UC Irvine

Performance statistics in centroid tasks are not the same as those used in classic decision tasks. In psychophysical experiments using decision tasks, signal detection theory and drift-diffusion models provide the frameworks for statistical inference from error rates and reaction times. However, neither of these frameworks are appropriate for psychophysical inference with centroid task data. In this dissertation, we explore a modeling framework for double-pass experiments with centroid tasks, and show its potential to (1) detect performance differences, and infer experimental effects without additional process model assumptions, (2) falsify properties of a latent process using nested model assumptions, (3) investigate neurocomputational models of the process, and (4) investigate properties of spatial attention at a deeper level than is possible using decision-based paradigms.