The review given in the preceding chapter of the opinions held by others on language and its science or philosophy will have prepared the way for an independent inquiry into the nature and objects of linguistic science. Before, however, we can discuss the limits and character of the science we must have a clear idea of the subject-matter with which it deals. Most of us, no doubt, have a rough-and-ready definition to give of language; but science requires something more than rough-and-ready definitions, and the discordant views as to the scope and meaning of the science of language which have come before us in the foregoing pages are plain evidence that an accurate definition of language is not so easy as would at first sight appear.

Provisionally, however, we may define language as consisting of certain modulations of the voice, variously combined and arranged, which serve as symbols for the thoughts or feelings we wish to express. The sounds that we utter must have a meaning before they can become language, otherwise they will be mere cries or gibberish, less worthy of the name of language than even the howling of the dog upon the prairie or the wild song of the forest-bird. Language is the outward expression and embodiment of thought—the garment, so to speak, with which the mind clothes itself when it would reveal itself to another or even, it may be, to itself. The words of a foreign tongue form a language only for those who understand what they signify: for those who do not they are but empty sounds, the idle murmur of a “barbarous” jargon. “The language of birds” was discovered to the Eastern sage alone: to all others the notes of the nightingale and the thrush were as the plashing of the waterfall and the drowsy humming of the bees. “Lessons in running brooks” may indeed be read by the mind, but it is the mind itself that puts them there, and only in so far as it creates a meaning for them does it create also the language in which they speak.

It is evident that our thoughts could be represented by other symbols than sounds. The first and most familiar instance that rises to our minds is writing, though writing symbolizes thoughts only indirectly, its immediate office being to symbolize sounds. There is a written language because there is previously a spoken language, and those who learn foreign tongues know well how detrimental the power of reading a language is when we wish to speak it: the language of the eye has to be translated into the language of the ear. Language can only be symbolized directly to the eye by hieroglyphics; but if our communication with one another depended upon hieroglyphic writing it would never be very extensive or progressive. To say nothing of its requiring time, writing materials, and skill in drawing, hieroglyphic writing can indicate objects alone with that clearness and certainty which language demands. It is hardly possible to represent in this way abstract ideas, verbs, or adjectives, so that what is denoted shall be recognized by another without previous instruction. Apart from these drawbacks, however, picture-writing has this advantage over spoken language, that its symbols are not mere arbitrary signs like sounds, but intelligible all the world over; and even the degenerated picture-writing of China, by preserving everywhere the same character for the same idea, has kept up a unity and spread a culture throughout the empire which would otherwise have been impossible among a people divided into many and diverse dialects.

Another means of symbolizing thought is “mathematical language,” which represents the calculations of the mathematician by written symbols such as 1, 2, 3, x, y, z. But such symbols are of late invention, and could not well be applied to express the daily concerns of life. Quite different is gesture-language, whereby our thoughts and emotions are represented by movements of the hands and other parts of the body. Most of our common needs could be expressed in this way, though gestures would be quite inadequate to represent the wants of a civilized community. Only such ideas as “I am hungry,” “let me drink,” “it is pleasant,” could be denoted by them. But, like picture-writing, gestures possess the great advantage of standing for the same ideas everywhere and among all men. The expression of pain or surprise, the threatening shake of the hand, the pointing of the finger, have the same message for the Negro as for the European. The traveller in a strange and unknown region is thrown back upon the language of gesture. Burton,[49] perhaps, exaggerates when he says that the Arapahos of North America, “who possess a scanty vocabulary, can hardly converse with one another in the dark,” and another reason may be given for this preference for the light; but the importance of gesture-language where other means of communication are wanting is too evident to need examples. Thus Fisher[50] tells us that the Comanches and neighbouring tribes have “a language of signs, by which all Indians and traders can understand one another; and they always make these signs when communicating among themselves.” To the same effect James[51] writes of the Kiawa-Kashaia Indians: “These nations, although constantly associating together and united under the influence of the Bear-Tooth, are yet totally ignorant of each other’s language, insomuch that it was no uncommon occurrence to see two individuals of different nations sitting upon the ground, and conversing freely by means of the language of signs. In the art of thus conveying their ideas they were thorough adepts; and their manual display was only interrupted at remote intervals by a smile, or by the auxiliary of an articulated word of the language of the Crow Indians, which to a very limited extent passes current among them.” Gesture-language is instinctive—the heritage, it may be, of the days before man acquired articulate language, or differed thus far from the brute beast: certain ideas call forth certain corresponding gestures, and we are not obliged to learn what gestures stand for particular ideas. Hence it is that even now spoken language is so largely accompanied by gesture. An excited speaker is likely to make much use of his hands; and we can often tell what a person is saying to us, though we do not hear him distinctly, by watching the play of his features. We know from the appearance of his face whether he is asking a question, whether he is angry, or whether he is dispirited. With the cultivation of articulate speech and confidence in the use of it, men become more phlegmatic in speaking, less inclined to have recourse to subsidiary helps. It is the awkward country girl whose “manners” have “not that repose which stamps the caste of Vere de Vere.” The preacher who addresses an audience of barristers does well to dispense with the gesticulation which is necessary to the mob-orator. According to M. Antoine d’Abbadie, an Abyssinian Galla marks the punctuation of his speech by the help of a leathern whip, a slight stroke denoting a comma, a harder one a semi-colon, a still harder one a full stop, while a note of admiration is represented by a furious cut through the air.[52] Even in this country, we have not to go far to find gesture-language employed in default of spoken language. Where the new system of observing the movements of the lips has not been introduced, the deaf and dumb can communicate with the outer world only by the help of gestures, though the gesture-language of the deaf and dumb, like phonetic writing, implies a previous spoken language. It is, therefore, to the instinctive gesture-language of the North American Indians what our system of writing is to hieroglyphics.

It will be noticed that under the general term of “gesture-language” we have included not only gesticulation, but also that play of feature and modulation of the voice which outlast gesticulation among a civilized people. Gesticulation can hardly form a universal language in the same way that play of feature and modulation of the voice can. Only in part have such gestures the same meaning for all men, and so serve to bridge over the gulf that divides articulate from inarticulate speech. Like play of feature and modulation of the voice, they are common to men and animals; but, unlike the latter, they are capable of receiving an arbitrary and conventional meaning. Helvétius, following in the track of Anaxagoras, asserted that we have become men through the possession of hands; had our arms terminated in a horse’s hoof, for instance, we should have been like the beasts that perish, wanderers and defenceless.[53] Indeed, it is quite conceivable that our forefathers would have remained contented with a gesture-language, had not the hands been wanted for other purposes. Food could not be prepared without them, whereas it was not until the desire of food was satisfied that the mouth was put to another use than that of asking for it.

