Workshop in Philosophy and Linguistics, University of Michigan, 22 November 2013

Reference Resolution: Grammar or Pragmatics?

Commentary to Una Stojnic, Matthew Stone and Ernest Lepore, ‘Discourse and Logical Form’

Kasia M. Jaszczolt University of Cambridge

http://people.pwf.cam.ac.uk/kmj21

1

Reference resolution for ‘demonstrative’ pronouns (‘he’) is

governed by linguistic rules (is ‘a function of linguistic context’, ‘grammaticised’)

√ No pure indexical/demonstrative distinction, pace Kaplan

2

? Grammaticized or pragmatic

3

‘[P]ronoun resolution is guided by interdependencies that structure the flow of information.’ SSL, p. 3

Attention-Coherence Approach • Centering Theory (Grosz, Sidner) • Coherence, e.g. Hobbs; Kehler; Asher & Lascarides (SDRT) • Dynamic Semantics (Dekker)

4

‘[T]he referent of a demonstrative pronoun on an occasion of use is fully determined by rules of language.’ (SSL, p. 2)

(1) A man1 walked in. He1 sat down. Therefore, some man sat

down. (2) Phil1 tickled Stanley, and Liz poked him1/2. (Result or Parallel)

[from Kertz et al 2008>> Kehler et al 2008?]

5

Context: sequence of individuals ordered by relative prominence

including the centre of attention; an assignment function cf. Dynamic Semantics

6

Meaning = updates, a relation between an input context and output context

xi discourse referent in the ith position of prominence (i=1 >>

@, centre of attention) <α> = update <α> ; [man (@)] ; [walk.in(@)] ; [sit.down(@)] Therefore <α> ; [man (@)] ; [sit.down(@)]

7

‘he’ always denotes the most prominent candidate referent Discourse context has a structure that dictates reference

resolution

8

My main question:

• Has it been demonstrated that reference is a function of linguistic context?

• Has it been demonstrated that attention-shifting

mechanisms are linguistic/grammaticised?

• How are the choices between coherence relations built into the LF?

9

Example:

deictic reading of ‘he’ [pointing to Bill] (1) A man1 walked in. He1 sat down. <πb> act of pointing at Bill √ Gestures are ‘integral to the linguistic utterance itself’ SSL, p. 6

? ‘Since this change of attention affects interpretation, it should

be reflected in the LF.’ SSL, p. 5 10

>> gestures contribute to some level of conceptual representation (SSL: LF) but in what sense exactly are they governed by the knowledge of language?

‘[c]ontext interacts with gesture to determine the form of demonstration’ SSL, p. 7

In what sense is it grammaticised?

11

Observation: We can try to make the change of attention be reflected in the

LF but the interpretation of the pointing gesture has to involve a pragmatic process of intention recognition at some level or other.

cf. Sam Cumming, “The Attentional Foundations of Coherence” underlying cognitive principles; stimulus >> questions

12

√ ‘...only if we restrict the theory’s use of beliefs, intentions and

other ‘private’ features of the participants’ cognitive states – such as individual memory organisation or processing effort – will the theory be a linguistic theory with linguistic generalisations and explanations.’

Asher & Lascarides (2003: 76)

13

Kaplan’s (1989b, Afterthoughts): directing intention >> Perry’s (2009) ‘forensic element’ in what is said (locutionary, intended content)

14

Context

• SSL: context as a sequence of individuals, an assignment function

• Kaplan: metaphysical context, index/parameters • Stalnaker: epistemological context, things that have been

said/sets of beliefs, two dimensions

15

cf. J. Merchant (2004) on incomplete sentences an ellipsis account Do fragmentary utterances have sentential syntactic

structures? ‘A nice dress’ >> ‘It is a nice dress’, ‘With scissors’ >> ‘Do it with scissors’ otherwise non-propositional utterances ?A: ‘Facebook!’

16

‘A man1 met Sam2. He2 greeted him1’ ‘The city council1 denied demonstrators2 a permit. They2

advocated violence1.’ >>we need Coherence (Coherence can override 0,1,2 ranking) >>In what sense is it built into grammar?

17

Rhetorical structure rules to be represented in the LF: • How many rules are there? • What is their epistemic status? • Can we propose a new rule for each new type of

connection discovered? the types ‘explanation’, ‘elaboration’ and ‘comparison/

contrasts’ adopted from Kehler (2002) or SDRT?

