"Some, not all" is an oft-referenced concept, yet there is no one word for it (that I know). We deftly understand when "some" may include "all" and when it may not. How?
The Barner paper examines how children pick up on scalar implicatures, finding that they would often fail to correctly interpret sentences which imply a difference between "only some" and "all" (so-called scalar alternatives). I was particularly fascinated by the implications of an example Barner et al. used for scalar interpretation. John asks, "Did you eat my cake?" and Mary responds, "I ate some of it." Pragmatically, Mary implies that she did not eat the whole cake, even though technically she could have eaten all of it. Now, say that Mary watched the entirety of some football game, and John asks, "Did you watch some of the game?" My intuition tells me that Mary could respond, "Yes" without implying that she did not watch the whole thing. The puzzle here is parsing out when "some, not all" is and is not implied.
When the responder specifies "some", we infer some purpose behind it. We go beyond interpreting sentences as mere logical statements, instead turning to a rough Bayesian model which uses as priors our intuitions as to why certain phrases should be included. If one had eaten the whole cake, unless one is being purposefully deceptive, there is no need to mention that one had eaten some of it; however, if one had only eaten some of the cake, they may want to convey that information. In the case of the football game, John may not seem particularly interested in how much Mary had watched. For terseness Mary may not convey that she had seen the whole game, though she could clarify that she had if she wanted to.
So, identifying what is meant by some (or the omission of some) is a high-order learned linguistic phenomenon, unrelated to an actual understanding of scales. This is to say that an understanding of scales is necessary for an understanding of lexical scalar implicature but it does not imply it. The Stiller paper finds that young children use context for scalar implicature, concluding that results like the Barner paper's are due to the children's lexical shortcomings. This conclusion agrees with my model. Putting the two results together, it seems that the class of inference necessary to understand why we in some cases interpret "some" to mean "some but not all" comes later in development. The human brain seamlessly integrates new information classes into its predictive model of the world, so perhaps understanding this distinction follows quickly from an advanced theory of the mind.
Jack, I found it particularly exciting that you drew the concept of Bayesian inference into this discussion of scalar implicature. It certainly seems correct that our previous experiences inform the conventional and maybe even the contextual definitions we give to "some" and "all." What's interesting to think about, on the completely different philosophical subject of "being purposefully deceptive," is, given your title that English isn't perfect but it doesn't matter, if a defendant on a witness stand maintains that he/she shot some of the victims (he/she indeed shot all of them), would that be considered "lying under oath?"
ReplyDelete