Popular Science Monthly/Volume 79/July 1911/The History and Varieties of Human Speech
| THE HISTORY AND VARIETIES OF HUMAN SPEECH[1] |
By Dr. EDWARD SAPIR
THE CANADIAN GEOLOGICAL SURVEY
PERHAPS no single feature so markedly sets off man from the rest of the animal world as the gift of speech, which he alone possesses. No community of normal human beings, be their advance in culture ever so slight, has yet been found, or is ever likely to be found, who do not communicate among themselves by means of a complex system of sound symbols; in other words, who do not make use of a definitely organized spoken language. It is indeed one of the paradoxes of linguistic science that some of the most complexly organized languages are spoken by so-called primitive peoples, while, on the other hand, not a few languages of relatively simple structure are found among peoples of considerable advance in culture. Relatively to the modern inhabitants of England, to cite but one instance out of an indefinitely large number, the Eskimos must be considered as rather limited in cultural development. Yet there is just as little doubt that in complexity of form the Eskimo language goes far beyond English. I wish merely to indicate that, however much we may indulge in speaking of primitive man, of a primitive language in the true sense of the word we find nowhere a trace. It is true that many of the lower animals, for example birds, communicate by means of various cries, yet no one will seriously maintain that such cries are comparable to the conventional words of present-day human speech; at best they may be compared to some of our interjections, which, however, falling outside the regular morphologic and syntactic frame of speech, are least typical of the language of human beings. We can thus safely make the absolute statement that language is typical of all human communities of to-day, and of such previous times as we have historical knowledge of, and that language, aside from reflex cries, is just as untypical of all non-human forms of animal life. Like all other forms of human activity, language must have its history.
Much has been thought and written about the history of language. Under this term may be included two more or less distinct lines of inquiry. One may either trace the changes undergone by a particular language or group of languages for as long a period as the evidence at hand allows, or one may attempt to pass beyond the limits of historically recorded or reconstructed speech, to reconstruct the ultimate origin of speech in general, and to connect these remote origins by means of reconstructed lines of development with historically attested forms of speech. Superficially the latter sort of inquiry is similar in spirit to the labors of the evolutionary biologist; for in both apparently heterogeneous masses of material are, by direct chronologic testimony, inference, analogy and speculation, reduced to an orderly historical sequence. As a matter of fact, however, the reconstruction of linguistic origins and earliest lines of development is totally different in kind from biological reconstruction, as we shall see presently.
Taking up the history of language in the sense in which it was first defined, we find that there are two methods by which we can follow the gradual changes that a language has undergone. The first and most obvious method is to study the literary remains of the various periods of the language of which we have record. It will then be found that not only the vocabulary, but just as well the phonetics, word morphology, and syntactic structure of the language tend to change from one period to another. These changes are always very gradual and, within a given period of relatively short duration, slight or even imperceptible in amount. Nevertheless, the cumulative effect of these slight linguistic changes is, with the lapse of time, so great that the form of speech current at a given time, when directly compared with the form of speech of the same language current at a considerably earlier time, is found to differ from the latter much as it might from a foreign language. It is true that the rate of change has been found to be more rapid at some periods of a language than at others, but it nevertheless always remains true that the changes themselves are not violent and sudden, but gradual in character. The documentary study of language history is of course the most valuable and, on the whole, the most satisfactory. It should not be denied, however, that there are dangers in its use. Literary monuments do not always accurately reflect the language of the period; moreover, orthographic conservatism hides the phonetic changes that are constantly taking place. Thus, there is no doubt that the amount of change that English has undergone from the time of Shakespeare to the present is far greater than a comparison of present-day with Elizabethan orthography would lead the layman to suppose, so much so that I am quite convinced the great dramatist would have no little difficulty in making himself understood in Stratford-on-Avon to-day. For some languages a considerable amount of documentary historical material is available; thus, the literary monuments that enable us to study the history of the English language succeed each other in a practically uninterrupted series from the eighth century A.D. to the present time, while the course of development of Greek in its various dialects can be more or less accurately followed from the ninth century B.C., a conservative date for the Homeric poems, to the present time.
For some, in fact for most languages, however, literary monuments are either not forthcoming at all or else are restricted to a single period of short duration. At first sight it would seem that the scientific study of such languages would have to be limited to purely descriptive rather than historical data. To a considerable extent this is necessarily true, yet an intensive study will always yield at least some, oftentimes a great deal of, information of a historical character. This historical reconstruction on the basis of purely descriptive data may proceed in two ways. It is obvious that the various phonetic and grammatical features of a language at any given time are of unequal antiquity, for they are the resultants of changes that have taken place at very different periods; hence it is reasonable to suppose that internal evidence would, at least within modest limits, enable one to reconstruct the relative chronology of the language. Naturally one must proceed very cautiously in reconstructing by means of internal evidence, but it is oftentimes surprising how much the careful and methodically schooled student can accomplish in this way. Generally speaking, linguistic features that are irregular in character may be considered as relatively archaic, for they are in the nature of survivals of features at one time more widely spread. Not infrequently an inference based on internal evidence can be corroborated by direct historical testimony. One example will suffice here. We have in English a mere sprinkling of noun plurals in -en, such as brethren and oxen. One may surmise that nouns such as these are but the last survivals of a type formerly existing in greater abundance, and indeed a study of Old English or Anglo-Saxon demonstrates that noun plurals in -en were originally found in great number but were later almost entirely replaced by plurals in -s. There is, however, a far more powerful method of reconstructing linguistic history from descriptive data than internal evidence. This is the comparison of genetically related languages.
