Lewis and the Semantics-Pragmatics Divide Ernie Lepore Matthew Stone Department of Computer Science Center for Cognitive Science Rutgers University Outline Rethinking semantics and pragmatics • coordination (Lewis 1969) • the conversational record (Lewis 1979) Linguistics and the social Implications for philosophical practice Background Last time: Intentions • Intention recognition is an important ingredient of understanding, collaboration BUT • Problematic to locate linguistic knowledge using intentions Need to look elsewhere to get clear on the scope of linguistic rules as social constructs Coordination Situations where agents’ actions must agree • but no intrinsic reason to prefer one joint strategy over another Coordination Examples from Lewis (1969) • Arrange a meeting • Restart an interrupted phone call • Row a boat • Share the road when driving • Perform a search as a group • Collude to set prices • Signal a choice Coordination as a game C1 C2 C3 R1 1;1 0;0 0;0 R2 0;0 1;1 0;0 R3 0;0 0;0 1;1 Games and reasoning If an agent chooses her action deliberately, she must consider her expectations about her partner. Idealization: equilibrium • no player would change his move, given the moves the others make Games and reasoning We generally expect equilibria in real life • where agents play repeatedly they can learn to match one another even if individual decisions are heuristic Equilibrium describes regularity in behavior • Lets us ascribe certain states to agents (perhaps tacit or implicit) State underlying equilibrium It’s common knowledge that (1) everyone conforms to R (2) everyone expects everyone else to conform to R (3) everyone prefers to conform to R on the condition that others do Lewis: this is a convention Objections: • Silent about how conventions are instituted That’s the point. • Lewis describes how we can get convention from salience, precedents and convergence – but this is just one story among many Games and reasoning A range of mechanisms explain equilibrium • Architecture – we’re only biologically capable of one equilibrium • Salience – we’re so constituted that preferred equilibrium leaps to mind • Experience – we’ve learned that others follow a given equilibrium All potentially relevant to language faculty Architecture: Universal Grammar Salience: Primitive preferences for particular patterns of interpretation Experience: Acquired lexical items, syntactic parameter settings, etc. Lewis: this is a convention Objections: • We don’t have to know all these things (Burge) Fair point – let’s talk about ‘social competence’ instead as whatever lets us solve (certain) coordination problems – not prejudging the actual status of skills, knowledge, choice & alternatives Lewis: this is a convention Objections: • Lewis’s conventions are always followed – this seems too strong (Gilbert, Millikan) • Examples: handing out cigars, using ‘bank’ to mean financial institution Lewis: this is a convention Objections: • Lewis’s conventions have to achieve preferred outcomes – this seems too strong (Gilbert, Millikan) • Examples: decorating for Christmas specifically with red and green Lewis: this is a convention Objections: • Seems to assume coordination problem exists antecedently of equilibrium – this seems too strong (Marmor) • Examples: playing chess by the rules Response In keeping with idea of social competence, let’s think of coordination problems are rational reconstructions Lewis on Signaling One party produces signal, knowing the state of the world The other acts, having seen the signal Explains information carried by signal, but a long way from meaning Signaling example ‘one if by land, two if by sea’ Sexton hangs certain patterns of lights conditional on what British are doing Revere prepares particular pattern of defense conditional on what lights he sees They want equilibrium Signaling games 1L2S 1S2L 1L2S 1;1 0;0 1S2L 0;0 1;1 Problems Meaning is underdetermined • 1 light: the British are coming by land or • 1 light: prepare the land defenses! or • both? Problems Meaning only present at equilibrium • So what about coordination that succeeds through other mechanisms – salience? good luck? partial or tentative precedents? Lewis on Convention Lewis attempts to generalize to language by directly understanding truth conditions as conventions for agents to use certain sentences in certain conditions (namely, when they are true) Lewis on Convention This is very cumbersome And doesn’t seem to get at the real difficulties Another idea Language combines social competence with specific institutions targeted at meaning Specifically: the conversational record (Lewis 1979) Conversational record List of propositions associated with discourse Specifies • interlocutors’ environment • what has been said already • what the purposes and plan is • what standards of meaning are in play • what issues are open • what conversation is committed to (Lewis 1979, Thomason 1990) Conversational record Record is dynamic • topic can change • meaning standards can be negotiated • presuppositions can be challenged • interlocutors can commit to new propositions or rescind previous commitments Utterances specify updates • generally, as a matter of meaning Conversational record Abstraction • Free to specify discourse referents, standards for vague predicates, other constructs from formal theories • Need not be tied to interlocutors’ knowledge or belief Record is arbitrary In key respects, it’s up to us how the record changes So it’s natural to think of the record as an object of coordination Coordinating on the Record One way of thinking • I have my version of what’s happened • You have your version of what’s happened • When it lines up, we’ve communicated Compare Neale’s presentation here Coordinating on the Record One way of thinking • I have my version of what’s happened • You have your version of what’s happened • When it lines up, we’ve communicated Drawback: describing cases of miscommunication, clarification, etc. Coordinating on the Record An indirect way of thinking • Each of us defers to practices • Meanings specify how to update the record • We coordinate on what we defer to Coordinating on the Record An indirect way of thinking • Each of us defers to practices • Meanings specify how to update the record • We coordinate on what we defer to Promises a better handle on miscommunication, clarification The Received View Semantics • linguistic specification of reference, truth • settles what the speaker is saying Pragmatics • general principles of inference and strategy • settles what speaker is doing Problems Fails to describe linguistic knowledge • more linguistic facts than supposes • e.g., rules for indirection, presupposition, information structure, etc. Fails to describe interpretive inference • wide range of practices for engaging with imagery, drawing insights Overall picture Key theoretical notion is inquiry • process (normally collaborative) in which interlocutors settle how things are Requires • public meanings • open to negotiation, debate Overall picture Inquiry privileges conventional meaning • depends on conversational record • depends on coordination Allows for a broad understanding of meaning Excludes insight or point of open-ended, idiosyncratic engagement with utterance Inquiry and CR Conversational record tracks inquiry • Assertion registers proposition on record as commitment of one party • Enables further follow up, such as clarification questions, arguments for or against, agreement or disagreement by other parties Record ensures a shared interpretation How will this affect philosophy? Better arguments, but more difficult ones. Close with case study: • Grice versus ordinary language philosophers Color: the dialectic Claim: color supervenes on appearance • Something is red just in case it looks red under normal conditions to appropriately endowed observers. Objection (Austin): • Hogwash! You’d only say something looked red if it wasn’t red! Color: the dialectic Reply (Grice): • Well, yes, you wouldn’t say something looks red unless it wasn’t red. • But, that’s not part of the meaning of “it looks red”, it’s an implicature. Us: • Sorry Grice, your linguistics is bogus. What next? Us: • Sorry Austin, your linguistics is bogus too! Often say “something looks red” when it is: • His fake tan looks orange. • Contaminated water still looks clear. • The distant shores look green and inviting. (examples after google searches) What next? Us: • Sorry Austin, your linguistics is bogus too! When “it looks red” means it’s not, it’s because of intonation. No objection to analysis of color, as long as you don’t use that intonation! Similar cases Believe and know. Try and succeed. Or and and. Ordinary language folks: 1st suggests not 2nd Grice: That’s just an implicature. Us: You’re both wrong. 1st suggests not 2nd if marked elsewhere lots of times 1st doesn’t suggest not 2nd. Basic point Our theories need to acknowledge the richness and complexity of language and communication. We can do the linguistics and philosophy we want without unhelpful categories like “conversational implicatures”.