Recent speech error research has challenged the traditional view that the temporal mislocation of a segment leaves its structural phonological integrity unaffected. Results of several studies suggest the existence of subsegmental, phonologically ill‐formed errors (e.g., Mowrey and MacKay, J. Acoust. Soc. Am. 88, 1299–1312 (1990)]. While Pouplier etal. [J. Acoust. Soc. Am. 106, 2242 (1999)] could demonstrate on the basis of articulatory data that errors involve systematic movements of individual speech gestures, their data are limited to tongue‐twister‐like stimuli. This leaves open the possibility that the errors they observe might arise at a low execution level given the repetitive nature of the task. The present study employs the SLIP elicitation technique [Motley and Baars, Commun. Quart. 24, 28–24 (1976)] which is traditionally hypothesized to reflect errors at an abstract phonemic planning level (subjects produce target word pairs in isolation). Results show that also in presumably abstract ‘‘planning errors,’’ often two gestures are coproduced (one erroneous, one appropriate), resulting in an illicit phonological structure. These findings call into question the assumption that segments are primary units of speech production. [Work supported by NIH.]