Inferential evidentiality in Hupda, Jarawara and Tariana: Evidence from narrative texts

Workshop Empirical Evidence for Evidentiality Nijmegen, 2014

Liisa Vilkki University of Helsinki

1. Introduction

• Many languages of North and South America have quite complex evidentiality systems

• For example, these languages may have several inferentials, expressing different types of inference as sources of information

• Inferentials of languages in North and South America may also be polysemous, indicating more than one sense or use

• Sometimes, inferential meanings may combine inferential properties with degrees of reliability of evidence and/or with epistemic degrees of certainty

• In this presentation, the focus is on the meanings of some inferentials of the endangered languages Tariana (Aikhenvald 2003), Hupda (Epps 2008) and Jarawara (Dixon 2004): semantic distinctions and polysemy patterns

• How could different kinds of encoding patterns be motivated? • Motivation: assumptions of the psychological theory of mental models

(e.g. Johnson-Laird 2006) and the notion of construal operations (e.g. Croft and Cruse 2004, Verhagen 2007, Langacker 2008)

• The presentation is based on my dissertation ‘Semantic typology of grammatically expressed epistemic modality and inferentiality’ (Vilkki forthcoming)

• Inferentiality: the types of inference (e.g. inference from observation, inference from results, inference from general knowledge, inference from previous experience)

• Epistemic modality: the speaker’s attitude to the factual status of the proposition in terms of the degrees of certainty

• A genealogically stratified variety sample of the languages of the world (Haspelmath & al. 2005)

• The survey of descriptive grammars and other descriptive material • All the grammatical expressions of epistemic modality and inferentiality from 130

languages • Not all the languages of the world have grammatical forms for epistemic and/or

inferential meanings • Every language of the sample have at least one grammatical form for epistemic

and/or inferential meanings

• In individual languages, inferential meanings may be conventional senses or contextual uses (based on the meaning potentials of inferential forms – cf. Langacker 2006)

• Typologists and other comparative linguists need standards for comparison (tertium comparationis) which involve semantic similarity and often formal similarity

• In this study, the notion of comparative concepts is used as the standards of comparison (cf. Haspelmath 2010)

• Comparative concepts are created by linguists for the specific purpose of cross-linguistic comparison

• They ignore the distinction between language-specific senses and uses • They are not psychologically real • They are more or less well-suited for the purpose of enabling cross-linguistic

comparison: classification, generalizations, motivations (explanations) • They are not necessarily universal, but cross-linguistically applicable

• The most important comparative concepts of this study are 33 functions which capture the relevant semantic similarity of language-specific meanings

• Most inferential and epistemic functions can be described by means of the two semantic parameters, indicating 1) the types of inference (and other sources of information) and 2) the degrees of the speaker’s certainty

1) The types of inference (included in the parameter ‘sources of information’):

P, Q: proposition; S: speech time 1. The speaker infers P on the basis of prior experience, general knowledge

or intuition (P is not after S) 2. The speaker infers P from observation of Q (P and Q are “simultaneous”) 3. The speaker infers P from observation of the results Q (P is before Q) 4. The speaker infers P from observation of Q (P is after Q) / The speaker infers P on the basis of prior experience, general knowledge or intuition (P is after S) 5. The speakers infers P from the report of Q

• The functions based on the values of the parameter 1)‘the types of inference’:

1. Inference from memory

2. Inference from simultaneous observation

3. Inference from results

4. Inference about future

5. Inference from report

2) The degrees of the speaker’s certainty:

1. Doubt

2. Weak possibility

3. Possibility

4. Probability

5. Certainty

6. Emphasized certainty

• The inferential functions may be purely inferential, based on the 1) parameter.

• They may also combine values of the parameters 1) and 2)

• (Purely epistemic functions are based on the parameter 2)

2. Some semantic distinctions and polysemy patterns of inferentials in Tariana, Hupda, and Jarawara:

Narrative texts

• Tariana (Aikhenvald 2003): Arawakan -> Arawakan, (family -> genus), Vaupés region of the Brazilian Amazon, 100 speakers

• Hupda (Epps 2008): Vaupés-Japurá -> Vaupés-Japurá, Vaupés region of the Brazilian Amazon, 1900 speakers

• Jarawara (Dixon 2004): Arauan -> Arauan, seven villages near the Purús River in the Brazilian Amazon, 170 speakers

