Presupposition Projection: the New Debate ACTL Spring School, London, April 21-25, 2008 Philippe Schlenker (Institut Jean-Nicod & NYU)
Why Study Presuppositions?
2
Presuppositions Approximation: A presupposition of S is a condition that must be met for S to be true or false.
Presuppositions a. John knows that he is incompetent. π: John is incompetent. b. Does John knows that he is incompetent? π: John is incompetent c. John doesn’t know that he is incompetent. π: John is incompetent.
Entailments a. John is French. b. Is John French? c. John isn’t French.
=> John is European. ≠> John is European. ≠> John is European.
3
Empirical Motivation John regrets that he is incompetent. π: John is incompetent.
John has stopped smoking. π: John used to smoke.
It is John who left. π: Someone left.
What John drank was vodka. π: John drank something.
She is clever! π: The person pointed at is female.
4
Empirical Motivation
John too was jailed. π: Someone other than John was jailed.
John was jailed again. π: John was jailed before.
Only John was jailed. π: Somebody was jailed.
5
Theoretical Motivation: Static vs. Dynamic Semantics Static View of Meaning Meaning = Truth Conditions
Dynamic View of Meaning (after the 1980’s) Meaning
= Context Change Potential
= potential to change beliefs Motivations for the dynamic view a. Pronouns, e.g. Every man who has a donkey beats it. b. Presuppositions. ☞ Is the ‘dynamic turn’ in semantics justified?
6
Modularity: The Semantics / Pragmatics Divide Semantics = study of meaning as it is encoded in words John is an American student => John is a student John is a former student ≠> John is a student
Pragmatics = study of the additional information that can be obtained by reasoning on the speaker’s motives Mr. Smith is unfailingly polite and always on time => Smith is a bad student ☞ Do presuppositions belong to semantics or pragmatics? 7
Presuppositions, Entailments, and Scalar Implicatures
8
Presuppositions vs. Entailments
Difference 1 (dubious) If an entailment of S is false, S is false, not weird.
-John is French. -No. He is South African.
-John knows that he is going to be fired. -No. He doesn’t know it. - No. He is going to keep his job.
9
Presuppositions vs. Entailments Difference 2 (very clear) Presuppositions ‘project’ differently from entailments.
a. Is John French?
≠> John is European b. John is not French. ≠> John is European c. None of these 10 students is French ≠> Each of these 10 students is European ≠> Some of these 10 students is European
a. Does John know that he is incompetent? => John is incompetent b. John does not know that he is incompetent => John is incompetent c. None of these 10 students knows that he is incompetent 10 => Each of these 10 students is incompetent
Presuppositions vs. Entailments a. Does John take care of his computer? => John has a computer b. John doesn’t take care of his computer => John has a computer c. None of these 10 students takes care of his computer => Each of these 10 students has a computer
a. Did John stop smoking? => John used to smoke. b. John didn’t stop smoking => John used to smoke c. None of these 10 students stopped smoking => Each of these 10 students used to smoke 11
Scalar Implicatures a. I will invite John or Mary =>? I won’t invite both b. John is a poet or a philosopher =>? He isn’t both.
Neo-Gricean Analysis H1 : forms a scale H2: All else being equal, speakers prefer to utter the more informative member of a scale.
S: I will invite John or Mary S’: I will invite John and Mary => If S is uttered, the speaker couldn’t assert S’ => If the speaker has an opinion about S’, the speaker believes that S’ is false.
12
Scalar Implicatures Some implicatures disappear in negative environments a. I won’t invite John or Mary. => denies that: I will invite John or Mary or both. b. Whenever I invite John or Mary, the party is disaster. => makes a prediction about: all cases in which I invite John or Mary or both.
Explanation a. I won’t invite both John and Mary is LESS informative than I won’t invite John or Mary (or both) b. Whenever I invite John and Mary, the party is a disaster. is LESS informative than Whenever I invite John or Mary, the party is a disaster. 13
Scalar Implicatures Other implicatures appear in negative environments... a. I won’t invite both John and Mary =>? I will invite one of them b. John isn’t both a poet and a philosopher =>? He is a poet or a philosopher c. None of my students has read Chomsky and Montague =>? Some of my students have read one of them
Explanation a. I won’t invite both John and Mary is LESS informative than I won’t invite John or Mary (or both) b. None of my students has read Chomsky and Montague is LESS informative than None of my students had read Chomsky or Montague 14
Scalar Implicatures Cancellability Unlike entailments, they can be cancelled.
Sensitivity to the logical context They ‘disappear’ in certain environments - and ‘appear’ in others.
Late Acquisition They are acquired relatively late by children.
Processing Cost They take time to compute.
15
Presuppositions vs. Implicatures An analysis of presuppositions as implicatures Hypothesis: If pp’ is a clause described as having presupposition p and assertion p’: (i) pp’ has as its meaning the conjunction of p and p’ (ii) but forms a scale
Examples a. b. c.
16
Predictions I pp’
entails
p
a. John knows that he is incompetent => John is incompetent b. I’ll invite John and Mary => I’ll invite John or Mary
not pp’
implicates
p
because (not p) is more informative than (not pp’) ! a. John doesn’t know that he is incompetent implicates: John is incompetent b. I won’t invite (both) John and Mary => I’ll invite John or Mary
17
Predictions II No student PP’ implicates Some student P because is more informative than hence the inference that i.e.
No student P No student PP’ not No student P Some student P
? The father of each of these students is about to receive a congratulation letter. =>? The father of at least 3 students is about to receive a congratulation letter.
None of these 10 students read the handout and did an exercise. =>? Each of these 10 students did (at least) one or the other =>? At least 1 of these 10 students did (at least) one or the other 22
Main Results (Chemla 2007)
Presuppositions display a different a behavior from scalar implicatures under no: -Non-universal inferences for implicatures -Universal implicatures for presuppositions
Not all quantifiers behave on a par: at least 3, more than 3, exactly 3 display an intermediate behavior (universal inferences half the time).
Not computing a presupposition takes time. 23
NO and Universal Inferences Left, from left to right 1. Every student stopped smoking => every student smoked 2. No student stopped smoking => at least one student smoked 3. No student stopped smoking => every student smoked Right, from left to right 1. Every student did A and B => every student did (at least) one 2. No student student did A and B => at least one student did (at least) one 3. No student did A and B => every student did (at least) one
24
NO and Universal Inferences
25
Less than three and Universal Inferences
26
Universal Inferences for Various Quantifiers
27
Reaction Times: Universal Inferences
28
Questions
Triggering Problem Why do some elementary clauses have presuppositions? a. John knows that it is raining π: It is raining. b. John rightly believes that it is raining π: none, or possibly: John believes that it is raining.
