One important theoretical question about word production concerns whether the phonemes of a word are retrieved in parallel or in sequential order. To address this question, Meyer and Schriefers (Journal of Experimental Psychology: Learning, Memory, and Cognition 17:1146-1160, 1991 used an auditory picture-word interference task and manipulated the position of the phonemes shared between a distractor word and a target picture. They found that begin-related distractors (e.g., boatbone) facilitated naming times when they were presented within 150 ms before or after the picture, whereas end-related distractors (e.g., cone-bone) were effective only if presented within 150 ms after the picture. This suggested that the word's end phonemes were activated later than the beginning ones. However, it remained unclear whether these effects genuinely reflected facilitation at the level of phonological retrieval. In this study, we examined later distractor presentation onsets, so that the distractors had little opportunity to influence earlier, lexical selection processes. At the latest onset tested, end-related-but not begin-related-distractors significantly facilitated naming. We concluded that late-presented distractors do indeed influence phonological encoding, and that their asymmetric effects support a sequential model of phoneme retrieval.Keywords Word production . Phonological encoding . Phonological retrieval . Word form retrieval . Picture-word interference task . Serial order Current theories propose that word retrieval involves at least two major stages: lexical selection, in which the speaker selects a word that matches the desired concept, and phonological retrieval, in which she or he then retrieves the word's phonological form. This information provides the input to subsequent, articulatory-motor programming processes (see, e.g., Dell, 1986Dell, , 1988Foygel & Dell, 2000;Levelt, Roelofs, & Meyer, 1999;Rapp & Goldrick, 2000;Roelofs, 1997Roelofs, , 2004). Most contemporary models propose that the phonemes of a word are retrieved simultaneously, irrespective of their position. Only at subsequent, articulatory-motor stages do operations become sequential (Levelt et al., 1999;Roelofs, 1997;Roelofs, 2004;Shattuck-Hufnagel, 1992). For example, Roelofs and colleagues proposed that all phonemes of a morpheme are activated in parallel, accompanied by number labels to identify their position (this is called segmental spellout). The phoneme string is then converted into a series of articulatory-motor commands by a process that operates sequentially (called syllabification and prosodification; Levelt et al., 1999;Roelofs, 1997Roelofs, , 2004. However, an alternative possibility is that phoneme retrieval is itself sequential: that is, the word's early phonemes initially receive more activation than later ones, and as retrieval unfolds, later phonemes receive progressively more activation (Dell, Juliano, & Govindjee, 1993;Hartley & Houghton, 1996;Houghton, 1990;Sevald & Dell, 1994;Vousden, Brown, & Harley, 2000; se...
