.2 contributes to this weak proposition.

(55) *aŋi* result *fogi=kalaka* find=EM23 *tï-iñambe=nïm-iŋo-koh* REF-drink-INTR-FUT.2=EM10 *ta-ti* qUOT-EV *i-feke* 3-ERG '"He could find *in itself* a kind of drink for us sometime", she kept saying about it, they say.' (said by Tufule at Aifa 1979)

From a conversation between a Kalapalo speaker and the present author about usage:

7 Interpersonal alignments and epistemic marking in Kalapalo

(56) a. *ege* 3PDEM *muku-gu* son-POSS *ipo-pïgï=kalaka* pierce-VPE=EM23 *egei-i* ADEM-COP 'She might have wanted someone to pierce her son (i.e. pierce her son's ears).'

b. *ah-la-tsï,* EXP-like.that-EV *ege* 3PDEM *muku-gu* son-POSS *i-po-pïgï* 3-pierce-PFV *its-a* EX-CONT *egei.* ADEM 'The fact is I've been telling you her son's ears have been already pierced.'

In this example, the speaker suggests his mother-in-law (whom he can't address directly) might have an opinion different from his own. In Line (57b) he uses *aka* marking information shared with his wife. Because of affinal avoidance, the speaker has no direct evidence of his mother-in-law's wishes and so uses **kalaka** in (57c).

(57) a. *ah* EXP *Ñuku* NAME *nïgifeke* qUOT *ai* EXP *efitsu* wife *ki-lï* utter-PNCT '"The fact is Ñuku", he said to her. "What?", his wife answered.' b. *ta-iku=nile ukwi-ta-ku=aka igei*

RQ-fully=wrong dual-CONT-fully=EM21 IDEM 'Why do we have to speak so mistakenly **to each other** about this?'

c. *aŋi=kalaka* result=EM23 *ukw-oto* dual-parent *iŋu-kugu-mbo-lï* join-fully-INTR-PNCT 'I'm not sure but maybe our parent has decided to come along **by**

**herself**.' (said by Kudyu at Aifa, 1979)

### **4.2.4** *nipa* **EM24**

The morpheme marks a contradiction involving an interlocutor's or 3rd person's role treated as different from what has been experienced, said, or seen. *nipa* is often used in examples that reference the distant past, but where the speaker has some historical evidence of positive impressions of 3rd persons. For example in (58) the narrator's use of *nipa* references impressions of temporally distant others, no longer present but mentioned in historical narratives.

	- nice fully=SE=EM24 1-people

'Except that my (ancestor) was thought very beautiful when that happened.' (Kofoño to EB at Aifa, 1979)

### Ellen B. Basso

Here the complement verb, 'see, think' in imperative mood (*iŋ-ge*), is used with the main verb for 'kill'. A warrior is addressing his relative, a person with little experience:

(59) *iŋ-ge* look/think-I *elu=mbe-tsï=nipa* kill=SE-EV=EM24 *a-pi-ga* 2-club-CONT *i-feke-ne* 3-ERG-PL 'Think carefully, don't you realize (**if you did that**) they would club us to death?'

In the context of a dispute about traveling, foreign visitors keep insisting upon going in a certain direction, even though others repeatedly show them (using arm motion) the direction to travel so as to avoid enemies. Possibly *nipa* is used as the participants are foreigners with no knowledge of the local geography.

(60) a. *la!* distant.place *e-te-ke* 2-go-IMP *efu-ta=nipa* canoe-in=EM24 *e-ge-tomi* 2-travel-PURP *Atatsinu-kai* NAME-by.means.of '**Except that's** the way you should go, **different from how you want/say**, you go by canoe on the Atatsinu.' b. *afïtï,* denial *la* distant.place *u-te-ta-ni* 1-go-CONT-FUT.1 'I won't, I'm going that (other) way.' c. *i-tagiñu-pe* 3-speech-ESS *kugu,* fully *i-tagiñu-pe* 3-speech-ESS *male!* too.much 'So much of their speech, too much of that speech of theirs!' d. *la!* distant.place *e-te-ke,* 2-go-I *Atatsinu-kai=nipa* NAME-DEST=EM24 *e-te-ke,* 2-go-I *la.* distant.place 'That's the way you should go, **different from what you said**, you should go on the Atatsinu (creek), way over that way (even though you don't want to).' (said by Kambe at Aifa 1982)

### **4.2.5** *nifa* **EM25**

The speaker invites interlocutor into a context to share knowledge, or asserts a conjoint decision regarding 2nd person or 3rd person (see also example (50a).

(61) *ukwatsa-ke-nifa* dual-run-I-EM25 *ku-nitsu-na* 1+2-grandmother-ALL 'We really should hurry over to that grandmother of ours.'

### 7 Interpersonal alignments and epistemic marking in Kalapalo

The following example is a discussion of the need for a conjoint decision despite an oracular message to the contrary. In line (62b) the speaker uses *kiŋi* (contraspective negation); and *aka* (confirmation of a 2nd person). A father is asked to help prepare his young son to be a warrior who will kill ravaging jaguars. Unlike the earlier examples (45)–(48) in which the oracular message is accepted, in this case it is at first rejected but as the need to act is far more important *nifa* is used (62c) together with the politeness *muke* 'hope in vain' as the speaker, a leader, is requesting something very special of the interlocutor that will benefit the community as a whole.

(62) a. *taloki=muke* useless=EM12 *figei* ADEM *eŋï=muke-fa* do=EM12-TOP *e-mugu* 2-son *tïi-lï* make-PNCT *ku-pehe* 1+2-ERG *nïgifeke.* qUOT

> '"It may be useless but even so we should try and make your son", they said.'


'It may be useless **but even so** in order to try and have someone avenge us we need to make our son.'

d. *eh he nïgifeke.*

> agree qUOT '"All right", he answered.' (said by Kambe at Aifa, 1979)

### **4.3** *papa* **EM26**

This marker is used when the speaker agrees the interlocutor should cooperate by joining in a proposed activity. Use of papa seems also to be a politeness strategy in speech about affinal relatives, as seen in both examples.

### Ellen B. Basso

The following conjoint example involves a mother agreeing to participate in her dauughter's new relationship of marriage, as seen by the pair's reciprocal use of the affinal civility register in which special politeness kin terms (daughter referring to her husband as 'your nephew' and her mother referring to her sonin-law as 'your younger brother' are used.

	- b. *ande* here/now *a-fatuwï.* 2-nephew, *ande* here/now *a-fatuwï* 2-nephew (=daughter's husband) 'I'm thinking your nephew's here now, I'm thinking your nephew's here now.'
	- c. *eh he* agreement *nïgifeke.* qUOT '"All right", she answered.'
	- d. *ah,* EXP *e-fisï-tomi=papa* 2-younger.brother-PURP=EM.26 *ege-na* 3-ALL *e-iña...* 2-BEN 'The fact is, if you want him to come be your younger brother (i.e. new husband) on your behalf, that's all right with me...' (Afanda told by Ugaki at Aifa 1979)

In an imperative construction, the speaker uses papa to agree with his wife:

(64) *ege-ke=papa* 2-take.away-IMP=EM26 *kuigiku* manioc.soup *nïgïfeke,* qUOT *ukw-oto-iña,* dual-parent-BEN *ukw-oto-iña.* dual-parent-BEN '"I agree you should take this soup with you", he said to her, "for our parent, for our parent".' (said by Ugaki at Aifa 1979)

### **4.3.1** *apa* **EM27**

This EM appears to be informal in comparison with EM26. The speaker wants a 2 nd or 3rd person to cooperate as asked. (1 cooperates with 2 as in (65a); 1 asks 2 to cooperate as in (66–67)).

	- b. *eh he* agreement *nïgifeke* qUOT '"All right" she said to him.'
	- c. *u-ki-ta-ni=apa=fofo* I-utter-CONT-FUT.1=EM27=IMM *i-feke* 3-ERG *i-fitsu* 3-wife *ki-lï* utter-PNCT '"I'll speak to him right away **just as you wish**", his wife said.' (said by Tufule at Aifa 1979)
	- b. *kule-mi-la=ale=keñi* take.care-ADV-QN-always=beware *e-mugu-ko* 2-son-PL *undu-pesi-kïgï* attack-ugly-PFV *i-feke-ni* 3-ERG-PL

'as we can never stop watching out in case they make an attack on your son.' (said by Madyuta at Tangugu, 1979)

(67) *ah,* EXP *e-ŋi-ke=apa* 2-see-I=EM27 *ande-la-iña* here/now-NEG-DAT *its-apï-gï-ko* 3-footprint-POSS-PL *its-a,* EX-CONT *tï-fitseŋe-ki-ñï* REF-stink-INTR-SN *akago* those.people

'The fact is, you can see their footprints here just as I did, those people over there are the Stinking People whom I want you to see.' (said by Madyuta at Tangugu, 1979)

(68) *i-ŋi-gote-fa* 3-bring-CONC-TOP *ku-pehe* 1+2-ERG *safake-fi-tsïgï-tsï=apa* 3-body.trunk-crush-IPE-EV-EM27 *e-ta...,* come-CONT 'If we want to bring (some of it), "come on let's go get 'it's crushed body" like you did...' (said by Kudyu at Aifa 1979)

### **4.3.2** *tima* **EM28**

A contrastive (1≠2)used when the speaker wishes to assert a strong disclaimer of responsibility, particularly a denial of fault after being blamed by an interlocutor. The rhetorical question prefix *tï* contributes to the sense of angry negation. Context: The story concerns a maned wolf father who is looking for his son's poisoner, going from one plant to another:

	- b. *u-ikeu-te-la=muke-ti* 1-be.angry-TRNS-QN=EM12-DES *igitse* unable *e-mugu* 2-son *ё-ta* come-CONT *u-fanu-tsoke-tiGi* 1-?-chew-ADV 'I never tried to get angry when your son came to chew open my fruit.'
	- c. *tï-tomi=tima* RQ-PURP=EM28 *u-ikeu-ŋalï* 1-be.angry-MAL *i=feke* 3-ERG

'**Why should I** be **blamed** for getting angry with him?' (said by Kudyu at Aifa 1979)

Context: The warrior comes to marry a young woman, but as he is always clutching his bundle of arrows to his chest, her mother is very frightened.

	- b. *ah* EXP *tafako* bow *oto* master *ale* CUM *elei,* PDEM *kukw-e-luiŋo* 1+2-kill-FUT.2 *fegei* ADEM *i-feke.* 3-ERG 'Believe me, they're all bow masters over there, as someone like that he's going to kill us.'
	- c. *eh he* agreement *ta-tima* RQ-EM28 *afïtï* contrastive 'How can I be blamed, that's wrong.'
	- d. *ta-te-ŋalï-ko* NEG-go-MAL-PL *u-feke,* 1-ERG *u.um,* EXP *ta-te-ŋalï-ko* NEG-go-MAL-PNCT-PL *u-feke.* 1-ERG '**Why should I** kill them? I'm trying to figure out why they think I should kill them.' (said by Apihũ at Aifa 1967)

### 7 Interpersonal alignments and epistemic marking in Kalapalo

### **5 Final observations**

These many examples show how the several rich grammaticalized epistemic subsystems in the language may only be revealed in full through a discoursecentered approach that examines the natural contexts in which these occur. A researcher depending primarily on elicitation or ordinary conversation might miss many of them altogether, as I discovered over time, particularly as EM so often occur during socially discordant speech events that speakers are reluctant to invent for the foreign listener. Narrative and ritual speech frameworks include marking of changes in the intersubjective relation when new information is received, positive acceptance of a conjoint activity is requested by the interlocutor, and when acceptance or outright rejection of the interlocutor's proposition occurs. In a conversational context, there are a variety of graded ways these can be marked, from strong assertion to a weak, inferential marking and denial or disjoint marking. Politeness or impoliteness occurs in many examples, particularly where the epistemic markers are used to mark judgments regarding conjoint activities. With regard to a third person, the marker used depends on whether the speaker is making an inference of the third person's stance based on common or historical memory (the logophroic *kalaka* versus *nipa*), as well as a stance marking based on conventional sociocultural knowledge regarding the behavior of a non-human actor (marked by *nafa*). Nuckolls & Michael (2012) note the importance of understanding sociocultural factors connected to the emergence of this complex type of linguistic feature. Kalapalo EM occur in ordinary conversation, greetings, affinal civility (a register involving constrained speech and gestural modesty) disagreement and even impolite description of a proposition , and most notably in quoted conversations that form important segments of extended narrative discourse. EM contribute pragmatically in these contexts to the "unity and the experience of stance as it emerges in dialogic interaction" (Du Bois 2007: 35), and there may be more person-to-person temporal and evidential features involved. EM show types of distribution of responsibility, including requested or agreed-upon alignment involving participation in an epistemic context, as well as a speaker's denial or rejection of co-participation, and denial of alleged responsibility. Quoted speech in narrative discourse exemplifies a speaker's need to shift between a focus on the epistemic object, marking of unique or shared perspectives, and reference to the interpersonal participation in epistemically situated discourses (which may also include deliberate impoliteness, involving the speaker's refusal to use the civility register as anticipated by interlocutors). As is often the case, such contexts themselves change throughout a conversation, and there are a number of EM (especially sets C and D) that mark the nature

of these person-to-person deictic shifts. This body of Kalapalo data forces us to look at interpersonal speech practice and experience, which challenges a simple contrast between private and public, micro- and macro-sociological events. Memories of such dialogical production of meaning and the effect of these processes on social roles and relationships are preserved in the Kalapalo narrative materials discussed here, an important source for understanding the social and historical forces that have led to the emergence of this complex set of epistemic markers.

