Oxford: Blackwell, 1994. 162pp. Page references are to this book; section references (s.) to Berkeley's Essay towards a New Theory of Vision.
Department of Philosophy U-54
University of Connecticut
Storrs, CT 06269-2054
The excavation of old battlefields can yield some surprises. The old muskets or catapults turn out to be, for the age, surprisingly lethal devices, and the issues which separated the contestants, as well as the alliances which joined some of them, are often found to differ from those described to us in the official histories, written by the victors. So it is too with intellectual history. Robert Schwartz has provided a delightful example of the joys of excavation in this book on Berkeleian themes in theories of vision. Some of the current battle cries will never again sound quite the same.
Berkeley's Essay towards a New Theory of Vision, published in 1709, has over the succeeding centuries been received quite differently by psychologists and philosophers. Russell summarized the theory as the claim that "everything looks flat" and thought this claim "disproved by the stereoscope". Endnote 1. Philosophers have taken the central conclusion of the New Theory to be the claim that distance cannot be immediately seen, and derided this as false or confused, but at any rate obviously unworthy of belief. It may come as a surprise to find that claim as the central tenet in a long tradition of psychological theories; and to learn that many psychologists, including Helmholtz and Wundt, took the stereoscope to provide important confirmation for the Berkeleian approach. Examination of Berkeley's text fails to endow the conclusion that distance cannot be immediately seen with any central importance; Berkeley seems instead to treat it as a premise he shares with the "optic writers", and the argument for it occupies a mere two sentences. Time has blurred the battle lines, burying some completely. Historical figures are often used as mere icons for current needs, but the distortions and abuse heaped upon Berkeley seem extreme even by those standards.
The book is organized as four essays on related themes: distance, size, perceptual inference, and Gibson. Berkeley's New Theory is often the starting point, but the essays range over the history of theories of vision, including contemporary developments. Schwartz does not aim to provide an historical narrative, but rather to clarify some contemporary conceptual disputes in theories of perception of distance and size, and in debates about "direct" perception and perceptual inference. Their origins can often be traced back to ideas in the New Theory. The result is not so much today's look at Berkeley as a Berkeleian look at today. As Schwartz puts it, his goal is not to give yet another listing of where Berkeley went wrong, but to ask instead "where did Berkeley go right?"--a delightful, if atypical, approach.
Schwartz disentangles three distinct claims about distance in Berkeley's account: (a) that distance is not an immediate idea of vision; (b) that strictly speaking ideas of distance are ideas of touch; and (c) that visual experience has no inherent spatial character. While attention has largely focused on the first claim, Berkeley it thought to be uncontroversial--something shared with the "optic writers". His contribution should rather be located in claims (b) and (c). Suppose the meaning of spatial ideas is not purely visual, or even purely sensory, but must be unpacked partly in tactile and motor terms. To perceive a place visually, perhaps one must have a concept of "place" defined in part by motor consequences; a "place" to which one can reach or move. Vision provides ideas that at best correlate with such places, and those correlations must be learned. From this account developed the long tradition of psychological models describing "cues" for depth: visual ideas that suggest or can be associated with depth, and from which one can eventually construct a concept of space in what are now visuo-motor terms. Even if one allows that visual experience has some inherent spatial qualities, the significance of those qualities can only be calibrated in motor terms; this claim too is prototypically Berkeleian.
Now Berkeley goes on to deny that visual experience has any inherent spatial character, and this claim has always required the most heroic measures of defense. Berkeley claims that we no more see distance than that we hear distance; that "ideas of space, outness, and things placed at a distance are not, strictly speaking, the object of sight" (s. 46), and that a disembodied intelligence that possessed only the sense of sight would have no "idea of distance, outness, or profundity, nor consequently of space or body" (s. 154). Schwartz provides a viable interpretation for such claims. Berkeley does not deny that vision has some inherent organization, and he does not deny that things appear to be at a distance. According to Schwartz, his point is that any such organization is not inherently a spatial organization, and that we must learn to treat such appearances as appearances of distance. In a famous passage, Berkeley describes how at all times we see the same number of minimally visible points (or minima visibilia). He goes on to say
by the distance between any two points nothing more is meant than the number of intermediate points. If the given points are visible the distance between them is marked out by the number of interjacent visible points. (s. 112)
The objects of vision also have position relative to one another. Comparing the objects of sight and touch, Berkeley says
To objects of either kind we indifferently attribute the terms high or low, right and left, and suchlike, denoting the position or situation of things; but then we must well observe that the position of any object is determined only with respect to objects of the same sense. (s. 111)
So we have visual distances and relative positions. Are not these ideas of space?
