The role of pragmatics in the definition of evidentiality

2.1 The semantic complexity of evidentials

An often cited example of a (large) paradigmatic set of evidential morphemes that evokes a definition in terms of information source is found in the Amazonian language Tuyuca (Barnes 1984, in Matlock 1989: 215 [our glossing]):

The evidential paradigm in Tuyuca appears to neatly illustrate evidential semantics in terms of perceptual and cognitive access, but it contains a large amount of complexity with respect to the use and meaning of forms. According to Barnes (1984), evidential suffixes in Tuyuca encode (subject) person, number and gender, alongside an evidential value, which contains the features ±visual, ±direct, and ±first hand (Barnes 1984: 267). The visual evidential, -wi is thus defined as +visual/+direct/+first hand. However, this evidential suffix is attested in contexts that do not reflect this assigned definition. According to the above definition, the sentence, diiga apewi (‘He played soccer’; ex. 1a), means that the speaker saw the person in question playing soccer. This is, however, not the only meaning attributable to -wi, which can also mark events denoting the speaker’s own actions, which the speaker (as a participant) would not be in a position to observe, only perform (Barnes 1984: 259). The semantic value of -wi thus includes the actions of the speaker and is not restricted to his/her visual perception. Participatory semantics of this kind is attested in an increasing number of descriptions of evidential systems, where it is sometimes afforded separate forms (see Section 3.2; cf. San Roque and Loughnane 2012; Bergqvist and Kittilä 2017, for a discussion).

The second defining value, +direct, is also less straightforward than the definition indicates. -Wi can be used to talk about the results of events, and not just events as they occur. In fact, such indirect access to events contradicts the definition of -wi in terms of directness and produces a semantic overlap with -yi (‘apparent’, ex. 1c, above, Barnes 1984: 258), which denotes indirect visual access (–direct, Barnes 1984: 267). This potential overlap suggests an ambiguity between -wi and -yi as two distinct forms of the same paradigm, but Barnes’ discussion of the two evidential categories suggests differences in terms of agency and privacy. Visual evidentials like -wi are exemplified by sentences describing activities involving people (including the speaker, e.g. studying, playing soccer, burning fields; Barnes 1984: 259), while the apparent, or indirect visual evidential, -yi is found with sentences describing resulting states of non-agentive processes (e.g. rotting, accidentally throwing something away) and the behavior of others (behaving crazy, behaving like an animal, see Barnes 1984: 260). The distribution of -wi and -yi in Tuyuca discourse thus appears to depend on considerations beyond the speaker’s perceptual access to events, which in both cases may target the results of events.

Lastly, -wi marks “‘timeless’ expressions that are within the realm of the speaker’s experience” (Barnes 1984: 259). Such expressions include the names of people and verified facts that the speaker can vouch for. In these cases, perceptual access is not a relevant parameter at all, and if we combine this use of -wi with the above outlined uses, we can only conclude that the meaning of -wi cannot be defined in terms of direct sensory access alone, in contradiction to Barnes’ (1984) proposal. An analysis of -wi based on actual usage would require attention to the above mentioned uses, and it would arguably push issues of perceptual accessibility to the analytical periphery.

A direct reflection of the semantic properties of direct evidentials in Tuyuca is also found in Cuzco Quechua (Quechuan, Peru), as described by Faller (2002). Cuzco Quechua has three evidential forms, namely =mi (‘direct’), =chá (‘conjectural/inferential’) and =si (‘report’). These are shown in ex. (2):

The direct evidential =mi is not limited to qualifying directly perceivable events; It can also mark encyclopedic information for which no sensory access is required, as well as reports about events, as long as they come from trusted sources and permit integration into the speaker’s knowledge (Faller 2002: 133). =mi is thus used to mark both direct and indirect access to events, in a way that is comparable to the direct evidential -wi in Tuyuca, which Faller also discusses in some detail (Faller 2002: 42). Faller proposes a definition of =mi in terms of the speaker’s “best possible grounds” for making an assertion and concludes that information source cannot be relied on to account for the semantics of evidentials in Cuzco Quechua (Faller 2002: 140). Faller’s account of evidentials in Cuzco Quechua anticipates our standpoint with regard to the pragmatically conditioned meaning of evidentials. However, even though Faller argues for novel terminology and classification, she maintains a traditional view of evidentials as reflecting the speaker’s (epistemic) evaluation. A term like “best possible grounds” does not permit a focus on the dynamics of the speech-situation and the dialogic negotiation of epistemic rights. We argue that such interactive concerns must be included in attempts to analyze evidentials and that the speaker’s access to events is insufficient for a consistent analysis of evidential forms.

