January 1969
A Reconsideration of Eye-Movements and Eye Postures Based on Ecological Optics
J. J. Gibson, Cornell University
The World Wide Web distribution of James Gibson’s “Purple Perils” is for scholarly use with the understanding that Gibson did not intend them for publication. References to these essays must cite them explicitly as unpublished manuscripts. Copies may be circulated if this statement is included on each copy.
The movements of the eyes have been recorded, measured, and classified. Sixty-five years ago, Dodge listed five types for the human eye which are still accepted: the saccadic movements, the pursuit movements, the convergences, the compensatory movements for active head turning, and those for passive head-turning. These were movements of the eyes relative to the head, caused by contractions of the six pairs of extraocular muscles. They were, and still are, measured relative to the head. They were taken to be either responses to stimuli, in accordance with the S-R formula, or the results of motor commands issued from the brain. Finally, they were thought to be a kind of behavior, analogous to the movements of the hands or other bodily extremities. A recent account of eye movements from this traditional point of view is given by Howard and Templeton (1966) Ch. 3.
It was admitted of course that non-movements of the eyes occurred as well as movements of the eyes. But this raises the point that a movement of the eyes relative to the head may be a fixation of the eyes relative to the world. The relativity of movement to the frame of reference, and the importance of the appropriate frame of reference were not emphasized. The movements of the eyes and the postures of the eyes between movements are clearly inseparable. Moreover the primary frame of reference for eye-movements is the environment of the observer, not his head. More exactly, eye movements should be conceived relative to the ambient optic array surrounding an observer, i.e., with respect to ecological optics. The failure to recognize that there is a steady state of light itself outside the observer’s body is a source of confusion in the study of vision.
The concept of an ambient optic array at every possible point of observation suggests the hypothesis that eye-movements, head-movements, and body-movements all fit together in an orienting system, and that the body, the head, and the eyes can all be oriented to the structure of the array (Gibson, 1966, Ch. 4). On this theory the observer does not have to feel the position of his eyes relative to his head, and of his head relative to his body, and of his body relative to the ground in order to “know” where his eyes are pointing. Once the concept of the ambient array is substituted for that of a chaotic flux of retinal images, the old puzzle of the “position sense” of the eye disappears.
The concept of the ambient array makes it understandable how the eye-movements can be adjustments of a system, the ocular system, to a fact of the environment. They are better understood as adjustmentsthan they are as reflexes or voluntary acts. The accommodation of the lens, the closing or opening of the pupil, and the adaptation of the retina are obvious cases of cybernetic adjustment rather than of responses to stimuli. The maximizing of the “detail” of the retinal image, the modulating of the luminous intensity of the image, and the adapting of the retina to the intensity level of the image are cases in which some optimal or equilibrium state is reached. These adjustments of the oculomotor system and the ocular organ do not, however, fit the usual formula for “homeostasis” since they serve the requirements of clear vision, not of behavior; of perception, not of bodily need.
Let us reconsider and sort out the types of eye movements and eye postures in terms of functional adjustments. Tentatively, there seem to be four levels of the adjustment of a visual system to the ambient optic array, (a) the stabilizing of the eyes to the array, (b) the successive sampling of the whole array by non-panoramic eyes, (c) the successive fixating of items in the array for foveated eyes, and (d) the binocular fixating of corresponding items in two arrays by conjugated eyes.
A. The Stabilizing of the Eyes to the Ambient Optic Array
A basic fact about vision, all vertebrate vision in fact, is that the eyes do not wander or drift relative to the environment. More exactly, they do not wander or drift when an optic array anchored to the environment is present. They tend to drift in the presence of homogeneous ambient darkness and (the evidence suggests) in the presence of the homogeneous ambient light provided by sky, dense fog, or a “Ganzfeld.” So the stabilizing of the eyes to the environment is actually an anchoring to the optical projections from the environment converging to the point of observation.
This stabilization of the eyes is not the same thing as fixation of the eyes. It is more basic, for stabilization occurs in animals with no tendency toward foveation, as in some fish. It is simply a prerequisite for vision since, if an eye moved continuously and aimlessly its retina could not register the array. Stabilization is an adjustment that maintains or keeps constant the animal’s orientation to its environment, but this is a special kind of orientation that occurs not by way of gravity, not by way of mechanical contact, but strictly by way of steady-state illumination. This adjustment is shown in pure form by experiments on the optokinetic reactions of animals, or on optokinetic systagmus. When the ambient array of a stationary animal is unnaturally made to rotate around his (as when a textured cylinder surrounding the animal is turned on an axis passing through the point of observation) the eyes track the array. Sometimes the eyes-and-head do so, or even the eyes-and-head-and-body. The phenomenal experience of the human observer in this situation should be the illusion of being rotated.
