Thefts were staged for pairs of witnesses who were subsequently given a photo-lineup from which to identify the thief. Individual witnesses (N = 159) were asked by a uniformed security officer to make an identification from either a perpetrator-present or perpetrator-absent lineup. Witnesses who made an identification (n = 122) were randomly assigned to receive either no information, or one of two types of extra-memorial information: (a) the co-witness had identified the same person from the photospread, or (b) the identified suspect had prior criminal involvement. The officer then asked witnesses to make confidence judgments and ecphoric similarity judgments (i.e., perceived resemblance of the witness's memory of the perpetrator to the identified suspect). Analyses were conducted on a subsample of witnesses (n = 96) selected such that all conditions contained the same proportion of witnesses who had viewed each of the three confederate thieves. Extra-memorial information was expected to inflate witnesses' confidence judgments but not their ecphoric judgments. Results indicated that prior-involvement information inflated the confidence of inaccurate witnesses, but not that of accurate witnesses relative to witnesses in the control condition. Co-witness information did not significantly inflate confidence for either accurate or inaccurate witnesses relative to control witnesses. Contrary to predictions, extra-memorial information had the same effect on witnesses' ecphoric similarity judgments as it did on their confidence judgments. Two-thirds of the witnesses who received extra-memorial information indicated that they relied on that information when making their confidence judgments, but there was no evidence that witnesses were able to accurately estimate the degree to which their confidence judgments were actually affected. Eyewitness confidence judgments can be problematic because they are susceptible to inflation effects due to witnesses' exposure to extra-memorial information. Unfortunately, ecphoric judgments are also susceptible to the same inflation effects. These findings highlight the critical importance of two characteristics of a properly conducted lineup. First, the person who conducts the lineup should not know which person in the lineup is the suspect. Second, confidence and similarity judgments should be secured from the witness at the time of the identification, prior to the introduction of any extra-memorial information.