THE STATUS OF LANGUAGE

What is the status of language in this sense in linguistics? How are we to understand it? Is it an approach that we adopt as linguists, an optional way of viewing the evidence like the scientific approach? Is it a definite object that can be studied scientifically? Is it an aspect of people (we can now ask), a possession of individuals or of groups? Is it an abstraction or model? Is it a theory based on evidence and subject to test against further data? Is it a special assumption that needs to be examined? All these views and more can be found in the literature.

Saussure appears at first glance to have embraced several of these different views in different places. But in a passage that is perhaps his most important contribution to our present topic, he remarks:

Other sciences work with objects that are given in advance and that can then be considered from different viewpoints; but not linguistics. . . . Far from it being the object that antedates the viewpoint, it would seem that it is the viewpoint that creates the object. [1916:23;1959:8]

This is not just a passing remark, but something that Saussure had considered very carefully. The critical edition (Engler 1967:24–26) collates a number of columns of his unpublished notes with this single paragraph, and Saussure’s point can clearly be seen to mean that other sciences study objects (like rocks, trees, and animals) that are given in advance in that they have an independent external reality and can thus be studied from different points of view. The physical properties of rocks are studied in physics, the chemical properties in chemistry, others in geology, crystallography, and mineralogy. Each science brings its own point of view to the study of the same physical objects, which continue to exist when we move from one order of ideas to another. But the objects of study in linguistics do not have this kind of reality. “Someone pronounces the French word nu ‘bare’,” explains Saussure:

a superficial observer would be tempted to call the word a concrete linguistic object; but a more careful examination would reveal successively three or four quite different things, depending on whether the word is considered as a sound, as the expression of an idea, as the equivalent of Latin nudum, etc. . . . ; besides, nothing tells us in advance that one way of considering the fact in question takes precedence over the others or is in any way superior to them. [1916:23; 1959:8]

Saussure is certainly right about this. The objects of language do not have an independent existence of their own; they are actually created by the very point of view taken. Some might object that sound has an external existence of its own. This is true enough, but Saussure answers that trying to reduce language to sound would be to fall into a trap: sound alone would not be speech without the ideas with which it combines. Physical sound, without the linguistic system, is not a linguistic object at all.

Let us consider this point more closely because it contradicts the impression sometimes expressed that the objects of study in linguistics, such as words, meanings, phonemes, formatives, signs, or sentences, are indeed objects that are given in advance like the objects of study in other sciences, or, alternatively, that the objects of study in other sciences, like those in linguistics, are also created by a point of view. The issue concerns the ontological and epistemological status of objects in a science, and the relation of a theory to what it is a theory of.

Let us recall some well-known points regarding how objects are studied in science by taking water as an example. One can observe that water is a colorless liquid and that it freezes and boils. Scientists have measured the temperatures at which it freezes and boils, and how these vary with the pressure. They have measured the volume of one gram of the substance at different temperatures and pressures. They have measured how much light of different wavelengths in the visible, infrared, or ultraviolet portions of the spectrum passes through it at different temperatures and pressures. They have measured how much heat it takes to melt it, to boil it, or to raise its temperature by one degree.

All these measurements, and others, have been compared through the years with a developing body of theory dealing with the properties of water. Properties such as boiling point, melting point, heat of fusion, heat of vaporization, specific heat, density, absorption of light, and others are postulated. These properties are theoretical constructs that enter into tentative theories of water. The theories interrelate the properties and explain them in terms of a larger picture.

The theories are capable of predicting the outcome of further experiments, and this is crucial, for by this means they can be tested against reality and the results can be verified by different scientists in different laboratories, often with different methods. In this way doubts can be resolved and agreement reached. If the predictions are not borne out in careful experiments, the theory may have to be revised. In this way the theory of water has grown and has been integrated into a broader network of physical, chemical, and biological theories. At any point there is a current version of the theory of water, our scientific view of what water is, and there is also the real object water in the real world against which our developing conceptual structure can be tested.

It is the same in other branches of science. For example, concepts of mass, charge, spin, force fields, probability densities, and the like are in the domain of theory subject to testing against careful observations of electrons and their effects in the real world.

The objects of study in linguistics, however, although they are often seen also as having properties as theoretical constructs in a linguistic theory, have no corresponding real objects in the real world against which the theories can be tested. Basically the objects of study in the other sciences are real objects but in linguistics they are not. An elaboration of Saussure’s point is that in the other sciences theories can be tested against physical reality; in linguistics they cannot, because there is no corresponding physical reality antedating the viewpoint to test them against. Rather than being objects given in advance that can be studied from a linguistic point of view, they are actually created by the very viewpoint taken.