Still the arbitrary element in gesture-language is very small compared with what it is in spoken language. Here beyond a few interjections, or possibly a few onomatopœic sounds, the whole body of symbols that stand for thought is purely conventional. The same combination of sounds may be used to denote very different ideas. There is no necessary connection between an idea and a word that represents it. It is as arbitrary as our making the sign 1 symbolize the idea of unity or the sign = the idea of equivalence. However well we may be acquainted with our own language, a foreign one will be wholly unintelligible to us until we have learnt it. Even natural sounds strike the ear of different individuals and nations in a totally different way. Exactly the same sound was intended to be reproduced in the “bilbit amphora” of Nævius,[54] the “glut glut murmurat unda sonans” of the Latin Anthology[55] and the puls of Varro; nay, as Dr. Farrar points out, even in the κόγξ and the βλώψ of the Greeks. The Persian bulbul has but little resemblance to the jugjug of Gascoigne, or the whitwhit of other writers; and yet all are attempts at imitating the note of the nightingale. The first word uttered by the children on whom Psammitikhus is said to have tried his famous experiment seemed to their keepers to be βέκ[ος], whereas we read in the great Papyrus Ebers, the standard work on Egyptian medicine compiled in the sixteenth century B.C., that if “a child on the day of birth ... says ni, it will live; if it says ba, it will die.” And only the last of these infantile cries bears any likeness to what we are told are the primitive and original utterances of childhood, ma, pa, and ta—utterances, by the way, which are only in part possible to the Mohawks and Hurons, who possess no labials.[56] So arbitrary and conventional must be the meanings we associate with the sounds of articulate speech, and so impossible is it to discover in them any signs of universal currency. There is no reason in the nature of things why the word book should represent what we mean when we look at the present volume; it might just as well be denoted by koob, or biblion, or liber; and if we chose we might always so denote it.

But although we might choose to do so, unless we could get other people to do the same, we should find ourselves unintelligible to our neighbours, and talking gibberish instead of a language. For the essential thing about a language is that it should be an instrument for the communication of our thoughts to others. There is no good in having symbols for our thoughts unless we wish our thoughts to become known to those about us. He who has no thoughts to communicate, no wants to be supplied, has no need of a language. But such a being, to use the words of Aristotle, is ἢ θηρίον ἢ θεός, “either a beast or a god;” or as we might, perhaps, render it in modern phraseology, either a hermit or an angel. The voiceless Yogi of India, or the Bernardine nun of southern France, is but as a dumb animal, or the hapless deaf-mute who has never been trained. The records of speech themselves testify to our instinctive recognition of the fact. The name Slave, for instance, by which so large a body of our Aryan kinsfolk have called themselves, means “the speaker,” in opposition to the “dumb” and unintelligible German; just as in Isaiah (xxxiii. 19), the Assyrians are a people “of a stammering tongue, that one cannot understand.” Man, indeed, comes from the root man, “to think;” but it is thinking for others, for the sake of embodying the thought in spoken utterance. The same root has produced μάντις, the “seer;” Μέντωρ and Minerva, whose counsels are for others, not for themselves; μηνύω, to “point out,” and moneo, to “advise;” μνή-μη, “recollection,” and memini, to “remember,” and mentio, “the bringing to mind” by mentioning in speech. Even in the Semitic idioms, zâcâr, “a man,” seems connected with zâcâr, “to remember,” just as the Latin mas is with μνήμη and memini. Language, in short, is the prerogative of man, distinguishing him from the brute beast, because it is the basis and bond of society. Man is “a social animal” in virtue of language; society could not exist without language any more than language could without society. The two are correlative terms, though it is for the sake of society that language has been formed. It is a social product, springing up with the first community, developing with the increasing needs of culture and civilization, and disappearing when the individual Robinson Crusoe is cast back on the island of primitive isolation.

But though it is a social product, it may also with strict truth be spoken of as growing up. A society never met together to make a language. To imagine this would be to revive the theories of the last century, which referred all society and government to a contract entered into by our remote forefathers. We do not call the present volume a book because we have made a formal agreement with our neighbours to do so, but because if we called it biblion or liber we should not be understood by the majority of them. The language which we speak is the heritage which has come down to us from the past, like the laws by which we are governed, or the habits and customs to which we conform. We represent our idea of a printed work by the word book, because we have been taught to do so by others, and those who taught us had been taught by others, and those again by others. But this process of teaching and learning implies a very slow and gradual change in the language that is being handed down. New words come into use as new objects and ideas have to be named, old words are forgotten, the pronunciation gets altered, and other changes hereafter to be described take place. And so, without any deliberate intention on the part of any individual or individuals, the whole character of a language comes in course of time to be transformed. Now and then, it is true, we can trace the invention of a wholly new word to an individual, like gas to the Dutch chemist van Helmont, or od force to Baron von Reichenbach; and still oftener of a new derivative like liberalize, introduced by the Marquis of Lansdowne, fatherland by Isaac Disraeli, incuriosité by Montaigne, urbanité by Balzac, or bienfaisance by the Abbé de Saint Pierre. But such words must be accepted by society, be ratified by the tacit agreement of the whole community, before they can become a part of living speech. Though gas has made its way into common use, blas, which van Helmont proposed at the same time to describe that property of the heavenly bodies whereby they regulate the changes of time, failed to commend itself to the general sense of the community, and so passed out of sight;[57] and such was also the fate of Balzac’s sériosité, of Malherbe’s dévouloir, and of Burke’s literator. In spite of his 262 works, and the grammars and vocabularies written to explain the jargon employed in them, Caramuel, a famous Spanish bishop of the seventeenth century, was unable to bequeath to posterity a single one of his numerous coinages. The “Cabalistic Grammar,” published at Brussels in 1642, and the “Audacious Grammar,” printed at Frankfurt twelve years later, remained unread and unknown, a monument of “cabalistic” dreams and “audacious” folly.[58] A paternal government may compel the acceptance of a foreign speech, in place of the familiar mother-tongue, like the rulers of Japan, who were said, a short time ago, to be meditating the substitution of English for the native language under pain of death. But even a government of this kind cannot invent a new grammar and a new dictionary; it can only borrow from others: and if we are to judge from the experiences of certain Oxford colleges where French was similarly enforced in the days of the Plantagenets, and Latin in those of the Commonwealth, the attempt, though backed by all the powers of State and Church, is likely to end in failure. Language must be the unconscious creation of the whole society, and the changes it undergoes must be equally that society’s unconscious work.

