Assessing model-based inferences in decision making with single-trial response time decomposition.

Area of Science:

• Cognitive psychology
• Computational neuroscience
• Psychophysics

Background:

• Quantitative models, particularly the drift-diffusion model (DDM), are widely used to study decision-making mechanisms.
• Model fit quality is often used to infer validity, but this can be insufficient for interpreting parameters.

Purpose of the Study:

• To formally assess the validity of the drift-diffusion model (DDM) using a novel cross-validation approach.
• To investigate the interpretability of DDM parameters by linking them to directly measurable physiological processes.

Main Methods:

• Recorded electromyographic (EMG) activity alongside response times (RTs) during perceptual decision tasks.
• Decomposed RTs into premotor time (PMT) and motor time (MT) to isolate motor processes.
• Manipulated stimulus strength, speed-accuracy trade-off, and response force to assess their effects on RT, PMT, and MT.

Main Results:

• All manipulated factors consistently affected motor time (MT), a direct measure of motor processes.
• The DDM's nondecision parameter generally reflected MT effects, except during the fastest responses.
• Observed discrepancies highlight limitations in the interpretability of DDM parameters under certain conditions.

Conclusions:

• The study provides new insights into the validity and interpretability of the drift-diffusion model (DDM).
• Direct measurement of muscular activity offers a valuable method for cross-validating cognitive models.
• Findings delineate the boundaries of DDM parameter interpretability, particularly concerning nondecision processes.