In this section, we have shown that direct evidentials in Tuyuca and Cuzco Quechua, which the literature treats as textbook examples of evidential paradigms encoding source of information, convey participation in events, the speaker’s consideration of their interlocutors’ knowledge state, and the integration of information that is regarded as generally accepted fact.

2.2 The grammatical status of evidentials

From early on in modern research on evidentiality, it has been noted that evidentials are found in different parts of grammar and the lexicon (e.g. Jacobsen 1986). Some evidentials are affixes on the verb (and sometimes, the noun), others are particles less closely tied to any specific clausal constituent, and yet others are lexemes belonging to word classes such as adjectives and adverbs (see contributions in Chafe and Nichols 1986 for various observations of formal differences between systems and languages). Aikhenvald (2003) advocates a narrow view of evidentiality, where only grammaticalized forms qualify for study. This narrow approach excludes the use of e.g. adverbs to signal how knowledge was acquired; these are called “evidential strategies” (Aikhenvald 2003: 2). Mushin (2013), by contrast, argues that there are no distinctive functional differences between grammaticalized evidentials and lexical ones; they are comparable in function and meaning, whereas the frequency and distribution of forms, of course, differ according to their grammatical status. We agree with Mushin that lexical evidentials are relevant to the study of evidentiality, but given the stated aim of the present paper, we take the opportunity to point out some distinct differences between a grammatical form like the purportedly visual -wi in Tuyuca and lexical means to express visual perception (i.e. “seeing”) in English.^[3] This comparison is offered in order to support our overall argumentation that (direct) grammatical evidentials encode the speaker’s claim of knowledge, and not their visual access to talked-about events (see Section 4). Example (1a) from Tuyuca (repeated here), is paraphrased in English (3a–d) in order to exemplify differences with respect to how direct (visual) access is conceptualized using a grammaticalized marker in Tuyuca and lexical means in English:

The first thing to note is that the constructions/lexical forms denoting visual access in (3) are speakers. The speakers are quite free to choose how they wish to formulate utterances in terms of lexical content. This constitutes an important difference between lexical morphemes in English and highly grammaticalized evidential morphemes in Tuyuca. Secondly, there are differences in meaning between (3a–d) resulting from syntactic and lexical choices. Although all constructions/forms in (3) feature a word related to seeing, they are not equivalent with regard to their primary denotational meaning, nor with respect to their connotations. The verbs see (3a, b) and watch (3c) can be distinguished by expressing an experience (to see) and an activity (to watch). The adverb, visibly, in (3d) modifies the verb playing, thereby conveying that the activity of playing soccer was plain to see and discernible to anyone present. The complement-taking construction in (3a) relates the experience of seeing to soccer-playing by the coordination of two finite clauses, one taking the other as a syntactic complement. The example in (3b) is a periphrastic construction where I saw has a syntactically and discursively secondary status, making it more of a meta-comment on the primary proposition, He played soccer. There are many semantic and grammatical nuances in the examples in (3), but not all of these would be prominent features of a grammatical form with the function of indicating visual access to an event.

More importantly, the examples in (3) cannot be used to describe situations where the act of seeing/watching is absent. If a speaker says, I saw that he played soccer, this utterance is only felicitous if the speaker indeed saw the person in question play soccer. The verbs see and watch do not denote involvement (i.e. doing), nor verified facts (cf. Barnes 1984: 259; Section 2). While lexical expressions related to seeing are semantically rich and allow for a number of ways to express this mode of sensory access, the grammaticalized evidential -wi in Tuyuca does not require seeing an event for the appropriate use of the form. -wi also signals involvement, resulting states, and things that the speaker knows for a fact (such as the names of people). Thus, the meaning of -wi in Tuyuca is quite different from what we would expect from a grammatical marker encoding ‘seeing’. From the discussion above, we see that it is problematic to base analyses of evidential markers in a language like Tuyuca on the basis of translation equivalents in a language that lacks grammatical evidentials (e.g. Spanish), or has different evidential distinctions.