29
Questions
Projection Problem How do the presuppositions of elementary clauses get transmitted to complex clauses ? a. If John is realistic, he knows that he is incompetent. π: John is incompetent b. If John is an idiot, he knows that he incompetent π: none, or possibly: if John is an idiot, he is incompetent 30
Questions
Architectural Question Where do presuppositions belong in the architecture of language: are they semantic or pragmatic?
31
The Projection Problem
32
Conjunction a. John knows that he is incompetent b. Is it true that John knows that he is incompetent? π: John is incompetent c. I doubt that John knows that he is incompetent π: John is incompetent d. None of these 10 students knows that he is incompetent. π: Each of these 10 students is incompetent.
a. John is incompetent and knows that he is. b. Is it true that John is incompetent and knows that he is? π: none c. I doubt that John is incompetent and knows that he is. π: none d. None of these 10 students is incompetent and knows it. 33 π: none
Conjunction a. John is depressed and his boss knows that he is incompetent b. Is it true that John is depressed and that his boss knows that he is incompetent? π: John is incompetent c. I doubt that John is depressed and that his boss knows that he is incompetent.
a. John is an idiot and his boss knows that he is incompetent. b. Is it true that John is an idiot and that his boss knows that he incompetent? π: if John is an idiot, he is incompetent (?) c. I doubt that John is an idiot and that his boss knows that 34 he is incompetent.
Conjunction p and qq’ presupposes p ⇒ q (... to be refined)
John is incompetent and he knows it / that he is π: none
John is an idiot and he knows that he is incompetent π: if John is an idiot, he is incompetent
John is depressed and his boss knows that he is incompetent Predicted π: If John is depressed, he is incompetent Actual π: John is incompetent Maybe because: the most plausible way to make the conditional true is to assume that its consequent is!
35
Conditionals a. If John is incompetent, he knows that he is. b. Is it true that if John is incompetent, he knows that he is? c. I doubt that if John is incompetent, he knows that he is.
a. If John is realistic, he knows that he is incompetent. b. Is it true that if John is realistic, he knows that he is incompetent? c. I doubt that if John is realistic, he knows that he is incompetent.
a. If John is over 65, he knows he can’t apply. b. Is it true that if John is over 65, he knows he can’t apply? c. I doubt that if John is over 65, he knows he can’t apply. 36
Conditionals a. if p, qq’ presupposes p ⇒ q b. if pp’, q presupposes p
a. If John knows that he is overqualified, he won’t apply. b. Is it true that if John knows that he is overqualified, he won’t apply? c. I doubt that if John knows that he is overqualified, he won’t apply.
a. If John knows that he is overqualified, he is depressed b. Is it true that if John knows that he is overqualified, he is depressed? c. I doubt that if John knows that he is overqualified, he is depressed.
37
Disjunctions a. p or qq’ presupposes (not p) ⇒ q b. pp’ or q presupposes p
a. If John is incompetent, he knows that he is. b. Either John is not incompetent, or he knows that he is.
a. If John is realistic, he knows that he is incompetent. b.Either John is not realistic,or he knows he is incompetent.
a. If John is over 65, he knows he can’t apply. b. Either John isn’t over 65, or he knows he can’t apply
a. If John knows that he is overqualified, he won’t apply. b. Either John doesn’t know that he is over qualified, or he 38 won’t apply.
Stalnaker’s Pragmatic Analysis
39
A Pragmatic Analysis p and qq’ presupposes p ⇒ q ‘... when a speaker says something of the form A and B, he may take it for granted that A (or at least that his audience recognizes that he accepts that A) after he has said it. The proposition that A will be added to the background of common assumptions before the speaker asserts that B. Now suppose that B expresses a proposition that would, for some reason, be inappropriate to assert except in a context where A, or something entailed by A, is presupposed. Even if A is not presupposed initially, one may still assert A and B since by the time one gets to saying that B, the context has shifted, and it is by then presupposed that A.’ Stalnaker, ‘Pragmatic Presuppositions’, 1974 40
Assumptions Assumption 1: Sentences may be true, false or # Assumption 2: A sentence S is a presupposition failure if it has the value # with respect to at least one of the states of affairs compatible with what the speech act participants take for granted. Definition 1: Common Ground = what the speech act participants take for granted. Definition 2: Context Set = set of worlds compatible with what the speech act participants take for granted.
Assumption 3: The Context Set is updated incrementally in discourse and in conjunctions. 41
Possible Worlds A possible world w = a complete specification of what is going on. It determines for every sentence S whether [[ S ]] w = true, [[ S ]] w = false, or [[ S ]] w = #.
Different clauses give rise to different functions, e.g.: The President of France is Chirac w1 → false w2 → true w3 → # w4 → # ...
The US President is Bush w1 → true w2 → false w3 → true w4 → # ...
Two plus two is four w1 → true w2 → true w3 → true w4 → true 42
Further Conditions Non-Contradiction A sentence S uttered in a Context Set C is deviant if S is true in no world of C.
Non-Triviality A sentence S uttered in a Context Set C is deviant if S is true in every world of C.
43
Stalnaker’s Analysis John knows that he is incompetent is: -true in w if John is incompetent and believes that he is -false in w if John is incompetent and doesn’t believe he is -# in w if John is not incompetent.
Suppose that the speech act participants do not know whether John is or isn’t incompetent. Suppose further that the Context Set C is C = {w1, w2, w3, w4} w1 : John is incompetent and believes that he is w2: John is incompetent and believes he isn’t w3: John is not incompetent but believes he is w4: John is not incompetent and believes he isn’t 44
Stalnaker’s Analysis T = John knows that he is incompetent uttered in C is a presupposition failure because this sentence is # in w3 and w4, which both belong to C
Suppose that the speech act participants do not know whether John is or isn’t incompetent. Suppose further that the Context Set C is C = {w1, w2, w3, w4} w1 : John is incompetent and believes that he is w2: John is incompetent and believes he isn’t w3: John is not incompetent but believes he is w4: John is not incompetent and believes he isn’t 45
Stalnaker’s Analysis S = John is incompetent is: -true in w if John is incompetent in w. -false in w in all other cases (i.e. the sentence does not have a presupposition)
a. Acceptability Clearly, John is incompetent uttered in C is not a presupposition failure. b. Update -Initially, the Context Set was C = {w1, w2, w3, w4} -After S is uttered, the new Context Set is: C’ = {w1, w2} (i.e. only the worlds compatible with S are retained)
46
Stalnaker’s Analysis John is incompetent. He knows it. = S. T.