Now the sum of knowledge possessed by a society increases the longer the society exists and the more civilized it becomes. This increase of knowledge is reflected in the language; and hence languages grow fuller and richer—more developed, as it is termed—the longer they last. The further back we can trace a language, the poorer it is seen to be. Not only are words, or rather derivatives and compounds, wanting, but the words that exist embody but a few out of the many meanings which afterwards cluster around them. The dictionary of the Anglo-Saxon, of the Ormulum, or even of Chaucer, is scant and meagre compared with that at the disposal of a modern English writer. The dialects of savages, which most resemble what all languages originally were, have few words, because they have few ideas to express, and such ideas as are expressed are wonderfully simple. Thus, the Tasmanians, when they wanted to denote what we mean by “tall” and “round,” had to say “long legs” and “like a ball” or the “moon” or some other round object, eking out their scanty vocabulary by the help of gesture.[59] So, too, the New Caledonians cannot be brought to understand such ideas as those conveyed by yesterday and to-morrow, and the jungle Veddahs of Ceylon are unable to remember even the names they give to their wives, unless the latter be present.[60] After this, it is not surprising that, like the Dammaras of South Africa, they are unable to count, and, consequently, have no numerals in their language. According to Mr. Galton,[61] indeed, the Dammaras are able to count as far as three, though he adds that they discover the loss of an ox, “not by the number of the herd being diminished, but by the absence of a face that they know.” If two sticks of tobacco are “the rate of exchange for one sheep, it would sorely puzzle a Dammara to take two sheep and give him four sticks.” “Once,” he goes on to say, “while I watched a Dammara floundering hopelessly in a calculation on one side of me, I observed Dinah, my spaniel, equally embarrassed on the other. She was overlooking half-a-dozen of her new-born puppies, which had been removed two or three times from her; and her anxiety was excessive, as she tried to find out if they were all present, or if any were still missing. She kept puzzling and running her eyes over them, backwards and forwards, but could not satisfy herself. She evidently had a vague notion of counting, but the figure was too large for her brain. Taking the two as they stood, dog and Dammara, the comparison reflected no great honour on the man.”

The number of abstracts possessed by a language is a good gauge of its development. It is difficult for us to realize the mental struggles and the ages of previous preparation required for the discovery of those ideas which now seem to us so familiar. The day on which, according to the ancient legend, Pythagoras struck out the idea of the world, and named it κόσμος, summed up all the labours of Eastern philosophy and Greek thought before which the law and order of the universe at last lay revealed. It is to Anaxagoras, to Herakleitus, to Xenophanes that we owe those ideas of mind, of motion, of existence which form the groundwork of modern science. Nay, our own generation has witnessed the creation of more than one great abstract idea, henceforth to be the common property of mankind, through the word by which it is expressed. To have won for the race a single idea like that of natural selection is a higher glory than the conquests of a Cæsar. Man’s first work, according to the old Hebrew writer, was to give names to “every living creature;” and the Assyrian story of the Creation, with the profound conviction that chaos is there where language is not, begins its record with the words:

The words by which we express such abstract and spiritual ideas as those of spirit, of virtue, or of intellect are all, when examined, found to have a purely sensuous origin. The spirit was but “the breath,” virtue “the quality of a man,” intellect “a choosing between.” We can only rise from the known to the unknown, from that which we perceive to that which is invisible. As the developing mind starts from the objects of sense, and passes over the bridge of analogy to objects of thought and reason, so, too, language, at the outset, had words only for the visible and the sensuous; and not until it called in the aid of metaphor could it express the higher imaginations of the soul. If we look closely into language, we may see how strewn it is with worn-out and forgotten metaphors. “They are,” as Carlyle has said, “its muscles and tissues and living integuments,” the aids whereby language can communicate something more than the things which we see and feel. Even among ourselves, there are few who can afford to dispense with the assistance of concrete illustration and metaphor when dealing with abstract subjects. They throw a halo of light around the impalpable objects of philosophic reasoning, and enable us to picture them before our minds. It is this picture-language, as we may call it, which gives so much of its charm to poetry, which made verse the first embodiment of literature, and lends to savage speech its poetical garb. The creations of mythology are in the main its work; and even modern science does not despise a “nature” which clothes itself with the attributes of humanity and of sex. It was the power possessed by language of rising from the concrete to the abstract that made the earliest hieroglyphic systems of writing possible, and which to this day enables the Chinaman to adapt his mode of writing to the introduction of new ideas. Like the Chinese lexicon, the multitudinous wealth of language can be traced back to a few and simple elements.

If we watch the first attempts of children to speak, we find that their wants and wishes are conveyed in a very small number of sounds, and that often a single word is made to express what we should represent by several. Now children, in spite of their inherited instinct of speech, are the best example we can have of the way in which the first men acquired their language, remembering only that the child nowadays has a complete language already framed for him, whereas the first men had to frame theirs for themselves. What the individual child now learns in a few years has been the laborious production of many a century and many a generation. But the child has still to learn it like his forefathers before him, and in learning it he may modify its sounds, its forms, or the meaning of its words, and so take part in bringing about what we call the growth of speech.

But it is not only by watching children that we can gain some idea of the way in which languages originally grew up. When we try to acquire a foreign tongue, not from books, but from conversation, we first pick up a few sentences and words, and then, by the help of these, endeavour to make our thoughts and wishes intelligible to others. But since the sentences and words we know are but few, we have to look about us for the simplest mode of expressing ourselves, and are obliged to make our expressions stand for many different ideas. Even then, however, our vocabulary is imperfect, and we often find ourselves wholly at a loss for any word by which to convey our meaning. Gestures are the only resource left to us, and it is by their help that we supplement our deficient knowledge of the spoken language. Indeed, the first words and sentences learnt at all may have been acquired by the same means. Travellers have drawn up vocabularies and phrase-books of the idioms of unknown tribes by pointing to objects or making use of gesticulations, and then observing what articulate sounds were associated with these movements by the persons addressed. It is a good example of the way in which gestures precede spoken language, and lead on to the latter. The same gestures are for the most part understood in the same sense among all the manifold races of men; a shake of the head signifies “no,” a pointing of the finger symbolizes “locality.” Gestures bridge over the gulf which separates inarticulate from articulate speech, and they are still a means of communication for the deaf-mute. But we must distinguish between gestures and that instinctive play of feature which Mr. Darwin has treated of in his work on the “Expression of the Emotions.” Gestures, in the proper sense of the term, are only partly the same for all races of men; no doubt the instinctive element preponderates in them, but we have to allow also for a certain element of conventionality. There is not the same physiological reason why a shake of the head should denote a negative as there is why a particular expression of the face should indicate pleasure, or pain, or surprise, or why a feeling of shame should bring a blush to the cheek. When we are told that the Veddahs of Ceylon are never seen to laugh, we at once infer that they have no sense of humour and no power of merriment. Gestures are rather a sign for the intellect than for the emotion, and since the same feeling must express itself similarly in the case of every one while the same thought need not, it is evident that that which expresses thought admits the element of conventionality more than that which expresses feeling. Pain must always be pain, and affect the nerves and muscles in the same way; what is thought of, on the contrary, may be conceived very differently, and represented in an equally varying manner. Hence it is that we share the play of feature with the brutes, whereas gestures—embodying as they do a rational rather than an emotional element—are for the most part peculiar to man. Man is man in virtue of language, and it was gestures that first made language possible.