We argue that the “additional” uses of -wi in Tuyuca, as reported by Barnes (1984), depend on a non-defeasible semantic feature inherent to such forms, namely the speaker’s ‘epistemic authority’ (see Section 1; Bergqvist and Kittilä 2020; Bergqvist and Knuchel 2019; Grzech 2020a). As we will show in Section 4, this notion is consistent with the attested uses of direct evidentials in Tuyuca and several other languages that have direct evidentials (see e.g. Curnow 2003; Grzech 2016). The ensuing exploration of issues, below, contributes additional evidence for the claim that epistemic authority is a relevant notion for the semantics of direct evidentials in the languages under discussion.

3.1 Evidentials and context: “mismatches”

Nuckolls and Michael (2014) argue that the use of evidentials is important for their description and we take this point further to argue that the pragmatics of evidentials should inform the definition of evidentiality. Pragmatic factors, such as perceptual-cognitive access in relation to the choice of evidentials and inter-speaker variability regarding the use of evidentials, are especially relevant in this regard. These parameters have been noted and discussed in the literature, but they have yet to figure in cross-linguistically viable attempts to define evidentiality. With respect to perceptual-cognitive access vis-à-vis choice of evidential, Sun (2018) observes that “[t]he speaker may construe a given situation in alternate ways and, when the need calls for it, avail herself of a range of non-typical evidential choices to convey additional semantic effects beyond the simple provision of information source” (Sun 2018: 58). While Sun’s formulation originates with a view of evidentials as primarily encoding information source, our analytical starting point is that the speaker’s construal of an event through the use of certain evidentials is not derived from how they may convey information source, but rather from how the speaker wishes to situate their knowledge against that of their interlocutor.

Mithun (2020) points out that speakers make different evidential choices even in seemingly identical situations. Perceptual and cognitive access to events does not allow one to predict the choice of an evidential form. Mithun also notes that “some evidentials tend to occur pervasively in certain kinds of speech events, more rarely in others, and not at all in still others” (Mithun 2020: 318). A single genre is unlikely to provide a full picture of all evidential strategies in a language and, if the aim is to identify and accurately describe all evidential forms in a language, then paying attention to genre is important, both methodologically and analytically.

Inter-speaker variation in the choice of evidentials has been studied using data from task-oriented speech, where speakers who are engaged in the same task use different forms to qualify otherwise identical utterances relevant to the task at hand (see Grzech et al. 2020, for a detailed discussion). Silva and AnderBois (2016) explore ways to stage (task-oriented) communicative events that prompt the use of evidential forms in naturalistic dialogue. Using the board game “Mastermind” as elicitation tool, they find previously undocumented epistemic/evidential markers with intersubjective, mirative, and dubitative semantics (Silva and AnderBois 2016: 71–72). Their investigation underscores the value of going beyond direct elicitation and the collection of narratives in the exploration of evidentials and epistemic marking strategies.

In a study of indirect evidentials in La Paz Aymara, Quartararo (2017) discusses the use of the non-experienced past form -tayna in task-oriented speech. With data from the Pear Story (Chafe 1980), Quartararo observes that speakers of La Paz Aymara routinely use -tayna to talk about the moving images of the story, despite their obvious appearance. Consider example (4) from Quartararo (2017: 150):

Quartararo remarks that the contents of the movie segment, which the speaker refers to in Example (4), are plain to see and that the use of the indirect evidential -tayna is unexpected for this reason. She goes on to suggest that speakers may use -tayna in the context of tasks such as the Pear Story to distance themselves from its contents in a way that is comparable to how speakers sometimes recount traditional Aymara narratives (Quartararo 2017: 150).

