Evidentiality in Nakh-Daghestanian Languages
Abstract and Keywords
This chapter is focused on the formal expression of evidentiality in Nakh-Daghestanian languages (Russia, Caucasus) and on the semantic distinctions available for evidentials. The vast majority of Nakh-Daghestanian languages express evidentiality in one way or another through grammatical means, but there are also many evidential strategies attested. Verbal evidentiality mainly shows up as one meaning of the perfect series and in inferential constructions with light verbs such as ‘find’, ‘stay’, or ‘be, become’. In addition, some languages have evidential enclitics and suffixes. The evidential meanings expressed are mostly indirect evidentiality, including hearsay and inference. The chapter also presents a short overview of related constructions such as conjunct-/disjunct-marking and epistemic modality.
The Nakh-Daghestanian (or East Caucasian) language family is the largest and the most diverse of the three autochthonous language families in the Caucasus, the other two being West Caucasian (or Abkhaz-Adyghe) and Kartvelian (or South Caucasian). The more than forty Nakh-Daghestanian languages are spoken in the southern parts of Russia, in northern Azerbaijan and in a few speech communities are found in Georgia (Map 23.1). The family can be divided into several sub-branches: Nakh (Chechen, Ingush, Tsova-Tush), Avar-Andic (Avar, Andi, Godoberi, Bagvalal, and more), Tsezic (Tsez, Hinuq, Khwarshi, Hunzib, and Bezhta), Lezgic (Lezgian, Agul, Tsakhur, Tabasaran, Kryz, Rutul, Budukh, Archi, and Udi), Khinalugh (sometimes grouped together with Lezgic), Dargi (traditionally considered to be one language, but consisting of several varieties that are mutually incomprehensible), and Lak (sometimes grouped together with Dargi).
The largest language of this family is Chechen with more than one million speakers and enjoying official status in the Autonomous Republics of Chechnya and Daghestan. Yet most of the Nakh-Daghestanian languages are rather small and exclusively used for oral communication within villages, e.g. Hinuq, Archi, or many Dargi varieties.
From a typological point of view, the languages have rather large consonant inventories, including ejectives and pharyngealization. Their morphology is agglutinating/fusional and the complexity strongly varies from language to language, with Ingush being the most complex, Lezgian the simplest language according to Nichols (2013).1 The languages are predominantly dependent-marking with rich case inventories, including ergative case and usually a vast array of spatial cases. In most of the Nakh-Daghestanian (p. 491) (p. 492) languages gender is an important grammatical category realized as gender/number marking on verbs and other parts of speech triggered by the absolutive argument. The languages have rich inventories of finite and non-finite verb forms (converbs, participles, infinite, and masdar—a deverbal noun). Common valency classes are: (i) intransitive (having one single argument in the absolutive), (ii) extended intransitive (one absolutive argument and a further argument in a spatial case), (iii) transitive (one absolutive and one ergative argument), (iv) extended transitive (one absolutive, one ergative, and one further argument in the dative or a spatial case), and (v) affective (one experiencer argument in the dative or a spatial case and one stimulus argument in the absolutive). The most common constituent orders are SV and AOV at the clause level and head-final order at phrase level. For recent overviews see van den Berg (2005); and Daniel and Lander (2011).
This chapter focuses on the formal expression of evidentiality in Nakh-Daghestanian languages and the semantic distinctions available for evidentials. The vast majority of Nakh-Daghestanian languages express evidentiality in one way or another through grammatical means. However, the Lezgic language Udi is a notable exception. In the other languages one finds grammaticalized evidentiality as well as many evidential strategies within the verbal paradigm (§§23.2.2–5). There are periphrastic light verb constructions (§23.2.4) and evidential enclitics and suffixes (§23.3). §23.4 contains a short overview of related constructions (e.g. conjunct-/disjunct-marking, epistemic modality).
There are no comparative studies of evidentiality in Nakh-Daghestanian languages, and older grammars often do not provide information about this topic. However, all recent comprehensive grammars include sections on evidentiality. In addition, a number of case studies of individual languages have been published in recent years, including Molochieva (2011) on Chechen; Tatevosov (2007b) on Bagvalal; Comrie and Polinsky (2007) on Tsez; Forker (2014) on Hinuq; Khalilova (2011) on Tsezic; Maisak and Tatevosov (2007) on Tsakhur; Tatevosov (2001a) on Archi, Bagvalal, and Dargwa; Mallaeva (2007) on Avar, and Maisak and Merdanova (2002) on Agul.
23.2. Verbal evidentiality
23.2.1. General characteristics
In Nakh-Daghestanian verbal evidentiality mainly shows up in two ways: (i) as one meaning of the perfect series, and (ii) in periphrastic light verb constructions. Furthermore, two languages have special constructions not attested in any other language of the family (evidential copula auxiliary in Chechen, past participle in Avar).