Step 1. -The initial Context Set is -After the first sentence is uttered, the new Context Set is
C = {w1, w2, w3, w4} C’ = {w1, w2}
Step 2. -The second sentence is evaluated with respect to C’ -By construction, in each world in C’, T has a value different from #. So T is not a presupposition failure in C’.
Step 3. C’ is updated to C” = {w1}.
47
Stalnaker’s Analysis Conjunction a. Treat S and T in the same way as the discourse S. T: the assertion of a conjunction is a succession of two assertions. b. Beautiful analysis of presupposition projection: every world in C that satisfies S must satisfy T. In other words: C |= S ⇒ T
Limitations a. How does the analysis extend to other operators? b. How does the analysis extend to embedded conjunctions? e.g. None of my students is rich and proud of it. 48
Heim’s Semantic Analysis (Heim 1983)
49
Heim’s Dynamic Semantics Notation: C[F] = update of the Context Set C with F Elementary Clauses a. C[John is incompetent] = # iff C = # = {w∈C: John is incompetent in w} otherwise b. C[John knows that he is incompetent] = # iff C=# or for some w∈C, John is not incompetent in w = {w∈C: John believes he is incompetent in w}, otherwise
Truth If C[S] ≠ # and w∈C, then: S is true at w iff w ∈ C[S]
Conjunction C[F and G] = C[F][G]
50
Heim’s Dynamic Semantics Negation C[not F] = # iff C[F] = # = C - C[F] otherwise
F a. not F = John doesn’t know that he is incompetent. b. C[not F] = # iff C[F] = #, iff for some w∈C, John is not incompetent in w = C - C[F] otherwise, i.e. = C - {w∈C: John believes he is incompetent in w} 51
Heim’s Dynamic Semantics Negation C[not F] = # iff C[F] = # = C - C[F] otherwise
This means that not(pp’) presupposes that p F
a. not F = John doesn’t know that he is incompetent. b. C[not F] = # iff C[F] = #, iff for some w∈C, John is not incompetent in w = C - C[F] otherwise, i.e. = C - {w∈C: John believes he is incompetent in w} 52
Heim’s Dynamic Semantics Negation C[not F] = # iff C[F] = # = C - C[F] otherwise
F a. not F = John doesn’t know that he is incompetent. b. C[not F] = # iff C[F] = #, iff for some w∈C, John is not incompetent in w = C - C[F] otherwise, i.e. = C - {w∈C: John believes he is incompetent in w} 53
Heim’s Dynamic Semantics Conditionals (analyzed as material implications) C[if F, G] = # iff C[F] = # or C[F][not G] = # = C - C[F][not G], otherwise
Worlds that refute if F, G
F
G 54
Heim’s Dynamic Semantics Conditionals (analyzed as material implications) C[if F, G] = # iff C[F] = # or C[F][not G] = # = C - C[F][not G], otherwise
This means that if pp’, q presupposes that p, and that if p, qq’, presupposes if p, q Worlds that refute if F, G
F
G 55
Heim’s Dynamic Semantics Conditionals (analyzed as material implications) C[if F, G] = # iff C[F] = # or C[F][not G] = # = C - C[F][not G], otherwise
Worlds that refute if F, G
F
G 56
Heim’s Dynamic Semantics if F, G = If John is incompetent, he knows it C[if F, G] = # iff C[F] = # or C[F][not G] = # But C[F] ≠ # and furthermore C[F] = {w∈C: John is incompetent in w} C[F][not G] = # iff C[F][G] = #, which is not the case (by construction). Furthermore, C[F][not G] = {w∈C: John is incompetent in w}[not G] = {w∈C: John is incompetent in w} - {w∈C: John is incompetent in w and John believes he is incompetent in w} = {w∈C: John is incompetent but doesn’t believe it in w} C[if F, G] = C - {w∈C: John is incompetent but doesn’t believe it in w}
57
Summary Meaning of an elementary clause = a CCP Conjunction C[F and G] = C[F][G]
Negation C[not F] = # iff C[F] = #;
= C - C[F] otherwise
Conditionals C[if F, G] = # iff C[F] = # or C[F][not G] = # = C - C[F][not G], otherwise
Disjunction C[F or G] = # iff C[F] = # or C[not F][G] = # = C[F] ∪ C[not F][G], otherwise
58
Disjunctions a. p or qq’ presupposes (not p) ⇒ q b. pp’ or q presupposes p
a. If John is incompetent, he knows that he is. b. Either John is not incompetent, or he knows that he is.
a. If John is realistic, he knows that he is incompetent. b.Either John is not realistic,or he knows he is incompetent.
a. If John is over 65, he knows he can’t apply. b. Either John isn’t over 65, or he knows he can’t apply
a. If John knows that he is overqualified, he won’t apply. b. Either John doesn’t know that he is over qualified, or he 59 won’t apply.
Heim’s Dynamic Semantics Disjunction C[F or G] = # iff C[F] = # or C[not F][G] = # = C[F] ∪ C[not F][G] otherwise.
F
G
a. John is not incompetent, or he knows that he is. b. C[not I or K] = # iff C[not I] = # or C[not not I][K] = #, i.e. iff C[I] = # or C[I][K] = #, which is never the case. Thus C[not I or K] = C[not I] ∪ C[I][K] 60
Heim’s Dynamic Semantics Disjunction C[F or G] = # iff C[F] = # or C[not F][G] = # = C[F] ∪ C[not F][G] otherwise.
F
G
This means that pp’ or q presupposes that p, and that p or qq’presupposes if (not p), q
a. John is not incompetent, or he knows that he is. b. C[not I or K] = # iff C[not I] = # or C[not not I][K] = #, i.e. iff C[I] = # or C[I][K] = #, which is never the case. Thus C[not I or K] = C[not I] ∪ C[I][K] 61
Heim’s Dynamic Semantics Disjunction C[F or G] = # iff C[F] = # or C[not F][G] = # = C[F] ∪ C[not F][G] otherwise.