We can thus conclude that language is not something that can be studied scientifically in any strict sense, for its objects have no independent reality; and it is not an abstraction or model or theory of anything, for there is nothing for it to be an abstraction, model, or theory of. In his unpublished notes Saussure exclaims that the illusion of things naturally given in speech is profound. Yes. Linguistic objects still seem very real to us, but it is an illusion. Forgetting this important point can lead us into fallacies, as will shortly become evident.

It is to Bloomfield that we owe a clear recognition of the true status of language in linguistics—the study of language rests on a special assumption.

Bloomfield points out that the individuals in a human society cooperate by means of sound waves which bridge the gap between the nervous systems of speaker and hearer (1933:26, 28). Setting aside the mechanisms in the speaker and the hearer as problems for physiology or psychology, he says the linguist deals only with the speech signal (1933:32).

But the study of language can be conducted without special assumptions only so long as we pay no attention to the meaning of what is spoken (1933:75). This phase of language study is known as phonetics. The phonetician finds that no two utterances are exactly alike (1933:76). As long as he ignores the meaning of what is said, he cannot tell us which features are significant for communication and which features are immaterial. A feature which is significant in some languages or dialects may be indifferent in others (1933:77).

Then Bloomfield points out that the difference between distinctive and nondistinctive features of sound lies entirely in the habit of the speakers. In the case of our own language we trust our everyday knowledge to tell us whether speech-forms are “the same” or “different.” In the case of a strange language we have to learn such things by trial and error, or to obtain the meanings from someone that knows the language (1933:77).

Then comes a crucial passage:

The study of significant speech-sounds is phonology or practical phonetics. Phonology involves the consideration of meanings. The meanings of speech-forms could be scientifically defined only if all branches of science, including, especially, psychology and physiology, were close to perfection. Until that time, phonology and, with it, all the semantic phase of language study, rests upon an assumption, the fundamental assumption of linguistics: we must assume that in every speech-community some utterances are alike in form and meaning. [1933:78]

Bloomfield discusses this assumption extensively here and in other chapters. On it rests not only phonology but the whole discipline of linguistics.

Bloomfield’s assumption is actually a compound of several assumptions. Together they effectively create and introduce into the discipline just those objects of study constituting language that are not given in advance. To understand this, consider that for Bloomfield a speech-community is a group of people who use the same system of speech-signals, i.e., the same language (1933:29). Thus, “in every speech-community” introduces by assumption both speech-communities and their languages. An utterance is an act of speech, and alike reflects judgments of speakers about acts of speech (1926). Thus, “some utterances are alike” introduces by assumption a segmentation not inherent in the sound waves and creates distinguishable units of speech, or speech-sounds. Forms are vocal features common to utterances judged the same or partly the same, and the corresponding stimulus-reaction features are meanings (1926). Thus, “alike in form and meaning” establishes by assumption that these created units are signs in the traditional sense.

From this we see that Bloomfield’s fundamental assumption of linguistics introduces precisely the traditional point of view and creates by assumption the very objects of study in linguistics. Without the assumption we have only the infinitely varying sound waves; with the assumption we have the speech-sounds, forms, and correlated meanings of a language. Without the assumption individuals in a society cooperate by means of sound waves; with the assumption we see them as using a language.

Since language is widely accepted as the domain of study of linguistics, some such assumption, overt or tacit, is widely held in the discipline. Though the phraseology and details of different positions may differ, this assumption lies at the foundation of every grammatical position today. There is no doubt that Saussure and Bloomfield are correct in this matter. The objects constituting language are not given in advance. They are created by a special assumption.

PROBLEMS OF JUSTIFICATION

In a true science all special assumptions are subject to doubt. Is there any scientific justification for an assumption like Bloomfield’s creating the objects constituting language? Bloomfield offered no justification. Considering his orientation toward science and his careful use of evidence, this might seem surprising and out of character. He introduced it almost apologetically as a temporary expedient until that time in the future when the other sciences are close to perfection. Perhaps in fact such an assumption cannot be justified. But Bloomfield had been trained in the study of the objects of language, and he and his predecessors and contemporaries viewed their discipline as the study of language. In other words he had received it from a tradition, and in his careful way was only making explicit what had traditionally been assumed.

