Eyewitness identification from lineups is fundamentally a signal detection task: the witness must discriminate between lineups containing the guilty suspect (target-present) and lineups where the suspect is innocent (target-absent). SDT analysis reveals that identification accuracy depends on both discriminability (the witness's memory strength) and criterion (their willingness to identify someone).
Diagnosticity and ROC Analysis
This ratio varies with criterion — not a pure measure of accuracy
ROC analysis: plot correct ID rate vs. false ID rate
AUC provides criterion-free accuracy measure
Traditional eyewitness research used diagnosticity ratios or individual hit/false alarm rates, which confound sensitivity and criterion. ROC analysis, advocated by Wixted and Mickes (2012), provides criterion-free accuracy measures that allow fair comparison of lineup procedures (simultaneous vs. sequential, different lineup sizes, and different instructions).
Policy Implications
SDT analysis has changed conclusions about lineup procedures. Sequential lineups, once thought to be more accurate than simultaneous lineups based on lower false identification rates, actually show similar or lower ROC performance — the apparent advantage was a criterion shift (more conservative responding) rather than improved discrimination. This finding has had direct implications for police lineup policy recommendations.