Six experiments are reported in which participants named pictures that were accompanied by auditorily presented distractors. Distractors that consisted of the first phonemes of the target names caused faster target-naming times than unrelated controls in a wide range of stimulus-onset asynchronies (SOAs), including SOA -300 ms. In addition, two variables that influenced the onset of the phonological effect obtained with word distractors were identified, one of which was the presence or absence of a semantic manipulation in the experiment. The results (a) show that data patterns that had seemed to support strict-serial models of language production (Schriefers, Meyer, & Levelt, 1990) are in fact not univocally interpretable and (b) highlight that the interpretation of (onsets of) context effects cannot be made without clear assumptions about both language production and language perception. © 2000 Academic Press.