Eyewitness Identification: Identification Accuracy and the Effect of Foil-Suspect Similarity

Abstract

The study involved of the participation of 297 undergraduate psychology students to determine how foils should be employed in eyewitness identification procedures. The aim was to investigate how the similarity between the perpetrator and fouls can impact recognition performance. Participants were tested in their ability to recognise target faces among each line-up under all four conditions: high similarity with target present, high similarity with target absent, low similarity with target present, low similarity with target absent and rated their confidence for each response. The result generated from the experiment supported the proposed hypotheses. Low similarity line-ups elicited more correct identification and high similarity line-ups elicited more false identifications, low similarity line-up had also elicited higher confidence in responses. The potential implication of this experiment’s result is discussed. Future studies should further examine the effect of foil-to-suspect similarity on identification accuracy.

Eyewitness Identification: Identification Accuracy and the Effect of Foil-Suspect Similarity

Eyewitness testimony frequently plays a vital role in the criminal justice system. The recollections of events and suspect identifications from eyewitnesses can have a major impact on the outcome of a trial. Eyewitness identification is among the most commonly used method to bring out compelling evidence against criminal defendants. The courts are often susceptible to placing great faith in the accuracy of an eyewitness’s recounts and many convictions made were depended heavily on the memory operation of eyewitnesses. Furthermore, eyewitness testimony can be the only evidence available in the verdict of certain criminal case. It is estimated that about 77,000 individuals are suspects involved in eyewitness identification parades per year in the United States alone (Goldstein, Chance & Schneller, 1989). The current guidelines for constructing identification line-up employed by police investigators in the United States (Technical Working Group for Eyewitness Evidence, 1999) recommend that the foils, individuals who are known to the police as innocent and strategically placed in the line-up, should generally match the appearance of the perpetrator as verbally described by the eyewitness. And when description of perpetrator provided by the witness is insufficient or when the given description differs greatly from the suspect’s appearance, foils should be of similar look in significant features to the suspect.

However, the substantial number of innocent individuals who were wrongfully convicted in which the mistaken eyewitness identifications largely contributed to (Scheck, Neufeld & Dwyer, 2003; Wells, Small, Penrod, Malpass, Fulero & Brimacombe, 1998) indicates that the standard eyewitness identification line-up procedures suffer from poor accuracy, and the need for better eyewitness identification line-up that would improve the accuracy of the identification decisions.

Studies have articulated concerns of the standard eyewitness identification line-up’s effectiveness and have made many recommendations for constructing and conducting better line-up (Wells, Small, Penrod, Malpass, Fulero & Brimacombe, 1998). Improved procedures are being implemented by judicial systems all over the world. This study aimed to investigate how the similarity between the perpetrator and fouls can impact recognition memory performance. To attempt to answer how should police employ foils in the line-up.

The hypotheses of this experiment were based on what is advised by Technical Working Group for Eyewitness Evidence (1999). It was hypothesised that line-ups that included high similarity foils would elicit more correct identifications of perpetrators and low similarity line-ups would elicit more false identification of innocents. Furthermore, it was hypothesised that participants would be far more confident in their identification decisions if the line-ups included foils who are low in similarity to the perpetrators.

Methods

Participants

The study involved two hundred and ninety-seven University of Adelaide undergraduate psychology students consisting of 85 males and 211 females.

Material

Participants were tested via an online survey: http://adelaidecogscifaces.appspot.com/.

The portraits of people used to construct the basic face recognition paradigm are face stimuli are taken from a database of faces (Glasgow Unfamiliar Face Database) that have been rated for their similarity to one another. The similarity ratings were used to construct each line-up, faces that were rated highly in similarity to one another were placed into high similarity line-ups, faces that were rated as low to one another were placed into low similarity line-ups.

Participants rated confidence via a ratio scale. It includes four responses from least confident to most confident: ‘Not confident at all, it’s a guess’; ‘Somewhat confident’; ‘Very confident’; ‘As certain as I can be’. Which indicates the numerical value from 0 to 3 respectively.

Procedure

Participants were accessed via an online survey, in which they were shown a series of target faces and were instructed to memorised and later recognise these faces among each line-up. They completed a brief activity involved change blindness task. Participants were then presented with 4 types of line-up, these included high similarity with the target present, high similarity with target absent, low similarity with target present and low similarity with target absent. Participants were told to identify the previously seen faces by recording the number of the person they thought was the target or by selecting ‘Not Present’. Participants then rated their confidence in each of their identification decisions via a ratio scale. Once the participants had completed the survey they were thanked for their participation.