This tradition, which Robins (1958) called the western grammatical tradition, has been studied extensively by historians of linguistics. Bloomfield himself had reviewed it in the first chapter of Language (1933), where he called it simply the study of language. Saussure explicitly acknowledged it when he said that “linguistics, having accorded too large a place to history, will turn back to the static viewpoint of traditional grammar but in a new spirit and with other procedures” (1916:122;1959:82–3). Perhaps if we reexamine from the present point of view the history of this tradition in its several aspects, we can come to understand where an assumption like Bloomfield’s might have come from. We may then be able to judge whether or not it has any scientific justification.

THE PROCESS VIEW

The history of the study of language can be traced back at least as far as the ancient Greeks. Already in Plato we find historical changes of names treated in terms of adding, transposing, subtracting, or changing letters (Cratylus 394 B), an early process view that seems to assume that letters (sounds) are elements of language that can be moved around like counters in a board game. This view may have come with the development of writing, for what can be written down can easily be objectified, and with alphabetic writing historical change can be observed, reinforcing the idea that there is some thing that has changed. It is of interest that a process view is also found in Pānini, but this seems not to have appreciably influenced the western tradition until modern times.

A process view has appeared throughout the study of language right up to the present day. The Stoics in the second century B.C. treated barbarism and solecism in similar process terms (Barwick 1922:96–99). Varro in about 45 B.C. treated derivation and inflection (declinatio) in process terms (De Lingua Latina VI 36, VIII 5, 6)(Kent 1938). The fourth century Latin grammarian Donatus treated two modes of barbarism: pronunciation and writing. Under these there were four species: addition, subtraction, change, and transposition, which were applied to letters, syllables, length, accent, and aspiration, giving forty types in all (Keil 1864:392). The spelling tradition has us change y to i and add ed. A process view has been accepted into modern linguistics. We find discussions of syncope, metathesis, ellipsis, and the like (using Greek terms). And we find affix hopping, extraposition, and other transformations.

Whatever its source, whatever the reasons for this objectification, no matter how natural it may have seemed to the ancients or seems to us now, we find no scientific justification here for assuming the elements of language that enter into the process descriptions. And since they are created by a point of view, they have differed in detail according to the different points of view of the different grammarians and linguists involved, thus adding to the confusions and disagreements in the discipline.

THE SEMIOTIC VIEW

In Plato’s Cratylus we also find concern for the relation of words to their meanings, an early view of a theory of signs. Aristotle is more explicit in On Interpretation 1, where we find a remarkably concise statement of the sign relation between written and spoken words, mental experiences, and the things of which our mental experiences are the images.

It was not until the Stoics, however, in the third and second centuries B.C., that a theory of signs, and with it the study of language, was brought together into a coherent doctrine and placed at the very center of philosophy.¹ The Stoics divided philosophy into three parts: physical, logical, and ethical. The logical part, which contained dialectic and rhetoric, was central, for without it physics and ethics could not express themselves. Stoic dialectic was in fact a highly developed philosophical theory of knowledge, a sophisticated elaboration of Aristotle’s theory of signs. It related signifier, signified, and referent through a series of levels of representation from phōnē ‘sound’ through lexis ‘diction’ and logos ‘speech’ to lekton, the highly developed logical deep structure or meaning that was the signified, and from this through levels of perception and sensation to real objects, the referents in the real world. Stoic grammar by 150 B.C. already had a distinctly modern look. The grammar comprised three of the levels: sound (a phonetic level), diction (a phonological and phonotactic level), and speech (a grammatical and syntactic level). Three of the parts of speech—the proper noun, common noun, and verb—were defined in terms of what they signified at the logic or lekton level. These were individual qualities like Diogenes, Socrates; general qualities like man, horse; and isolated predicates like write, speak. The other two parts of speech were defined grammatically and syntactically.

The whole structure of Stoic dialectic, of which their grammar was but a part, focused on how one knows the truth about referents in the real world in terms of logical propositions, and how these are signified by vocal sounds that can be written down. In this the criterion of truth was the apprehending presentation, that which comes from a real object, agrees with that object itself, and has been imprinted on the mind—and is conveyed through the senses. Epistemology has made changes and advances since the Stoics, but in modern science we still test our theories of what is true about the world by applying observational criteria.

