Introduction to the science of language, Volume 1 (of 2)

by A. H. Sayce

Besides flectional suffixes, then, classificatory or formative suffixes also may ultimately be due to the process of composition. Upon them, too, analogy will have worked its influence, assimilating them to the other suffixes which in course of time they had come to resemble. When composition had once reduced a word to the condition of a mere adjunct of another word, there was no reason why it should not be put to the same uses as other similar adjuncts. When the root bhar, “to bear,” in such Latin compounds as leti-fer could no longer be distinguished from the suffix -tio(n) in words like na-tio, it was naturally treated in the same way.

But it does not follow, as a good number of writers on language have assumed, that because some of the classificatory suffixes are examples of composition, all of them are so, any more than in the case of flection and the flectional suffixes. Indeed, we have only to glance at the numerous suffixes employed by our own Aryan family of speech in forming or “building” words to see how impossible it would be to trace back a large proportion of them to independent words. How, for instance, could we claim any such origin for the suffixes -la- and -ra- in querela and λαμπρὸς, or the suffixes -ana-, -na, and -an in pecten, donum, and ἱκανός? With such suffixes all we can do is to watch the changes they have undergone, or caused other sounds to undergo, through the action of phonetic decay and false analogy. Thus in Latin where the combination sr changes into the softer br, stems like ceres (Sanskrit śiras), “head,” and fes (as in festus) have turned into cerebrum and Februus when combined with the suffix -ra; and if we take the suffix as itself, we shall find its sibilant passing into r before another vowel, and so originating a long series of curious transformations. The r which we get in the genitive of temporis was transferred by analogy to the nominative also, where no vowel followed it, and though there was a struggle at first between the twin forms in s and r, traces of which survive in the twin arbos and arbor, the later and incorrect form with r finally carried the day, and classical Latin knows only of a sopor, not a sopos. But it may be asked why should the penultimate syllable of sopōris be long whereas it is short in tempŏris and arbŏris, and why, too, should sopor be masculine while tempus is neuter? Here, again, false analogy has been at work. A certain number of masculine nouns terminating in -tor and denoting agents, like dator or victor, existed in the language, and when sopos was changed to sopor, it was assimilated to these both in gender and in declension. Even victor, however, had passed under the action of false analogy. When we compare the Latin victor with pater, or the Greek σωτήρ with πατήρ, it is at once clear that we are dealing in each case with the same suffix, although in victor the vowel has been thickened into the fuller o. But while victor and σωτήρ have a long vowel in the oblique cases, this is not the case with the much older words pater and πατήρ (accusative πατέρα). It is evident, therefore, that this long vowel must have been a sort of after-thought; and so, in fact, it was. First of all the vowel of the nominative was lengthened to compensate for the loss of the final sibilant (paters), and the quantity of the vowel in the nominative was then analogically extended to the other cases as well. How far this was from having been originally the case may be gathered from another form of the same suffix which we have in the Sanskrit patram, the Greek πτέρον, and the Latin ara-tr-um. Here the vowel between the two consonants of the suffix has disappeared altogether, as it has also in words like the Latin sæclum for sæ-culu-m, or the Gothic nê-thla, our needle, where the suffix, in spite of the change it has suffered, really goes back to tar. The latter group of words (in tar), however, is distinguished from the former (in trum) in both signification and gender, the masculine agent being replaced by a neuter noun of instrumentality. We can easily see how such a transition of meaning must have come about. The agent presupposes the act just as much as the act presupposes the agent. Agent and act, in fact, are co-relative terms, and the parent-Aryan distinguished them, not by the classificatory suffix—for they both belonged to the same class—but by the flectional suffix, which was in the one case -s in the nominative singular, and in the other -m. The Latin trucidator and the English murder (formerly murther, like slaugh-ter and laugh-ter) have precisely the same suffix, and it is only a recollection of the difference in meaning in the flectional suffixes which has survived their loss that prevents them from being used with the same signification. Even these flectional suffixes themselves—as we shall see hereafter—did not originally imply that difference of meaning to the expression of which they were afterwards adapted. In nouns like the Latin virus or the Sanskrit śiras-, the final sibilant denoted a neuter rather than a masculine or a feminine, while servum or humum show that the final labial might characterize the objective case of both masculine and feminine nouns.