But gestures alone are often but a poor resource for either the child or the traveller. They fail to express the meaning intended. Let us suppose a child, for example, to have been scratched by a cat, or frightened by a herd of cows. It can represent the pain it has suffered, or the terror it has experienced by gestures, but if it be unacquainted with the names of cat and cow, it can only point out those animals by imitating the sounds they utter; and miow and moo-moo become the nursery names for “cat” and “cow.” And what still goes on in the nursery was a general procedure in the childhood of mankind. The domestic cat was introduced into Egypt from Nubia in the time of the eleventh or twelfth dynasty, and the Egyptians forthwith called it the miau, a name which it still bears in China. Indeed, the French and German equivalents of “puss,” mimi and mitz, have the same origin as the miow of the nursery or of Egypt, though German could not refrain from borrowing the unmelodious ending of katz. Dr. Comrie states[62] that the natives of the north-east coast of Papua call the dog a “bow-wow,” and when first shown an iron axe named it din-din, from the sound which it seemed to make.[63] This imitation of natural sounds goes by the long and barbarous name of Onomatopœia, and though an attempt has been made to substitute “Imson” (imitatio son-i) for “onomatopœic word,” it has failed.[64] Now if we are to infer anything from the habits of the nursery, and of those savage tribes which best represent the infancy of mankind, onomatopœia must have played a large part in the formation of language. Its advocates have done much harm to what Professor Max Müller has happily termed “The Bow-wow Theory,” by endeavouring to trace back words as we now find them to an onomatopœic origin; but this does not prove that the theory when scientifically applied is false. It is true that there are few words like miow which can be immediately referred to an onomatopœic source; it is true also that articulate language begins with roots, from which its scientific student must derive its words; but it is equally true that a large proportion of these roots—or rather of what these roots presuppose—was formed by the help of onomatopœia. It is not only objects like a dog or an iron axe that the Papuans met by Dr. Comrie named from the sounds they made upon his ear; an action like that of “eating” was equally called nam-nam from the noise produced by the process. We who speak a highly developed language, the worn-out débris of which are more than sufficient for the creation of new words and forms, can hardly realize the influence of onomatopœia upon rude and uncivilized jargons. Of course it is not necessary that the imitation of natural sounds should be an exact one; indeed, that it never can be: all that is wanted is that the imitation should be recognizable by those addressed. The same natural sound, consequently, may strike the ear of different persons very differently, and so be represented in articulate speech in a strangely varying manner. Thus, as has been noted before, bilbit, glut-glut, and puls, are all attempts to represent the same sound. Just as colours strike differently upon the eyes of different men, so also do sounds upon their ears, and the poverty of primitive languages in terms to denote the colours is parallel to the imperfection with which they represented natural sounds.[65]

Besides gestures and onomatopœia, there is a third way in which we can make ourselves intelligible without knowing the articulate language of those to whom we are speaking. This is by making use of interjectional cries. Like the play of feature, interjectional cries are the same for all men; we all make much the same kind of exclamation when hurt, or angry, or surprised. They express our emotions, not our ideas; and since the main object of language is to express ideas, interjectional cries can have had but a small share in its formation. Here and there we can point to a few roots, like agh (ach) in Aryan, which seem to have this derivation; but before the root agh could become a root in the linguistic sense of the word, and give rise to a number of derivatives, it was needful for it to cease to be an interjection; that is to say, it had to express an idea, and not an emotion. Many of our modern interjections, like alas, lo, are words that once possessed a full conceptual meaning, but have lost their original signification, and been degraded to the level of mere emotional cries. So hard is it for language to admit anything which was not from the first significant in thought. Interjections remind us of the animal side of our nature, and they have forced their way into language only because that animal side must be represented to the mind. But in thus forcing their way they have ceased to be the simple utterances of pleasure and pain, and become expressive of conceivable states of feeling. Only in so far as the first men approached the brutes more nearly than we do, were interjectional cries likely to help them in building up the structure of speech. We may, however, include under the head of interjections those instinctive cries uttered by men when engaged in a common work, to which Professor Noiré would trace all roots whatsoever.[66] The sense of life and power that makes the child shout or the bird sing, and is the ultimate motive of human speech, causes us to beat time by the help of rhythmical utterances. And though the utterance be but a monotonous sing-song, it becomes a symbol and sign of the action it accompanies to all those who have taken part in it, and in course of time may pass into a word. How many of the roots of languages were formed in this way it is impossible to say, but when we consider that there is no modern word which we can derive from such cries as the sailor makes when he hauls a rope, or the groom when he cleans a horse, it does not seem likely that they can have been very numerous. Still they were probably more numerous than the roots formed from other interjectional cries.

The origin of language, then, is to be sought in gestures, onomatopœia, and to a limited extent interjectional cries. Like the rope-bridges of the Himalayas or the Andes, they formed the first rude means of communication between man and man. Onomatopœic words and interjections came to be metaphorically applied to denote other ideas than those for which they properly stood, while the relations of grammar were pointed out by the help of gesticulation. Thus, by imitating the gurgling of water and pointing to the mouth, a man could signify what we express by the sentence, “I wish to drink,” or, “I am thirsty;” and by uttering a cry of pain and pointing to a knife, he could show that he had been cut by it. In course of time a collection of words would be formed, each of which represented what we now call a sentence. For a sentence, it must be remembered, is the name given by the grammarian to what the logician would call a proposition or a judgment, and though a judgment may be analyzed into subject and object and connecting copula (or mental act of comparison), we cannot, if we wish to be intelligible, separate its elements one from the other. The whole sentence, the whole Λόγος, as the Greeks would have termed it, is the only possible unit of thought; subject and object are as much correlated as the positive and negative poles of the magnet.

