Lexical parsing strategies in two languages: Constraints on language selection in word recognition

Cross-linguistic research demonstrates that languages differing in spelling-to-sound consistency use distinct lexical parsing strategies. These findings tend to support models like the Dual Route Cascaded Model (Coltheart, Rastle, Perry, Langdon, & Ziegler, 2001) in which lexical access occurs via two routes, one in which phonology is directly retrieved in parallel, and a second in which phonology is computed serially. This model predicts that consistent languages rely more on the computational route, whereas less consistent languages rely more on the direct route. If lexical parsing preferences require that languages differing in consistency engage distinct processing, for bilinguals, two languages that differ in consistency should function independently. However, bilingual word recognition research demonstrates that both languages are active and competing, even when they differ in consistency, a finding captured by the Bilingual Interactive Activation model (Dijkstra & van Heuven, 1998). This creates a paradox for word recognition in that cross-language interactions occur, but for lexical parsing, both languages appear to function independently. To examine this paradox, three experiments were performed using tasks shown to reveal parsing preferences. In each experiment, the performance of native English speakers and proficient non-native speakers of English was compared in English. In Experiment 1, participants made lexical decisions to stimuli including pseudohomophones, to compare native and non-native spelling-to-sound computation. In Experiment 2, participants performed clustered lexical decision in which word and nonword stimuli were parsed after the onset or the first vowel to compare the importance of the rime unit in processing. Finally in Experiment 3 participants named words and nonwords including English-German cognates to examine how second language processing strategies differ for German-English bilinguals in the presence of strong first language cues.; In Experiments 1 and 2, results were similar for native English and native German speakers. In Experiment 1, native Japanese speakers also produced similar results in English, although their performance was more dependent on English proficiency. However when native German speakers named English words in Experiment 3, both first and second language processing strategies were revealed. Implications for current models of word recognition are discussed, and a bilingual Dual Route model is proposed.