Can people detect eyewitness-identification accuracy within and across situations?

Thefts were staged 108 times for as many witnesses who were subsequently given a photo lineup for identifying the thief. The thefts were staged under conditions designed to yield low (33%), moderate (50%), or high (74%) proportions of correct identifications of the thief. Corroborating past research, the relationship between witnesses' identification accuracy and witnesses' confidence was negligible within conditions. There was no evidence that the confidence-accuracy relationship changed across conditions or that witness confidence changed across theft conditions. A representative sample of 48 witnesses (8 accurate-iden tification and 8 false-identification witnesses from each of the three theft conditions) was cross-examined. Subjects (n = 96) viewing the cross-examinations showed no ability to detect accurate- from false-identification witnesses within conditions as measured by subjects' belief of witnesses. Although subjects changed their rate of belief of witnesses as a function of the theft conditions (62%, 66%, and 77%, respectively), the rate at which subjects discounted witnesses' testimony was insufficient across conditions. Subjects were shown to be especially overbelieving of witnesses when the rate of witness accuracy in that condition was low.