Language, then, we may lay down, begins with sentences, not with single words. The latter exist only for the lexicographer, and even the lexicographer has to turn them into sentences by affixing a definition if he would render them intelligible. We are accustomed to see sentences divided into their individual words in writing, and so we come to fancy that this is right and natural. But the very accent which we lay upon our words ought to show us how far this is from the truth. The accent of a word varies according to its place in a sentence; for purposes of accentuation, we regard not the individual words, but the whole sentence which they compose. And this outward fact of accentuation is but an indication of the inward fact of signification. All language must be significant; but until the whole sentence is uttered, until the whole thought which lies behind it is expressed, this cannot be the case. The expression of the thought may be faulty and imperfect, but unless the thought be sufficiently expressed to be intelligible to another, it has not yet embodied itself in the form of language. The Greek Λόγος was not the individual word, which, apart from its relation to other parts of the sentence, has no meaning in itself, but the complete act of reasoning, which on the inward side is called a judgment, and on the outward side a sentence or proposition. The single word is to the sentence what syllables and letters are to the single word. We may break up a word into the several sounds of which it is composed, but this is the work of the phonologist, not of the speaker. So, too, we may break up a sentence like “Don’t do that” into the four words Do-not-do-that, but this, again, is the conscious procedure of the grammarian. Sentences may be of any length; they may consist of a single syllable, like go! or yes, or they may have to be expressed by a large number of separate “words”; what is essential is that they should be significant to another, should adequately convey to his mind the whole thought that is intended to be expressed. Unless the sounds we utter are combined into a sentence, they have no more meaning than the cries of the jackal or the yelping of the cur; and until they have a meaning, and so represent our thought, they do not constitute language. The sentence, in short, is the only unit which language can know, and the ultimate starting-point of all our linguistic inquiries.

It is not necessary that the sentence should be divided into its component words in writing any more than it actually is in speech. The French je le vois is as much a single, undivided group of sounds as the Basque dakust or the Latin amatur. In the polysynthetic languages of America, in which the separate words of a sentence are cut down to their bare stems and fused into a single whole, the sentence can as little be split up into its elements as an ordinary compound in Greek or German. The ancient Hindu grammarians, with that wonderful insight into language which has made their labours the basis of modern scientific philology, treated the several words of a sentence just as we treat syllables and letters. A number of single words are run into one, the sounds at the end of each word being modified to suit those that follow, in accordance with the so-called rules of Sandhi, and the whole group of words is then written without division. Thus the word trinairguṇatwamâpannairbadhyante must be analyzed into trinais, “with grass blades” (an instrumental pl.), guṇatwam, “a rope’s state” (acc. sing), â-pannais, “having attained” (part. pass. of the compound verb â-pad, agreeing with trinais), and badhyante, “they are bound” (3rd pl. pres. pass. of the verb bandh). In fact, a little attention will convince every one that even in our own language not only does the accent of a word depend upon its place in the sentence, but that the sound with which it terminates equally depends upon the sound which follows. We pronounce “of” in one way when it stands by itself in the dictionary, in another way when it precedes “the” or “that.”

If the sentence is the unit of significant speech, it is evident that all individual words must once have been sentences; that is to say, when first used they must each have implied or represented a sentence. And this is borne out by an examination of the records of speech. We shall see hereafter that words may be divided into conceptual or presentative, and pronominal or representative, and that wherever we can trace back the latter to their source, we find them to have been originally presentative. Thus words like “and” or “because” are now purely symbolic and representative; there was a time, however, when they denoted the very definite ideas of “a going further,”[67] and “by the cause.”[68] Now, if we look carefully into the nature and essence of these presentative words, it becomes clear that they were at the outset so many shorthand notes or summaries of various sentences. Take, for example, the word memorandum. Before it can form a part of language, memorandum must be significant. This can come about only in two ways. Either we must accompany the utterance of the word memorandum with gestures which imply “This is a memorandum,” or “Write a memorandum,” or something similar, or else we must express the meaning of these gestures by equivalent words. That is to say, the isolated word memorandum must be incorporated into a sentence by being brought into relation with other words, before it can become part and parcel of living speech. Taken by itself, it belongs to the dictionary-maker only, and even he has to add a definition, that is to say, to make it the subject of a sentence, if his dictionary is to be something more than a mere catalogue of unmeaning sounds. Before a definition is supplied by the lexicographer or the reader, a word is not yet a word; it has no meaning.

The student of language, then, cannot deal with words apart from sentences. The significant word—that combination of sounds which represents a thought—is really a crystallized sentence, a kind of shorthand note in which a proposition has been summed up. Each advance in philosophy and science is marked by the acquisition of a new idea or fact, the result of a long train of previous observations and reasonings: and the more complex the idea or the fact, the more numerous will be the reasonings, the sentences or judgments, which underlie it. What a multitude of judgments, which when expressed in language we call sentences, are implied by the two simple words humanity and gravitation! It is a truism in psychology that the terms of a proposition, when closely interrogated, turn out to be nothing but abbreviated judgments. The ordinary theory of modern comparative philologists traces all languages back to a certain number of abstract roots, each of which was a sort of sentence in embryo, and though this theory is scarcely tenable in the form in which it is usually presented, it is yet certain that there was a time in the history of speech when the articulate or semi-articulate sounds uttered by primitive man were made the significant representatives of thought by the gestures with which they were accompanied. And this complex of sound and gesture—a complex in which, it must be remembered, the sound had no meaning apart from the gesture—was the earliest sentence. The isolating languages of Further India still express a new concept by the juxtaposition of two words which denote that it is the species of a higher genus. Thus, in Taic or Siamese kin is to “eat,” but when nam, “water,” is added, kin-nam means “to drink;” mi is “rich,” mi din, mi nám, “earthy,” and “watery,” that is to say, “rich (in) earth” and “water.”[69]

These examples from the far East show us the way in which our words first came into existence. They have grown out of sentences by a process of comparison and determination. Two or more sentence-words, referring to the same object or idea viewed under different relations to the speaker, might be set over against one another, and the phonetic part in which they agreed taken to denote the object or idea considered by itself. Thus in Semitic kâtal is “he killed,” kotêl, “killing,” k’tol, “to kill” and “kill,” kâtûl, “killed,” and katl, kitl, or kutl, “a killing,” where the difference of signification is marked by a difference of vowel, and co-existing forms of this kind, when compared with each other, would determine that the three consonants k-t-l had the general sense of “killing.” But an inflectional language does not permit us to watch the word-making process so clearly as do those savage jargons in which a couple of sounds like the Grebo ni ne signify “I do it” or “you do it,” according to the context and the gestures of the speaker. Here by degrees, with the growth of consciousness and the analysis of thought, the external gesture is replaced by some portion of the uttered sounds which agrees in a number of different instances, and in this way the words by which the relations of grammar are expressed come into being. A similar process has been at work in producing those analogical terminations whereby our Indo-European languages adapt a word to express a new grammatical relation. Thus, in English, the Greek termination ize (or -ise) has been abstracted from the words to which it properly belonged by comparing them together, and has been instinctively, as it were, invested with a particular meaning, so that we can now turn any word we like, whether of Greek origin or not, into a transitive verb by attaching to it this suffix. In humanize, for instance, it is added to an adjective of Latin origin, in jeopardize to a Romanic compound. When once a sentence-word had been broken up into single words by comparing it with other sentence-words relating to the same subject, it was easy to extend the operation to other sentence-words, which were accordingly broken up and analyzed without being compared with related sentences. The phonetic expression of the verbal copula by which the subject and object were connected together, was the last result of this analytic process; it was long left to be supplied by the mind, the simple juxtaposition of subject and object being considered sufficient to suggest the mental act by which they were compared or contrasted, and to this day many languages, those of Polynesia, for example, still remain without a verb. Thus, in Dayak kutoh ka-halap-e arut-m, “thy boat is very beautiful,” is literally “very its-beauty thy-boat,” andi-m handak imukul-ku, “thy brother will be struck by me,” means properly, “thy-brother my striking-being,” while to express “he has a white jacket on,” the Dayak must say, ia ba-klambi ba-puti, “he with-jacket with-white.”[70]