Berkeley says strictly speaking not. They are ideas of the organization of visual experience, but not inherently ideas of space. This rests partly on the claim that ideas of distance strictly speaking are ideas of touch. If we also accept the Berkeleian premise that ideas of sight and of touch are totally distinct, and have different objects, it follows that we cannot have any visual ideas of (tactile) space. Berkeley here suffered from his inability to read Frege. Even if one allows that ideas with distinct senses might co-refer, one version of this conclusion remains viable. Those visual ideas are not inherently ideas of space. That they manage to latch onto the objects they do could not be derived from the facts of vision alone, or as a matter of reason, but only from contingent associations with facts of touch.
What then of the claim (a), that no visual ideas of distance are immediate? The Berkeleian argument for this claim is the famous "one point" argument:
Distance, of itself and immediately, cannot be seen. For, distance being a line directed endwise to the eye, it projects only one point in the fund of the eye, which point remains invariably the same, whether the distance be longer or shorter. (s. 2)
Schwartz says this argument does not depend critically on stimulus ambiguity--that stimuli anywhere along that line affect only that one point in the eye. Instead the claim is that there is no quality of visual sensation corresponding to differing distances along the line of sight:
When the three-dimensional world is mapped onto our two-dimensional retina, nothing in the retinal image "directly" presents or represents this distance itself. There is, in turn, no extent in the visual sensation corresponding to difference in distance along a line of sight. (p. 22)
This inference, from retinal image to sensation, is based on the "sensory core" model popular at the time, under which qualities of sensation more or less directly reflect qualities of the retinal image. That model is perhaps one of the areas in which Berkeley (and everyone else at the time) went more wrong than right, but Schwartz offers another defence. Even if the apparent brightness of a stimulus correlated perfectly and unambiguously with its distance, Schwartz says that Berkeley would not treat apparent brightness as an immediate idea of distance:
these different sensations of brightness would not, in and of themselves, be ideas of distance. They would remain immediate ideas of brightness that could serve only as signs of distance. Line-of-sight distance itself would not be displayed in the retinal image or its accompanying sensation. (p. 22)
Some very interesting moves are made in this argument, and thankfully their details are a large part of the subject matter for the rest of the book. But this argument
convinced Berkeley, along with almost everyone else, that seeing distance could not be immediate . . . Through most of the history of theories of vision it has been generally accepted that distance perception is not a direct or immediate process and that some version of a cue model is the correct approach. (p. 23)
With such a model, the stereoscope could be read as a vindication, rather than a disproof, of the Berkeleian approach: it was taken to show that the disparity between images could be an effective stimulus or cue for distance. Theorists continued to disagree about what sorts of visual processes were produced by the stereoscope. Helmholtz thought the stereoscope confirmed the importance of learned non-organic processes, and so showed that binocular fusion was a "psychic act". Ironically, the stereoscope was most widely thought to refute "Kantian" theories of vision, under which appearances of spatiality were treated as a priori intuitions imposed by the mind. The stereoscope, showing that cues in the stimulus could produce the appearance, thereby cast doubt on a priori accounts. Thus is one misunderstanding cancelled by another!
Schwartz describes a variety of vindications of Berkeleian insights in current accounts of binocular disparity and of the calibration of visual space. The chapter on size is essentially devoted to this theme as well. As Schwartz puts it
important aspects of Berkeley's position may remain on firmer scientific ground that is often supposed. For the claim that the distance meaning of visual experience is derivative, that vision must be scaled, and that this calibration is in units of motion constitute the main theses of Berkeley's account of distance perception. (p. 47)
These indeed are places where Berkeley went right.
Berkeley shared the early empiricist model of sensations as "natural signs" of their causes, produced organically by the respective sense organs, and requiring no mental elaboration. They are the inputs to inferential processes, the raw data, the given. Unfortunately this conjunction of features has not held together very well over the years; the different parts of it have come unstuck. In the central essay of the book, Schwartz examines some different ways of drawing the line between the immediate and the derived. The distinction is of central importance. Berkeleian accounts of distance and size depend upon it. The stereoscope does not constitute a disproof of Berkeley, in that if the processes caused by it which produce an appearance of voluminousness show the influence of learning, or cannot be given an organic explanation, or, in more recent terms, require computational explanations, then they are not purely "sensory" processes, and the ideas they produce are not "immediate" ideas. The distinction is vital still, in interpreting the work of Gibson and the fierce controversy over the question of "direct" perception. Do perceptual systems indulge in the computational elaboration of "cues", or do they directly "pick up" invariants in the optic array? Before charging into the breach it might be advisable to know what the terms of the dispute mean.
The map that Schwartz provides of this particular minefield will be of interest and enduring value to any investigator who saunters anywhere in the neighborhood. He describes five different interpretations of the contrast between the immediate and the inferred. (It is fine that theoretical terms have several meanings, but awkward to find five.) Part of the task is to show that these interpretations are logically independent, and do not hang together; this Schwartz does with some trenchant examples.
