Awareness of implicit attitudes reexamined: Large-scale tests in two experimental paradigms

Influential classic and contemporary accounts have posited that implicit attitudes reside beyond conscious awareness. This idea has been challenged by robust evidence demonstrating that participants can predict their own implicit attitudes highly accurately. However, past tests have all been conducted using well-known targets (e.g., racial groups), about which participants possess ample relevant knowledge. As such, predictions of implicit attitudes may have relied on inferential mechanisms drawing on sources of knowledge other than privileged introspective access, most prominently explicit attitudes. In this project we conducted eight experiments (five preregistered; total N = 6,794) in which implicit attitudes and their explicit counterparts were manipulated to shift in opposite directions, thus eliminating the latter as an obvious source of inference. Across experiments, mean-level implicit attitude predictions (β = 0.31) were directionally consistent with explicit (β = 0.49) rather than implicit attitudes (β = 0.20). Moreover, although predictions and implicit attitudes were associated with each other at the level of individual participants (β = 0.20), the relationship was considerably stronger among participants for whom explicit and implicit attitudes were aligned with each other (β = 0.38) than among participants for whom the two were dissociated (β = 0.11). These results generalized across two learning paradigms (impression formation and attribute conditioning), two implicit attitude measures (IAT and EPT), and two types of experimental design (between-participant and within-participant). Echoing classic and contemporary accounts, the present data suggest that implicit attitudes are largely unconscious, and successful implicit attitude predictions are subserved primarily by inference rather than introspection.