In making a survey of the spoken languages of the world, we soon find that though they differ from each other, they do so in quite varying degrees. In some cases the differences are not great enough to prevent the speakers of the two languages from understanding each other with a fair degree of ease, under which circumstances we are apt to speak of the two forms of speech as dialects of a single language; in other cases the two languages are not mutually intelligible, but, as in the case of English and German, present so many similarities of detail that a belief in their common origin seems warranted and indeed necessary; in still other cases the two languages are at first glance not at all similar, but reveal on a closer study so many fundamental traits in common that there seems just ground for suspecting a common origin. If other languages can be found which serve to lessen the chasm between the two, and particularly if it is possible to compare them in the form in which they existed in earlier periods, this suspicion of a common origin may be raised to a practical certainty. Thus, direct comparison of Russian and German would certainly yield enough lexical and grammatical similarities to justify one in suspecting them to have diverged from a common source; the proof of such genetic relationship, however, can not be considered quite satisfactory until the oldest forms of German speech and Germanic speech generally have been compared with the oldest forms of Slavic speech and until both of these have been further compared with other forms of speech, such as Latin and Greek, that there is reason to believe they are genetically related to. When such extensive, not infrequently difficult, comparisons have been effected, complete evidence may often be obtained of what in the first instance would have been merely suspected. If all the forms of speech that can be shown to be genetically related are taken together and carefully compared among themselves, it is obvious that much information will be inferred as to their earlier undocumented history; in favorable cases much of the hypothetical form of speech from which the available forms have diverged may be reconstructed with a considerable degree of certainty or plausibility. If under the term history of English we include not only documented but such reconstructed history as has been referred to, we can say that at least in main outline it is possible to trace the development of our language back from the present day to a period antedating at any rate 1500 B.C. It is important to note that, though the English of to-day bears only a faint resemblance to the hypothetical reconstructed Indogermanic speech of say 1500 B.C., there could never have been a moment from that time to the present when the continuity of the language was broken. From our present standpoint that bygone speech of 1500 B.C. was as much English as it was Greek or Sanskrit. The history of the modern English words foot and its plural feet will illustrate both the vast difference between the two forms of speech at either end of the series and the gradual character of the changes that have taken place within the series. Without here going into the actual evidence on which the reconstructions are based, I shall merely list the various forms which each word has had in the course of its history. Starting, then, with foot—feet, and gradually going back in time, we have fūt—fīt, fōt—fēt, fōt—fḗte, fōt—fȫ́te, fōt—fȫ́ti, fōt—fōti, fōt—fṓtir, fōt—fṓtiz, fōt—fṓtis, fōt—fṓtes, fōd—fṓdes, and finally pōd—pōdes, beyond which our evidence does not allow us to go; the last forms find their reflex in Sanskrit pād—pādas.
All languages that can be shown to be genetically related, that is, to have sprung from a common source, form a historic unit to which the term linguistic stock or linguistic family is applied. If, now, we were in a position to prove that all known forms of speech could be classified into a single linguistic stock, the apparent parallel above referred to between linguistic and biological reconstruction would be a genuine one. As it is, we must content ourselves with operating with distinct and, as far as we can tell, genetically unrelated linguistic stocks. The documentary evidence and the reconstructive evidence gained by comparison enable us to reduce the bewildering mass of known languages to a far smaller number of such larger stock groups, yet the absolute number of these latter groups still remains disquietingly large. The distribution of linguistic stocks presents great irregularities. In Europe there are only three such represented: the Indogermanic or Aryan, which embraces nearly all the better known languages of the continent; the Ural-Altaic, the best known representatives of which are Finnish, Hungarian and Turkish; and the Basque of southwestern France and northern Spain. On the other hand, that part of aboriginal North America which lies north of Mexico alone embraces fifty or more distinct linguistic stocks. Some stocks, as, for instance, the Indogermanic just referred to and the Algonkin of North America, are spread over vast areas and include many peoples or tribes of varying cultures; others, such as the Basque and many of the aboriginal stocks of California, occupy surprisingly small territories. It is possible to adopt one of two attitudes towards this phenomenon of the multiplicity of the largest known genetic speech aggregates. On the one hand one may assume that the disintegrating effects of gradual linguistic change have in many cases produced such widely differing forms of speech as to make their comparison for reconstructive purposes of no avail, in other words, that what appear to us to-day to be independent linguistic stocks appear such not because they are in fact historically unrelated, but merely because the evidence of such historical connection has been so obscured by time as to be practically lost. On the other hand, one may prefer to see in the existence of mutually independent linguistic stocks evidence of the independent beginnings and development of human speech at different times and places in the course of the remote history of mankind; there is every reason to believe that in a similar manner many religious concepts and other forms of human thought and activity found widely distributed in time and place have had multiple origins, yet more or less parallel developments. It is naturally fruitless to attempt to decide between the monogenetic and polygenetic standpoints here briefly outlined. All that a conservative student will care to do is to shrug his shoulders and to say, "Thus far we can go and no further." It should be said, however, that more intensive study of linguistic data is from time to time connecting stocks that had hitherto been looked upon as unrelated. Yet it can hardly be expected that serious research will ever succeed in reducing the present Babel to a pristine unity.
Although we can not demonstrate a genetic unity of all forms of human speech, it is interesting to observe that there are several fundamental traits that all languages have in common. Perhaps these fundamental similarities are worthy of greater attention than they generally receive and may be thought by many to possess a high degree of significance. First of all, we find that in every known language use is made of exactly the same organic apparatus for the production of speech, that is, the glottal passage in the larynx, the nasal passages, the tongue, the hard and soft palate, the teeth and the lips. The fact that we are accustomed to consider all speech as self-evidently dependent on these organs should not blind us to the importance of the association. There is, after all, no à priori reason why the communication of ideas should be primarily through sound symbols produced by the apparatus just defined; it is conceivable that a system of sound symbols of noises produced by the hands and feet might have been developed for the same purpose. As a matter of fact, there are many systems of thought transference or language in the widest sense of the word, as a moment's thought will show, that are independent of the use of the ordinary speech apparatus. The use of writing will occur to every one as the most striking example among ourselves. Among primitive peoples we may instance, to cite only a couple of examples of such subsidiary forms of language, the gesture language of the Plains Indians of North America and the very highly developed drum language of several African tribes. From our present point of view it is significant to note that these and other such non-spoken languages are either, as in the case of practically all systems of writing, themselves more or less dependent on a phonetic system, that is, speech in the ordinary sense of the word, or else are merely auxiliary systems intended to replace speech only under very special circumstances. The fact then remains that the primary and universal method of thought transference among human beings is via a special articulating set of organs. Much loose talk has been expended by certain ethnologists on the relatively important place that gesture occupies in the languages of primitive peoples, and it has even been asserted that several so-called primitive languages are unintelligible without the use of gesture. The truth, however, is doubtless that the use of gesture is associated not with primitiveness, but rather with temperament. The Russian Jew and the Italian, for instance, non-primitive as they are, make a far more liberal use of gestures accompanying speech than any of the aborigines of North America.
