Abstract: One of the most important issues in the study of bilingual language processing is cross-language interference. From the perspective of development-oriented researchers, less proficient bilinguals lack the cognitive control to suppress activation of non-target language (L1) while highly-proficient bilinguals have. Contradictory to this proficiency-dependent assumption, it is found that certain context could help less proficient bilinguals to constrain the accessibility of non-target language by means of top-down processing mechanism, though the constraint does not spread to lexical level. According to Duyck et al. (2007), the lexical level activation could be attributed to inadequate contexts. While previous research focused on the lexical access, the primary objective of the present study is to determine whether the effect of context would also apply to semantic access of sentences. We assume that less proficient bilinguals could as well suppress the activation of words and sentences of non-target language if given appropriate context.
176 college students (44 in each of the 4 experiments) in their junior year, none majored in English, were asked to report number of years they had spent in learning English and their proficiency in English. The results showed that their proficiency in second language has no significant difference compared to subjects of other studies. Tasks were programmed with and displayed by E-prime 2.0. The basic procedure involved 1) a sentence comprehension task (e.g., The man is chasing a thief.) following a context (or no context); 2) a decision task in which a picture accompanied with a L1 verb was presented and participants were asked to judge whether the verb had the same meaning with that conveyed by the picture as quickly and accurately as possible. Different contexts characterized the four experiments. Reaction time and accuracy were measured automatically by computers and analyzed with SPSS 13.0.
Participants responded faster and more accurately in the sentence-verb consistent condition compared to the non-consistent condition—at least this is the case in the first three experiments: 1) when there was no context at all, t1(37)=3.74, p<0.001, t2(38)=4.08, p<0.001; 2) when a 6-minute-long animated cartoon bearing no relationship with the tasks was provided, t1(38)=3.15, p=0.002, t2(38)=4.81, p<0.001; 3) when a short passage with corresponding speech semantically related with the following sentences is given as a context, t1(40)=4.09, p<0.001, t2(38)=5.76, p<0.001. However, in experiment 4), the context being an animated cartoon that also predicted the meaning of the following sentences, this previous bias was eliminated, i.e., no significant difference in the performance of the two conditions was observed. t1(38)=0.22, p=0.828, t2(38)=0.45, p=0.659.
The current study provided the effect of context with some new evidence. The first experiment confirmed non-selective access of sentences across bilinguals’ two languages. Specifically, the comprehension of semantically-related sentence activated verbs of the non-target language and thereby facilitated the picture-verb judgment task. This non-selectivity persisted in experiment 2 and 3 irrespective of the given contexts, but disappeared in the final experiment, suggesting that appropriate context, such that combine both image and meaning, allow for selectivity of target-language access. Generally, the more consistent (with the tasks) and vivid the context is, the easier it is for less proficient bilinguals to inhibit activation of non-target language; because such context leads to the enhancement of relevant components of the target-language thus increases the recognition ability in the working memory of less proficient bilinguals. What we have found may provide some insight into the theoretical explanation and future research direction of second language learning.

Key words: bilingual, semantic access, proficiency, context, cognitive control