Test-retest reliability of summary measures derived from the dot-probe task

Abstract

The concept of attentional bias to threat is important in studying clinical anxiety, and the dot-probe task is the most commonly used paradigm used to examine this attentional bias. However, the task has faced inconsistent and unsatisfactory findings in its psychometric properties. Specifically, studies have found poor test-retest reliability of the task, especially in non-clinical samples. Although it is important to understand attentional biases in people with anxiety, it is also of significance to understand how the concept works in normal populations. Since then, many have proposed approaches to increase the reliability of the task, but inconclusive results remain. Conflict drift-diffusion models have been proposed to model the dynamics of the cognitive process in two-choice conflict tasks based on reaction time distributions and accuracy data. This study implements the diffusion model of conflict to study whether such computational methods can help significantly increase the test-retest reliability of the task. 33 participants were invited to attend two data-collection sessions and complete the same dot-probe task at a one-week interval. Results indicate that the bias index calculations and parameters derived from the model all had unacceptably low reliability and they had little relation with self-report social anxiety scores. No vigilance to threat was found in this non-clinical sample, and a negative bias index in both sessions indicates a slight avoidance of threat. The findings confirm that the dot-probe task, with or without the diffusion model added, is not a suitable sole measure of attentional bias to threat within non-clinical samples.