• Examples of purely inferential functions: Aikhenvald (2003:300): Tariana 1. walikasu-nuku hiku-sina di-keña beginning-TOP.NON appear-REM.PST.INFR SG3.NF-start hĩ Yapiɾikuɾi walikasu-naku nawiki-naku DEM.ANIM God beginning-TOP.NON people-TOP.NON di-ñaleta-tha-sina nawiki sede-sina SG3.NF-disappear-CAUS.FR-REM.PST.INFR people NEG.EXIST-REM.PST.INFR ’In the beginning this was how it started. The (traditional) god destroyed the people at the beginning, there were no people.’ (-sika/-sina: ’I infer it on the basis of general knowledge, prior experience, intuition, or common sense’) - Inference from memory:

The speaker infers P on the basis of prior experience,

general knowledge or intuition (P is not after S)

Aikhenvald (2003:302): Tariana

2. Diha-mha

he-PRS.NONVIS/INFR

‘It is him (nonvisual) (we infer it because of his

characteristic gait).’

- Inference from simultaneous observation:

The speaker infers P from observation of Q (P and Q are

“simultaneous”)

Aikhenvald (2003:301): Tariana 3. nese diha di-sawi ithani-nhina then ART SG3.NF-root near-REM.PST.SPEC.INFR paita ma:mi na:-nite one.ANIM tinamou call.PL3-TOP.ADV.ANIM di-swa-nhina di-ehweta-ka SG3.NF-stay-REM.PST.SPEC.INFR SG3.NF-lay.egg-SUB ’Then at the root of this (tree) one (bird) called tinamou was staying laying eggs.’ (-nihka/-nhina: ’I infer it on the basis of obvious evidence’) - Inference from results: The speaker infers P from observation of the results Q (P is before Q)

• In Hupda (Epps 2008:650-651), the enclitic

-cud covers all the functions, illustrated in

examples (1)-(3)

• (Some of the Maricopa (Gordon 1986:189-190) inferential constructions make fine-grained semantic distinctions, such as inference of prior events that is based either on the subject’s state, or the subject’s actions, or the way the subject sounds)

• Examples of the combination of an inferential property with epistemic properties: Aikhenvald (2003:388): Tariana 4. nu-dalipa i-nu-ka-tha-sika-da SG1-near PL2-come-DECL-FR-REC.PST.INFR-UNCERT ihie i-nu enter.PL2 PL2-come ’You probably came to visit me, to my frustration, (but is this so?), come in.’ - Inference from memory and weak possibility-possibility-probability

Epps (2008:111): Hupda

5. húp kǝwǝg-pog-cudʔũh

person eye-EMPH-INFR.EPIST

‘It must have been a person’s eye (I saw)’

6. wᴐh-ʔȋh-cudʔũhníy

River.Indian-M-INFR.EPIST

‘I guess he was a River Indian’.

- Inference from memory and weak possibility-possibility-probability

• Examples of purely inferential functions (Jarawara): Dixon (2004:205): Jarawara 7 a. moto ati ka-tima-re-ka motor.boat.M noise in.motion-upstream-IMM.PST.FIRSTH.M-DECL.M ‘the noise of the motor boat was coming upstream (the noise could be heard)’ b. moto ati ka-time motor.boat.M noise in.motion-upstream ’the noise of the motor boat comes upstream’ c. moto ka-time-no motor.boat.M in.motion-upstream-IMM.PST.NONFIRSTH.M ’the motor boat was coming upstream (although it could not yet be seen)’ - Inference from simultaneous observation

Dixon (2004:204): Jarawara

8. Wero kisa-me-no,

Wero get.down-back-IMM.PST.NONFIRSTH.M

ka-me-hiri-ka

in.motion-back-REM.PST.FIRSTH.M-DECL.M

‘Wero got down from his hammock (which I didn’t see, and

went out (which I did see)’

- Inference from results

• An example of the combination of an inferential property with epistemic properties:

Dixon (2004:222): Jarawara 9. otaa to-ka-tima-habone, away-in.motion-upstream-INT.F moto otara kaa-kosa-ka-no boat APPLIC.in.motion-MIDDLE-DECL.M- IMM.PST.NONFIRSTH.M ‘we need to go upstream, the boat might have already come and left us (here)’

- Inference from memory and weak possibility-possibility

• -(ha)ni/-(hi)no , as well as -(he)te/-(hi)ta and -(he)mete/-(hi)mata: non-firsthand evidence • In specific contexts, more specific uses: inferential uses, ’situation not

seen by the speaker and in which she did not participate’, ‘intrinsically uncontrollable situations’, ‘narrator dissociates herself from responsibility’, ‘mirativity’

