Sentential context apparently has a more influential role in the selection of meanings for lexically ambiguous words (e.g., bank) than the identification of spoken words. Priming studies show that the preceding context of an ambiguous word can have an immediate (but often weak) effect on the activation of its meanings, suggesting some interactive processing (Lucas, 1999). On the other hand, sentential context is often viewed as unimportant for the identification of spoken words, with any effects restricted to the integration stage of spoken word recognition (Connine, Blasko & Wang, 1994; Marslen-Wilson, 1989).
This distinction between the effects of context in the two processes can be interpreted in different ways. It is normally assumed to reflect an architectural difference, such that lexical identification and meaning retrieval are located at different levels, with different rules for the involvement of semantic information. But it is also plausible that the distinction can be accommodated by a single system, in which the only difference between the domains is in the availability of disambiguating information—sentential context will generally be unnecessary for spoken word identification because bottom-up information will normally resolve any ambiguity, but it will always be needed for the resolution of lexical ambiguity.
Three experiments are reported, which examine spoken word identification but in a situation more like lexical ambiguity. The perceptual effects of utterances such as "A quick rum picks you up" are examined, in which the underlined sequence is ambiguous. It could be simply the word rum, but it could also be a token of run, in which the /n/ has fully assimilated to /m/ because of the influence of the following segment. In this case, there is no single lexical item (like bank) with two possible meanings, but nonetheless the acoustic input is insufficient to decide between the two lexical alternatives (run and rum). The repetition priming studies show that sentential context does indeed have a significant immediate effect on word activations in this ambiguous case. These results can be explained by a model of spoken word recognition in which sentential cues are integrated with bottom-up sources of information throughout the course of word recognition, but that the probabilistic nature of sentential context effects means that they are normally overridden by more reliable bottom-up cues.
References
Connine, C. M., Blasko, D. G., & Wang, J. (1994). Vertical similarity in spoken word recognition - multiple lexical activation, individual-differences, and the role of sentence context. Perception & Psychophysics, 56, 624-636.
Lucas, M. (1999). Context Effects in Lexical Access: A meta-analysis. Memory and Cognition, 27, 385-398.
Marslen-Wilson, W. D. (1989). Access and integration: Projecting sound onto meaning. In W. D. Marslen-Wilson (Ed.), Lexical representation and process. Cambridge, MA: MIT Press.