Purchase this article with an account.
Chris Sims, Rachel Lerch; A new vocabulary for understanding limits on perception. Journal of Vision 2016;16(12):1435. doi: 10.1167/16.12.1435.
Download citation file:
© 2017 Association for Research in Vision and Ophthalmology.
Human perceptual processing, like any physical information channel, is by necessity capacity-limited. However, limits on channel capacity are only one factor that is relevant for understanding what makes for an 'optimal' or 'efficient' communication system. Perceptual systems exist not merely to transmit information, but rather to allow organisms to accomplish behavioral goals. Hence, an optimal perceptual system is one that minimizes the task-relevant costs of perceptual error (defined by a particular cost function) subject to a limit on capacity. The mathematical field of rate–distortion theory provides the theoretical tools necessary for studying this problem. Importantly, in this framework the optimality of a given perceptual system depends critically on three factors: (1) limits on capacity, (2) limits on statistical learning, and (3) the match between the implicit and explicit cost function for a given task. Research on visual perception and visual memory has focused almost exclusively on understanding capacity or resource limits, while the latter two constructs have been largely overlooked. Hence, rate–distortion theory provides a much richer vocabulary for understanding perceptual processing, and yields novel and unique predictions for human performance. We applied this framework to two benchmark domains in the study of perception: the discrete categorization of perceptual signals (also known as absolute identification) and perceptual working memory. In each case rate–distortion theory provides a quantitative fit to the data that is better than or comparable to the best published models in the literature. We argue that human perceptual performance is explained by three equally important factors, not just limits on capacity. Perceptual processing must also be understood in terms of the implicit cost function it seeks to minimize, as well as limits in implicit statistical learning.
Meeting abstract presented at VSS 2016
This PDF is available to Subscribers Only