### **Phonological symbols**

Where my symbols are different from IPA conventions, the latter are placed in brackets. Alto Xingu conventions are marked (AX). *Consonants*: voiceless stops: *p*: bilabial; *t*: post-alveolar; *k*: velar; *kʷ* : dorso-velar; voiced stops: *d*: post-alveolar; *g*: velar; post-alveolar voiced stop: *dy* [d ]; prenasal voiced bilabial stop: *mb* [ b]; voiced uvular flap: *ɢ*; voiceless fricative: *f* [ɸ]; voiceless alveolar fricative:*s*; voiceless glottal fricative: *h*; voiced post-alveolar affricate: *ts*; voiced velar affricate: *č*; bilabial nasal: *m*; velar nasal: *n*; post-alveolar nasal: *ñ* [ɲ] *nh* (AX) ; velar nasal: *ŋ ng* (AX);post-alveolar lateral: *l*; bilabial semivowel: *w*; velar semivowel: *y*. Vowels: front: medial: *e*; unrounded: *i*; Central: medial: *ï* [ɨ] *ü* (AX); unrounded: *a*; Back: rounded: *u*; medial: *o* [ɤ]; stressed vowels show nasalization.

### **Abbreviations**


### 7 Interpersonal alignments and epistemic marking in Kalapalo


### **Acknowledgements**

My thanks go first to the organizers of this conference for inviting me to participate and to the participants who kindly contributed comments and suggestions regarding my presentation. Anonymous reviewers of my original draft provided many important suggestions for improving the description and analysis of these complex data, though I take full responsibility for the final version. Above all, I wish to express my unending gratitude to the three generations of Aifa and Tanguro communities in the Alto Xingu region of central Brazil. As always I acknowledge their continuing to allow me to learn from them in so many different and lasting ways.

### **References**

Aikhenvald, Alexandra Y. 2004. *Evidentiality*. Oxford: Oxford University Press.


### **Appendix**


**Table: Kalapalo Epistemic Clitics/Particles (5 pages)**



### Ellen B. Basso



### 7 Interpersonal alignments and epistemic marking in Kalapalo

# **Chapter 8**

# **Epistemic uses of the** *pretérito pluscuamperfecto* **in La Paz Spanish**

### Geraldine Quartararo

Department of Romance Studies and Classics, Stockholm University

This paper explores epistemic-evidential uses of the pluperfect, i.e. *pretérito pluscuamperfecto*, in La Paz Spanish. The *pretérito pluscuamperfecto* displays functions of a reported evidential form, conforming to results from previous studies on Argentinian Spanish (Bermúdez 2008; Speranza 2014) and, furthermore, is used according to a previously unnoticed inferential evidential function. Using theoretical frameworks from Kockelman (1957) and Bergqvist (2018), this paper describes the configuration of participant roles and event types implied in the different evidential functions of the *pretérito pluscuamperfecto*.

### **1 Introduction**

Romance languages do not possess grammaticalised evidentials and express the evidential domain through verbal inflection. In peninsular Spanish, for instance, the evidential domain is expressed by means of the simple future, the past imperfect, the present conditional and the past conditional. The simple future and the present conditional are used to express inference in the present (1) and in the past (2), respectively.

(1) Peninsular Spanish (Squartini 2001: 317; gloss added) *Ahora* now *serán* be.FUT.3PL *las* ART.F.PL *cuatro.* four 'It must be (lit. will be) 4 o'clock.'

Geraldine Quartararo. 2020. Epistemic uses of the *pretérito pluscuamperfecto* in La Paz Spanish. In Henrik Bergqvist & Seppo Kittilä (eds.), *Evidentiality, egophoricity, and engagement*, 239–262. Berlin: Language Science Press. DOI: 10.5281/zenodo.3975809

### Geraldine Quartararo

(2) Peninsular Spanish (Squartini 2001: 317; gloss added, translation modified) *Serían* be.CON.3PL *las* ART.F.PL *ocho* eight *cuando* when *salimos.* go.out-PST.1PL 'It was (lit. would be) 8 o'clock when we left.'

Whereas the imperfect (3) and the past conditional<sup>1</sup> (4) are used to express reported evidentiality.

	- a. *¿Qué tal* how *sigue* follow-textscprs.3sg *Ana?* PN 'How is Ana doing?'
	- b. *Mejor* better *me* 1SG.DAT *parece.* seem-PRS.3SG *No* not *la* 3SG.F.ACC *v-i,* see-PST.1SG *porque* because *cuando* when *llegu-é* arrive-PST.1SG *dorm-ía.* sleep-PST.IPFV.3SG *Pero* but *hab-ía* have-PST.IPFV.3SG *com-ido* eat-PTCP *algo,* something *y* and *tenía* have-PST.IPFV.3SG *menos* less *fiebre.* fever *Esta* this *noche* night *la* 3SG.F.ACC *veía* see.PST.IPFV.3SG *el* the *médico* doctor *de nuevo.* again

'I think she's better. I did not see her, because when I arrived, she was sleeping. But she had eaten something and had lower fever. Tonight, the doctor is supposedly going to see (lit. saw) her again.'

(4) Peninsular Spanish (Squartini 2001: 317; gloss added, translation modified) *Según* according.to *fuente-s* source-PL *políticas* politic-PL.F *consult-ad-as* consult-PTCP-PL.F *por* by *este* this *periódico,* newspaper *Milosevic* NP *hab-ría* have-COND.3SG *aceptado* accept-PTCP *que* that *la* ART.F.SG *fuerza de interposición* peacekeeping.force *en* in *Kosovo* Kosovo *est-é* be-SUBJ.1SG *compue-sta* form-PTCP.F *por* by *un* ART *30% de* of *efectivo-s* troop-PL *de* of *la* ART.F.SG *OTAN.* NATO

'According to political sources consulted by this newspaper, Milosevic accepted (lit. would have accepted) the Kosovo peacekeeping force to be composed of 30% NATO troops.'

<sup>1</sup>The evidential use of the past conditional is restricted to journalistic or more formal prose (Reyes 1996: 33).

### 8 Epistemic uses of PPl in La Paz Spanish

With regard to the *pretérito pluscuamperfecto*, Spanish grammars (Hernández Alonso 1986, Cartagena 1999) describe it as the tense that points out the *consecutio temporum* 'sequence of tenses' between two past actions: the more recent action is conjugated in imperfect, simple past or present perfect, while the preceding action is in *pretérito pluscuamperfecto.*

(5) Peninsular Spanish (10\_SP\_TASK: 10)

*luego* then *mi* my *la* ART.F.SG *mi* my *mujer* woman *fue* go.PST.3SG *a* to *vend-er* sell-INF *lo* 3SG.ACC *que* that *hab-ía* have-IPFV.3SG *cosech-ado* harvest-PTCP *y* and *yo* I *me* 1.REFL *fui* go.PST.1SG *a* to *sent-ar* sit-INF 'Then my wife went to sell what she had harvested and I went to sit.'

Moreover, the *Nueva Gramática de la Lengua Española* (Española 2010: 542) indicates two other uses of the *pretérito pluscuamperfecto*, such as the expression of habitual actions (6) and politeness (7).

(6) Peninsular Spanish (RAE 2010: 452; gloss added, translation added) *A* at *esa* that *hora,* hour *los* the *viernes* friday *Eugenio* NP *hab-ía* have-IPFV.3SG *sal-ido* go.out-PTCP *del* from.ART.SG *trabajo.* work

'At that time, every Friday, Eugenio went (lit. had gone) out from work.'

(7) Peninsular Spanish (Hernández Alonso 1986: 355; gloss added, translation added) *Hab-ía* have-IPFV.3SG *pens-ado* go.out-PTCP *yo* I *ped-ir-le.* ask-INF-3SG.DAT

'I was thinking (lit. had thought) to ask him/her.'

In addition to these normative uses, studies on Latin American Spanish varieties (Laprade 1981; Mendoza 1991; Callisaya Apaza 2012; Adelaar & Muysken 2004; Speranza 2014; Bermúdez 2008) have also attested evidential uses of the *pretérito pluscuamperfecto* (8).

(8) La Paz Spanish (Laprade 1981: 224; gloss added, translation modified) *Me* 1.REFL *hab-ía* have-IPFV.3SG *tra-ído* bring-PTCP *est-a* this-F *puntabola.* pen 'I (accidentally) brought (lit. had brought) this pen with me.'

### Geraldine Quartararo

The specialized literature on the evidential use of the *pretérito pluscuamperfecto* in Argentinian Spanish is limited to two studies, i.e. Bermúdez (2008) and Speranza (2014). While questioning the temporal function of the *pretérito pluscuamperfecto*, Bermúdez (2008) shows four evidential functions of this tense: an 'external source', which expresses the perspective of a third party; 'shared access to information', marking information also known by the addressee; 'endophoric source', marking information that does not come from sensory experience; and finally, 'mirative', which marks information that goes against speaker's expectations. Speranza (2014), in turn, proposes a sociolinguistic analysis of the uses of the *pretérito pluscuamperfecto* and observes a higher number of epistemicevidential uses in the varieties of Argentinian Spanish that are in contact with languages with grammaticalised evidential-epistemic systems, such as Quechua and Guaraní (Speranza 2014: 26). With respect to the expression of the commitment to the truth of information provided by the *pretérito pluscuamperfecto* as evidential-epistemic form, Bermúdez (2008) and Speranza (2014) arrive at different conclusions. Bermúdez (2008: 217) states that the use of an indirect evidential does not necessarily imply a low commitment to the truth of information:

Assigning information to an external source may mean either a weakening, or a strengthening of the reliability of the utterance, this depends on the level of reliability given to the source by the participants involved in the language exchange.<sup>2</sup>

Whereas Speranza (2014: 111) states more precisely that the use of the *pretérito pluscuamperfecto* implies the speaker's low degree of commitment to the truth of the information provided.

The appearance of the PPl (*pretérito pluscuamperfecto*) is related to utterances where there is the possibility of greater ambiguity in the attribution of what is mentioned […] the sender, then, expresses a lower degree of reliability by selecting a subordinate tense.<sup>3</sup>

I am not aware of a separate study focusing on the epistemic-evidential function of the *pretérito pluscuamperfecto* in Bolivian and Peruvian Spanish, even

<sup>2</sup>El asignar una información a una fuente externa puede significar tanto una debilitación como un fortalecimiento de la credibilidad de la afirmación, lo cual depende del nivel de credibilidad concedido a la fuente en cuestión por los participantes del intercambio lingüístico.

<sup>3</sup>La aparición del PPl se vincula a emisiones en las que existe la posibilidad de mayor ambigüedad en la atribución de los dichos […] El enunciador, entonces, expresa su menor grado de confiabilidad a través de la selección del tiempo verbal dependiente.

### 8 Epistemic uses of PPl in La Paz Spanish

though this use has been noted in the literature (Laprade 1981: 222–225; Mendoza 1991: 196–203; Callisaya Apaza 2012: 306–308; Adelaar & Muysken 2004). Laprade (1981: 223) notices that in La Paz Spanish the *pretérito pluscuamperfecto* can have mirative function or indicate absence of direct knowledge. Along this line, Mendoza (1991: 199) adds a further observation based on phonology, arguing that whenever the *pretérito pluscuamperfecto* has evidential-epistemic function in La Paz Spanish, the auxiliary verb *haber* 'to have' shows an accent shifting from *había* to *habiá*. Finally, Callisaya Apaza (2012: 307) states that epistemicevidential uses of the *pretérito pluscuamperfecto* are also found in other regions of Bolivia, although the author does not specify which ones. The contributions of the present study are three-fold. First, it details the epistemic-evidential uses of the *pretérito pluscuamperfecto* in La Paz Spanish. As already mentioned, the use of the *pretérito pluscuamperfecto* as an indirect evidential form has been already described for Argentinian Spanish (Speranza 2014; Bermúdez 2008), but its use in other varieties of Latin American Spanish and, specifically, in Bolivian Spanish has not been accounted for. The second contribution consists of further data that highlights the pragmatic functions of the form in interaction. It is argued that, in its evidential function, the *pretérito pluscuamperfecto* signals the distancing of the speaker from the propositional content of the utterance. Such a distancing, however, is not necessarily related to a low degree of commitment to the truth of the information and, in this regard, the evidential uses of this tense, i.e. inferential or reported, display different outcomes. The third contribution is to give better insights on the configuration of the pragmatic features involved in the use of *pretérito pluscuamperfecto* as evidential form. Following the theoretical framework proposed by Kockelman (1957) and Bergqvist (2018), the pragmatic features – such as event types and participant roles – involved in the different evidential functions of the form as well as their distribution are discussed. The first-hand data used in the study were collected in La Paz, Bolivia during 2014 and 2015.