These constructions usually express only indirect evidentiality (though see the discussion in §23.2.5 on direct evidentials); that is, predominantly hearsay and inference from sensory evidence. Occasionally, one can find examples illustrating inference from general knowledge or pure reasoning.
Evidential systems belonging to (i), i.e. having at least a formal connection with the perfect, are often evidential strategies since they also have non-evidential readings (p. 493) (resultative/perfect).2 Most of the systems share a couple of properties so that we can describe a typical Nakh-Daghestanian verbal evidential system as being:
• small with marked indirect versus neutral (i.e. unmarked) verb forms
• confined to the past tenses and conflated with the tense system
• not conflated with modality, i.e. the use of an indirect evidential does not mean or imply that the speaker is uncertain or not committed to the truth of his/her utterance
• interaction with person (‘first-person effect’, see §23.2.2)
• negation has the proposition in its scope, not the evidential
• in questions, the same forms are available as in indicative clauses, and the information source of the addressee is at issue
Speakers are highly aware of the evidential semantics and comment on it (cf. Nichols 2011: 243 on Ingush). Evidentials are found in various speech styles, though some forms largely seem to function as genre markers for traditional folktales. The origins of the evidentials vary: evidentials originate in perfects, the past participle (as in Avar), and complement construction (e.g. an inferential construction, as shown in §23.2.4).
23.2.2. Evidential strategies: The perfect series
Most if not all of the Nakh-Daghestanian languages have verb forms comparable to perfects in other languages, both from a formal as well as from a functional point of view. These verb forms are typically analytic, made up of a (perfective/past) converb or participle and a copula or a similar auxiliary (e.g. ‘be in’ in Agul and Lezgian). If the copula/auxiliary can itself be inflected for other tenses we get a series of verb forms that I will refer to as perfect series.3 This series is mostly in opposition to another series of past tenses that do not express evidentiality, often called aorist.
In many of the surveyed languages (Bagvalal, Godoberi, Avar, Lak, Dargwa varieties such as Sanzhi, Ashti, Kubachi, Icari, and Standard Dargwa, Ingush, Agul, and Tsakhur) a perfect-like verb form and eventually other verb forms from this series (e.g. pluperfect) express past time reference with some additional meaning components including indirect evidentiality. Which additional meaning component is at stake depends on the individual verb forms. For perfect-like forms it is normally resultativity and/or perfect meaning. In the case of the pluperfect, it is relative past time reference, and for other verb forms it may be imperfective aspect. Only in a few languages some or all verb forms of the perfect series have developed into grammaticalized indirect evidentials (see the discussion of Tsezic in §23.2.3). (p. 494)
Whether indirect evidential meaning is expressed by the perfect or not depends on the lexical semantics of the verb in the utterance as well as on the context, i.e. telic verbs that describe actions with results may enhance the resultative reading. For instance, in Bagvalal (Tatevosov 2007b) verbs can be divided into three groups:
i. verbs with no evidential meaning (q’očã ‘want, love, fall in love’, b-ič’ã ‘look, wait, try’, b-iɬɬi ‘concern, seem’). When q’očã is inflected for the perfect, the result is a reading with present time reference. The same is observed for verbs with similar meanings in Hinuq (Forker 2013: 222) and Avar (Forker in preparation (a)).
ii. verbs with only evidential meaning (around two-thirds of the verbs, e.g. eššẽː ‘put on’ hats, headscarves, etc.)
iii. verbs with a resultative or indirect evidential meaning (around 33% of the verbs, e.g. heƛ’ã ‘dress’ for lower-body cloths and shoes, b-išši ‘keep, catch, hold’). For instance, depending on the context (1) has two different readings.
Similarly, the Kubachi perfect of the intransitive verbs ‘sit’, ‘stand’, ‘get tired’, ‘get hungry’, ‘become full’, and ‘convalesce, recover’ has only the resultative reading and no evidential value whatsoever (Magometov 1963: 196). The same is true for certain Avar intransitive verbs (e.g. ‘get sick’, ‘die’) which can therefore freely be used with first person subjects (2). These verbs never express indirect evidentiality when inflected for perfect. See also Maisak and Merdanova (2002, 2016) for similar observations on Agul.
For Lak, Friedman (2007: 362) claims that the meaning expressed depends on the alignment. The ergative construction triggers the evidential reading (3a). In contrast, the bi-absolutive construction triggers the resultative meaning (3b). In these two sentences, the constructions are only evident from the form of the copula since the first person singular pronoun does not distinguish ergative from absolutive. In the ergative construction, the copula agrees with O (prefix b-, no person marker) in gender and number. In the bi-absolutive construction, the copula agrees with the agent in person (suffix -ra) and in gender and number (zero prefix for masculine singular gender).