Zeisler (2016) notes for Ladakhi (Western Tibetic, India/China) that socially conditioned uncertainty resulting from the presence of figures of authority may influence speakers’ choice of evidentials. The use of an inferential marker in such situations may thus serve as a disclaimer and may be triggered by considerations of deference and politeness (Zeisler 2016: 62). The use of inferential evidentials may also signal a reference to knowledge as shared between equals. Like in the deferential use, the speaker thus refrains from posing as the sole authority and invites the addressee to share the knowledge of the speaker (Zeisler 2016: 63)

Gipper (2018) also points out substantial variation in the way speakers use evidentials and related epistemic markers in Yurakaré (isolate, Bolivia), which cannot be accounted for by sociolinguistic variables, such as age or gender, alone. The individual preferences of speakers cannot be discounted in charting usage patterns of evidentials in Yurakaré. Evidence from staged, communicative elicitation tasks suggest that the methodologies we develop for eliciting and annotating evidentials must be informed by their contextualized use and by how speakers’ evidential choices are affected by the inter-personal context (see Grzech et al. 2020, for details).

Mushin (2001) argues that the primary function of evidentials is to express the speaker’s epistemological stance. This consists of a “conceptualisation of information in terms of a speaker’s assessment of her knowledge, and the internal structure of these conceptualisations that result in a variety of mappings onto linguistic structures” (Mushin 2001: 52). She further argues that “the adoption of an epistemological stance is an automatic response to attending to some body of information”, but that the choice of this stance, and the form used to convey it, is not automatic, even when evidentials are obligatory (Mushin 2001: 55). The speaker’s individual stance and that speaker’s status relative to other speakers in the community has an effect on the choice of evidentials that goes beyond perceptual and cognitive access to the event in question. The purported mismatches observed in the literature on evidentials are only mismatches if evidentiality is conceptualized in terms of information source. This problem disappears if we instead conceptualize evidentials in terms of their epistemic function: to claim or defer knowledge in an interactive setting. In our attempt to formulate a revised analytical framework for evidentials, we aim to identify meaning features that persist across uses and grammatical contexts. We regard semantic content that only can be found in certain grammatical and discourse contexts as an implicature that stems from the encoded meaning of a given form. For grammatical forms, we do not expect such encoded meaning to be salient to, or available for commentary by the speakers that use such forms, given that these have more functional than lexical/conceptual content (cf. Mithun 2020, for a discussion). Direct evidentials encode the speaker’s claim of knowledge, not their sensory/perceptual experience, which can only be viewed as grounds for acquiring knowledge. This is why we think that “information source” is a flawed definition of evidentiality, and why a pragmatically grounded notion like epistemic authority is a better candidate (see Section 4).

3.2 Evidentials and person

The interpretation of evidentials is often conditioned by the subject person of the utterance they occur in (Aikhenvald 2004; cf. Curnow 2002, 2003). The interaction between evidentials and subject person may consist of restrictions with regard to which evidentials may co-occur with certain subject persons, but it can also pertain to the meaning of evidential forms (see Curnow 2003). An example of a subject person-based restriction is found in Tuyuca, where there is no first/second person form for the present tense inferential evidential (Barnes 1984: 258). This means that stating an inference about ongoing actions/events that primarily involve the speaker is not possible using a grammaticalized evidential form (see also Malone 1988: 120).

The combination of direct evidentials and first person subjects may produce interpretations not found with third person subjects. An example of this comes from Tucano, a language closely related to Tuyuca, where visual and non-visual evidentials (as defined against the context of a third person subject) in first person contexts prompts the interpretation that the speaker performed a volitional action (5a), or an unintentional/accidental action (5b):

A traditional view of evidentiality as signaling information source treats this apparent change in meaning as pragmatically induced, and secondary to the encoded, core meaning of an evidential, which is found in contexts with third person subjects. We do not think this is an appropriate mode of analysis. The fact that the interpretation of evidential forms may depend on grammatical distinctions that reflect aspects of the context (e.g. the speaking subject), is a significant fact that has a role to play in the analysis of evidentials. There are comparable observations made with respect to the interaction between subject person and epistemic modals, which we take to be indicative of the strong connection between epistemic modality and evidentiality (e.g. Lehmann 2012).