As we shall see hereafter, all the facts at our disposal tend to show that the roots of speech, or at all events the earliest sentence-words out of which the later languages of mankind have sprung, were polysyllabic, and other facts go equally towards proving that the terminations of these primitive roots or sentence-words displayed a wearisome monotony of agreement. Survivals, as Mr. Tylor has happily termed them, are among the most valuable means we have of arguing back to an earlier state of things, and we can only treat as a survival the habit of a child whom I know, who in her first essays at speech affixed a final ö to almost all her words, saying for instance, come-ö and dog-ö for “come” and “dog.” The older a speech is, the more it has suffered from the wasting and wearing effects of time, and a language like the Chinese, which stands out as some weather-beaten granite peak among the languages of a later day, has so concealed all traces of the originally pluriliteral character of its vocabulary, that it is only within the last few years that Sinologues, like Dr. Edkins and M. de Rosny, have detected it. So, we may infer, will it also be found with all the other languages of the world; the first utterances of mankind were polysyllabic, though not perhaps of such monstrous length as the sentence-words of Eskimaux or Algonquin. In the friction and comparison of these utterances similar terminations came in some instances to be set apart to denote the relations of grammar; in other instances the grammatical relations which lay implicit in the sentence-word were made explicit by its being set over against another sentence-word similarly employed elsewhere; and so it came in course of time to be what the Chinese would call an “empty word” with no presentative meaning of its own. Thus, on the one side, as M. Bergaigne has shown, the old adjectival suffix bha (bhi) in our own family of speech has become the sign of the dative and genitive cases (Latin ti-bi, dat., Old Slavonic te-be, gen.) just as the adjectival termination sya or tya (as in δημόσιος, “belonging to the people”) has become the sign of the genitive (ἱππο[σ]ιο); while, on the other side, the Chinese tsĭ̥ h‘ai, “to be hurt,” is literally “eat hurt,” and tshyeu tha̤n, “autumn,” “harvest-heaven.” The Chinese word can still be used indifferently as a noun, a verb, an adverb or the sign of a case much like such English words as silver and picture, and its place in the sentence alone determines in what sense it shall be construed. This is an excellent illustration of the early days of speech, when the sentence-words contained within themselves all the several parts of speech at once—all that was needed for a complete sentence; and it was only by bringing them into contact and contrast with other sentence-words, that they came to be restricted in their meaning and use, and to be reduced into mere “words.” Language never forgot the mode in which it had framed its first vocabulary, and the Greek and Roman, as much as the Red Indian of America, in framing their compounds instinctively stripped off the so-called inflections, and reduced the word they placed first to its simple stem. That part alone of the word which remained unchanged and unchangeable, could be made use of when the word was to be treated as simply a word and nothing more. The North American languages reflect more faithfully than the languages of the Old World the primitive condition of speech, and the North American languages can possess from six to eight thousand different verbal forms or sentences without having abstracted from them a single word which will express the sense of the verb out of all relation to anything else.[71] Thus, the Cheroki has thirteen verbs to denote particular kinds of “washing,” such as “washing the head,” or “the hands,” or “myself,” and each of these verbs has a multitude of forms, but no isolated word to denote “washing” in general has as yet been extracted from them.[72] The difficulty has often been noticed of getting a savage or barbarian to give the name of an object without incorporating it into a sentence or bringing it into relation with something else. Thus, a Kurd who supplied Dr. Sandwith with a vocabulary of the Zaza dialect, was so little able to conceive of words like “head,” “father,” “hair,” except as related to himself or some one else, that he had to combine them with a personal pronoun, saying sèrè-min, “my head,” piè-min, “my father,” porè-min, “my hair.” The Hoopah and Navaho vocabularies, published by Schoolcraft,[73] similarly prefix the possessive pronoun h’, hut to all their words, as hotsintah, hut-tah, “forehead,” huanah, hunnah, “eye,” hoithlani, hutcon, “arm;” and Dr. Latham points out the same fact in Wallace’s vocabularies from the river Uapes, where eri-bida, eri-numa in Uainambeu, tcho-kereu, tcho-ia in Juri, and no-dusia, no-nunia in Barrè, literally “my head,” “my mouth,” are given as the equivalents of simple “head” and “mouth.” He also states that he has noticed the same peculiarity among the English Gipsies.[74] The making of words as distinct from sentences was a long and laborious process, and there are many languages like those of North America in which the process has hardly yet begun. A dictionary is the result of reflection, and ages must elapse before a language can enter upon its reflective stage. Our children still learn the languages they speak by first acquiring the knowledge of certain phrases and sentences, and then gradually analyzing them into words, and the adult who wishes to gain a successful acquaintance with another tongue must pursue the same plan. What Steinthal says of the Chinese, that its “smallest real whole is a sentence, or at least a sentence-relation,”[75] is true of other languages as well, and the words of which a sentence is composed have no actual existence apart from that sentence, except for the phonologist and the lexicographer. Until the whole sentence is completed the individual words of which it consists have no more signification than the syllables ful and ness or cy and ly which occur so plentifully in English. The first condition of language is that it should be significant, and words are only significant when they stand in relation to one another. The logos, the true word, said Aristotle, was the cause of knowledge; the individual words of which it was composed were but symbols and tokens of the impressions of sense.