This claim needs to be checked because it would go against the general meaning of the bi-absolutive construction. Normally, bi-absolutive constructions topicalize the transitive subject, and the object (semantic patient) is often an indefinite NP and backgrounded (Forker 2012). To the contrary, for a resultative reading the patient NP is central since it is the resulting state of the referent of this NP that is expressed.
The resultative meaning is semantically close to an impersonal construction. In written languages such as Avar or Standard Dargwa it is frequently used in newspapers and journalistic writings. In this genre, transitive verbs are used without an overt agent and the focus of the described action is on the result. Example (4) is taken from a newspaper article about the opening of a new kindergarten.
When the perfect-like and related verb forms express indirect evidentiality, then it is inference based on sensory (visual) evidence or hearsay. The sensory evidence can be a result of the action referred to by the verb or some other evidence that does not automatically count as a result. Thus, (5a) exemplifies inference from visual evidence and (5b) hearsay since it is part of the oral knowledge about the history of the Godoberi village. I did not find any clear examples of indirect evidentiality expressing inference based on pure reasoning or general knowledge.
The indirect evidential function of the perfect shows up in different genres, especially in traditional folktales, accounts of historical events, religious texts, anecdotes, etc. In many examples, the indirect evidential reading is only an implicature and can be cancelled. For example, in Sanzhi Dargwa the perfect can be used when the speaker is judging from some (p. 496) traces (e.g. a lot of water on the kitchen floor and a wet towel) that somebody has washed the dishes. To express or better imply that s/he did not observe this event but inferred it based on evidence s/he would use the perfect. However, the same verb form can be used with a follow-up clause stating explicitly that the speaker saw the dishwashing event, in which case the evidential implicature is cancelled (6).
In most of the languages, the perfect series can only occur in independent main clauses. This is to be expected because cross-linguistically evidentials are often restricted to main clauses and may have scope over, but cannot occur themselves within subordinate clauses (Aikhenvald 2003a: 17). Yet there is at least one exception to this rule. In Ashti Dargwa the perfect can be used as part of periphrastic conditional clauses and preserves its indirect evidential reading. When these clauses are used without the apodosis they express wishes. Example (7) can be used by a speaker who gave his clothes away in order to have them washed, but he does not see himself the process of washing.
In the same language, we find a correlation between indirect evidentiality and aspect that is at the first glance surprising. In Ashti Dargwa, as in all other Dargwa varieties, aspect is marked through stem allomorphy. Almost all verbs have an imperfective and a perfective stem. Imperfective verbs preferably express non-firsthand evidentiality (cf. Sumbatova and Mutalov 2003 on Icari; Belyaev 2012 on Ashti). This is typologically unexpected since it is frequently the perfect or perfective tenses that acquire evidential meanings or develop into indirect evidentiality markers (Aikhenvald 2004a: 112–16: 264). For instance, in Icari only the perfect and the pluperfect (called ‘evidential present’ and ‘evidential past’ respectively) of imperfective verbs convey indirect evidential meaning (Sumbatova and Mutalov 2003: 88–9). If the same paradigmatic forms are used with perfective verbs the evidential meaning is absent. Belyaev (2012) proposes the following explanation: the paradigmatic verb form (i.e. suffix(es) plus copula) conveys perfectivity as part of its meaning and optional non-firsthand evidentiality as it is typical for the perfect. When it is used with an imperfective verbal stem perfectivity cannot be conveyed, due to the imperfective aspectual semantics of the stem and only the non-firsthand evidential value remains and is therefore more prominent with imperfective verbs. This explanation is plausible for Ashti since in this language the perfect and the pluperfect of imperfective and perfective verbs express indirect evidentiality and the perfect/resultative meaning can be expressed only with perfective verbs. Nonetheless, this analysis does not work for Icari because it would (p. 497) imply that the perfect and the pluperfect when used with perfective verb stems also carry indirect evidential meaning. This does not seem to be the case according to the full grammar of the language (Sumbatova and Mutalov 2003: 86–8). For Icari, it rather seems that perfect and pluperfect when used with imperfective verbs have evolved a special resultative meaning that has acquired an evidential extension. The latter is unavailable for the perfective verbs.
When the perfect series is used with the evidential meaning we get the ‘first-person effect’. Normally, indirect evidentials cannot be used with first-person arguments. If they are used, the reading must change and the first-person argument is interpreted as an unconscious or involuntary actor that only post factum got to know about his/her actions (8).
In Sanzhi Dargwa, the referent of the first person pronoun is not necessarily the subject or a subject-like argument, but can be an oblique argument. This appears to contradict Curnow’s (2003: 43) claim on the correlation between indirect evidentiality, person, and volitionality. Curnow states that ‘The notion of non-volitionality arises only in sentences which have a first person subject’. However, the first person singular pronoun in the comitative case in (9) cannot be analysed as the subject, but only as an involuntary, unintentional participant in the situation.
