Evaluating eyewitness identification procedures: ROC analysis and its misconceptions

John Wixted, Laura Mickes

Research output: Contribution to journalComment/debate

90 Downloads (Pure)

Abstract

ROC analysis is a straightforward but non-intuitive way to determine which of two identification procedures better enables a population of eyewitnesses to correctly sort innocent and guilty suspects into their respective categories. This longstanding analytical method, which is superior to using the diagnosticity ratio for identifying the better procedure, is not in any way compromised by the presence of fillers in lineups and is not tied to any particular theory of memory or discrimination (i.e., it is a theory-free methodology). ROC analysis is widely used in other applied fields, such as diagnostic medicine, and this is true even when the medical procedure in question is exactly analogous to a lineup (e.g., a detection-plus-quadrant-localization task in radiology). Bayesian measures offer no replacement for ROC analysis because they pertain to the information value of a particular diagnostic decision, not to the general diagnostic accuracy of an eyewitness identification procedure.
Original languageEnglish
Pages (from-to)318–323
Number of pages6
JournalJournal of Applied Research in Memory & Cognition
Volume4
Issue number4
Early online date5 Sept 2015
DOIs
Publication statusPublished - Dec 2015

Cite this