The meaning of direct evidentials as signaling the speaker’s actions was noted for visual evidentials in Tuyuca and Cuzco Quechua (Section 2). As noted in Section 2, some languages have separate evidential forms conveying involvement and participation, which are called ‘participatory’ and ‘performative’ evidentials (Bergqvist and Kittilä 2017; San Roque and Loughnane 2012). Such forms imply subject identity, but may occur with any subject person. In third person contexts, participatory evidentials are interpreted as signaling factuality, a feature that they have in common with visual evidentials in Tuyuca (Section 2; see also Aikhenvald 2018: 32). An example of a performative evidential from Central Pomo is found in ex. (6):

Data from genetically diverse languages such as Tuyuca, Cuzco Quechua and Central Pomo point to the importance of first person contexts (i.e. indexing the speaking subject) in the analysis of direct evidentials. The speaker’s “volitional action” and “performance” as conveyed by direct evidentials may be viewed against the speaker’s role as epistemic authority, not as an extension of visual access, which is an irrelevant semantic parameter in these cases.

Aikhenvald claims that evidentials are most frequently used in third person contexts (Aikhenvald 2018: 27), but this claim is not supported by reference to actual frequency counts, or publications with such data. We are unaware of any substantiated claims of this kind in the literature, and we suspect that it has more to do with unquestioned remnants of traditional, western philosophical inquiry, than with the actual distribution of evidentials in languages that have them. The inherent meaning of linguistic categories is traditionally conceived of in terms of truth-value, and investigated forms should occur in “objective” contexts, void of any trace of subjectivity (i.e. speaker evaluations; see e.g. Searle 1987). If it can be demonstrated that evidentials predominantly occur in third person contexts, then this is an interesting fact, but not necessarily an argument to treat such contexts as basic, or primary. Dahl (2000) observes that egophoric pronominal forms denoting the speech-act participants are more frequent than allophoric (third person) pronouns in spoken Swedish. For non-prompted, spontaneous conversations, Dahl concludes that speakers talk more about themselves and their interlocutor(s) than they talk about people who are not present (i.e. third persons; Dahl 2000). Dahl finds support for his observations concerning spoken Swedish in comparable corpora of English and Spanish. The statistical tendency for egophoric subject pronouns to be more frequent in everyday talk should not be discounted in attempts to analyze parts of grammar that are sensitive to person distinctions. We consider it likely that frequencies of co-occurring evidentials and subject persons differ according to genre (narrative/dialogue), modality (spoken/written), and the meaning of the evidential in question (e.g. direct/indirect). We should not assume that there is a privileged combination of evidential and subject person without really having looked at the co-distribution of evidentials and subject person in these differing contexts.

In sum, the interpretation of evidentials is systematically conditioned by subject person, and the claim that evidentials predominantly occur in third person contexts is unsubstantiated. The co-distribution of evidentials and (subject) person is expected to depend on contextual factors that make generalizing predictions concerning such pairings complex and challenging. If we are to look for a genre of speech that may serve as prototypical in a wider sense (i.e. in terms of frequency, as well as in an evolutionary sense), it would have to be spontaneous, non-prompted, face-to-face conversation. This is also where we should start to look for default uses and meanings of forms. Given the predominantly occurring egophoric person reference in this mode of speech (genre), first and second person contexts cannot be regarded as secondary to third person contexts. If there is a bias towards using evidentials in third person contexts, as Aikhenvald claims, then this purported bias can only be explained in terms of function, not by ad-hoc presumptions of primary meaning.

3.3 Evidentials and assessor shift

Evidentials can be anchored to the speaker or the addressee. This means that the epistemic qualification found with an evidential can be attributed to the addressee and reflect the expectations of the speaker concerning the addressee’s epistemic perspective in talking about an event. This is a well-known feature of evidentials and epistemic modals alike (San Roque et al. 2017; cf. Lehmann 2012). Sun (2018) points out the importance of addressee in the analysis of evidentials:

The addressee is […] a critical person factor in shaping evidential formation and selection. In evidentially marked interrogatives, one may couch a question in the addressee’s perspective, presupposing an information source likely to be available to the latter. Aside from deferring the authority of assertion to the addressee in questions, one’s perspective in assertions may also be shifted to, or combined with, the addressee’s perspective to reflect the latter’s sources of knowledge, or adduce shared evidence to support one’s verbal claims. (Sun 2018: 62–63)