F
G
a. John is not incompetent, or he knows that he is. b. C[not I or K] = # iff C[not I] = # or C[not not I][K] = #, i.e. iff C[I] = # or C[I][K] = #, which is never the case. Thus C[not I or K] = C[not I] ∪ C[I][K] 62
Heim’s Dynamic Semantics Definition of Truth If w∈C, a. F is # in w relative to C iff C[F] = # b. If ≠ #, F is true in w relative to C iff w∈C[F]
John is incompetent. He knows it. = S. T. C = {w1, w2, w3, w4} C[S] = C’ = {w1, w2} C[S][T] = C” = {w1}.
a. Relative to w1, C, the discourse is true, since w1∈C[S][T] b. Relative to w2, C, the discourse is false , since w2∉C[S][T]
63
Global Accommodation My sister is pregnant. '... it's not as easy as you might think to say something that will be unacceptable for lack of required presuppositions. Say something that requires a missing presupposition, and straightway that presupposition springs into existence, making what you said acceptable after all.' I said that presupposition evolves in a more or less rule-governed way during a conversation. Now we can formulate one important governing rule: call it the Rule of accommodation for presupposition If at time t something is said that requires presupposition P to be acceptable, and if P is not presupposed just before t, then - ceteris paribus and within certain limits presupposition P comes into existence at t." (D. Lewis) 64
Local Accommodation a. The king of France is not wise because there is no king of France. b. None of my students takes good care of his car because none of my students has a car! c. John doesn't know that he is incompetent because he just isn't incompetent!
a. It's not the case there is a king of France and he is wise because ... b. None of my students has a car and takes good care of it because... c. It's not the case that John is incompetent and knows it ...
Question: can we do without Local Accommodation by appealing to meta-linguistic uses of various operators?
65
Global vs. Local Accommodation C[not F]= # iff C[F]=# = C - C[F], otherwise.
Global Accommodation: C' = {c∈C: France is a monarchy at the time and in the world of c}. We then compute C'[the king of France is not powerful].
Local Accommodation: Instead of computing C - C[F] (which wouldn't even be defined, since C[F]=#), we compute: C - C'[F], where C'={c∈C: France is a monarchy at the time and in the world of c} 66
Local Accommodation is a Last Resort
Local accommodation is permissible whenever global accommodation would contradict a. the literal meaning of a sentence b. or an implicature of a sentence [or possibly: certain types of implicatures, e.g. primary implicatures]
67
The Proviso Problem a. If the problem was easy, it is not John who solved it. b. John knows that if the problem was easy, someone solved it (Geurts 1999)
Predicted presupposition of (a) and (b): If the problem was easy, someone solved it Actual presupposition of (a) Someone solved the problem Actual presupposition of (b) If the problem was easy, someone solved it 68
The Proviso Problem John is an idiot and he knows that he is incompetent π: if John is an idiot, he is incompetent
John is depressed and he knows that he is incompetent Predicted π: If John is depressed, he is incompetent Actual π: John is incompetent Maybe this is because the most plausible way to make the conditional true is to assume that its consequent is! ... but this kind of reasoning fails to address the minimal difference between: -If the problem was easy, it is not John who solved it -John knows that if the problem was easy, someone solved it (Geurts 1999).
69
The Proviso Problem Direction 1 (van der Sandt 1992, Geurts 1999) -This problem refutes the standard dynamic approaches as well as all approaches that make similar predictions. -A different analysis must be proposed, in which presuppositions are treated in a more syntactic fashion (‘Discourse Representation Theory’) This is a major contender among current theories.
Direction 2 (still promissory) With enough pragmatic reasoning, we can stick to Heim’s predictions - which in any event seem to be correct in other cases, e.g. If John is over 65, he must know that he is too old to apply
70
Heim’s Explanatory Problem
Problem: is the account explanatory? (Soames 1989) C[F and G] = (C[F])[G] C[F and* G] = (C[G])[F] When F and G are not presuppositional, C[F and G]=C[F and* G]={w∈C: F is true in w and G is true in w}
71
Heim’s Explanatory Problem There are many ways to define the CCP of or... C[F or1 G] = C[F] ∪ C[G], unless one of those is # C[F or2 G] = C[F] ∪ C[not F][G], unless one of those is # C[F or3 G] = C[not G][F] ∪ C[G], unless one of those is #
F
G 72
Summary
73
Satisfaction Basic Account -A presupposition must be entailed by the context. -Context = context set, i.e. what the speech act participants take for granted.
Problem: the Basic Account doesn’t work. a. John knows that he is incompetent => presupposes that John is incompetent b. John is incompetent and he knows that he is => presupposes nothing.
Dynamic Approach: the Basic Account is almost correct, but there are local contexts. C[A and B] = C[A][B] Debate of the 1990’s: Heim 1983 vs. van der Sandt/Geurts 74
The Dynamic Approach I: Pragmatics Stalnaker's Analysis: a pragmatic solution a. It is raining and John knows it. Step 1: Update the Context Set C with It is raining C[It is raining]={w∈C: it is raining in w}=C' Step 2: Update the intermediate Context Set C’ with John knows it [=that it is raining] C'[he knows it]={w∈C: it is raining in w and J. believes in w that it is raining} b. #John knows that it is raining and it is (raining).
Problem: the account is explanatory but not general (i) Assert(p and q) ≈ Assert(p) + Assert(q) (ii) This doesn’t extend to other connectives.
75
The Dynamic Approach II: Semantics Heim's Analysis: semantic reinterpretation of Stalnaker’s rules. a. Rule: C[F and G] = (C[F])[G], unless C[F]=# b. Results: same as before, except that they can be extended: -to other connectives -to quantifiers.
Problem: this is general but not explanatory (Soames) C[F and G] = (C[F])[G] C[F and* G] = (C[G])[F] When F and G are not presuppositional, C[F and G]=C[F and* G]={w∈C: F is true in w and G is true in w}
76
The Dynamic Approach III: Initial Error ? ‘... when a speaker says something of the form A and B, he may take it for granted that A (or at least that his audience recognizes that he accepts that A) after he has said it. The proposition that A will be added to the background of common assumptions before the speaker asserts that B. Now suppose that B expresses a proposition that would, for some reason, be inappropriate to assert except in a context where A, or something entailed by A, is presupposed. Even if A is not presupposed initially, one may still assert A and B since by the time one gets to saying that B, the context has shifted, and it is by then presupposed that A.’ Stalnaker, ‘Pragmatic Presuppositions’, 1974 77
The Dynamic Approach III: Initial Error ? Uncontroversial vs. dubious updates Bush is a genius. Everyone knows it. a. Needed: After the first sentence is uttered, it is common belief that Bush is a genius - so the presupposition of the second sentence can be satisfied. b. In fact: After the first sentence is uttered, something much weaker is common belief: it is common belief that the speaker believes that Bush is a genius. But this doesn’t suffice to satisfy the presupposition of the second sentence. c. Conclusion: Viewing presupposition projection in terms of belief update might be an error.
78
Predictive Theories Explanatory Depth • Attempt to find an algorithm which predicts how presuppositions are projected by various operators once their syntax and their bivalent truth-conditional behavior has been specified. • Or: show that such an attempt is doomed.
