Eyewitness Identification: A Signal Detection Theory Lab (2D)

Every eyewitness lineup decision is, underneath, a single number crossing a threshold: a felt strength of recognition compared against a personal decision criterion. This 2D companion renders that model as two overlapping Gaussian curves — one for an innocent lure, one for the true culprit — separated by Δμ and spread by σ, so the real signal-to-noise ratio d′ = Δμ/σ is always visible. Drag the criterion line to see the classic speed–accuracy trade-off of eyewitness testimony: a lenient witness racks up hits but also false identifications of innocent people, while a cautious one protects the innocent at the cost of missed guilty suspects. Run batches of simulated witnesses to watch the hit rate, false-alarm rate, and an empirically recovered d′ converge toward the values the sliders predict, then flip to biased ("suggestive") lineup instructions to watch the criterion shift while the recovered d′ does not — the real, well-documented mechanism behind eyewitness-identification reform.