• An example of a mirative use: Dixon (2004:206): Jarawara 10. bahi to-ke-hino sun away-in.motion-IMM.PST.NONFIRSTH.M ’the sun is (surprisingly to me) going away (i.e. setting)’

3. Motivations: the mental models and construal operations

• From the typological perspective, language-specific meanings that fall under the functions ‘inference from results’ and ‘inference from memory’ are often formally distinguished, and this suggests that human beings are sensitive to this kind of distinction

• We saw above that Tariana is one of the languages that treat these functions differently

• The basic inferential markers are enclitics fused with recent past and remote past: -nihka/-nhina indicates that the inference is based on visible results of some previous situation (‘inference from results’), whereas

-sika/-sina indicates inference grounded on general knowledge or prior

experience with similar situations (‘inference from memory’)

• Why do languages so often make this kind of distinction?

• Suggestion: Johnson-Laird’s (2006) psychological theory of mental models can be used in accounting for this differentiation

• In this theory, perception and knowledge that is stored in long-term memory show up as the most important sources for everyday inference

• Since language is arguably part of human cognitive faculty, it is quite natural that meanings indicating perceptual (e.g. -nihka/-nhina) and memorial (e.g. -sika/-sina) grounds for inference are frequently differentiated in linguistic inferential forms

• However, languages may have more fine-grained distinctions in the domains of perception and memory, when these domains are used as bases for inference

• For instance, Tariana also has present tense evidential enclitic -mha which primarily indicates nonvisual evidence but which can also be used to make perceptually grounded inferences with reference to the present (‘inference from simultaneous perception’)

• (Maricopa has several specific constructions, corresponding to ‘inference from results’)

• On the other hand, there are languages with inferential forms that do not make these kinds of distinctions

• For example, the Hupda enclitic -cud covers all the three functions (1)-(3)

• The identical encoding of language-particular meanings matching with these functions suggests that there is some conceptual similarity between these meanings

• Of course, these meanings are similar in that they all indicate inference as a source of information

• But how can we relate, from the psychological point of view, the form-meaning pattern found in languages like Hupda to the different kind of pattern exhibited by languages like Tariana?

• Suggestion: the Hupda pattern is also in conformity with the positions of the theory of mental models

• According to this theory, conclusions can be based not only on one type of ground, but also on several grounds at the same time

• In fact, the latter possibility, “heterogeneous reasoning” (Johnson-Laird 2006) is more usual in everyday reasoning situations

• Since reasoning starts sometimes with one kind of ground and sometimes with some mixture of grounds, it is not surprising that these related types of inference may be conveyed by the same linguistic inferential form

• The specific interpretation of the inferential form may be activated in actual contexts and situations

• It is possible to examine the relationship between different patterns in even more detail

• Why are there these kinds of differences?

• The theory of mental models does not provide any obvious assumptions which could be applied in trying to answer this question in a more profound way

• Suggestion: the notions of granularity and focusing, which are aspects of salience and attention, are important in accounting for diverse encoding possibilities both in the domain of inferentiality (and epistemic modality)

• Salience and attention: types of construal operations

• Construal is a cover term, used in cognitive linguistics for different possibilities of viewing experiences or phenomena (e.g. Croft and Cruse 2004:40-73, Verhagen 2007, Langacker 2008:55-89)

• The diverse encoding patterns in the semantic domains of epistemic modality and inferentiality can be partly explained by means of a set of construal operations

• Different kinds of classifications of construal operations have been proposed (e.g. Talmy 2000, Croft and Cruse 2004:45, Langacker 2008:85-86, Verhagen 2007:55)

• The classifications are motivated by the requirement that they should correspond closely to psychological processes and abilities postulated by cognitive psychologists and phenomenologists

• Granularity can be defined as the level of detail and precision at which a phenomenon is conceptualized or characterized (e.g. Croft and Cruse 2004:51-53, Langacker 2008:55-57)

• For example, the inferential forms of Tariana, (Maricopa), Hupda and Jarawara conventionally impose different construals on similar phenomena

• The speakers of Tariana have to pay attention to quite fine-grained distinctions in the domain of inferentiality when they use the forms

-nihka/-nhina ‘inference from results’, -sika/-sina ‘inference from memory’,

and –mha ‘inference from simultaneous perception’

• (The Maricopa speakers must make still finer distinctions, when they indicate ‘inference from results’)

• These distinctions are ignored by the speakers of Hupda in their use of the form –cud: this form indicates inference as a source of information in a less granular way, although it cannot be selected to express just any kind of inference, such as inference about future situations

• In order to use -cud appropriately, the speakers of Hupda need to attend to the possibility of expressing broadly conceived inference as an evidential specification

