The Scope and Flexibility of Statistical Summary Representations

One of the most fundamental questions in cognitive science is how the mind copes with the massive amount of perceptual information in the environment, giving rise to our seeming high-fidelity perception of the world. The answer to this question is surely complicated, but at a basic level seems to involve a combination of both law-resolution snapshots of large portions of the environment (which I'll term Statistical Summary Representations - SSRs) and high-resolution representations of only highly constrained areas or objects (obtained through selective attention). A wealth of research over the past 50 years has been devoted to understanding how certain regions or objects are selected for further processing and whether this selection is driven by internal or external factors. Comparatively little research to date has focused understanding the nature and feasibility of SSRs. This dissertation focuses on understanding the scope and flexibility of SSRs and specifically whether this process is well suited to different features of our real-world perception. The following experiments, when read as a whole, serve as case studies aimed at answering this question by showing that SSRs: (1) are computed over continuously changing stimuli; (2) operate just as effectively over auditory and visual stimuli; and (3) can operate efficiently over heterogeneous shape input. Taken together, these studies all support the assertion that SSRs are in fact a fundamental aspect of the way in which our minds overcome this capacity limit.