UC San Diego

A difficult learning problem for both children and artificial language learning systems is knowing what is intended to be conveyed based on what is literally said. For example, adults usually take "Some teas contain caffeine" to also convey that "Not all teas contain caffeine", an inference known as a scalar implicature. The present work investigates the role of language-specific knowledge in such inferences through three studies on scalar implicatures in American Sign Language (ASL). The first study illustrates a new experimental paradigm and compares prototypical scalar implicatures in ASL and English. The second study includes the first investigation of general use coordinators in ASL that can be interpreted as either conjunction ("and") or disjunction ("or"). This provides a test case for the role of language-specific lexical contrast in scalar implicatures, with results showing that lexically non-contrastive scales (i.e., lexical scales whose items differ in meaning but not in form) trigger less scalar implicatures than prototypical lexically contrastive scales, which are based on contrasting lexical items. In the third study, both lexically contrastive and non-contrastive scales are interpreted by deaf native signers and also deaf signers who learned ASL at later ages. Result show that later ASL learners calculated less implicatures than early learning signers, but only on the lexically contrastive scale. Together, these studies support a view that despite their context dependence, scalar implicatures are most likely to be triggered by lexical items which contrast with each other in form to create a context-independent "scale", and that there may even be advantages to learning scales early in life. The dissertation concludes with suggestions for incorporating lexical contrast into theories of implicature and for further study of the semantic/ pragmatic interface in sign languages