Abstract
Taking advantage of the pictorial priority effect, a new type of person-authentication system called picture-authentication is currently being proposed. However, some issues necessary for constructing a safe and easy-to-use picture-authentication system remain unresolved, e.g., effects of repetition on recognition tests or using the same pictures repeatedly as dummies, as well as differences between autobiographical versus non-personal pictures. A long-term recognition experiment was executed with two kinds of pictures: autobiographical (AB; brought in by the participants), and self-selected (SS; chosen from a given set of pictures by a participant). Results of two recognition tests (3 or 6-week-delay from learning), AB pictures showed slightly higher performance than SS, and confidence rate was much better for AB than SS. Both AB and SS showed large improvements in performance on the second test. However, recognition performance decreased when the same dummies were used from previous tests, compared to when the novel dummies were used.
