The theory that follows is more easily illustrated with examples than formulated in general terms. The difficulty with general formulations is that, for want of a precise quantification, vague phrases like much and a great deal have to be resorted to, and the result sounds banal. The interest will lie, then, in showing, by way of examples—of samples—that a proposition admitting only of a banal general formulation can have applications that are fresh and important.
It proves convenient to enter into the subject via a discussion of metaphor and metonymy.
Metaphor and metonymy are familiar concepts in diachronic semantics. In particular, Roman Jakobson has given them renewed prominence by taking up the thought of the Kazan School (Baudouin de Courtenay and Kruszewski, especially the latter) that metaphor and metonymy are the two basic steps in semantical change. For them, and for him, the thesis is deduced from the psychological theses that (a) the basic psychological process is association and (b) the two varieties of association are association by similarity and association by contiguity.
Of course one might adopt the linguistic thesis without intending any commitment to the psychological thesis that gives it a pedigree, and that is what I do here. Not that I reject the psychological thesis, but in order to make my linguistic points I do not need to take any stand on it. Neither am I committed to treating the class of metaphors and the class of metonymies as constituting a strict dichotomy, i.e., as two mutually exclusive and jointly exhaustive species of the genus-class semantical changes. All that I need for my present purpose is that (a) both classes are very large—in other words, that there are many occurrences of metaphor and many of metonymy—and (b) the classes are mutually exclusive—in other words, that no occurrence of metonymy is also an occurrence of metaphor. Joint exhaustiveness is not necessary for my purpose; there may be semantical changes that are neither metaphorical nor metonymic.
It will turn out that many occurrences of metaphor and of metonymy are hypothetical; not that they could not be observed, but that (relative to the body of data with which we are working) they have not been observed.
One last preliminary remark. Jakobson uses ‘metaphor’ and ‘metonymy’ in a sense considerably wider than what is usual. With ‘metonymy’ this does not matter very much, since the term has no currency outside of linguistics and is not much used even there. With ‘metaphor’ it is quite different. ‘Metaphor’ has both a well-established ordinary or popular sense and a well-established sense among literary critics; Jakobson’s sense is different from both of these. Here I need do no more than call attention to the fact that there is a difference.
Between metaphor and metonymy (taking these terms in the broad sense given them by Jakobson) there is a difference of great importance to the diachronic linguist: metonymy much more than metaphor is bound to an extralinguistic situation. Most metaphors can be understood fairly well without knowing anything about the circumstances, i.e., the extralinguistic situations in which the process of metaphor occurs, whereas with metonyms a knowledge of the circumstances is much more often required.
There are exceptions on both sides; I speak only of prevailing tendencies. A few examples will show what I mean.
Waiter. In the literal or etymological sense, anyone whose profession or occupation or habitual practice it is to wait is a waiter. (This involves one of the two meanings of the English suffix -er; the other meaning, inanimate, applies to instruments; as it happens, English uses this meaning too in forming a derivative from wait. Because -er1 and -er2 are homonyms, we would correspondingly treat waiter1 [in a restaurant] and waiter2 [occurring only in ‘dumb-waiter’] as homonyms.) In actual usage, waiter is used in a narrower sense. This is because, in actual fact, in recent Western culture, the only people who waited as a matter of occupation or of habitual practice were (with the exception of those who waited under other circumstances, but who were not called waiters because they were called by other names, such as valet) people who waited ‘at table’, i.e., who waited around while other people ate. The narrowing from literal to actual meaning would count as metaphor in Jakobson’s sense, though not in the popular sense or in that of literary critics. The reason is that narrowings of sense just as much as widenings are founded on similarity. Waiters on table would be found in the houses of the wealthy. In private clubs, public restaurants, and the like, the ratio of guests or patrons to waiters would be higher, perhaps so high that the patrons would wait for the waiter as much as the waiter would wait on the patrons. Nevertheless, there is a striking similarity—namely, a similarity of function—between the waiter (flunky, garçon) in a great house and the employee in a public restaurant who carries in food and carries off empty plates; and the sense that will be found current in the United States in the 1970s and the literal sense, whenever and wherever it was first employed, are related as wider and narrower, respectively. It follows that the step from the original literal meaning of waiter 1 to the narrower meaning ‘waiter at table in a great house’ as well as the step from this to the wider meaning now current in the United States both count as metaphor.
The foregoing account has, at every turn, mentioned circumstances, i.e., extralinguistic situations. The point I have wished to make is not that no reference to the extralinguistic situation is needed when the linguist is describing metaphor; it is, rather, that appropriate situations are readily imagined, so that special Wörter-und-Sachen research is not needed.
There are hard instances of metaphor and easy instances of metonymy, as a few examples will show. Bloomfield (1933:400 §22.6) quotes Othello III, 3, in which Desdemona is compared with a hawk. A reader today is completely baffled by the passage because he does not know (a) anything about falconry or (b) the technical terms used by Shakespeare. If his second-mentioned ignorance were corrected, his first ignorance would still prevent him from making sense of Shakespeare’s metaphor. Similarly a person must understand a game (petteia, a board game, and astragalos, a dice game, for the ancient Greeks; baseball for present-day Americans) to understand metaphors based on that game.