• The meaning potential of the Jarawara -(ha)ni/-(hi)no (etc.) is still broader, since it covers several kinds of non-inferential meanings, in addition to meanings, corresponding to ‘inference from simultaneous observation’ and ‘inference from results’

• Focusing, another important aspect of salience and attention, concerns the alignment of the selected conceptual content into foreground versus background (cf. Langacker 2008:57-65)

• A foreground is salient relative to a background

• Focusing is especially important in the explanation of the differences between the main types of epistemic and inferential functions: pure epistemic, predominantly epistemic, pure inferential, and predominantly evidential functions

• The uses of the Tariana forms nihka/-nhina ‘inference from results’ and

-sika/-sina ‘inference from memory’, discussed above in terms of

granularity, also provide an example of focusing

• The distinction between the meanings of these forms is real, but not sharp

• Aikhenvald (2003:287-288): the difference between these meanings lies in access to direct unmistakable evidence and to the degree of reasoning, based on general knowledge

• The less obvious the direct evidence available to the speaker and the more the speaker needs to apply reasoning, grounded on general knowledge, the more likely it is that -sika/-sina will be selected

• For instance, if the speaker sees a man with quite obvious signs of a dog’s teeth on his hand, she will use -nihka/-nhina in an utterance, indicating ‘The dog bit him’

• If the signs on the man’s hand are not unmistakable, the speaker will select -sika/-sina

• Suggestion: it is possible to interpret this example in terms of foreground and background

• In situations, where direct evidence is foregrounded in the inferential cognitive domain, the speaker selects nihka/-nhina

• General knowledge as a basis for inference remains in the background

• In contrast, when sika/-sina is selected, general knowledge is foregrounded, whereas direct evidence stays in the background

• Moreover, Aikhenvald’s description clearly indicates that the speakers of Tariana use direct evidence and general knowledge as grounds for inference to varying degrees

• This example thus shows that focusing comes in degrees

4. Conclusions

• Languages of North and South America provide a wide variety of encoding possibilities in the semantic domain of inferentiality

• In this presentation, I have focused on semantic distinctions and polysemy patterns of some inferentials in the endangered languages Tariana (Aikhenvald 2003), Hupda (Epps 2008) and Jarawara (Dixon 2004), and on the question of how these encoding possibilities could be motivated

• I have argued that different kinds of possibilities (those that were discussed) are in conformity with the assumptions of the psychological theory of mental models (e.g. Johnson-Laird 2006)

• In addition, I have proposed that especially the construal operation of granularity is important in accounting for diverse encoding possibilities in the domain of inferentiality

References:

• Aikhenvald, Alexandra 2003. A Grammar of Tariana. Cambridge: Cambridge University Press.

• Croft, William and Alan D. Cruse 2004. Cognitive Linguistics. Cambridge Textbooks In Linguistics. Cambridge: Cambridge University Press.

• Dixon, Robert M.W. 2004. The Jarawara Language of Southern Amazonia. Oxford Linguistics. Oxford: Oxford University Press.

• Epps, Patience 2008. A Grammar of Hup. Mouton Grammar Library 43. Berlin/New York: Mouton de Gruyter.

• Gordon, Lynn 1986. Maricopa Morphology and Syntax. University of California Publications in Linguistics 108. Berkeley/Los Angeles/London: University of California Press.

• Haspelmath, Martin 2010. Comparative concepts and descriptive categories in cross-linguistic studies. Language 86:663-687.

• Haspelmath, Martin, Matthew Dryer, David Gil, and Bernard Comrie, eds. 2005. World Atlas of Language Structures. (Book with interactive CD-ROM). Oxford: Oxford University Press.

• Johnson-Laird, Philip 2006. How We Reason. Oxford: Oxford University Press. • Langacker, Ronald W. 2006. On the continuous debate about discreteness. Cognitive

Linguistics 17:107-151. • Langacker, Ronald W. 2008. Cognitive Grammar: A Basic Introduction. Oxford/New York:

Oxford University Press. • Talmy, Leonard 2000. Toward a Cognitive Semantics. Concept Structuring Systems, Vol.1.

Language, Speech, and Communication. Cambridge, MA: The MIT Press. • Verhagen, Arie 2007. Construal and perspectivization. In Geeraerts, Dirk and Hubert

Cuyckens, eds. The Oxford Handbook of Cognitive Linguistics. Oxford/New York: Oxford University Press, 48-81.

• Vilkki, Liisa (forthcoming). Semantic typology of grammatically expressed epistemic modality and inferentiality.