Now, if language be the embodiment of thought, and if thought can only express itself under the form of the complete sentence, it is plain that we must look to the sentence for a true classification of languages. The sentence expresses the way in which we think, and the different forms assumed by the sentence—that is to say, the different modes in which the relations of subject, object, and verb are denoted will constitute the only sound basis for classifying speech. The particular relation between the several ideas summed up in a judgment or sentence agrees with the manner in which we regard the objects about which we think and speak. If, for instance, we have no clear idea of any distinction between ourselves and the objects around us, in talking about them any reference to ourselves will be left out of sight. Instead of saying, “I am running,” where the speaker distinguishes himself from the act in which he is engaged, we should say like the Romans curro, where the personal pronoun has no separate and independent mark of its own. Different races of men do not think in the same way; and, consequently, the forms taken by the sentence in different languages are not the same. Thus in the so-called isolating languages, the separate terms or ideas which make up the sentence are not subordinated to each other, and fused into a single whole, but every word remains a separate and distinct sentence. The Chinaman has to say, “thya̤n-hi le̥ṅ tsyaṅ-s̆aṅ-lei”-literally, “heaven-air cold begin-rise-come,”—if he wants to state that “the weather began to be cold;” and the Burman’s way of expressing “we are going,” is by saying, “ṅā dō dhwā kra dhań”—“I multitude go multitude which.” In cases such as these, the ideas are each set down independently, instead of being subordinated one to another, and the words which embody them are accordingly contrasted with each other like so many independent sentences. On the other hand, in the agglutinative languages, the ideas which make up the sentence, though still kept distinct and independent, are no longer set over against one another, but brought into mutual relation and harmony, and regarded as of equal force and meaning. The root or stem still stands out clearly and separately, and the suffixes of relation are marked with equal distinctness; But for all that, the inward fact of the incipient subordination which exists between them is denoted by the outward fact of vocalic harmony, whereby the vowels of both stem and suffix have to belong to the same class. The Turkish sign of the infinitive, mak, has to become mek after a root like sev, “love,” though both root and suffix still retain their own individuality; and while at-lar is “horses,” ev-ler is “houses.” The grammatical relations expressed in the Aryan class of languages by case-endings and person-endings, or by prefixed pronouns and prepositions, have to be represented, as a general rule, by postfixes, since in no other way can sufficient emphasis be laid upon them, and the danger avoided of their being swallowed up in the verb or noun. Our “I love,” or “the man,” look but little different in writing from the Turkish sev-r-im, or the Basque gizoná, gizonák; the case is quite altered, however, when we try to pronounce these words, the accent falling on the verb in our “I love,” but allowing the distinction between verb and pronoun to be clearly felt in the Turkish sevrim. It is among the inhabitants of mountainous and cold regions in the Aryan and Semitic families of speech—among Albanians, Bulgarians, Scandinavians, and Aramæans—that the definite article is postfixed instead of being prefixed; and we can see at once what an emphasis and distinctness would be given to it by such a position. Only where foreign influences have been at work do the agglutinative languages change the order of the words in the sentence and, as in the case of the Hungarian definite article a, az, prefix the words expressive of the grammatical relations, instead of postfixing them. Still further, to mark out the several parts or terms of the sentence, the objective pronoun may be inserted between the subjective pronoun and the verbal root or stem; and so we may have a sentence-word like the French je vous donne, as in the Basque zamaztet (from eman, “to give”), or the converse arrangement of the terms, as in n-aza-zu-n, “that you may have me” (“me-have-you-may”). The incorporating languages, as they are called, are the oldest examples of the agglutinative class, for they go back to the time when the speaker had not yet begun to analyze his sentences, and when he could not say simply, “I give,” without finishing the sentence with the objective pronoun. Hence it is that in Basque we must say dituzte beren liburnac, “they have them their books,” instead of simply “they have their books;” and in Accadian, the language of primitive Chaldea, “I built a house” would be ê mu-n-rû,[76] literally “house I-it-built.”

Very distinct from these incorporating tongues are the polysynthetic or incapsulating dialects of America, in which the words that make up a sentence are stripped of their grammatical terminations, and then fused into a single word of monstrous length and appearance. Thus the Algonquin would say, wut-ap-pé-sit-tuk-quś-sun-noo-weht-unḱ-quoh, if he wished to express the sentence “he, falling on his knees, worshipped him;” and this cumbrous compound denotes exactly what we split up into seven words. These polysynthetic languages are an interesting survival of the early condition of language everywhere, and are but a fresh proof that America is in truth “the new world.” Primitive forms of speech that have elsewhere perished long ago still survive there, like the armadillo, to bear record of a bygone past. The conception of the sentence that underlies the polysynthetic dialects is the precise converse of that which underlies the isolating or the agglutinative groups. The several ideas into which the sentence may be analyzed, instead of being made equal or independent, are combined like a piece of mosaic into a single whole. The sentence has not passed beyond its primitive form, or rather that primitive form has been retained in spite of the growth and development of the languages to which it belongs. It is possible that the Eskimaux may be the descendants of the savage races who inhabited the caves of southern France, when the rivers were stiff with ice for half the year, and the reindeer roamed freely through the woods and meers; at all events, among the icebergs and dark winters of the North, they have preserved their old habits of thought, their old mode of viewing the world about them, almost unchanged. And yet our own class of speech, that class to which we give the name inflectional, and which we sometimes think is the crown and standard of all other kinds of language, is not so far removed in usage from the Eskimaux or the Algonquin as are the isolating dialects of China and the agglutinative jargons of Mongol and Turk. In the inflectional group the words or suffixes which denote grammatical relations are subordinated to the words which express objects or actions—that is to say, to nouns and verbs. The termination of the Latin currit has lost all distinct and independent meaning of its own; apart from the verbal stem to which it is subordinated, it is a mere flatus vocis, a mere empty sound. In flection proper, which we may see best exemplified in the Semitic tongues, the relations of grammar are denoted by internal vowel change—adamu, “man,” for instance, being nominative, adami genitive, and adama accusative. It was only afterwards, and by the force of analogy, that first unmeaning suffixes and then agglutinated words which were gradually assimilated to them, came to take the place of internal vowel change. What we may term the inflectional instinct sought to express the various relations of the sentence, as they successively rose to consciousness, out of the original sentence-word itself. When separate words like wards or ly (like) were afterwards employed for the same purpose, they first had to lose their own individuality, to become empty words, representative and not presentative, and as such to be engrafted upon the old stems. The Greek φη-μί, or the Sanskrit ad-mi, “I eat,” are single wholes; the first personal pronoun ma, weakened to mi, has lost all life of its own, and its sole right to existence lies in its absorption into the stems φη- and ad-. But an inflectional language cannot carry out its fundamental principle with logical completeness. All the subordinate relations of a sentence cannot be brought into the same close connection with the principal idea as in φημί and admi. Sentences like “I speak” or “I eat” may be comprehended under a single word; but there are many sentences where this is impossible, and where the attempt to express in language the relation between the principal and the subordinate, between the subject and the attribute, has to be given up. In the Latin poeta bonus, for example, the subject and the attribute appear as separate words; and there is nothing in the flection attached to each to show that they stand in any relation whatsoever one to the other. So far as the form goes there is nothing to tell us whether the two words mean “a good poet” or “the poet is good.” The fundamental principle of flection has been violated, and the language is on the high road to that more developed condition in which, as in Chinese, the two ideas are set plainly and distinctly one against the other, and the mind is left to supply the relation between them. This impossibility of carrying out thoroughly the principle of flection brings about an analytic tendency in all inflectional forms of speech. The longer an inflectional language lives the more analytic it becomes. The Englishman says “I will go,” and the Frenchman le monde, where the Latin was contented with ibo and mundus. One by one the grammatical relations implied in an inflectional compound are brought out, as it were, into full relief, and provided with special forms in which to be expressed; but the change that has taken place is but an apparent one, the inflectional spirit of the language still remains; and though we write “he runs,” “I will go,” we pronounce as if they were single words. The pronoun and the verb, taken apart and by themselves, convey no meaning to our minds; we have to combine them before they become significant, and (the order of the words excepted) there is but slight difference between an English sentence like “never to be sufficiently relied upon,” and the Tamil sārndāykku, “to thee that hast approached,” to be analyzed into sār, “approach,” d sign of the past, āy, “thee,” and ku, “to.”