If we examine in a large way the structure of any given language, we find that it is further characterized by the use of a definite phonetic system, that is, the sounds made use of in its words are reducible to a limited number of consonants and vowels. It does not seem to be true, certain contradicting statements notwithstanding, that languages are to be found in which this phonetic definiteness is lacking and in which individual variation of pronunciation takes place practically without limit. It is of course freely granted that a certain amount of sound variation exists in every language, but it is important to note that such variation is always very limited in range and always takes place about a well-defined center. All known forms of speech, then, operate with a definite apparatus of sounds; statements to the contrary will in most cases be found to rest either on a faulty perception on the part of the recorder of sounds unfamiliar to his ear or on his ignorance of regular sound processes peculiar to the language. Naturally the actual phonetic systems found in various languages, however much they may resemble each other in this fundamental trait of definiteness, differ greatly in content, that is in the sounds actually employed or neglected. This is inevitable, for the vast number of possible and indeed existing speech sounds makes an unconscious selection necessary. Even so, however, it is at least noteworthy with what persistency such simple vowel sounds as a and i and such consonants as n and s occur in all parts of the world.
Even more than in their phonetic systems languages are found to differ in their morphologies or grammatical structures. Yet also in this matter of grammatical structure a survey from a broad point of view discloses the fact that there are certain deep-lying similarities, very general and even vague in character, yet significant. To begin with, we find that each language is characterized by a definite and, however complex, yet strictly delimited grammatical system. Some languages exhibit a specific type of morphology with greater clearness or consistency than others, while some teem with irregularities; yet in every case the structure tends to be of a definite and consistently carried out type, the grammatical processes employed are quite limited in number and nearly always clearly developed, and the logical categories that are selected for grammatical treatment are of a definite sort and number and expressed in a limited, however large, number of grammatical elements. In regard to the actual content of the various morphologies, we find, as already indicated, vast differences, yet here again it is important to note with what persistence certain fundamental logical categories are reflected in the grammatical systems of practically all languages. Chief among these may be considered the clear-cut distinction everywhere made between denominating and predicating terms, that is between subject and predicate, or, roughly speaking, between substantive and verb. This does not necessarily imply that we have in all cases to deal with an actual difference in phonetic form between noun and verb, though as a matter of fact such differences are generally found, but simply that the structure of the sentence is such as to show clearly that one member of it is felt by the speaker and hearer to have a purely denominating office, another a purely predicating one. It may be objected that in Chinese, for instance, there is no formal distinction made between noun and verb. True, but the logical distinction of subject and predicate is reflected in the form of the Chinese sentence, inasmuch as the subject regularly precedes the predicate; thus, while the same word may be either noun or verb, in any particular sentence it necessarily is definitely one and not the other. Other fundamental logical categories will, on a more complete survey, be found to be subject to grammatical treatment in all or nearly all languages, but this is not the place to be anything but merely suggestive. Suffice it to remark on the wide-spread systematizing of personal relations; the wide-spread development of ideas of tense, number and syntactic case relations; and the clear grammatical expression everywhere or nearly everywhere given to the largely emotional distinction of declarative, interrogative and imperative modes.
Granted that there are certain general fundamental traits of similarity in all known languages, the problem arises of how to explain these similarities. Are they to be explained historically, as survivals of features deep-rooted in an earliest form of human speech that, despite the enormous differentiation of language that the lapse of ages has wrought, have held their own to the present day, or are they to be explained psychologically as due to the existence of inherent human mental characteristics that abide regardless of time and race? If the latter standpoint be preferred, we should be dealing with a phenomenon of parallel development. It is of course impossible to decide categorically between the two explanations that have been offered, though doubtless the majority of students would incline to the psychological rather than to the historical method. At any rate, it is clear that we can not strictly infer a monogenetic theory of speech from the fundamental traits of similarity that all forms of speech exhibit. Yet even though these are of psychologic rather than historic interest, it is important to have demonstrated the existence of a common psychological substratum, or perhaps we had better say framework, which is more or less clearly evident in all languages. This very substratum or framework gives the scientific study of language a coherence and unity quite regardless of any considerations of genetic relationship of languages.
In spite of the fact that, as we have seen, no tangible evidence can be brought to bear on the ultimate origin or origins of speech, many attempts have been made, particularly in the first half of the nineteenth century, when it was more common for historical and philosophical problems of extreme difficulty to be attacked with alacrity, to point out the way in which human speech originated or at least might have originated. From the very nature of the case, these attempts could not but be deductive in method; hence, however plausible or ingenious in themselves, they have at best a merely speculative, not a genuinely scientific interest. We may therefore dispense with anything like a detailed inquiry into or criticism of these theories. Two of the most popular of them may be respectively termed the onomatopoetic or sound-imitative and the exclamatory theories. According to the former, the first words of speech were onomatopoetic in character, that is, attempts to imitate by the medium of the human organs of speech, the various cries and noises of the animate and inanimate world. Thus, the idea of a "hawk" would come to be expressed by an imitative vocable based on the actual screech of that bird; the idea of a "rock" might be expressed by a combination of sounds intended in a crude way to reproduce the noise of a rock tumbling down hill or of a rock striking against the butt of a tree; and so on indefinitely. In course of time, as these imitative words by repeated use became more definitely fixed in phonetic form, they would tend to take on more and more the character of conventional sound-symbols, that is of words, properly speaking. The gradual phonetic modifications brought on in the further course of time would finally cause them to lose their original onomatopoetic form. It may be freely granted that many words, particularly certain nouns and verbs having reference to auditory phenomena, may have originated in this way; indeed, many languages, among them English, have at various times, up to and including the present, made use of such onomatopoetic words. It is difficult, however, to see how the great mass of a vocabulary, let alone a complex system of morphology and syntax, could have arisen from an onomatopoetic source alone. The very fact that onomatopoetic words of relatively recent origin are found here and there in sharp contrast to the overwhelmingly larger non-onomatopoetic portion of the language accentuates, if anything, the difficulty of a general explanation of linguistic origins by means of the onomatopoetic theory.
