Title: Humpty Dumpty's Problem: putting complex words back together again

Over the past 15 years, considerable evidence from a range of different languages and methodologies has converged to provide clear evidence that the early stages of visual word recognition involve a mechanism of form-based morphological parsing, which operates across all potentially morphologically complex words, regardless of formal or semantic opacity (Rastle and Davis 2008, Lewis et al 2011, Royle et al 2012, Fruchter et al 2014, inter alia). Comparatively little attention, however, has been focused on how linguistic processing proceeds once morphological constituents have been identified.

In this talk I'll discuss the results of a number of recent and ongoing experiments using MEG to investigate how we rapidly access information about the constituents of morphologically complex words including two (saw, jumped, outrun, thinkable) or three (outran, unthinkable, reflatten) morphemes, and how we make use of this information to reassemble the pieces and evaluate their syntactic and semantic wellformedness. I'll focus much of the talk on 'fresh from the lab' data from a project with Alec Marantz (NYU) and Christina Manouilidou (UPatras) that we are just now analysing, in which we are investigating the neural spatio-temporal dynamics of access to the lexical category vs. argument structure representations of verbal stems. I'll argue that by focusing on the apparently simple question of how we detect and make use of information about morphological constituents, we can gain significant insight into the overall architecture of the human linguistic system.