In the remainder of the paper, I will refer to the *pretérito pluscuamperfecto* by the acronym PPl (cf. Speranza 2014) 4 .

### **2 Evidentials, epistemic modality, participant roles and event types**

Aikhenvald (2004: 3) argues that the core meaning of evidentiality is the expression of the source of information, but she notes epistemic extensions for both

<sup>4</sup>This paper is based on chapter 10 of my PhD thesis (Quartararo 2017).

### Geraldine Quartararo

reported (Aikhenvald 2004: 180) and inferential evidential markers (Aikhenvald 2004: 176). Such extensions are usually related to the expression of the speaker's degree of commitment to the truth of the information, i.e. epistemic modality, and are attested in languages in which the two domains, i.e. evidentiality and epistemic modality, are expressed by the same forms (Plungian 2001: 354, cf. Romance languages). The overlap between the two domains is visible in how indirect evidentials may indicate both the speaker's direct contact with the results of an event, and the lack of such results. This is the case of inferring evidentiality<sup>5</sup> (Willett 1988: 57), as in the use of the Italian future tense in (9), and reported evidentiality (Willett 1988: 57), as in the use of the Italian conditional mood in (10). In both cases, the speaker may express different degrees of reliability regarding the verisimilitude of the state of affairs due to the lack of direct evidence.


'According to Luca the train left (lit. would have left) at 5 yesterday.'

In recent years, some studies on the pragmatic properties of evidentials (Curnow 2002, 2003; Clift 2006; Faller 2006; Hengeveld & Hattnher 2015) have significantly contributed to the description of semantic extensions acquired by evidentials in specific communicative contexts. Such studies have also provided better insights into the pragmatic features involved in the use of evidentials. Hanks (2012: 172) summarizes three pragmatic dimensions that affect the use of evidentials: *source of knowledge*, i.e. the source on which the information rests; *source of statement*, i.e. the source of the utterance provided; and *expressivity/interaction force*, i.e. the "subjective relation between the speaker and some element of the utterance context" (Hanks 2012: 174). The first and third pragmatic dimensions (i.e. source of knowledge and expressivity) have been detailed in studies on evidentials, both from a typological perspective (Willett 1988; DeLancey 1997;

<sup>5</sup>Through the term "inferring" Willett indicates both inference, in Willett's terms "results", and assumption, in Willett's terms "reasoning".

### 8 Epistemic uses of PPl in La Paz Spanish

Plungian 2001; Aikhenvald 2004) and in language specific descriptions (Curnow 2002, 2003; Clift 2006; Babel 2009), but the second pragmatic dimension, *source of statement*, has received less attention in the literature. According to Hanks (2012: 174), two kinds of possible pragmatic effects belong to the *source of statement*, i.e. the *discourse modality* and the *participant roles*. *Discourse modality* refers to the perspective that speakers adopt in shaping their utterance. In this respect, Nuckolls (2012) demonstrates that, in Pastaza Quichua, the use of different evidential markers in conversational context does not necessarily indicate the access to information, but it can also specify the perspective adopted by speakers towards the information.

	- a. *Ñuka-ta* I-ACC *ña* now *kai* this *ruya-ta* tree-ACC *rikwi-i* look-IMP *chi* that *sʰapi-mi* base-EV *siri-u-n.* lie-DUR-3SG 'Look at me (up in) this tree! It's lying right at that base!' (Context: The speaking self of the narrative event (-mi sⁿ) where Luisa becomes the voice of Tito talking to his friend)
	- b. *Ni-sha-shi kapari-ni.*
		- say-COR-EVD shout-1SG

'Saying (according to my husband) I shout.' (Context: The voice of the other (-shi), where Luisa specifies the perspective of Tito who was asserting something to her)

In examples (11a) and (11b), the use of two markers, i.e. the direct evidential -*mi* and the reported evidential -*shi*, does not signal a difference in the way Luisa has gained access to information. Since in both cases Luisa heard Tito's words, it rather points out the two different perspectives from which Luisa is providing the information. In (11a), by using the evidential marker -*mi,* she provides information from Tito's perspective who *de facto* pronounced the words that she is reporting. In (11b), on the other hand, by using the evidential marker -*shi,* Luisa maintains her perspective and distances herself from Tito's words.

The change of perspective implied by the use of different evidentials, as shown for Pastaza Quichua, is essential to clarify the relevance of the second class of pragmatic effects established by Hanks (2012), i.e. *participant roles*. Drawing on Goffman's (1981) classification, speakers can be said to occupy three roles, namely, *principal*, *author* and *animator*. The *principal* is the one responsible for the utterance, i.e. the last person who committed to the information provided.

### Geraldine Quartararo

The *author* is the person who has chosen the words of the utterance in the narrated world (Goffman 1981), i.e. who has pronounced it for the first time. Finally, the *animator* is the person who physically produces the utterance. The three roles generally overlap within the same speaker, e.g. in the sentence "I am fine", the speaker is *principal* since s/he is taking responsibility for his/her own emotional and health status, as well as *author* and *animator* since s/he chooses the words of the statement and physically utters them.

Further elements of the description of the pragmatic features involved in the use of evidentials are provided by Kockelman (1957) and Bergqvist (2018). Kockelman (1957) proposes an implementation of Jakobson's (1957) classification of event types by adding a *commitment event*, and by expanding the *narrated speech event* to apply to all evidential notions, calling it *source event*. The resulting set of event types is composed by the *speech event*, the *narrated event*, the *source event* and the *commitment event*. The *speech event* corresponds to the world in which the utterance is made. The *source event* corresponds to the "spoken-about world in which speaking occurs" (Kockelman 1957: 128), and may be distinguished according to the type of contact that a speaker has with a source (Kockelman 1957: 143). The *narrated event* indicates the world described in the utterance. Finally, the *commitment event* is the world where the speaker commits to the truth of the proposition expressed (Kockelman 1957: 127). In addition to this proposal, Kockelman (1957) also establishes a correlation between Goffman's (1981) participant roles, i.e. *animator*, *author* and *principal*, and the new set of event types, i.e. the *speech event*, the *narrated event* and the *commitment event*. Within this framework, Bergqvist (2018) formulates another relation that connects *source event* (Kockelman 1957: 128) with a new participant role defined as *cognizer* (Bergqvist 2018: 22), i.e. the person who perceives the event. This set of correspondences is illustrated in 8.1.


Table 8.1: Correspondence between event types and participant roles

If one takes the model of correspondences shown in 8.1, and applies this to Example (11), above, it becomes possible to provide an analysis of the pragmatic

### 8 Epistemic uses of PPl in La Paz Spanish

features relevant to the use of evidentials. In (11a), Luisa produces the utterance as if Tito was pronouncing it. This strategy results in two series of consequences in the configuration of the correlation between event types and participant roles: first, by reproducing Tito's voice, Luisa creates an artificial overlap between the participant roles of the two speakers, playing simultaneously the *animator* (Luisa is indeed the last who pronounced the utterance), the *author* (Tito has chosen the words of the information), the *cognizer* (Tito witnessed the event) and the *principal* (Tito committed to the truth of his statement); second, by impersonating Tito's voice, Luisa fictitiously matches the world in which she is pronouncing the utterance with the world in which Tito pronounced the utterance, i.e. the *speech event* overlaps with the *source event*, since they fictitiously occur in the same world. Given the use of the direct evidential -*mi,* the *commitment event* coincides with the *source event*. Finally, considering that the *narrated event* does not make any reference to the world in which the *speech event* occurs (e.g. in "I promise"), it will be kept separate from the others. In (11b), the configuration of event types and participant roles is different. By using the reported evidential -*shi*, instead, Luisa specifies the separation between the *speech event* and the *source event*, the *narrated event* is kept distinguished from the previous two event types and, finally, the configuration of the *commitment event*, as for the correlated participant role, cannot be established.

### **3 Material, participants and method**

Thirty Spanish-Aymara bilingual speakers participated in the study (17 males and 13 females, age range: 18–64). All participants first learned Aymara and then acquired Spanish during their childhood. The L2 proficiency in the standard variety of La Paz Spanish varied among the speakers depending on age and level of education. About 60% of the speakers had university level education, 26,6% had secondary education and 13,4% had primary education.

The data was collected mainly in La Paz and El Alto (Bolivia). The corpus consists of fully transcribed recordings lasting 8 hours and 24 minutes in total. The transcription convention employed (Briz & Universidad de Valencia 2000) has also been used for the transcription of colloquial Spanish corpora and allows for a faithful representation of speech.

The corpus is divided into three parts: the first and largest part consists of the recordings of the "Family Problems Picture Task" (San Roque et al. 2012), the second part consists of five recordings of the task "The Pear Story" (Chafe 1980), and the third part consists of four recordings of personal narratives.

### Geraldine Quartararo

The "Family Problems Picture Task" (FPPT) was created to activate the use of cognitive categories such as evidentiality and mirativity (San Roque et al. 2012: 140). Its two-fold nature of problem-solving and interactive task allows the activation of inferential processes and, therefore, supports the analysis of the use of evidentials in interactive settings. The task consists of 16 pictures in black and white that follow a defined order. The temporal sequence and content of the pictures are not always clear. Inferential processes are required to understand the order and development of the story.

The FPPT was developed in five steps: in the first step, speakers were asked to describe five of the sixteen pictures randomly selected by the fieldworker; in the second step, speakers ordered all the pictures according to the story that they believed it was represented; in the third step, one of the two participants in the first two steps was asked to describe the story built in the first person singular; in the fourth step, the other participant was asked to tell the story in the third singular person to a third person who did not participate in the task until then; finally, during the fifth step, the third participant was asked to tell the story s/he had been told. In order to facilitate data analysis, the internal organization of the transcriptions follows the same structure of the FPPT, i.e. each transcription is divided into five parts.

### **4 The epistemic-evidential functions of the PPl**

The corpus features 78 tokens of the PPl. The analysis reveals that in most of the cases, 68%, the PPl is used according to its temporal function (see example 5), i.e. it indicates the temporal relation between two past actions; nevertheless, in a significant number of cases, 32%, the PPl seems to operate as an epistemicevidential form, i.e. it specifies the epistemic relation between the speaker and the event. When the PPl is used as an epistemic-evidential form, it may display inferential evidence, reported evidence or mirativity. As an inferential evidential, the PPl signals inference based on observable evidence (13 cases). As a reported evidential, it signals second-hand report, i.e. the speaker has directly heard the words of someone else (9 cases). Finally, as mirative form, it indicates surprise (2 cases). Table 8.2 summarizes this.

In the corpus, over 90% of the occurrences of the PPl (72 cases out of 78) comes from the transcriptions of the FPPT; the remaining cases come from the transcriptions of the personal narratives. By analyzing the distribution of these 72 cases among the steps of the FPPT, it turns out that the PPl appears 8 times during the first step, 5 times during the second step, 24 times during the third step, 22 times

### 8 Epistemic uses of PPl in La Paz Spanish


Table 8.2: Function of the PPl in the data

during the fourth step, and 13 times during the fifth step. By further narrowing down this analysis to the cases in which the PPl seems to operate as an epistemicevidential forms, it is notable that all cases of PPl with evidential-epistemic functions occur within the transcriptions of the FPPT. Secondly, no case of inferences based on observable results occurs in the third (i.e. narrative in the first person) and the fifth step of the task (i.e. report of the story to a third participant). Lastly, the largest number of cases of the second-hand reported function appears in the fifth step. This distribution is presented in Table 8.3.

Table 8.3: Evidential functions of the PPl (columns) for the steps of the FPPT (rows)


### **4.1 Mirative or inferential function?**

The debate on the relation between evidentiality and mirativity is still an open one. On the one hand, some scholars discuss the nature of the relation between the two domains: DeLancey (1997) and Aikhenvald (2004) consider them as separate, while Lazard (1999) and de Haan (2012) consider them to be related. On the

### Geraldine Quartararo

other hand, recent studies (Hill 2012) entirely reject the description of mirativity as an independent category, arguing that direct evidence ("sensory evidence" in Hill's terms) is an adequate category to account for most of the markers described as miratives in the literature.

Taking into account cases in which the use of the PPl is related to the expression of a direct, visual contact with the discourse object, the data shows three situations: (i) the PPl can simultaneously signal both an inferential and a mirative function (example 12), (ii) the PPl can signal only inference (example 13) and, (iii) the PPl can signal only mirativity (example 14).

In a few cases, it is not possible to establish a clear distinction, or a hierarchy between the inferential (Willett 1988) and the mirative function of the PPl; the two functions, indeed, seem to co-exist and overlap within the same form, i.e. the PPl [cf. 12].

