Generalizing parametric models by introducing trial-by-trial parameter variability: The case of TVA

We identify two biases in the traditional use of Bundesen’s Theory of Visual Attention (TVA) and show that they can be substantially reduced by introducing trial-by-trial variability in the model. We analyze whole and partial report data from a comprehensive empirical study with 347 participants and elaborate on Bayesian model selection theory for quantifying the advantage of trial-by-trial generalization in general. The analysis provides strong evidence of trial-by-trial variation in both the VSTM capacity parameter and perceptual threshold parameter of TVA. On average, the VSTM capacity bias was found to be at least half an item, while the perceptual threshold parameter was found to be underestimated by about 2 ms.