Result

A summary of the accuracy of participants in identifying target faces among line-ups under the four conditions: high similarity with target present, high similarity with target absent, low similarity with target present, low similarity with target absent and confidence rating for their response.

Table 1. Summary of the condition of line-up, mean proportion correct and confidence rating in each condition

Note. The accuracy data is reported as proportion correct.

The data in Table 1 shows the mean proportion correct of two hundred and ninety-seven participants under each condition. And the mean confidence rating of all participants.

Figure 1. Column chart showing the portion of accurate response by conditions

Figure 2. Column chart showing confidence of response by similarity

Figure 3. ROC plot showing identification performance

Note. Each point on the ROC plot represents the rate of correct and false identification at each level of confidence.

Figure 4. Boxplot of high similarity line-ups with target present (SD= 0.17)

Figure 5. Boxplot of high similarity line-ups with target absent (SD= 0.27)

Figure 6. Boxplot of low similarity line-ups with target present (SD= 0.19)

Figure 7. Boxplot of low similarity line-ups with target absent (SD= 0.28)

Figure 8. Boxplot of confidence in high similarity line-ups (SD= 0.43)

Figure 9. Boxplot of confidence in low similarity line-ups (SD= 0.44)

Discussion

The result generated from this experiment did not support the proposed hypotheses. It was shown that low similarity line-ups demonstrated noticeable higher mean proportion correct in target identification than line-up with high similarity foils. Additionally, high similarity line-ups elicited slightly higher proportion of false responses than low similarity line-up regardless of target presence. Moreover, the result indicates that participants were marginally more confident in their response when the line-ups were low in similarity, whereby it was hypothesised that participants would be more confident in low similarity line-ups. This somewhat supported the hypothesis. However, the difference in confidence between high and low similarity are minor. The ROC plot has also shown that low similarity line-up yields better performance in higher proportion correct.

The result generated from this experiment contradicts the findings of Fitzgerald et al (2013). Whereby it was concluded by Fitzgerald et al that when compared with line-ups that included foils of high foil-suspect similarity, the rate of innocent misidentifications nearly doubled when line-ups included foils of low foil-suspect similarity. Similarly, Police investigators were advised by Technical Working Groups for Eyewitness Evidence (1999) to place suspect in line-up with foils of similar look in significant features. Furthermore, the findings of Fitzgerald et al (2013) suggested that participants would be far more likely to identify target regardless of whether it was correct or false. This may be represented by the confidence in identification decisions. This is somewhat consistent to what was found in this experiment.

Anyhow, the contradictions between the result of this experiment and findings from previous studies may be explained by the fact that the face recognition test employed here differs significantly from eyewitness line-up procedures utilised in previous studies. The varied distribution of datasets and the presence of many outliners could invalidate the result of the experiment all together. It is possible that a similar result may not be obtained if the experiment is replicated. Overall, this experiment does not dismiss the rule that has been viewed as the best practice in eyewitness identification procedures, it is still recommended to implement foils highly similar to suspects into line-ups.

References

• Fitzgerald, R., Price, L. & Charman, S. (2013), The effect of suspect-filler similarity on eyewitness identification decisions: a meta-analysis. Psychology, Public Policy, and Law.
• Goldstein, A., Chance, J. & Schneller, G. (1989), Frequency of eyewitness identification in criminal cases: A survey of prosecutors, Bulletin of the Psychonomic Society, vol. 27, no. 1, pp. 71-74.
• Scheck, B., Neufeld, P. & Dwyer, J. (2003), Actual innocence: When justice goes wrong and how to make it right, New York, NY: New American Library.
• Technical Working Groups for Eyewitness Evidence. (1999). Eyewitness evidence: A guide for law enforcement. Washington, DC: US Department of Justice, Office of Justice Programs.
• Wells, G. L., Small, M., Penrod, S., Malpass, R. S., Fulero, S. M. & Brimacombe, C. A. E. (1998), Eyewitness identification procedures: Recommendations for lineups and photospreads, Law and Human Behavior, vol. 22, no. 6, pp. 603-647.