I recently took the Implicit Association Test on racial impressions at Harvard.  While I don’t necessarily disagree with my results, I still have strong doubts about the way the test was constructed.  I was shown to have a moderate preference for my own race, white, which doesn’t really surprise me.  After all, I grew up in a predominantly white community, so my early exposure to other races was limited.  I presume it’s probably biologically natural to prefer one’s own race, and there is also a clear historical advantage to being white in America which could quite logically, even as it is unfortunate, strengthen an unconscious desire to associate with white people.  What really bothered me about the test was that it seemed from watching the Dateline interview that the researchers had an agenda.  Every time the test was given, black faces were grouped with negative characteristics in the first part of the test.  Why?  Why don’t the researchers randomize the test so that some participants receive the grouping of black faces with positive characteristics early in the test and then switch the categories in the 2^nd half?  Does this reveal a bias on the part of the researchers?  Rather than testing the preconceptions held by participants, is it possible that the test actually places an association of black with bad in the participants’ minds that then impacts the 2^nd half of the test?  This would correspond well with the observation recorded by Gladwell in Blink that participants who were primed to think about old age in a sentence construction test subsequently behaved in line with that association.  Is the IAT priming participants to form associations between pictures of black people and negative terms?  I would like to know if the results change at all when the order of the test is changed.  William James makes a point of explaining how memory enables us to learn in Talks to Teachers.  Through recalling a series of reactions, we can substitute one behavior for another, “eliminating all the intermediary steps” (p. 21).  For James, one interaction can be instructive, so it is logical to conclude that a test item could potentially teach participants how to behave rather than simply measure that behavior.