The exclamatory theory, as its name implies, would find the earliest form of speech in reflex cries of an emotional character. These also, like the hypothetical earliest words of imitative origin, would in course of time become conventionalized and sooner or later so modified in phonetic form as no longer to betray their exclamatory origin. The criticisms urged against the onomatopoetic theory apply with perhaps even greater force to the exclamatory one. It is, if anything, even more difficult here than in the former case to see how a small vocabulary founded on reflex cries could develop into such complex linguistic systems as we have actually to deal with. It is further significant that hardly anywhere, if at all, do the interjections play any but an inconsiderable, almost negligible, part in the lexical or grammatical machinery of language. An appeal to the languages of primitive peoples in order to find in them support for either of the two theories referred to is of little or no avail. Aside from the fact that their elaborateness of structure often seriously militates against our accepting them as evidence for primitive conditions, we do not on the whole find either the onomatopoetic or exclamatory elements of relatively greater importance in them than elsewhere. Indeed the layman would be often surprised, not to say disappointed, at the almost total absence of onomatopoetic traits in many American Indian languages, for instance. In Chinook and related dialects of the lower course of the Columbia, onomatopoesis is developed to a more than usual extent, yet, as though to emphasize our contention with an apparent paradox, hardly anywhere is the grammatical mechanism of a subtler, anything but primitive character. We are forced to conclude that the existence of onomatopoetic and exclamatory features is as little correlated with relative primitiveness as we have found the use of gesture to be. As with the two theories of origin we have thus briefly examined, so it will be found to be with other theories that have been suggested. They can not, any of them, derive support from the use of the argument of survivals in historically known languages; they all reduce themselves to merely speculative doctrines.
So much for general considerations on language history. Returning to the gradual process of change which has been seen to be characteristic of all speech, we may ask ourselves what is the most central or basic factor in this never-ceasing flux. Undoubtedly the answer must be: phonetic change or, to put it somewhat more concretely, minute or at any rate relatively trivial changes in pronunciation of vowels and consonants which, having crept in somehow or other, assert themselves more and more and end by replacing the older pronunciation, which becomes old-fashioned and finally extinct. In a general way we can understand why changes in pronunciation should take place in the course of time by a brief consideration of the process of language learning. Roughly speaking, we learn to speak our mother-tongue by imitating the daily speech of those who surround us in our childhood. On second thoughts, however, it will be seen that the process involved is not one of direct imitation, but of indirect imitation based on inference. Any given word is pronounced by a succession of various more or less complicated adjustments of the speech organs. These adjustments or articulations give rise to definite acoustic effects, effects which, in their totality, constitute speech. Obviously, if the child's imitative efforts were direct, it would have to copy as closely as possible the speech articulations which are the direct source of what it hears. But it is still more obvious that these speech articulations are largely beyond the power of observation and hence imitation. It follows that the actual sounds, not the articulations producing them, are imitated. This means that the child is subject to a very considerable period of random and, of course, wholly involuntary experimenting in the production of such articulations as would tend to produce sounds or combinations of sounds approximating more or less closely those the child hears. In the course of this experimenting many failures are produced, many partial successes. The articulations producing the former, inasmuch as they do not give results that match the sounds which it was intended to imitate, have little or no associative power with these sounds, hence do not readily form into habits; on the other hand, articulations that produce successes or comparative successes will naturally tend to become habitual. It is easy to see that the indirect manner in which speech articulations are acquired necessitates an element of error, very slight, it may be, but error nevertheless. The habitual articulations that have established themselves in the speech of the child will yield auditory results that approximate so closely to those used in speech by its elders, that no need for correction will be felt. And yet it is inevitable that the sounds, at least some of the sounds, actually pronounced by the child will differ to a minute extent from the corresponding sounds pronounced by these elders. Inasmuch as every word is composed of a definite number of sounds and as, furthermore, the language makes use of only a limited number of sounds, it follows that corresponding to every sound of the language a definite articulation will have become habitual in the speech of the child; it follows immediately that the slight phonetic modifications which the child has introduced into the words it uses are consistent and regular. Thus if a vowel a has assumed a slightly different acoustic shade in one word, it will have assumed the same shade in all other cases involving the old a-vowel used by its elders, at any rate in all other cases in which the old a-vowel appears under parallel phonetic circumstances.
Here at the very outset we have illustrated in the individual the regularity of what have come to be called phonetic laws. The term "phonetic law" is justified in so far as a common tendency is to be discovered in a large number of individual sound changes. It is important, however, to understand that phonetic law is a purely historic concept, not one comparable to the laws of natural science. The latter may be said to operate regardless of particular times and places, while a phonetic law is merely a generalized statement of a process that took place in a restricted area within a definite period of time. The real difficulty in the understanding of phonetic change in language lies not in the fact of change itself, nor in the regularity with which such change proceeds in all cases affected, but, above all, in the fact that phonetic changes are not merely individual, but social phenomena; in other words, that the speech of all the members of a community in a given time and place undergoes certain regular phonetic changes. Without here attempting to go into the details of this process of the transformation of an individual phonetic peculiarity into a social one, we will doubtless not be far wrong in assuming that uniformity is at first brought about by a process of unconscious imitation, mutual to some extent, among the younger speakers of a restricted locality, later, perhaps, by the half-conscious adoption of the new speech peculiarity by speakers of neighboring localities, until, finally, it has spread either over the entire area in which the language is spoken or over some definite portion of it. In the former case the historic continuity of the language as a unit is preserved, in the latter a dialectic peculiarity has asserted itself. In the course of time other phonetic peculiarities spread that serve to accentuate the dialectic division. However, the ranges of operation of the different phonetic laws need not be coterminous, so that a network of dialectic groupings may develop. At least some of the dialects will diverge phonetically more and more, until in the end forms of speech will have developed that deserve to be called distinct languages. It can not be denied that, particularly after a considerable degree of divergence has been attained, other than purely phonetic characteristics develop to accentuate a difference of dialect, but every linguistic student is aware of the fact that the most easily formulated and, on the whole, the most characteristic differences between dialects and between languages of the same genetic group are phonetic in character.