The suffix tar (ter) brings us back to those classificatory suffixes which trace their descent from independent words, if, as is very probable, we have to connect it with the root found in our through, the Latin trans and ter-minus, the Zend tarô, “across,” the Sanskrit tar-âmi, “I pass over,” and perhaps, too, the numeral tri, tres, three.[258] It is not difficult to understand how a word signifying “to go through with a thing,” could be taken to form nouns of agency. What more suitable description could be given of “a giver” than “one who goes through with giving,” dator(s)? The antiquity of this use of the suffix in our family of speech may be gathered from the fact that it is employed to form those nouns of relationship which are the first to require a name. Brother, sister, daughter, mother, father, all contain this ancient suffix. Brother (bhrâ-tar) is “the bearer,” from the root bhar, daughter, “the milker” or rather “sucker,” from the root dugh, while the Sanskrit grammarians derive father (pitar) and mother (mâtar) from the roots pâ and mâ, which respectively mean “to defend” and “to create.” It is obvious, however, that both “father” and “mother” must have received names long before it was necessary to speak of “going across” or “passing through,” and that our Aryan ancestors would not have waited to compound two words together before giving names to the nearest and dearest of relationships. As a matter of fact, in almost all languages names have been found for the parent in the two simple labial utterances pa and ma; and the identity of these with the Aryan roots pâ and mâ must be a pure accident. What seems to have happened in the case of our names of relationship was this. When the Aryan family first comes before us in the records of speech, it is as a civilized clan with a vast but indeterminate background of unknown history lying behind them. They had long since entered upon what may be termed the epithetic stage, when man discovered that he was a poet, and began to invent epithets for the objects about him, and to form compounds. It was at this stage of culture and civilization that the Aryan community coined compound epithets for brother, for daughter, and for sister, which succeeded in driving out and replacing the older words that had preceded them. The new compounds in tar took the fancy of the community, and were widely extended by the force of analogy. The old labials which had done duty for the ideas of “father” and “mother” followed the fashion set by the younger names of relationship, and so just as bhrâ-tar had come to signify “brother,” pa-tar and mâ-tar came to signify “father” and “mother.”

Languages do not begin with composition. If the sentence is anterior to the word, a considerable time must elapse between the first beginnings of a language and the piecing together of two independent words. Isolating tongues like the Chinese or the Burman, where so much use is made of composition in order to create new conceptions or to define old ones, are shown by this very fact to have passed into a decrepit stage of existence. The epithetic stage is one far advanced in the history of a speech; it implies poetic imagination, a certain measure of culture and civilization, and the germs of a mythology. The new compounds of this epithetic stage follow the genius and analogy of the language to which they belong. If the formation of words depends largely on the use of suffixes, the newly coined words will in time adapt themselves to the old rule; what were once independent words will become suffixes, and be employed in exactly the same way as the other suffixes of the language.

The very existence, then, of classificatory suffixes due to composition in our Indo-European idioms implies the existence of earlier suffixes for which we cannot claim a similar origin. We have already seen that this is the case with many of the suffixes which serve the purposes of flection; though the person-endings of the verb go back to separate words, every attempt to discover such a derivation for the principal case-endings has ended in failure. What is true of the case-endings is pre-eminently true of those suffixes which are neither flectional nor classificatory. If we analyze the Latin alumnus, we find first of all the flectional suffix -(u)s, then the classificatory suffix mino, which relegates the word to the same class of middle participles as the Greek τυπτόμενος, and lastly, the suffix u, which intervenes between the root al and the classificatory suffix mino. We may call this u a “connecting-vowel,” or “an euphonic vowel,” or anything else we choose, but the fact remains that it is a suffix which can be separated from the root al. It is a suffix, however, which is neither flectional nor classificatory, and may be termed secondary for want of a better name. Secondary suffixes play an important part in our family of speech, and just as a flectional suffix often appears as a classificatory one, so, too, a classificatory suffix may appear as a secondary one. If, for example, we compare a word like civitas (civ-i-ta-t-s) with sec-ta, we may not only get the secondary suffix -i-, following immediately upon the root, but also a reduplication of the classificatory suffix ta, which here at least can have no classificatory sense. We may accordingly define a secondary suffix as one which does not refer the word of which it forms a part to any particular class; and where we have several classificatory suffixes amalgamated together the first of these have generally become secondary. Thus the English songstress is a combination of two suffixes, one Saxon and the other Romanic, which equally denoted the feminine. By the side of sang-ere, “the singer,” stood in Anglo-Saxon sang-estre, “the songstress;” it was only when the classificatory significance of the termination had died out that a new one which really went back to the Greek -ισσα through the Latin issa (as in abbatissa), and the French -esse (as in justesse),[259] was attached to it, and so the old classificatory suffix became a merely secondary one. In fact, as soon as the force of a classificatory suffix has been weakened in a word, a fresh classificatory suffix is always ready to be attached to it, just as children will talk of more-er and most-est, or as Lord Brougham introduced the equally anomalous worser.

