Segmentation: Learning How to ‘Hear Words’ in the L2 Speech Stream

We ‘hear words’ when we can segment prosodic units from the speech stream and activate associated lexical entries. Segmentation is sometimes regarded in SLA as a perceptual problem, not a grammatical one. I argue here that this view is wrong: segmenting formatives results when we construct prosodic units on the basis of phonetic cues to their edges. The learner's first task is to acquire the relevant cues to these edges. The problem of segmentation is discussed within the framework provided by the Autonomous Induction Theory.