Predicting the timing of wrong decisions with LATER.

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Decision-Making Models

Background:

• Response time (latency) is a key metric for understanding neural decision processes.
• Existing models like LATER (Linear Approach to Threshold with Ergodic Rate) have successfully predicted response latencies but not errors.
• A comprehensive decision model must account for both correct and incorrect response distributions.

Purpose of the Study:

• To evaluate the LATER model's ability to predict errors in decision-making tasks.
• To investigate how response latency distributions differ for correct and incorrect responses.

Main Methods:

• Utilized a decision task involving visual targets of specific colors, designed to elicit a high error rate.
• Applied a modified LATER model, incorporating LATER units, to analyze latency distributions and response proportions.
• Compared model predictions against empirical data from human subjects performing the task.

Main Results:

• Initially, faster responses were equally likely to be correct or incorrect.
• Latency distributions for correct and incorrect responses diverged over time, with errors becoming less frequent.
• The LATER model accurately predicted both latency distributions and the overall proportion of correct and incorrect responses.

Conclusions:

• The LATER model, when applied to a task generating errors, successfully accounts for both correct and incorrect response latencies.
• Color information, delayed in arrival, appears to influence response cancellation and initiation, which the LATER model captures.
• This study validates the LATER model as a robust tool for analyzing neural decision-making, including error patterns.