18

Rhetorical cooperativity vs. Gricean cooperativity Asher and Lascarides 2013, ‘Strategic conversation’ ‘an implicature is safe when it can be reliably treated as a

matter of public record’ (p. 1)

19

Coherence relations necessitate representing shifts of

attention and inference. vs. building in post-Gricean heuristics (OT Pragmatics)

20

‘Coherence relations are being represented in the logical form because, crucially, their effect is delivered by the grammar, constrained by linguistic rules, and interrelated with other aspects of meaning.’ SSL, p. 15

? Is this ‘going beyond the structure as the source of

information restricted to filling in syntactic slots? minimalism/indexicalism/radical contextualism

21

SSL: Prominence and scale of attention arguably have a stronger explanatory power than pragmatic inference.

• I can’t see how it helps with examples (3) & (4):

(3) ‘A man1 met Sam2. He2 greeted him1’ (4) ‘The city council1 denied demonstrators2 a permit. They2

advocated violence1.’

lexical rules needed for assessing which rhetorical structure rules apply.

Is this really very different from post-Gricean heuristics? 22

• eo and es (explanation via subject/object) simply translate a

choice that is inherently pragmatic

• Do topic markers support the Attention-Coherence view? (overtly present in the grammar)

23

‘[G]rammar specifies a diverse set of resources for raising entities to prominence.’ SSL, p. 7 fn2

?? Possible mechanism: Demonstratives provide slots >>

argument structure provides defaults? >> the lexicon directs towards the right rhetorical structure rule?

24

Compositionality

A Big Assumption: Representing reference determination as a linguistic rule is to

be preferred over a pragmatic account. • Would compositionality be affected if we didn’t make this

assumption?

25

SSL’s Linguistic rules vis-à-vis A&L’s logics of conversation

• Rhetorical structure rules rely on ‘modular architecture of discourse interpretation’ (Asher and Lascarides 2003: xvii)

!Different logics for different kinds of reasoning: constructing LF (information content) and inferences (information packaging)

rhetorical relations (Narration, Explanation, Contrast, Parallel...) and the glue language, logic of information packaging (non-

monotonic)

26

A neo-Gricean alternative: Rethink content-character distinction allowing for characters

that correspond to the inferential base of the length that is appropriate for the context at hand

(‘fluid characters’, Jaszczolt 2012, ‘Pragmaticizing Kaplan’)

27

Challenge for intention-based contextualist accounts: Endless flexibility of meaning?

Interactive compositionality ‘Meaning eventually stabilizes, making compositionality possible, because the (linguistic as well as extralinguistic) context, however big, is always finite’. Recanati (2012: 190-1)

28

On the other hand...

SSL are correct in saying that pragmatic approaches have too little to say about the interaction of semantics and pragmatics.

But this is because pragmatic approaches assume ‘top-down’

modification of the LF (Recanati 2005) or even interactive building up of the representation (Jaszczolt 2005, 2010)

29

Q: Do post-Griceans resort to ‘brute coindexing’? (SSL, p.20)

A: No. Reference resolution intrudes in the semantic representation ‘top-down’. (Recanati)

Or: free merger of information, producing a representation

that treats all the sources of information on an equal footing (Jaszczolt)

30

How is ambiguity avoided on SSL’s account? √Reference is a function of context. But: The structure of the context relies on rhetorical structure the

recognition of which is a pragmatic process. Rhetorical structure rules capture standard conceptual links

between propositions (inferred in a pragmatic process or automatically assigned as defaults)

31

Is the flow of information accounted for ‘bottom-up’ or ‘top-down’?

‘he’: anaphoric/deictic Top-down >> Gricean pragmatic

32

Definite/indefinite distinction

• Does <α>; [man (@)] introduce a new d.r. For indefinites as well as definites?

• How is a cataphoric introduction of a referent to be handled?

33

• The promised discussion of the parallel account of ‘I’ and ‘he’ that would contrast it with Kaplan proceeds without a discussion of the indexical (‘I’)

While The properties of ‘I’ as an indexical should not be taken for granted.

34

?

‘he’ is like ‘I’ (grammar-driven) or ‘I’ is like ‘he’ (pragmatic-inference-driven)

35

? Grammar produces the self-referring function

Chierchia (1989: 28): The cognitive access to oneself is

‘systematically excluded from the interpretation of (non-pronominal) referential expressions. It is systematically present in the interpretation of overt pronouns. It is systematically and unambiguously associated with the interpretation of PRO the null subject of infinitives and gerunds. It is associated with the interpretation of long-distance reflexives (at least in some languages)’.

36

? Grammar produces the self-referring function

Chierchia (1989: 28): The cognitive access to oneself is

‘systematically excluded from the interpretation of (non-pronominal) referential expressions. It is systematically present in the interpretation of overt pronouns. It is systematically and unambiguously associated with the interpretation of PRO the null subject of infinitives and gerunds. It is associated with the interpretation of long-distance reflexives (at least in some languages)’.

