1. Gazzara / ENG 252 *with thanks to Dr. Seth Cable, MIT

2. Our Goal Thus Far To develop a theory of the combinatory system that derives the assertions of a complex (declarative) sentence from the meanings of its component expressions Larson (1995: 361-368); Partee (1995: 311-316), chierchia & McConnell-Ginet (2000: 1-33); Heim & Kratzer (1998: 1-3); Cable (2010: 1-9)

3. Some new terminology The ‘truth conditions’ of a sentence S are the conditions under which S is true – Canonical Truth-Conditional Statement: ‘S is true if and only if p’ Some Consequences: The ‘truth conditions’ of S are another name for the ‘assertions’ of S

4. Our Goal (Restated Again) Develop a theory of the combinatory system that derives the truth conditions of a sentence from the ‘meanings’ of its component parts

5. A quick review of how we got here We want to know how the ‘meaning’ of a sentence is computed from the ‘meaning’ of its parts This requires us to make more precise what we mean by ‘meaning’ This leads us to the notion of the information that a sentence conveys This requires us to make more precise what we mean by ‘conveying information’ This leads us to the notion of the information ‘asserted’ by a sentence This notion can be recast as the truth conditions of a sentence Thus, we want to know how the truth conditions of a sentence can be derived from the ‘meanings’ of its component parts…

6. Psychological Reality Our restated goal carries a certain presupposition regarding the nature of our ‘language faculty’, which it is worth pausing to reflect on: The ‘Psychological Reality’ of Truth-Conditions Our goal presupposes that part of our cognitive capacity as speakers of a language is a system that derives truth conditions

7. This isn’t so far fetched a claim… … to see this, let us first consider the following crude but ‘common-sense’ vision of how linguistic communication works:

8. A Naïve – but Commonplace – Theory of Communication First Step: Speaker utters a string of sounds ( / d ə ha ʊ s ɪ z an faj ɹ / ) Second Step: Listener hears those sounds Third Step: Listener’s linguistic system interprets the sounds and puts a ‘picture’ in their head

9. A Few Problems for this Naïve View While this ‘picture theory’ of meaning continues to be a common folk theory of communication, it has long been known to be inadequate as a serious analysis (Wittgenstein, Philosophical Investigations) Problem 1: Pictures are Ambiguous – If the meaning of “my house is on fire” is a picture of my house being on fire,what makes it a picture of my house rather than a physically identical one? Problem 2: Pictures are Too Concrete – What is the ‘picture’ which is the meaning of “1+1=2” or “There are billions of stars in the Milky Way”

10. So what is it?... So if the meaning of a sentence isn’t a picture or something similar, what is it?... … well, the following more plausible model of communication suggests that T-conditions might play an important role…

11. A More Plausible Model of How Communication Operates Speaker’s Utterance: / d ə ha ʊ s ɪ z an faj ɹ / Listener’s Computations: – (i) Syntax: The string /d ə ha ʊ s ɪ z an faj ɹ / has the following structure: [[the house][is[on[fire]]]] – (ii) Semantics: [[the house][is[on[fire]]]] is true iff the house is on fire – (iii) Pragmatics: The speaker is an honest guy, so he believes what he says... The speaker is smart, so what he believes is true... – So “[[the house][is[on[fire]]]]” must be true… So, given its truth conditions, the house must be on fire… oh my god the house is on fire!...

12. Observations Observation 1: This model of communication isn’t obviously fraught with the difficulties faced by the naïve ‘picture theory’ presented earlier Observation 2: In this model of human communication, a central step in comprehension is the computation of the truth-conditions of the speaker’s utterance: – If listeners have reason to believe that a speaker’s utterance is true (e.g. because the speaker is honest and knowledgeable)…. – then knowing the truth conditions of a sentence allows the listener to thereby deduce facts about the world!

13. CONCLUSION of Truth Conditional Semantics To the extent to which this last point is an accurate characterization of the kind of information our systems compute when we ‘understand’ an utterance, then part of what our language systems ‘do’ during sentence comprehension is compute truth- conditions

14. Brief Terminological Aside: ‘Object Language’ and ‘Metalanguage’ a. The Object Language: The language we are describing (not using) – We want our theory to associate sentences of the object language with their truth conditions b. The Metalanguage: – The language we are using to characterize the truth conditions of the sentences of our object language We want our theory to associate sentences of the object language with their truth conditions, as characterized in the metalanguage

15. SAMENESS/DIFFERENCE The object language and metalanguage can both be the same language…. – “The house is on fire” is T (true) iff the house is on fire. Object language: English / Metalanguage: English …or, they can be two different languages: – “Het huis brandt” is T (true) iff the house is on fire. Object language: Dutch Metalanguage: English

16. Larson (1995: 361-368); Partee (1995: 311-316), Chierchia & McConnell-Ginet (2000: 1-33); Heim & Kratzer (1998: 1-3); Cable (2010: 1-9)

17. Obtaining a System That Derives Truth Conditions The Goal of Our Enterprise To develop a system that, for every sentence S of English, derives the truth-conditions of S from (i) the meanings of the component pieces of S, and (ii) the syntax of S

18. Outline Sketch of a System Satisfying THIS SEMANTICS(‘Barack’) + SEMANTICS(‘smokes’) = SEMANTICS(‘Barack smokes’) = TRUTH CONDITIONS(‘Barack smokes’)

19. Some Important Questions How do we construct a theory that obtains this result for us? What are the ‘meanings’ of “Barack” and “smokes” such that combining them together can yield for us the truth conditions of “Barack smokes”?

