An interval based semantics for negative degree questions MÁRTA ABRUSÁN (Institut Jean Nicod, ENS) & BENJAMIN SPECTOR (Harvard Society of Fellows) JSM :: Toulouse :: April 4, 2008
Introduction Goal: To give a new account for negative islands with degree questions: (1) Negative Islands a. Who didn’t John invite to the party? b. *How tall isn’t John? ..and their obviation (Fox and Hackl 2005): (2) How much radiation are we not allowed to expose our workers to? (3) * How much radiation are we not required to expose our workers to?
The core ideas to be developed: INTERVALS: degree predicates denote relations between individuals and intervals: (Schwarzschild and Wilkinson (2002), Schwarzschild (2004), Heim (2006)) If degree questions range over intervals, the presupposition that a question should have a MAXIMALLY INFORMATIVE TRUE ANSWER (Dayal 1996) can never be met, and (1)b and (3) result in a presupposition failure.
1
The plan 1
A brief overview of the analysis
2
Comparison with Fox and Hack’s (2005) and Rullmann’s (1995) account
3
Our analysis in more detail
4
A problem? Apparent cases of overgeneration
5
The solution to our problem
6
Conclusion
7
Appendix 1: cases of undergeneration: the Π-operator
8
Appendix 2: fast vs. slow
2
Overview of the account Dayal (1996) has proposed that a question presupposes that it has a most informative true answer. (4) Maximal Informativity Hypothesis (Dayal 1996) A question presupposes that it has a maximally informative true answer Intervals: Schwarzschild and Wilkinson (2002), Schwarzschild (2004), Heim (2006): degree predicates denote relations between individuals and intervals (sets of degrees): (5) A set of degrees D is an interval iff: For all d, d’, d’’: if d∈D & d’’∈D & d≤d’≤d’’, then d’∈D (6) tall=λI. λxe. x’s height ∈I (7) How tall is John?w = λp.∃I [I∈DI ∧ p=λw’. John’s height ∈I in w’] ‘For what interval I, John’s height is in I?’ We predict that examples of negative islands such as (9) result in a presupposition failure, given the above two assumptions: (8) Who didn’t John invite to the party? (9) *How tall isn’t John 3
Positive and Negative degree questions Positive degree questions (10) How tall is John?w
= λp. ∃ I [I∈Dintervals & p=λw’. John’s height ∈ I in w’] ‘For what interval I, John’s height is in I?’
(11) -------------------------------[---(---J’s height ---)---]--------------------------------Most informative answer : take I = {John’s height} Negative degree questions (12) *How tall isn’t John?w = λp. ∃ I [I∈Dintervals & p=λw’. John’s height ∉ I in w’] ‘For what interval I, John’s height is not in I?’ (13) ---------[----------------------] I2-- J’s height ----------{-------------------------}I3--Intuitively: We are looking for the maximal interval among the intervals in which John’s height is not contained. Given the negative context, if an interval K covers interval I, the truth of John’s height ∉ K will entail the truth of John’s height ∉I. The problem is that there is no such maximal interval. 4
Negative degree questions, more precisely (14) *How tall isn’t John?w = λp. ∃ I [I∈Dintervals & p=λw’. John’s height ∉ I in w’] ‘For what interval I, John’s height is not in I?’ (15) ---[----------------------] I2------dj ---------{--------------}I3--More precisely: (i) Let John’s height be any non-zero degree d. (ii) The set of all intervals that do not include John’s height (=N) contains exactly two exclusive sets of intervals: all the intervals fully below d, contained in [0, d[ (=A) and all the intervals fully above d, contained in ]d, +∞) (=B). (iii) It is easy to see that for any interval I included in A, the (true) proposition that John’s height is not in I, does not entail that John’s height is not in B, and vice versa. (iv) Hence, there is no interval I in N such that the true proposition that John’s height is not in I entails all the true propositions of the same form in N. Dayal’s (1996) condition cannot be met, and we predict a presupposition failure. 5
Modal obviation Fox and Hackl (2005) observe that existential modals below negation improve negative islands: (16) a. b.
How fast are we not allowed to drive? For what I, it is not allowed that our speed be in I?
