Recognition receiver operating characteristic asymmetry: Increased noise or information?

Area of Science:

• Cognitive Psychology
• Neuroscience
• Psychophysics

Background:

• Recognition memory Receiver Operating Characteristic (ROC) curves typically show asymmetry.
• This asymmetry has been explained by two competing models: unequal variance signal detection (uvsd) and dual process signal detection (dpsd).

Purpose of the Study:

• To test the underlying assumptions of the uvsd and dpsd models regarding ROC asymmetry.
• To determine whether ROC asymmetry is caused by noisier evidence or more informative evidence.

Main Methods:

• Fitting the uvsd and dpsd models to old/new recognition data.
• Using model evidence parameters to predict performance on a novelty, three-alternative forced-choice (N3AFC) recognition task.
• Conducting simulations to analyze model predictions and mispredictions.

Main Results:

• The dpsd model's predictions were supported by experimental data.
• The uvsd model failed to predict observed patterns, showing mispredictions.
• Simulations revealed that the dpsd model could predict the uvsd model's errors.

Conclusions:

• ROC asymmetry in recognition memory is driven by increasingly informative target evidence, not noisier evidence.
• The findings invalidate the uvsd model, highlighting the construct validity of the dpsd model.
• This research clarifies the mechanisms underlying recognition memory asymmetry.