In the first experiment I ever conducted, I gave people a placebo and told some of the subjects it would cause heart palpitations, rapid breathing, and sweaty palms. These are the symptoms people experience when they're undergoing strong emotion. I then gave subjects a series of steadily increasing electric shocks, with instructions to tell me when the shocks became too painful to bear. I anticipated that subjects who were told the pill would cause arousal would mistakenly attribute their shock-produced arousal to the pill. They would consequently find the shock less aversive and would be willing to take more of it than control subjects who could only assume their arousal was being produced by the shock. And that was indeed the finding. After removing the electrodes I asked the subjects in the arousal-instruction condition who had taken a great deal of shock why they had taken so much. A typical answer would be, “Well, I used to build radios, and I got a lot of shocks so I guess I got used to it.” I might then say, “Well, I can see why that might be. I wonder if it occurred to you that the pill was causing you to be physiologically aroused.” “Nope, didn't think about the pills and didn't think about the arousal.” I would then tell them what the hypothesis was. They would nod politely and say they were sure that would work for a lot of people. “But see, I used to make radios … ”

It was perfectly clear that subjects had no idea of what had gone on in their heads. At the time, believe it or not, this claim seemed to most people to be quite radical. There was a bedrock presumption that thought is basically linguistic. To show that this assumption was mistaken, and that quite elaborate cognitive processes can go on without people's awareness of them, I began to do experiments with Tim Wilson in which we would manipulate some aspect of the environment that would affect subjects’ behavior in some way. For example, we might have people examine an array of nightgowns and tell us which they preferred.