Combining sources of evidence in speech perception
Dominic W. Massaro · The Journal of the Acoustical Society of America · 1981
A central issue in speech perception research is how the listener evaluates and integrates the multiple sources of evidence in the message. Bottom-up theories stress the contribution of acoustic featural information specifying the identity of small segments of the speech signal. Top-down theories stress the contribution of higher-order linguistic and semantic constraints in the disambiguation of the message. An alternative model is proposed to account for both bottom-up and top-down contributions to speech perception. The critical assumption of the model is that acoustic featural information and higher-order constraints provide independent sources of information. These sources are evaluated and integrated together by the listener. Evidence from experiments varying acoustic featural information, phonological constraints, lexical context, and sentential constraints will be discussed in the framework of the model. [Work supported by NIMH.]