In most of the languages, the indirect evidential meaning does not imply that the speaker is uncertain about the uttered situation or does not vouch for the truth of his/her utterance. Hence, it is not the degree of speaker commitment that is at stake but only the source of the information. Yet Tsakhur seems to be an exception to this rule. The verb forms from the perfect series are used in certain contexts with direct evidence, for instance when referring to unexpected situations. Example (10) was uttered by someone who personally attended the situation but was surprised and maybe even shocked about the fact that he received only a scarf and not woollen socks, because woollen socks are considered to be more valuable than a scarf and would be an appropriate gift for a close relative.
Maisak and Tatevosov (2007) claim that what both contexts (i.e. indirect and direct evidentiality) have in common is a distancing effect that has also been described by Slobin and Aksu (1982: 196–7) for Turkish -miş. The speaker wants to distance herself/himself from the event referred to. At the same time, s/he implies that the proposition does not denote undoubted and well-established facts. In other words, these verb forms also express epistemic modality. In the appropriate context, they can also have mirative connotations as is the case for (10). In addition, they are used for recounting dreams which are described as seen during sleep and with negated verbs of knowledge, thought, perception, or liking (Maisak and Tatevosov 2007).
23.2.3. Grammaticalized verbal evidentiality
In Nakh-Daghestanian languages, grammaticalized evidentiality as part of the verbal paradigm is rather rare and all of the discussed verb forms up to now are instances of evidential strategies. However, a few languages have verbal forms conveying indirect evidentiality. These are the Tsezic languages, some Dargwa varieties, Ingush, Chechen, and Avar.
In the Tsezic languages, most or even all verb forms from the perfect series have developed into genuine indirect evidentials.4 In Bezhta, Khwarshi, Hinuq, and Tsez, the perfect is formed by simply adding an inflectional suffix to the verb without making use of the present tense copula.5 In Bezhta and Khwarshi, the copula can be added to the inflected verb, but then the meaning is perfect/resultative and the indirect evidential reading is lost (Khalilova 2011). The meaning of the indirect evidentials is inference, usually from visible results or traces (11a,b), and hearsay. When used with first person subjects we get the expected shift in the semantics towards an involuntary agent, often with a mirative flavour. For instance, (11b) is part of a traditional story about Mullah Nasreddin who much to his surprise discovers that he did not die after his donkey had brayed three times, though he expected this to happen.
For Khwarshi, Bezhta, and Tsez, the opposition between the perfect series and the other past forms has been described as marked indirect versus marked direct (Khalilova (p. 499) 2011)6 whereas for Hinuq an analysis by Forker (2014) shows that it is in fact marked indirect versus unmarked neutral.
In all four languages, the marked direct/unmarked forms occasionally occur in traditional fairy tales and other unexpected places, and similar switches from indirect to unmarked have been reported for Bagvalal (Tatevosov 2007b: 372). Comrie and Polinsky (2007) argue for Tsez that the switch between indirect evidentials and direct evidentials is at least partially motivated and can be explained through the assumption of several layers of witnesses. For instance, the marked direct may be used to add more vividness, and when a referent that is part of the story has seen the narrated situation and later talks about it. However, their claim is not supported by the data that they present in the appendix: one story is told from a first-person perspective, so obviously the marked direct is used exactly in the expected way. In the other story, there appears to be an arbitrary variation between indirect and direct evidentials with no clearly identifiable rules.
The Ingush Non-witnessed Inferential and the Non-witnessed Inferential Past have indirect evidentiality as part of their core meaning and can only occur in non-firsthand contexts. They are not only used in inferential contexts, but also express other types of indirect evidentiality. Example (12) has been uttered by an interviewer in a reply to a description of a place from A’s childhood. The interviewer (=B) infers from A’s description that the brother must have been at the place they were talking about and has seen it firsthand.
Similarly, the pluperfect in Sanzhi Dargwa and in Ashti Dargwa when formed with imperfective verbs (see example (9) in §23.2.2 and subsequent discussion) always expresses indirect evidentiality (Belyaev 2012).
Avar has an evidential construction not attested in any other Nakh-Daghestanian language. The past participle when heading an independent clause and other periphrastic forms derived from it can be used to express hearsay evidentiality.7 The indirect evidential meaning of the past participle is reserved for planned and organized stories such as traditional fairy tales, and is not very common even there (13).