(12) La Paz Spanish (5\_SP\_TASK: 6)


```
c. ...
```

```
...
...
```
d. *aaaa* INTERJ *hab-ía* have-PST.IPFV.3SG *sido* be.PTCP *agricultor* farmer *el* the *cuate* guy 'Aaaa he is (lit. had been) a farmer, the guy'

```
e. est-án
   be-PRS.3PL
              cosech-ando
              harvest-GER
                            no
                            no
   'They are harvesting, aren't they?'
```
f. *est-án* be-PRS.3PL *cosech-ando* harvest-GER *sí* yes *el* the *campo* countryside *es* be.PRS.3SG 'They are harvesting, yes, it is in the countryside'

In (12), the speakers are describing picture (e) in Figure 8.1, which is the fifth image shown to participants during the first step of the FPPT, showing a man and a woman that are picking pumpkins in a garden. The remaining four pictures, previously shown to the participants during this stage of the task, depict the man

### 8 Epistemic uses of PPl in La Paz Spanish

Figure 8.1: Pictures used during the first step of FPPT

in a cell (picture 8.1a), drinking alcohol (picture 8.1b), hitting his wife (picture 8.1c), and standing in a courtroom (picture 8.1d).

From observing the first four pictures, speakers are not expected to be able to guess the man's profession. Picture 8.1e appears to present new and partly surprising information, indicated by rising intonation and the interjection "aaaa" (see Example 12, above). Although a mirative function is implied by this specific use of the PPl, it is clear that an ongoing inferential process is at the foundation of the information provided. There is no doubt, that in (12) the sentence where the PPl occurs is an inference, given the fact that no pictures in the task clearly show that the man's profession is farming. A further instance of this evidential function of the PPl is found in (13).

	- a. *aquí* here *qué* what *est-á* be-PRS.3SG *hac-iendo* do-GER *este* this *señor* man *ya* already *le* 3SG.DAT *empiez-a* start.PRS.3SG *a* to *cont-ar* tell-INF *ha* have.PRS.3SG *deb-ido* must-PTCP *est-ar* be-INF *lejos* far.away *trabaj-ando* work-GER *este* this *señor* man *tal vez* maybe *le* 3SG.DAT *empiez-a* start-PRS.3SG *a* to *cont-ar* tell-INF *su* 3.POS *señora* wife *a* to *su* 3.POS *hijo* son *todo* all *el* the *suceso* happening *como* how *hac-ía* do-PST.IPFV.3SG *como* how *trabaj-aba* work-PST.IPFV.3SG *no* no

'Here, what is this man doing? Aaa now he starts to tell. He must be far away, this man, maybe. He starts to tell his wife and his son everything happened. How he did, how he worked, no?'


3POS-PL pumpkin-PL

'And here they start to work, it must be the day after or later, no? both work, they are picking pumpkins'


8 Epistemic uses of PPl in La Paz Spanish

*vend-er* sell-INF *a* to *la* ART.F.SG *feria* market *allí* there *es* be.PRS.3SG *con* with *su* 3.POS *hij-ito* son-DIM *es* be.PRS.3SG *más* more *pequeño* small 'Those pumpkins that they picked. They are carrying to the market. There he is with his little son, he is younger' (glossed) h. *más* more *pequeñ-ito* small-DIM *Yola* Yola.PN 'Younger Yola' (glossed) i. *hab-ía ten-ido dos hijo-s.*

have-PST.IPFV.3SG have-PTCP two child-PL 'He must have (lit. had had) two children' (glossed)


Example (13) is an extract from the second stage of the FPPT. Here, the PPl is the main verb of the utterance *había tenido dos hijos* 's/he must have had two children'. The speakers placed the pictures of the story in the order shown in Figure 8.2.

In picture 8.2a, a man sits talking to a woman and a boy. In picture 8.2b, the same man is picking pumpkins in a garden with a woman. Finally, in picture 8.2c, the man, the woman and a small child are walking together down a road, carrying two baskets full of pumpkins. After putting in order the three pictures, the speakers imagine that the actions depicted in them are performed in a few days, *debe ser al día siguiente o más tarde* 'it must be the day after or later'. Furthermore, by comparing picture 8.2a to picture 8.2c, they cannot help but notice the presence of two children with different ages. This visually available evidence produces the inference made by B that the couple has two children (*había tenido dos hijos*).

If in the previous cases (example 12 and 13) the use of the PPl is related to an ongoing inferential process, in two cases, the PPl seems to operate exclusively as a mirative form indicating the surprise of the speaker with respect to something drawn in the pictures.

### Geraldine Quartararo

Figure 8.2: Pictures used during the second step of FPPT

(14) La Paz Spanish (5\_SP\_TASK: 5) *uuu* INTERJ *qué* what *pas-a* happen-PRS.3SG *aquí* here *a* to *su* his *mujer* wife *le* 3SG.DAT *hab-ía* have-PST.IPFV.3SG *peg-ado* hit-PTCP *ese* that *maricón* wimp 'Uuu, What's happening here? That wimp has hit (lit. had hit) his wife!'

In Example (14), taken from the first stage of the task, the speaker is describing what is drawn in picture 8.1c. The use of the PPl, in this case, does not signal an inference, nor is it possible to consider this use of the form as related to other documented uses of the PPl in Spanish such as the expression of *consecutio temporum*, politeness or habitual aspect. Given the linguistic elements that co-occur with the PPl in Example (14), i.e. the interjection "uuu", the appellative *ese maricón* 'that wimp', and the exclamatory form of the utterance, the function of the PPl aligns better with the speaker's (negative) surprise of the man hitting the woman in the picture. This use of the PPl could therefore be said to be an instance of the mirative function, also conforming to mirative uses of this tense as noticed for Argentinian Spanish (Bermúdez 2008).

8 Epistemic uses of PPl in La Paz Spanish

### **4.2 Secondhand reported evidential function**

As a reported evidential, the PPl always signals second-hand reports, meaning that the speaker reports something that s/he has heard directly from the *author* of the utterance.

According to Spanish grammars (Maldonado 1999), indirect speech is constructed through a conjugated reporting verb followed by the complementizer *que* 'that' and a subordinate clause, whose verb is conjugated according to specific tense agreement rules. If the reporting verb is in the present tense, then the verb of the subordinate clause will also be in the present tense, the simple past/present perfect/imperfect, or in the future tense. In the subordinate clause, the use of one tense rather than another depends on the original tense of the verb of the reported utterance.


'He says that some relatives went to leave him clothes'

Examples (15) and (16) are from the fourth and the fifth stage of the FPPT, exemplify the change from direct to indirect speech in Spanish. In (16), speaker A is reporting to the fieldworker what speaker B told him during the previous stage of the task (15).

The data contains few examples of the PPl as a reported evidential. In such cases, the reporting verb is in present tense, as in (18).

(17) La Paz Spanish (10\_SP\_TASK: 10), speaker C *Dos* two *person-as* person-PL *van* go.PRS.3SG *trabaj-ando* work-GER */ una* a *pareja* couple 'two people are working, a couple'

### Geraldine Quartararo

(18) La Paz Spanish (10\_SP\_TASK: 10), speaker D *ee bien* well *dic-e* say-PRS.3SG *¿no?* no *un* a *día* day *hab-ía* have-PST.IPFV.3SG *hab-ido* have-PTCP *una* a *pareja* couple 'Well he says, doesn't he? one day there was (lit. had been) a couple'

Examples (17) and (18) present a similar situation to the one already discussed for Examples (15) and (16). Example (18) is a reported representation of what said by the speaker C in (17). However, unlike Example (16), the use of the PPl in Example (18) cannot be analyzed in terms of tense agreement, which is clearly violated, but rather responds to the need of the speaker to distance her/himself from the reported utterance. This distancing is linguistically expressed by a removal in time of the reported utterance and the greater temporal distance between the verb of speaking, *dice* 's/he says', and the verb of subordinate clause, *había habido* 'had had'.

The use of two different tenses in the reported speech of (16) and (18), i.e. the simple past and the PPl, respectively, depends on the epistemic-evidential function of the PPl. In both examples, the presence of the reporting verb *decir* 'to say' makes explicit that the information provided comes from another speaker and that there is a subsequent epistemic distance between the speaker and the source of information. By using the simple past (16), the speaker B does not add any further pragmatic information to the story and presents it as a mere outcome of a report. In contrast, the PPl in example (18) creates a greater distance between the speaker and the information provided. This allows speaker D to (i) signal that the story provided is a report, and (ii) maintain her/his own perspective by specifying that what s/he is reporting does not represent her/his own words nor her/his view of the story. In other words, the use of the PPl as the main verb of the subordinate clause in (18) indicates that the speaker D does not want to commit to the story told by the speaker in example (17).

### **4.3 Event types and participant roles in the evidential uses of the PPl**

The analysis of the configuration of the pragmatic features involved in the use of the PPl as an evidential produces two separate outcomes depending on the evidential function expressed by the PPl.

When the PPl signals inference, as in Example (13), the participant roles of *animator*, *author* and *cognizer* are placed with the speaker, since s/he (i) pronounces the utterance in the real word, (ii) choses the words of the utterance and (iii) makes the inference. The role of *principal* in Example (13), needs some further discussion, however. Despite what is generally stated in the literature on

### 8 Epistemic uses of PPl in La Paz Spanish

the evidential use of the PPl (Speranza 2014), in the data from La Paz Spanish there are no clear instances in support of the hypothesis that the inferential use of the PPl encodes the speaker's low, or high commitment to the truth of the information, i.e. there are no instances of additional linguistic elements, e.g. *tal vez* "maybe" or *ciertamente* "certainly", that indicate the epistemic stance of the speaker towards the information provided. This observation aligns with what Cornillie (2009) states with respect to evidentials in Italian (examples 9 and 10) that the inferring function of the simple future in Italian is not strictly tied to the expression of commitment to the truth of information. Likewise, the inferential use of the PPl does not appear to signal degree of commitment, but rather specifies the presence of an intermediary step, i.e. a cognitive process, in the acquisition of information by the speaker. Consequently, the use of the PPl as an inferential evidential does not specify the participant role of *principal* and the form could thus be considered as epistemically neutral. With respect to the configuration of event types, in Example (13) there is a clear distinction between the *narrated event* and the *speech event*, since the action described in the utterance refers to a world that is not related to the one in which the *speech event* took place. The *source event*, instead, seems to coincide with the *speech event*; this overlap is due to the fact that the process that leads the speakers to state their inference is simultaneous to the pronunciation of the utterance. Finally, the configuration of the *commitment event* – as already mentioned for the related participant role, i.e. *principal –* does not seem to be specified in the inferential use of the form.

With regard to the evidential second-hand reported function of the PPl (Example 18), as already mentioned, the difference determined by the use of the PPl in Example (18) and the simple past in Example (16) lies in the different stance from which the speakers produce their narratives. In (16), by using the simple past, the speaker reports the story without adding an epistemic qualification. In (18), by contrast, the speaker adds epistemic information to the utterance by using the PPl. Such temporal distancing, allows the speaker to reduce her/his commitment to the truth of the proposition. Regarding the configuration of participant roles, it is relevant to notice that Examples (16) and (18) show different configurations. In (16), the role of*cognizer* coincides with the role of *animator*, while *author* remains separate and *principal* is unspecified. In (18), on the other hand, the configuration of the participant roles of *animator*, *cognizer* and *author* is the same as in (16), but the participant role of *principal* is present and coincides with that of *author*. With respect to the configuration of the event types in Examples (16) and (18), it is important to note that in reported speech, the *speech event* and the *source event* are always separated. In Examples (16) and (18) the *narrated event* is also located separately. The main distinction between the two examples, therefore, relates to the

### Geraldine Quartararo

specification of the *commitment event*. While the use of the canonical reported speech in Example (16) does not imply the speaker's commitment, the use of the PPl in Example (18) features a low degree of commitment by the speaker. Table (8.4) summarizes the configuration of the event types discussed for the evidential function of the PPl.



### **5 Conclusions**

On the basis of first-hand data from La Paz Spanish, this study details the uses of the PPl and demonstrates that in La Paz Spanish, beyond its normative uses, this tense is also used to express epistemic-evidential functions. In the existing literature on the Latin American varieties of Spanish, the PPl has been described as a tense that can serve as mirative, reported speech, and endophoric evidentials. The present paper provides new findings and demonstrates that in La Paz Spanish the PPl is also used accordingly as an inferential evidential that has been previously unnoticed. The analysis also reveals that the form can convey more than one epistemic-evidential function simultaneously, meaning that in these cases, it is actually not possible to establish a sharp distinction between the evidential function and the mirative function, since both seem to play an important role in this use of the form. It is important to note, moreover, that in the data both the mirative and the inferential uses of the PPl are strictly related to visual access of the source of information. This last statement, to a certain extent, supports Hill's (2012) analysis of the mirative forms as markers related to sensory contact with the source event. However, given the PPl's attested epistemic-evidential functions, it is not appropriate to consider it as a "sensory evidential" (Hill 2012), but more like the Turkish -*mIş*, i.e. an instance of a "mediative" (Lazard 1999) form.

### 8 Epistemic uses of PPl in La Paz Spanish

A further consideration concerns the speaker's commitment to the truth of the proposition marked by an evidential. In its inferential use, the PPl does not signal the speaker's commitment to the truth of the proposition and in these cases, it only expresses inference without any further epistemic connotations. By contrast, its evidential use as a (second-hand) reportative evidential, is related to the expression of a lower commitment to the truth of the information provided. I believe that this difference is basically due to the type of contact that the speaker has with the source. In the first case, the speaker has visual contact with the source that activates an inferential process based on the speaker's own logic and interpretation; in the second case, the contact with source is mediated and the speaker is aware of telling the story that another individual has formulated and whose accuracy s/he cannot confirm.