37

An argument from non-pronominal expressions (but not the one you expect)

x Pace Chierchia, cognitive access to oneself is not so ‘systematically’ excluded from the interpretation of non-pronominal expressions:

‘Sammy wants a biscuit.’ ‘Mummy will be with you in a moment.’

38

Honorifics:

Japanese and Thai: the first-person marker has the

characteristics of both a pronoun and a noun. Pronouns and nouns are not morphologically different: like nouns, pronouns do not form a closed class; like nouns, they form the plural by adding a plural morpheme;

also e.g. Burmese, Javanese, Khmer, Korean, Malay, or

Vietnamese. Typically: ‘slave’, ‘servant’, royal slave’, ‘lord’s servant’,

‘Buddha’s servant’ are used for self-reference with self-denigration;

Thai: 27 forms of first person (cf. ‘mouse’) Siewierska (2004) and Heine and Song (2011)

39

Conflation of the nominal with the pronominal:

Acoma (New Mexico), Wari’ (Brazil): no personal pronouns;

Generic one and arbitrary PRO: ‘One can hear the wolves from the veranda.’ ‘It is scary PRO to hear the wolves from the veranda.’ Generic one and arbitrary (non-controlled) PRO express

‘generalizing detached self-reference.’ Moltmann (2010: 440)

40

Spatial deixis:

Thai phŏm1 nii2 ( ‘one male this’); Japanese kotira, Korean yeogi, and Vietnamese hây (‘here’)

used for self-reference;

41

Ambiguity of the 1st person pronoun?

Kratzer (2009): pronouns can be ambiguous between a

referential and a bound-variable interpretation ‘I’m the only one around here who can take care of my children.’ ‘Only I admitted what I did wrong.’ ‘Only you can eat what you cook.’

42

Cross-linguistic differences: bound-variable uses are rare, restricted, and differ from language to language.

Tylko ja jeden przyznałem się do błędu. only 1Sg soleSgMNom admit1SgPastM Refl to mistakeSgMGen Tylko ja jedna tutaj potrafię zajmować się

Only 1Sg soleSgFNom here can1SgPres careInf Refl swoimi dziećmi. ReflPronPl Instr childPl Instr

43

Kratzer:

bound variable pronouns are underlyingly referential

pronouns whose meaning can be accounted for through context-shifting.

or: they are unspecified and obtain the meaning through

feature transmission from their binders in functional heads.

44

Ambiguity and underlying ‘I’ reference

‘Alice wants what Lidia wants.’ underlying ‘I’-reference (self-attribution of property) But: ‘Lidia’s mother wants what Lidia wants and that’s why she is

buying her lots of scientific books.’ no underlying ‘I’-reference ( propositionalism)

45

Argument from first-person pronoun: A summary

Self-referring that involves cognitive access to oneself defies any attempt to fit it squarely into the mould of a single, systematic morphosyntactic device.

46

Argument from first-person pronoun: A summary

Self-referring that involves cognitive access to oneself defies any attempt to fit it squarely into the mould of a single, systematic morphosyntactic device.

Instead, the device standardly used for this purpose in

English, the first-person singular pronoun, can have other uses as well, and devices that specialise for other uses, such as common nouns and proper names, can adopt the function of reference de se.

47

Argument from first-person pronoun: A summary

This suggests that formal semantics that relies on the rigid distinction between an indexical and non-indexical expression (Kaplan 1989) needs ‘pragmaticising’ .

(Jaszczolt 2012a, b; 2013a, b)

48

To conclude

• I agree that ‘it is not the pronoun that should be disambiguated, but rather the particular mechanisms that affect its interpretation’ SSL, p. 19

But: It has not been demonstrated that these mechanisms are

linguistic (cf. the selection of a coherence relation; ‘two kinds of logic’).

49

• Post-Gricean accounts of reference resolution are slightly misrepresented: they do not postulate ambiguity of demonstratives (cf. Modified Occam’s Razor, pragmatic enrichment of the propositional content)

‘[C]ontext is not as powerful as the traditional model seems to

presume’ SSL, p. 20

?What is ‘the traditional model’?

50

• The promised discussion of the parallel account of ‘I’ and ‘he’

that would contrast it with Kaplan proceeds without a discussion of the indexical (‘I’)

51

Open questions: • At what level of representation should compositionality be

sought? (cf. interactive composition) • What happens to a pronoun in a dialogue? • Is building in pragmatic rules into the LF compatibile with

indexicalism on a large scale? • How do we extend the attention-coherence view to other

domains (quantifier domain restriction, lexical meaning, ...)?

• Perhaps there is a need for corpus data in order to obtain an exhaustive account of rhetorical relations.

