2008 Context repetition benefits are dependent on context redundancy

What sources of statistical information do humans leverage to organize the mental lexicon? Recently, accounts emphasizing the importance of word frequency have been challenged by accounts emphasizing contextual diversity (CD). The latter suggest that words will be processed faster if they occur in a greater number of unique semantic contexts. Previous corpus studies have operationalized CD by counting the number of documents that each word appears in. However, there is no guarantee that each document corresponds to a unique semantic context, as documents may address similar topics. We develop a measure of CD that takes into account the semantic similarity of documents, and show that for words appearing in many contexts, appearing in multiple contexts is most beneficial if the contexts are high in semantic distinctiveness. In Experiment 2, we induce a similar effect experimentally using an artificial-language paradigm, demonstrating that repetition of a word does not speed processing unless the repetition is accompanied by a change in context