Abstract
In spite of the intuition of many bilinguals, a review of empirical studies indicates that during reading under many circumstances, possible words from different languages temporarily become active. Such evidence for "language non-selective lexical access" is found using stimulus materials of various kinds: interlingual homographs (words that are identical in orthography between languages, such as the English-Dutch word BRAND, meaning "fire" in Dutch), cognates (words that have an orthography and a meaning that are similar or identical across languages, such as TOMATO in English and TOMAAT in Dutch), and interlingual neighbors (words from two languages that differ in only one letter position, such as STEAK and STERK, meaning "strong" in Dutch). However, although there is parallel lexical activation in both languages during bilingual word recognition, the actually observed result patterns also appear to be task-dependent. A distinction must therefore be made between factors affecting the word identification system directly (such as sentence context) and factors affecting the task/decision system (non-linguistic context and task demands). Recent models of bilingual word recognition are discussed with respect to these two types of factors.
