Drawing on research on subjective confidence, we examined how the confidence and speed in responding to personality items track the consistency and variability in the response to the same items over repeated administrations. Participants ( = 57) responded to 132 personality items with a true/false response format. The items were presented five times over the course of two sessions. Consistent with the Self-Consistency Model, the confidence and speed with which an item was endorsed at its first presentation predicted the likelihood of repeating that response across the subsequent presentations of the item, thus tracking test-retest reliability. Confidence and speed also predicted the likelihood that others will make the same response, thus tracking inter-person consensus. However, confidence and speed varied more strongly with within-person consistency than with inter-person consensus, suggesting some reliance on idiosyncratic cues in response formation. These results mirror, in part, findings obtained in other domains such as general knowledge, social attitudes, and personal preferences, suggesting some similarity in the decision processes underlying the response to binary items: responses to personality items are not retrieved ready-made from memory but constructed at the time of testing, based on the sampling of a small number of cues from a larger population of cues associated with the item's content. Because confidence is based on the consistency with which the cues support a response, it is prognostic of within-person consistency and cross-person consensus. Theoretical and methodological implications are discussed.