Helmholtz’s need to invoke another factor in such cases must therefore be due to his embracing a theory which ruled out such explanations as I have mentioned. And this is indeed the case. For he embraced what has become best known through the criticisms of the Gestalt psychologists as the ‘constancy hypothesis’ – that there should, unless other factors intervene, always be a constant relation between how we perceive things and the pattern of stimulation. I suspect that the constancy hypothesis has at its back another doctrine – that of ‘specific nervous energies’ – put forward by Johannes Müller in the 1830s, although he himself acknowledged its source in another physiologist named Bell. On this theory each nerve has its own function and cannot take over another function. It was inferred from this that the excitation of any given nerve attached to a sense-organ produces its own experience. Such a theory is naturally associated with and could be used to bolster up, firstly, a sensory atomism – the thesis that in vision, for example, excitation of the eye produces a mosaic of discrete and atomic visual experiences. Secondly, it could be used to provide the backing of the constancy hypothesis to which I have already referred, since the hypothesis of discrete neural functioning naturally has as a corollary the thesis that what takes place at the level of nerve-endings must be repeated at the level of resulting experience. And if the neural functioning at the level of nerve-endings is discrete so must be the array of experiences which result. Discrete functioning rules out interactive processes. One can see from this why the Gestaltists, who were so opposed to the idea of discrete functioning in the nervous system, were equally opposed to the constancy hypothesis. But the thesis of isomorphism – that there is an identity or complete similarity of structure between the phenomenal (how things appear) and the cortical (how things are in the brain) – is nevertheless a theory in the same line of country. For it assumes that there is a correspondence between how things appear and the pattern of neural excitation, even if it rejects the idea that the correspondence must be at the level of what happens at nerve-endings.

Helmholtz thought that the deviations of perceptual experience from what is to be expected on the basis of the constancy hypothesis are to be explained not in terms of anything further in the process of stimulation of sense-organs (for how could it be on the assumptions in question?) but in terms of ‘intellectual processes’ carried out by the perceiver. In particular the perceiver makes some kind of judgment or carries out some kind of inference, which, since he is not aware of it, must therefore be unconscious. The role that this might play is most obvious in the case of space-perception where the sizes and shapes that things seem to have often deviate from what might be expected from a consideration of the pattern of stimulation on the retina alone. Helmholtz also used the idea in connection with such phenomena as colour contrast. The Helmholtz three-cone theory of colour vision allows a deviation from the thesis of one experience – one kind of nerve being stimulated, in that it allows that some experiences, the perception of yellow, for example, may be the product of the excitation of more than one kind of cone simultaneously in a given proportion. But it does not allow for interactive processes between areas of the retina stimulated in the ordinary way. That is to say that a set of cones being stimulated by light of a certain frequency should produce a definite experience whatever might be going on in adjacent areas of the retina. But this does not seem to be the case. Hence it has to be explained, given these presuppositions, in a quite different way. It is assumed that a grey patch seen in the middle of a red area will tend to look green because the perceiver makes an inference about the colour of the inset patch. That is to say that there is a tendency to infer that the inset colour must be opposed to the colour of the surrounding area, and this inference sometimes overrides the given facts of sensory experience. Helmholtz thought that it does this because we have the tendency to take the inset patch as seen through the surrounding colour.

My phrase ‘the given facts of sensory experience’ is important because it reveals that there is presupposed in all this a thesis about the relationship of sensation to perception. The constancy hypothesis in fact assumes that, at least over the range of phenomena where it applies, perception is just a matter of sensation. Thus where the hypothesis cannot get application the deviations that are apparent cannot be due to anything sensory and must therefore be explained by reference to processes that are not sensory.¹ In fact the rejection of the constancy hypothesis by the Gestalt psychologists and its replacement by the thesis of isomorphism is, despite first appearances, a kind of reinstatement of the role of the sensory in perception, except that it is recognised that the nature and structure of the sensory is not determined simply by what happens at the level of nerve endings. At all events it rejects the idea that anything non-sensory or intellectual plays a part in perception, the gap between what might be expected on the basis of what happens at the level of nerve-endings and what appears to be the case from a consideration of the phenomenal being closed by appeal simply to further or different causal physiological processes. It is perhaps of some interest in this connection that J.J. Gibson claimed in his The Perception of the Visual World to be resuscitating the constancy hypothesis by providing a more comprehensive account of what must be taken as happening at the level of the retina; he thereby ruled out any role for anything like unconscious inference. He also claimed that his theory was a psychophysical one. In his more recent book The Senses Considered as Perceptual Systems, which in many ways extends the account of the earlier book, he makes an explicit rejection of the notion of sensation as having any part to play in a theory of perception, as well as a rejection of any intellectual processes such as those involved in conception and belief. Yet in most ways the theory of the later book is just an extension of that of the earlier. How can he in that case reject the notion of the sensory in an account of perception?

