Three experiments and a simulation study investigate competing featural and phonemic views of the representation of the speech input in access to the mental lexicon. Auditory lexical decision and gating tasks show that the processing consequences of subcategorical mismatches (conflicts between phonetic cues to speech segment identity) depend on the lexical status of the conflicting cues, such that conflicts that only involve nonwords do not disrupt performance. A further study, using a phonetic-decision task with the same stimuli, found the same pattern. A simulation study shows that the interactive activation model TRACE, with top-down feedback to a prelexical phonemic level, does not model these effects successfully. The authors argue instead for a direct access featural model, based on a distributed computational substrate, where featural information is mapped directly onto lexical representations.

This experiment explores the role of prosodic cues in resolving temporary morphosyntactic ambiguities in spoken language comprehension. Using a cross-modal naming task, we find that prosodic cues are as effective as overt lexical cues in controlling how the listener resolves attachment ambiguities. This suggests that prosodic factors can affect the early stages of parsing and interpretation.