|Zusammenfassung||Bilingual individuals read in their second language (L2) and frequently make decisions based on information perceived in their L2. Compared to monolinguals, visual word recognition in bilinguals bears additional aspects that may also affect later stages of processing. Their integration into neurocognitive models of single word reading and decision making still leaves many open questions. In particular, bilingual word recognition entails the ability to recognize the language membership of a word, and to map letter strings to lexical representations and phonology of the appropriate language, whereby the underlying mapping rules can be different and even contradictory across languages. Furthermore, following these perceptual aspects of word recognition, the perceived information is often used as a basis for higher level, non-perceptual processes, such as decision making under risk. A prominent example for the effects of linguistic framing on decision making is the stronger preference for risky choices when negative outcomes are emphasized than when positive consequences are explicitly stated (the framing effect). The effects of foreign language use on processes influencing the framing effect, most prominently affective processing and cognitive control, has been postulated, but direct evidence for these effects is scarce.
This dissertation's research investigates cognitive and neuronal aspects of bilingual visual word recognition and decision making, using the case of bilingual individuals with native language German (L1) and second language English (L2). Study 1 isolated the effects of word and pseudoword length (i.e. number of letters) on the dynamics of sublexical encoding and lexical access in German and English, using a diffusion model of reaction times (RTs) in lexical decisions. Study 2 investigated whether language membership decisions and naming in a language-ambiguous context are influenced by continuous sublexical (bigram frequencies) and lexical (orthographic neighborhood size) similarity of a letter string to German and English words. This was extended to an investigation of the neuronal correlates of language similarity statistics and language membership representations in Study 3, using functional magnetic resonance imaging (fMRI). Study 4 investigated whether a reduced framing effect during foreign language use is due to reduced affect or increased cognitive control.
Study 1 showed that sublexical encoding of words and pseudowords during lexical decision takes longer for longer stimuli, whereas lexical access is accelerated for words but slowed for pseudowords. Sublexical length effects in English were larger than previously found in native speakers of English, whereas the effects of length on lexical access in English were similar to those previously reported for English. These findings suggest that sublexical but not lexical processing remains tuned to the orthographic structure of the L1.
In Studies 2 and 3, graded sublexical and lexical language similarities to each of the two languages biased language membership decisions. Moreover, sublexical processing and phonological encoding were faster for letter strings with an L1-typical sublexical structure. fMRI data of Study 3 revealed that brain activity in left ventral occipito-temporal cortex, previously associated with processing of letter strings, was positively correlated with sublexical similarity L1. The level of ambiguity in lexical similarity to L1 and L2 determined activation in bilateral angular gyri, providing direct evidence for language-unselective lexical access in a region previously associated with lexical access in the L1. Brain activations in the left supramarginal gyrus and temporo-parietal junction, previously associated with phonological encoding, were positively correlated with lexical similarity to L1. This finding suggests that phonological representations of the L1 but not L2 are automatically activated during language decisions on visual input. Finally, multivariate fMRI analyses revealed that language membership information is contained in distributed activation patterns throughout the visual word processing network, as well as in additional right parietal areas.
Study 4 showed that in a consistent language setting framing effects are comparable in native and foreign languages and at all foreign language proficiency levels. However, it also revealed that the framing effect is reduced following a language switch, not only into a foreign language but also into the L1. This finding is best interpreted in terms of increased cognitive control following language switching, rather than as altered affective processing during foreign language use, as was claimed in previous studies.
Based on these findings I propose to extend neurocognitive models of bilingual visual word recognition in several aspects. First, sublexical mapping of orthography should prioritize the orthographic structure of the L1. Second, the summed activation propagated from sublexical representations to the lexicon should increase with stimulus length, which would lead to faster lexical access for longer words. Third, lexical representations should be clustered by language, to constitute a basis for language membership information within the core visual word processing network. Fourth, input to language membership representations should stem from all sublexical and lexical representations, weighted by their typicality for each language. Finally, our findings imply that the input language affects non-linguistic decision processes through a modulation of cognitive control levels and not affective processing. This link needs to be included in integrated models of perceptual linguistic and higher-level processing.