Beyond one model per phenomenon.

study, taking truly independent ratings to see which verbs participants considered to be denominal or deverbal (i.e. exocentric or not). The results of this new experiment were clear. The semantic similarity (or otherwise) of verbs-incontext to ordinary irregular usage predicted the acceptability of past-tense forms. Whether or not verbs were perceived to be exocentric was irrelevant to this. It is worth noting that these results were obtained using the verbs put forward by Kim et al. [5] to try to demonstrate their theory. Thetheorythathomophoneverbprocessingcanbeexplained by exocentricity fails completely in the face of examples like shoe–shoo, where it is the denominal verb that is irregular (shoe–shod) and the deverbal verb regular (shoo–shooed). (Further support for these findings comes from on-line studies that haveshown that semantics – and not exocentric form – predictcomprehensiontimeforhomophoneverbs[7].) What is the relevance of these data to the past-tense debate? First, they demonstrate that the ‘in-principle’ objection to single-route models supposedly posed by homophone verbs is wrong. Inflections of homophone verbs are determined by complex semantic and phonological patterns, not grammatical status. Second, they show that children’s ability to process homophone verbs is not indicative of their innate sensitivity to nouns and verbs [8], but rather is indicative of children’s ability to learn languages. Finally, because the exocentricity thesis is false, the dual-route theory itself cannot account for the processing of homophone verbs. Providing an account of homophone-verb inflection is vital to explaining past-tense processing [1–4]; the evidence suggests that it is two routes that this phenomenon rules out, not one.