When normative grammar grew out of Stoic dialectic it did not carry over the propositional-logic level of lekton nor the levels of perception and sensation. Some of the parts of speech, however, were still defined partially in terms of what they signified at a logic or meaning level. Thus the theory-of-signs idea of language as a relation between sound and meaning given by grammar and lexicon was retained in normative grammar. We find such a treatment and definitions in the works of the second-century grammarian Apollonius Dyscolus, in the grammar usually attributed to Dionysius Thrax, in the early Latin grammars of Donatus and Priscian, and in the various grammars of European nationalism right up to modern times. Meanwhile the theory of signs was kept alive by philosophers to whom epistemology and the theory of knowledge were of special interest. Although modern linguistics has moved away from prescriptivism, it has generally retained the ancient intellectual standpoint. Grammarians and linguists who have read the classics or the writings of philosophers and logicians have been able to rediscover there the ancient theory that had always been implicit in grammatical writings, particularly in the familiar traditional definitions of the parts of speech, the form-meaning organization of dictionaries, and the standard definition of language as a relation between sound and meaning. Thus to many the theory of signs, when encountered for the first time, has seemed eminently natural and inevitable. To others, however, it has seemed strange and counterintuitive.

Although this semiotic view of language is widely held among modern linguists, as we have seen, and Stoic grammar as a precursor of modern linguistics has been acknowledged by Jakobson (1965), Sebeok (1974), and others, still we cannot find a scientific justification here for assuming objects conceived as signs. In fact, in a theory of knowledge, signs and their sounds and meanings are explicitly not objects given in advance that can be studied through the senses like the objects of study of the other sciences. They are not objects in the real world at all. They are treated not as objects to be known scientifically but rather as part of the means of knowing.

We have to conclude that we do not find here a scientific justification for an assumption of signs as objects of study in linguistics. Furthermore the ancient semiotic standpoint has certain disabilities as a foundation for a modern scientific linguistics. In the first place a philosophical or theory-of-knowledge concern for expressing the truth, if it actually is a concern of modern linguistics, is much too narrow a standpoint. It leaves out of account the many other presumed uses of language mentioned earlier as being of interest in modern linguistics. Designed for a narrow philosophical purpose, the theory is ill-suited as a foundation for treating issues in the broader domain, for example topics in pragmatics. Second, the focus on truth in a theory of knowledge has tended to take as the standard case one word one meaning and unambiguous propositions independent of context. This has raised the problem of ambiguity, for everywhere we find utterances depending heavily on the situational context, and the theory is not equipped to handle the situational context. These points and some of their consequences will be elaborated further in a later section.

THE TEXT-BASED VIEW

The Alexandrians in the third and second centuries B.C. had a somewhat different approach to grammar. Their concerns were Homeric studies, rejuvenating poetry, organizing knowledge, and preserving the great literary heritage of the past (Pfeiffer 1968; Sandys 1915). From this effort came text-based methods in grammar and literary criticism which involved, for example, collecting all the parallels in the Iliad and Odyssey. Aristophanes of Byzantium worked out recurrent patterns of inflection, stated general rules of analogy, paid attention to current spoken dialects of Greek, and tried to preserve and indicate the original pronunciation in the reading of ancient poetry by inventing special diacritics and inserting them in texts to indicate accent and quantity. Aristarchus of Samothrace recognized eight parts of speech (Quintilian, Inst.Orat. I iv 18–20)(Watson 1856), and is credited with adding a sixth rule of analogy to the five of Aristarchus (Barwick 1925:149). The rules of analogy provided tests for classifying words:

If two words were of the same ‘kind’, e.g. both of them nouns or verbs in the same ‘case’ or ‘inflection’, and identical in termination. number of syllables and sound, they were ‘analogous’ to one another; i.e. they belonged to the same declension or conjugation. [Sandys 1915:50–1]

According to the sixth test, both words were to be simple or both compound. One can see here the forerunner of modern text-based methods, including the methods of sames and differents. However, we still do not find here any justification for assuming that the objects compared are given in advance in any scientific sense.

THE NORMATIVE VIEW

The normative tradition, lasting two millennia from the time of the Stoics, had considerable influence on the conception of language and grammar, a point to which we will return in a later section. In connection with the present topic, however, it should be noted that a prescriptive norm would likely be perceived by speakers and writers as something external to them, something out there that is powerful and has to be dealt with and reacted to, thus enhancing the illusion that language is a concrete reality with a natural existence or “life” of its own.² We could investigate scientifically how or why people may perceive language as an external reality. But of course their perception, belief, intuition, or illusion, even if it were universal, or even innate, could not be taken as a scientific justification for an assumption creating the objects of language or language itself.

