Calculation of signal detection theory measures

Stanislaw, Harold; Todorov, Natasha · 1999 · OpenAlex-citations

DOI: 10.3758/bf03207704

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the underutilization of signal detection theory (SDT) in psychological research, attributing this gap to a lack of accessible instructional material in standard textbooks. The authors aim to demonstrate that SDT analysis is now as computationally straightforward as a t-test due to modern software, thereby encouraging broader application across disciplines such as perception, memory, lie detection, and medical diagnosis. The core motivation is to provide researchers with the mathematical formulas and practical methods necessary to calculate SDT measures, which separate sensitivity (discriminability) from response bias (decision criterion). The authors provide a comprehensive overview of SDT concepts and present specific mathematical formulas for calculating key measures across three primary experimental designs: yes/no tasks, rating tasks, and forced-choice tasks. For yes/no tasks, the paper details formulas for sensitivity measures $d'$ and $A'$, as well as bias measures $\beta$, $c$, and $B''$. It explicitly defines the necessary mathematical functions, including the $\Phi$ (phi) function for converting z-scores to probabilities and the $\Phi^{-1}$ (inverse phi) function for converting probabilities to z-scores. The text provides step-by-step derivations for these calculations, including simplified single-formula equations for $A'$ and $B''$ to avoid common computational errors. For rating tasks, the authors describe methods for constructing receiver operating characteristic (ROC) curves and calculating the area under the curve ($A_z$) as a bias-free measure of sensitivity. For forced-choice tasks, the paper explains how proportion correct serves as a sensitivity metric. The findings are presented as a set of validated computational procedures rather than empirical data results. The authors illustrate these methods using hypothetical data and graphical examples, such as Figure 1, which depicts signal and noise distributions to explain hit rates, false-alarm rates, and criterion placement. They clarify the assumptions required for parametric measures like $d'$ (normality and equal variance of distributions) and contrast them with nonparametric alternatives like $A'$, which do not require these assumptions. The paper also highlights the utility of ROC analysis in rating tasks to test the validity of the equal-variance assumption. The significance of this work lies in its role as a practical guide that lowers the barrier to entry for SDT application. By providing explicit formulas and discussing implementation via lookup tables, specialized software, and general-purpose spreadsheets, the authors enable researchers to accurately quantify performance without conflating sensitivity with response bias. This facilitates more rigorous analysis in fields ranging from jury decision-making to industrial quality control, ensuring that observed performance differences are correctly attributed to changes in discriminability or decision criteria.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-25; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• signal detection theory