Sporer, S. L., Penrod, S., Read, D., & Cutler, B. (1995). Choosing, confidence, and accuracy: A meta-analysis of the confidence-accuracy relation in eyewitness identification studies. Psychological Bulletin, 118, 315-327.
Sporer et al., (1995) present an in depth meta-analysis conducted using results and data from previous studies examining the relation between confidence-accuracy. Thirty studies were used in the meta-analysis which used some kind of mock crime event in their methodology, used both target present and target absent lineups, had collected both confidence and accuracy measures from participants, and still had raw data to use for further analyses. The main addition to the existing research on the CA relation in the meta-analysis is the examination of the moderator variable of whether or not people identify someone out of a lineup or do not make an identification.
Clearly the processes would be different for those who choose compared to those who do not choose someone from a lineup. The authors cite evidence from Bem's (1972) self-perception theory regarding the notion that the act of making some kind of decision, rather than indecisiveness, should result in greater confidence in the ability to have made a worthwhile decision (i.e., in the case of eyewitness identification, an accurate one). Thus, stemming from the notion of these inherent differences between choosers and non-choosers, Sporer et al., (1995) ran separate analyses for choosers (those who chose either a correct or incorrect person from the lineup) versus non-choosers (those who rejected the lineup). The effect sizes for choosers were much higher than for non-choosers, even for weighted effect size estimates. The real interesting thing is that the effect sizes, weighted, were higher than the overall effect size with both choosers and non-choosers included. Additionally, those who chose were much more confident compared to those who did not make an identification from the lineup. This indicates the strong role of the moderator variable of being a chooser or a non-chooser when faced with making a lineup identification decision.
Being a meta-analysis, the article provides a variety of information from the various studies examined, including whether or not the mock crime event was live or recorded to video, gender of the target, scale range used for confidence statements, proportion of correct identifications, and correct decisions broken down for choosers and non-choosers. This is an excellent review of the CA relation by some of the biggest and well known researchers in the field of eyewitness identification. If you only had time to read one article about the relation between eyewitness confidence and accuracy, I would recommend you read Sporer et al., (1995).
--By DJP
Penrod, Sporer, Read, and Cutler’s meta-analysis of 30 studies concerning confidence-accuracy was my choice of reading from Dwight’s selected articles. In essence, the article addresses the widely accepted (in the research field, at least) notion that confidence of selection accuracy and actual selection accuracy have a weak correlation, and rounds up numerous studies in order to get a better grasp on the accuracy of this idea. One study polled sixty three psychologists in the field of eyewitness testimony, and one question addressed the weak relation between witness confidence and accuracy. Of those polled, only six felt the research was inconclusive, one stated that the research didn’t support the statement, one did not respond, and none thought that the reverse was true. So this survey, when broken down, essentially states that fifty five of the sixty three psychologists at minimum “tend to agree” that an eyewitness’s confidence is not a good predictor of their identification accuracy. On the flip side, though, while no true statistics were provided to my knowledge, “surveys of the lay public…reveal substantial, cross-cultural belief that confidence predicts accuracy.” This seems like common sense to most folks, but in this situation, research seems to indicate otherwise.
A number of factors have been identified with regards to moderator variables and their effect on confidence-accuracy, including (but not limited to) timing of the confidence judgment, other co-witness’s identifications, how distinctive the suspect is, and the similarity in the suspects appearance between encoding and retrieval. To me, while its good that these have been identified, it seems that most of these variables are next to impossible to account for, much less act upon. I know this isn’t a major point of the meta-analysis, but I think it’s worth noting that no matter how many advances are made in the implementation of lineups and decreasing the effects of system variables, that variables as significant as those listed above can never really be accounted for.
SB