AX-CPT Measures Are Mostly Reliable and Share Variance, but Fall Short as Valid Measures of Attentional Control

Attentional control refers to the ability to maintain a goal and goal-relevant information in the face of distraction. Previous research has highlighted the lack of reliability and shared variance in attentional-control measures. This may result from the heterogeneity of the control processes induced by the tasks. The aim of the present study was to reduce or control for this heterogeneity in two ways. First, we used one task – the AX-CPT – with different stimulus materials (letters, words, dot patterns, matrices, and pictures). In this task, a target response is required in trials where two specific stimuli occur in sequence (e.g., the letter “A” followed by the letter “X”). In all other trials, a non-target response is required. Second, we disentangled proactive control (goal maintenance) from reactive control (goal reactivation) by computing separate indices for each process. Using data from 667 young adults, we obtained good reliability for most indices. Structural equation modeling identified models with factors representing shared variance across stimulus materials. In a subsequent analysis, we bypassed the indices by modeling proactive and reactive control directly from error rates and response times. Using a bifactor modeling approach, we found shared variance across stimulus materials for error rates only. However, in both analyses, the pattern of latent correlations contradicted the expectations. For example, a positive and very high correlation was observed between indices representing proactive and reactive control. Overall, these findings question the validity of AX-CPT measures, suggesting that the task cannot resolve the measurement problem in attentional control.