Does SDT imply a homunculus fallacy?

Sense datum theory does not commit the ‘homunculus fallacy’. What I mean by this is that of the numerous theories which fall under the rubric ‘sense datum theory’ there is at least one which is not committed to the fallacy. There are more than one SDT’s which are free of the problem, but one is enough to make the logical point. Admittedly, some SDT’s seem designed to fall prey to the fault. However, simply being a sense datum theory does not incur the fallacy.

Naive Sense Datum Theory (NSDT), the preferred family of theories in the Official View of this site, and other versions of SDT are free of the problem. Of the others, several which are prey to the problem seem to be as much the artefacts of hostile critics of SDT as of their advocates. As we shall see, some versions of SDT have the superficial appearance of having the difficulty, but careful analysis shows that things are not so straightforward as they may seem.

For a quick example of what is at stake in the issue, consider Susan Blackmore's discussion a subject’s experience of rotation of mental imagery (arguably a companion process to visual perception). She writes:

If you are tempted to think that there must be a mental screen on which the rotated image is projected, and that “you” either do or do not consciously look at the screen, then ask yourself where and what you and the screen could be. If you are a conscious entity looking at the screen, then the classic homunculus problem arises. The inner “you” must have inner eyes and brain, with another inner screen looked at by another inner you and so on – to an infinite regress.

Blackmore, Susan, Consciousness. (2nd edn.) New York / Oxford: O.U.P, 2014, p. 56.

Blackmore does not make explicit any of the steps in arguing from the assumption of inner screen to the infinite regress, but it shows the direction of argument. Even classic historical sources of the theory of ideas, for our purposes a prototype of SDT, have been aware that an incautious formulation of the view invites the beginnings of the homunculus critique.

Thus, Descartes wrote:

Now, when this picture thus passes to the inside of our head, it still bears some resemblance to the objects from which it proceeds. As I have amply shown already, however, we must not think that it is by means of this resemblance that the picture causes our sensory perception of these objects—as if there were yet other eyes within our brain with which we could perceive it. Instead we must hold that it is the movements composing this picture which, acting directly upon our soul in so far as it is united to our body, are ordained by nature to make it have such sensations.

AT VI:130, CSM I:167 (emphasis added).

Although more is going on in this text than simply the mobilizing of the first step of the homunculus argument, that key element is unmistakably there.

Representative of the employment of the homunculus argument by modern analytic philosophers, we may find in Gareth Evans’ (in this case posthumously published) writing an explicit identification of what the problem is — in the case of the perception of visual structure.

We may start by taking note of the fact that many of the spatial descriptions of visual experience which we are inclined to give are obviously metaphorical. For example, I myself have been speaking of cortical stimulation ‘producing a pattern of phosphenes arranged in a square’, and the literature is replete with such metaphorical talk. It is clear that such descriptions cannot be taken literally: there are not literally four points of light, or indeed four things of any kind, arranged in a square. To think that when a subject seems to see four points of light arranged in a square, there really are four (mental) items actually arranged in a square is to commit the sense-datum fallacy. It might be better to call it ‘the homunculus fallacy’, to which it inescapably gives rise. One commits the homunculus fallacy when one attempts to explain what is involved in a subject’s being related to objects in the external world by appealing to the existence of an inner situation which recapitulates the essential features of the original situation to be explained – by introducing a relation between the subject and inner objects of essentially the same kind as the relation existing between the subject and outer objects.

Evans, Gareth, "Molyneux's Question", in Collected Papers. Oxford: Clarendon Press, 364-399, p. 397.

Consider last the brief statement of the Cartesian Theatre problem by the perceptual psychologist, Nicholas Humphrey, invoking the position of the philosopher Daniel Dennett and arguing by rhetorical question:

Daniel Dennett, the leading critic, writes: “The persuasive imagery of the Cartesian Theater keeps coming back to haunt us—laypeople and scientists alike—even after its ghostly dualism has been denounced and exorcized.” He is right, of course, to reject the idea that there could be a place inside your head where one part of your brain creates a faithful replica of the world for another part of your brain to look at (and what part of your brain would look at the replica of the replica?). (Emphasis added)

Humphrey, Nicholas, Soul Dust: The Magic of Consciousness. Princeton / Oxford: Princeton U.P., 2011, p. 40.

The quotations above together pick out a cluster of issues around the homunculus fallacy. I shall return to these texts and tease apart some of the different strands that are present in their commentaries. First, however, I shall make a short statement of the NSDT position in regard to its liability to the alleged problem.

NSDT is a form of direct realism. What distinguishes it from traditionally so called ‘Direct Realism’ is that it says that what is ‘directly’ accessed is a volume or surface in the brain. What we usually take to be the outside world, presented in consciousness, is a brain construct according to NSDT. In the Direct Realist conception, the phenomenal array is taken to be the distal world. In that conception, seeing and being visually consciously aware of are coextensive, if not identical. In NSDT, we might be tempted to preserve that coextensiveness, but that would have the terminological consequence that we will say both that we do not see the distal visual world and that we are not consciously aware of the distal world. That would be one suggestive way to go with the terminology; it would emphasize our experiential remoteness from the world outside our bodies. However, there is some convenience in preserving locutions which are familiar to us, and, consequently, in this formulation of NSDT we reserve the term ‘see’ for a relation of the subject to the distal stimuli. ‘Being consciously aware of’ is reserved for a relation of the subject to the phenomenal array, understood as a brain construct. That is dictated by the substantive content of NSDT. Accordingly, for NSDT we are consciously aware of a brain state, and not consciously aware of the external world. On the other hand, we see objects and/or properties in the external world of distal stimuli. Thus, our conscious awareness of the phenomenal array is a component of a more comprehensive process of seeing. It might properly be called a modular element in that process.

