Signal Detection Analysis

Signal Detection Theory quantifies d' and criterion bias by disentangling discriminability from response strategy using hit and false alarm rates. You apply SDT decision logic to binary behavioral responses, then use correction rules for extreme rates before calculating sensitivity and bias measures.

Signal detection analysis requires log-linear or 1/(2N) adjustments when hit or false alarm rates hit floor or ceiling. These corrections prevent infinite d' values, ensuring bias-corrected metrics like d', c, and beta remain computable and valid for publication-ready reporting.

SDT rating scale paradigms use d' for equal-variance assumptions, da for unequal variance, and Az for nonparametric ROC-based sensitivity. Choose based on your distribution assumptions and paradigm design, applying ROC guidance to select the appropriate bias-corrected sensitivity metric.

Signal detection analysis constructs ROC curves by plotting hit rates against false alarm rates across criterion levels. For recognition memory, report d' or Az, bias measures like c and beta, confidence intervals, and statistical tests using canonical SDT reporting checklists.

Signal detection theory supports ABX and same-different paradigms with specific decision logic and d' conversions tailored to each design. SDT provides bias-corrected sensitivity metrics across these psychophysics tasks, including correction rules and reporting guidance for publication.

Raw accuracy conflates perceptual sensitivity with response criterion shifts, misleading cognitive neuroscience findings. Signal detection analysis disentangles discriminability from bias settings before producing summary statistics, requiring bias-corrected metrics whenever criterion shifts or base rate differences exist.