On the side of metonymy, the polysemy by which one and the same word signifies (i) the act of doing something that is relatively transient, perhaps momentary, and (ii) the result, relatively enduring, perhaps even permanent, of the act—the polysemy that Ogden and Richards (The Meaning of Meaning, 1930:134) eccentrically style the Utraquistic Subterfuge—is easy to understand, because the situations that give rise to it are so ubiquitous; the ease of understanding results from everybody’s familiarity with the extralinguistic situation.
Many euphemisms are based on metonymy, e.g., loins for genitals and bathroom for the room used not only for bathing and washing but also for eliminating. Insofar as euphemism is a response to a verbal taboo and insofar as a verbal taboo is widespread (cross-cultural), the extra-linguistic situation will be self-explanatory.
Here are some examples of things that in universal experience go together: (1) ‘Right and skillful’; ‘right and straight’. The etymology of Latin dexter, etc., reveals the former and that of English right, etc., reveals the latter rhetonymy. The basis in universal experience is that in all societies a heavy majority of the population are right-handed. (2) Latin vertex is manifestly derived from vertere (‘turn’). In the sense of ‘top of the head’ and ‘peak of a hill or mountain’—even more clearly with the latter—it is obvious that this sense is metonymically connected with the literal sense ‘turning point’: If you are passing across a hill, the turning point is where you stop climbing up and start climbing down. Even if (Nida, 1951:197) the people of Yucatan are unacquainted with hills, they are acquainted with pyramids and with gravity, and so with the experiences that underlie the metonymic step from ‘turning point’ to ‘summit’.
In all these situations, a semantical change can be understood by one who correctly supposes the situation and not by anyone else, and correct supposition may be caused by direct experience of similar situations, or by a vivid imagination (including empathy), or by a lucky guess, or no doubt in other ways. It is an empirical fact that, broadly speaking, more metaphors than metonymies are intelligible (durchsichtbar, as German felicitously says) without special instruction; the explanation must be that (speaking again in broad terms) the felt similarities on which metaphors are based are fairly universal, in marked contrast to the contiguities on which metonymies are based. Some contiguities are more universally, others more restrictedly experienced, and—here is the salient point—a meaning-change based on a more restricted contiguity is nearly as likely to win eventual currency as one based on a more nearly universal one.
We can be more specific. To understand a semantical change, say the change of expression e1 from having sense s1 to having sense s2, one must understand both s1 and s2. One circumstance in which we would say that so-and-so did not understand e(s1) is that in which he was simply unacquainted with the referent; another is that in which he was acquainted with the referent but did not know that it was the referent of e(s1). (It sounds self-contradictory to say that he was acquainted with the referent of e(s1) but not with the fact that it was the referent of e(s1); to make it clear that it is not a self-contradiction, the medievals would have employed the concept of ‘insofar as’ (Latin qua or inquantum), and said that he was acquainted with the referent of e(s1) but not qua referent of e(s1). Present-day philosophers, notably W. V. Quine, would put the same point in terms of the failure of substitutivity in the case of ‘opaque’ reference.) An example: Most Americans today do not know that Homer’s word knemis, and the now obsolete English word greave, have approximately the same sense as current American English shin guard (and also as the word puttee, which was used in World War I). This is the situation obtaining when a hearer does not understand a foreign language (or a foreign dialect) whose grammar is the same as that of his own language or dialect. For in that case the differences in question are purely lexical. All normal speakers/hearers of English have experienced children, but relatively few know that in German the word to express that experience is Kind. In the example from Othello, the obstacles are partly linguistic and partly nonlinguistic. They are more linguistic than nonlinguistic, as is shown by the fact that if Shakespeare’s words were translated, or paraphrased, even roughly, into modern English, an average speaker/hearer—who in this day and age would not have any direct experience of hawks and falcons, except perhaps from watching educational television, but would have direct experience of various other birds (pigeons and sparrows in Central Park)—would be enabled thereby to have a pretty good grasp of the purport of Othello’s extended metaphor.
It may already be plain from these few examples that there is a continuous gamut from utterly commonplace and familiar experiences to extremely unusual ones. One would expect this gamut to be reflected in a gamut of degrees of ease with which a student of semantical change can guess (surmise, conjecture, hypothesize) the situation in which a given change took place. And my proposition that by and large the circumstances of metaphors are comparatively easy to guess, those of metonymies comparatively hard, would seem to introduce a dichotomy, a gulf, a difference of kind where the general considerations given above would lead us to expect a continuum, a difference of degree.
If there should be any clash I would withdraw the proposition and stick with the general considerations. The proposition has no basis but low-level induction; it reports an impression gained from study of cases, but will not be insisted upon. But one of the points I hope to make is that experience, effort, and attention can increase a student’s ability to make good guesses, such as he can make about metaphors, even without these three inputs. However, even with metaphor, experience, effort, and attention make him able to do better than he would without them. Let me give an example or two.
I will define two types of problem. Suppose that there is a general theory of semantical change such that every problem, of whichever type, is the problem of finding one or more explanations that explain in terms of this theory; in other words, one or more explanations that apply this theory.