A final consideration is related to the absence of examples of the PPl expressing other evidential functions, such as inferential reasoning, third-hand report, and folklore (see Aikhenvald 2004). This absence could depend either on the nature of the form that does not convey all the indirect evidential functions or on the nature of the materials used to elicit the evidential forms used in this study. In relation to this second possibility, I believe that two elements of the FPPT may have influenced my results: (i) the predominant role played by the visual contact in the development of the first four stages and (ii) the fifth stage producing mainly second-hand reported speech. Although these two elements do not preclude the use of the PPl with other epistemic-evidential functions in the whole corpus, they certainly favor certain uses rather than others. More studies of first-hand data are needed in order to improve our understanding of the epistemic-evidential uses of the PPl in both La Paz Spanish and other varieties of Latin American Spanish.

### **Abbreviations**



### **References**


8 Epistemic uses of PPl in La Paz Spanish


# **Chapter 9**

# **Same same but different: On the relationship between egophoricity and evidentiality**

### Manuel Widmer

University of Zurich

Egophoricity (a.k.a. "conjunct/disjunct") is a grammatical phenomenon whose grammatical status generated heated discussions in recent years. While some scholars have analyzed egophoricity as a subcategory of the well-established grammatical category of *evidentiality*, others have treated egophoricity as an independent grammatical category. This study aims at assessing the relationship between egophoricity and evidentiality from a functional-typological perspective. The chapter first discusses the varying structural complexity of egophoric systems against the backdrop of typological models that treat egophoricity as an evidential subcategory (e.g. Plungian 2010; San Roque & Loughnane 2012). It is argued that such models fare well with complex egophoric systems of the Lhasa Tibetan type (see Tournadre & Dorje 2003) but fall short of accommodating binary egophoric systems of the Kathmandu Newar type (see Hargreaves 2005). In a second step, the chapter takes a closer look at systems of the Lhasa Tibetan type, arguing that there are Lhasa Tibetan-type systems in which egophoric and allophoric markers display a considerable degree of independence, both from a functional and a structural point of view. These observations provide substantial evidence for the claim that egophoricity constitutes a functional domain distinct from evidentiality.

### **1 Introduction**

Egophoricity (a.k.a. conjunct/disjunct) is a cross-linguistically rare phenomenon that has so far only been documented for a comparatively small number of languages spoken in the Himalayas, the Caucasus, South America and Papua New

Manuel Widmer. 2020. Same same but different: On the relationship between egophoricity and evidentiality. In Henrik Bergqvist & Seppo Kittilä (eds.), *Evidentiality, egophoricity, and engagement*, 263–287. Berlin: Language Science Press. DOI: 10.5281/ zenodo.3975811

### Manuel Widmer

Guinea (cf. Creissels 2008). The grammatical status of egophoricity is a controversial issue that generated much discussion in recent years. While some scholars have analyzed egophoricity as a subcategory of the well-established grammatical category of *evidentiality*, which serves the primary function of marking one's source of information (Aikhenvald 2004), others have maintained that egophoricity has a different functional motivation and should be considered as a grammatical category in its own right. This paper contrasts these two approaches, arguing that egophoricity and evidentiality are best conceived of as two independent grammatical categories but still interact with each other in various ways because they share a conceptually related functional motivation.

The chapter is structured as follows: §2 discusses some terminological issues and gives a brief overview of different descriptive approaches towards egophoric systems. §3 focuses on the distinction between egophoric systems consisting of two markers and egophoric systems comprising three or more markers. It is argued that binary systems cannot easily be reconciled with existing typologies of evidentiality, which suggests that egophoricity may in fact constitute a grammatical category distinct from evidentiality. §4 further develops this argument by demonstrating that there are languages in which egophoricity and evidentiality manifest themselves as structurally independent subsystems. §5 summarizes the results of this study and highlights some directions for further research.

### **2 Preliminaries**

### **2.1 Terminology**

What is referred to as *egophoricity* in this chapter has been described under different names in the past. Initially, the phenomenon was known under the name "conjunct/disjunct", a term that was introduced by Hale & Watters (1973) and subsequently popularized by Hale (1980) and DeLancey (1990). Since the 1990s, various scholars have proposed a range of additional terms, either to refer to the phenomenon itself, its subcategories, or both. Table 9.1 below gives an overview of some terminological proposals (see San Roque et al. 2018 for a detailed overview).

The designation *egophoricity*, which is derived from Tournadre's (1991) term *egophoric*, has gained considerable acceptance in the course of the past decade and is now the most widely used term. The term "heterophoric", which was introduced as an antonym of the term *egophoric* by Tournadre (1991), never gained much currency in the literature. This is most probably due to the fact that Tournadre himself abandoned the term "heterophoric" early on when he gave up his dichotomous analysis of the Lhasa Tibetan egophoric system (cf. Tournadre 2008:

### 9 Relationship between egophoricity and evidentiality


Table 9.1: Selected terminological approaches to egophoricity

301, fn. 48). However, various scholars have subsequently reintroduced antonyms to describe the antonymic value of *egophoric* in binary egophoric systems. These antonyms are "non-egophoric" (San Roque et al. 2018), "alterphoric" (Post 2013), and "allophoric" (Widmer 2017). In what follows, the terms *egophoric* and *allophoric* are used to refer to the two subcategories of egophoricity.<sup>1</sup>

It is generally assumed that egophoric systems revolve around an epistemic role that comprises the different speech act roles to which the relevant markers may relate in different pragmatic contexts (see below for discussion). This epistemic role has been given a range of different names such as "locutor" (Curnow 1997), "epistemic source" (Hargreaves 1991, Hargreaves 2005), "informant" (Bickel & Nichols 2007), or "assertor" (Creissels 2008). The present chapter follows Creissels (2008) in using the term *assertor*.

<sup>1</sup>Note that the terms *egophoricity*, *egophoric*, and *allophoric* were already used by Dahl (2000), albeit in a different sense. Dahl does not use the term *egophoricity* to refer to an epistemic grammatical phenomenon but to distinguish between two types of reference: *egophoric reference* and *allophoric reference*. The first includes reference to speech-act participants, generic reference, and logophoric reference, whereas the latter comprises all other types of reference.

### Manuel Widmer

### **2.2 Approaches to egophoricity**

In the course of the past decades, scholars have proposed different analyses of egophoricity. In early studies (e.g. Hale & Watters 1973; Hale 1980), egophoricity was analyzed as a peculiar type of person agreement, an approach that has occasionally been invoked by recent typological work (e.g. Aikhenvald 2004; Bickel & Nichols 2007). Person-based analyses have in common that they focus on the characteristic distribution of egophoric / allophoric markers in declarative and interrogative contexts (see Table 9.2 below), which they contrast with the distribution of person endings in ordinary person indexation systems.

Table 9.2: The prototypical distribution of egophoric / allophoric markers across clause types


Whether or not it is feasible to treat egophoricity as a manifestation of the grammatical category of *person* essentially hinges on how one defines the concept *person*. A traditional notion of person that is solely based on the roles of *speaker*, *addressee*, and *other* is not particularly useful for describing the distribution of egophoric and allophoric markers. However, within the framework of a broader definition that conceives of *person* as "the grammaticalization of conceptual distinctions between participants involved in speech activities" (Bickel & Nichols 2007: 220), it is possible to establish a link between person agreement and egophoricity. Under such an approach, egophoric systems can be analyzed as person agreement systems that make a binary distinction between the assertor, viz. the speech act participant against whose knowledge a proposition is profiled, and everybody else (see Bickel 2008; Bickel & Nichols 2007: 223; Creissels 2008). However, this comes at the cost of blurring the fundamental distinction between syntactic agreement (determined by syntactic roles of arguments) and pragmatic agreement (determined by the identity of speech act participants and the type of speech act).

Person-based approaches towards egophoricity offer valuable insights into the different ways in which the participants of speech acts may be conceptualized across languages. At the same time, scholars working in such frameworks have often interpreted the assertor-based conceptualization of speech acts as the

### 9 Relationship between egophoricity and evidentiality

core function of egophoric systems. In other words, they have taken agreement with the assertor as the primary function of egophoricity (see Bickel & Nichols 2007; Creissels 2008). This generalization is problematic for two reasons. First, a person-based analysis may work well for languages that have highly syntacticized egophoric systems, but it runs into difficulties when dealing with languages that have egophoric systems that are based on pragmatic rather than syntactic principles (cf. San Roque et al. 2018). Second, there is some evidence that egophoric markers may not be invariably tied to the role of the assertor. One piece of evidence for this claim comes from the Chibchan languages Ika and Kogi. For both languages, Bergqvist (2012, 2016) has described verbal affixes that are functionally reminiscent of egophoric / allophoric markers but appear to be sensitive to the speech act roles of *speaker* and *addressee* rather than the epistemic role of *assertor*. Another piece of evidence comes from the Tibeto-Burman language Bunan. Bunan possesses the volitive suffixes -*te* (SG) / ℎ (PL), which indicate one's intention to perform an action (Widmer 2017: 555–556). These suffixes are clearly modal in nature and accordingly have a different functional motivation than egophoric markers (see below for a discussion of this aspect). However, they still revolve around the epistemic role of the assertor (Widmer 2017: 453– 459). These facts suggest that the epistemic role of the assertor is not a part of the functional definition of egophoricity but rather represents a distinct grammatical phenomenon that can occur independently of egophoric systems.

In sum, person-based analyses of egophoric systems are rarely propagated in the literature nowadays, most probably due to the analytical challenges outlined above. In contemporary studies, egophoricity is commonly analyzed as an epistemic phenomenon, that is to say, a grammatical means to relate the knowledge that is conveyed in a given speech act to speech act participants. These epistemic approaches can be broadly classified into two types: (i) those that conceive of egophoricity as a subcategory of evidentiality in the sense of Aikhenvald (2004), i.e. as having "source of information" as their primary meaning, and (ii) those that treat egophoricity as an epistemic category with a different functional motivation.

Approaches that treat egophoricity as a subcategory of evidentiality in a more or less explicit manner are encountered both in cross-linguistic and languagespecific studies. Typological studies that are relevant here are Plungian (2010) and San Roque & Loughnane (2012). Plungian (2010: 34) postulates an evidential subcategory "participatory evidence", which indicates that a speaker knows about an event because she / he was directly involved in it.<sup>2</sup> Although this definition is reminiscent of the functional properties of egophoric markers, Plungian

<sup>2</sup>Note that Plungian (2010) adopts the term *participatory* evidence from Oswalt (1986).

### Manuel Widmer

does not identify participatory evidence with egophoricity (which he refers to as "conjunct/disjunct"). Rather, he states that egophoric systems are "not so much related to the grammatical expression of evidentiality but to the expression of person (i.e. that of the speaker)" (2010: 43), thus suggesting that evidentiality and egophoricity are distinct categories. At the same time, he notes that the Trans-New Guinea language Oksapmin possesses participatory evidentials (2010: 34). Indeed, Loughnane (2009) describes a "participatory-factual" or "personalfactual" evidential category for this language. However, Loughnane (2009: 253) also explicitly compares this evidential subcategory to egophoric marking, stating that "[t]he conjunct [i.e. egophoric] term in conjunct/disjunct systems [i.e. egophoric systems] appears, at least in some languages, to be a personal evidential." Accordingly, there is a link between Plungian's (2010) notion of "participatory evidence" on the one hand and egophoric marking on the other, even though Plungian does not make this explicit.

In their survey of evidential systems in languages of the New Guinea Highlands, San Roque & Loughnane (2012) follow Plungian (2010) in postulating an evidential subcategory "participatory evidence", which is used to mark events that the speaker performed herself / himself or which are generally known to her / him. Unlike Plungian (2010), San Roque & Loughnane explicitly link this category to egophoricity (which they call "conjunct/disjunct"), stating that the languages of the New Guinea Highlands "highlight the relevance of participatory and visual-sensory evidentials to conjunct/disjunct systems, and suggest that in certain cases "conjunct" and "disjunct" terms can be analyzed as participatory and visual(-sensory) evidentials, respectively" (2012: 158).

Like Plungian (2010) and San Roque & Loughnane (2012), the recent typologically oriented study by Tournadre & LaPolla (2014) treats egophoricity as an evidential subcategory, which they refer to as "egophoric" or "personal". However, unlike the above cited authors, Tournadre & LaPolla (2014) additionally extend Aikhenvald's (2004) original definition of evidentiality from simply marking one's "source of information" to marking both one's "source and access to information".

Some recent language-specific studies that analyze egophoricity as an evidential subcategory are Loughnane (2009), whose work has already been mentioned above, as well as Norcliffe (2018) and Floyd (2018), who describe egophoric markers as a part of complex evidential systems of the Barbacoan languages Namtrik/ Guambiano and Cha'palaa, respectively.