With these commitments in mind, we may readily see that the homunculus problem, in the guise of Dennett’s Cartesian Theatre, does not arise for NSDT. That is because, on NSDT’s account, we do not suppose, when we look at an external object, that we look at the phenomenal array in the brain, although we are consciously aware of it. (For present purposes I suppose ‘look at’ to be a cognate term with ‘see’.) The essential point is that looking at or seeing external things is not grounded in looking at or seeing inner things, but rather (partly) grounded in being consciously aware of them.

Similarly, there is no homunculus fallacy / sense datum fallacy as described by Gareth Evans, in which we “ explain what is involved in a subject’s being related to objects in the external world by appealing to the existence of an inner situation which recapitulates the essential features of the original situation to be explained – by introducing a relation between the subject and inner objects of essentially the same kind as the relation existing between the subject and outer objects.” In NSDT, conscious awareness, the relation between subject and inner objects, is not of the same kind as seeing, the relation between subject and outer objects.

Are there indeed SDT’s that commit the homunculus fallacy? Consider the following invitation of Michael Tye, a hostile critic of sense datum theory:

Focus your attention on the scene before your eyes and on how things look to you. You see various objects; and you see these objects by seeing their facing surfaces. Sense-datum theorists claimed that the facing surfaces of the objects are themselves seen by seeing further immaterial surfaces or sense data.

Tye, Michael, Consciousness, Color, and Content. Cambridge, MA.: M.I.T. Press, 2002, p. 45.

Is it true that sense datum theorists account for seeing the facing surfaces of external bodies by appealing to the seeing of inner objects and attributes? Obviously not, if, for example, NSDT is a consistent theory. However, there are well-known versions of SDT that seem, at first sight, to explain seeing external things by seeing (internal) sense data. A plausible example is the sense datum theory of Frank Jackson.

According to Jackson (1977), the objects in the (external) world that we see are mediate rather than immediate objects of perception, where x is a mediate object of (visual) perception (for S at t) iff S sees x at t, and there is a y such that (x ≠ y and) S sees x in virtue of seeing y.

Jackson explains ‘in virtue of’ by appeal to paradigm cases, including:

  • my car being red in virtue of its body being red
  • a sentence’s being true in virtue of a given proposition’s being true
  • someone’s being strong in virtue of his body being strong
  • someone’s living in Australia in virtue of living in Melbourne

His fundamental “analytic thesis” for the concept of seeing is:

… that to see a reasonable-sized, opaque material object is to see something distinct from that object, the relevant immediate object of perception (whatever the ontological status of the latter may turn out to be).

Jackson, Frank, Perception: A Representative Theory. Cambridge: C.U.P, 1977, p 28.

On Jackson’s version of “Representationalism” or “Representative Theory”, which denies that colours ‘properly speaking’ are instantiated in external objects, when we see a (so-called) white wall, for example, the immediate object of seeing is a coloured patch, the white surface. We see opaque objects, such as the wall, by seeing parts of them. The crucial thing for Jackson is that the part in this example is a white surface that is a mental, coloured patch. In Jackson’s own terms:

the wall is seen in virtue of seeing the white expanse; for seeing material objects can, according to [the Representative theorist], be analysed in terms of seeing mental entities belonging to them. That is, he holds that the application of ‘S sees —‘ to something material can be defined in terms of its application to something mental, but not conversely – because, for instance, the mental thing might exist and be seen without the material one existing or being seen.

This account is rather different from NSDT. It proposes a sort of intimacy between mental immediate objects of seeing, such as the sense datum that is the white surface, and the material mediate objects of seeing, such as the wall. That intimacy is constituted by the white surface being part of the wall and belonging to it, in the appropriate sense. (Aristotelian scholars may see in this theory something that accords with the conception of perception as (part of) the object existing in the mind.)

It seems to me that what follows from these proposals is that there is a certain tension between the verbal formula of Jackson’s thesis and the details of the seeing of an object. At first sight, one has the impression that seeing an object is explained by seeing sense data. The sense data are the immediate objects of seeing; the external object is the mediate object of seeing. If ‘seeing’ is being used univocally, we seem to have a situation of the kind that Gareth Evans described as the sense-datum fallacy or homunculus fallacy.

However, we learn from the application of Jackson's account to concrete cases such as seeing the white wall, that we do not actually see the wall. We never see opaque visual objects, only their parts. And the parts that we see are not, for example, the facing surfaces of walls, but rather mental coloured surfaces which, presumably, cannot be part of the very physical geometry of the wall itself. For one thing, nothing in the material wall is coloured. For Jackson, the only thing that is white is the sense datum and, so, what we see is a white surface which is distinct from the non-white wall. So, contrary to Jackson’s verbal formula for mediate and immediate perceptual objects, we do not actually see the mediate object. If that’s so, then Jackson’s SDT seems not to be committed to the explanation of outer seeing by inner seeing, where ‘seeing’ is understood univocally. (Jackson regarded his proposals about seeing sense-data and seeing external objects as using 'seeing' univocally. However, he conceived of his proposals, not as a reductive explanation of seeing objects by seeing sense-data, but as a philosophical analysis that constituted an "analytical expansion" of our talk about seeing objects. His own claim that seeing an object is seeing something distinct from that object seems to me to force speakers into apparent contradictions at odds with adequate philosophical analysis, if there is indeed any such thing sui generis. For this reason, I think it proper to judge that what make Jackson's theory seem vulnerable to the homunculus complaint are the tensions in his attempt needlessly to represent his claims in the format of a narrowly-conceived technical device.)


(to be continued)