In both problem types, it is a datum that a certain word or expression, e1 has senses s1 and s2, and in both types the problem is to find one or more explanations, applying the general theory, that explain the datum. (It may be as well to caution the reader that the word datum does not signify that the proposition so called is true, but only that it is functioning as a premise in reasoning.) The types differ in that in Type I, the datum includes a proposition relating the two s’s chronologically, i.e., a proposition telling us that the word has the first-mentioned sense before, after, or at the same time as it has the second-mentioned sense. In Type II, by definition, the datum does not include information of relative chronology.
It is obvious that solutions of Type II problems are easier, and (for the same reason) riskier, i.e., more likely to be refuted, than solutions of Type I problems. Type I has all the data of Type II, and more; so Type I imposes an added constraint in advance of explanation, and Type II offers more possibilities of constraint after explanation, some of which will be incompatible with the explanation.
In terms of these two types, I can describe some differences between metaphor and metonymy. Each type can be divided in various ways; consider subclass I-P, defined as the subclass of I subject to the condition (=constraint) that the explanation is explanation by metaphor, and subclass I-N subject to the condition that the explanation is by metonymy. (The acronymous letters P and N appended by hyphens to the Roman numeral I are meant mnemonically, to recall the words metaPhor and metoNymy respectively; the ordinary English pattern of forming acronyms would yield the same acronym, M, for both.) Correspondingly there are subclasses II-P and II-N. One more refinement—we want to be able to compare problems in pairs, of this sort: The two members of each pair are the same, except that the second member includes some extralinguistic data. Using E as an acronym for extralinguistic, we may describe each such pair of problems of Type I as consisting of a member of Type I-P and a member of I-P-E, and similarly for the other three types. These expressions as I am introducing them here are loose, not rigorous; e.g., the expression I-P, in the absence of any mention of E, is intended to signify that the datum does not include any extralinguistic propositions, but it would take a very elaborately constructed formalization to make a plausible version of such a supposition available; in lieu of that, I will simply explain that I mean not ‘absolutely nothing extralinguistic’ but ‘no special extralinguistic propositions’.
Now I can state in these terms what I want to say. Given a pair I-P, I-P-E, the problem I-P-E is not appreciably easier to solve than the problem of I-P, whereas given a pair I-N, I-N-E, it often is appreciably easier. Extralinguistic data aid the solution when the sought connection is metonymical more than when it is metaphorical.
My formulations above, though loose in some respects, are meticulous about not assuming that the data are true. The requirements of the formulations are as well satisfied by false data as by true ones. Then we can treat truth as a simply added condition. According to standard philosophy of science, an explanation is (provisionally) regarded as true insofar as (1) it is able to explain data that are themselves (provisionally) true and (2) it is not in rivalry with another explanation that satisfies requirement (1). The question of what to say of explanations failing condition (2) need not concern us here.
The question was before us whether the proposition that metaphors are easier to see through than metonymies without extralinguistic help could be reconciled with the uninterrupted continuum from universal to extremely unusual experiences. In answering yes, I lean on the fact that similarity (likeness, resemblance), which underlies metaphor, is divisible into levels, such that a person who does not himself feel a similarity between A and B may yet in certain cases by understanding higher-level similarities conjecture that someone else may feel A and B to be similar; and I lean also on the fact that there is far greater agreement, among all humans who have experienced both A and B, as to how similar A is to B, than there is as to whether A occurs together with B. The philosopher’s distinction between a priori and a posteriori knowledge is involved here: similarities between experienced things are not themselves merely experienced, as contingent, but are known a priori, as necessities; whereas contiguities, in time or in space, between objects of experience are themselves merely experienced, i.e., are learned by experience.
An interesting example of an opaque metonymy is offered by the phrase kick the bucket. The explanation given in the New English Dictionary, s.v. bucket (substantive, 2) tells us that in Norfolk the beam from which pigs are suspended by their heels for slaughtering is called a bucket; reasoning from this lexicographical, linguistic datum we might readily go on to the hypothesis that the animals in their death throes literally kick the bucket (in this to us new sense of bucket) to which they are tied, and that the contiguity of the events of struggling in this way and of dying is the extralinguistic basis of the metonymie step from the literal meaning of kick the bucket (i.e., ‘kick the slaughtering-beam’) to the meaning of dying. There is still much remaining to be told: how and when, in Norfolk or wherever, kick the bucket subsequently came to be transferred, this time by metaphor, from the dying of pigs in the slaughterhouse to the dying of any animal, including human beings, under any circumstances; how long the transfer took (years? decades? centuries?); whether, at any stage in the transfer, there was emotive meaning involved—such as a feeling that human beings are in this regard no different from brute animals: they both go to their death unwillingly. All this remains to be told, and the hypothesis-forming linguist is no better able to supply the details for a metaphor than for a metonymy.
But there is room for some skepticism as to whether the NED account is linguistically adequate. Assuming that (apart from questions about the adequacy of the way in which linguistic facts are distinguished from extralinguistic ones) the account is correct in its facts, one can still detect, just from reading that account, the amateur’s mistake of failing to distinguish the intension of a word, which governs its range of possible uses, from the set of properties characterizing that specific situation, just one of its possible uses, which gives rise to the discussion. We are not told whether, at the time and place in question, bucket was applied to a beam whether or not it was used for slaughtering. If bucket was simply the Norfolk dialect word for beam, then learning this lexicographical fact would not help one to pass from the meaning ‘kick the beam’ to the meaning ‘die’; the metonymy would be, as I said, opaque. And if in that dialect it meant specifically ‘beam used for slaughtering’, so that the lexicographical fact itself supplies the linguist with the basis for metonymy, then the linguist would want to press back to the anterior metonymy that is obviously presupposed.
