A Two-Stage Process Model of Sensory Discrimination: An Alternative to Drift-Diffusion

Area of Science:

• Cognitive Neuroscience
• Decision Science
• Psychophysics

Background:

• Sensory decision-making under uncertainty is often explained by the drift-diffusion model, which posits slow accumulation of noisy evidence to a threshold.
• This model predicts multiplicative interactions between signal strength/value and reaction time, and independence of decision time from evidence strength when the process is interrupted.

Purpose of the Study:

• To test the predictions of the standard drift-diffusion model for sensory discrimination under uncertainty.
• To develop and validate an alternative model that accurately captures human behavioral data in these tasks.

Main Methods:

• Two experiments were conducted with human participants performing a sensory discrimination task involving noisy stimuli.
• Behavioral data on reaction times were collected and analyzed.
• A novel 'estimate-then-decide' model was formulated and compared against the drift-diffusion model.

Main Results:

• Neither of the two key predictions of the drift-diffusion model held true for the experimental data.
• Reaction times did not show the predicted multiplicative interaction between signal strength/value.
• Decision time after an interruption was not independent of the accumulated sensory evidence strength.

Conclusions:

• The standard drift-diffusion model is insufficient to explain human sensory discrimination under uncertainty.
• The proposed 'estimate-then-decide' model provides a better account of the observed behavioral data.
• This new model suggests a two-stage process involving evidence estimation followed by decision-making based on that estimate.