Decomposition of compound words: an MEG measure of early access to constituents

The psychological reality of the morphological complexity difference between compounds (teacup) and single words (crescent) is highly controversial, and the conditions and time course of morphological decomposition remain contested in the psycholinguistic literature (McQueen & Cutler, 1998). Decomposition is argued to occur early or late (Andrews, 1986), in novel but not lexicalized compounds (Van Jaarsveld & Rattink, 1988), and in long but not short compounds (Bertram & Hyona, 2003). This study investigates decomposition in compounds using magnetoencephalography (MEG) in a visual lexical decision paradigm comparing compounds (CW) and single words (SW). The results suggest early decomposition, regardless of lexicalization or constituent length. Response time (RT) is sensitive to both early and late processes, whereas the latency of the MEG component at 300-400ms after word onset (M350) indexes early lexical activation but not post-lexical processing (Embick et al. 2001, Pylkkanen et al. 2002). When the first and second CW constituents have higher log frequency (1.96/1.96 vs. 0.455) and shorter letter-length (3.82/4.0 vs. 7.8) than the SW, early decomposition predicts faster RT and earlier M350 than SW, reflecting constituent over whole-word properties. Late decomposition predicts faster RT but not earlier M350. Since the CW are lexicalized and have short constituents, lexicalization and length constraints predict no RT or M350 differences. Whole-word-only accounts likewise predict no RT or M350 differences.