Sixteen Ss were trained as to the affective signal value of two symbols. One symbol served as a signal for taboo words, and the other as a signal for neutral words. The results showed that when words in the test phase were preceded by the taboo signal a higher recognition threshold was obtained than when words were preceded by a neutral signal. This effect was found to be independent of the affect of the word that followed the signal. The results were interpreted as casting doubt on the possibility that data supporting perceptual defense can be interpreted on the basis of some form of set disruption.