Eyewitness evidence is the often the strongest evidence against a defendant. Unfortunately, eyewitnesses identification errors are also the most common cause for wrongful convictions. The study of reliability and accuracy of eyewitness has taken two approaches. The first one studies systems variables, which are variables that can be controlled by the legal system (eg. Lineup procedure and so forth). The second approach deals with estimator variable. Estimator variables involve things like the eyewitnesses confidence in his/her decision, amount of time taken to ID the perpetrator from the lineup, etc. Estimator variable are also called "postdictors" b/c the eyewitness IDs first and then the postdictors are used to estimate the likely accuracy of the ID. There are a number of postdictors and the purpose of this article is to see if the postdictors are work for both within race and between race IDs.

The authors used the specific postdictors of (1) eyewitness confidence, (2) decision time, and (3) judgement strategy. These are postdictors that have been shown to be useful in own-race experiments but have never been used for other-race eyewitness accuracy studies. Eyewitness Confidence has been the most studied postdictor and is generally accepted as a poor postdictor of eyewitness accuracy. The authors also talk about two types of judgment strategies. Relative judgments occur when a person is presented with a lineup and they compare and contrast the lineup members in order to make their decision.  Absolute judgments occur when the witness makes an all or none decision for each lineup member individually.  Past research has shown that absolute judgment strategies lead to considerably more accurate responses. Therefore, knowing the type of strategy the eyewitness used (either relative or absolute) we can better predict their accuracy or the likeliness of them making false positives. Decision time is related to response latency (how long it takes for the eyewitness to make the id when they look at the lineup). The decision should be easier to make, and therefore quicker, when the identified face is the guilty indv. Vs. when the identified face is another innocent indv.

The authors’ predicitons:

the factors will be useful in postdictors of eyewitness accuracy

decision time will be the most powerful postdictor of accuracy

Methods-

Participants: 249 intro psych. Students (128 Asian and 121 White).

The Crime: 6 videos with one early 20s male; 3 White and 3 Asian. They stole money from a woman’s unattended purse. Woman confronted perp and perp spills purse and runs off. Ends with lady picking up the contents of her purse.

Lineups: The participants viewed target-absent (no confederate/perp) and target-present (confederate/perp) photo lineups. The foils were chosen from large pools of pictures and were chosen b/c they were similar looking to the perp.

Response Latency: the time it took from presentation of the lineups to the actual decision.

Measures: Relative-absolute judgements were m. using 2 self-report items. Confidence was m. on a 7 point Likert scale. This m. how certain the eyewitness was in their identification decision.

Procedure-

Participants randomly assigned to a particular video. Participants were separated and then asked to describe the perp (open-ended) and then fill out a description form (close-ended). Participants randomly assigned to target-absent or target-present and then asked if the perp was present or absent. If present then which position was he? Then thye did their confidence item and the relative-absolute questionaire (the order of this was all randomized for diffferent participants).

Results-

Analyses for choosers and nonchoosers in own- and other race ids were conducted to m. overall accuracy. Made sure that the postdictors replicated previous findings in the own-race category, otherwise they couldn’t have meaningful results for the other-race condition. Used a regression analysis and determined that the postdictors used were useful in determining accuracy.

Accuracy Rates: Neither choosers nor nonchoosers were significant in the difference btw identification of own-race or other-race perps. Nonchoosers were significantly more accurate than choosers.

Choosers: Choosers were correct 44% of the time. The pattern of effect expected were found for white perps ided by white participants. Confidence and accuracy are moderately correlated and there is no difference btw the strength of the relationshiop for own- and other-race categories. Decision time is correlated with accuracy for choosers – suggesting that the less time the identification took the more accurate it was. This is somehwat stable across the own- and other-race situations.

They broke the decision time down into two group: 1. 1-15s and 2 15s or more. The first group were significantly more accurate than the slower people. Therefore, in the real world eyewitnesses who make their identification quicker are more likely to be accurate and make fewer errors. Then to drive the point home that 15s was an arbitrary number they broke the decision time down into 9 categories. Refer to figure 1 in the article- this shows a dramtic decline in accuracy after 15s has elapsed.

Correlation btw type of judgement used and accuracy was found for white witnesses identifying white perps. Suggests that absolute judgements are more likely to be accurate than relative judgements. Correlation didn’t hold across conditions- in the other-race condition the correlation disappears. Judgement is a good postidictor in own-race but not other-race conditions.

Regression showed significance in only one postdictors accuracy: confidence. However, when the data was collapsed into own- vs. other-race identifications, only the own-race id was significant. For own-race confidence and judgment were useful postdictors. This suggests that postdictors that are useful in own-race identification are not necessarily useful in other-race identification.

The discriminant function analyses showed 6\that 69% of the chooser were classified into accurate or inaccurate groups correctly. The chi-square supported the functions category assignments. Own-race identification was accurate 67% of the time and other-race was accurate 66% of the time. Again postdictors are useful for own-race, not other-race, identification accuracy.

Nonchoosers: 60% of nonchoosers were accurate. No significant correlation were found for confidenc and accuracy or relative vs. absolute judgment. Interestingly, Asains who aswered with shorter response latency were more accurate but the opposite was true for the White. The shorter their decision time, the more inaccurate their response. Whites identifying Asains show the same pattern as they do with Whites but Asains identifying Whites had no relation btw latency of response and accuracy of id. Regression equs\ation shows nothing significant and the discriminant function analysis didn’t predict beyond chance.

Discussion-

Non of the variables were good postdictors for nonchoosers. The three postdictors were all useful for the choosers in own-race situations. Confidence and judgement were found to significantly postdict accuracy. These factors could also be used to categorize the participants 67% of the time. Findings suggest that eyewitness accuracy could be enhanced if the police used such simple strategies in own-race conditions. However, they will prove fruitless in the other-race condition and can’t postdict beyond chance. These postdictors methods must be collected at the time of identification or they may change over time and be unreliable. The findings also suggest that the process of id in own-race is very different from the process of id in other-race but what is that difference?

Conclusion-

Factors commonly used to postdict accuracy in own-race identification do not work in other-race id. Wrongful convictions are related to the extent the police use estimator variables to assess accuracy of other-race identifications. Findings suggest that the legal system should prevent false ids to begin with and not rely on accuracy assessment.