52

Select References

Asher, N. and A. Lascarides. 2003. Logics of Conversation. Cambridge: Cambridge University Press.

Asher, N. And A. Lascarides. 2013. ‘Strategic conversation’. Semantics and Pragmatics 6. 1-62.

Chierchia, G. 1989. ‘Anaphora and attitudes de se’. In: R. Bartsch, J. van Benthem and B. van Emde Boas (eds). Semantics and Contextual Expression. Dordrecht: Foris. 1-31.

Chierchia, G. 2004. ‘Scalar implicatures, polarity phenomena, and the syntax/pragmatics interface’. In: A. Belletti (ed.). Structures and Beyond: The Cartography of Syntactic Structures, vol. 3. Oxford: Oxford University Press. 39-103.

Groenendijk, J. and M. Stokhof. 1991. ‘Dynamic Predicate Logic’. Linguistics and Philosophy 14. 39-100.

Heine, B. and K.-A. Song. 2011. ‘On the grammaticalisation of personal pronouns’. Journal of Linguistics 47. 587-630.

53

Jaszczolt, K. M. 2005. Default Semantics: Foundations of a Compositional

Theory of Acts of Communication. Oxford: Oxford University Press. Jaszczolt, K. M. 2010. ‘Default Semantics’. In: B. Heine and H. Narrog (eds).

The Oxford Handbook of Linguistic Analysis . Oxford: Oxford University Press. 215-246.

Jaszczolt, K. M. 2012a. ' 'Pragmaticising' Kaplan: Flexible inferential bases and fluid characters'. Australian Journal of Linguistics 32. 209-237.

Jaszczolt, K. M. 2012b. 'Context: Gricean intentions vs. two-dimensional semantics'. In: R. Finkbeiner, J. Meibauer and P. B. Schumacher (eds). What is Context? Linguistic Approaches and Challenges. Amsterdam: John Benjamins. 81-103.

Jaszczolt, K. M. 2013a. 'First-person reference in discourse: Aims and strategies'. Journal of Pragmatics 48. 57-70.

Jaszczolt, K. M. 2013b. 'Contextualism and minimalism on de se belief ascription'. In: N. Feit and A. Capone (eds). Attitudes De Se: Linguistics, Epistemology, Metaphysics. Stanford: CSLI Publications. 69-103.

Jaszczolt, K. M. in progress. Interactive Semantics. Oxford: Oxford University Press.

54

Kaplan, D. 1989a. ‘Demonstratives’. In: J. Almog, J. Perry, and H. Wettstein (eds). Themes from Kaplan. New York: Oxford University Press. 481-563.

Kaplan, D. 1989b. ‘Afterthoughts’. In J. Almog, J. Perry & H. Wettstein (eds). Themes from Kaplan. New York: Oxford University Press. 565-614.

Kehler, A. 2002. Coherence, Reference, and the Theory of Grammar. Stanford, CA: CSLI Publications.

Kratzer, A. 2009. ‘Making a pronoun: Fake indexicals and windows into the properties of pronouns’. Linguistic Inquiry 40. 187-237. Lewis, D. 1979. ‘Attitudes de dicto and de se’. Philosophical Review 88. 513-543.

Merchant, J. 2004. ‘Fragments and ellipsis’. Linguistics and Philosophy 27. 661-738.

Moltmann, F. 2010. ‘Generalizing detached self-reference and the semantics of generic one.’ Mind and Language 25. 440-473. Perry, J. 2009. ‘Directing intentions’. In J. Almog & P. Leonardi (eds). The

Philosophy of David Kaplan. Oxford: Oxford University Press. 187-201. Recanati, F. 2004. Literal Meaning. Cambridge: Cambridge University Press.

55

Recanati, F. 2012. ‘Compositionality, flexibility, and context dependence’. In: M. Werning, W. Hinzen & E. Machery (eds). The Oxford Handbook of Compositionality. Oxford: Oxford University Press. 175-191.

van der Sandt, R. A. 1992. ‘Presupposition projection as anaphora resolution’. Journal of Semantics 9. 333-377.

Siewierska, A. 2004. Person. Cambridge: Cambridge University Press. Stalnaker, Robert C. 1978. ‘Assertion’. Syntax and Semantics 9. New

York: Academic Press. Reprinted in Robert C. Stalnaker, 1999, Context and Content: Essays on Intentionality in Speech and Thought, 78-95. Oxford: Oxford University Press.

Stalnaker, Robert C. 2011. ‘The essential contextual’. In Assertion: New Philosophical Essays, Jessica Brown & Herman Cappelen (eds), 137-50. Oxford: Oxford University Press.

56