Simultaneous segmentation and generalisation from nonadjacent dependencies

Rebecca, F. & Padraic, M.

Lancaster University, UK

One of the primary challenges facing language learners is identifying words in speech, and learning the grammatical rules that apply to these words. Past research has provided conflicting accounts regarding the separability of these tasks. One possibility is that words are first identified and then relations between words are computed (Peña, Bonatti, Nespor & Mehler, 2002), meaning different processes may apply to these tasks. Alternatively, it may be that these tasks are resolved in combination during language learning, and that instead of distinct operations, the same type of learning process may apply to both word identification and grammar acquisition. We examined adults? ability to acquire specific words and generalise structure from non-adjacent dependencies in continuous speech, to establish whether segmentation and generalisation may occur in tandem. Contrary to prior suggestions (Peña et al., 2002), using a new test of generalisation, we found that segmentation and generalisation can occur simultaneously from the same signal. Interestingly however, in a separate study, we found that sleep differentially affects performance on these tasks, with an enduring benefit of sleep for segmentation, but only a short-term benefit of sleep for generalisation, highlighting the possibility for sleep-driven computational differences.