The format of abstract perception: causes and consequences

Abstract: How does perception represent abstraction? How does the mind coalesce across domains? Why is learning sometimes easy and sometimes difficult? In this talk, I propose that the answer to all of these questions can be gleaned from considering the representational format of mid-level perception, and especially number perception. I show that number perception is experience-independent, bipartite, and probabilistic. These facts carry consequences for the rest of the mind. For example, I show that (1) number perception is not derived from experience with mid-level perceptual objects (contrary to recent machine learning models suggesting otherwise); (2) that number perception readily integrates with other domains where their formats match (e.g., number and area perception) but not where they mismatch (e.g., number and long-term memory); and (3) that the bipartite format of number perception allows for radical adjustments in what counts as a unit of enumeration, allowing even children to showcase perceptual division and multiplication before exposure to symbolic mathematics. The study of representational formats, therefore, can tell us a lot about how facets of the mind sometimes quickly come together, and when they do not, and can help answer long-standing questions in cognitive science about nativism, modularity, and learning.