This fact is straightforwardly predicted by the present account: While wrt. (9) it was a fact about the world that John’s height is a single degree, the degrees of speed with which we are allowed to drive might correspond to an interval, e.g. ]0, 120km/h], if the law states that our speed must be at most 120 km/h. Now any I that is entirely above 120km/h is such that it is not allowed that our speed be in I. In this case, there is a strongest true proposition among the H/K alternatives to (16): the true proposition that the speed with which we are not allowed to drive is ∈ ]120, +∞). Dayal’s condition is met Same for universal modals above negation: How fast are we required not to drive? 6
Modal obviation (contd.) Existential modals above negation however do not obviate negative islands: (17) a. # How fast are we allowed not to drive? b. For what I, it is allowed that our speed be not in I? Suppose we have some obligations as to what our speed should be: Call S the set of all the speeds such that our speed is required to be one of them: E.g: (i)
0------ allowed not--------[d--------[--allowed not--]------------>
S: required For any subset of S which does not equal S, it is allowed that our speed is not in it. Also, for any subset of the complement set of S, it is allowed that our speed is not in it. The interval which covers all the intervals such that our speed is allowed not to be in it is therefore [0,+∞[. However this interval cannot be the maximally informative interval, because it will also cover the interval(s) for which it is required that our speed be in it, therefore Dayal’s presupposition is not met. If we have no obligations for our speed, the maximal interval for which it is allowed that our speed is not in it is [0,+∞[. Dayal’s presupposition is met, however, the answer means that there is a possible world in which our speed is not in [0,+∞[, which is impossible. 7
Comparison with Fox & Hackl (2005) Replacing Maximality with Maximal Informativity (cf. Beck and Rullmann 1989) ( this is similar to our account) Dayal’s assumption A question presupposes that it has a unique maximally informative answer, i.e. a true answer that entails all the other true answers. A question asks for the maximally informative answer But given a monotonic semantics for degree predicates negative islands are not predicted: (18)
a.* How tall isn’t Jack? b. For what d, Jack isn’t d-tall
The maximally informative answer should ask for the smallest degree d such that Jack is not dtall. There should be no problem if there is such a degree. F&H’s solution: Dense scales The smallest degree d such that Jack is not d-tall does not exist.
8
Fox & Hackl (2005) (Contd.) Dense scales Let d be Jack’s height: Suppose that d’ = d + ε is the smallest degree such that Jack is not d’-tall. Take d’’ = d +ε/2. Clearly, Jack is not d’’-tall; but d’’ is smaller than d’, which is contradictory. Intuitively: Let Jack’s height be 180 cm. The set of all true propositions of the form Jack is not dtall is the following: (19) {…,Jack is not 180,000001 cm –tall, ..., Jack is not 180,05 cm-tall,…., Jack is not 181 cm-tall,….}
It will be apparent that there is no minimal degree d such that Jack is not d-tall. This is simply because for any d > 180cm, there is a d’ such that d > d’ > 180cm. Dayal’s condition is not met. Discrete scales F & H must extend this account even to cases where the domain of degrees is not “intuitively” dense, such as cardinality measures (a certain sort of degrees), as in: (20) *How many children doesn’t Jack have? 9
Fox and Hackl (2005) on Discrete scales, modularity and blindness to contextual information (21) *How many children doesn’t Jack have? Account: suppose Jack has exactly 3 children. Then he does not have 4 children, but he also does not have 3.5 children, or 3.00001 children… A natural objection: even granting that it makes sense to say that Jack has 3.5. children, yet the exact number of children someone has is always an integer. So Jack does not have 3+ε children is known to be equivalent to Jack does not have 4 children, and there is in fact a true answer that entails all the other ones. Modularity F & H have to assume a very strong modularity assumption: presumably, the knowledge that the number of children someone has is an integer is a form of lexical/encyclopedic knowledge. According to them however, this knowledge is not purely logical, given some reasonable notion of logicality (one that is blind to lexical semantics/encyclopedic knowledge). F & H’s central claim is that Dayal’s condition is computed only on the basis of the purely logical meaning of the question, i.e. is blind to contextual, encyclopedic or lexical information. while there might be arguments for this view, we would like to explore an alternative 10
Fox and Hackl (2005): Accounting for the modal obviation facts Existential modal under negation (22) How fast are we not allowed to drive? Before, it was a fact about the world that predicates such as λd. we are d-fast denote a closed interval, hence its complement denoted an open interval given the assumption that scales are dense. As a result, Dayal’s condition could not be met. Here, the predicates of the form λd. POSSIBLE (P(d)) can denote open intervals, and therefore their complements, i.e. predicates of the form λd.¬POSSIBLE(P(d)) can denote closed intervals. Likewise for λd.NEC(¬P(d)). Illustration Suppose that the law states that our speed should be lower than 65 km/h, and says nothing more. It follows that the set of worlds compatible with the law is {w: our speed is lower than 65 km/h}. So for any speed v below 65 km/h (however close to 65 km/h), there is a permissible world in which our speed is v. Hence for any speed lower than 65km/h, we are allowed to drive at that speed. On the other hand, we are not allowed to drive at 65km/h. Hence 65 km/h is the lowest speed v such that we are not allowed to drive at speed v. Dayal’s condition can be met. 11
Returning to our account: Discrete scales Nothing special needs to be said about discrete scales. (23) A. b.
