Stockall, L. 1, 2
1 School of Languages, Linguistics and Film
2 Queen Mary, University of London
Two basic questions have dominated investigations on the processing of morphology for the past 30+ years: 1. are transparently morphologically complex words like 'teacher' and 'jumped' decomposed into their constituent morphemes at an early, automatic, form based stage of processing? 2. are irregular allomorphs like 'sang' and 'taught' processed using the same mechanisms as regulars, or by a distinct system? The first question has been conclusively answered by a string of experiments beginning with Longtin, Segui, and Hallé (2003) using a masked priming paradigm and comparing regular stemallomorph pairs with pseudo related pairs like 'brother-broth'. More than 20 experiments using this paradigm have now been reported (see Rastle & Davis 2008 for a partial review), and overwhelmingly the results support an early, automatic form based stage of morpho-orthographic decomposition for all strings that can plausibly be parsed into a stem and affix. This evidence from behavioral masked priming is further supported by a growing number of experiments using MEG to investigate single word reading. Zweig & Pylkkänen (2009), Lewis, Solomyak & Marantz (in press), and Lewis & Marantz (under review) provide evidence for an evoked response peaking ~150ms after the onset of a visually presented word that indexes the morpho-orthographic decomposition stage that the masked priming data argues for. Crepaldi et al (2010), using behavioral masked priming, show that stem targets are rapidly primed by their irregularly inflected allomorphs, consistent with an early, automatic, form based parsing mechanism for irregulars too. I'll discuss a series of experiments investigating the processing of regular and irregular allomorphs using EEG (exp1) and MEG (exp2&3) and comparing masked priming (exp1&2) to single word reading (exp3). These experiments reveal that regular and irregular allomorphy are associated with equivalent early evoked responses in the 150-250ms time range (with interesting differences between EEG and MEG, and between priming and single word reading), thereby arguing for a single mechanism model for morphological parsing.