Whiting, C. M. 1, 2 , Shtyrov, Y. 2 , Hauk, O. 2 & Marslen-Wilson, W. D. 1, 2
1 Department of Experimental Psychology. University of Cambridge. Cambridge, UK.
2 MRC Cognition and Brain Sciences Unit. Cambridge, UK.
Extensive behavioural evidence points to a process of automatic segmentation for any visual form that is potentially morphologically complex (Rastle, Davis & New, 2004; Longtin & Meunier, 2005). It is argued that the presence of morphological structure drives this process since both a stem and an affix are necessary to trigger segmentation, regardless of word meaning. Our goal was to delineate the role of morphological structure in modulating the processing of morphologically simple and complex words in English using magnetoencephalography (MEG). Test words contrasted the presence of a potential stem and affix and the semantic relationship between the embedded stem and whole form (farm-er, corn-er, scan-dal). A single-trial approach using multiple linear regression analysis was employed (Hauk et al., 2006), which can take advantage of spatiotemporal information from individual words. Six orthogonal variables associated with morphological processing and visual word recognition were submitted to linear regression analysis at each sensor at every millisecond: word length/N size, bigram/trigram frequency, word frequency, stem frequency, affix frequency and semantic relatedness. A hierarchical processing stream emerged moving from bilateral occipitotemporal cortex to left anterior temporal cortex: early sensitivity to length/N size and orthographic structure at 100 ms was followed by effects of stem and affix frequency between 220 and 300 ms, with word frequency and semantic transparency effects appearing after 400 ms. The results support claims for early morphological processing based on the presence of orthographic cues to morphological structure, and suggest that the application of multiple linear regression to MEG along with standard factorial analyses can aid in further elucidating the linguistic variables that modulate neural activity across space and time.