Explaining how learners extract ‘formulae’ from L2 input

Abstract

Studies of L2 production have shown that both children and adult learners make use of ‘formulae’, putatively ‘unanalysed’ sequences of words. In this paper I discuss how formulae may arise in L2 acquisition by processes of segmentation. Carroll and MacDonald (Ms. 2009), Carroll et al. (2009) show that even ab initio learners can rapidly segment sound forms from continuous strings. The data are consistent with two approaches to the segmentation of words: words are segmented by tracking co-occurrence statistics over adjacent syllables (transitional probabilities or TPs); the left edges of words are placed just before a strong syllable (a Metrical Segmentation Strategy). In my contribution to this special issue, I address the question of how strings of syllables can be re-analysed as morpho-syntactic categories, their phrasal projections and dependencies. I do this in terms of the Autonomous Induction Theory (Carroll 2001) discussing formulae in particular in terms of correspondences across autonomous and modular representational systems: prosodic, morpho-syntactic, and conceptual.