Cognate Effects in a Sentence Context

Becoming a bilingual can change a person’s cognitive functioning and language processing in a number of ways. This study focused on how knowledge of a second language influences how people read sentences written in their native language. We used the cognate-facilitation effect as a marker of cross-lingual activations in both languages. Cognates (e.g., Dutch-English schip [ship]) and controls were presented in a sentence context, and eye movements were monitored. Results showed faster reading times for cognates than for controls. Thus, this study shows that one of people’s most automated skills, reading in one’s native language, is changed by the knowledge of a second language. Research in cognitive psychology has shown that becoming a bilingual can change one’s cognitive system in several ways, even beyond the language domain. For instance, Bialystok and her colleagues have shown that bilinguals are more efficient in tasks that tap into cognitive control (e.g., Bialystok, Craik, & Ryan, 2006). Most research, however, has focused on how bilingualism influences language processing in general. For instance, Ameel, Storms, Malt, and Sloman (2005) showed that linguistic category boundaries in each language can move toward one another in bilinguals, at least for the concrete objects used in their study. Also, it has been observed that bilinguals are slower than monolinguals at naming pictures in their first language (Gollan, Montoya, Cera, & Sandoval, 2008; Ivanova & Costa, 2008). However, picture naming is a relatively controlled task and not as highly automated as native-language reading. As Reichle, Pollatsek, Fisher, and Rayner (1998, p. 125) argue, reading is ‘‘the most important and ubiquitous skill that people acquire for which they were not biologically programmed.’’ Native English speakers read with an impressive speed of three to five fixations per second, and saccades go forward about five to nine character positions in the sentence (Reichle et al., 1998). We investigated whether knowledge of a second language can influence this highly automated skill of reading in one’s native language. Are bilinguals able to restrict lexical access to representations in the (native) language of the text, or is their other (nonnative) language activated strongly enough to influence reading? Studies on isolated (out-of-context) word recognition demonstrated interactions between a bilingual’s two languages (e.g., Dijkstra, Grainger, & Van Heuven, 1999; Duyck, 2005; Jared & Kroll, 2001; Van Heuven, Dijkstra, & Grainger, 1998). For instance, bilinguals are faster in reading cognates (translation equivalents with full or partial form overlap, e.g., DutchEnglish: sport-sport; Dutch-German: dier-Tier) than control words (Dijkstra et al., 1999). This cognate facilitation effect, observed in second-language reading (e.g., Dijkstra et al., 1999) and in native-language reading (Van Hell & Dijkstra, 2002), has typically been explained by assuming language nonselective activation in the two languages. The presentation of a word in one language activates orthographic, phonological, and semantic representations of all known languages. The cross-lingual activation spreading from these three codes speeds up the activation of cognates compared to control words and results in faster word-recognition times. However, a strong theory of language nonselective lexical access requires a more stringent and ecologically valid test. Obviously, people rarely read words presented in isolation. Instead, words are usually encountered in meaningful sentences. The fact that people read a coherent set of words in one language may influence lexical access and the degree of cross-lingual activations in the two languages. Using the language of the sentence as a cue to guide lexical access for upcoming words would indeed be a very efficient strategy to speed up word Address correspondence to Eva Van Assche, Department of Experimental Psychology, Ghent University, Henri Dunantlaan 2, B-9000 Ghent, Belgium, e-mail: [email protected]. PSYCHOLOGICAL SCIENCE Volume 20—Number 8 923 Copyright r 2009 Association for Psychological Science recognition, because this would limit lexical search to lexical entries of only one language. We addressed this issue by investigating cross-lingual activations in native-language sentence reading by bilinguals. This provides a very conservative test of a profoundly nonselective language system, because we tested for an influence of the weaker second language, learned in adolescence, on nativelanguage sentence reading, which is a highly automated skill. Although this situation has never been explored, a few studies have investigated the reverse situation, namely native-language influences on nonnative-language sentence reading. These studies show that lexical access during second-language sentence reading does not seem to operate in a language-selective way (Altarriba, Kroll, Sholl, & Rayner, 1996; Duyck, Van Assche, Drieghe, & Hartsuiker, 2007; Schwartz & Kroll, 2006; Van Hell & de Groot, 2008), although sentence context seems to constrain second-language reading in bilinguals somewhat (depending on sentence constraint and degree of orthographic overlap between cognates). This may not be very surprising because it is likely to be extremely difficult for unbalanced bilinguals, like the majority of bilinguals who participated in these studies, to ‘‘turn off’’ their native and dominant language. Indeed, many word-recognition studies reported much stronger influences of the native language on second-language processing than of the second language on native-language processing (e.g., Dijkstra & Van Heuven, 2002; Duyck, 2005). Because influences of the native language are stronger than influences of the second language, testing bilinguals reading in their native language is a good way to demonstrate the existence of a lexical system that is profoundly not language-selective. Therefore, we took this challenging approach in the study described here. In a pretest, we replicated the native-language cognate-facilitation effect for words presented in isolation (Van Hell & Dijkstra, 2002) with a set of 40 cognates and control words that were matched on word class (all words were nouns), word length (identical), number of syllables, word frequency, neighborhood size (Coltheart, Davelaar, Jonasson, & Besner, 1977), and bigram frequency. Forty-two Dutch-English bilinguals performed a Dutch (native-language) lexical decision task (word/nonword decision) on these words, Dutch filler words, and nonwords. Linear mixed-effects model analyses in which frequency was included as a control variable showed significantly faster reaction times for cognates (M 5 493 ms) than for controls (M 5 507 ms), F(1, 3067) 5 7.70, p < .01. We also used a continuous measure of cognate status by defining cross-lingual overlap between each word and its translation (e.g., piloot-pilot: .95; schaap-sheep: .52; eend-duck: .08) using the word similarity metric developed by Van Orden (1987). This analysis showed a gradual effect of cross-lingual overlap on word processing: Recognition of Dutch words was facilitated when words had higher degrees of orthographic similarity with English, F(1, 3067) 5 4.45, p < .05. This pretest demonstrated that secondlanguage lexical representations become active when bilinguals read native-language words in isolation, and constitutes a validation of these materials for the actual sentence study. We investigated whether knowledge of a second language may still influence lexical access in a native-language sentence context, even though the language of the sentence provides a highly efficient cue for lexical search. We presented the exact same cognates and controls in a native-language sentence context, and monitored participants’ eye movements. This methodology, which taps into early stages of word recognition, is a very good test of naturalistic reading because it does not require an experimental task with a decision component (e.g., lexical decision).