Using reaction times and binary responses to estimate psychophysical performance: an information theoretic analysis

Area of Science:

• Psychophysics
• Cognitive Neuroscience
• Information Theory

Background:

• Psychometric functions describe accuracy, while chronometric functions describe reaction times (RTs) in response to stimuli.
• Traditionally, RTs are overlooked in psychophysical parameter estimation, despite potentially containing significant Shannon information.
• Existing diffusion models do not fully incorporate RT data or lapse rates.

Purpose of the Study:

• To extend the proportional-rate diffusion model (PRD) to incorporate RTs and other parameters.
• To estimate psychophysical parameters using both binary responses and RTs.
• To quantify the Shannon information gained by observers about stimulus strength.

Main Methods:

• Extended the PRD model (EPRD) to fit RTs to an inverse Gaussian distribution and include lapse rates and point-of-subjective-equality (PSE) parameters.
• Utilized a two-alternative forced choice (2AFC) design.
• Employed maximum likelihood estimation to fit binary responses and RT data separately and in combination to the EPRD model.

Main Results:

• Parameter estimates from binary responses alone were comparable to those from RTs alone, supporting the diffusion model's validity.
• The EPRD model successfully estimated mutual information between responses/RTs and stimulus strength.
• Human observers gained an average of 2.68 to 3.55 bits/trial of Shannon information.

Conclusions:

• RTs provide valuable information for psychophysical parameter estimation, complementing accuracy data.
• The extended diffusion model offers a more comprehensive approach to understanding sensory information processing.
• The study quantifies the information observers gain, introducing the 'Shannon increment' as a measure of detection capability.