Friday, May 9, 2008, 1:00 – 3:00 pm Royal Palm 5

Organizer: Denis G. Pelli (New York University)

Presenters: Patrick Cavanagh (Harvard University and LPP, Universit� Paris Descartes), Brad C. Motter (Veterans Affairs Medical Center and SUNY Upstate Medical University), Yury Petrov (Northeastern University), Joshua A. Solomon (City University, London), Katharine A. Tillman (New York University)

Symposium Description

Crowding is a breakdown of object recognition. It happens when the visual system inappropriately integrates features over too large an area, coming up with an indecipherable jumble instead an object. An explosion of new experiments exploit crowding to study object recognition by breaking it. The five speakers will review past work, providing a tutorial introduction to crowding, and will describe the latest experiments seeking to define the limits of crowding and object recognition. The general question, including �integration�, �binding�, �segmentation�, �grouping,� �contour integration�, and �selective attention�, is a burning issue for most members of VSS.


Crowding: When grouping goes wrong

Patrick Cavanagh

Early visual processes work busily to construct accurate representations of edges, colors and other features that appear within their receptive fields, dutifully posting their details across the retinotopic landscape of early cortices. Then the fat hand of attention makes a grab at a target and comes up with an indecipherable stew of everything in the region. Well, that�s one model of crowding. There are others. Whatever the model of crowding, it is clear that the phenomenon provides a rare window onto the mid-level process of feature integration. I will present results on nonretinotopic crowding and anticrowding that broaden the range of phenomena we include in the category of crowding.

Correlations between visual search and crowding

Brad C. Motter

Visual search through simple stimulus arrays can be described as a linear function of the angular separation between the target and surrounding items after scaling for cortical magnification. Maximum reading speeds as a function of eccentricity also appear to be bound by a cortical magnification factor. If crowding can explain these visual behaviors, what is the role of focal attention in these findings?

Locus of spatial attention determines inward-outward anisotropy in crowding

Yury Petrov

I show that the locus of spatial attention strongly affects crowding, inducing inward-outward anisotropy in some conditions, removing or reversing it in others. It appears that under normal viewing conditions attention is mislocalized outward of the target, which may explain stronger crowding by an outward mask.

Context-induced acuity loss for tilt: If it is not crowding, what is it?

Joshua A. Solomon and Michael J. Morgan

When other objects are nearby, it becomes more difficult to determine whether a particular object is tilted, for example, clockwise or anti-clockwise of vertical. “Crowding” is similar: when other letters are nearby, it becomes more difficult to determine the identity of a particular letter or whether it is, for example, upside down or mirror-reversed. There is one major difference between these two phenomena. The former occurs with big objects in the centre of the visual field; the latter does not. We call the former phenomenon “squishing.” Two mechanisms have been proposed to explain it: lateral inhibition and stochastic re-calibration. Simple models based on lateral inhibition cannot explain why nearby objects do not impair contrast discrimination as well as tilt acuity, but a new comparison of acuities measured with the Method of Single Stimuli and 2-Alternative Forced-Choice do not support models based on stochastic re-calibration. Lateral inhibition deserves re-consideration. Network simulations suggest that many neurones capable of contrast discrimination have little to contribute towards tilt identification and vice versa.

The uncrowded window for object recognition

Katharine A. Tillman and Denis G. Pelli

It has been known throughout history that we cannot see things that are too small. However, it is now emerging that vision is usually not limited by object size, but by spacing. The visual system recognizes an object by detecting and then combining its features. When objects are too close together, the visual system combines features from them all, producing a jumbled percept. This phenomenon is called crowding. Critical spacing is the smallest distance between objects that avoids crowding. We review the explosion of studies of crowding � in grating discrimination, letter and face recognition, visual search, and reading � to reveal a universal law, the Bouma law: Critical spacing is proportional to distance from fixation, depending only on where (not what) the object is. Observers can identify objects only in the uncrowded window within which object spacing exceeds critical spacing. The uncrowded window limits reading rate and explains why we can recognize a face only if we look directly at it. Visual demonstrations allow the audience to verify key experimental results.