Beyond Face Value: Assessing the Factor Structure of an Eye-Tracking based Attention Bias Task

Abstract

Behavioral measurement of attention bias for emotional information has been hampered by poor psychometrics. While there is a precedent of researchers improving the measurement of questionnaires through assessment of the factor structure of the measure, this approach has not yet been applied to behavioral tasks. In the present study, we assess the psychometric properties of a free viewing, eye-tracking task designed to measure extended dwell time for emotional stimuli. We introduce a novel application of factor analysis to assess the unidimensionality of the tasks’ trials, a necessary requirement to derive mean bias scores. We also used thresholds across omega hierarchical, explained common variance, and percent uncontaminated correlations to evaluate whether the data could be considered unidimensional. Undergraduate students (N = 130) viewed two counterbalanced tasks of a 4 x 4 matrices of sad/neutral, or happy/neutral facial expressions for 10 seconds each across 60 trials. Dwell time for sad (alpha = .95, omega = .96) and neutral faces (alpha = .95, omega = .95) in the sad/neutral version showed excellent internal consistency. However, in contrast to several other traditional bias metrics, only dwell time for sad faces and neutral faces from the sad/neutral task met the three standards for interpreting the data as unidimensional. This study demonstrates the importance of assessing the psychometric properties of behavioral tasks’ outcomes, including the factor structure, as not all metrics are equivalent, even when derived from the same data.