The transition model (i) of classical phonetics distinguished between inevitable transitions between momentary target configurations (called coarticulation by Menzerath and Lacerda in 1933) and assimilation extending over a larger domain than a transition, implying reorganization of the input. By the 1950s it was clear that coarticulation extended beyond target configurations and involved more than two phonemes, and sub‐cortical tugofwar models (ii) came to be preferred, based on competition between phonemes for muscles and articulator movement [e.g., Öhman, J. Acoust. Soc. Am. 39, 151–168 (1966)] and seeing coarticulation and assimilation as synonymous. Finally (iii) gesture queuing models [e.g., Kozhevnikov and Chistovich, Speech Articulation and Perception (1965)] delay gestures that are antagonistic to on‐going activity, implying a cortical scanning procedure to survey on‐coming input. Examples of articulator timing, analyzed from x‐ray motion films of speech, are presented that favor (iii) rather than (ii). It is argued that much current controversy over coarticulation can be avoided if a cortical level of motor control and the distinction between coarticulation and assimilation were accepted again. The procedures and some data are presented in Wood [J. Phon. 19, 281–292 (1991)].

This content is only available via PDF.