THE HYPOTHETICAL CONSTRUCT VIEW

What about the many linguistic insights that have been obtained within frameworks assuming the objects of language? Perhaps there is a justification here: call the assumption a hypothesis supported and strengthened by all those valuable results. But recall also the many problems facing linguistics with which we started. The prognosis under the current regimen does not seem very hopeful. Besides, it is not sufficient for a body of theory to be adequate, say, 80 percent of the time. Applications of the traditional Latin grammatical categories to indigenous languages in previous centuries also gave some worth-while results; the grammars are still useful even today. But that wasn’t sufficient. The theory failed in various respects and thus had to be discarded.

Perhaps the objects of language are hypothetical constructs, theoretical entities postulated to account for the evidence. Language would then be in the realm of theory. The trouble with this is that there is nothing for language in this sense to be a theory of. True, grammar is often taken to be in the realm of theory—a theory of language—but this does not help, for as we have seen, the objects of language, such as speech-sounds, phonemes, syllables, morphemes, words, phrases, sentences, meanings, sememes, signs, or others, are not objects naturally given and existing independently of theory.

THE LACK OF JUSTIFICATION

Careful search has not turned up a scientific justification for an assumption creating the objects of language. These objects are not given in advance, but created by a point of view, as Saussure once realized, and their introduction into linguistics requires a special assumption, as Bloomfield emphasized.

An inability to justify special claims or foundational assumptions on the basis of solid evidence is the criterion by which science has rejected all sorts of false theories, mistaken ideas, erroneous superstitions, unwarranted speculation, and rampant mysticism. In all such cases the burden of providing justification rests on those who would continue to maintain the assumption or theory. It must be this way; otherwise science would be led to accept all sorts of guesses, hunches, and wild ideas that could not be justified. Without proper justification how do we know that it makes any more sense to have a science that studies the objects of language than it would to have a biology of unicorns, griffins, and chimeras even if we had found tracks in the mud by the river bank?

It is sometimes suggested that Bloomfield’s assumption, though admittedly false, should be retained anyway, since it has proved fruitful in stimulating research: “Ask of a theory, ‘What has it done for me lately?’ ” An argument like this could be used to perpetuate any false lead that had turned up clues. Better that we should seek the truth than preserve past delusions. Science is tolerant of hypotheses in their early stages and tries to find evidence for their truth or falsity. But when a hypothesis proves to be unfounded it must be abandoned forthwith lest it misdirect our research and distort our understanding of nature.

If Bloomfield’s assumption or its equivalent cannot be justified, which appears to be the case, a linguistics based on the assumption cannot be a science. Note that this is true independently of what may have been the origin of the assumption, its history, or its sometime intuitive appeal. Without adequate scientific justification by evidence from the senses, it is a special assumption over and above the minimum assumptions of all science, and therefore it has no place in science.

Thus we see that with the usual conception of language and the objects of language, and the standard conception of science based on observation and theory, the frequently expressed definition of linguistics as the scientific study of language is an oxymoron. It involves a contradiction in terms. There is no hope of having a linguistics of language that is scientific. The two goals of linguistics, that it study language and that it be a science, are incompatible.

But business as usual in linguistics has accepted these two incompatible goals. It might be urged that we should continue to accept both goals, that there are benefits in retaining inconsistent theories, and besides, we have nothing better. But the benefits of inconsistent theories are illusory, and there are at least two alternatives, one of which will be examined in some detail in later chapters. A linguistics governed by business as usual is living on borrowed time. Simple intellectual honesty dictates that we recognize that the goal of studying language is incompatible with the goal that linguistics be a science, and business as usual, in accepting both goals, is incoherent. Besides, if we ask what this position has done for us lately, we would have to say that it has led us astray. It has led to the chaotic proliferation of inevitably inadequate grammatical theories.

Thus it appears that we cannot properly continue with business as usual. There is a fork in the road ahead. Either we can have a linguistics of language that retains Bloomfield’s assumption or its equivalent and cannot be a science, or we can have a scientific linguistics, in which case we will have to give up Bloomfield’s assumption. We cannot have it both ways.

On the first alternative, linguistics would study language and be disciplined by grammar. Resting ultimately on assumptions, it would relinquish the possibility of testing central aspects of its theory against reality. When theories differed in these aspects there would be no objective criteria for choosing among them. Under this alternative, linguistics would not be a science in the strict sense even if it continued to call itself a science.

On the second alternative the phenomena identified above in the section on linguistics would be studied scientifically. Linguistics would be disciplined by science. But without the assumption of the objects of language, the theoretical structure of linguistics would have to be rebuilt. This could be a large undertaking.

We must choose one or the other of these incompatible alternatives: that linguistics should retain its goal of studying language, or that it should retain its goal of being a science.