I will now define another problem type, illustrating it with the word fine. In present-day English it is not obvious whether fine1, as in ‘before signing the agreement, be sure to read the fine print’ and in ‘a fine-tooth comb’, is connected with fine2 as in ‘it’s a fine day’, whatever we mean by ‘connect’. The diachronic problem is, then, whether to hypothesize for every stage of English (prior to the present) in which fine1 and fine2 occurred that they were mere homonyms—the same in sound, but unconnected—or whether, the other chief possibility, to hypothesize that at some stage they expressed different but connected senses of one word. To put the point less formally, were they always unconnected or were they once connected?
(The entire discussion of the problem will be relative to a theory. I am not saying that there is no occasion for discussing theories—the present lectures are just such an occasion—but only that by a fiction or idealization I will assume that in the problem types as I set them up, the general theory remains constant, and the only question is, what data to apply it to.)
In problems of Type III, which I am now introducing, the datum includes less in one way and more in another way than in Type I; also, less and more than Type II. Types I and II dealt with one later sense coming from one earlier sense. Type III deals with two later senses and one earlier sense. In it, the problem is: Given two contemporaneous senses of what is either one word or two distinct but homonymous words, is there an earlier sense with which each of the two senses later than it but contemporaneous with each other can be connected by the theory?
Type III, like Types I and II, is defined in terms of single-step connections. Naturally, complex types have to be acknowledged; for example, the semantical change culminating in the present-day phrase kick the hucket is, according to the account adapted from the NED and given above, a complex of at least three component simpler steps: (1) metonymy from ‘kick the beam’ (while tied to it and being slaughtered) to ‘die’ (when tied to the beam and slaughtered); (2) metaphor from ‘die by slaughter in a slaughtering-house’ to ‘die’; and (3) misunderstanding of bucket (‘beam’ in Norfolk dialect) as bucket (‘pail’ by speakers of standard English). The present lectures do not attempt to work out a list of either the possible or the useful complex types.
Offhand there seem to be in present-day English two adjectives fine, with no synchronic connection detectable: fine1, as in ‘fine print’, ‘fine-tooth comb’; and fine2 as in ‘fine day’. Historically (diachronically) considered, either they were always mere homonyms or else there was a time when one of them arose from the other, or both arose from some third word. To restate slightly, either there was never a time or else there was once a time when they had a common source (either one of themselves or a third thing).
The two alternatives involve a point of logic and a point of scientific method. The first alternative is a universal, the second a particular ( =existential) proposition. A universal proposition is inherently incapable of decisive, conclusive verification by induction, whereas an existential proposition is capable. This consideration dictates the heuristic order that, as between the two alternatives, we try first to verify, i.e., either to confirm or to disconfirm, the existential proposition, and, if we fail in this, turn by default to the universal proposition. The decision when to conclude that our effort to confirm the existential proposition has gone on long enough, and that the time has come to give up, does not admit of being subjected to any rule. And it itself is equivalent to a universal proposition, because equivalent to If I were to search further for a confirming instance, I wouldn’t find one’, which is in turn equivalent to ‘None of the instances not yet examined by me is a confirming instance’.
Suppose in our investigation of fine we have reached the point of posing our problem thus: Can fine1 and fine2 be connected by metonymy? To answer affirmatively is to hypothesize, generically, that there is a connection, and, specifically, that the connection is metonymie. It is not often that the linguist has at his disposal evidence sufficiently detailed to let him confirm such a hypothesis conclusively. But often he can confirm it with less or with more probability. The first point I particularly want to emphasize here is that if there is a substantial collection of pretty well confirmed hypotheses, all applying the same general theory, then these confirmations strengthen one another and confer probability of confirmation on the general theory itself. My second point of emphasis is that experience with a collection of pretty well confirmed hypotheses that hypothesize metonymy has heuristic value for the linguist in teaching or cultivating the knack of guessing where, if at all, situations appropriate for metonymy would be found.
With the word fine, here is such a situation. Cloth that is fine in the first sense (‘fine-grained’, ‘made of fine threads’) is also fine in the second. For most purposes, cloth of finer grain, i.e., with more threads to the inch, woven of finespun thread, is more desired than cloth of coarser grain, i.e., with fewer threads to the inch. It is also more expensive, partly because the labor cost in its production is greater. Here, then, are three properties associated by contiguity: (1) fine-grained; (2) desired, prized; (3) expensive. I say no more about (3) because it has not affected the diachronic semantics of the word fine. Property (1) underlies the sense of fine1 and property (2) of fine2... If fine1 and fine2 are connected historically through this contiguity, their connection is metonymic.
