Argument from Illusion

Descartes observed that not everything we sense is as it appears. He used many of the same examples that Ayer does. He noted, though not on exactly the same grounds as Ayer, that there is no way to tell just from our sensations, which ones are sensations of things as they actually are and which ones aren't. He concluded that we can't be sure, on the obvious basis of sensations that things are the way we think they are. He raised, for the first time, the question of knowledge of the external world, and related issues.

Ayer has inherited a modern refinement of the same problem. He takes Descartes argument and converts it into an argument that we never see physical objects at all, only sense data.

The argument has two main parts.

1. We sometimes see sense data, not physical objects. (Sense data exist.)
2. We never see physical objects, only sense data.

The problem of knowledge of the external world becomes the problem of how our sense data are evidence for the existence of an external world.

Part 1. Sense data exist

What do we sense when we what we are sensing is not a physical object? Sense data. That is, sense data is just a made-up philosopher's term for whatever answers that question.

Thus, claim that there is sense data is just the fact that what we seem to sense isn't always what is in fact there put in slightly different terms.

Part 2. We never sense anything other than sense data

• Sense data are of a very different kind from physical objects, and yet we cannot ever sense the difference directly—we infer it from experiences of a different sort. Claim: Since we do sometimes experience sense data, and since we never experience anything of a very different kind from sense data, we never sense physical objects, and so all we ever sense is sense data.
• We have experiences that vary continuously from ones that are clearly of sense data to ones of the sort we might have thought were experiences of physical objects, but there is no dividing line. Thus, there is no shift, and so all we experience is sense data.
• None of our experiences that might be taken to be of an object are of the object alone: there always need to be appropriate other conditions on us and the environment. Thus, we never have an experience of a thing in itself. But that means none of our experiences are of objects, and so they must be of sense data.

---

That is what Ayer calls the argument from illusion, which is an argument from cases in which our sensations do not correspond to what there is ("delusive" or "nonveridical" perceptions) to the claim that all we ever sense is sense data. Of course, the Cartesian problem of knowledge of the external world immediately follows, since if we have never seen the external world, there is obviously a problem about how we know that its there.

Let me point out that the problem arises with the fall of Aristotelianism: according to Aristotle's theory of vision, objects emit little "copies" of themselves that are received by our eyes. There can't be any illusion in the normal case because the color, shape, whatever, we perrceive comes directly from the object itself. The new theory of vision, of course, has images on retinas, which are, in fact, always distorted, and so, even in the best case, what we get is indirect and distorted.

It is important to note that sense data and retinal images have nothing to do with each other: the new theory of vision makes the idea the perception of physical objects is never direct plausible, but sense data are what we "see," they are the answer to the question above, which has nothing to do with retinas.

The argument that all we see is sense data is completely bogus

In part 1 of the argument, we assumed that things can't look other than the way they are: if I see a bent stick, but there is only a straight stick, then I inferred that I'm not really seeing the actual stick. But I could instead have said, among many other possibilities, that I am seeing the straight stick, it just looks bent. Matters are harder for sensations of things that don't exists at all, like phantom limbs, but perhaps that can be finessed, at least, no argument has been given to the contrary.

In part 2 of the argument we assumed that qualitatively different things must produce qualitatively different sensations. Maybe sense data and physical objects really are different, but we can't sense the difference.

We also just assumed that if our sensations are in part determined by something other than an object, they can't be sensations of the object. But why not? Perhaps, for example, things really look the color that they are when they are seen in good light by someone with normal vision not on drugs.

The upshot is that the argument from illusion is completely unconvincing: we made all sorts of unjustified assumptions in giving it.

However, none of the unjustified assumptions under discussion are empirical assumptions: we didn't assume that you can tell when a perception is delusive, we didn't assume that bent sticks look straight in air, and so forth. None of the unjustified assumptions has anything to do with "the facts." Someone who says that we always see physical objects, they just aren't always how they look, doesn't disagree with the advocate of sense data about anything about our actual sensation, only about the best way to describe them.

The linguistic move

Talk of sense data is one perfectly good way to describe certain problematic features of sensation. There may be others. The argument from illusion shows, not that we never see anything but sense data, but rather that putting things in that way is one way to describe certain problematic features of perception. To adopt the "sense data language," we need to argue that it is a good and fruitful way to describe the phenomena for our purposes.

This kind of ascent to linguistic considerations is characteristic of modern analytic philosophy, and Ayer gives a wonderfully clear example of how and why it is done.

-- ShaughanLavine - 16 Jan 2007