Thirty-six subjects in a cold pressor experiment were asked to periodically rate the unpleasantness of the experience and estimate the water temperature, under three attention conditions. Subjects were given (1) instructions to monitor their sensations, (2) instructions to monitor their emotions, or (3) no monitoring instructions. With their hands immersed in the cold water, subjects had to learn and later recognize a list of words related in varying degrees to either the sensory or the emotional features of coldness. As predicted, subjects who monitored their sensations rated the experience as less unpleasant, but their estimates of the water temperature did not differ from the other two conditions. In addition, these subjects had a higher rate of "hits" and "false alarms" for words associated with the sensory features of coldness. These results are seen as providing tentative support for the parallel-processing model of pain perception and specifically, the hypothesis that pain reduction can result from a modification of a pain schema.