The importance of articulatory and phonetic training in the production of L2

The PAM model was later reviewed by Best and Tyler, this time taking into account the perception of second language sounds by more experienced L2 learners. Best and Tyler assume that a common L1-L2 system will be created by the speaker, and that this system, combining both the phonological and the phonetic level, will be flexible to change. This is due to the idea that L1 sound system learning mechanisms and processes can adapt fairly easily to changes in the language environment by perceptual learning of new higher-order invariants over age. L2 learning is then seen as a functional extension of this process. Best and Tyler (2007) move on to present the possible scenarios of how L1 and L2 sounds relate at the phonological level. From this perspective, they predict the likelihood of success of perceptually learning an L2 phonetic category in each case. The main issue is then whether or not the learner has perceived equivalence between an L2 and an L1 phonological category, that is, has perceptually assimilated the L2 phone to this L1 phonological entity. This implies that an L2 phone could be assimilated to an L1 phone if the phonological category has a similar contrastive relationship to surrounding categories in the phonological space. Best and Tyler present the example of French uvular /r/: although this phone is perceptually assimilated to English liquid /r/, the two phones do not resemble each other on a phonetic level (2007: 24-25).

Best and Tyler then illustrate the possible assessment scenarios looking at how L2 minimal contrasts are perceived by L2 learners. These scenarios are extensions of possible 1995 PAM non-native speech perception scenarios to L2 learning: i) Single Category assimilation, in which there is no acquisition of a new category, unless listeners can perceive phonetic differences between the native phonemes and, at least, one of the L2 phones; ii) Category Goodness assimilation: in which there is a possibility of acquisition of a new phonological category only for the more deviant phone; iii) Two Category assimilation, in which there is no acquisition of new categories, if the L2 phones are perceived as good exemplars of L1 categories (as a shift to new categories in similar cases is not probable); iv) Uncategorised-Categorised assimilation, in which a new category is not likely to be learned, if the phone is perceived as a good instance of L1; on the other hand, the acquisition of the ‘uncategorized’ phone depends on both phonetic and phonological characteristics of the set of native phonemes with which the L2 phone is assimilated; v) Uncategorised-Uncategorised, in which depending on the similarity of the two phones to the same or to different sets of L1 native phonemes one category can be created, as both phones may “converge into a single new but undifferentiated phonetic and phonological category”, but if they are different, and thus they are distant in the L1 system, two categories could be created; vi) Non-Assimilable, in which sounds may either be incorporated in the phonological space as uncategorized or be always considered non-linguistic.
Nevertheless, despite the adjustment made to adapt the model to L2 learning, it still presents some critical aspects, such as the lack of criteria to establish how non-native contrasts are assimilated to native categories and of explications on how L1 phonological system would change itself through the exposition to non-native contrasts.