We find here a typical situation in which metonymy occurs. It may not be the only situation, but still the situations (all taken together) in which (1) and (2) are both present make only a very small subset of the situations, however we may enumerate them, in which human beings find themselves. The task confronting the diachronic semanticist, when he is trying to verify the hypothesis that fine1 and fine2 are connected in one step by metonymy, is to find situations belonging to that subset, however small it may be. He can say a priori, before searching, that if there was metonymy there must have been such a situation.
Metonymy, however, is not a process. We may loosely call it a step, but if we examine what we mean we will find it more appropriate to explain that we mean the result of a process rather than a process.
The theory proposed in the present lectures is that metonymy is a result of a process of misunderstanding. It involves acts of communication in which S (the speaker) intends his utterance U in one way and H (the hearer), hearing U correctly, takes it incorrectly: mis-takes it, misinterprets it, misunderstands it. In conformity with prevailing theories about speech and language, I assume further that a change in the language is some sort of result or effect of a large number of individual changes. In the simplest possible case, only one S and one H would be involved; S would speak of fine cloth, intending (=signifying), as was customary up until then, cloth with a high thread count (by the way, note that even today, spooled thread is identified as to thickness by the number of threads it takes, laid side by side, to make a width of one inch; thus, Number 12 is a coarse thread, suitable, say, for sewing on overcoat buttons; Number 50 is very fine, suitable for mending a tear as invisibly as possible); H would hear S, would be with him in the extralinguistic situation of examining cloth—S is, perhaps, a salesclerk, and H a prospective buyer; would be unfamiliar with the word fine; would judge from S’s gestures, expression, and manner that S admired the cloth and expected that H would admire it, and would (silently, to himself) advance it as his first hypothesis about the word fine that S (1) was using the word in the way standard with those who were familiar with it and (2) intended to signify by it that the thing called ‘fine’ was desirable and desired.
Here are two ways of describing what happened, (i) For S, the property (1) fine-grained is central, (2) desired is marginal; H takes (2) to be central, and takes (1) as marginal if he considers it at all. (ii) For S, (1) is figure, (2) is background; for H, (2) is figure, and (1) is in the background, if present at all. Description (i) is the way that Bloomfield (1933:149 §9.8; 430-31 §24.4; 440 §24.7) adopts from Hermann Paul; description (ii) is cast in terms of Gestalt theory. I believe that the reason why (i) and (ii) are so close is that (i) was gotten by Paul from the proto-Gestaltist Wilhelm Wundt, whereas (ii) comes from full-fledged (Koffka; Köhler) Gestalt theory. There is, however, beneath the surface, another difference. For Bloomfield, the difference between central and marginal is quantitative: the more frequent (for a given individual) is central for him, the less frequent is marginal; Bloomfield, in the behavioristic manner, views centrality and marginality summistically, in marked contrast to the gestaltist view.
As I will soon explain more fully, my own theory, if it has to choose between a more summistic and a more gestaltist theory, will choose the latter, because it assumes that in a single act of communication a misunderstanding can take place that eventually leads to metonymy. However, my theory does not accept either of the two descriptions presented above, but rather a third one. Before I give it, let me elucidate my imaginary account of how fine might have acquired the meaning ‘desirable’. H, I said, would silently and to himself advance a certain hypothesis. Experience has shown that such a statement regularly draws certain objections that I wish to forestall. There are those who will say that H does not advance a hypothesis, or ratiocinate, or infer, or anything like that; H just goes ahead and takes fine in a certain way. For such people, it is part of the meaning of such expressions as ‘infer’ and ‘advance a hypothesis’ (=’hypothesize’) that the acts they apply to are deliberate, or, at the very least conscious, so that such a phrase as Helmholtz’s ‘unconscious inference’ is a contradiction in terms. There is a way of neutralizing such people without going into the controversy. It is to insert the modifier virtually, or its synonym in effect. The result of virtually doing x is the same as the result of actually doing x; and here it is only the result that I care about, not the means. The inferential steps I hypothesize are virtual steps; as long as the result is the same, it does not matter whether it is a native user of the language who—as it were—goes through them almost instantaneously, or someone learning the language as a second language who goes through them laboriously, or a scientist (linguist) studying the language who goes through them analytically, or whoever.
This account has considered a single episode of communication. We have reason to think that it is very rare that a single episode effects a change in a language. With the word fine, the likelihood would be that a large number of episodes was involved. We may use the concepts ‘microscopic’ and ‘macroscopic’ to describe the relation. A change in a language is a macroscopic event, being an aggregate of microscopic events. We only have occasion to invoke the microscopic-macroscopic contrast when the number of aggregated events is quite large. (In language phenomena the number is much smaller than in physical phenomena.) In some fields, one has been able to determine macroscopic minima; for instance, in perception, there are just-noticeable differences. Linguistics has not as yet established minima of macroscopic semantical change; the present lectures neither attempt to determine such a minimum nor assume that there are any. Most work in diachronic semantics—indeed, most work in all of diachronic linguistics—is simply observational, not experimental; each macroscopic change is a brute given, not subject to direct experimental modification. It is true that, starting, say, from the data in the NED, we may propose an explanation that hypothesizes certain intermediate stages, and then, looking further, we may find new documents—previously unpublished diaries, say—that confirm or disconfirm our hypotheses; but although these new documents shorten the time interval, they will never or hardly ever be so close together but that still shorter directly observable intervals must have existed. The sparsity of data in his sample is a basic fact confronting the diachronic linguist, which his methodology must take due account of.
