At this point, it is undeniable that much of our behavior is either determined or heavily influenced by factors that are below the level of our awareness or that we are not able to consciously control. There are many examples of this in the literature. For example, our arousal level, which in most contexts is beyond our control (unless we're master yogis), can influence our perceptions of the attractiveness of others. Thus, studies have shown that we will rate people as being more attractive if we view their photos (or encounter them in person) immediately after a strenuous exercise session, or while we are standing near the edge of a high ledge. In these situations, our arousal levels are high, and we can mistakenly attribute the increased level of arousal to a person rather than their actual causes.
Another example of the unconscious influences on our behavior is the way our current goals affect our perceptions of value. The effect of current goals on intertemporal choice is well documented, but they can also influence current choices in ways of which we are not always aware. Imagine that we are shopping for cars, and after viewing a first car, we become hungry (though we don't consciously notice this) just before viewing a second car. If we do not satisfy this goal before viewing the second car, we will perceive it as less valuable than we otherwise would. Thus, even if the first and second car would ordinarily be viewed as equally valuable, viewing the second car while we have a strong goal that is not related to car-buying (in this case, hunger) will tend to cause us to perceive it as less valuable than the first car. This is because goal-irrelevant objects tend to be "devalued" relative to their default value.
As these and many other examples show, unconscious influences on our behavior are pervasive. Recently, however, researchers have begun to research "unconscious attitudes," and in particular "unconscious prejudices." Several methods for testing for these implicit attitudes have been developed, including implicit priming, the implicit association test (and its more recent incarnation, the Go/No-Go Association Test), and the evaluative movement assessment. The impetus for these tests was the recognition that explicit tests of unsavory attitudes and prejudices would be overly influenced by demand characteristics. Implicit tests, it was thought, would tap into these attitudes in ways that would not be influenced by conscious beliefs about things like the popularity of the attitudes. And originally, there was good reason to think that the IAT, the most popular of these tests, was doing just this. IAT performance is not completely dependent on intention, and because IAT results (in the form of rank-order preferences) don't correlate well with explicit measures, there's good reason to believe that they are showing us things that explicit measures couldn't.
It's very "sexy" (to use the social psychologists' own buzz-word) to talk about unconscious prejudices and implicit attitudes, and for this reason, it doesn't surprise me that IAT results showing that white participants seem to harbor unconscious prejudices towards African Americans, or other out-groups, are now receiving attention in the popular press. As a quote from the above-linked book demonstrates, these results are striking:
One of the reasons that IAT has become so popular in recent years as a research tool is that the effects it is measuring are not subtle the IAT is the kind of tool that hits you over the head with its conclusions. When theres a strong prior association, people answer in between four hundred and six hundred milliseconds, says [psychologist Anthony G.] Greenwald. When there isnt, they might take two hundred to three hundred milliseconds longer than that which is in the realm of these kinds of effects is huge. One of my cognitive psychologist colleagues described this as an effect you can measure with a sundial. (p.77)Recent work,however, has shown that conclusions of "prejudice" from IAT scores may be unwarranted. First, there appear to be several methodological problems with the IAT (e.g., the evaluations of one category are influenced by those of the opposing category; see the discussion in this paper). In addition, people routinely evaluate nonwords more negatively than words. Can we infer, then, that people have an implicit prejudice against nonwords? In reality, it's likely that things like familiarity and arousal level are influencing IAT results. There are new implicit tests (e.g., the one in the linked paper) that are designed to avoid some of these problems, but even with these tests, inferring "prejudice" or even "attitudes" in general is discouraged. Thus, while it is safe to conclude that unconscious factors influence our perceptions and behavior, it is premature to conclude that we can test for the existence of unconscious prejudices.
Will the popular press take heed of the methodological and theoretical problems with IAT scores? Probably not. While Lakoff's framing analysis has been criticized by many conservatives, neither conservatives nor liberals have paid any attention to the fact that the conceptual metaphor theory upon which it is based has been largely discredited by empirical research. IAT and implicit prejudices are even sexier than framing analysis, and seem to hold appeal for both liberals and conservatives. Hopefully, like most fads, this one will eventually just fade away, at least until the research catches up with the conclusions that have sparked the popular interest in the first place.