Perception and Visual Cognition

Peter Bex

Department of Psychology

Approaches to Visual Perception

Ecological - e.g. Gibson (1950)

Constructionist - e.g. Hochberg (1970)

Computational - e.g. Marr (1980)

Ecological View

Founded by Gibson (1904-1979)

Primary goal of perception is to support actions and identify important events, based on the information available in spatial and temporal changes in the optic array.

Ecological View

Direct theory of perception without mediation by complex cognitive processes. Texture alone can reveal surface shape and slant:

Ecological View

Depth can be perceived directly from the amount of texture an object occupies, without complicated computations based on retinal size and prior knowledge:

Ecological View

Objects have affordances that may be perceived directly e.g. an apple is ÒeatableÓ, a chair affords ÒsittingÓ, a picture of an apple affords only inspection, never grasping or eating.

Affordance is identified from resonance, not from matching with memory traces.

Traditional experiments and visual illusions tell us nothing about Perception, only how observers cope with artificially impoverished inputs.

Constructionist View

What is important is not how perceptions support action, but how perceptions support understanding

19th and early 20th century Structuralists proposed that perceptions are the sum of many elementary sensations

Gestaltists and Constructionists instead argue that representations of reality are assembled from fragments of sensory information - the whole percept is often greater than the sum of the parts

Muller-Lyer Illusion

from: Gregory, R. L. (1973) Eye and brain (2nd ed.). New York, McGraw-Hill

Ponzo or Railway Illusion

Impossible Objects

Constructionist View

Perceptions are influenced by experience, knowledge and expectations, this distinguishes them from Gestaltists

Ames Room

Computational View

Describe computations within the visual system that convert sensory stimulation into representations of the world.

The main job of vision is the representation of shape by encoding edges, corners, texture etc.

Computational Model

Edges are located at zero crossings in the 2nd spatial derivative of an image:

Zero crossings in the 2nd spatial derivative of an image:

Physiological substrates for encoding edges

Comparison of predicted responses of on- and off-centre X cells with electrophysiological recordings

from: Marr, D. (1982) Vision. New York, Freeman