True, some one will say, changes of a purely phonetic character can be shown to be of importance in the history of language, but what of changes of a grammatical sort? Are they not of equal or even greater importance? Strange as it may seem at first blush, it can be demonstrated that many, perhaps most, changes in grammatical form are at last analysis due to the operation of phonetic laws. Inasmuch as these phonetic laws affect the phonetic form of grammatical elements as well as of other linguistic material, it follows that such elements may get to have a new bearing, as it were, brought about by their change in actual phonetic content; in certain cases, what was originally a single grammatical element may in this way come to have two distinct forms, in other cases two originally distinct grammatical elements may come to have the same phonetic appearance, so that if circumstances are favorable, the way is paved for confusion and readjustment. Briefly stated, phonetic change may and often does necessitate a readjustment of morphologic groupings. It will be well to give an example or two from the history of the English language. In another connection we have had occasion to briefly review the history of the words foot and feet. We saw that there was a time when these words had respectively the form fōt and fōti. The final i-vowel of the second word colored, by a process of assimilation which is generally referred to as "umlaut," the ō of the first syllable and made it ö, later unrounded to ē; the final i, after being dulled to an e, finally dropped off altogether. The form fōti thus step by step developed into the later fēt, which is the normal Anglo-Saxon form. Note the result. In fōti and other words of its type the plural is expressed by a distinct suffix -i, in fēt, as in modern English feet, and in words of corresponding form it is expressed by an internal change of vowel. Thus an entirely new grammatical feature in English, as also in quite parallel fashion in German, was brought about by a series of purely phonetic changes, in themselves of no grammatical significance whatever.
Such grammatical developments on the basis of phonetic changes have occurred with great frequency in the history of language. In the long run, not only may in this way old grammatical features be lost and new ones evolved, but the entire morphologic type of the language may undergo profound modification. A striking example is furnished again by the history of the English language. It is a well-known feature of English that absolutely the same word, phonetically speaking, may often, according to its syntactic employment, be construed as verb or as noun. Thus we not only love and kiss, but we also give our love or a kiss, that is, the words love and kiss may be indifferently used to predicate or to denominate an activity. There are so many examples in English of the formal, though not syntactic, identity of noun stem and verb stem that it may well be said that the English language is on the way to become of a purely analytic or isolating type, more or less similar to that of Chinese. And yet the typical Indogermanic language of earlier times, as represented say by Latin or Greek, always makes a rigidly formal, not merely syntactic, distinction between these fundamental parts of speech. If we examine the history of this truly significant change of type in English, we shall find that it has been due at last analysis to the operation of merely phonetic laws. The original Anglo-Saxon form of the infinitive of the verb kiss was cyssan, while the Anglo-Saxon form of the noun kiss was cyss. The forms in early middle English times became dulled to kissen and kiss, respectively. Final unaccented -n later regularly dropped off, so that the infinitive of the verb came to be kisse. In Chaucer's day the verb and the noun were still kept apart as kisse and kiss, respectively; later on, as a final unaccented -e regularly dropped off, kisse became kiss, so that there ceased to be any formal difference between the verb and noun. The history of the Anglo-Saxon verb lufian "to love" and noun lufu "love" has been quite parallel; the two finally became confused in a single form luv, modern English love. Once the pace has been set, so to speak, for an interchange in English between verbal and nominal use of the same word, the process, by the working of simple analogy, is made to apply also to cases where in origin we have to deal with only one part of speech; thus we may not only have a sick stomach, but we may stomach an injury (noun becomes verb), and conversely we may not only write up a person, but he may get a write up (verb becomes noun). It has, I hope, become quite clear by this time how the trivial changes of pronunciation that are necessitated by the very process of speech acquirement may, in due course of time, profoundly change the fundamental characteristics of language. So also, if I may be pardoned the use of a simile, may the slow erosive action of water, continued through weary ages, profoundly transform the character of a landscape. If there is one point of historic method rather than another that the scientific study of language may teach other historical sciences, it is that changes of the greatest magnitude may often be traced to phenomena or processes of a minimal magnitude.
On the whole, phonetic change may be said to be a destructive or at best transforming force in the history of language. Reference has already been made to the influence of analogy, which may, on the contrary, be considered a preservative and creative force. In every language the existing morphological groups establish more or less definite paths of analogy to which all or practically all the lexical material is subjected; thus a recently acquired verb like to telegraph in English is handled in strict analogy to the great mass of old verbs with their varying forms. Such forms as he walks and he laughs set the precedent for he telegraphs, forms like walking and laughing for telegraphing. Without such clear-cut grooves of analogy, indeed, it would be impossible to learn to speak, a corollary of which is that there is a limit to the extent of grammatical irregularity in any language. When, for some reason or other, as by the disintegrating action of phonetic laws, too great irregularity manifests itself in the morphology of the language, the force of analogy may assert itself to establish comparative regularity, that is, forms which belong to ill-defined or sparsely represented morphologic groups may be replaced by equivalent forms that follow the analogy of better-defined or more numerously represented groups. In this way all the noun plurals of English, if we except a few survivals like feet and oxen, have come to be characterized by a suffixed -s; the analogical power of the old -s plurals was strong enough to transform all other plurals, of which Anglo-Saxon possessed several distinct types. The great power exerted by analogy is seen in the persistence with which children, whose minds are naturally unbiased by tradition, use such forms as foots and he swimmed. Let us not smile too condescendingly at the use of such forms; it may not be going too far to say that there is hardly a word, form, or sound in present-day English which was not at its first appearance looked upon as incorrect.