Form of the ‘New Theories’ True False
Projection
True False Failure 79
Dynamic semantics Transparency Theory Trivalent Approaches Reconstruction of Local Contexts
must be satisfied in its local context [semantic] must be articulated as a separate conjunct failure can be ignored if it doesn’t affect meaning must be satisfied in its local context [pragmatic]
Lexicalist?
A presupposition ...
Dynamic?
Main ideas
Trivalent?
Four Theories (among others...)
Yes Yes Yes No No No Yes No No Yes Yes No /No 80
The Transparency Theory (Schlenker, ‘Be Articulate’, 2007)
81
The Transparency Theory John knows that it is raining Meaning ≈ It is raining and John believes this [and ...]
Principle 1: Be Articulate! If possible, articulate as separate members of a conjunction parts of a meaning that are conceptually heterogeneous. e.g. if possible, say: … it is raining and John knows it … rather than: … John knows that it is raining…
Principle 2: Least Effort Don’t utter words that could be eliminated without changing the informational content of the sentence given what is assumed. E.g. do not say : … it is raining and … if these words can be determined to be idle as soon as they have been uttered.
82
The Intuition Example 1: John knows that it’s raining Be Articulate If possible, say: rather than:
It’s raining and John knows it John knows that it’s raining
Least Effort Do not say: if for all p,
It’s raining and … C |= it’s raining and p ⇔ p
which happens if... Presupposition: C |= it’s raining
Example 2: It is raining and John knows it.
83
The Intuition Example 3: If it’s raining, John knows it Be Articulate If possible, say: knows it rather than:
If it’s raining, it’s raining and John If it’s raining, John knows it
Least Effort Do not say: if for all p,
If it’s raining, it’s raining and … C |= if it’s raining, it’s raining and p ⇔ if it’s raining, p which is always true. Presupposition: None!
84
Least Effort >> Be Articulate
85
Two Principles of Manner Be Articulate If possible,
say rather than
… (p and pp’) … … pp’ …
e.g. Say: John is incompetent and he knows it / that he is.
Least Effort Do not say (p and pp’) rather than pp’ Version 1: Incremental if for each c’ of the same type as d and for each acceptable sentence completion b’ C |= a (d and c’) b’ ⇔ a c’ b’ e.g. #If John is incompetent, he is incompetent and sad.
86
Two Principles of Manner Be Articulate If possible,
say rather than
… (p and pp’) … … pp’ …
e.g. Say: John is incompetent and he knows it / that he is.
Least Effort Do not say (p and pp’) rather than pp’ Version 2: Symmetric if for each c’ of the same type as d and for each acceptable sentence completion b’ C |= a (d and c’) b ⇔ a c’ b e.g. #John is incompetent and sad, if he is incompetent.
87
Symmetric Transparency
Transparency = Be Articulate + Least Effort Thus a dd' b is acceptable in C if a (d and dd') b is not acceptable in C, i.e. if for each acceptable c’ , C |= a (d and c’) b ⇔ a c’ b
88
Incremental Transparency
Transparency = Be Articulate + Least Effort Thus a dd' b is acceptable in C if a (d and dd') b is not acceptable in C, i.e. if for each acceptable c’ and for each acceptable sentence completion b’, C |= a (d and c’) b’ ⇔ a c’ b’
89
How to Compute Incremental Principles To make things manageable, we work with structurally disambiguated sentences. There are 2 equivalent solutions: Solution 1:
Consider an enriched language with brackets that encode constituency (as is done here).
Solution 2 (Fox, Stabler): Talk directly about derivation trees. When we discuss the ‘end’ of a sentence, we really mean a completion of the left-most part of a derivation tree.
☞ In general, when the presupposition trigger comes ‘at the end’ of the sentence, there is no difference between 90 the symmetric and the incremental principle.
(if p. qq') If John is an idiot, he knows that he is incompetent Prediction: C |= John is an idiot ⇒ John is incompetent
Transparency: for all syntactically acceptable c’, C |= (if p . (q and c’)) ⇔ (if p . c’) Claim: Transparency is satisfied ⇔ C |= p ⇒ q [We treat conditionals as material implications] ⇐ : Straightforward ⇒ : Taking c’ to be some tautology, we get: C |= (if p. (q and c’)) ⇔ (if p . c’), hence C |= (if p. q) 91
(p and qq') John is an idiot and he knows that he is incompetent Prediction: C |= John is an idiot ⇒ John is incompetent
Transparency: for all syntactically acceptable c’, C |= (p and (q and c’)) ⇔ (p and c’) Claim: Transparency is satisfied ⇔ C |= p ⇒ q ⇐ : Straightforward. ⇒ : Taking c’ to be some tautology, we have: C |= (p and (q and c’)) ⇔ (p and c’), hence C |= (p and q) ⇔ p, hence in particular C |= p ⇒ q
92
(p or qq') John is not an idiot or he knows that he is incompetent Prediction: C |= John is an idiot ⇒ John is incompetent
Transparency: for all syntactically acceptable c’, C |= (p or (q and c’)) ⇔ (p or c’) Claim: Transparency is satisfied ⇔ C |= (not p) ⇒ q ⇐ : Straightforward because p or F ⇔ p or (not p and F) ⇒ : Taking c’ to be some tautology, we have: C |= (p or (q and c’)) ⇔ (p or c’), hence C |= (p or q), or in other words C |= (not p) ⇒ q
93
(No P. QQ’) None of these 10 students has stopped smoking Prediction: C |= Each of these 10 students smoked.
Transparency: for all syntactically acceptable c’, C |= (No P. (Q and c’)) ⇔ (No P . c’) Claim: Transparency is satisfied ⇔ C |= (Each P . Q) ⇐ : Straightforward ⇒ : Suppose, for contradiction, that i is a P-individual which is not a Q-individual. Taking c’ to be true of i and nothing else, we get: C |= (No P. (Q and c’)) ⇔ (No P . c’). The right-hand side is false but the left-hand side is true!
94
The Incremental Bias Common Wisdom: there is a sharp contrast between (p and qq’) vs. (qq’ and p) a. John used to smoke, and he has stopped smoking. b. #John has stopped smoking, and he used to smoke.
Problem: general deviance when the 1st conjunct entails the 2nd (but - oops - this doesn’t follow from Be Brief!) a. John resides in France and he lives in Paris. b. #John lives in Paris and he resides in France.
Still, the incremental bias is (arguably) real: a. John used to smoke five packs a day, and he has stopped smoking. b. ?John has stopped smoking, and he used to smoke five 95 packs a day.