Approaches that treat egophoricity as an epistemic category with a functional motivation different from "source of knowledge" have so far predominantly been

### 9 Relationship between egophoricity and evidentiality

applied in the description of individual languages, in particular languages spoken in the Greater Himalayan region. Table 9.3 gives an overview of these approaches.<sup>3</sup>

As the table illustrates, most of the above cited approaches characterize the functional motivation of egophoricity in terms of a binary contrast between two different types of knowledge, viz. privileged / personal / old / internal knowledge vs. non-privileged / objective / new / external knowledge. While individual approaches focus on different functional aspects of this dichotomy, it is clear that they essentially all describe the same phenomenon, viz. an epistemic grammatical category that is conceptually related yet functionally distinct from evidentiality (in the sense of Aikhenvald 2004) and distinguishes between two types of perspectives on knowledge: an epistemically privileged and epistemically nonprivileged perspective.

### **3 Different manifestations of egophoricity**

In the languages of the world, we encounter two different structural manifestations of egophoric systems. First, there are binary egophoric systems, which instantiate a dichotomic contrast between an egophoric and an allophoric marker. Such a system can be found in Kathmandu Newar (Tibeto-Burman, Nepal). The endings of the Kathmandu Newar egophoric system are illustrated in Table 9.4 below.

Second, there are egophoric systems in which egophoric markers directly contrast with two or more evidential markers. Such a system has been described for Lhasa Tibetan, whose imperfective and perfective endings are illustrated in Table 9.5 below.<sup>4</sup>

It is this variability in the structural organisation of egophoric systems that has given rise to controversies about the grammatical status of egophoricity and its relationship to evidentiality (cf. Gawne & Hill 2017: 295–296). The interpretation

<sup>3</sup> It is important to note that these descriptive models differ strongly from each other in terms of terminology and the grammatical status that is assigned to egophoricity. Accordingly, it would be misleading to think of these different models as a unified approach. However, it is still justified to group them together since they all define egophoricity in a way that is not compatible with Aikhenvald's (2004) definition of evidentiality.

<sup>4</sup>The egophoric markers -*pa yin* and -*byung* are tied to different participant roles. The first expresses an epistemically privileged perspective in combination with agents, while the latter expresses an epistemic privileged perspective in combination with undergoers (Widmer & Zúñiga 2017). The relevance of participant roles for egophoricity is not discussed in the detail in the present chapter but briefly touched upon in § 6.

### Manuel Widmer


Table 9.3: Non-evidential epistemic approaches towards egophoricity

### 9 Relationship between egophoricity and evidentiality


Table 9.4: The Kathmandu Newar system (Hargreaves 2005)

Table 9.5: The Lhasa Tibetan system (DeLancey 1990)


of egophoric markers as evidentials is sensible and practical from the perspective of ternary systems of the Lhasa Tibetan type, in which an egophoric marker like the imperfective ending -*gyi yod* directly contrasts with the direct evidential ending -*gyis* and the indirect evidential ending -*gyi yod*-*pa red*. However, such an evidential analysis is more difficult to uphold for a binary system of the Kathmandu Newar type, where one only encounters a binary contrast between the egophoric markers -*e* (NPST) / -*ā* (PST) and the allophoric markers -*i* (NPST) / -*a* (PST). Consider the following sentences from Kathmandu Newar below.

	- a. *Ji* 1SG *wan-ā.* go-PST.**EGO** 'I went.' (Hargreaves 2005: 12)
	- b. *Ji* 1SG *mhiga* yesterday *then-a.* arrive-PST.**ALLO** 'I arrived yesterday.' (Hargreaves 2005: 13)
	- c. *Cha* 2SG *wan-a.* go-PST.**ALLO** 'You went.' (Hargreaves 2005: 12)
	- d. *Wa* 3SG *wan-a.* go-PST.**ALLO** 'S/he went.' (Hargreaves 2005: 12)

### Manuel Widmer

As the examples illustrate, the egophoric ending -*ā* is used whenever the assertor describes an event that she / he intentionally and consciously performed herself / himself, while the allophoric form -*a* is used whenever the assertor describes an event that she did not instigate intentionally and / or consciously or that was performed by some other person. Hargreaves (1991, 2005) has argued that egophoric and allophoric endings in Kathmandu Newar essentially encode an opposition of "privileged access" vs. "non-privileged access" to the mental state that is associated with intentional actions. Egophoric markers indicate that the assertor willfully instigated the event in question, while allophoric markers indicate that this is not the case.

Hargreaves' analysis can be contrasted with the evidential approaches mentioned in § 2.2, which analyze egophoricity as an evidential subcategory usually referred to as "participatory evidence" (Plungian 2010; San Roque & Loughnane 2012). If one were to apply such a descriptive approach to the Kathmandu Newar system, egophoric markers would have to be analyzed as evidential markers that mark a proposition as being based on knowledge that was acquired through personal participation in the relevant event.

As far as egophoric markers are concerned, such a model can be implemented without difficulties, as it provides an adequate functional characterization of egophoric marking. However, matters become more complex once we turn to allophoric markers. If we consider the Kathmandu Newar egophoric system presented in Table 9.4 as an evidential system and analyze the endings -*e* 'NPST.EGO' and -*ā* 'PST.EGO' as expressing "participatory evidence", we also have to come up with an adequate functional characterization of the contrasting endings -*i* 'NPST.ALLO' and -*a* 'PST.ALLO'. Egophoric and allophoric endings stand in a salient functional opposition and constitute a functionally self-contained system within the verbal domain. If the egophoric endings are analyzed as evidential markers, it appears conclusive that the corresponding allophoric endings should also be assigned an evidential function. However, such an analysis poses difficulties, as it is hard to assign clear evidential semantics to these markers. It is misleading to characterize allophoric endings as direct evidentials, as they can be used in contexts in which the assertor did not directly witness an event. At the same time, it is problematic to describe allophoric endings as indirect evidentials, since they can occur in contexts in which the assertor directly observed an event.

Current typological models of evidentiality are not helpful in resolving this issue. Consider the following table, which gives an overview of the typological models proposed by Aikhenvald (2004), Plungian (2010), San Roque & Loughnane (2012), and Hengeveld & Hattnher (2015).

### 9 Relationship between egophoricity and evidentiality


Table 9.6: Selected typological models of evidentiality

None of the models listed above postulates an evidential subcategory that accurately characterizes the function of allophoric forms in Kathmandu Newar or other languages with binary egophoric systems. Most notably, Plungian (2010) and San Roque & Loughnane (2012), which postulate participatory evidence as an evidential subcategory and apply this notion to markers that have been described as egophoric elsewhere in the literature, do not postulate another evidential subcategory that would correspond to the allophoric forms in binary egophoric systems. To be sure, San Roque & Loughnane (2012: 158) argue that allophoric endings appear to be functionally equivalent to "visual(-sensory)" evidentials in certain languages. However, as noted above, the allophoric endings of Kathmandu Newar cannot be described as direct evidentials. One might maintain an evidential analysis by augmenting Plungian (2010) and San Roque & Loughnane's (2012) models with an additional allophoric / "non-participatory" evidential subcategory. However, as discussed earlier, this subcategory would be difficult to define with respect to its evidential value. These considerations illustrate that it is difficult to maintain an evidential analysis of allophoric endings and, more generally, of a binary egophoricity contrast of the Kathmandu Newar type.

Proponents of an evidential approach towards egophoricity might argue that allophoric markers could be analyzed as "non-evidential" forms that do not express any epistemic semantics. Such an analysis, however, is at odds with the fact that allophoric markers serve an epistemic function in the sense that they express the assertor's non-privileged epistemic perspective with regard to an event. One might further argue that the epistemic construal of allophoric forms only arises as a consequence of a generalized conversational implicature (see Levinson 2000), that is to say, because they stand in direct contrast with ego-

### Manuel Widmer

phoric forms. However, such a pragmatic explanation would presuppose that allophoric forms are semantically less marked than egophoric forms. This asymmetric relationship would be expected to consistently manifest itself in terms of both form (the allophoric category should be morphologically less marked than egophoric category) and / or function (the egophoricity distinction should be neutralized in favor of the allophoric form in certain contexts) across languages. However, there is no compelling cross-linguistic evidence for a privative semantic markedness relationship between the domains of egophoric and allophoric marking.

In sum, binary egophoric systems pose a challenge to evidential approaches. If one analyzes egophoric markers as expressing "participatory evidence", one is confronted with the difficulty that allophoric markers, which constitute an integral part of binary egophoric systems, cannot be adequately described under such an analysis. This observation suggests that egophoricity may not be an evidential category in the sense of Aikhenvald (2004) but rather represents a functionally independent epistemic "package" that can be combined with an evidential system, as the example of Lhasa Tibetan demonstrates. If this assumption were correct, however, we would expect this functional independence to correlate with a certain degree of structural independence in ternary systems of the Lhasa Tibetan type. This hypothesis is further explored in §4 below.

### **4 Structural evidence for an independent status of egophoricity**

### **4.1 Egophoricity as an independent functional domain**

There are languages with ternary egophoric systems of the Lhasa Tibetan type in which egophoricity and evidentiality manifest themselves as independent functional domains. Such a system is attested in Bunan, a Tibeto-Burman language of North India. The language possesses a simple egophoric system in the present tense, in which there is a binary contrast between an egophoric and an allophoric form, and a more complex ternary system in the past tense, in which the egophoric form is part of paradigm that also comprises evidential markers. Table 9.7 and the example sentences in (2) illustrate the present tense endings.

	- a. *gi* 1SG *dzamen* food *noj* much *dza-k-ek.* eat-INTR-**PRS.EGO.SG** 'I am eating a lot.' (Widmer fieldnotes)

### 9 Relationship between egophoricity and evidentiality

Table 9.7: The present tense egophoric system of Bunan (Widmer 2017)


b. *han* 2[SG] */* / *tal* 3[SG] *dzamen* food *noj* much *dza-k-are.* eat-INTR-**PRS.ALLO.SG** 'You are eating a lot.' / 'She / he is eating a lot.' (Widmer fieldnotes)

Table 9.8 and the example sentences in (3) illustrate the past tense endings.<sup>5</sup>

Table 9.8: The past tense egophoric system of Bunan (Widmer 2017)


### (3) Bunan, Tibeto-Burman


<sup>5</sup>The egophoric past tense marker has the allomorphs -*et* ~ -*men*, whereas the allophoric inferential past tense marker has the allomorphs -*dʑi* (SG) / -*tɕʰok* (PL) ~ -*ta*. The allomorphs are lexically conditioned and depend on the transitivity class of the relevant predicate. Verbs belonging to the intransitive and middle class take the allomorphs -*et* and -*dʑi* (SG) / -*tɕʰok* (PL), while verbs belonging to the transitive class take the allomorphs -*men* and -*ta*.

### Manuel Widmer

As Table 9.9 illustrates, Widmer (2017) proposes a layered analysis of the past tense system, with an evidentiality distinction being nested inside an egophoricity distinction. This is not the only possible analysis of this past tense paradigm, however. Non-nested approaches towards structurally similar systems have been proposed in the literature as well (cf. Tournadre & Dorje 2003, 2008; San Roque & Loughnane 2012; *inter alia*). Under a non-layered approach, the egophoric form directly contrasts with a direct evidential and an inferential form. The nonlayered analysis of the Bunan past tense system is illustrated in Table 9.9.

Table 9.9: A non-layered analysis of the Bunan past tense system


At first, a non-layered analysis of the Bunan system seems more parsimonious, as it allows one (i) to reduce the number of epistemic subcategories in the past tense from four to three and (ii) to reduce the number of epistemic categories from two to one. However, on closer examination it becomes clear that this analysis poses a number of difficulties in the case of Bunan.

If one adopts a non-layered approach and postulates *egophoric* (EGO), *direct* (DIR), and *inferential* (INFER) as directly contrasting values, it becomes impossible to relate the past tense system to the present tense system, which merely distinguishes the values *egophoric* (EGO) and *allophoric* (ALLO). In particular, such an approach fails to explain how the value *allophoric* relates to the values *direct* and *inferential*. Since a non-layered analysis presupposes that all of the aforementioned markers stand in a direct functional contrast, it entails that the values *egophoric*, *allophoric*, *direct*, and *inferential* all belong to the same functional domain. However, this analysis is at odds with the structural organization of the system, in which the values *allophoric* and *direct* / *inferential* are in complementary distribution. In order to maintain a non-layered analysis, one would have to argue that the egophoric markers are "semantically hybrid" in the sense that they can both contrast with evidential markers, which specify one's source of knowledge, as well as with allophoric markers, which express one's epistemic outside perspective with regard to a certain event. However, this entails that egophoricity has to be defined as a "fuzzy" semantic concept, as its subcategory of *egophoric* can serve two different functions, depending on the grammatical context in which it occurs. A non-layered approach towards egophoricity in Bunan thus raises a

### 9 Relationship between egophoricity and evidentiality

number of difficulties that can be avoided if one considers egophoricity and evidentiality as two distinct categories that are independent of each other but stand in a layered relationship.

It is important to note that the paradigm described in Table 9.8 is not the only paradigm in which egophoricity and evidentiality manifest themselves as two distinct functional domains. A structurally similar system is attested in the assumptive present tense, which occurs in propositions that are based on one's overall knowledge of the world. In the assumptive present tense, the egophoricity distinction is nested inside of the evidential subcategory rather than the other way around. This is illustrated by the following table.

Table 9.10: The assumptive present tense endings


The following sentences illustrate the use of the assumptive present tense.

