The Decomposition and Recomposition of Eyewitness Identifications: Eyewitness Reliability, Guessing and Lineup Bias

Abstract

Traditional measures used by eyewitness researchers often paint an incomplete picture of the complex process that is an eyewitness identification. We seek to answer the following questions: 1) What proportion of suspect identifications are simply guesses, and what is the source of these guesses? Are they rooted in blind luck or do they stem from a poorly constructed (biased) array, in which the suspect (guilty or innocent) may be the only lineup member who resembles the perpetrator and is selected by someone who has a weak memory for the perpetrator’s appearance and merely selects the best match to their memory. 2) How biased is an array? While there are procedures which attempt to estimate lineup bias, they are often expensive (in participants and/or time), difficult to reproduce (live arrays), may be indirect (such as using mock witnesses to select “suspects” merely on the basis of suspect descriptions) and/or may not reflect the performance of the array in its natural circumstances as presented to actual witnesses. 3) What proportion of correct identifications are reliable ones? In particular, in a given set of eyewitness parameters, what proportion of eyewitness are able to identify the perpetrator when present and also correctly reject the perpetrator when absent? We describe a procedure for taking eyewitness summary data as input, and decomposing those numbers into the proportion of reliable eyewitnesses (witnesses who are accurate in both perpetrator-present and perpetrator-absent arrays), cautious eyewitnesses who never make a selection, and eyewitnesses who can be induced by variations in identification procedures to make guess. From these terms we calculate what percentage of hits in a given set of lineup parameters can be attributed to reliable eyewitnesses, how many are caused by random guessing, and how many result from guessing due to bias in presentations of arrays.