Subjects selected by test-anxiety level were presented pairs of words and asked to make judgments of physical identity, acoustic matching, or taxonomic category membership. Experiment 1 varied type of judgment between subjects; in Experiments 2 and 3, each subject made all three decisions. There was no consistent support for the hypothesis that, relative to low-anxiety subjects, high-anxiety subjects would show a more pronounced decrement in reaction times for semantic decisions than for nonsemantic decisions.