### (4) Bunan, Tibeto-Burman


The difference between the egophoric ending -*men* and the allophoric ending -*mendʑi* is so subtle that it is difficult to capture in the English translation. The use of the egophoric ending indicates that the speaker makes this statement on the basis of knowledge that she / he considers to be personal and intimate. The use of the allophoric ending, in turn, indicates that the speaker makes this statement on the basis of knowledge that she / he considers to be equally accessible to other persons.

The assumptive present tense historically derives from a periphrastic construction consisting of the infinitive -*men* and an inflected form of the equative copula *jen*-. The originally periphrastic nature of the construction can still be seen

### Manuel Widmer

in negated contexts. When the assumptive evidential present tense is negated, the equative copula occurs as a separate phonological word and expresses the negation.<sup>6</sup> Consider the following example.

(5) Bunan, Tibeto-Burman

*Khjak* here *ra-men* come-**INF** *men* **NEG.EQ.EGO** *manɖi=astok* Mandi=TERM *ra-men.* come-**ASSUM.EGO**

'(These trees) do not grow here (in Kullu), they can be found up to Mandi.' (Widmer 2017: 223)

Accordingly, one might argue that assumptive evidentiality in Bunan cannot be considered as an evidential subcategory in its own right, as the relevant construction is still partly periphrastic in negative contexts. However, in affirmative contexts the nonfinite verb form and the copula have phonologically fused to the extent that native speakers are no longer aware that the form originally consisted of two independent syntactic constituents. From a strictly synchronic point of view, it is thus legitimate to analyze the endings -*men*, -*mendʑi*, and -*mentɕʰok* as nonsegmentable morphemes.

One might further argue that the notion of layeredness in itself is not a sufficient criterion to distinguish between the functional domains of egophoricity and evidentiality, as evidential subcategories can themselves display a layered structure. The Tibetan variety spoken in Spiti, North India, for example, displays a set of perfect markers that combine the semantics of direct perception evidentials with the semantics of inferential evidentials. Consider the following examples:

	- a. *khō ārāk tʰūŋ-wanuk*
		- 3SG liquor drink-PERF.**INFER.VIS**

'He must have had liquor.' (I can see it) (Hein 2001: 46)

b. *khō* 3SG *pētɕā* book *sīl-anak* read-PERF.**INFER.NONVIS** 'He must have read the book.' (I can hear it) (Hein 2001: 46)

However, it appears that layered constructions like the one illustrated in (6), in which one grammatical form simultaneously expresses two different evidential subcategories, are rare and combinatorial possibilities of such constructions are

<sup>6</sup> In Bunan, verbal negation is commonly expressed by the prefix *ma*-, which attaches to the verb root (Widmer 2017: 429–432). In the case of the equative copula *jen*-, the negative prefix has phonologically fused with the copula, yielding the negative copula stem *men*-.

### 9 Relationship between egophoricity and evidentiality

highly restricted. These restrictions appear to be a consequence of the fact that two evidential subcategories have to relate to conceptually compatible information sources in order to complement each other. For example, it is difficult to imagine how a direct evidential contrast of visual vs. non-visual evidence could be nested inside of an assumptive evidential construction, because the two subcategories presuppose conceptually distinct information sources, viz. one's sensory perception and one's overall knowledge of the world. One's overall knowledge of the world cannot be accessed through any of one's five senses. Accordingly, there is no cognitive basis for implementing a direct evidential contrast inside of an assumptive evidential construction.

Egophoricity, in turn, does not appear to be restricted in the same way and can be combined with evidential subcategories that presuppose different types of information sources. As has been demonstrated above, egophoricity oppositions can host evidential contrasts that presuppose the direct perception of an event (i.e. direct evidence) or the resultant state of an event (i.e. inferential evidence) and they can also be nested inside of evidential subcategories that presuppose world knowledge about an event. One could of course argue that egophoricity is combinable with so many different evidential subcategories because it presupposes an unspecific type of information source. However, given the fact that there is independent evidence that egophoricity constitutes an autonomous functional domain that is distinct from evidentiality, it is much more sensible to assume that egophoricity is so versatile because it does not indicate the source of one's knowledge but rather specifies the quality of one's knowledge in terms of a binary distinction between an epistemically privileged and an epistemically non-privileged perspective.

### **4.2 Egophoricity as an independent morphological category**

In the languages of the world, we also encounter epistemic systems in which egophoricity and evidentiality manifest themselves as morphologically independent subsystems. Such a system has been reported for the Barbacoan language Tsafiki, which displays two different sets of morphemes to encode egophoricity and evidentiality. As a consequence, the Tsafiki systems looks like a ternary system of the Lhasa Tibetan type from a functional perspective, but differs from the latter in so far as egophoricity and evidentiality are realized as distinct morphological categories. This is illustrated in (7a)–(7c) below. In these examples, the markers -*yo* 'EGO' and -*i* 'ALLO' occur in the morphological slot after the verb root, while the inferential evidential marker -*nu* 'INFER' occurs in the morphological slot following them.

### Manuel Widmer

	- a. *La* 1MASC *kuchi=ka* pig=ACC *tote-yo-e.* kill-**EGO**-DECL 'I killed the pig.' (Dickinson 2000: 412)
	- b. *La* 1MASC *kuchi=ka* pig=ACC *tote-i-e.* kill-**ALLO**-DECL 'I killed the pig (unintentionally).' (Dickinson 2000: 412)
	- c. *La* 1MASC *kuchi=ka* pig=ACC *tote-i-nu-e.* kill-**ALLO**-**INFER**-DECL 'I must have killed the pig (unintentionally).' (Dickinson 2000: 412)

In these examples, the egophoric marker -*yo* 'EGO' expresses that the assertor intentionally and consciously killed the pig, whereas the use of the allophoric marker -*i* indicates that the assertor caused the animal's death by accident. Within the domain of allophoric marking, there is an additional distinction between direct and inferential evidence. The presence of the inferential marker -*nu* in (7b) suggests that the assertor did not directly witness the pig's death but only infers what has happened upon finding the animal's dead body, while the absence of the inferential marker in (7c) indicates that the assertor observed the animal's death (cf. Dickinson 2000: 412). The combination of the egophoric marker -*yo* 'EGO' and the inferential marker -*nu* is not allowed (Dickinson 2000: 410–413). This restriction does not come as a surprise, however. If a predicate takes an egophoric marker, this presupposes that the speaker directly witnessed the respective events. In other words, the privileged epistemic perspective that is implied by egophoric marking entails direct perception of the relevant facts.

One might argue that syntagmaticity is not a valid argument to distinguish the two categories because syntagmatic relationships may also hold between evidential markers (cf. Aikhenvald 2004; Hengeveld & Hattnher 2015) and egophoric markers (cf. Widmer 2017; Floyd 2018). However, the claim being made here is not that markers belonging to the same epistemic category should not be able to stand in a syntagmatic relationship with each other. Rather, the point is that one would expect to find systems in which egophoricity and evidentiality are encoded separately if the two categories are conceptually independent grammatical phenomena, and this is exactly what one encounters in Tsafiki.

### 9 Relationship between egophoricity and evidentiality

### **5 Discussion**

The preceding sections have discussed the relationship between egophoricity and evidentiality from a structural perspective and argued that there is substantial evidence for the claim that egophoricity constitutes a grammatical category in its own right. The arguments that have been brought forward in favor of this position focus on different aspects of egophoric systems. §3 explores the functional motivation of egophoricity by contrasting Hargreaves' (2005) analysis of the Kathmandu Newar egophoric system with an alternative "evidential approach". The section argues that such systems cannot be adequately described in terms of an evidential framework and makes a case for Hargreaves' (2005) analysis, under which egophoricity can be conceived of as a distinct epistemic category with a different functional motivation. §4.1 discusses the structural organization of complex epistemic systems that combine egophoric / allophoric and evidential markers. It is demonstrated that egophoricity and evidentiality constitute two functional layers in Bunan. These layers closely interact but are ultimately independent of each other. §4.2 adduces further evidence for this claim by discussing the case of Tsafiki, a language in which egophoricity and evidentiality stand in a syntagmatic relationship, constituting two independent morphological subsystems.

When considered in isolation, each of the three aspects discussed above may not provide sufficient evidence to warrant the analysis of egophoricity as a grammatical category different from evidentiality. However, when taken altogether, they make a compelling case for treating egophoricity as a grammatical category in its own right. After all, egophoricity meets the minimal requirements to be considered as an autonomous grammatical category. First, egophoricity manifests itself as a functionally and formally self-contained grammatical subsystem in languages with binary egophoric systems (e.g. Kathmandu Newar). Second, egophoricity may manifest itself as a self-contained grammatical subsystem in languages with complex epistemic systems. This may either be in the form of an autonomous functional domain (as in Bunan) or in the form of an autonomous morphological system (as in Tsafiki).

One could make an even stronger case for treating egophoricity as an autonomous grammatical category if one could demonstrate that egophoricity and evidentiality categories combine into paradigms in which they constitute two orthogonal grammatical categories whose subcategories can be freely combined. Such a hypothetical paradigm is given in Table 9.11 below.

However, evidence from natural languages suggests that egophoricity and evidentiality do not combine into paradigms in which they stand in an orthogonal

### Manuel Widmer

Table 9.11: A paradigm with egophoricity and evidentiality as orthogonal categories


relationship. Rather, it appears that egophoricity and evidentiality represent two distinct yet closely related functional domains that can combine into a number of different hierarchical relationships with each other. Evidence for this claim comes from the epistemic systems of Bunan and Tsafiki, which have been described in §4.1 and §4.2. In both languages, we encounter an egophoric system in which the allophoric domain hosts an additional evidential contrast between *direct* vs. *inferential*. This is illustrated in Figure 9.1 below.

Figure 9.1: An egophoric system hosting an evidentiality distinction

In Bunan, we additionally encounter an evidential subcategory hosting an egophoric system. This is illustrated in Figure 9.2 below.


Figure 9.2: An evidential subcategory hosting an egophoric system

The fact that egophoricity and evidentiality fail to combine into paradigms in which they represent orthogonal categories is a consequence of their closely related semantics. Both categories essentially serve the primary function of relating the knowledge that is conveyed in propositions to specific speech act participants. However, they diverge with regard to which aspect of this knowledge

### 9 Relationship between egophoricity and evidentiality

relation they profile. Egophoricity specifies the quality of one's knowledge in terms of a dichotomic opposition that distinguishes between an epistemically privileged and an epistemically non-privileged perspective, while evidentiality specifies the source of one's knowledge.

The relationship between egophoricity and evidentiality is thus comparable to the relationship between the well-established grammatical categories tense and aspect (Gawne & Hill 2017: 299). In some of the world's languages, temporal and aspectual distinctions are intertwined in such complex ways that it is difficult to tease them apart. At the same time, there is compelling cross-linguistic evidence that tense and aspect are two distinct grammatical phenomena, even though they may not be recognizable as such in all languages of the world. Accordingly, few scholars would argue that one should abandon the distinction between tense and aspect and, for example, treat aspect as a subcategory of tense or vice versa. In my view, the relationship between egophoricity and evidentiality should be conceptualized in the same manner. In some languages, egophoricity and evidentiality may be formally and functionally integrated in one single system to the extent that there is no reason to analyze them as separate phenomena. When describing such languages, it may be feasible and justified to describe egophoricity and evidentiality as exponents of one unified grammatical subsystem. However, when treating such languages from a cross-linguistic perspective, it is necessary to make a clear distinction between egophoricity and evidentiality, as there is compelling cross-linguistic evidence that they constitute two distinct grammatical phenomena. That is not to say that egophoricity and evidentiality should only be investigated in isolation of each other. After all, the two categories are closely related from a functional point of view. However, the two categories should then not be treated as one unified grammatical category but as two autonomous epistemic categories.

### **6 Conclusion**

This chapter has discussed the relationship between egophoricity and evidentiality and argued that there is substantial evidence for the claim that egophoricity constitutes a grammatical category in its own right. The arguments brought forward in favor of this claim are based on both functional and structural evidence. First, it was demonstrated that binary egophoric systems are difficult to describe in the framework of an evidential approach, both from a languagespecific as well as from a typological perspective. Second, it was demonstrated that there are complex epistemic systems in which egophoricity manifests itself as an independent functional layer or as an autonomous morphological system.

### Manuel Widmer

Several issues could not be addressed in this paper for lack of space. For example, the study has not addressed the frequently observed sensitivity of egophoric markers to participant roles (see Bickel 2008; Post 2013; *inter alia*). In many languages with egophoric systems, egophoric markers can only be used in contexts in which the assertor assumes a specific participant role, most often the role of an agent (cf. Widmer & Zemp 2017; Widmer & Zúñiga 2017). While such "epistemic argument marking" (Bickel 2008) also appears to be relevant for evidential markers to some extent, it is clear that the phenomenon plays a much more important role in the case of egophoricity. This suggests that the strong tendency of egophoric markers to be tied to certain participant roles is a further characteristic that sets them apart from evidential markers. However, further research is needed to explore this topic.