Lastly, the verbal evidential system of Chechen deserves a short discussion since it is unique among the Nakh-Daghestanian languages. Chechen has a specialized copula xilla that conveys indirect evidentiality (hearsay, inference from visible evidence) as part of its core meaning (Molochieva 2011: 213), thus representing grammaticalized verbal evidentiality. The copula is inflected for the perfect or the remote past, but these inflectional forms are described as expressing direct evidentiality that is overridden by the lexical meaning of the copula. The periphrastic verb forms generated by employing the evidential copula occur in main clauses and can also be used in three different types of subordinate clauses: (i) in the protasis of irrealis conditionals; (ii) in certain adverbial clauses such as temporal (simultaneous), concessives, comparative, locative, and a few other adverbial clauses (see Molochieva 2011: 234 for the complete list); and (iii) in headless relative clauses. Sentence (14a) illustrates a headless relative clause with the nominalized participle of xilla. The use of xilla indicates that the speaker did not see the shooting. Molochieva (2011: 239) also notes that indirect evidentiality in Chechen can be expressed in imperatives of causativized verbs. For example (14b) can be uttered in a situation in which the speaker requests the addressee to prepare the cheese bread while she (the speaker) is absent. The process of making cannot be observed by the speaker, but the result should be obtained before her return.
Many Nakh-Daghestanian languages have periphrastic indirect evidentials with an auxiliary or light verb meaning either ‘find, come across, discover’ or ‘become, be, be at, stay, remain, stand’. The precise morphosyntactic properties, the functional range and the frequency with which the constructions are attested in natural texts differ from language to language. But since there is nevertheless a sizable number of common properties across the various languages, I assume that it is possible to speak of a construction type and treat the language-specific instances together. In Avar and the Tsezic languages the construction is not very common. In contrast, in Dargi languages it is rather frequent and a characteristic stylistic device in folktales. Examples of the languages and the involved light verbs are: Ingush (the auxiliary xu(r)g-, called delimited ‘be’ in the grammar by Nichols (2011)), the Lezgic languages Archi (χos ‘find, come across, discover’: Kibrik 1977: 238–43, 1994: 338–9) and Agul (xas ‘become, stand, get’: Maisak and Merdanova 2002), the Andic language Bagvalal (-isã (p. 501) ‘find’: Maisak and Tatevosov 2001), Avar (-at- ‘find, happen, be, turn up’: Forker in preparation (a)), Khwarshi (-us -/-ɨs - ‘find’: Khalilova 2009: 231–7), Hinuq (-aši- ‘find (out), come across’: Forker 2014), and the Dargi languages Icari (Sumbatova and Mutalov 2003: 109–10), Sanzhi, Kubachi (Magometov 1963: 196), Kajtag (Temirbulatova 2004) that have the verb -už- ‘be, be at, stay, remain’, and Standard Dargwa (van den Berg 2001: 45), and Mehweb Dargwa that have -uʔ- ‘be, be at, stay, remain’ (Magometov 1982: 96).8
The light verbs are also used as lexical verbs with their literal meaning. Verbs translating with ‘find, come across, discover’ belong to the valency class of affective verbs. This means that they typically take an experiencer subject marked with the dative case or another oblique case, and a stimulus or theme-like object in the absolutive case. In many languages, the light verbs are additionally used in epistemic probability constructions (Forker in preparation (b)), in conditionals and in concessives (e.g. Archi, Hinuq, Tsez, Bezhta, Avar, Bagvalal, Ingush).
Typically, the indirect evidential constructions involving these light verbs have the meaning of inferentiality from direct, visible evidence. This means that the speaker directly observed or discovered the result of an event and then made an inference about that event (15a,b).
In some languages such as Bagvalal the construction can therefore not be used for hearsay evidentiality. It is possible, but by no means necessary, that the observation or discovery of the evidence and the connected inference correlates with surprise on the part of the speaker. This is reflected in the standard Russian translation of this construction with the verb okazyvat’sja ‘find (oneself), turn out, prove, appear’ which can also indicate surprise alongside its indirect evidential meaning. Thus, in some descriptions it is called ‘admirative’ (cf. the Archi grammar by Kibrik (1977) and the account of Agul by Maisak and Merdanova 2002).
In the Dargi languages, the situation is slightly different, probably due to the literal meaning of the light verb. For Standard Dargwa the construction has been described by van den (p. 502) Berg (2001: 45) as expressing that ‘the activity is inferred from general knowledge or from hearsay’. In other Dargi varieties, we can observe that this construction expresses hearsay evidentiality and inference from general knowledge or observable results. It is frequently used in traditional narratives (16a), but also in historical narratives about (presumably) real events. It regularly occurs at the beginning of fairy tales and traditional stories, e.g. in the phrase ‘once upon a time’ (16b) and in similar formulaic expressions. Then the story can continue without the auxiliary.
Ingush has two special verb forms involving not the normal present tense copula, but the future tense/finite conditional form of delimited ‘be’ as auxiliary (‘delimited’ is a special Aktionsart type). These forms express inference not from sensory evidence but based on pure reasoning and logic (17). They are often used in consequence (apodosis) clauses of irrealis conditionals.