Symmetric Readings a. There is no bathroom, or the bathroom is well hidden. b. The bathroom is well hidden, or there is no bathroom.
a. If there is a bathroom, the bathroom is well hidden. b. If the bathroom is not hidden, there is no bathroom.
If p, q ≈ If not q, not p If not (p and q), not p ≈ If p, p and q ≈ If not q, not p
a. ? John has stopped smoking and he used to smoke five packs a day. b. ? Is it true that John has stopped smoking and that he used to smoke five packs a day? c. ? I doubt that John has stopped smoking and that he used to smoke five packs a day. 96
General Results Propositional case: Incremental Transparency is equivalent to dynamic semantics. not pp’ presupposes p (p and qq’) presupposes p ⇒ q (p or qq’) presupposes (not p) ⇒ q (if pp’. q) presupposes p (if p . qq’) presupposes p ⇒ q
Quantificational case: Incremental Transparency is equivalent to dynamic semantics when some technical conditions are met (Schlenker 2007). [Q P] RR’ presupposes [Every P] R [Q PP’] R presuppose [Everything] R (bad result!)
97
Unless Unless John didn’t come, Mary will know that he is here.
a. Prediction of Heim 1983: No prediction (unless is not discussed) b. Prediction of Transparency: There should be no presupposition (if: John came ⇒ John is here) This follows from the equivalence:
⇔
Unless John didn’t come, q Unless John didn’t come, John came and q. 98
While While John worked for the KGB, Mary knew that he wasn’t entirely truthful about his professional situation.
a. Prediction of Heim 1983: No prediction (while is not discussed) b. Prediction of Transparency: Given knowledge that a spy is not entirely truthful about his professional situation, there should be no presupposition. This follows from the equivalence: While John worked for the KGB, q ⇔ While John worked for the KGB, he worked for the KGB and q 99
Explanatory ?
Costly Acc. ?
Yes Yes
Exactly 5
Dynamic semantics
At least 5, less than 5
No
Universal Inferences and Processing
Yes
Yes No
Transparency Theory Yes Almost Almost No? Yes
100
Summary: Transparency Inferential Predictions The predictions for at least 5, less than 5, exactly 5 might be too strong.
Processing Predictions ? If presuppositions are implicatures of manner, one might expect that they should be processed online like other implicatures. But presuppositions differ from scalar implicatures in this respect: • it takes time to process a scalar implicature • not processing a presupposition is what takes time [What are the facts about other implicatures of manner ?] 101
Trivalent Approaches (Peters, Beaver & Krahmer, George, Fox)
102
Dynamic semantics Transparency Theory Trivalent Approaches
Lexicalist?
Dynamic?
A presupposition ...
Trivalent?
Main ideas
must be satisfied in its Yes Yes Yes local context [semantic] must be articulated as a No No No separate conjunct failure can be ignored if Yes No No it doesn’t affect meaning
-The Basic account is right: presupposition gives rise to a failure if it is not satisfied in the (global) context. -But if we view ‘failure’ as uncertainty, there are cases in which the failure doesn’t have semantic consequences.
103
The Intuition Example 1: John knows that it’s raining -If there is any world w of C in which it is not raining, the clause gets the value # at w, which we interpret as uncertainty: we don’t know whether the clause is true or false at w. -We have no way to ‘recover’ from this uncertainty (i.e. it does have consequences), so the sentence is assessed as being deviant. Presupposition: It is raining. 104
The Intuition Example 2: It is raining and John knows it. -If there is any world w of C in which it is not raining, the second conjunct gets the value # at w, which we interpret as uncertainty: we don’t know whether the second conjunct is true or false at w. -Consider any world w’ of C: • if it is raining at w’, no failure arises; • if it is not raining at w’, the value of the second conjunct is indeterminate. But this doesn’t matter, because the first conjunct is false and hence the entire sentence is false anyway. Presupposition: None. 105
The Intuition Example 3: If it’s raining, John knows it -If there is any world w of C in which it is not raining, the consequent gets the value # at w, which we interpret as uncertainty: we don’t know whether the consequent is true or false at w. -Consider any world w’ of C: • if it is raining at w’, no failure arises; • if it is not raining at w’, the value of the consequent is indeterminate. But this doesn’t matter, because the antecedent is false and hence the entire sentence is true anyway. Presupposition: None. 106
Supervaluations Idea: The logic is trivalent, but an argument with the value # is ignored in case no matter how # is resolved [as 0 or as 1], this won’t affect the final value of the formula.
Implementation • [[ . ]], which is trivalent, is defined in terms of a plurality of bivalent valuations [[ . ]] i, [[ . ]] k, [[ . ]] l, etc. • [[F]](w) = 1 iff for all i ∈ I, [[ F ]] i(w) = 1 [[F]](w) = 0 iff for all i ∈ I, [[ F ]] i(w) = 0 [[F]](w) = # otherwise 107
Supervaluations how to define the bivalent interpretations [[ pp’ ]](w) = 1
[[ pp’ ]](w) = 0
[[ pp’ ]](w) = #
[[ p ]](w) = 1
[[ pp’ ]]i(w) = 1
[[ p’ ]](w) = 1
[[ pp’ ]]i’(w) = 1
[[ p ]](w) = 1
[[ pp’ ]]i(w) = 0
[[ p’ ]](w) = 0
[[ pp’ ]]i’(w) = 0
[[ p ]](w) = 0
[[ pp’ ]]i(w) = 1 [[ pp’ ]]i’(w) = 0
Note: When a formula (e.g. p, p’) is not presuppositional, it has the same value under all i’s. 108
Supervaluations vs. Strong Kleene Supervaluations are not compositional [[ (pp’ or (not pp’)) ]](w) = 1 if [[pp’]](w) = # [[ (pp’ or (not qq’))]](w) = # if [[pp’]](w) = [[qq’]](w) = #
The closest compositional system is the Strong Kleene Logic -Traditionally, the rules of the Strong Kleene logic are stipulated (see Peters 1979; George 2008 gives a solution) -Other solution: ‘generalization to the worst case’. Index the expressions that get repeated in formula - e.g. turn (pp’ or (not pp’)) into (pp’1 or (not pp’2)) - and apply Supervaluations. The result is Strong Kleene. 109
No student No student has stopped smoking. => Every student used to smoke.
