Scalar Implicature

Both of the two readings this week are research papers on scalar implicature, i.e. the implication of sentences that make assertions about some objects that other objects are excluded. For instance, the sentence “some students passed the test” pragmatically implies not all students passed the test. More specifically, each paper examines scalar implicature as a matter of language acquisition in young children.

The paper by Barner, Brooks, and Bale presents experimental evidence that failure of children to compute scalar implicature results from ignorance of scalar alternatives. In the sole experiment, children were shown to appropriately assert strengthened alternatives to scalar sentences when the word “only” was used and when the scales were presented contextually, whereas children failed to do so with context-independent utterances. This suggests that neither “pragmatic immaturity”, lack of processing resources, nor limited working memory are responsible. Crucially, children are able to determine the lexical meaning of sentences, informativeness, and appropriateness of statement strength, although they experience difficulty with context-independent scales.

At the end of the paper, the authors suggest that acquiring context-independent scalar implicature could be a matter of slow, repeated exposure to examples of contrasting scales, which prompts “grouping of lexical items.” I would be interested to know if such learning is natural in the sense that children find it easier to learn scalar implicature than contradictory pragmatic rules. This goes to the heart of whether the pragmatics we practice is principled, arbitrary, or somewhere inbetween. We can imagine all sorts of pragmatic rules - good and bad - but what makes it the case we adopt some and not others? For instance, we could adopt the rule that saying “some animals are asleep” should not imply that some animals are awake, because if the speaker knew whether some animals were awake, he could have said “only some animals are asleep.” That is, we assume that ambiguity in an utterance is due to ignorance of the speaker. This rule, although it contradicts scalar implicature, is not totally irrational. It would be intriguing (but not without ethical concerns) to try to teach such “incorrect” pragmatic rules to children.

The paper by Stiller, Goodman, and Frank discussed three experiments. The first showed that children could succeed, though not as well as adults, at ad-hoc implicature, suggesting either that failure of children to compute scalar implicature could be due to lack of lexical knowledge to a greater extent than pragmatic immaturity. Further, the second and third experiments respectively show that implicature is not a straightforward counterfactual and is sensitive to the rarity of features as a measure of informativeness. These results are taken to show that implicature integrates real world knowledge.

It seems to me the results of the second paper fit with the discussion in the conclusion of the first. Both emphasize real-world lexical knowledge as an important component of scalar implicature.

The second paper also prompted me to think about the claim that we operate under the Gricean maxim that speakers should be exactly as informative as required. It strikes me that this isn’t a complete scientific hypothesis until we posit precisely how we determine what information is required. One possibility is that “what information is required” is also determined pragmatically or learned by experience, e.g. we learn that when stating whether animals are asleep or awake, we should give the listener enough information to distinguish between the particular scalar alternatives “all” and “some.”