The stabilization of the eyes is facilitated and assured by compensatory eye-movements. Head-turning is compensated for by eye-turning so that the eyes tend to be fairly stable even in the absence of an ambient array. As Dodge knew, there are two parallel causes of compensatory eye turning, coordinate compensation and vestibular compensation.
Coordinate compensation for active head-turning is an expression of the fact of reciprocal innervation of the neck-muscles and the eye-muscles. The motor hookup is such that when the head goes one way the eyes go in the other. This eye-movement is in no sense a response to a stimulus.
Vestibular compensation for passive head-turning is an automatic counterrotation of the eyes caused by the input from the cupulae of the semicircular canals, in isolation. This is a famous reflex. It can only be elicited by imposing a rotation on the observer’s head and body, as in a rotating chair (or by artificially stimulating the canals in other ways). And it is observed in pure form only when the observer is deprived of his optic array by darkness or a blindfold or by diffusing eye-caps. Pure vestibular nystagmus of this sort, being a response to angular acceleration, can be manifested as after nystagmus when passive rotation has been experimentally prolonged. In this case the eyes are rendered unstable relative to the ambient array, and the result is vertigo and distortion. But normal vestibular nystagmus is simply a component of ocular stabilization.
B. The Successive Sampling of the Array by the Head and Eyes
Ecological optics explains how it is that an animal has an optical environment as well as a mechanically solid environment, a vibratory environment, and a volatile environment. It surrounds him on all sides, which is to say that the optic array is ambient.
The fact is that no species of animal has a completely panoramic ocular system and no animal can apprehend the whole ambient array at once. The animal must therefore sample it over time in order to perceive the whole environment. The degree to which successive sampling is required instead of simultaneous grasping depends on the solid-angular size of the sample accepted by the system (the scope of the “field of view,” with both eyes). In primates, where the orbits of the eyes have migrated to the front of the head, the sample is only about a hemisphere (a discussion of the field of view of a head, an eye, and a fovea is found in Gibson, 1966, Ch. 12, p. 257 ff.).
It is important to realize that the sampling of the ambient optic array is not the same thing as the fixation of details in a sample (a field of view). Sampling is carried out by movements of the head (the eyes being stabilized most of the time). The need for looking around arises whether or not an animal has eyes with concentrated foveas. Even in the fish, the angular sample accepted by each “window” in the head must be swept over the spherical array if it is to perceive the environment.
Sample-taking in visual perception does not have to be treated as a mental process. Each successive sample has a large overlap with the preceding sample and there is a great deal of common structure during such a sequence. In this respect the sample-taking when looking around at one point of observation is like the sample-taking during locomotion at successive points of observation. In both cases the perceiver seems to be aware of the scene, not of the sequence of samples.
C. The Successive Fixation of Details of the Array by Foveated Eyes
We now come to the consideration of ocular adjustments which depend on a concentration of the resources of the eye in a macula or fovea, instead of a dispersing of them over the whole retina. A foveated eye has to fixate on one detail of the optic array at a time. The advantage is that it can register each detail much more acutely than a non-foveated eye could. Successive fixating is called scanning. Fixating is an adjustment that centers the fovea on a significant detail and maintains the centering long enough for the information to be registered. Actually the narrowing down of the fovea may progress from a large item in the array to smaller and smaller parts of it. The adjustment is not a single response; it is more like “hunting” or “homing.” And the posture of the fixated eye is not perfectly steady; it vibrates in accordance with the tremor of the eye-muscles.
Scanning is the complement of fixating, the shifting of the eye’s posture as distinguished from the maintaining of its posture. The shift of the eye from one fixation to another is called a saccadic movement. It is very rapid, beginning and ending within a few hundredths of a second. Hence, even with frequent saccadic movements, the proportion of time spent by the visual system in fixating is very high relative to the time spent in shifting. The saccadic eye-movement has the same form as the “fast phase” of nystagmus.
There are several types of fixation and several types of scanning. Consider first fixation.
1. Ordinary fixation with a stationary observer. In this case the optic array is “frozen,” that is, unchanging in time and the items or details of the array fixated can be thought of as either “forms” or as the “features” of forms. This is the familiar kind of ocular fixation which has been experimentally recorded with a picture or a page of print.
2. Fixation with a moving observer. In this case the ambient optic array undergoes transformation, both as a whole and in every part, and also undergoes the accretion and deletion of details that accompany changing occlusion, that is, the concealing and revealing of things at edges. Presumably the items fixated are “motions” or “transformations” or “transitions” in the array as well as “forms” in the array. But a theory of fixation is underdeveloped for this case, since it has been so little studied.
