Challenges to pooling models of crowding: Implications forvisual mechanisms

Journal of Vision

article

Ruth Rosenholtz¹  Dian Yu¹  Shaiyan Keshvari¹ 

[1] Computer Science and Artificial Intelligence Laboratory, Massachusetts Institute of Technology;Department of Brain & Cognitive Sciences, Massachusetts Institute of Technology

关键词: peripheral vision;    crowding;    poolingmechanism;    high-dimensional pooling models;   

DOI  :  10.1167/jov.19.7.15

来源: Association for Research in Vision and Ophthalmology

【 摘 要 】

A set of phenomena known as crowding revealperipheral vision’s vulnerability in the face of clutter.Crowding is important both because of its ubiquity,making it relevant for many real-world tasks and stimuli,and because of the window it provides ontomechanisms of visual processing. Here we focus onmodels of the underlying mechanisms. This reviewcenters on a popular class of models known as poolingmodels, as well as the phenomenology that appears tochallenge a pooling account. Using a candidatehigh-dimensional pooling model, we gain intuitionsabout whether a pooling model suffices and reexaminethe logic behind the pooling challenges. We show thatpooling mechanisms can yield substitution phenomenaand therefore predict better performance judging theproperties of a set versus a particular item. Poolingmodels can also exhibit some similarity effects withoutrequiring mechanisms that pool at multiple levels ofprocessing, and without constraining pooling to aparticular perceptual group. Moreover, we argue thatother similarity effects may in part be due tononcrowding influences like cuing. Unlikelow-dimensional straw-man pooling models,high-dimensional pooling preserves rich informationabout the stimulus, which may be sufficient to supporthigh-level processing. To gain insights into theimplications for pooling mechanisms, one needs acandidate high-dimensional pooling model and cannotrely on intuitions from low-dimensional models.Furthermore, to uncover the mechanisms of crowding,experiments need to separate encoding from decisioneffects. While future work must quantitatively examineall of the challenges to a high-dimensional poolingaccount, insights from a candidate model allow us toconclude that a high-dimensional pooling mechanismremains viable as a model of the loss of informationleading to crowding.

【 授权许可】

CC BY|CC BY-NC-ND