Ten ‘good’ visual imagers and ten ‘poor’ imagers, representing the upper and lower halves of an unselected sample of adult volunteers, were given a task of detecting differences between altered pictures presented in pairs, either simultaneously or successively. Performance was measured by reaction time. ‘Good’ imagers reacted faster than ‘poor’ imagers, and the mode of presentation (simultaneous v. successive) did not affect their performances. ‘Poor’ imagers, on the other hand, reacted significantly slower when the pictures compared were presented successively. The results suggest that self-rated visual imagery can be used to predict the successful recall of visual information required in a discrimination task.