Empiricism and Semantic Knowledge
Empiricism tells us that all knowledge worthy of the name derives from the senses. In Hume’s formulation, every idea has its origin in an impression, such as an impression of red. This is a psychological theory to which empirical evidence is relevant (what if we came across a whole batch of ideas that exist without benefit any prior sensory impression?). It is not intended as a logical or conceptual truth: there is nothing in the concept of an idea that entails that ideas must derive from impressions. That is why Hume conducts a survey of ideas to determine whether his general principle is correct. Notoriously, he runs into apparent counterexamples: not only the missing shade of blue, but also ideas of causation, the self, space and time, persisting bodies, and number. But he never (so far as I know) faces up to the case of semantic knowledge—what he would call ideas of meaning. We clearly have such knowledge: we know what the words of our native language mean and we can learn the meanings of words used by foreigners. Meanings are objects of cognition: we can think about them, ascribe them to marks and sounds, reason concerning them, have arguments over them. We know them as well as we know colors and shapes (the empiricist’s favorite examples). I know that “snow is white” means that snow is white—does anyone contest that? But how do I have such semantic knowledge—do I have it by means of sensory impressions of meaning? I hope the answer to that is a resounding No: when I hear someone speak and know his meaning I have no sense impression of meaning (I do have impressions of the sounds he makes). If I did, I could understand people speaking a foreign language without laboriously learning it—I could just sense what they mean. It would be like seeing a new combination of colors or hearing a new series of sounds—the senses would have it covered. There is thus no such thing as an impression of meaning in the sense intended by empiricists. This is as true for one’s own meaning as much as it is for other people’s meaning: I don’t have an impression of what “red” means to me. There are no sense impressions of senses, my own or other people’s. I don’t have sensations of meaning. I know about these things, but not because I sense them with my senses. If “impression” means “sense impression”, and these are limited to what the five senses convey, then we have no impressions of meaning. No child embarking on the business of learning language is ever confronted by an impression of meaning from which he or she derives ideas of meaning. Nor has anyone ever suggested such a thing—because it is obviously mistaken. So empiricism is false for the acquisition of semantic knowledge. Meaning belongs with those other ideas that resist the empiricist dogma: causation, the self, persisting bodies, etc. But in the case of meaning it is even harder to dispute whether the ideas in question are really possessed (do we really have an idea of the self?): we indisputably do have semantic knowledge, semantic concepts, and semantic thoughts—but without antecedent semantic impressions to ground them.
You might try claiming that our knowledge of meaning is theoretical: we infer meaning from other types of impression. It is hard to see how this will go but a suggestion that has found some traction is ostension: we have impressions of pointing. The index finger extends in a perceptible dog-containing environment as “dog” is intoned, and an observer can have an impression of this performance—you see dog-oriented pointing going on and infer the meaning of “dog”. There is no need for me to critique this theory, given the obloquy it has been forced to endure, but I will say that an impression of pointing (the stiffened index finger in line with an object) is hardly sufficient to generate an idea of the meaning intended. The impression is far removed from the semantic information it is supposed to impart; you may as well say the expression on the pointer’s face is the basis of the meaning he intends. Many beings (e.g. dogs) could be witness to such a performance and yet have no knowledge of the meaning of “dog” as a result of it. As the native extends his finger while uttering “gavagai”, the field linguist is still in doubt about what that word means. But even if knowledge of meaning could be gained from such impressions, the fact remains that meanings themselves are not perceptible. The first-person knowledge the linguist has of her own meanings is not impression-based either. Moreover, there is a strong feeling that meanings could not give rise to impressions: meanings are not the kind of thing that produces impressions—like numbers (and unlike molecules). It is a necessary truth that meanings are imperceptible. In any case our actual knowledge of them is not impression-based, i.e. acquired by means of direct observation of their nature (by “acquaintance”). Rather, ideas of meaning are brought to perceived utterances, not derived from such utterances. Where these ideas do come from is not so clear: they could be innate, or they could be created by the developing child in some way. They could also be a mixture of the two, as much of our knowledge undoubtedly is. But what an idea of meaning can’t be is a “faint copy” of an antecedent impression, or be “abstracted” from an impression of meaning. There are no impressions of meaning that could form the basis of our knowledge of meaning in the manner envisaged by the empiricists. Meanings themselves are right there before us, so to speak, but they don’t show up in consciousness as Humean impressions.
Accordingly, knowledge of analytic truth can’t be impression-based: you don’t have an impression of the meaning of “bachelor” and an impression of the meaning of “unmarried male” and notice that these impressions are identical. Yet you have genuine knowledge here, just not empirical knowledge. If it were based on semantic impressions, it would be empirical: but as it isn’t, it ain’t (as Lewis Carroll once said). So knowledge of meaning is a powerful counterexample to empiricist epistemology, yet seldom if ever cited in this connection. And if the correct theory of it is that such knowledge is either innate or created (or a combination of the two), then this theory might well be the one to adopt for other types of knowledge too, such as knowledge of color and shape. Impressions of red might merely elicit our innate concept of red, rather than being the basis on which we acquire that concept ab initio. At any rate, we don’t come to know meaning by having impressions of meaning fed into our minds.