The Theory of Visual Attention (TVA) provides a well-defined stochastic framework for understanding how attention is allocated in different tasks. An implicit assumption of TVA is that the parameters describing attention remain related across tasks, at least if the conditions are comparable (i.e. stimuli share the same color, eccentricity, etc.). Under these assumptions, TVA offers an interesting tool for studying the reproducibility of attention effects. Here, we focus on the reproducibility of the mean total processing capacity from the Temporal Order Judgment task - a comparably new experimental paradigm to estimate TVA parameters - to the Whole Report task - a well-established standard task.
For the analysis, we developed a Bayesian Model that implements a "traditional-K"-like likelihood for the Whole Report task and the TVA/TOJ likelihood for the Temporal Order Judgment task. We decided on a "traditional-K"-like likelihood for the Whole Report task because it has a compact form that fits well in the Bayesian framework.
The posterior distributions of the mean processing capacity assessed via a well-powered within-subjects experiment overlap. However, we cannot conclude with sufficient confidence that they are equal as there is still much variance within the parameters, leaving differences of medium to large size possible. The talk will discuss different theoretical and modeling-related explanations for this result.