Ex. (7) illustrates how the visual evidential -tia, in Duna (Trans-New Guinea, Papua New Guinea) can signal the speaker’s (7a) and the addressee’s (7b) direct access to the pig’s visit to the speaker’s garden:

Primarily, this is a feature of evidentials in questions (regardless if they are truly interrogative, or not), but there are also examples of addressee-including evidentials that do not require a change of sentence-type, as discussed by San Roque (2015) for “impersonal” evidentials in Duna:

According to San Roque, the -yanua and -nua forms denote potential sensory and visual access to some discourse entity by “someone” (hence the term, impersonal), perhaps most likely, the speaker and the addressee, but even though San Roque is hesitant to offer a definite characterization of the impersonal evidential suffixes, it seems clear that given their generic reference, the perspective of the addressee is not excluded. San Roque’s example of impersonal evidentials signals perceptual-cognitive accessibility per se without connecting this accessibility to a given cognizer, a conceptual difference between “I saw that x” and “x was visible”. Mushin (2013) comments on general sensory availability as a conceptual building block for (direct) evidentials, “linguistic evidence as the basis for one’s knowledge is thus to draw on (potentially) publicly verifiable evidence – what anyone might have seen or heard if they had been in the speaker’s place” (Mushin 2013: 642). However, this conception is less well attested than one where evidentials are anchored to a specific cognizer, i.e. the speaker and/or the addressee. We are open to the possibility that non-deictic evidentials of the kind outlined above should be regarded as functionally different from the kinds of evidentials that are anchored to the speech-act participants. We leave this possibility for another discussion, however.

Tuyuca has designated interrogative-evidential forms, which according to Malone (1988: 22), convey a strong expectation that the addressee will be able to provide an appropriate response to the evidentially-laden question.

Lehmann calls the anchoring of evidentials and modals with the addressee “assessor shift” (Lehmann 2012: 10), a term that denotes a shift of epistemic perspective from the speaker to the addressee. Lehmann exemplifies assessor shift in modals with data from German, Yukatek, and Korean, but also notes the same kind of shift in evidentials (Lehmann 2012: 37).

A related notion that also features this kind of perspective-shift between the speaker and the addressee is egophoric marking (a.k.a. “conjunct/disjunct”, Hale 1980; “assertor’s involvement”, Creissels 2008; ‘egophoricity’, Floyd et al. 2018). Egophoric marking may be defined as signaling the epistemic authority of the speaker and/or the addressee, subject to their respective involvement in an event (see Bergqvist and Knuchel 2017 for a discussion; cf. Floyd et al. 2018). Egophoric marking thus conveys the involvement of one, or both, of the speech-act participants, as indicated by sentence-type. In statements, the egophoric marker conveys the speaker’s involvement and their resulting epistemic authority, and in questions, the egophoric marker targets the speaker’s deferral of epistemic authority to the addressee given their involvement in the talked-about event. The actions of third parties and events with third person subjects are non-egophoric. An example of egophoric marking in Kathmandu Newar is found in example (10), where ā is egophoric with the past tense form of the verb wane ‘go’:

From the literature on egophoric marking and epistemic modality, we can tell that placing an epistemic qualification with the speaker, or the addressee, is a trait shared by at least two related forms of epistemic marking that may be distinguished from evidentiality (see Boye 2012 for a discussion of the domain of epistemicity in language with respect to epistemic modality and evidentiality). We do not find equivalent shifts in other verbal categories (e.g. tense-aspect marking), which means that this shift of viewpoint must be considered as a defining property of evidentiality and related epistemic notions. To-date, this property has yet to figure in attempts to provide a functional definition of evidentiality. San Roque et al. (2017) observe differences in how evidentials signal a shift in perspective, from their discussion of evidentials in interrogatives, it appears that this feature is mostly present in direct evidentials, and less likely for indirect evidentials. The authors suggest a possible explanation for this by drawing, partly, on the role of epistemic authority as a relevant notion for evidentials and egophoric markers alike (San Roque et al. 2017: 140).