How many children doesn’t Jack have? For what interval I of integers, the number of children Jack has is not in I?
Suppose Jack has exactly 4 children: Then the set of intervals that correspond to a true answer are all the intervals strictly below 4 and every interval strictly above 4. But no interval in this set corresponds to a maximally informative answer (as before)
12
Argument for the interval reading That the interval reading indeed exists is shown by the following example: (24) How fast are we required to drive? Suppose that one the highway we should drive between 60km/h and 120 km/h. Then the complete answer is predicted to state exactly this. Scenario Jack and Peter are devising the perfect Republic. They argue about speed limits on highways. Jack believes that people should be required to drive at a speed between 60km/h and 120km/h. Peter believes that they should be required to drive at a speed between 60km/h and 140 km/h. Therefore… (25) Jack and Peter do not agree on how fast people should be required to drive on highways The disagreement is about the maximal speed, not the minimal speed, as the standard account would have it
13
The overgeneration problem Problem # 1 (26) How fast are we required to drive? • Case a: there is a minimal speed s: hence the maximally informative answer is the interval [s, +∞). This corresponds to our intuitions • Case b: there is a maximal speed s. There is a maximally informative answer, namely the one based on [0, s]. But in fact (26) is unnatural if it is known that there is a maximal speed and no minimal speed. This is an overgeneration problem, in the sense that our account so far predicts a question like (26) to be felicitous in too many contexts. • (Note that the corresponding question is perfectly fine in a language like French, where instead of how fast/how slow one finds At what speed) Problem #2 (27) How fast are we not allowed to drive on this highway? Normally, one understands from (27) that one is not allowed to drive too fast. But suppose that in fact the only obligation we have is to drive at a speed higher than 60 km/h, and there is no maximal speed. Then the most informative interval I such that we are not allowed to drive at a speed in I is [0, 60], and so Dayal’s condition is met. Again we fail to predict that a question like (27) imposes a further restriction, namely implicates that there is a maximal speed. 14
Solving the overgeneration problem Assumption : Let HowI{S,<} φ(I) ? be a degree question, where I is an interval based on a scale S whose ordering relation is <. (28) HowI{S,<} φ(I) ? presupposes the following: There is a true answer φ(I) with I being an interval such that its lowest bound x is not 0 and for any y > x, y ∈ I. (29) Fast:
0 <1 <2<3<4<5<6<7<8<9< ∞
maximal fast
I.e.: if S is open scale upward (fast), I must be of the form [x, +∞), with x distinct from 0 Solving problem #1 (30) How fast are we required to drive? The presupposition: there is an interval of the form [x, +∞) such that we are required to drive at a speed in [x, +∞). This entails that there is a minimal speed. It does not entail that there is no maximal speed. For suppose the actual regulation is that our speed should be between x and y. Then it follows that our speed must be above x, i.e. in the interval [x, +∞).
15
Solving the overgeneration problem Solving problem #2 (31) How fast are we not allowed to drive on this highway? The additional presupposition now says that for some interval [x, +∞), we are not allowed to drive at a speed in this interval. This amounts to saying that the regulations impose a maximal speed. This is the desired result Nothing is lost: (32) How fast did Jack drive? Presupposes that for some interval [x, +∞), with x distinct from 0, Jack’s speed was in this interval. This is true as long as Jack has a non-null speed. French A quelle vitesse vs. English how fast/how slow (33) A quelle vitesse a-t-on l’obligation de rouler sur cette autoroute? No overgeneralization problem for A quelle vitesse, presumably because there is no ordering. 16
Comparison with Szabolcsi and Zwarts (1993)
Similarities with the present account: The difference between the island sensitive and non-island sensitive extractees resides in the structure of the domain of quantification of these elements The explanation for negative islands is connected to the complement not being defined in certain domains. Differences: Their explanation does not extend to obviation effects by modals. The interval account is independently motivated. Under our account, the unacceptability of negative islands results from the maximally informative answer never being defined (while for Sz&Z the unacceptability results from the fact that it cannot always be defined)
17
Summary Degree questions range over intervals The presupposition that a question should have a maximally informative true answer (Dayal 1996) can never be met, and negative islands with degree questions are the result of a presupposition failure. Modal obviation is explained in a straightforward manner
18
Acknowledgements We would like to thank Danny Fox, Emmanuel Chemla, Gennaro Chierchia, Jon Gajewski, Irene Heim, Roni Katzir, Giorgio Magri, David Pesetsky, Anna Szabolcsi for comments questions and discussions.
The first author would like to acknowledge financial support by the European Science Foundation (Euryi project on presupposition, to P. Schlenker).
19
Appendix 1: The undergeneration problem Suppose we are required to drive between 80 and 120 km/h: The following utterance should express a contradiction: (1) John knows how fast we are required to drive, but does not know what the maximal speed is. Problem: (1) above seems to be fine. Existential modals: (2) A. b.