So what are five ways to distinguish the immediate from the inferred? First, postulate a difference in kind between sensation and perception, or between sensory processes and perceptual processes, and identify the immediate with sensation, the inferred with perception. Second, ask whether learning is required, and identify the immediate as that which requires no learning, the inferred as that which does. Third, draw a distinction between impoverished stimuli and the processes which elaborate those stimuli, and identify the impoverished beginnings as the contribution of the world, the elaborations as ours. Fourth, distinguish processes in the brain which have purely organic explanations from others, which are "mental". The latter earn the honorific because they (variously) involve consciousness, or reasoning, or computation. Fifth, invoke an epistemological distinction between that which can be perceived directly and that which can only be perceived indirectly. The former provide premises for inferences yielding the latter.
In this wash of ambiguity there is no hope of legislating a single correct answer, and the desire to do so betrays a certain political naivete. Schwartz predicts that "no single yes or no answer to the question of visual inference is likely to prove satisfactory" (p. 111). No single answer will be satisfactory because we do not have one question; we have at least five. Furthermore, since these interpretations are logically independent, unless one is purposely seeking camouflage, there is little to be gained by lumping them all under one big banner. Similar morals apply to terms describing the inputs to inferences: the data, the immediate ideas, the given. With no single correct way to draw the line where inferences begin, there is also no single correct way to draw the line where the data end:
The problem with the notion of "the given" is not that we cannot make sense of specific notions of input or data to a system. It is with the term "the" and its implications of uniqueness. (p. 119)
If you like the style of this passage, you will like the book. In the end Schwartz questions the "sense, value, or need" of taking sides in the debate on perceptual inference, and recommends we simply abandon the term. He says the debates "lack substance": they lack empirical substance, but if anything they are stuffed with too many philosophical ones.
It is testament to Berkeley's contribution to the theory of vision that the first significant challenge to cue models of distance perception arose only in the 1950's, with the work of James Gibson. Gibson has some sharp disagreements with that tradition. For example, the one-point argument shows there are no retinal variations corresponding to varying distances only by confining itself to static and monocular vision--what Gibson calls "snapshot" vision. Allow the retina to move, or bring two into play, and one could immediately register such differences. Perhaps higher level relations and ratios in the retinal images--gradients, texture units, disparities, optic flow patterns and variations over time--provide information sufficient unambiguously to determine distance. Gibson aimed to show that one can immediately sense distance, and that perception of space does not require cue interpretation, perceptual inference, or hypothesis formation.
Schwartz points out a surprising number of agreements between Berkeleian and Gibsonian approaches. Both emphasize the role of vision in guiding movement. Both doubt the psychological reality of computational processes, and agree that in some sense there is direct perception of the visual world. Schwartz claims that Berkeley would not be dismayed by the discovery of retinal disparities, texture gradients, or optic flow patterns; he would treat them as additional cues of distance. Whereas Gibson would claim that
the traditional approach goes wrong at just the point at which it conceptualize these features of the light array as cues. Being sensitive to the higher-level invariant features of the stimulus array is not to sense "signs" of the environment, but to sense spatial properties of the environment themselves. (p. 143)
So has Gibson shown that distance can be sensed directly? The conflict is seen to rest not so much on empirical differences between the theories, as on how those theories are described. Once again we fall into the pit of five-ways ambiguity. If one is convinced that there is no single correct answer to the question of how to interpret "immediate" or "direct" perception, then there is no need or value in resolving this dispute, "for the question itself is vague and subject to alternative interpretations" (p. 132).
A similarly deflationary conclusion is found to apply to Fodor and Pylyshyn's attack on "direct perception". Undoubtedly the Gibsonian notion of "direct pickup" suffers from multiple interpretations, but the same is true of Fodor and Pylyshyn's notion of "transduction", and only grand arguments of principle can be provided for treating everything beyond transduction as inference. The problem with such arguments is that they
go through (or fail) pretty much independent of most of the empirical and theoretical issues that separate Gibsonian and conflicting accounts of visual phenomena. (p. 141)
Philosophy of psychology should try to achieve better acuity than this. Schwartz recommends that we give up as hopeless the ambiguous distinction between the immediate and the inferred; or failing that, at least provide a non-tendentious description of which version of it we favor. This suggestion will undoubtedly irritate all sides of the dispute; but insofar as the goal is to resolve the substantive issues, that resolution can only proceed in the way that Schwartz suggests. His book disentangles issues in theories of vision that had been hopelessly confused, and provides a clean map of concepts and categories some of whose borders had been blurred by controversies dating to 1709. The clarity it brings to discussions of perception of distance, of size, and of perceptual inference is exemplary. Run your theories through its sieve; they will emerge the better for it.
1. Bertrand Russell, Human Knowledge: Its Scope and Limits. (New York: Simon and Schuster, 1964), p. 51.
Return to Austen Clark's online papers .
Return to the Philosophy Department home page.