3.4 Evidentials and intersubjectivity

Evans et al. (2018a, 2018b) discuss the notion of ‘engagement’, as encoding asymmetries in the mental disposition of the speech-act participants (Evans et al. 2018a: 118). Engagement refers to intersubjective configurations of knowledge/access, which can be exclusive to the speaker or shared with the addressee. In their discussion of how systems reflect this notion, Quechuan languages are noted to feature engagement semantics as part of the meaning of evidential forms. In Sihuas Quechua, South Conchucos Quechua and Huamalíes Quechua, all spoken in Peru, the evidential paradigms consist of more than the three forms discussed by Faller (2002) for Cuzco Quechua (see Section 2.1). South Conchucos has five evidential markers and Sihuas and Huamalíes Quechua each have six (Hintz and Hintz 2017; Howard 2012). In addition to the three markers shown in example (2) above, these languages also have markers encoding individual and shared knowledge. South Conchucos has markers of “mutual knowledge” and “shared conjecture” (Hintz and Hintz 2017: 88), and Sihuas has a system of three contrastive pairs, encoding distinctions between: i) mutual versus individual knowledge, ii) individual versus shared conjecture and iii) individual versus generalized knowledge based on a report (Hintz and Hintz 2017: 96). In Huamalíes, there are markers indicating speculation, assertion of knowledge co-constructed by the speaker and the addressee, and negation of such knowledge (Howard 2012). Examples of intersubjective semantics in South Conchucos and Sihuas Quechua are shown in examples (11) and (12), below (Hintz and Hintz 2017: 91):

The evidential systems of South Conchucos and Sihuas Quechua diverge from current conceptions of evidential systems, since evidentials in these languages convey not only the speaker’s basis for claiming knowledge, but also whether this is shared with the interlocutor or not. Hintz and Hintz (2017) state that “[i]n terms of evidentiality, the source of information for mutual knowledge has to do with the contributions of conversational participants as well as their shared nonlinguistic experiences and the beliefs and assumptions they hold in common” (Hintz and Hintz 2017: 92). Thus, aspects of the interpersonal context are important, not only for the choice of evidentials, but also for the semantics encoded in some evidential forms. Reflexes of =mi (‘direct’) and =chi (‘conjectural’) in South Conchucos and Sihuas Quechua have “the same underlying semantics as the cognates -mi/-n and -chá in Cusco Quechua” (Hintz and Hintz 2017: 93; cf. Section 2.1). It is likely that Hintz and Hintz were able to pin down the epistemic distinctions related to ownership of knowledge because, unlike many previous descriptions of Quechuan evidentiality (including Faller’s work on Cuzco Quechua), their analysis is based on conversational data. In fact, they observe that two of the South Conchucos evidentials, =cha: (‘confirmation/assertion of mutual knowledge’) and =cher (‘mutual conjecture/appeal for consensus’) only occur in verbal interaction (Hintz and Hintz 2017: 93). They conclude that “evidential markers are essentially interactional devices for the packaging and negotiation of information in discourse” (Hintz and Hintz 2017: 92), and, as such, are best analyzed on the basis of how they occur in conversations.

Nuckolls’ (2018) account of evidentials in Pastaza Quichua, spoken in the Ecuadorian Amazon, supports this view of evidentials as interactional devices. The use of evidentials in Pastaza Quichua is motivated by the speech-act participants’ wishes to adopt a certain perspective, rather than by the source of information for their claims (Nuckolls 2018: 203). This can be seen in example (13; Nuckolls 2018: 208):

Example (13) shows how -mi occurs in a conversation about a baby who is the size of a newborn, despite being seven months old. The speaker is aware that all of the baby’s older siblings have died in infancy, so the claim she makes is not based on established facts. Nuckolls observes that -mi is used to signal the perspective of the speaker, distinguishing it from the perspectives of other discourse participants (cf. Nuckolls 2018: 205–206). Signaling the speaker’s perspective is part of the semantics of -mi in Pastaza Quichua and Nuckolls claims that indicating the source of knowledge is a less salient feature than accurately expressing with whom knowledge originates, and with whom the rights to information resides (cf. Nuckolls 2018: 206).

Evidentials in Pastaza Quichua, South Conchucos, Sihuas, and Huamalíes Quechua are analyzed in slightly different terms, but the context of their use and their respective functions in interaction are strikingly similar. In these Quechuan varieties, the use of evidentials is not primarily related to sources of information, but to the epistemic rights and responsibilities of the speech-act participants.