Caught in the act: False on-task contingency beliefs reverse the color-word contingency learning effect.

A key question in learning research concerns the role of awareness in human contingency learning. While a large portion of the contingency learning (CL) effect can be attributed to the episodic retrieval of previous SR bindings, a significant residual CL effect remains, which reflects the impact of global contingencies on behavior. In a high-powered (N = 168), pre-registered study, we implemented an on-task measure of awareness that prompted participants to actively search for underlying contingencies. This allowed for tracking fluctuations in contingency beliefs and relating these to the emergence of residual CL effects. We found that the residual CL effect is sensitive to propositional reasoning processes, as participants who completed the additional on-task assessment showed a larger residual CL effect than those in an off-task control group. This amplification was explained by increased levels of awareness in the on-task group, which were directly linked to a more pronounced residual CL effect. Notably, no residual CL effect emerged in the absence of awareness, regardless of whether awareness was assessed on-task or retrospectively. Interestingly, participants who repeatedly selected the same low contingency color in the on-task assessment showed a reversed residual CL effect, characterized by faster responding to low contingency color-word combinations that aligned with their false beliefs. Taken together, the present findings suggest that genuine human contingency learning - captured by residual CL effects that reflect the impact of environmental regularities on behavior - is not the product of automatic association formation but is fundamentally driven by active propositional reasoning processes.