The answer to this question lies, I think, in the tradition of thinking against which the psychology of perception has developed. For while it is natural perhaps to think that the stimulation of a sensory organ must eventually result in a sensation or group of sensations, it is not so clear exactly what a sensation is; and for that there is a whole epistemological tradition to fall back on. When I say that it is not clear exactly what a sensation is, I mean that this is not clear in the case of the majority of the senses. The suggestion that the stimulation of one kind of nerve ending may produce a sensation of pain is relatively unproblematic. But what counts as analogous to this in the case of, say, vision? The empiricist tradition in epistemology has generally had recourse to the notion of sense-impressions or sense-data as what is given in perception. It is thus natural, once given this notion, to suppose that these sense-impressions are what are caused by the stimulation of our senses – and the term ‘sense-impression’ is such that its very etymology supports that idea. Any suggestion that what we perceive and how we perceive it must be a function of construction or other activity on our part correspondingly undermines the view that perception is ultimately a matter of things being given via sense-impressions. And that is how it is with Gibson. I say nothing here about the rightness or wrongness of his view; I wish merely to make clear its ancestry and general character.

I have tried elsewhere, particularly in my Sensation and Perception, to make clear the role of sensation within perception. It may be enough now to say that what I have described above as natural – that it is sense-impressions that are the result of the direct stimulation of our sense-organs – is, when one comes to think about it, not natural at all. The supposal that it is comes from the pervasiveness of a tradition, and the way in which the psychology of perception has emerged out of a long course of philosophizing about perception with epistemological problems in mind. Though I find it an increasingly unpopular view these days, I am still of the opinion that distinguishing philosophical from psychological questions is a desirable prolegomenon to any fruitful thinking in this and other areas. Progress does not always come from everybody trying to tackle the same issues in the same way, without distinction of function, and without, to mention a more important point, getting clear what the problems are. Co-operation between philosophers and psychologists may be more fruitful if it comes from delimitation of function rather than an attempt to answer any question that presents itself without first sorting out what kind of question it is. However that may be, the question whether there is anything given in perception is the question whether the use of our senses provides us with any knowledge which is both incorrigible and independent of any knowledge of any other and prior kind. The empiricist who wishes to claim that knowledge is in the end totally dependent on the use of our senses must claim that there is knowledge of this kind. And such knowledge must not be concept-dependent in the sense that it presupposes, whether simply logically or also temporally, any conceptual understanding in terms of which what the senses provide is to be construed. For such conceptual understanding would involve a form of knowledge – a knowledge of how what is provided by the senses is to be construed, or, to put it in another way, what it is to be construed as. And in that case what the senses provide could not constitute a ‘given’.

The issues here are about the foundations of knowledge and whether there are such foundations and if so where they are to be sought. The empiricist view that such foundations are to be sought in sense-impressions or sense-data does not necessarily have as a corollary that these must be caused by the stimulation of our senses, however natural it is to think that this must be the case. For one thing, there are alternative theories, however plausible or implausible they may seem, e.g. the view put forward by Berkeley that sensory ideas can only be caused by spirits. Such a view would be absolutely excluded if the thesis that sense-impressions are caused by the stimulation of our sense-organs was a necessary corollary of the thesis that sense-impressions provide the foundations of our knowledge. If it is natural to look in such a connection to the stimulation of our sense-organs, it is because a causal process of this kind seems to exclude the possibility of its results presupposing the application of knowledge derived from any other source. It remains however a question whether those results can be construed as knowledge at all, and if they cannot they are of no use in providing the foundations of knowledge. There would in that event be a gap between the causal story and the epistemological pursuits. In fact the ramification of these factors and the kinds of philosophical influences that have come to bear on psychological theories of perception have been very complicated; but basically the issues turn in the end on considerations of the kind that I have mentioned.