23.2.5. Direct evidentiality within the verbal system
According to accounts of Khwarshi (Khalilova 2009: 221–9, 240–1), Tsez (Comrie and Polinsky (2007), Ingush (Nichols 2011: 249–50), Chechen (Molochieva 2011: 216–18), Archi (Kibrik 1977: 238–43, 1994: 338–9), and Bagvalal (Maisak and Tatevosov 2001: 307–12), these languages have verbal forms that express direct evidentiality. The meaning is often described as giving preference to visually acquired knowledge, i.e. the speaker is an eyewitness of the described situation (cf. Khalilova 2009: 221; Nichols 2011: 249).
For Khwarshi, Tsez, Ingush, and Chechen, the direct evidential verb forms are part of the verbal paradigm and have past time reference. They are predominantly used in everyday conversations, autobiographical narrations (18), and in reported speech within traditional narratives. (p. 503)
Remarkably, Tsez as well as Hinuq have a special suffix used only with verbs inflected for the witnessed/unmarked past series in questions (cf. Comrie and Polinsky 2007 on Tsez, and Forker 2014 on Hinuq).
Nonetheless, the analysis of the mentioned verb forms as markers of direct evidentiality remains slightly doubtful. In Khwarshi, Tsez, and Chechen, the direct evidential forms occasionally occur in contexts where one would expect indirect evidentials, e.g. in traditional folktales (see the remarks in §23.2.3 on Comrie and Polinsky 2007), accounts of historical events that the speaker did not witness personally (Khalilova 2009: 224–5), or when drawing inferences from evidence. For instance, the following example (19) from Chechen has been uttered by a speaker who was sitting at the table drinking tea with another person. The speaker turned away from the table, and then back again after a few moments, when s/he saw that his cup was empty. Therefore, it might be possible to analyse these forms as semantically not expressing direct evidentiality, but as strongly implying such a meaning that nevertheless can be overridden.
The verbal forms labelled ‘direct evidentials’ (or marking direct evidentiality) in Archi and Bagvalal are reminiscent of the inferential evidentials since they make use of the same auxiliaries χos ‘find, come across, discover’ (Archi) and -isã ‘find’ (Bagvalal), but the two constructions differ in a few morphosyntactic properties. For example, if the lexical verb takes a converb suffix, the future participle suffix, or the preterite suffix (20), then the construction expresses direct evidentiality (Maisak and Tatevosov 2001: 308).
However, the analyses of both languages can be called into question. Some of the provided examples contain overt subjects of the auxiliaries that are distinct from the subject of the lexical verbs. Consequently, they may better be analysed as complement constructions of (p. 504) the verb ‘find’ (cf. Kibrik 1994: 339). For other examples it is unclear if the speaker actually witnessed the event or rather its result(s). In the latter case we would have an instance of the inferential construction. Thus, example (20) permits both interpretations. In fact, Kibrik (1994: 338) provides the following description of the Archi construction ‘Someone is witnessing part or the result of P’ (= the event/situation). Therefore, a plausible hypothesis seems to be to suppose that Archi and Bagvalal like the other languages discussed in §23.2.4 have only one construction with the light verb ‘find’ conveying indirect evidential semantics based on visual evidence. As a borderline case, its meaning can include the end of the relevant situation and always includes its later visual traces.
23.3. Evidential enclitics and suffixes
A number of Daghestanian languages have evidential enclitics and suffixes in addition to their verbal evidential systems (Avar, Godoberi, Tsakhur, Archi, Kryz, Mehweb Dargwa, Hinuq, Tsez, and Khwarshi). The origins as well as the functions of these enclitics and suffixes are quite diverse. Therefore, most of them will be discussed individually.
Archi (Lezgic) has the indirect evidential suffix -li that can be added to past tenses only (e.g. to the aorist) leading to the meaning ‘speaker and/or addressee were not eyewitness to the action X before the moment of speech’ (Kibrik 1977, 1994: 329). The dominant meaning is inference, but it is frequently used in traditional narratives with hearsay evidentiality (21).
The language also has a perfective converb marker -li, which is used in adverbial clauses and for the formation of the perfect. According to Tatevosov’s (2001a) analysis, there is only one suffix -li with the (perfective) converb meaning that occurs as part of the perfect series with the copula dropped or, if one likes, headed by a zero copula. Thus, the Archi evidential suffix is also an exponent of the perfect series that was presented in §23.2.2 as the typical verbal evidential strategy attested in many Nakh-Daghestanian languages.
Under certain circumstances the suffix has only the meaning of a proper perfect and can therefore be used with first persons: if the speaker participates in an action/situation that is unknown to the addressee or whose reasons or causes are unknown to the addressee. For instance, (22a) can be uttered when the addressee does not know that the speaker hates her, and (22b) as an explanation to the audience who does not know why the speaker brought the people.