20. More about the Meaning of ‘Meaning’: An Excursus of ‘Extensions’ We’ve already seen that the everyday word ‘meaning’ is vague and ambiguous in a number of ways…here’s another:

21. The Meaning of the Phrase ‘The President’ In one sense, the meaning of the NP “the president of the USA” changed about a year ago. It went from ‘meaning’ George Bush to ‘meaning’ Barack Obama. (‘denotation’, ‘reference’) But, in another sense, the meaning of the NP has stayed the same. It still means (roughly) ‘the person who holds the office of the presidency of the United States’. (‘sense’, ‘concept’) Instead of using the word ‘meaning’ in this vague and ambiguous fashion, let’s introduce two different terms to refer unambiguously to these two different ‘senses’ of the word “means”

22. Extension vs. Intension (of NPs) a. The extension of an NP is the thing in the actual world that the NP refers to The extension of “the president” is Barack Obama b. The intension of a phrase is the ‘general concept’ behind the phrase, which determines (for a given time/situation) what the extension of the phrase is The intension of “the president” is the person who holds the office…

23. So, the ‘meaning’ of an NP can be broken up into its ‘extension’ and its ‘intension’… …can the meaning of a sentence likewise be broken up in this way?

24. Intension of a Sentence = ‘Truth Conditions’ We might take the ‘intension’ of a sentence to be (something like) its truth conditions… As we’ve already seen, the truth conditions of a sentence are akin to what we might vaguely describe at the sentence’s ‘conceptual/propositional content’ … But if the ‘intension’ of a sentence is its truth conditions, what is its ‘extension’?...

25. Extension of a Sentence = Truth Value If we take the ‘intension’ of a sentence to be its truth conditions, then we should take the ‘extension’ of a sentence to be its truth value Why? Recall that the ‘intension’ determines for a given time/situation what the extension is Truth conditions determine for a given time/situation what the truth value is The (crucial) idea that the extension of a sentence is its truth value is one of those ‘weird ideas’ that you just have to get used to…

26. The General Picture a. Intension of “X”: A kind of ‘concept’/ ‘definition’ which – for any given time/situation determines what “X” ‘picks out’ at that time/situation b. Extension of “X” The thing which, at a given time/situation “X” ‘picks out’ c. Illustrative Paradigm: i. Intension of “the president” = whoever holds the office of the presidency ii. Extension of “the president” = Barack Obama iii. Intension of “Barack smokes” = “Barack smokes” is T iff Barack smokes iv. Extension of “Barack smokes” = TRUE

27. What’s the point of all of this? As we will presently see, you can actually build a decent theory of how meanings can ‘compose’ to yield truth- conditions by paying attention to extensions (defined as above)!

28. Towards a Compositional Semantics Based on Extensions Some New Notation: [[ X ]]: the extension of “X”; (function from X to its extension) The Compositionality of Extensions a. We’ve broken down the ‘meaning’ of a S/NP into its extension and its intension b. Recall, however, that our semantic system must be such that the ‘meaning’ of a complex expression should be derived from the ‘meaning’ of its component parts c. Thus, the extension of a complex expression should be derived from the extensions of its component parts!

29. In a Picture: [[ Barack ]] + [[ smokes]] = [[ Barack smokes ]] = True The Extension of a Predicate = Function In order to make the picture above work, the extension of the predicate “smokes” must be such that ‘combining’ it with the extension of “Barack” yields the extension of “Barack smokes”, which is the value ‘TRUE’

30. How can we model this? Well, we know the following: (i) [[ Barack ]] = Barack (ii) [[ Barack smokes ]] = T(rue) Thus: [[ smokes ]] + Barack = T The extension of “smokes” ‘combines’ with Barack to yield T So, the extension of “smokes” is like a ‘device’ that takes Barack as input and yields T as output… So, the extension of “smokes” is like a FUNCTION!

31. The core idea: [[ smokes ]] = A function from entities to truth values, which yields T iff that entity smokes Your final exam WILL be open note, though it’s on you to either access this PPT during the exam or to print out these slides (six to a page). Expect short-answers, multiple choice and T or F questions only on this content. I’ll go over some example questions with you.