Through the Back Door to Theory

One of my all time favorite psychology articles is Nisbett and Wilson's (1977) Psychologicial Review article entitled Telling More Than We Can Know: Verbal Reports on Mental Processes. Their point was that we enjoy very little direct access to our cognitive processes, at least not enough to directly know the real reasons that we think and do the things we think and do. One study they reported offered a memorable example. Participants entered a room with elastic ropes hanging from the ceiling, each one with a weight attached to the end. If the participant grabbed the end of one rope and walked out as far as he or she could go, he or she could not reach any of the other ropes. Yet the participants' task was to get a hold of two ropes and tie them together. After participants had struggled with this puzzle for a while, an experimenter would walk through the room and bump one of the ropes causing it to swing. Within seconds, virtually all participants solved the puzzle. They would push one rope causing it to swing and then leave that rope to grab another rope, which they then pulled out as far as it would go toward the swinging rope. Catching the swinging rope as it swung toward them, they would then tie the two ropes together. What interested Nisbett and Wilson (1977) were participants' descriptions of how they solved the puzzle. Almost all of them missed the critical cue of the experimenter causing a rope to swing. They offered other reasons: that they were good problem solvers in general; that as students of psychology, they were given to flash insights; etc. Not only were their answers most often wrong, they were also a bit self-enhancing. It is tremendously flattering to be asked how you got an idea and developed it into an experiment-as the editors of Psychological Inquiry have done in this issue. However, as one tries to answer that question, the Nisbett and Wilson (1977) research stands as a chastening lesson. One could easily do what the participants in the aforementioned experiment did: construct a sensible, slightly flattering story that fits the available evidence but that has little to do with the real causes. So there is my caveat. Whatever I say henceforth has to be regarded as some kind of construction from available evidence and from a memory that is itself constructed. We are talking detective work here, not direct knowledge. There is a roughly verifiable record of events to work with, and two collaborators without whom I know this work never would have happened-Joshua Aronson, my coauthor on Steele and Aronson (1995), the article in question, and Steven Spencer, who codeveloped earlier stereotype threat research with me that was indispensable to the Steele and Aronson research. The first point about research ideas that I would stress is that in the period leading up to the Steele and Aronson (1995) research, we were not trying to have a research idea. Perhaps because I had just changed jobs, I remember feeling less pressure than usual to move into a program of research. Don't get me wrong. I was doing conventional research on self-affirmation processes. However, perhaps because that work was going on, I felt freer to think more generally about other things. And into that opening came a compelling and practical problem that soon captured most of my attention: the academic underperformance of minority students at the University of Michigan. I discovered this problem while serving on a faculty committee for minority student recruitment and retention. I saw a chart that showed very clearly what has now become recognized as a national phenomenon: that at each level of SATs that students had when they entered Michigan, Black students got lower subsequent grades than other students. I had always thought that any racial gap in grades was due to differences in how well the groups were prepared for college work-reflecting racial differences in, for example, access to education. However, these data showed different: that a racial gap in grades persisted even among students who were equally well prepared for college work, among students who had the same entering SAT scores. Clearly, beyond poor preparation was depressing the grades of Black students. So that was the problem that I and my students began thinking about. It was self-evidently important. However, what made it especially interesting was that the something depressing Black students' grades seemed like it might be social psychological. When I arrived at Michigan in the late 1980s, Dick Nisbett was doing research on the way that one's postgraduate discipline-for example, chemistry, law, or psychology-affects one's use of statistical rules of inference. One of the things that impressed me about this