If we return to the constancy hypothesis, it is worth noting that what is supposed to be constant is the relation between the pattern of stimulation and the resulting pattern of sensations; yet those sensations must be thought to constitute forms of perception. For, anyone who embraces the constancy hypothesis ought to suppose that how we see the world can be explained ultimately in terms of perceptions corresponding to the sensations brought about by strictly sensory processes, and any other factors that are introduced simply build on the foundations of the original perceptions. The Helmholtzian notion of unconscious inference in a way makes this clear, for inference must take place from what is known or believed; hence if the inference is to be the explanation why we do not see things as the constancy hypothesis suggests it must work by moving from what is known or believed as a result of the processes that the constancy hypothesis supposes hold good. In other words the stimulation of the senses must on that account produce a form of perception which involves knowledge or belief. Yet neither that knowledge or belief nor the process of inference from it to the perception that finally results are such that the perceiver is in any way aware of them.

It might be objected that the process of stimulation itself need not be thought of as producing knowledge or belief but that the perceiver has to regard whatever it does produce in such a way as to make inference from it. That is to say that the perceiver has to believe something about whatever the stimulation of the senses produces as a basis for subsequent inference, but that the product of the process of stimulation does not itself have to be a belief. It sometimes happens that we infer things from what are in a genuine sense sensations. We may for example infer from the quality of a bodily sensation something about the identity of whatever is causing it, e.g. that the prick is being caused by a prickle or splinter. Such inferences, which need not be in any way unconscious, depend however on further knowledge, presumably acquired through experience, about the kinds of sensations that certain kinds of thing produce when they affect our bodies. Given this knowledge the inference to the identity or nature of the thing causing the sensation may be a genuine inference in every way. But the Helmholtzian situation is not like that (although it is perhaps worth noting that the local-sign theory about the location of sensations – a theory that Helmholtz also embraced – may involve something of this kind). Consider the Helmholtzian theory of colour contrast, which depends, as I have already noted, on the thesis that we tend to make an inference about the colour of an inset area from that of the surrounding area on the basis that we are seeing the inset area through the surrounding area. The principle of inference here must presumably be that seeing something through an area of a given colour must inevitably make the something look more like the colour of the surrounding area than it would otherwise have done. The situation would be supposedly like seeing things through coloured spectacles. But a grey patch set in the middle of a red area does not look reddish. Hence what we must infer is that its real colour is not plain grey but of a colour that is shifted more towards the opposite of red. Hence it looks greenish.

There are several points of interest in this, and a discussion of them will bring out certain general considerations about the relation between how things look and any putative intellectual processes that are involved in perception. The first thing to note is that in the story that I have given a reference to how things look appears twice. There is, first, the suggestion that something seen through another colour must inevitably look tinged with that colour. There is, second, the final report about how the thing in question actually looks, and it in fact looks different from how it would look on the first hypothesis. There is, in fact, of course the same deviation in looking at things through coloured spectacles, in that as a matter of fact things look or at least come to look more like their actual colours than one might expect at first sight. The explanation clearly lies in certain features of context. Hence the principle that is supposed to give rise to the inference is not itself founded on any confirmed law about the looks of things in these circumstances. It is based on an a priori belief that seeing things through a coloured medium must make them look tinged with that colour. I say that this is an a priori belief to indicate that it cannot in the circumstances be based on any exact empirical findings. Wearing rose-tinged spectacles may well alter our view of the world in various subtle ways but it may not make things actually look rose-coloured. Whether it will probably depends on a number of other factors, including how long we have been wearing the spectacles. What will undoubtedly be the case is that the wave-lengths of light reaching the eyes will be affected by wearing spectacles of this kind. If we believe that this must necessarily affect how things look, it must be because we think that there is a uniform correlation between the look of a thing and the wave-length of light which reache...