New limits to automaticity: Context modulates semantic priming.

It is widely assumed that semantic priming in visual word recognition is automatic when the task requires word-level analysis. The present experiments show that this conclusion is too strong. Whether brief-duration primes facilitated the processing of related targets in lexical decision depended on the context in which the primes were seen. Semantic priming occurred if subjects saw only brief primes (blocked condition) but was minimal if longer primes were presented as well (mixed condition). Converging operations indicate that this modulation of semantic priming reflects operations beyond the lexical level rather than early encoding deficits. Rather than being automatic, semantic priming depends on the context in which a word is read.