Person Identification in a Legal Setting

Strathie, Ailsa (2010). Person Identification in a Legal Setting. PhD thesis Glasgow Caledonian University.

URL: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos...

Abstract

In a criminal prosecution, CCTV footage from the crime scene may play a central role in establishing the culprit's identity, yet it is well established that unfamiliar face matching is highly error-prone. In an attempt to optimise identification from CCTV images, expert evidence based on forensic image comparison may be provided in court. However, forensic image comparison techniques suffer from a major shortcoming. There is a lack of scientific evidence to support their use. The focus of this thesis is on supplying the data to address this issue, and three individual themes are explored.

The first theme examines three techniques that are used to demonstrate facial image comparison in court. Results show that none of the techniques increase accuracy. Instead, they appear to have a detrimental effect on identity matching decisions. Critically, these techniques also increase false positive responses, and, therefore, the likelihood that an innocent person will be convicted.

The second theme investigates an alternative strategy for identifying a person from imagery evidence. This investigation focuses on the possibility that the ears, which are subject to less variation in appearance than faces due to their rigid internal structure, may usefully inform identity matching decisions. In a series of five experiments, the possibility that images of the ears may optimise identification accuracy is explored. The results show that matching decisions based on images of ears alone are less accurate than decisions based on faces alone, but when images of ears are shown with images of faces, overall accuracy is increased.

The third theme examines how top-down influence such as emotional crime information, and expert evidence based on CCTV identification, influences unfamiliar face matching decisions. In a series of experiments, expert opinion evidence on identity is manipulated, and the effect on face matching accuracy is examined. The results show that inconsistent information significantly reduces matching accuracy, but consistent information fails to confer any benefit, which undermines the practice of providing expert evidence on CCTV identification in court.

Collectively, these three lines of investigation provide substantial new empirical evidence that questions the use of facial image comparison evidence in court, and the second theme offers a simple alternative method of optimising identity verification. Future work could extend upon these findings using more ecologically valid conditions, which would enhance the applied value of this work.