Phonetics and Phonology in Language Comprehension and Production: Differences and Similarities

The rationale for this volume is that much research in the past has focused either on speech perception or on production, with surprisingly little interaction between these two fields of investigation. The book contains nine chapters by expert researchers in perception and production, with the stipulated requirement that each contribution should make reference to both processes in order to highlight how they are similar and how they differ. Of course, the processes are rather different, so it is not possible just to turn the speech production process backwards to understand how perception works. Listeners must make guesses about what is being said, so crucially they have to maintain and constantly re-evaluate parallel hypotheses on the basis of the incoming stream of sounds until they have determined which one is most likely to be correct, while production involves no such parallel guesswork. And speakers must achieve an extraordinarily intricate synchronisation of a whole range of different articulators, a muscular coordination effort which is not necessary for perception. However, presumably both processes access the same store of knowledge in the brain, so one assumes that the phonological representation of words must be in a form that both recognition and production can utilise. Furthermore, it seems likely that at least some of our speech handling faculties are shared by production and perception, so the question then is: at what level do the two processes become separate, and to what extent do they make use of shared utilities? Some of the chapters in this book concentrate on the mental representation of phonology, while others focus more on the processes involved in speech production and perception, particularly considering models that attempt to simulate the processing of speech. One chapter that seeks to provide a summary of the current state of thinking in all these areas, including phonological representation as well as models of speech processing for both perception and production, is the one by James McQueen, Delphine Dahan and Anne Cutler. It consists of a substantial but rather densely packed overview of research into whether phonological information is categorical (dealing with phonemes) or more finely graded and whether phonological and lexical processing in perception and production proceeds in a serial or a cascaded manner. Their comprehensive summary concludes that, for recognition the processing is cascaded, with multiple candidate words being evaluated in parallel, and the information is finely graded, with access to far more detailed information than just phonemes, but in speech production the processing is mostly serial, with the possibility of only very limited cascading between the levels, and the information is categorical, dealing with phonemes and perhaps sometimes with an inventory of complete syllables. This fundamental contrast between perception and production makes sense, of course, as only in recognition is there a need for detailed goodness-of-fit scores to be maintained while multiple candidate parses are being evaluated in parallel, and moreover as semantic processing is sometimes essential for the successful identification of lexical items, strict serial processing would not work for speech comprehension. However, the conclusion that simultaneous parallel processing occurs for the various stages during perception is challenged by Miranda van Turennout, Bernadette Schmitt, and Peter Hagoort, who report on the performance of WEAVER in modeling the temporal interaction of phonological, syntactic and semantic processing in the production and perception of speech, particularly in the light of electrophysiological research involving the lateralized readiness potential (LRP) and also the negative going potential N200 found in certain electrode sites near the front of the brain during a variety of picturenaming tasks. This neural imaging research indicates that, in speech production, there is mostly serial processing of information, from semantic concept to syntactic form and finally to phonological encoding, and this is well represented by WEAVER . And even though this model was Libri