Much remains to be done with the linguistic theory of reducing macroscopic to microscopic events, but even with presently prevailing theory we may say what I have said already, that with very few exceptions, a macroscopic linguistic change is not caused by a single microscopic change, or even by a small number. In the present example there must have been a number of episodes, numerically distinct but qualitatively alike, and thus a number of hearers who misunderstood the intention of their respective speakers as regards the word fine. Let us consider further whether it is plausible that there should have been a number of such episodes.
If fine was a common word, how would there have been a number of such hearers? If we suppose that they were children, this commits us to supposing that a children’s misunderstanding became current; Jespersen (1922, Book II) has urged this commitment, but he leans heavily on rather general arguments, with little hard support. If it was adults who misunderstood, is it that they were simply unfamiliar with the word? There is another possibility: In each episode of the set that caused the semantical change, H heard the word fine and was familiar with it in the sense ‘fine-grained’, but he inferred that S, using the word fine, must be intending it in some other sense. S intends ‘fine-grained’, H is familiar with ‘fine-grained’, and yet H thinks that S does not intend ‘fine-grained’. My next task is to make this comedy of errors plausible.
Consider a customer looking at bolts of cloth or articles of clothing. One bolt, or one jacket, is made of coarse-grained, the other of finegrained, cloth. The clerk or salesperson first shows the customer the coarse-grained cloth, then shows him the fine-grained cloth and says: ‘Now here is a very fine cloth’. The clerk means more than he says (pregnant meaning’). All he says is that the cloth is fine, but he means both this and, additionally, that it is also desirable and good. The customer, H, hearing the clerk, S, say this, and supposing (=inferring; here is where the misunderstanding enters) that S is saying all that he is meaning, is obliged (in order to back up his supposition) to infer further that fine has a different meaning for S than it has for H himself. He does not need to be told that the cloth is fine-grained; he can see that for himself. If, then, S takes the trouble to mention it, it must be that S is telling H something that S could not see for himself. H’s inference here is not wholly wrong, but it is partly wrong.
It may be that H made a linguistic mistake as well as a nonlinguistic one. It is not universally true that what is fine-grained is good. In some cases it is, in some cases it is not. We would not consider a fine-toothed comb or a fine file either better or worse, speaking absolutely, than a coarse-toothed comb or a coarse file; it is better for some purposes, worse for others, and indifferent for others. Sugar and salt can be ground coarse or fine (sugar finer than salt); for most culinary purposes it does not matter, within a certain range, whether the grain of one’s salt and pepper is coarse or fine. With nails, pins, and needles, the situation is different: it generally does matter whether they are coarse or fine. And as for cloth, if the customer H did not already know that with fabrics fine is considered better than coarse, then naturally the sous-entendu of the clerk would escape him; and so, mistakenly failing to infer that there is a sousentendu, and confronted with an utterance that (if fine be assigned the meaning familiar to H) would seem incomplete, H virtually considers assigning it a sense that would make the utterance complete.
The utterance would seem incomplete in the respect that something more would need to be said to give it point. By itself, as it stands, it seems pointless. The ordinary-language phrases by itself and as it stands are analyzed in linguistics with the help of the concept of the discourse. The discourse is the next unit above the sentence. Morphemes compose words, words compose phrases, phrases compose sentences, sentences compose discourses. The possibility exists that a discourse will consist of a single sentence, a sentence of a single phrase, and so on. In an episode of the sort I am now imagining, ‘Now here is a very fine cloth’ occurs as a one-sentence discourse. If H takes this sentence in the way that would first occur to him, he will find it pointless because it tells him what he can see for himself, and he will be surprised that S does not follow it (or precede it) by one or more other sentences, in the same discourse, that would make clear what its point was. And, surprised, he will virtually infer that the way in which it first occurred to him to take the sentence is not the way S intended it.
Now H might be making the nonlinguistic mistake of not knowing that fine cloth is generally considered to be desirable and good cloth. (If anyone objects to calling simple ignorance a mistake, let him find some other ordinary and convenient label for it.) But he might know this and yet misunderstand S. For even if he knew this, he might suppose S to be saying all that he meant; and this supposition, this hypothesis, this inference, would be a mistake, and of a kind that we can fairly call a linguistic mistake. The nonlinguistic mistake makes the linguistic mistake more likely, but is not a sine qua non. H might think it was pointless for S to tell H something that H could see for himself, and yet not think it was pointless for S to tell H something—viz., that the cloth in front of both of them was desirable and good—that would follow deductively from something (that cloth is fine-grained) that H could see for himself when conjoined with something (fine-grained cloth is desirable and good) known to H to be the general opinion. In that case, the misunderstanding on H’s part can be described as follows: S treated the connection between fine grain and desirability as going without saying, and H did not realize that S was treating it that way. Note that it is irrelevant to H’s misunderstanding whether or not he would treat the connection as going without saying; what is in question is not whether H agrees with what S is doing, but whether he is aware of what S is doing.
