Psycholinguistic and neurolinguistic investigations of scalar implicature

Abstract

The present study examines the representation and composition of meaning in scalar implicatures. Scalar implicature is the phenomenon whereby the use of a less informative term (e.g., some) is inferred to mean the negation of a more informative term (e.g., to mean not all). The experiments reported here investigate how the processing of the implicature-based aspect of meaning (e.g., the interpretation of some as meaning not all) differs from other types of meaning processing, and how that aspect of meaning is initially realized. The first three experiments measure event-related potentials (ERPs) to examine whether inferential pragmatic aspects of meaning are processed using different mechanisms than lexical or combinatorial semantic aspects of meaning, and whether inferential aspects of meaning can be realized rapidly. Participants read infelicitous quantifiers for which the semantic meaning (at least one of) was correct with respect to the context but the pragmatic meaning (not all of) was not, compared to quantifiers for which the semantic meaning was inconsistent with the context and no additional pragmatic meaning is available. Across experiments, quantifiers that were pragmatically inconsistent but not semantically inconsistent with the context elicited a broadly distributed, sustained negative component. This sustained negativity contrasts with the N400 effect typically elicited by nouns that are incongruent with their context, suggesting that the recognition of scalar implicature errors elicits a qualitatively different ERP signature than the recognition of lexico-semantic errors. The effect was also distinct from the ERP response elicited by quantifiers that were semantically inconsistent with a context. The sustained negativity may reflect cancellation of the pragmatic inference and retrieval of the semantic meaning. This process was also found to be independent from lexico-semantic processing: the N400 elicited by lexico-semantic violations was not modulated by the presence of a pragmatic inconsistency. These findings suggest there is a dissociation between the mechanisms for processing combinatorial semantic meaning and those for inference-based pragmatic meaning, that inferential pragmatic meaning can be realized rapidly, and that the computation of meaning involves continuous negotiation between different aspects of meaning. The next set of experiments examined how scalar implicature-based meanings are realized initially. Default processing accounts assume that the interpretation of some of as meaning not all of is realized easily and automatically (regardless of context), whereas context-driven processing accounts assume that it is realized effortfully and only in certain contexts. In two experiments, participants' self-paced reading times were recorded as they read vignettes in which the context did or did not bias the participants to make a scalar inference (to interpret some of as meaning not all of). The reading times in the first experiment suggested that the realization of the inference was influenced by the context: reading times to a target word later in the vignette were facilitated in contexts in which the scalar inference should be realized but not in contexts where it should not be realized. Importantly, however, reading times did not provide evidence for processing cost at the time the inference is realized, contrary to the predictions of context-driven processing accounts. The results raise the question of why inferencing occurs only in certain contexts if it does not involve extra processing effort. In the subsequent experiment, reading times suggested that the inference may not have been realized when participants engaged in a secondary task that increased processing load. These results, together with the results of other recent experiments, suggest that inferencing may be effortless in certain contexts but effortful with other contexts, and not computed at all in still other contexts, depending on the strength of the bias created by the context. These findings may all be accountable for under a recently-proposed constraint-based processing model of scalar implicature.