Learning the meanings of words from contexts and definitions: ERP Evidence

Abstract

Purpose: We conducted an ERP experiment on the learning of word meanings by context and definition aimed at two related hypotheses: (1) A word is better integrated into the meaning of a sentence when it is learned through sentence contexts rather than definitions. (2) Integration is also affected by variation in sentence contexts. Method: Adults learned the meaning of unfamiliar words by reading sentences that contained the word or a definition. Sentence contexts were repeated or varied over 4 occurrences. Following learning, participants read sentences that contained a key word, either one they had learned or a matched control word, and judged whether the sentence made sense. Results: The N400, an index of word-to context integration, was reduced when a learned word made sense. However, this reduction was greatest for words that had been experienced in context, and reliably smaller for words experienced in definitions. The N400 was also reduced more following 4 varied sentence contexts than a single repeated context. Conclusions: The results provide clear evidence for both hypotheses. The advantage for learning word meaning from contexts is explained by the convergence of meaning features across various sentences, which allows stronger overlap of episodic memory traces. A similar explanation holds for the positive effect of context variability.