How fast are we allowed to drive? For what I, ∃w’ACC( w, w’). our speed is in I in w’?
Given that our speed corresponds to a single degree, it seems that the question should presuppose that there is a single speed such that we are required to drive at that speed. Problem: This seems wrong.
20
Appendix 1: Solution to the undergeneration problem The PI-operator (S & W, Heim) (3) [[ fast1 ]]> = λd. λx. x’s speed is at least d (4) [[ fast2 ]]<, > = λD.λx. x’s speed is in D • The PI (point-to interval)-operator (reversing the order of arguments – this is harmless) (5) Π=λP. λI. max(P)∈I (6) Π[λd. Jack is at least d-fast] = λI. MAX{d: Jack is at least d-fast} ∈ I = λI. Jack’s speed is in I. (7) [[ fast2 ]]<, > = λD.λx. (Π [λd. fast1(d)(x)])(D) The crucial point is that Π can in principle appear higher than just above the lexical degree predicate. Extension to degree questions (8) A. B
How fast is John? For what interval I, Π.(λd. John is d-fast) (I) ?
21
Appendix 1: Solution to the undergeneration problem (contd.) Two possible LFs for the question below: (9) How fast are we required to drive? (i)
For what interval I, it is required that (Π.(λd. we are d-fast) (I) = For what interval I, it is required that our speed be in I? this is the interval reading discussed before
(ii) For what interval I, Π(λd. it is required that we drive at least d-fast) (I)? ‘Π(λd. it is required that we drive at least d-fast)’ denotes the set of intervals that include the maximal speed s such that we are required to drive at least s-fast. This is the reading predicted by the standard treatment ILLUSTRATION: Suppose that on the highway we must drive between 45mph and 75mph; then this maximal speed is 45mph; and therefore‘Π(λd. it is required that we drive at least dfast)’ denotes all the intervals that include 45mph. The most informative answer turns out to be the one you get by just taking I = {45mph} Crucially, Π(λd. Jack didn’t drive d-fast) is not defined, because there is no maximum in ‘λd. Jack didn’t drive d-fast’. So Π cannot scope above negation, and negative islands are still accounted for. 22
Appendix 2: Fast vs. Slow On the interval reading the two are equivalent—but the presupposition in (28) makes different predictions for them: recall: (10) HowI{S,<} φ(I) ? presupposes the following: There is a true answer φ(I) with I being an interval such that its lowest bound x is not 0 and for any y > x, y ∈ I. Fast: Slow:
0 <1 <2<3<4<5<6<7<8<9< … <9 <8<7<6<5<4<3<2<1<
∞ 0
maximal fast maximal slow
I.e: - if S is open scale upward (fast), I must be of the form [x, +∞), with x distinct from 0 - if S is closed upward (slow): I must be of the form [x, y], where y is the maximal degree in S (here: y=0). On the “classic” (high Π) reading how fast is not equivalent to how slow : (11) How slow are we required to drive on this highway? For what interval I, Π(λd. it is required that we drive at least d-slow) ? ‘Π(λd. it is required that we drive at least d-slow)’ denotes the set of intervals that include the maximal degree of slowness d such that we are required to drive at least d-slow, i.e. the minimal speed such that we must be at most d-fast. 23
REFERENCES Abrusan, M., 2007, Contradiction and Grammar: the case of weak islands, PhD Diss, MIT Beck, S., and H. Rullmann. 1999. A Flexible Approach to Exhaustivity in Questions. Natural Language Semantics 7:249-298. Dayal, V. 1996. Locality in WH quantification: Kluwer Academic Publishers Boston. Fox, D., and M. Hackl. 2005. The Universal Density of Measurement. Unpublished manuscript. Heim, I. 2006. Remarks on comparative clauses as generalized quantifiers. ms. MIT. Rett, J. 2006. How Many maximizes in the Balkan Sprachbund, SALT 16 Rett, J. 2007, Antonymy and Evaluativity, SALT 17 Rullmann, H. 1995. Maximality in the semantics of wh-constructions, University of Massachusetts at Amherst Amherst, Mass. Schwarzschild, R., and K. Wilkinson. 2002. Quantifiers in Comparatives: A Semantics of Degree Based on Intervals. Natural Language Semantics 10:1-41. Spector, B. 2004. Intervention Effects in How-Many Questions and the Syntax/Semantics Interface. 3 page abstract for Sinn und Bedeutung 9. Spector, B. 2005/2006. Aspects de la pragmatique des operateurs logiques. PhD Diss, Paris 7. Szabolcsi, A., and F. Zwarts. 1993. Weak islands and an algebraic semantics for scope taking. Natural Language Semantics 1:235-284.
24