Experimental data (Chemla): Universal inferences Prediction: Weak inferences Reason: In order to refute the claim, it is enough to find one student who used to smoke and doesn’t - no matter what the status of the other students is, this will suffice to guarantee that the sentence is false. Hence it doesn’t take much for the sentence not to be a failure (...but the price is that the sentence must be false!) 110
No student Chemla 2007: strong universal inferences. Maybe the inference in [No P]QQ’ stems from: -failure conditions, together with -truth conditions: check for every P-individual that it makes QQ’ false, and hence that it makes Q true and Q’ false. hence: [every P] (Q and not Q’) (Fox 2007)
Problem a. Did no student stop smoking? b. No student stopped smoking. -I doubt it.
George 2007: ‘Preference for truth’ [≠ George 2008] A sentence is deviant if its presupposition alone rules out the possibility that it is true [i.e. true ‘with certainty’]
111
More than 3 students, Exactly 3 students a. More than 3 students have stopped smoking. b. Exactly 3 students have stopped smoking.
Chemla 2007: universal inferences are at chance here. More than 3: Trivalence derives weak presuppositions. Exactly 3: -The initial trivalent account predicts weak presuppositions. -With ‘Preference for Truth’, very strong presuppositions. Intuition: To have any chance of establishing (with certainty!) that an exactly statement is true, we cannot tolerate any uncertainty about any individual, since it could turn out to ‘tip the balance’. 112
Incremental vs. Symmetric Treatments Peters 1979 and George 2007: compositional treatments George 2007 offers a general rule that derives trivalent semantics of any operator from a specification of its classical semantics and its syntax.
Left-Right Bias F uttered in C results in a presupposition failure unless for every w ∈ C for every a which is an initial string of F, for every sentence completion b which only includes bivalent material, [[ a b ]](w) ≠ #
Incremental Kleene = Incremental Supervaluations Symmetric Kleene > Incremental Supervaluations
113
Explanatory ?
Costly Acc. ?
Yes Yes
Exactly 5
Dynamic semantics
At least 5, less than 5
No
Universal Inferences and Processing
Yes
Yes No
Transparency Theory Yes Almost Almost No? Yes Trivalent Approaches ? No ? ? Yes 114
Summary: Trivalence Inferential Predictions -The predictions for no student might be too weak. -One way of fixing the problem leads to presuppositions for exactly 3 which might be too strong (George).
Processing Predictions ? Unclear - because it is not obvious yet how ‘local accommodation’ or the non-generation of a presupposition should be analyzed in this framework. 115
Local Contexts Revisited (Schlenker, ‘Local Contexts’, 2008)
116
Dynamic semantics Transparency Theory Trivalent Approaches Reconstruction of Local Contexts
must be satisfied in its local context [semantic] must be articulated as a separate conjunct failure can be ignored if it doesn’t affect meaning must be satisfied in its local context [pragmatic]
Lexicalist?
Dynamic?
A presupposition ...
Trivalent?
Main ideas
Yes Yes Yes No No No Yes No No Yes Yes No /No
-A presupposition must be satisfied in its local context. -Local contexts are computed by a predictive algorithm.
117
The Intuition -We preserve the idea that a presupposition must be satisfied in its local context. -We abandon the view that local contexts are the result of belief update.
Instead, we take the local context of an expression E (whose type ‘ends in t’) in a sentence S uttered in C to be the narrowest domain that one may restrict attention to without semantic risk when assessing the contribution of E to the conversation.
Example 1: John knows that it’s raining -Local context = C (all worlds in C matter) -Thus C should entail that it is raining.
118
The Intuition Example 2: It is raining and John knows it. -When we assess John knows it, we can restrict attention to those worlds in C in which it is raining. This is because all other worlds are either irrelevant to the conversation because • they are outside of C, or • they are inside C but they make the first conjunct false, and thus we are not interested in the value of the rest of the sentence in those worlds (the sentence is false anyway). -It can be shown that any stronger restriction does carry a semantic risk. So the local context of the second conjunct is: C ∧ it is raining - which satisfies the presupposition.
119
The Intuition Example 3: If it’s raining, John knows it -When we assess John knows it, we can restrict attention to those worlds in C in which it is raining. This is because all other worlds are either irrelevant to the conversation because • they are outside of C, or • they are inside C but they make the antecedent false, and thus we are not interested in the value of the rest of the sentence in those worlds (the sentence is true anyway). -It can be shown that any stronger restriction does carry a semantic risk. So the local context of the second conjunct is: C ∧ it is raining - which satisfies the presupposition.
120
Symmetric Local Contexts The symmetric local context of a propositional or predicative expression d that occurs in a syntactic environment a _ b in a context C is the strongest proposition or property x which guarantees that for any expression d’ of the same type as d, if c’ denotes x, then C |= a (c’ and d’) b ⇔ a d’ b (If no strongest proposition or property x with the desired characteristics exists, the local context of d does not exist).
p and qq’
=> qq’ and p =>
p and • => find narrowest domain at • • and p => find narrowest domain at • 121
Incremental Local Contexts The incremental local context of a propositional or predicative expression d that occurs in a syntactic environment a _ b in a context C is the strongest proposition or property x which guarantees that for any expression d’ of the same type as d, for any b’ which is syntactically acceptable, if c’ denotes x, then C |= a (c’ and d’) b’ ⇔ a d’ b’ ...
p and qq’
=> qq’ and p =>
p and • => find narrowest domain at • • ... => find narrowest domain at • 122
Example 1: (p and qq’) Symmetric Local Context of qq’ It is the strongest proposition c’ such that, for any d’ [i.e. no matter the value of qq’ turns out to be], C |= (p and (c’ and d’)) ⇔ (p and d’)
Certainly the condition will be satisfied if c’ denotes C ∧ p Any further restriction will be semantically risky. Suppose c’ excludes some p-world w of C. If d’ is true at w, then: w |= (p and d’) w |≠ (p and (c’ and d’)) so the equivalence is not satisfied. 123
Example 2: (if p, qq’) Symmetric Local Context of qq’ It is the strongest proposition c’ such that, for any d’ [i.e. no matter the value of qq’ turns out to be], C |= (if p, (c’ and d’)) ⇔ (if p, d’)
Certainly the condition will be satisfied if c’ denotes C ∧ p Any further restriction will be semantically risky. Suppose c’ excludes some p-world w of C. If d’ is true at w, then: w |= (if p, d’) w |≠ (if p, (c’ and d’)) so the equivalence is not satisfied. 124
Example 3: ([No P] QQ’) Symmetric Local Context of QQ’ It is the strongest property c’ such that, for any d’ [i.e. no matter the value of QQ’ turns out to be], C |= ([No P] (c’ and d’)) ⇔ ([No P] d’)
Certainly the condition will be satisfied if c’ denotes CP, i.e. the property which is true of individual d at world w just in case w ∈ C and d satisfies P at w.