Now these secondary suffixes play a most important part in a large number of languages, and more especially in our own Aryan ones. It is seldom that a classificatory or flectional suffix can be added immediately to the root, as in the Sanskrit ad-mi, “I eat;” a secondary suffix has usually to intervene, by means of which the root is raised to what has been variously termed a base, a theme, or a stem. So far as the Indo-European family of speech is concerned, it is probable that even such exceptions to the general rule as that of ad-mi are really due to phonetic decay, which has worn away the original stem to a simple monosyllable, as it has done in so many English words like man or fall. When we come to deal with roots, we shall see good reason for believing that they were all or for the most part once dissyllabic, and the tendency that many children show to turn the monosyllables of modern English into dissyllabic words may be but an instinctive reversion to the early type of speech. No doubt it is very possible that just as classificatory suffixes have been changed into secondary ones, so on the other hand secondary suffixes may have come in course of time to assume a classificatory character. A conspicuous example of this may be found in the suffix ya, which in Greek words like φέρουσα for φερο-ντ-yα, or δότειρα for δοτ-ερ-yα, has become a mark of the feminine gender. A distinction of gender is by no means engrained in the nature of things, and the majority of spoken languages, such as most of those which are agglutinative or isolating, know nothing at all of it. In some idioms, those of the Eskimo, Chocktaw, Mushtogee, and Caddo, for instance,[260] the place of gender is taken by the division of objects into animate and inanimate, while elsewhere they are divided into rational and irrational. In the Bâ-ntu dialects of South Africa, nouns are separated into a number of classes, in one case as many as eighteen, by means of prefixes which were originally substantives like our -dom, -ship, or -hood; and the agreement of the pronoun, adjective, and verb with the substantive is denoted by the employment of the same suffix. Bleek has not inaptly compared these classes of the Bâ-ntu noun with the genders of our own family of speech. Thus if we were to take a noun like I-SI-zwe, “nation,” which belongs to the si-class or gender, in order to express the sentence “our fine nation appears, and we love it,” the Kafir would have to say I-SI-zwe S-etu E-SI-χ’le SI-ya-bonakala si-SI-tanda, literally “nation ours appears, we-it-love.” Similarly the noun U-LU-ti, “stick,” would require a corresponding change of prefix in the words in agreement with it; and the sentence would run: U-LU-ti LW-etu O-LU-χ’le LU-ya-bonakala si-LU-tanda.[261] There are many indications that the Aryan language, or rather the ancestor of that hypothetical speech which we term the parent-Aryan, was once itself without any signs of gender. We have only to turn to Latin and Greek to see that the words which denote “father” and “mother,” pater and mater, πατὴρ and μητὴρ, have exactly the same termination, while so-called diphthongal stems as well as stems in i (ya) and u (like ναῦς and νέκυς, πόλις and λῖς) may be indifferently masculine and feminine. Even stems in o and a, though the first are generally masculine and the second generally feminine, by no means invariably maintain the rule, and feminines like humus and ὁδός or masculines like advena and πολίτης show us that there was a time when these stems also indicated no particular gender, but owed their subsequent adaptation, the one to mark the masculine and the other to mark the feminine, to the influence of analogy. How analogy came to act seems to have been as follows. First of all the idea of gender was suggested by the difference between man and woman, male and female, and, as in so many languages at the present day, was represented not by any outward sign, but by the meaning of the words themselves. Thus in the Hidacha of North America we are told that “gender is distinguished by using, for the masculine and feminine, different words, which may either stand alone or be added to nouns of the common gender,”[262] and in the Sonorian languages further south it can only be denoted by the addition of words which signify “man” and “woman.”