The disintegrating influence of phonetic change and the leveling influence of analogy are perhaps the two main forces that make for linguistic change. The various influences, however, that one language may exert upon another, generally summed up in the word borrowing, are also apt to be of importance. As a rule such influence is limited to the taking over or borrowing of certain words of one language by another, the phonetic form of the foreign word almost always adapting itself to the phonetic system of the borrowing language. Besides this very obvious sort of influence, there are more subtle ways in which one language may influence another. It is a very noteworthy phenomenon that the languages of a continuous area, even if genetically unrelated and however much they may differ among themselves from the point of view of morphology, tend to have similar phonetic systems or, at any rate, tend to possess certain distinctive phonetic traits in common. It can not be accidental, for instance, that both the Slavic languages and some of the neighboring but absolutely unrelated Ural-Altaic languages (such as the Cheremiss of the Volga region) have in common a peculiar dull vowel, known in Russian as yerí, and also a set of palatalized or so-called "soft" consonants alongside a parallel set of unpalatalized or so-called "hard" consonants. Similarly, we find that Chinese and Siamese have in common with the unrelated Annamite and certain other languages of Farther India a system of musical accent. A third very striking example is afforded by a large number of American Indian linguistic stocks reaching along the Pacific coast from southern Alaska well into California and beyond, which have in common peculiar voiceless l-sounds and a set of so-called "fortis" consonants with cracked acoustic effect. It is obvious that in all these cases of comparatively uniform phonetic areas embracing at the same time diverse linguistic stocks and types of morphology we must be dealing with some sort of phonetic influence that one language may exert upon another. It may also be shown, though perhaps less frequently, that some of the morphologic traits of one language may be adopted by a neighboring, sometimes quite unrelated, language, or that certain fundamental grammatical features are spread among several unrelated linguistic stocks of a continuous area. One example of this sort of influence will serve for many. The French express the numbers 70, 80 and 90, respectively, by terms meaning 60-10, 4 twenties and 4 twenties 10; these numerals, to which there is no analogy in Latin, have been plausibly explained as survivals of a vigesimal method of counting, that is counting by twenties, the numbers above 20, a method that would seem to have been borrowed from Gallic, a Celtic language, and which still survives in Gaelic and other modern Celtic languages. This example is the more striking as the actual lexical influence which Celtic has exerted upon French is surprisingly small. So much for the influence of borrowing on the history of a language.
We may turn now to take up the matter of the varieties of human speech. One method of classifying the languages of the world has been already referred to; it may be termed the genetic method, inasmuch as it employs as its criterion of classification the demonstrable relation of certain languages as divergent forms of some older form of speech. As we have already seen, the linguistic stocks which we thus get as our largest units of speech are too numerous to serve as the simplest possible reduction of the linguistic material to be classified. One naturally turns, therefore, to a psychological classification, one in which the classificatory criterion is the fundamental morphological type to which a particular language or stock is to be assigned. Such a classification of morphological types may proceed from different points of view, varying emphasis being laid on this or that feature of morphology. It is clear at the outset that we have to distinguish between what we may call the subject-matter or content of morphology and the mere form pure and simple. Any grammatical system gives formal expression to certain modes or categories of thought, but the manner of expression of these categories or the formal method employed may vary greatly both for different categories and for different languages. Not infrequently the same logical category may be expressed by different formal methods in the same language. Thus, in English, the negative idea is expressed by means of three distinct formal methods exemplified by untruthful, with its use of a prefix un-, which can not occur as a freely movable word; hopeless, with its use of a suffix -less, which again can not occur as a freely movable word; and not good, in which the negative idea is expressed by an element (not) that has enough mobility to justify its being considered an independent word. We have here, then, three formal processes illustrated to which may be assigned the terms prefixing, suffixing and juxtaposing in definite order. While the same logical category may be grammatically expressed by different formal methods, it is even more evident that the same general formal method may be utilized for many different categories of thought. Thus, in English, the words books and worked use the same method of suffixing grammatical elements, the one to express the concept of plurality, the other that of past activity. The words feet and swam, furthermore, respectively express the same two concepts by the use of an entirely distinct formal method, that of internal vowel change.
On the whole one finds that it is possible to distinguish between two groups of grammatically expressed logical categories. One group may be characterized as derivational; it embraces a range of concepts expressed by grammatical elements that serve to limit or modify the signification of the word subjected to grammatical treatment without seriously affecting its relation to other words in the sentence. Such merely derivational elements are, in English, prefixes like un-, suffixes like -less, agentive suffixes like -er in baker, and numerous others. The second group of logical concepts and corresponding grammatical elements may be characterized as relational; they not merely serve to give the word affected a new increment of meaning, as is the case with the first group, but also assign it a definite syntactic place in the sentence, defining as they do its relation to other words of the sentence. Such a relational grammatical element, in English, is the plural -s suffix; a word, for instance, like books differs from its corresponding singular book not merely in the idea of plurality conveyed by the suffix -s, but assumes a different grammatical relation to other words in the sentence—a book is, but books are. Such relational elements are, furthermore, the case and gender suffixes of nouns and adjectives in Indogermanic languages; furthermore, the personal endings and tense suffixes of verbs. On the whole it may be said that derivational elements are of relatively more concrete signification than the relational ones and tend to become more thoroughly welded into a word unit with the basic word or stem to which they are attached or which they affect. This statement, however, is only approximately of general application and is subject to numerous qualifications. The greatest degree of concreteness of meaning conveyed by derivational elements is probably attained in many, though by no means all, American Indian languages, where ideas of largely material content are apt to be expressed by grammatical means. To this tendency the name of polysynthesis has been applied. Thus in Yana, an Indian language of northern California, such ideas as up a hill, across a creek, in the fire, to the east, from the south, immediately, in vain and a host of others are expressed by means of grammatical suffixes appended to the verb stem; so also in Nootka, an Indian language of Vancouver Island, so highly special ideas as on the head, in the hand, on the rocks, on the surface of the water, and many others, are similarly expressed as suffixes. It is important to note that, although the distinction between derivational and relational grammatical elements we have made is clearly reflected in some way or other in most languages, they differ a great deal as to what particular logical concepts are treated as respectively derivational or relational. Such concepts as those of sex gender, number and tense, which in Indogermanic are expressed as relational elements, are in other linguistic stocks hardly to be separated, as regards their grammatical treatment, from concepts treated in a clearly derivational manner. On the other hand, demonstrative ideas, which in most Indogermanic languages receive no relational syntactic treatment, may, as in the Kwakiutl language of British Columbia, serve an important relational function, analogous, say, to the Indogermanic use of gender; just as in Latin, for instance, such a sentence as "I saw the big house" is expressed by "I-saw house-masculine-objective big-masculine-objective," with a necessary double reference to the concepts of case relation and gender, so in Kwakiutl the sentence "I saw the house" would have to be expressed by some such sentence as "I-saw-the-objective-near-you house-visible-near-you," with an analogous necessary double reference to the demonstrative relations involved. If, now, it has been shown that no necessary correlation exists between particular logical concepts and the formal method of their grammatical rendering, and if, furthermore, there can not even be shown to be a hard and fast line in grammatical treatment between concepts of a derivational and concepts of a more definitely relational character, what becomes of the logical category per se as a criterion of linguistic classification on the basis of form? Evidently it fails us. Of however great psychological interest it might be to map out the distribution in various linguistic stocks of logical concepts receiving formal treatment, it is clear that no satisfactory formal classification of linguistic types would result from such a mapping.