Any further restriction will be risky. Suppose c’ excludes w, x s.t. x satisfies P at w. If at w d’ is true of x only, w |≠ ([No P] d’) because x satisfies P and d’ at w w |= ([No P] (c’ and d’)) because at w c’ excludes x so the equivalence is not satisfied. 125
Equivalence with Transparency The Transparency Theory was conceived as being antidynamic. Local Satisfaction as reconstructed here seeks to preserve as much as possible from the dynamic approach. And yet... the two theories are almost equivalent.
Equivalence When all the relevant local contexts exist... Incremental Transparency= Incremental Satisfaction Symmetric Transparency = Symmetric Satisfaction
Why? Consider pp’ in a certain syntactic environment. -Transparency: ‘make sure that p is redundant!’ -Satisfaction: 1. Start by computing the strongest restriction which is innocuous in the environment; then: 2. Check that p follows from it.
126
Equivalence Proof (Symmetric Case) a. Local Contexts: d is entailed by the symmetric local context (=lc) of dd’ in the sentence a dd’ b b. Transparency: a (d and dd’) b violates the symmetric version of Be Brief
(a) ⇒ (b): c’ denotes the local context of dd’; for any g 1. C |= a (c’ and g) b ⇔ a g b . Since c’ entails d, we have: 2. C |= a (c’ and g) b ⇔ a (c’ and d and g) b. c’ is a lc, so: 3. C |= a (c’ and (d and g)) b ⇔ a (d and g) b. By 1, 2, 3: 4. C |= a (d and g) b ⇔ a g b.
(b) ⇐ (a): For any appropriate g, C |= a (d and g) b ⇔ a g b. This means that d is an innocuous restriction for the interpretation of dd’. Since c’ 127 is the strongest such restriction, it must entail d.
A Theory of Triviality Local Felicity (Stalnaker) For an expression E to be felicitous, it should neither be the case that its local entails it, nor that it entails its contradiction.
Problem: general deviance when the 1st conjunct entails the 2nd (but - oops - this doesn’t follow from Be Brief!) a. John resides in France and he lives in Paris. b. #John lives in Paris and he resides in France.
The deviance of (b) immediately follows from Local Felicity and our reconstruction of local contexts.
128
Explanatory ?
Costly Acc. ?
Yes Yes
Exactly 5
Dynamic semantics
At least 5, less than 5
No
Universal Inferences and Processing
Yes
Yes No
Transparency Theory Yes Almost Almost No? Yes Trivalent Approaches ? No ? ? Yes Local Contexts Yes Almost Almost Yes Yes (A more systematic formal comparison is attempted in the Appendix of Schlenker 2008)
129
General Results A > B means: A predicts stronger presuppositions than B Incremental Kleene
Incremental Transparency ‘=’ Incremental Satisfaction ≈ Dynamic Semantics
>
‘=’ Symmetric Satisfaction
Incremental Supervaluations
>
> Symmetric Transparency
=
✕
Symmetric Kleene > Symmetric Supervaluations
130
Alternative Directions Constraining Dynamic Semantics a. Unger and van Eijck 2007: deriving dynamic semantics from epistemic logic. b. LaCasse 2007: constraining the type of lexical entries that can be defined within Dynamic Semantics
Trying something completely different Chemla 2008 offers a unified theory of presuppositions and implicatures, based on very simple but very different principles. It makes very subtle predictions with respect to presupposition projection in quantified cases. 131
Conclusion An embarrassment of plenty At least 5 new theories provide plausible and predictive accounts of projection.
Importance of new data a. Fine-grained data about quantificational cases are crucial. b. Reaction times would be helpful.
Left-right asymmetry -it can be made to follow from the syntax -it might be a (processing?) bias rather than a purely semantic fact. 132
Perspectives: the Triggering Problem
133
The Triggering Problem Intuition: -There couldn’t be a verb know* which was bivalent and had the same truth conditions as know. x knows* p ≈ p and x believes p [with no presupposition]
-But current accounts don’t rule out know*. The Holy Grail of Presupposition Generation Find an algorithm that predicts [in a context] the truth and failure conditions of E from its truth conditions alone.
True Non-true
Triggering
True False Failure
Alternative: the Holy Grail doesn’t exist - know* exists e.g. difference between be aware that p vs. be right that p
134
Directions Pragmatic intuitions [Grice, Stalnaker; Simons; Abbott] Presuppositions are triggered by pragmatic considerations e.g. x knows p asserts 2 things at once [about x / the world] Problem 1: No predictive account has ever been given. Problem 2: Presuppositions project differently from implicatures; this should be derived.
Abusch: Lexical Alternatives x is a aware that p ≈ p and x believes that p. Alternative: x is unaware that p ≈ p and x doesn’t believe that p Reasoning: one of the alternatives is true. Problem: The alternatives are stipulated (+ projection ?)
Chemla: Presuppositions = Scalar Implicatures with more alternatives - e.g. know p has alternatives {p, not p}
135
Triggering vs. Projection Standard Position: there are 2 independent problems: Triggering: how elementary presuppositions are generated Projection: how they are inherited
Transparency: Presuppositions = Manner Implicatures Triggering: [entirely promissory]: an elementary presupposition is just a distinguished part of a bivalent meaning that ‘wants’ to be articulated as a separate conjunct, one should say It’s raining and John knows it. Projection: follows from the definition of a presupposition when some more highly ranked principles are taken into account: Be Brief >> Be Articulate e.g. do not say It’s raining and if it is already assumed that it is raining. Result: algorithm that derives most results of Heim 1983. 136
Triggering vs. (Local) Contexts Meaning
True
True Triggering False Non-true Context C Failure Announce p presupposes p when C |= (announce p) => p Meaning relative to the Global Context? She has just announced to her parents that she is pregnant. a. Talking about an 11-year old: => she is playful. b. Talking about a 30-year old: => she is pregnant.
Meaning relative to the Local Context? At a costumed party, we encounter someone with a mask. We do not know whether this is Ann, an 11-year old, or Mary, a 30-year old.
If this is Mary, the person in front of us has / has not announced to her parents that she is pregnant.
137
Appendix. Abusch’s Challenge A case of lexical arbitrariness a. John is right that Mary is pregnant. Presupposition: John believes that Mary is pregnant. b. John is aware that Mary is pregnant. Presupposition: Mary is pregnant.
Possible reply: there is a hidden syntactic difference John is right in thinking that Mary is pregnant a. (?) Which of these individuals is your mother aware that you invite home? b. * Which of these individuals is your mother right that you invited home?
138
