Exposure to an accented production of a particular phoneme in word contexts induces a shift in listeners' representations of the inclusiveness of that phoneme category. In a lexical decision experiment, the same ambiguous phoneme (between /f/ and /s/) replaced /f/ in 20 words ending with /f/ (e.g. carafe) for some listeners, while for others it replaced /s/ in 20 words ending with /s/. A subsequent phonetic categorisation experiment showed that the /f/ category had become more inclusive for the former group, while the /s/ category became more inclusive for the latter group (Norris, McQueen & Cutler, 2003). Importantly, exposure to the same ambiguous sound in a nonword context had no effect on category boundaries. The observed plasticity could not be accounted for by adaptation or contrast effects; Norris et al. argued that the plasticity occurred in the service of word recognition. Adjusting category boundaries allowed more rapid recognition of an unusual speaker's speech.
Adjustment would be useless if it did not generalise to other words. In further research the same exposure conditions were used, but the phoneme categorisation test phase was replaced by a test phase involving another lexical task, capitalising on cross-modal identity priming effects: recognition of a written word is faster if the same word has just been heard. The critical words in this case were /f/-/s/ minimal pairs (e.g. knifenice). The spoken form ended with the ambiguous sound, and at issue was how much priming this form produced for recognition of the two written words. For listeners with initial-phase exposure to the ambiguous sound replacing /f/, more priming resulted for words ending with /f/ (knife), while for listeners with /s/- exposure more priming resulted for /s/-words. Thus the learning generalised from the 20 words used in the first phase to the rest of the lexicon, as predicted.
This suggests that the utility of phoneme-category plasticity is indeed facilitation of word recognition. An open question is how stable listeners' lexical representations need to be for such rapid adjustment of category boundaries to occur. For instance, does less lexical stability imply less plasticity in phonemic representations? Less lexical stability might be expected, for example, in young children's lexical representations, in L2 learners' lexical representations, or in the representations for low-frequency words in the adult lexicon. In each case we might predict less induced adjustment of category boundaries.