A. Psychology as the study of ecological information.

The traditional view of visual perception, dating to Descartes, is that sensory impressions or sensations, accessible to conscious awareness (perhaps with suitable training), are the building blocks of perception of the wholistic objects. Our perceptions of objects like chairs, tables and people are constructed from an inventory of sensory atoms. The emphasis is on atoms provided at the interface between the mind and the environment.

Instead, Gibson sought a biological framework for thinking about vision. Vision (like the other sensory systems) is a capability that serves a particular function - to support animal survival by permitting improved information about what is going on out there in the world. So we should expect the eye (and ear, etc) to have evolved in order to exploit useful `information' (in some biological sense). That `information' (defined in some practical `embodied' way quite unrelated to Shannon's disembodied, context-independent concept also called `information') is what Gibson says we should be seeking in a theory of psychology. Psychology should provide an ecological theory of what the nervous system is about. This is what it means to take our animal nature seriously and to treat us as intrinsically situated in a specific environment and embodied in a particular body.

So the discovery that the eye is like a camera may have historically encouraged a focus on this `screen' and on its discreteness in time (eg, from saccades and blinks) and in space (eg, cell-based `pixels'). Gibson says we should look at where the info is, rather than at what the channels are like that the info comes in on. Information about what is happening in the environment is to be found in the OPTIC ARRAY - that pattern of light approaching the animal. It is the result of ambient light (bouncing between objects on the earth's surface). At any point (in ambiently illuminated locations), there is light from all directions approaching that point. If there is an eye there, it can collect and interpret the ambient light for information about locations where the light came from. To understand visual perception, we should first discover what information is available that is useful. Until we understand this, we cannot begin to understand the actual realtime mechanisms of neural cognition, says JJ.

So, Gibson sought a theory of `potential knowledge', things about stimulus patterns in light that, IF known, would be useful for understanding the layout of surfaces (which would be useful for those who move around on the surface of the planet). ``Information is simply there, '' says JJ, and is neither transmitted nor stored (1979, p. 243). Perception does not resemble a communication process in the slightest. He proposed that it is the `higher order invariants', eg, ratios of several direct (iconic) measurements, that are ``picked up'' since they are the more informative properties, patterns or invariants of stimulation.

B. Some consequences of this point of view:

1. Animal eyes are sensitive to just the range of electromagnetic radiation that happens to bounce off the surfaces that our body can't pass through. Given particular problems specific to terrestrial navigation, this light (when available) is informative about remote places, things and events.

2. For certain specific purposes, like using the direction of the sun for navigation, the radiomagnetic frequencies that pass through the clouds are especially useful. So its not surprising bees and other navigating animals have evolved sensitivity to these frequencies. Some nocturnal animals (like snakes) have evolved sensitivity to infrared (however infrared is not so easily refracted with a lens, so the snake's infrared sensors do not provide such good direction information as visible light in the snake's eye).

3. The design of the eye blocks light from most directions in order to use a lens that collects light from a narrow angle so that the optic array in one general direction can be explored with both eyes. By collecting light through a tiny point, more detail can be accurately obtained.

4. Gibson proposes that humans have Five Perceptual Systems (based on 1966, p. 50):

1. Basic Orienting (bodily orientation): using the vestibular organ - based on gravity and inertial effects on otoliths in the inner ear.
2. Auditory System (listening): using the external ear and internal organs in a moving head and body -- depends on travelling vibrations in air or bone.
3. Haptic System (touching): using pressure sensors in skin, joints and muscles --based on tissue deformation.
4. Taste-Smell System (smelling and tasting) - using chemo-receptors in nose and tongue -- based on volatility (for smell) and the nutritive and biochemical values of solutions.
5. Visual System (looking) - using photoreceptors for ambient light in moving eye, head and body.

Note how these differ from the traditional `Five Senses'!

5. The notion of a `stimulus' (as in `the thing that causes a Response'), from this perspective, is revealed to be a cagey notion. It is used variously (and typically quite inconsistently) for either a remote `object' (eg, something perceived and identified by some unknown or unspecified mechanism) or for something more like a primitive `physical goad' (whose iconic distinctness and causal properties are obvious). This was one basis of Chomsky's attack on Skinner too. But Gibson replaces countable `stimuli' with `stimulation' and accepts responsibility to show just exactly what it is that could be usefully extracted from the patterns presented on sensory surfaces. Of course, this generally implies a mathematical formulation of the invariant. It may perhaps be as simple as a ratio of two lengths, but is sometimes mathematically quite complex.

C. Some examples of invariant properties in ecological optics.

1. As objects move relative to their background, they wipe out the background texture at leading edges and unwipe texture at trailing edges.

2. Textures exhibit a gradient from larger to smaller size that correlates negatively with distance.

3. HORIZON RATIO invariant: the ratio of the part below to part above the horizon is constant no matter what the distance.

4. OPTIC FLOW is the way visual texture elements move smoothly across the visual field when an animal is moving, except for the singularity toward which one's motion is directed. They expand away from the singularity. As texture elements get closer, they move faster across the visual field. They collapse toward another singularity behind the moving animal. The OPTOMOTOR RESPONSE is the tendency of insects to move in a way that keeps optic flow minimized - thereby assuring that they move in a consistent way despite buffeting by air currents, etc.

5. TIME-TO CONTACT (tau) invariant - despite variability in rate and distance. David Lee discovered this invariant while studying gannets, a North Atlantic diving bird. Tau = distance/velocity. Or it is a length (perpendicular to direction of motion) divided by the rate of change in that length. This number specifies the number of seconds til contact if locomotion rate is kept constant. The surprise is that it can be `computed' from very local information (on a map of the visual field) at any distance and without even estimating the velocity of forward motion using minimal assumptions (eg, that motion continues at a constant velocity).

NOTE: For further reading about Gibson's ideas, look at the papers by Bingham and by Turvey and Carello in our text, `Mind as Motion.' Both are classically Gibsonian in conception. And both are clearly dynamical models.

Robert Port, 1999