Having thus disposed of the subject-matter of linguistic morphology as a classificatory criterion, there is left to us the form pure and simple. Here we are confronted first of all by a number of formal grammatical methods or processes. These, being less numerous than the logical categories which they express themselves, and, furthermore, being on the whole more easily defined and recognized, would seem to lend themselves more easily to classificatory purposes. The simplest grammatical process is the juxtaposing of words in a definite order, a method made use of to perhaps the greatest extent by Chinese, to a very large extent also by English; the possibilities of the process from the point of view of grammatical effectiveness may be illustrated by comparing such an English sentence as "The man killed the bear" with "The bear killed the man," the actual words and forms being identical in the two sentences, yet definite case relations being clearly expressed in both. A somewhat similar process, yet easily enough kept apart, is compounding, that is, the fusion of two words or independent stems, into a firm word-unit; the process is particularly well developed in English, as illustrated by words like railroad and underestimate, and indeed is found widely spread among the most diverse linguistic stocks. In some languages, as in the Sioux and Paiute of our own country, compounding of verb stems is frequent, as illustrated by such forms as to eat-stand, that is to eat while standing; on the other hand, in not a few linguistic stocks, as the wide-spread Athabascan stock of North America and the Semitic languages, compounding as a regular process is almost or entirely lacking. Perhaps the most commonly used formal method of all is affixing, that is, the appending of grammatical elements to a word or to the body or stem of a word; the two most common varieties of affixing are prefixing and suffixing, examples of which have been already given from English. Probably the majority of linguistic stocks make use of both prefixes and suffixes, though they differ greatly as to the relative importance to be attached to these two classes of elements. Thus, while both in Indogermanic and in the Bantu languages of Africa prefixes and suffixes are to be found, we must note that the greater part of the grammatical machinery of Indogermanic is carried on by its suffixes, while it is the prefixes that in Bantu take the lion's share of grammatical work. There are also not a few linguistic stocks in which suffixing as a process is greatly developed, while prefixing is entirely unknown; such are Ural-Altaic, Eskimo, and the Kwakiutl and Nootka languages of British Columbia. On the other hand, languages in which prefixes are used, but no suffixes, seem to be quite rare. A third variety of affixing, known as infixing, consists in inserting a grammatical element into the very body of a stem; though not nearly so wide-spread as either prefixing or suffixing, it is a well-attested linguistic device in Malayan, Siouan, and elsewhere. Still another wide-spread grammatical process is reduplication, that is, the repetition of the whole or, generally, only part of the stem of a word; in Indogermanic we are familiar with this process in the formation, for instance, of the Greek perfect, while in many American Indian languages, though in far from all, the process is used to denote repeated activity. Of a more subtle character than the grammatical processes briefly reviewed thus far is internal vowel or consonant change. The former of these has been already exemplified by the English words feet and swam as contrasted with foot and swim; it attains perhaps its greatest degree of development in the Semitic languages. The latter, internal consonant change, is on the whole a somewhat rare phenomenon, yet finds an illustration in English in at least one group of cases. Beside such nouns as house, mouse, and teeth we have derived verbs such as to house, mouse around, and teeth; in other words a certain class of verbs is derived from corresponding nouns by the changing of the final voiceless consonants of the latter to the corresponding voiced consonants. In several non-Indogermanic linguistic stocks, as in Takelma of southwestern Oregon and in Fulbe of the Soudan, such grammatical consonant changes play a very important part. As the last formal grammatical process of importance may be mentioned accent, and here we have to distinguish between stress accent and musical or pitch accent. An excellent example of the grammatical use of stress accent is afforded in English by such pairs of words as cónflict and conflíct, óbject and objéct, the verb being accented on the second syllable, the noun on the first. Musical accent is a far more prevalent phonetic characteristic than is perhaps generally supposed; it is by no means confined to Chinese and neighboring languages of eastern Asia, but is found just as well in many languages of Africa and, as has been recently discovered by Mr. J. P. Harrington and the writer, in a few North American Indian languages. As a process of definite grammatical significance, however, musical accent is not so wide-spread. It is found, to give but one example, in the earlier stages of Indogermanic, as exemplified, among others, by classical Greek and by Lithuanian.
Having thus briefly reviewed the various grammatical processes used by different languages, we may ask ourselves whether the mapping out of the distribution of these processes would be of more service to us in our quest of the main types of language than we have found the grammatical treatment of logical concepts to be. Here a difficulty presents itself. If each linguistic stock were characterized by the use of just one or almost entirely one formal process, it would not be difficult to classify all languages rather satisfactorily on the basis of form. But there are great differences in this respect. A minority of linguistic stocks content themselves with a consistent and thoroughgoing use of one process, as does Eskimo with its suffixing of grammatical elements, but by far the larger number make use of so many that their classification becomes difficult, not to say arbitrary. Thus in Greek alone every one of the processes named above, excepting consonant change, can be exemplified. Even if we limit ourselves to a consideration of grammatical processes employed to express the relational concepts, we shall find the same difficulty, for the same language not infrequently makes use of several distinct processes for concepts of this class.