Another aspect that could only be touched upon briefly in this chapter is the status of the epistemic role of *assertor* in egophoric systems. As noted in §2.1, there is reason to believe that the assertor is a grammatical phenomenon that is independent of egophoric systems and, accordingly, is not part of the functional definition of the category. At the same time, it is a fact that the vast majority of the egophoric systems that have been described so far revolve around this notion. This suggests that the two phenomena are strongly connected. The nature of this connection will have to be clarified by future research.

Finally, this paper has primarily focused on the relationship between egophoricity and evidentiality, thereby neglecting potential relationships to other epistemic categories such as mood, mirativity, etc. Evidence from the languages of the world suggests that egophoricity may closely interact with these categories as well. Future research into these aspects will further enhance our understanding of egophoricity and related grammatical phenomena, thus contributing towards an ever more fine-grained picture of epistemic categories and their mutual interaction.

### **Abbreviations**


### 9 Relationship between egophoricity and evidentiality


### **References**

Aikhenvald, Alexandra Y. 2004. *Evidentiality*. Oxford: Oxford University Press. Bergqvist, Henrik. 2012. Epistemic marking in Ika (Arwako). *Studies in Language*


### Manuel Widmer


Abraham, Werner, 14 Adelaar, Willem F. H., 26, 34,166, 241, 243 Agha, Asif, 73 Ahmad, Ali, 98, 103, 107, 120, 121, 127, 128 Aikhenvald, Alexandra Y., 1, 2, 4, 7, 9, 27, 66, 118, 153, 202, 243– 245, 249, 259, 264, 266–269, 272–274, 280 Åkerman, Vesa, 110 Aksu, Ayhan, 4, 5 Ameka, Felix K., 198 Andvik, Erik E., 14 Asher, Nicholas, 48 Asher, Ronald E., 119 Baart, Joan L. G, 154, 158 Babel, Anna, 245 Bashir, Elena, 144, 151, 153, 154 Basso, Ellen B., 198, 199, 213 Berger, Hermann, 154 Bergqvist, Henrik, 1–3, 6, 9–15, 24, 28, 30, 31, 51–54, 110, 239, 243, 246, 267 Bermúdez, Fernando, 239, 241–243, 254 Berthelin, Signe Rix, 14 Beyer, Artur, 75 Bickel, Balthasar, 10, 265–267, 284 Boye, Kasper, 4, 7, 10, 27, 28, 34 Braun, Friederike, 75, 76

Bühler, Karl, 79 Cacopardo, Alberto, 153 Cacopardo, Augusto, 153 Cain, Bruce D., 96, 98, 100–103, 105– 107, 111, 118, 120, 123, 128, 133 Callisaya Apaza, Gregorio, 241, 243 Cardona, George, 151 Carlin, Eithne, 199 Cartagena, Nelson, 241 Chafe, Wallace, 2, 26, 32, 44, 247 Chuquín, Carmen, 34 Clark, Herbert H., 31, 86 Clift, Rebecca, 244, 245 Cole, Peter, 26, 31, 33, 34, 36 Cornillie, Bert, 4, 27, 257 Creissels, Denis, 3, 62–65, 83, 96, 110, 123, 167, 264–267 Curnow, Timothy J., 5, 8, 9, 65, 110, 165, 166, 178, 179, 188, 194, 244, 245, 265 Cusihuamán, Antonio G., 33, 34 Dahl, Östen, 62, 72, 145, 153, 265 Dasher, Richard B., 52 David, Anne Boyle, 158 Davidsen-Nielsen, Niels, 14 de Haan, Ferdinand, 4, 7, 144, 200, 202, 204, 249 Decker, K. D, 153

Briz, Antonio, 247

Degener, Almuth, 156, 158 DeLancey, Scott, 2, 5, 110, 141, 244, 249, 264, 271 Demuth, Kathryn, 76 Dendale, Patrick, 27 Desclés, J.-P, 3 Deshpande, Madhav M., 151 Dickinson, Connie, 3, 5, 110, 165, 166, 270, 280 Dixon, R. M. W., 1 Döring, Sophia, 32 Dorje, Sanje, 118, 263, 276 Du Bois, John W., 15, 87, 198, 229 Elian, Naomi, 84 Evans, Nicholas, 1, 3, 4, 12–15, 52, 53, 76, 82, 84, 87, 118 Faller, Martina T., 34, 35, 37, 38, 49, 244 Fanshel, David, 11 Floyd, Rick, 34, 37 Floyd, Simeon, 1, 3, 10, 96, 101, 110, 166, 185, 191, 268, 280 Fox, James J., 87 Franchetto, Bruna, 199 François, Alexandre, 69–71 Fried, Robert Wayne, 110 Fritz, Sonja, 96, 98, 103–107, 111, 120, 121, 124–128, 133 Gaby, Alice, 102 Gair, James W., 96, 98, 100–103, 105– 107, 111, 118, 120, 123, 126, 128, 133 García-Ramón, Amparo, 29 Gawne, Lauren, 1, 122, 269, 283 Geiger, Wilhelm, 98, 103, 127, 128 Gioria, Rachel, 87

Givón, Talmy, 15, 30 Gnanadesikan, Amalia E., 98, 100– 103, 107, 108, 116–120, 129 Goffman, Ervin, 245, 246 Gonzales Castaño, Geny, 166 González Ruiz, Ramón, 27 Grzech, Karolina, 4, 23, 25–27, 29, 33, 46, 47, 49, 53 Guentchéva, Z., 3 Guerreiro, Antonio, 198 Hagège, Claude, 62 Haider, Naseem, 142 Hale, Austin, 3, 10, 63, 65, 96, 101, 102, 110, 112, 113, 116, 123, 129, 264, 266 Haller, Felix, 265, 270 Halliday, M. A. K., 15, 69 Hanks, William F., 244, 245 Hargreaves, David, 2, 10, 43, 64, 65, 110, 263, 265, 270–272, 281 Hattnher, Marize Mattos Dall'Aglio, 244, 272, 273, 280 Hayano, Kaoru, 31, 54 Heath, Jeffrey, 76 Heegård Petersen, Jan, 152, 154 Hein, Veronika., 278 Hengeveld, Kees, 2, 244, 272, 273, 280 Heritage, John, 11, 12, 15, 29, 54 Hernández Alonso, César, 241 Hill, Nathan W., 1, 2, 250, 258, 269, 283 Himmelmann, Nicolaus P., 26 Hintz, Daniel J., 14, 34–36, 49 Hintz, Diane M., 14, 34–36, 49 Hobson, Peter, 84 Hock, Hans Henrich, 110, 127 Hoff, Berend J., 199, 203

Hornberger, Nancy H., 25 Howard, Rosaleen, 35, 36 Huber, Brigitte, 265, 270 Hyland, Ken, 3, 12 Jake, Janice, 34 Jakobson, Roman, 7, 87, 246 Jespersen, Otto, 15 Johanson, Lars, 141 Jones, Eunice, 143, 154, 156 Kamio, Akio, 42, 49 Kelly, Barbara F., 110 King, Kendall A., 25 Kiss, K. E., 46 Kittilä, Seppo, 2, 3, 5, 9 Knuchel, Dominique, 1, 11 Kockelman, Paul, 239, 243, 246 Kohistani, Razwal, 156 Koptjevskaja-Tamm, Maria, 72 Kostov, Kiril, 75 Kratochvíl, František, 52 Krifka, Manfred, 32 Kroeber, Alfred, 74 Kroeker, Menno, 3 Kumari, T. C., 119 Labov, William, 11 Lambrecht, Knud, 32 Landaburu, Jon, 31, 84 LaPolla, Randy J., 268 Laprade, Richard A., 241, 243 Lazard, Gilbert, 5, 249, 258 Lehmann, Christian, 69, 71, 81 Lehmann, Thomas, 119 Lehr, Rachel, 153 Leiss, Elisabeth, 14 Levinson, Stephen C., 6, 273 Lewis, M. Paul, 24, 25, 97

Liddicoat, Anthony, 165, 166, 194 Liljegren, Henrik, 142, 144, 147, 150, 159 Lindström, Jan, 14 Loughnane, Robyn, 4, 9,110, 263, 267, 268, 272, 273, 276 Lubberger, Beate, 144, 154, 156 Luis, Ana R., 33 Lum, Jonathon, 102 Lyons, John, 28 Maldonado, Concepción, 255 Masica, Colin P., 119, 144 Matić, Dejan, 48, 54 Maumoon, Yumna, 98, 103, 106 Meira, Sergio, 199 Mendoza, José G., 241, 243 Mendoza, Martha, 213 Merlan, Francesca, 68, 72, 73, 76, 84 Michael, Lev, 202, 229 Mithun, Marianne, 9 Moore, Lesley, 76 Moseley, Christopher, 25 Mushin, Ilana, 8 Muysken, Pieter C., 25, 26, 34, 37, 46, 49, 52, 241, 243 Nichols, Johanna, 2, 10, 265–267 Nikolaeva, Irina, 27 Norcliffe, Elisabeth,165,167–169,176, 187, 190, 194, 268 Nuckolls, Janis, 34, 36, 202, 229, 245 Nuckolls, Janis B., 36 Nuyts, Jan, 81 Olbertz, Hella, 2

Ortíz Ricaurte, Carolina, 13 Oswalt, Robert L., 267

Pabón, Marta, 166, 167, 169, 170

Palmer, Bill, 102 Palmer, F. R., 1, 2, 4, 27 Parker, Gary John, 33 Perder, Emil, 154, 155, 158, 159 Perry, John R., 144, 153 Plungian, Vladimir A., 1, 4, 7, 244, 245, 263, 267, 268, 272, 273 Post, Mark W., 101, 110, 265, 270, 284 Pyrard, François, 98 Quartararo, Geraldine, 8, 243 Radloff, Carla F., 154 Ramirez, Henri, 8 Raymond, Geoffrey, 54 Renzi, Lorenzo, 75 Repp, Sophie, 32, 46, 47 Reyes, Graciela, 240 Reynolds, Christopher, 98 Rojas Curieux, Tulio, 166, 181, 183, 184, 191 Rossano, Federico, 11 Rule, W. M., 14 Rumsey, Alan, 68, 77, 84, 85, 87 San Roque, Lila, 1, 4, 8, 9, 26, 31, 61– 63, 66, 68, 71, 81, 83, 96, 101, 110, 112–115, 123, 124, 129, 166, 174, 188, 247, 248, 263– 265, 267, 268, 272, 273, 276 Sánchez, Liliana, 34, 46, 52 Sandman, Erika, 110 Sapir, Edward, 198 Saudiq, Abdullah, 98, 103 Schieffelin, Bambi B., 61–63, 66, 71, 76, 81, 83, 110, 114, 123 Schmidt, Ruth Laila, 156 Schöttelndreyer, Burkhard, 110 Schultze-Berndt, Eva, 14

Searle, John R., 11 Shakil, Shakil Ahmad, 154 Sillitoe, Paul, 14 Silverstein, Michael, 87 Skopeteas, Stavros, 26 Slater, Keith W., 110, 270 Slobin, Dan, 4, 5 Spencer, Andrew, 33 Speranza, Adriana, 239, 241–243, 257 Squartini, Mario, 239, 240, 244 Stalnaker, Robert, 31, 48 Stivers, Tanya, 11, 24, 28, 41, 54 Strand, Richard F., 156 Sun, Jackson T.-S., 265 Svärd, Erik, 159 Tasmowski, Liliane, 27 Tatevosov, Sergei, 5 Tedlock, Dennis, 86 Tomasello, Michael, 84 Tournadre, Nicholas, 5, 101, 110, 118, 263, 264, 276 Tournadre, Nicolas, 3, 264, 265, 268 Traugott, Elizabeth Closs, 30, 52 Triviño Garzón, Lilia, 166, 167, 169, 190 van Driem, George, 270 Vásquez de Ruiz, Beatriz, 166, 167, 169, 176, 189, 190 Verstraete, Jean-Christophe, 69 Watters, David E., 63, 65, 113, 264, 266, 270 Weber, David, 33, 34, 37 Widmer, Manuel, 95, 97, 101, 118, 125, 129, 131, 132, 265, 267, 269, 270, 275–278, 280, 284 Wijesundera, Stanley, 98, 103

Willett, Thomas, 2, 6, 27, 244, 250 Windfuhr, Gernot, 153 Wroblewski, Michael, 25

Zemp, Marius, 95, 97, 101, 118, 125, 129, 131, 132, 156, 284 Zúñiga, Fernando, 269, 284

# Evidentiality, egophoricity, and engagement

The expression of knowledge in language (i.e. epistemicity) consists of a number of distinct notions and proposed categories that are only partly related to a well explored forms like epistemic modals. The aim of the volume is therefore to contribute to the ongoing exploration of epistemic marking systems in lesser-documented languages from the Americas, Papua New Guinea, and Central Asia from the perspective of language description and cross-linguistic comparison. As the title of the volume suggests, part of this exploration consists of situating already established notions (such as evidentiality) with the diversity of systems found in individual languages. Epistemic forms that feature in the present volume include ones that singal how speakers claim knowledge based on perceptual-cognitive access (evidentials); the speaker's involvement as a basis for claiming epistemic authority (egophorics); the distribution of knowledge between the speechparticipants where the speaker signals assumptions about the addressee's knowledge of an event as either shared, or non-shared with the speaker (engagement marking).