A last point about fine. The linguistic mistake—overlooking S’s sousentendu, i.e., supposing that S said all he meant—entails a second linguistic mistake on H’s part, viz., ascribing to fine a new sense. Conceptually, this second linguistic mistake is a distinct step from the first; and we must take note of it, because we will then go on to hypothesize that H, equipped now with a new sense of the word fine will on future occasions when functioning not as hearer but as speaker use this word fine (=‘desirable’), and sooner or later will use this new word (or this old word in a new sense, whichever it is for him) in a situation, such as when speaking about the weather, where it would never occur to him to use the old word. When that happens, the new word is launched. Whether the newly launched word floats or sinks is another matter.
Using the example fine, I have sketched a theory of processes causing metonymy. The example can be used to illustrate both why and how metonymy differs from metaphor as regards intelligibility. To guess successfully the metonymic connection between fine1 and fine2, it is not enough to know the signification (=connotation) and the denotation of fine1 (i.e., what it means and what things it would be predicated of) and the signification and the denotation of fine2; one must know where the two denotations overlap. Of course with perfect knowledge of the denotations one would know their overlap, but one might have a fair, ordinary knowledge of their denotations and yet a poor knowledge of their overlap. It is part of the training of the diachronic semanticist to become skilled at finding, for any two given words with possibly overlapping denotations, whether there is any probable overlap.
We may now resume the discussion of the microscopic-macroscopic relation. So far I have shown (not with high probability but with initial plausibility) some conditions under which metonymy could take place. But this does not suffice to show that it would and did take place. Supposing that the circumstances for fine were more or less as I imagined them, it remains to be shown that these repeated misunderstandings of fine1 (used with a sous-entendu) as fine2 (used straightforwardly) would become current in the language. The standards for adequate confirmation are not as high in linguistics as in the physical sciences, but still there is a difference between a merely possible and an adequately verified hypothesis. In the present lectures I do not address myself to that part of a hypothesis that is concerned with thorough testing, but only to that prior part in which the hypothesis is put forward and is given a first, preliminary testing to see that it has enough plausibility and promise to be worth testing further.
One of the uses of the English noun will is its legal use, as in ‘This is my last will and testament’. In connecting this with the use as in ‘It is my will that my property be given to my church’ we may well make reference to Charles Peirce’s distinction between a Type and a Token. The relation between will1 (as in ‘It is my will that. . .’) and will2 (as in ‘This is my last will. . .’) is not that between type and token, but is derived from it. When we speak of the Bible, we sometimes mean the type and sometimes a token. If someone wants to take a solemn oath and says ‘Bring me a Bible to swear on’, he means a token, as also in the colloquial ‘I wouldn’t believe him if he swore to it on a stack of Bibles’. If he says ‘I want to take a solemn oath; bring me the Bible’, he may mean a token, for example he might be referring to the large Bible kept on the parlor table, which is not the only copy but is the largest copy in the speaker’s household, or he might mean the type, without caring to stipulate any particular specimen (any particular token) of that type. The only way to bring a type to a person is to bring a token of it to him, but if the requester means the type and not a token, then any token of it would do.
The type-token relation is a general pattern, freely extendible. Therefore we cannot ask which came first, the type or the token, as we can ask which came first, fine1 or fine2. Every word that signifies a type signifies a token and vice versa, and if a sentence containing such a word is unambiguous, it is because of the other words in it that determine whether, as used in that sentence, the word is intended to signify a type or a token. For some words, it will have happened that the first speaker to use that word will have intended a token by it; for others, the type; but these are merely new instances of the pattern that yields for every type a token (more precisely, a word signifying a token) and for every token a type. The origin of the pattern does not concern us here; the pattern is found in ancient Greek, and may have diffused to the speakers of English or may have arisen independently among them.
Besides types and tokens, there is a third mode, in the mind, otherwise described as in the memory (the tablets of the heart, Proverbs 3:3, 7:3; Jeremiah 17:1; 2 Corinthians 3:3), where Jews and Christians carry the Ten Commandments and Christians the Lord’s Prayer. Any recitation, silently to oneself or aloud, is a token, but the ability (disposition) to produce a token at will is the third mode of existence.
Now thanks to this pattern, one may say ‘I read in the Bible that. . .’ or 1 read in the Bible I carry around with me that. . .’; these are not quite the same, because the latter suggests the possibility of discrepancy between one copy of the Bible and another, but otherwise they are the same. With will, the facts are appreciably different. One’s will in the legal sense is not any token of a type; it is a certain singular document, standing in an asymmetrical relation to copies of it. Speaking of copies of it, we would say of any of them that it is a copy of the will but not, vice versa, that the will is a copy of it, nor that it and the will are copies of the same thing. The status of a will is rather like that of a holograph or of an autograph manuscript in that it is epistemically unique, with the further feature that because of its unique epistemic status it is also given unique legal status. The status of the Magna Carta and of the Declaration of Independence is similar.
There is, nevertheless, a type determined by any legal will whose tokens are (i) that will itself and (ii) all copies of it. The question I am considering here is whether there is a connection between will in the legal sense and will in some other sense. I answer with the hypothesis that there is a connection, that (given a certain understanding of ‘metonymy’) it is metonymic, and that the process causing this metonymy involves certain ambiguities.
