Previous studies of briefly presented images have demonstrated primary process transformations. Two rebuses (made of pictures suggesting syllables of words) were exposed briefly to thirty subjects; they could not identify the exposed pictures. The subjects were then exposed to lists containing the words. A 'clang' effect was found in their responses to the lists when they were asked for spontaneous associations—there were associations with one rebus and not the other. The authors postulate that the correct identification may serve to protect perception against primary process transformations.