Internal reliability of measures of substance-related cognitive bias

Aims There is growing interest in cognitive biases related to substance use, but evidence from the anxiety literature suggests that tasks commonly used to assess these may suffer from low internal reliability. We examined the internal reliability of the visual probe and modified Stroop tasks. Design Secondary analysis of visual probe and modified Stroop task data collected across seven independent studies. Setting Human laboratory study. Participants Healthy volunteers (n = 408 across seven independent studies) recruited from the general population on the basis of alcohol or tobacco use. Measurements Visual probe and modified Stroop task measures of substance-related cognitive bias. Findings Measures of cognitive bias for substance-related cues, as assayed by the visual probe and the modified Stroop tasks, may not be reliable. In particular, the visual probe task showed poor internal reliability, as did unblocked versions of the modified Stroop task. Conclusions The modified Stroop task is preferable to the visual probe task as a measure of substance-related cognitive bias, on the basis of its psychometric properties. Studies using cognitive bias tasks should not assume they are reliable, and should routinely report reliability estimates where possible.