Laboratory experiments can create working definitions of manipulated race, randomly assign participants to interact with black or white confederates, and measure a variety of proposed psychological mechanisms (implicit and explicit attitudes) to determine their effect on various types of discriminatory behavior. For example, Dovidio et al. (2002) conducted a multiphase experiment on how whites’ explicit and implicit racial attitudes predict bias and perceptions of bias in interracial interactions. At the beginning of the term, white college students completed a 20-item standardized measure of prejudice, the Attitudes Toward Blacks Scale. Later in the semester, 40 students (15 male and 25 female) participated in what they believed to be two separate studies. In the first, a decision task required participants to respond as quickly as possible—after the letter P or H was displayed on a computer screen—as to whether a given word displayed for each trial could ever describe a person or a house. Unbeknownst to them, on critical trials versus practice trials the letter P was preceded by a standardized schematic sketch of a black or white man or woman, presented at subliminal speeds (0.250 seconds). This level of presentation has been shown repeatedly to prime relevant associations in memory and, in particular, stereotypes. As in countless other studies (e.g., see Fazio and Olson, 2003), the findings in this study revealed subtle forms of stereotypic association when people responded more quickly to negative words (“bad,” “cruel,” “untrustworthy”) preceded by a black face and to positive words (“good,” “kind,” “trustworthy”) preceded by a white face, and more slowly to the converse combinations. As is typical with this method, no participant reported being aware of the subliminal faces. Such studies show how researchers can measure automatic and unconscious racial bias, regardless of expressed levels of prejudice (Devine, 1989). At this point, then, the experimenters had access to two kinds of attitudes—the explicit ones expressed on the questionnaire and the implicit ones suggested by the participants’ speed of stereotypic associations. These are the psychological causes of different kinds of discrimination hypothesized in the next step.

In what participants assumed to be a separate study focused on acquaintance processes, the participants met separately with two inter-