Shaul Hochstein¹, Noam Khayat¹, Marina Pavlovskaya¹, Haya Abtan¹, Daniella Koyfman¹, Merav Ahissar¹; ¹Hebrew University, Jerusalem

Since Dan Ariely (2002) reignited interest in ensemble perception, also known as scene summary statistics perception, hundreds of papers have been published on this subject. It is now well established that ensemble perception includes low and high level parameters, objects, and scenes, and that observers perceive ensemble statistics including average, range, skewness, and distribution. Here we present new findings concerning the broad extent, neural computational process, cerebral site(s), and function of ensemble perception. Surprisingly, we found that ensemble means are represented in mid-level cortical areas specializing in object perception, including centrally the Lateral Occipital Complex, leading to the suggestion that ensembles are perceived as single objects rather than as groups of objects: a forest is not a group of trees, a traffic jam is not a bunch of cars, and a VSS conference is not a bunch of scientists, lectures and posters. Rather, they are, indeed, a forest, a traffic jam, and a VSS conference. Considering the very limited capacity of short term working memory, such aggregation and redefinition is essential for perception, computation, and memory. Yet, the process of aggregating ensembles without first perceiving each individual needs clarification. I shall show that direct population encoding solves the issue elegantly and simply. In summary, we now have a better handle on answering the three essential questions regarding ensemble perception: Where in the visual brain are these summary statistics computed; why is this important for perception and memory; and, how is this computation performed.

Acknowledgements: grant from ISF Israel Science Foundation