Encoding of Inferences into Long-Term Memory in Bilinguals

When reading and listening to information, people often make inferences to facilitate their comprehension and the creation of a discourse representation. However, because such inferences are not explicitly stated, they can lead to distortions in the memory representation known as inference-based memory errors. The present study investigated the encoding of inferences in bilinguals while considering the effects of language proficiency, presentation modality, and working memory load. In Experiment 1, Spanish-English bilingual participants read and listened to narratives composed from common schemas (e.g., going to eat at a restaurant) before completing a recognition test on the details from narratives. In Experiment 2, Spanish-English bilinguals listened to narratives while under an additional load on memory or under no load before completing a recognition test. The critical measure was how well participants could discriminate presented schema-relevant details from inferred (non-presented) schema-relevant details. Experiment 1 showed that discrimination of presented information from non-presented inferences was better when the narratives were read versus listened to. Experiment 2 showed that imposing an additional load on working memory decreased the discrimination of presented items from non-presented inferences. Furthermore, because no effects of language proficiency emerged, the results of both experiments suggest that inference-based memory errors occur at a level of representation that is conceptual and not language-dependent. Future directions for research and implications for the legal system are also discussed.