Kryz, another Lezgic language without grammaticalized verbal evidentiality, has borrowed the Turkic evidential suffix -miš from Azeri for the expression of inference and hearsay. The suffix is added to verbs. In converb constructions it is only suffixed to the finite verb and has scope over the whole utterance. It is compatible with most verb forms, but not with the aorist, perfect resultative and progressive constative, which generally have direct evidential value. The suffix is almost exclusively found at the margins of texts (beginning or end), or to report narrative setbacks. The indirect evidential suffix can also convey surprise (23).
Tsakhur (Lezgic) has an evidential enclitic =ji that indicates the acquisition of knowledge about a situation on the part of the speaker. It does not imply doubt. To the contrary, the speaker is committed to the truth of the proposition marked with =ji. It is compatible with past and present time reference. The enclitic has two evidential meanings: (i) indirect evidentiality with an obligatory mirative interpretation, and (ii) direct evidentiality.
The first meaning is only available in combination with perfective verb forms. When used with the first meaning the enclitic expresses hearsay (24) or inference based on tangible consequences. There is a clear first-person effect to the extent that a first person subject is reinterpreted as unconscious or involuntary agent.
When used with the second meaning, =ji highlights the resultant state after the acquisition of knowledge (25). The information expressed in such an utterance must be within the personal knowledge sphere of the speaker. For example, the enclitic can only be employed to denote situations that took place during the lifetime of the speaker.
This combination of contradictory meanings (direct and indirect evidentiality) expressed by one and the same enclitic seems to be somewhat inconsistent and unusual, and it is not attested in any other Nakh-Daghestanian language (though see the discussion in §23.2.5 on the possible relationship of the inferential construction with direct evidentiality). However, it can occasionally be found in other languages outside of the Caucasus. For instance, the St'át'imcets particle lákw7a expresses a direct non-visual information source and indirect inference from evidence (Matthewson 2011). Similarly, Korean -te combines direct and inferential evidentiality (Lim 2012).
The evidential enclitics in Avar, West Tsezic, Mehweb Dargwa, and Lezgian express only hearsay. They are clearly distinct from the quotative markers and can co-occur with them. The enclitics in Lezgian, West Tsezic, and Mehweb Dargwa probably originate from inflected forms of the verb ‘say’. For instance, the Mehweb Dargwa hearsay enclitic =k’ʷan has grammaticalized from -ik’ʷ- ‘say.imperv’ plus general tense suffix for third person -an. In the traditional folk tales published in Magometov (1982), =k’ʷan occurs frequently at the beginning of the stories (26).
The Lezgian hearsay marker is =lda (Haspelmath 1993: 148) and goes back to the verb luhun ‘say’, just like the quotatives luhuz and lahana (Haspelmath 1993: 367). The West Tsezic hearsay evidential enclitics are =ƛo in Khwarshi, =ƛax in Tsez, and =eƛ in Hinuq. In Hinuq and Tsez, the enclitics often occur together with the unmarked/direct evidential verb forms (27). According to speakers of Hinuq, this adds more vividness to the narrative. As can also be seen in (27), the Tsez quotative enclitic =ƛin can freely co-occur with the hearsay evidential and only marks reported speech.
Khwarshi has an additional hearsay construction with the fossilized general tense form č’aːl of the affective verb ‘to inform, to hear’ (28). The construction forbids the use of certain verb forms such as the witnessed past (a direct evidential form with past time reference) and the definite future. The quotative particle can optionally occur together with the hearsay construction.
The Avar enclitic =ila can be added to all verbal forms, including the verb forms that already express indirect evidentiality (perfect series, past participle) and verb forms not having past time reference and/or not having an evidential meaning. According to Charachidzé (1981: 135), it indicates that the congruence between the assertion and the reality is uncertain. It is frequently found in traditional folk tales (29), but also in other contexts expressing hearsay evidentiality. It is probably a cognate of the quotative particle =ilan.
Finally, for a number of Nakh-Daghestanian languages what looks like quotative particles have been analysed as markers of hearsay evidentiality. To these languages belong Agul (Maisak and Merdanova 2002), Archi (Chumakina 2011), and Ingush (Nichols 2011: 249, 279–80, 559–60). However, since in all these languages the same particles are also used as markers of reported speech without any necessary implication of indirect evidentiality, and in Archi and Ingush the markers still inflect like other verbs because they originate from verbs of speech, I do not consider them to have hearsay evidentiality as part of their meaning.
23.4. Expressing knowledge by other means
Some Nakh-Daghestanian languages have constructions whose meaning does not directly evoke the information source, so they cannot be said to express evidentiality. However, they make reference to the state of knowledge of the speech act participants (speaker and addressee) and their possible status as epistemic authority. These constructions can be said to partially overlap with evidentiality and therefore deserve a short discussion.
