Persons high and low in non-clinical paranoia (based on scores from the Paranoia Scale) were administered two implicit learning tasks that comprised information regarding the covariation between a stimulus (e.g. a face) and a specific characteristic (e.g. ''fairness''). To assess whether persons high in non-clinical paranoia were particularly sensitive to learning social information, both social (faces) and non-social stimuli (cars) were used. Results showed that the group high in non-clinical paranoia demonstrated implicit learning to all stimuli, irrespective of content. The group low in non-clinical paranoia showed greater implicit learning for non-social relative to social stimuli. The results partially support a content-specific bias since there were differences in social ratings relative to non-social ratings between the two groups. Finally, the group high in non-clinical paranoia was significantly more confident in their ratings relative to the group low in non-clinical paranoia for all stimuli. The implications of these findings for non-clinical paranoia are discussed.