Barner and Stiller

Barner: It is easy to see that intended meaning can, and often does, go beyond literal meaning. For instance, if I say “I read some of the book”, this utterance is typically taken to imply that I did not read ALL of the book. In such a case, the implication that is drawn is not a logical consequence of the utterance, I.e., the implication that is drawn is not contained within the semantics of the utterance. Children are usually pretty bad at making these inferences; for example, when they are shown a picture of 3 out of 3 horses jumping a fence and asked whether the description “some of the horses jumped over the fence” is a good one, children will almost always say “yes”, whereas adults will almost always say “no”. However, children are usually pretty good at making inferences when it comes to utterances containing numbers, so if someone said “2 horses jumped over the fence” to describe a picture of 3 horses jumping over the fence then children will be able to identify it as a bad description. One of the questions explored in the paper is why children display this asymmetry in pragmatic ability, I.e. Why they make this pragmatic inference in the case of numbers, but not in the case of “all” and “some” - the hypothesis proposed in the paper is that whereas children have explicit knowledge of numbers as belonging on a scale, they are not yet able to see that “some” and “all” likewise form a scale. As a result, when children hear a sentence like “3 horses jumped over the fence” they are able to recognize sentences like “4 horses jumped over the fence” or “5 horses jumped over the fence” as scalar alternatives; then they can negate all of these (more informative) scalar alternatives to arrive at the strengthened interpretation, viz. “3 and ONLY 3 horses jumped over the fence”. But when they hear the sentence “some horses jumped over the fence” they do not see that “all horses jumped over the fence” is a relevant alternative and thus they do not negate it in strengthening their interpretation of the original utterance.

Stiller: This paper is also about scalar implicature. The experiments described in this paper are designed to test several different theories of implicature. These theories are: (1) The Gricean Theory which says that people derive implications from two fundamental principles, viz., contributions are as informative as possible and contributions are not more informative than they need to be. (2) The Linguistics Alternative model which says that implications are derived from a process of negating a set of alternatives of a limited sentential complexity. The Gricean model is problematized by the second experiment presented in the paper. The experimenters had participants look at a set of three faces. Each face had exactly 2 features. The possible features a face could have were: (1) blue hat, (2) top hat, (3) a moustache, (4) glasses. The first face had (1) and (3), the second face had (1) and (4), and the third facce had (2) and (4). Participants were asked to pick out “the one with glasses”. The Gricean model would predict that participants would go for the second face, since Gricean agents would be able to see that if someone meant to point out one of the other two faces, they would have used a different description instead. The third experiment demonstrates that knowledge of the distribution of features influences the implicatures drawn. (So if A knows that top hats are very rare, then A will tend to infer that B did not mean to pick out the person with the top hat if B does not include information about the top hat in her description.)

Children are said to be bad at making scalar implicatures, and the evidence for this is roughly something like the following: when children are shown pictures where N out of N objects have feature X, and they are asked “is ‘Some objects have feature X’ a good description of the picture?” they will usually say “yes”. I wonder though if this result really demonstrates that children are bereft of pragmatic ability - it seems to me that the children could have simply interpreted the question as “is ‘Some objects have feature X’ a TRUE description of the picture?” (while the adults probably interpreted it as “is ‘Some objects have feature’ a FELICITOUS description of the picture?”) - but if the children really did interpret the question in this way, then their answer does not by itself demonstrate that they are not sensitive to the felicity of a statement. It would have been more telling in my mind if the experimenters told the children participants “Your friend says: ‘some of the horses are jumping over the fence’, which picture do you think your friend is looking at?” and at the same time the experimenter holds up two pictures: one with 2 out of 3 horses jumping over a fence and the other with 3 out of 3 horses jumping over the fence. If the participants usually pick the first picture, this would show that they are sensitive to the felicity of the description ‘some’, i.e., that they recognize that ‘some’ pragmatically excludes ‘all’; on the other hand, if they picked both pictures with roughly equal frequency, then this would show that they are not aware that ‘some’ has any pragmatic import that excludes ‘all’.