The non-linearity of speech production

The traditional linear conception of speech production shared by the majority of formalists and functionalists alike envisages a step-by-step succession of stages on the way from conceptual intention to speech chain product. However, the alternative conception of parallel distributed processing has for some time been gaining favour in such quarters. This contribution presents facts from a number of languages, including English, that support the idea that speech processing must be in large part non-linear, including at the lower levels involved in lexical access and phonological spell-out. Top-down contextual processesmust constrain bottom-up generation, even when the relevant contextual trigger appears to lie further ahead in the linear speech chain. A Whiteheadian perspective is adopted to help gain insight into the non-linear processes involved in speech production and comprehension.