Each of the leading classes of speech naturally comprises various species or subdivisions. Thus the isolating Chinese differs from the isolating dialects of Further India, in that the Chinese mode of expressing the relations of the sentence by position is replaced in these by the use of words like prū, “do;” khã, “suffer;” khōṅ, “possession,” mha, “from.” So, again, in the agglutinative class, the Bâ-ntu languages of Southern Africa prefix the same substantive, worn down, it may be, to a mere unmeaning symbol, to each of the words in a sentence which have to be brought into relation with each other; o-ka-ti k-etu o-ka-ua, for instance, being “our fine stick,” or literally, “stick ours fine.” The Malayo-Polynesian dialects have not yet attained to the conception of the verb; thus yaku imukul olo (“I smitten people”) is “I am smitten by the people;” iṅga̤ra̤-ku ia̤ tatau (“my-thought he rich”) “I thought he is rich;” ia̤ baklambi baputi (“he with-jacket with-white”) “he has a white jacket on.” Basque grammarians generally hold that the Basque has but two verbs, “to be” and “to have,” while, on the other hand, there are many languages which lack precisely these two.

But in all these sub-classes, just as in the main classes of speech, it is the different conception of the sentence and the form it takes which characterize the whole language. However much alike may have been the circumstances by which the first communities of men were surrounded, they yet viewed the world without them and their own relation to it with different eyes. The idea they formed of the sentence and its parts was not the same everywhere. When with the growth of consciousness came also the formal expression in utterance of the relations of the several parts of the sentence, it was inevitable that this expression should clothe itself in essentially various forms. And the psychological peculiarity which originated each of these forms—a peculiarity itself the result of previous experiences and tendencies—became continually more definite, more confirmed, more unalterable. The logician may reduce all forms of the affirmative proposition or judgment to the single “A is B,” but the grammarian knows that this is like the jus gentium of the Romans, a mere abstraction from a limited number of observed instances. It may be the right form for the sentence to take in the manifold languages of the world, but as a matter of fact it has never been taken in any one of them. The form of the sentence as shaped by the primitive language-builders of each human community has imprinted itself indelibly upon the linguistic consciousness of their successors. Racial type and characteristics will change as soon as the conception of the sentence. Many of the agglutinative languages have approached so nearly to the phænomena of inflection, as to make it difficult to determine why they should not be classed with the inflectional tongues; and yet for all that they remain agglutinative, and have remained so as far back as we can trace them. Our own language is agglutinative, and even isolating in many respects, while the French je vous donne seems a clear instance of incorporation. The Chinese, on the other hand, shows much that is agglutinative, much even that resembles inflection, and it is only the polysynthetic languages of America that remain true to their stereotyped primæval character. Nevertheless, in spite of all this apparent confusion and overlapping, this borrowing, as it were, of characteristics from other families of speech, the great types of language stand out each of them visibly and distinctly. Their broad characteristics can be clearly sketched, their essential diversity easily felt. It is only when we come to map out the boundaries between them, to determine where isolation ends and agglutination begins, that we find ourselves at fault. Here as elsewhere in nature there is no sharply-defined line of division to be drawn; species passes gradually and insensibly into species, class into class. But in spite of this, species and classes really exist, each with its own type and characteristics, each founded upon its own conception of the sentence and its parts. When we remember that the sentence, and not the isolated word, is the starting-point of philology—when we make it what the logician would term the fundamentum divisionis for our classification of speech—there is no longer any difficulty in distinguishing between the several families of speech, and assigning to each its character and place. The Finnic idioms have become so nearly inflectional as to have led a recent scholar to suggest their relationship to our Aryan group; nevertheless, they have never cleared the magical frontier between flection and agglutination, hard as it may be to define, since to pass from agglutination to inflection is to revolutionize the whole system of thought and language and the basis on which it rests, and to break with the past psychological history and tendencies of a speech. There are South American butterflies whose colours have come to resemble so closely those of the plants on which they are found as to be indistinguishable from them; for all that, the butterfly still remains a butterfly, and the plant a plant.

Such, then, is language in its origin and its nature. It is significant sound, the outward embodiment and expression, however imperfect, of thought. Before sound can become significant it must express the whole thought or judgment; that is, it must take the form of a sentence. Historically, the sentence and not the word comes first. The sentence consists of two factors, one the external sound, the other the internal thought, and neither of these factors can be disregarded by a true science of language.

Now, science is accurate knowledge. The statement may seem a truism, but it is a truism which has sometimes been forgotten. For that which is accurate is only that which can be defined and limited, that of which all the boundaries, as it were, are distinctly mapped out and known. But the boundaries of knowledge can only be discovered by the help of comparison. It is, in fact, the comparative method that constitutes the very life of inductive science; it is the application of the comparative method to any subject which brings that subject within the domain of scientific knowledge. Our knowledge that night and day follow one another alternately, or that if we put our hands into the fire they will be burnt, is not yet scientific. In order to know anything scientifically we must be able to compare it with something else, and so determine its size, or weight, or character. Our feelings may tell us that the atmosphere is hot or cold, but we have no scientific knowledge of either fact until we can measure one degree of heat or cold against another by means of the thermometer. As soon as we know the exact amount and character of each degree of heat or cold, we have laid the foundations of a science of thermology. It is just the same in the case of language. Here, too, as soon as we can compare languages and the elements of languages together, and so measure and determine their character, we shall have the beginning of a science of language. But the comparison must be made by the aid of a common standard. The old attempts to compare Latin with Greek, or both with Hebrew, were failures because the test applied was a capricious one, depending on the subjective fancies and prejudices of the inquirer. We cannot compare two things together without having a third term—a common standard by which to measure them. We must not have one rule and measure for one set of words or languages and another rule and measure for another set. The comparative method we employ must be alike in all cases.