Language and information

Lecture 3: Information

3.2. How words carry meaning

[Audio recording—Opens a separate, blank window. Return to this window to read the transcript.]

Now, a reconsideration of how words carry meaning. At what point do words get meaning? One should first point out something that may not be immediately obvious, and that is, that meanings do not suffice to identify words. They can give a property to words that are already identified, they don't identify words. Another way of saying this is to say that, as everybody who has looked through Roget's Thesaurus knows, there is no classification and categorization and structure of meanings per se, such that we could have the words of a given language assigned to those meanings in particular ways. The meanings are not arranged or organized or structured, that is to say, meanings over the whole scope of language cannot be arranged independently of the words. Of course they can in kinship terms, and in numbers, or a few very strictly organized parts of the world. The processes that did give us words—I mean the procedures for determining words—are in fact not semantic, that is to say, did not rely on meaning. The direct procedure which distinguishes one word from another, is based on the distinguishability of sounds, and is not able, for instance to distinguish homonyms one from the other. The stochastic procedure which I mentioned last week, which exhibits the boundaries of the word in an utterance does not rely in any way on the meanings, only relies on the sequences of sounds—of phonemes or letters. In fact, correlating phoneme-sequences with meanings is not enough to determine that something is a word. For instance, the phonemes sl have a certain hard-to-distinguish meanings in a host of words, initially, as in slide, and slither, and slimy, and slosh, and slick, and so forth—in a large number of words;gl—again initial, so it is very easy to look them up in a dictionary—has a certain meaning, which again is some kind of common meaning, in very many words like glimmer, and gleam, and glow, and gloaming, and so forth. But we can't make words with them, because the regularity of their combination with other things, which in turn do not have to be separate words but at least have to be separate in some other respect, is not sufficient. There are even things—affixes in this case—which used to be words, I mean used to be affixes anyhow, like le, which appears in dazzle, and nuzzle, and juggle, and frazzle, and so forth, which indeed once was a morpheme. But it appears in such an irregular way, and has spread to so many other words where one cannot connect it with anything, that it would be very hard to define it as a morpheme in English grammar today.

There are also other difficulties, of course. The homonyms are very disturbing, and cannot be easily distinguished. The three words for sound—the three words for sound, "noise", and "healthy", and "a body of water", for instance—are all from three different sources, etymologically, and presumably they are different words, even though they have the same phonemes. But the two words will—or there are more than two, actually—one of them "futurity" and the other "intent" or "desire"—are not normally treated in dictionaries as a single word, they are too different from each other. Nevertheless, they go back to the same ultimate source, and of course they have a  very close relation in meaning nevertheless; it's only their grammar that is different.

Now, fortunately, the problem of deciding words is easier because words don't appear irrespective of environments; they always appear in environments, and characteristically in environments. And some of this was discussed in this matter of, first of all, whether they are operators or arguments of each other, and second of all with what likelihood they are such. The words have different meanings, very often, in different operator-argument combinations—under different operators. The word set has a different meaning connected with tennis, or with silverware, or with a theater set, with the stage. The word divide had virtually the same meaning as the word multiply if it is said of a cell: for a cell, to divide is to multiply. The meanings of words vary in different environments. This is not just an observation. It obviously has to do with structure in language, because words which have no restriction on what words they occur with—the ones which I have said have total high frequency, like the indefinites, which can occur freely under any verb—these do not have different meanings under different operators. So that, this matter of having different meanings under different operators is not just something that happens, because it can also be stopped by something else that has a grammatical character of the word combination.

Now, the meanings of words change, of course. Much of this change, however, is involved in the extension of a word into new combinations. We should therefore consider the extension of words into new combinations. This thing has been shown—more strongly by Henry Hoenigswald than by anyone else, but it is not an unknown thing in historical linguistics—it has been shown that the extension of a word to new combinations is determined not simply by the meaning of the word, but by the way the combinations of this word, in which this word has already entered, relate to the way the other words combine. In other words, it has to do with the niche which the given word already holds in the word-combinations of partially similar and partially related words. And related not only in meanings, but related by the kind of combinations that they commonly enter into. In fact, meaning alone would never suffice to determine how to extend the environment of a word. If we look at a great number of words which have moved into further environments, and so have gotten slightly different meanings, it will be clear that the meaning of the word before it entered into the new environment does not sufficiently explain why it got into that environment. The use of the word sail before it was used for zeppelins, would not quite account for why sail was used for zeppelins. The use of the word fly before it was used the first time for flying a flag—I don't know when that was; it's easier with zeppelins—the use of the word fly other than for flag does not readily explain why it came to be used for flag rather than various other words, and so on. Therefore, to go from meaning to the extension of selection, the extension of environments of a word, would not be very satisfactory. In contrast, we can go from the environment of a word to its meaning, but only from the meaning of the environment—meaning, if we know the meanings of the words with which a word occurs, we can guess very very closely what the meaning of that word is. In fact, it is sufficient to determine it precisely, more precisely as a matter of fact than anything else that can be done. And no dictionary can go into real detail on the meanings of words without giving examples of its use with other words. It is not possible to calculate, to figure out, the meaning of a word only by knowing what words it occurs with. It is possible to know the meaning of a word if you know the meanings of the words that it occurs with. But of course, those words in turn can be figured out with the help of that word [together with others], so that there is something circular here, but at the same time, the extension of selection, let's say, what combinations the word has, obviously plays a major role.