In Axaxdərə Akhvakh (Andic), the affirmative perfective participle -ada is used in independent declarative clauses with first person agentive subjects and in independent interrogative clauses with second person agentive subjects. In contrast, second and third persons in assertions as well as first and third persons in questions take the regular affirmative perfective suffix -ari (30a, b). According to Creissels (2008b) in assertive clauses the affirmative perfective implies that the speaker has direct knowledge of the situation.
This is commonly called egophoric marking or a conjunct/disjunct system (Creissels 2008b). In conjunct/disjunct systems, the speaker is the epistemic authority in assertions and the addressee is the epistemic authority in questions. A similar system is attested for Mehweb Dargwa (Magometov 1982: 119–20) and Zakatal Avar (Saidova 2007).
The Nakh languages Chechen and Ingush have phonologically reduced second person (and in Chechen even first person) pronouns in the dative and in Ingush also in the genitive that have morphosyntactic and semantic properties not typical for dative and genitive case (Nichols 2011: 280–3, Molochieva 2011: 244–8). The pronouns bear some similarity to ethical datives and have evolved from free non-argument benefactives (Molochieva and Nichols 2011). They are used to announce something new or important and unexpected for the addressee or the speaker or to indicate an important generalization that is known to both speaker and addressee but not in the addressee’s immediate consciousness. Thus, the information is usually not new for the addressee and the speaker seeks confirmation of her/his assumptions. For instance, (31a) has been uttered by a man who informed his wife that he did not like the future bride of his son. Example (31b) states a fact about livestock breeding that is part of the general knowledge. By uttering this sentence the speaker invites the addressee to agree with her/him on that fact.
As a final point, I briefly mention epistemic modality because it is often discussed together with evidentiality. In Nakh-Daghestanian languages, evidential constructions are usually formally and functionally distinct from epistemic modals. The latter occur in a number of different constructions (see Forker in preparation (b) for a short overview):
• epistemic probability with a light verb/auxiliary (in some languages this is the same verb also used in the inferential construction)
• epistemic necessity with the verb ‘must’
• within the verbal paradigm (irrealis verb forms, future-in-the past, etc.)
As was mentioned throughout this chapter, evidentials normally do not imply any doubts on the part of the speaker concerning the truth of the proposition, and the speaker is not less (p. 509) committed to his/her utterance when s/he uses an indirect evidential. The only exception seems to be the perfect series in Tsakhur (see §23.2.2).
In a nutshell, the expression of evidentiality in Nakh-Daghestanian can be summarized as follows: grammaticalized evidentiality as well as evidential strategies show up as part of the verbal inflectional system, usually conflated with tense. Most languages have fairly standard indirect evidentials based on the perfect series. Another common trait is inferential constructions with light verbs ‘find’, ‘stay’, or ‘be, become’ and some languages have evidential enclitics or suffixes. The verbal evidential systems usually express indirect evidentiality (hearsay and inference), though some authors claim that a number of languages also have direct evidentiality. Promising topics for future research include, among others, the relationship between the imperfective aspect and indirect evidentiality attested in Dargwa varieties and the connection between alignment and evidentiality found in Lak.
I thank Timur Maisak and Alexandra Aikhenvald for comments, suggestions, and corrections. The first version of this paper was written during my period as a Feodor-Lynen Fellow at James Cook University (Cairns). I am grateful to the Humboldt Foundation for financial support and to Alexandra Aikhenvald for hosting me at the inspiring Language and Culture Research Centre.
(1) Nichols measures the overall complexity, including phonology, morphology and syntax.
(2) By resultative, I mean reference to the state that obtains as a result of a preceding action, and by perfect I mean reference to a past action that is relevant at the moment of speech.
(3) Note that the individual descriptions and grammars may use different labels for these forms. Thus, in Agul the verb forms treated here as belonging to the perfect series are called resultative and past resultative general factive (Maisak and Merdanova 2002).
(4) Nevertheless, even in these languages one or two verbs represent exceptions to this rule (e.g. ‘want’ plus perfect indicating a present state of wanting in Hinuq (Forker 2013: 222), Bezhta and Hunzib).
(6) Khalilova (2011) analyses Hunzib along the same lines. This contrasts with the Hunzib grammar by van den Berg (1995) according to which not the perfect itself but only verb forms containing the perfect form of the copula zuq’on lo as auxiliary have indirect evidential value.
(7) Other functions of the past participle are the formation of (i) relative clauses, (ii) interrogative clauses, (iii) term focus constructions, and (iv) assertive modality.
(8) In Icari and Standard Dargwa, there are other light verbs in addition to the light verbs described that can also convey evidential meanings: Icari elɣ- ‘remain’, Standard Dargwa kal- ‘remain’ (Sumbatova and Mutalov 2003: 109–10). For Standard Dargwa, Mutalov (2002) writes that there is a small difference in the semantics between the two constructions depending on the auxiliary, but he does not provide arguments or examples in favour of his claim.