Whether the connection between will1 as in ‘It is my will that. . .’ and will2, as in ‘This is my last will and testament. . .’, is metonymic depends on whether a thought and a document expressing that thought are contiguous. Most discussions of contiguity are vague on this point. They make it clear that it makes sense to say, whether truthfully or falsely, that two physical things are contiguous in space or in time, that two processes (whether physical or mental) are contiguous in space or in time, and that a physical thing is contiguous in space or in time with a process. But as to whether it makes sense to say that a mental thing is contiguous with a physical thing, they are silent. Without giving an elaborate discussion of the question, I think we may reasonably agree to call a thought, or other state of mind, contiguous in time and in space with any physical manifestation of it. Under that agreement, will1 and will2 are metonymically connected.
I had two reasons for introducing will as an example. One was that it raises a question about the meaning of contiguity, and so of metonymy. The other is that the process causing metonymy involves language in a way I have not hitherto mentioned.
I hypothesize as with fine an extralinguistic situation where an utterance is intended in one way by its speaker and is taken in another way by its hearer. But with fine, the hearer’s mistake consists of ignoring a sous-entendu. With will it must be of a different sort. What utterances could have been taken as referring either to will1 or to will2? Such utterances as ‘Who are the witnesses to his will?’, ‘Can you prove that that was his will?’, ‘He changed his will twice’. There are many other utterances that one would expect that could not be taken either way, at least not with equal readiness, and in attempting to carry my hypothesis beyond the stage of formulation (which is all that I do here) to the stage of testing it, one would tackle the question of whether utterances of the sort that could be taken either way would probably have caused will2 to become current in the standard language. In leaving the example, I will hazard two suggestions. One is that a legal will will commonly have at the head of it, and/or on the outside back or cover of it, the simple word WILL, which has just the ambiguity that we are looking for. The other is that no matter when the semantical change took place, even if it was in the Middle Ages, there was never a time when copies of documents were not made, so that at all times of English history when there was any literacy at all, there must have been a distinction between documents of which there was only one official copy—royal proclamations, letters patent, letters of identification, licenses, etc.—and other documents; and wills belonged to the former class. For all such documents, the contiguity between state of mind and manifestation of it—and therefore the possible metonymy between a word signifying that state of mind and the same word signifying a manifestation of it—would have been simple and straightforward, because it would have been one- to-one: within the class of documents that I am now speaking of, each state of mind would have at most one manifestation of it. The effect of this fact would be to make it unnecessary to speak of an instance or a specimen or a token of the manifestation, or even to speak of a manifestation at all.
We speak of the barometer as rising and falling. Many of us will admit that strictly speaking it is not the entire barometer but the mercury column englassed in it that rises and falls. But the fact is that we do not speak strictly. Let us consider the hypothesis that once people spoke only strictly about the barometer, and that the current loose speaking may be viewed as a semantical change.
The most important part of a barometer is the glass tube containing mercury. It is the most expensive part, the most fragile, and the hardest to repair. Moreover, it is the most indispensable part of the entire instrument. This fact is commonly expressed by various metaphors: it would be called the heart, the core, or the essence of the instrument. Most people who use barometers have only a superficial acquaintance with how they work and why they are designed as they are. The word barometer is Greek to them; if required to define it, they would say something like ‘a thing for telling what the weather will be’.
All this being so, a hearer H may be very familar with the word barometer and yet very shaky in his grasp of it. Familiar, in that he has heard it a hundred times, and yet shaky, in that these hundred times have left him with a wide latitude of interpretation. In an extreme case, all hundred occurrences might have been merely in so many different tokens of the one sentence type, ‘Look at the barometer’. A case as extreme as this is unlikely, but if the hundred occurrences were all in one or another of the following sentence types: (1) ‘The barometer is rising/falling’, (2) Is the barometer rising/failing?’, (3) ‘Look at the barometer’, (4) ‘What does the barometer say?’, (5) ‘The barometer says... the case would be not at all improbable, and yet the hearer H with only these hundred occurrences of the word at his disposal would find himself presented with a considerable latitude of interpretations any one of which would fit his data. Whether we think of him as inferring, or supposing, or guessing, our explanation should reckon with the possibility, and should estimate the likelihood, that the interpretation he settles on is not the interpretation the speaker intended; in that case it will be a misintepretation, or misunderstanding. The misunderstanding will not come to light as long as H uses the word only in the sentence types he has heard, or in certain others, but only when he uses it in sentence types such that the word as intended by the previous speakers would not have been used in those sentence types. (It would have been meaningless there, or meaningful but somehow wrong.) The same will be true, for example, for sentences such as The barometer has a gauge next to it’, where under older usage the proper thing to say would be ‘The barometer includes a gauge next to its mercury-filled tube’, or the like.
Examples could be multiplied.
Bloomfield, Leonard. 1933. Language. New York: Holt, Rinehart & Winston.
Jespersen, Otto. 1922. Language: Its Nature, Development, and Origin. London: Allen and Unwin.
Nida, Eugene. 1951. A system for the description of semantic elements. Word 7:117-99.
Ogden, Charles K., and Richards, I. A. 1930. The Meaning of Meaning. 5th ed. New York: Harcourt, Brace and World.