So far, we have examined Russell’s theory of descriptions as a view about a particular class of expressions in English. We will now turn our attention to the importance of that theory for the philosophy of mind and epistemology.

1 Sense data, acquaintance, and the possibility of thought

One prominent question in the philosophy of mind, which has been around at least since Kant, is: How are certain kinds of thoughts possible?

One plausible kind of answer to this question might be called empiricism about the contents of thoughts. This holds that, for every constituent of thought, an agent is able to have thoughts involving that thing just in case it was presented to him in experience. (This should be distinguished from empiricism in epistemology, thought of as a view about justification.)

Why empiricism about the contents of thoughts is plausible.

Russell holds a version of this view. He begins by explaining the key notion in his philosophy of mind, acquaintance:

He then moves to discussion of the kinds of objects with which we can be acquainted. These fall into three classes, along with a possible fourth:

1. Sense data. We are familiar from our reading of Ayer with the arguments behind the sense datum theory of perception, which holds that in perception we are immediately presented with mental particulars. Russell held a version of this view; hence it was natural for him to think that sense data are among the things with which we can be acquainted.
2. Universals. Russell: “...we have also ...what may be called awareness of unniversals. ...Not only are we aware of particular yellows, but if we have seen a sufficient number of yellows and have sufficient intelligence, we are aware of the universal yellow.” (212)
3. Relations. Sentences like ‘if one thing is before another, and that is before a third, then the first thing is before the third’ as showing that awareness of relations cannot be analyzed as awareness of relata.
4. Oneself.

Russell then states his version of empiricism about the contents of thoughts as follows:

What follows from this principle, along with Russell’s views about the sorts of things with which we can be acquainted, is that the contents of our thoughts must only include sense data and properties, but never material objects. There is thus an important sense in which the external world is beyond the reach of our thought.

2 Names, understanding, and the possibility of thought

In ‘Descriptions’ Russell distinguished between names, which directly designate objects in the world, and definite descriptions, which do so indirectly. This gives us two ways in which the thought expressed by a sentence may concern an object: the object may itself be a part of the thought (as would happen if the sentence contained a Russellian name), or the object may be singled out indirectly by properties which are themselves a part of the thought (as would happen if the sentence contained a description satisfied uniquely by the object in question).

This distinction corresponds to an intuitive distinction between two ways of referring to, or thinking about an object. (Consider, on the one hand, demonstratives, and on the other hand, descriptions like ‘the point midway between my left foot and the surface of the sun’.)

This gives rise to the same question discussed above: what kinds of objects can we think about directly, as opposed to merely indirectly, via some properties which they happen to have?

Given Russell’s view of names, we can raise this question in a linguistic form by asking, ‘What sorts of objects can we name?’ The purpose of this section is to show that, using Russell’s distinction and some plausible theses from the philosophy of language, we can reach the same result as in the previous section: that the constituents of our thoughts include only sense data, universals, and relations, but never other external objects. The interesting point is that the present argument seems to show this without relying on the sense datum theory of perception.

The first principle is about our access to our own thoughts:

This is closely linked to another principle:

These principles are linked because, you might think, if [2] were false then [1] would be as well. For if one could be mistaken about whether a sentence of one’s own language was meaningful, then one might say that sentence to oneself, thereby taking oneself to be thinking a thought. But if the sentence was meaningless, one would not be thinking a thought, and so one would, contra [1], be mistaken about whether one was thinking a thought. This is some reason to think that [1] entails [2].

Consider now the following plausible principle about the meaningfulness of words and sentences of which they are a part:

Consider, e.g., ‘They’re serving flibbertyflam in the cafeteria today.’ It’s plausible that this sentence fails to express a thought because ‘flibbertyflam’ lacks a meaning.

From [4], it is natural (though it does not strictly follow) to infer the following principle:

But it follows from [3] and [5] together that

But from [2] and [6] together it follows that

From which it is natural to infer

But now recall the parallelism between thought and language with which we began:

If this is right, this means that we can derive from thesis [8] a claim about the sorts of objects with which we can be acquainted, or think directly about: those objects about whose existence we cannot be mistaken.

And from this it seems to follow that we cannot name, or be acquainted with, physical objects, since we can always be mistaken about their existence. And, from Russell’s ‘fundamental epistemological principle’, it then follows that physical objects can never be constituents of our thoughts.

3 The role of the theory of descriptions

These two arguments both purport to show that the contents of our thoughts are limited to sense data and universals and relations. But this gives rise to an obvious problem. There are many sentences which we seem to understand, but which appear to contain names for material objects. Russell’s theory of descriptions as comprehensive solution to this problem. The example of ‘Julius Caesar.’ (221)