Experimental control of conflict in a predictive visual probe task: Highly reliable bias scores related to anxiety

Concerns have been raised about the low reliability of measurements of spatial attentional bias via RT differences in dot-probe tasks. The anticipatory form of the bias, directed towards predicted future stimuli, appears to have relatively good reliability, reaching around 0.70. However, studies thus far have not attempted to experimentally control task-related influence on bias, which could further improve reliability. Evoking top-down versus bottom-up conflict may furthermore reveal associations with individual differences related to mental health. In the current study, a sample of 143 participants performed a predictive Visual Probe Task (predVPT) with angry and neutral face stimuli online. In this task, an automatic bias is induced via visually neutral cues that predict the location of an upcoming angry face. A task-relevant bias was induced via blockwise shifts in the likely location of target stimuli. The bias score resulting from these factors was calculated as RTs to target stimuli at locations of predicted but not actually presented angry versus neutral faces. Correlations were tested with anxiety, depression, self-esteem and aggression scales. An overall bias towards threat was found with a split-half reliability of 0.90, and 0.89 after outlier removal. Avoidance of threat in blocks with a task-relevant bias away from threat was correlated with anxiety, with correction for multiple testing. The same relationship was nominally significant for depression and low self-esteem. In conclusion, we showed high reliability of spatial attentional bias that was related to anxiety.