3. Pursuit fixation with a moving object. In this pure case the ambient array is unchanging except for a superimposed figure that moves over it. If the eyes fixate the figure instead of the background there occurs a “pursuit movement” in Dodge’s classification. Essentially, however, it is a case of fixation. In this case, as in the others, fixation is not a resting state of the eye-muscles or of the eye-head-body system; it requires continual corrections (microsaccades) and must be actively maintained, as any bodily posture must be.
Next consider types of scanning, the shifts of fixation by means of saccadic eye-movements. The verb “to scan” means simply to look at parts in succession.
1. Random scanning. Theoretically, the significant items and details of an optic array might be scattered evenly over the array. The shifts of fixation might then be jumps of random amounts in random directions.
2. Free scanning. Actually the information-bearing arrays from natural environments have a biased distribution of information, and regularities of structure. An array of this sort can be explored in some regular way. However the sequence of eye-movements makes no difference for the perception of the environment. For most situations (including pictures) the order in which the parts of the situation are looked at is never twice the same, and the perceiver is not even aware that he has made a series of fixations. Free scanning is to a considerable degree “zigzag scanning.”
3. Compulsory scanning. The special optic array from written text has parts that must be looked at in an arbitrary order (e.g., left to right and top to bottom). If the array is not scanned in this order, it is not intelligible. Its parts are of a radically different sort from those of an ecological array, or a picture, or a map. This kind of scanning, being entailed in reading, has received much study, but the learning of it must depend on the prior development of natural zigzag scanning.
If the above description is correct, a saccadic eye-movement is typically not a response to a “stimulus” falling on the periphery of the retina that brings the fovea to bear on it. A “stimulus” is a spot of light. But an optic array is not a collection of spots of light. And the exploration of an optic array is not a chain of responses to spots of light.
D. The Fixation of Both Eyes on the Same Item at the Same Time by Conjugated Eyes
When the eyes have concentrated foveas and are also in the front of the head so that they admit overlapping samples of slightly different ambient arrays, the stage is set for a very complicated kind of adjustment. This involves convergent fixation and conjugated scanning.
In terms of ecological optics, convergence means the effort of the binocular system to eliminate disparity of structure at the foveal center of the two samples of the two arrays (see Fig. 9.14, p. 179 in Gibson, 1966). The eliminating of disparity at the center creates gradients of disparity outward from the center. And disparity is information for the slants and the occluding edges of the surface-layout. If the binocular system can register disparities, it can detect layout. (Nothing whatever is said in this formula, about the “fusion” of “images.”).
Along with convergent fixation there has to go a yoking together of the saccadic movements between fixations, called conjugation. In man (and probably other primates) convergent fixation is not optional but automatic, and conjugated saccadic eye-movements have become compulsory. The eyes have lost the power to fixate independently and to explore independently, a power that they possess in the case of many mammals. The advantage gained is that the binocular system can register disparity. It can register the difference between two perspective projections of the nearby environment at two points of observation, more particularly the layout of the hands and things grasped in the hands.
The reader may have noticed the absence in the above discussion of the common textbook assertion that what the eyes fixate is an external object. The reason is that we cannot make this assertion. It begs the question of visual perception. It represents an illegitimate confusion of objects and light. All an eye can do is center its gaze-line on the peculiar structure in an optic array that specifies an object (or center its fovea on the same structure in a “potential retinal image over time”).
Summary
We have considered the following types of eye-movements and eye-postures: optokinetic nystagmus, coordinate compensatory movements, vestibular nystagmus, vestibular after nystagmus, the various kinds of fixation, including pursuit fixation, the saccadic movements that go to make up the various kinds of scanning, and finally the convergent fixations and conjugated saccades that specially characterize our own human eyes.
Granting certain assumptions of ecological optics all the above can be understood as mechanisms to facilitate the pickup of information in light. The eyes are stabilized to the ambient array; they sample the array in time; they foveally fixate items of the array and scan or explore it; and finally they must show convergent foveal fixation with conjugated saccadic movements if binocular disparity is to be utilized.
Bibliography
Gibson, J. J. The Senses Considered as Perceptual Systems, 1966, Ch. 9, 10, 11.
Howard, I. P. & Templeton, W. B. Human Spatial Orientation, 1966, Ch. 3.
Woodworth, R. S. & Schlosberg, H. Experimental Psychology. Ch. 17.
Yarbus, A. L. Eye Movements and Vision, Tr. 1967.