On a closer study of linguistic morphology, however, we find that it is possible to look at the matter of form in language from a different, at the same time more generalized, point of view than from that of the formal processes employed themselves. This new point of view has regard to the inner coherence of the words produced by the operation of the various grammatical processes, in other words, to the relative degree of unity which the stem or unmodified word plus its various grammatical increments or modifications possesses, emphasis being particularly laid on the degree of unity which the grammatical processes bring about between the stem and the increments which express relational concepts. On the basis of this formal criterion we may classify languages, at least for the purposes of this paper, into the three main types of linguistic morphology generally recognized. The first type is characterized by the use of words which allow of no grammatical modification whatever, in other words the so-called isolating type. In a language of this type all relational concepts are expressed by means of the one simple device of juxtaposing words in a definite order, the words themselves remaining unchangeable units that, according to their position in the sentence, receive various relational values. The classical example of such a language is Chinese, an illustration from which will serve as an example of the isolating type of sentence. wọ̄o̯3 (rising from deep tone) pū2 (rising from high) pʽā4 (sinking from middle) tʽā1 (high) may be literally translated "I not fear he," meaning "I do not fear him"; wọ̄o̯3 "I" as subject comes first; pʽā4 "fear" as predicate follows it; pū2 "not," inasmuch as it limits the range of meaning given by the predicate, must precede it, hence stands between the subject and predicate; finally tʽā2 "he" as object follows the predicate. If we exchange the positions of wọ̄o̯3 and tʽā1 we change their syntactical bearing; wọ̄o̯3 "I" becomes "me" as object, while tʽā1, which in our first sentence was best translated as "him" now becomes "he" as subject, and the sentence now takes on the meaning of "he does not fear me."
In the second main type of language, generally known as the agglutinative, the words are not generally unanalyzable entities, as in Chinese, but consist of a stem or radical portion and one or more grammatical elements which partly modify its primary signification, partly define its relation to other words in the sentence. While these grammatical elements are in no sense independent words or capable of being understood apart from their proper use as subordinate parts of a whole, they have, as a rule, their definite signification and are used with quasi-mechanical regularity whenever it is considered grammatically necessary to express the corresponding logical concept; the result is that the word, though a unit, is a clearly segmented one comparable to a mosaic. An example taken from Turkish, a typical agglutinative language, will give some idea of the spirit of the type it represents. The English sentence "They were converted into the (true) faith with heart and soul" is rendered in Turkish džan u gönül-den iman-a gel-ir-ler[2] literally translated, "Heart and soul-from belief-to come-ing-plural." The case-ending -den "from" is here appended only to gönül "soul" and not to džan "heart," though it applies equally to both; here we see quite clearly that a case-ending is not indissolubly connected with the noun to which it is appended, but has a considerable degree of mobility and corresponding transparency of meaning. The verb form gel-ir-ler, which may be roughly translated as "they come," is also instructive from our present point of view; the ending -ler or -lar is quite mechanically used to indicate the concept of plurality, whether in noun or verb, so that a verb form "they come," really "come-plural," is to some extent parallel to a noun form like "books," really "book-plural." Here we see clearly the mechanical regularity with which a logical concept and its corresponding grammatical element are associated.
In the third, the inflective, type of language, while a word may be analyzed into a radical portion and a number of subordinate grammatical elements, it is to be noted that the unity formed by the two is a very firm one, moreover that there is by no means a mechanical one-to-one correspondence between concept and grammatical element. An example from Latin, a typical inflective language, will illustrate the difference between the agglutinative and inflective types. In a sentence like videō hominēs "I see the men," it is true that the verb form videō may be analyzed into a radical portion vide- and a personal ending -ō, also that the noun form hominēs may be analyzed into a radical portion homin- and an ending -ēs which combines the concepts of plurality with objectivity, that is, a concept of number with one of case. But, and here comes the significant point, these words, when stripped of their endings, cease to have even a semblance of meaning, in other words, the endings are not merely agglutinated on to fully-formed words, but form firm word-units with the stems to which they are attached; the absolute or rather subjective form homō, "man," is quite distinct from the stem homin- which we have obtained by analysis. Moreover, it should be noted that the ending -ō is not mechanically associated with the concept of subjectivity of the first person singular, as is evidenced by such forms as vīdī "I saw" and videam "I may see"; in the ending -ēs of hominēs the lack of the mechanical association I have spoken of is even more pronounced, for not only are there in Latin many other noun endings which perform the same function, but the ending does not even express a single concept, but, as we have seen, a combined one.
The term polysynthetic is often employed to designate a fourth type of language represented chiefly in aboriginal America, but, as has been shown in another connection, it refers rather to the content of a morphologic system than to its form, and hence is not strictly parallel as a classificatory term to the three we have just examined. As a matter of fact, there are polysynthetic languages in America which are at the same time agglutinative, others which are at the same time inflective.
It should be carefully borne in mind that the terms isolating, agglutinative and inflective make no necessary implications as to the logical concepts the language makes use of in its grammatical system, nor is it possible definitely to associate these three types with particular formal processes. It is clear, however, that on the whole languages which make use of word order only for grammatical purposes are isolating in type, further, that languages that make a liberal grammatical use of internal vowel or consonant change may be suspected of being inflective. It was quite customary formerly to look upon the three main types of morphology as steps in a process of historical development, the isolating type representing the most primitive form of speech at which it was possible to arrive, the agglutinative coming next in order as a type evolved from the isolating, and the inflective as the latest and so-called highest type of all. Further study, however, has shown that there is little to support this theory of evolution of types. The Chinese language, for instance, so far from being typical of a primitive stage, as used to be asserted, has been quite conclusively proven by internal and comparative evidence to be the resultant of a long process of simplification from an agglutinative type of language. English itself, in its historical affiliations an inflective language, has ceased to be a clear example of the inflective type and may perhaps be said to be an isolating language in the making. Nor should we be too hasty in attaching values to the various types and, as is too often done even to-day, look with contempt on the isolating, condescendingly tolerate the agglutinative, and vaunt the superiority of the inflective type. A well-developed agglutinative language may display a more logical system than the typically inflective language. And as for myself, I should not find it ridiculous or even paradoxical if one asserted that the most perfect linguistic form, at least from the point of view of logic, had been attained by Chinese, for here we have a language that, with the simplest possible means at its disposal, can express the most technical or philosophical ideas with absolute lack of ambiguity and with admirable conciseness and directness.