Projects in Previous Years

1st Semester 2017/18: Pragmatic reasoning on vague semantics

Alexandre Cremers
If you are interested in this project, please contact the instructor(s) by email.
Pragmatic reasoning is a very active field in theoretical and experimental linguistics. A particularly well-studied case is that of implicatures. These are inferences which can be drawn from an utterance by assuming that the speaker followed some standard maxims of conversation (Grice, 1967), as in example (1):

(1) Mary isn't much taller than John

From (1), we tend to infer that Mary is still somewhat taller than John, but this does not follow from the semantics. The inference can even be cancelled if the speaker follows up with "she might even be shorter than him". The standard explanation is that we consider alternative sentence (1') below.

(1') Mary isn't taller than John

The speaker could have used (1') to describe the situation, because it would have shorter and more informative than (1). A plausible reason why she chose not to utter (1') would be that she did not think that it is true. If we take the speaker to be well informed on Mary's height, this means she would even know that it is false.
The maxims at play here are Quality (only say things that you know to be true), Quantity (make your contribution as informative as is required), and possibly Manner (be brief).
The reasoning leading to the inference of the negation of (1') from (1) is very general, and has been shown to apply to a wide range of sentences. There are however exceptions. (2) is such an exception.

(2) Mary isn't very tall for a 12-year-old.

With neutral prosody, (2) does not convey that Mary is somewhat tall for a 12-year-old. It may even suggest that she is rather short (Horn, 1989).
In a recent paper (Leffel et al, in prep), we argue that the contrast between (1) and (2) comes from the fact that "tall for a 12-year-old" is vague, whereas "taller than John" is not. Vagueness is a widespread phenomenon in natural languages, by which predicates do not have clear membership boundaries. For instance, it is possible to find someone who wouldn't count as "tall" but would also not count as "not tall".
While vagueness has been well-studied, its interaction with implicatures have not. We proposed an informal account for the lack of implicature in (2) along the following lines: (2) and its implicature ("Mary is tall for a 12-year-old") are both vague statements about Mary, of opposite polarity (one conveys a vague upper-bound on her height, the other one a vague lower-bound). Because there isn't much distance between the thresholds for 'tall' and 'very tall', the combination of (2) and its implicature isn't fully compatible with any possible heights for Mary.
Students participating in this project will be free to choose among a variety of possible developments (or suggest their own):
  • Explore possible ways to formalize the idea above. A good candidate would be to extend the idea of innocent exclusion developed by Fox (2007) for implicatures to a semantics based on fuzzy logic. An alternative would be to look into recent developments in Bayesian pragmatics.
  • Reach a better understanding of the semantic and pragmatic contribution of intensifiers such as 'very' and 'much'. This could be done using data from Leffel et al.
  • Look at other cases of interactions between vagueness and implicatures.
  • Experimental work may be considered if funding is available