Signal Detection Theory: The Decision Process

Signal Detection Theory (SDT) is a framework for understanding how decisions are made under conditions of uncertainty. Originally developed for radar and communications engineering, it was adapted for psychology by Wilson Tanner and John Swets in the 1950s. SDT revolutionized psychophysics by separating an observer's ability to detect signals (sensitivity) from their willingness to report them (bias). This theory is fundamental to cognitive psychology and is related to the absolute threshold.

Table of Contents

1. What Is Signal Detection Theory?
2. The Signal Detection Problem
3. Key Concepts
4. Measuring Sensitivity and Bias
5. Applications
6. Try the Experiment
7. Key Takeaways

What Is Signal Detection Theory?

Signal Detection Theory provides a mathematical framework for analyzing decisions made under uncertainty. Unlike classical threshold theory, SDT recognizes that detection involves both the sensory process and the decision process.

The Core Insight

Two Separate Processes:

1. Sensitivity (d'): The ability to discriminate signal from noise
2. Response Bias (β or c): The tendency to say "yes" or "no"

Why This Matters:

• Same sensitivity can yield different performance based on bias
• Bias can change without sensitivity changing
• Separating them provides clearer understanding

The Signal Detection Problem

The Four Outcomes

In any detection task, there are four possible outcomes:

Hit: Correctly detecting a signal
Miss: Failing to detect a signal
False Alarm: Reporting a signal when none exists
Correct Rejection: Correctly saying no signal

The Challenge

Uncertain Evidence:

• Sensory evidence is noisy
• Signal and noise distributions overlap
• Perfect detection is impossible
• Must set a decision criterion

Key Concepts

Signal and Noise Distributions

Noise Distribution:

• Sensory activity when no signal present
• Normally distributed
• Mean = 0 (by convention)

Signal + Noise Distribution:

• Sensory activity when signal present
• Also normally distributed
• Mean shifted by signal strength
• Same variance as noise

The Decision Criterion

Setting the Threshold:

• Observer sets criterion on evidence axis
• Above criterion → "Signal present"
• Below criterion → "Signal absent"
• Criterion placement determines bias

Factors Affecting Criterion:

• Payoffs for different outcomes
• Base rates of signals
• Costs of errors
• Instructions and motivation

ROC Curves

Receiver Operating Characteristic:

• Plot of Hit Rate vs. False Alarm Rate
• Shows trade-off between hits and false alarms
• Curve shape indicates sensitivity
• Position on curve indicates bias

Perfect Detector:

• ROC curve goes through top-left corner
• 100% hits, 0% false alarms
• Theoretical ideal

Chance Performance:

• Diagonal line from bottom-left to top-right
• Equal hits and false alarms
• No sensitivity

Measuring Sensitivity and Bias

d' (d-prime): Sensitivity

Definition:
The distance between signal and noise distribution means, measured in standard deviations.

Formula:

d' = Z(Hit Rate) - Z(False Alarm Rate)

Interpretation:

• d' = 0: No sensitivity (chance performance)
• d' = 1: Weak sensitivity
• d' = 2: Moderate sensitivity
• d' = 3: Strong sensitivity
• d' = 4: Very strong sensitivity

Properties:

• Unaffected by response bias
• Pure measure of discriminability
• Comparable across conditions

β (beta): Likelihood Ratio Bias

Definition:
The ratio of the heights of the signal and noise distributions at the criterion.

Formula:

β = f_SN(criterion) / f_N(criterion)

Interpretation:

• β = 1: Neutral bias
• β > 1: Conservative bias (prefer "no")
• beta < 1: Liberal bias (prefer "yes")

c: Criterion Location

Definition:
The distance of the criterion from the neutral point, in standard deviations.

Formula:

c = -0.5 × [Z(Hit Rate) + Z(False Alarm Rate)]

Interpretation:

• c = 0: Neutral criterion
• c > 0: Conservative criterion
• c < 0: Liberal criterion

Applications

Sensory Psychology

Threshold Measurement:

• More accurate than classical methods
• Separates sensitivity from bias
• Reveals true perceptual abilities

Sensory Systems:

• Vision: Contrast sensitivity
• Hearing: Tone detection in noise
• Touch: Pressure detection
• Taste and smell: Chemical detection

Memory Research

Recognition Memory:

• Separates memory strength from response bias
• "Old" vs. "New" decisions
• Remember vs. know judgments

Eyewitness Testimony:

• Line-up identifications
• Separating accuracy from willingness to choose
• Improving identification procedures

Medical Decision Making

Diagnostic Testing:

• Sensitivity and specificity of tests
• Setting diagnostic criteria
• Balancing false positives and negatives

Radiology:

• Tumor detection in X-rays
• Setting detection thresholds
• Training radiologists

Cognitive Psychology

Attention:

• Divided attention effects
• Vigilance and monitoring
• Search tasks

Implicit Memory:

• Subliminal perception
• Unconscious processing
• Priming effects

Real-World Applications

Security Screening:

• Airport security
• Baggage screening
• Balancing security and convenience

Quality Control:

• Manufacturing inspection
• Defect detection
• Setting inspection criteria

Weather Forecasting:

• Rain prediction
• Hurricane warnings
• Balancing false alarms and misses

Try the Experiment

Experience signal detection yourself:

Try the Signal Detection Theory Experiment →

You'll discover:

• Your own sensitivity (d')
• Your response bias
• How criterion placement affects outcomes
• The trade-off between hits and false alarms

Key Takeaways

• Two Processes: Sensitivity and bias are separable
• d' Measures Sensitivity: Pure measure of discriminability
• β or c Measures Bias: Tendency to say yes or no
• ROC Curves: Visualize sensitivity and bias trade-offs
• Wide Applications: From perception to memory to medical diagnosis
• More Accurate: Better than classical threshold methods
• Practical Value: Improves real-world decision-making

References

• Green, D. M., & Swets, J. A. (1966). Signal Detection Theory and Psychophysics. Wiley.