Testing Evidence Accrual Models by Manipulating Stimulus Onset.

Area of Science:

• Cognitive Psychology
• Psychophysics
• Computational Neuroscience

Background:

• Response time models often assume evidence accrual over time.
• Stimulus presentation can vary in temporal dynamics (e.g., stepped vs. ramped).
• Dominance describes a temporal relationship where one stimulus precedes another.

Purpose of the Study:

• To formalize and test evidence accrual models using stimulus dominance.
• To investigate whether dominating stimuli consistently yield faster and more accurate responses.
• To identify conditions under which evidence accrual models may fail.

Main Methods:

• Formalization of the class of evidence accrual models.
• Introduction of the concept of stimulus dominance (ramped vs. stepped stimuli).
• Extensive computer simulations of random walk and accumulator models.

Main Results:

• Evidence accrual models generally predict more accurate responses to dominating stimuli.
• Anomalous cases show slower mean correct response latencies to dominating stimuli.
• These anomalies occur with highly asymmetric response criteria and extreme response biases.

Conclusions:

• Stimulus time course manipulation via dominance offers empirical tests for evidence accrual models.
• Standard evidence accrual models predict improved accuracy and speed with dominating stimuli under normal conditions.
• Extreme response biases are critical for observing counterintuitive latency predictions in accumulator models.