In theories of spoken language production, the gestural code prescribing the movements of the speech organs is usually viewed as a linear string of holistic, encapsulated, hard-wired, phonetic plans, e.g., of the size of phonemes or syllables. Interactions between phonetic units on the surface of overt speech are commonly attributed to either the phonological encoding stage or the peripheral mechanisms of the speech apparatus. Apraxia of speech is a neurogenic disorder which is considered to interfere with the mechanisms of phonetic encoding. Analyses of apraxic speech errors have suggested that phonetic representations have a non-linear, hierarchically nested structure. This article presents a non-linear probabilistic model of the phonetic code, which embraces units from a sub-segmental level up to the level of metrical feet. The model is verified on the basis of accuracy data from a large sample of apraxic speakers. (Contains 3 figures, 3 tables, and 1 footnote.)