When is a Direct Test of Memory More Sensitive Than an Indirect Test?

A direct test of memory such as recognition memory is typically considered to be less sensitive than an indirect test such as lexical decision (e.g., Schacter, 1987). However in previous research we found that manipulations which eliminated semantic priming of lexical decision did not eliminate semantic priming of recognition memory. In the present experiment we replicated and extended this result by including an additional indirect memory test - semantic categorization. Subjects initially performed a lexical decision task, and were then given two memory tests for the words which had appeared in the lexical decision task: a direct recognition memory test and an indirect semantic categorization test. In lexical decision, performing letter search on a prime eliminated semantic priming of related targets. However, semantic priming was preserved in recognition memory: Primes were better remembered if they had been followed by related as compared to unrelated targets during lexical decision. In contrast, semantic categorization performance was unaffected by whether the lexical decision items had initially appeared in a related or an unrelated context. The dissociation suggests that different processes underlie priming in lexical decision and in recognition memory. The former is best explained in terms of spreading activation, whereas the latter is best explained in terms of integrative processing involved in the production of an episodic trace.