Subjects were presented with a list of pictures and words and performed tasks that oriented processing toward the concept as an image, the concept as a verbal item, or toward underlying referential information associated with the concept. Recongnition (Experiment 1) for concepts presented as pictures was superior only when the task required subjects to orient to the concept as a verbal item per se`a word superiority effect was observed when orientation was to the concept as per se. The mode of presentation did not influence concept recongnition when subjects had focused on the referential meaning of the items. Memory for form was also influenced by the task, with subjects more likely to claim they saw a word as a picture than vice versa, but only in the referent tasks. When subjects were asked to recall, rather than recognize the concepts (Experiment 2), there was a similar, though not identical,pattern of results. The sensory-semantic model of D.L Nelson, V.S. Reed, and C.L.McEvoy (Journal of Experimental psychology : Human learning and memory,1977,3,485-497) had difficulty with some aspects of these data. The availability of information regarding cognitive operations performed on the input seems to be an important component of memory traces.