When discriminating pairs of stimuli from an acoustic voice‐onset‐time (VOT) continuum, English‐speaking listeners typically show a performance peak in the region of the phonetic category boundary. An ongoing series of experiments demonstrates that the location and height of this peak are affected by preceding phonetic context. While any preceding context appears to lower discrimination performance, perhaps by interfering with auditory memory, a preceding [s] also shifts the peak towards the short‐VOT end of the continuum and, correspondingly, increases “voiceless” responses in a labeling task. This latter effect (trivial within a syllable because of English spelling conventions) occurs even when a word boundary intervenes and does not seem to be mediated by the duration of the silent closure interval following the [s]. A hypothesis currently being pursued is that phonological voicing decisions for word‐initial stop consonants are sensitive to the voicing status of preceding phonetic segments. [Work supported by NICHD.]