The what and when of word retrieval in spoken language production

Semantically driven lexical access in spoken language production is the process of transposing a communicative intention into phono-articulatory codes. At each stage of processing, fromlexical-semantic retrieval through to phono-articulatory processing,more representations are active than are minimally necessary to produce the target word. In recognition of that fact, research over the past three decades has been driven by a (putative) need to answer a core question:How is conflict resolved at the lexical level? The ‘standard view’ involves some form of the idea that selection of the target word is governed by a stochastic process implemented via a biased competitionmechanism, with the nature of the ‘bias’ differing quite widely across proposals.Here we argue that the standard view is out of synchrony with a range of findings; interestingly, and as a matter of historical precedent, a number of those findings pre-date the earliest formulations of the standard view. Stepping back, we argue for a new approach to framing the question of word retrieval: nomechanism is needed to resolve conflict at the lexical level, because there is no conflict at the lexical level. Selection occurs at the semantic/ message level. After selection of the intendedmeaning representation, the next opportunity in the processing pathway to re-evaluate which word to produce is at the level of the pre-articulatory responsemonitor.We refer to this as a ‘ballistic’model of word retrieval to emphasize there is no uncertainty about the identity of the target word at the lexical level, and that retrieval of the correct representation at that level can be sped up, but not slowed down.