[263] Then when the conception of gender had once been arrived at it was extended to other objects besides those to which it properly belongs. The primitive Aryan had not yet distinguished the object thought of from the subject that thought of it; he was still in the stage of childhood, and just as he transferred the actions and attributes of inanimate objects to himself, so too he transferred to them the actions and attributes of himself, and endowed them with a life similar to his own. The same age which saw the creation and growth of a mythology saw also the origin of gender in nouns, and the distinction of gender in the demonstrative pronouns, due to their reference to animate beings, reacted on the nouns expressive of inanimate objects to which they likewise referred. As soon as the preponderant number of stems in o in daily use had come to be regarded as masculine on account of their meaning, other stems in o, whatever might be their signification, had to follow the general rule and be classed as masculine nouns. How readily the gender of a word may be determined by its termination has been already seen in the history of the Latin stems in -os. Here and there the constant use of a word with particular pronouns or its obvious and natural meaning resisted the common tendency, and hence the preservation of such anomalies as ὁδός, humus,[264] and advena mentioned above. The suffix ya, however, like the suffix -ιδ- (as in αὐλητρίς) in Greek or the suffix -ic- (as in victrix) in Latin, formed part of a class of words which all followed the dominant type; neither use nor meaning interfered with the appropriation of them all to express the feminine gender. The accident by which the suffix was attached to words which chiefly denoted female agents eventually caused it to become a classificatory instead of remaining a mere secondary suffix. But the Aryans were not contented with only two genders, as the Semites and some other races were. A time came when the Aryan awoke to the consciousness that he was essentially different from the objects about him, that the life with which he had clothed them was really but the reflection of his own. He began to distinguish the agent from the patient, and to turn his middle conjugation into a passive one. The first sign of this new-grown consciousness was the formation of a nominative for the first personal pronoun; ego, ἐγών, the Sanskrit aham, is a far later creation than the objective me or mâ, and whether it be a compound or not, as some scholars believe, at all events it marks the epoch when the “me” became an “I.” The discovery had been made that a difference existed between the nominative and the accusative. But this difference existed only in the case of animate beings, or of those objects which the custom of language and the habits of thought it had produced regarded as animate; there was another class of objects and ideas which were beginning to require a name and yet could not be reckoned as coming under either of the two genders with which the language was already acquainted. The same development of thought which had revealed the distinction between subject and object brought with it also the conception of abstracts or general terms. Besides the individual trees which had long ago received their names, the idea of “tree” itself now needed a word to express it, and the speaker was no longer contented with detailing his single utterances one by one, but wanted a general term like “word” or “speech” wherein to sum them up. And so the new class of neuter nouns came into existence, which were really nothing more than old accusative cases or bare stems used as nominatives and given a separate life of their own. So far as form goes, the Greek δένδρον and ἔπος cannot be distinguished from λόγον and ὄπες, the Sanskrit vâchas representing both ἔπος and ὄπες alike, any more than the Latin regnum and vulgus can be distinguished from dominum and reges. In the pronouns the bare stem in t or d, which had once served for all cases and all genders, was set apart for neuter nouns, and the Aryan declension was made complete with its encumbrance of three genders, which it has needed the practical genius of the English language to shake off. The further changes that took place in the distribution of these three genders must be described by the historical grammars of the special languages of the Aryan family: the age came when their original meaning and intention was as much forgotten as that of mythology; they were looked upon as the functions of certain suffixes which thus became classificatory, and, as in Latin stems in -as or French nouns like mer which owe their gender to the confusion of the plural nominative maria with the singular nominative of musa, they became the sport and puppet of false analogy. The mixture of dialects which varied as to the genders they assigned to particular nouns completed the confusion, and modern German is an instance of a language which still clings to an outward excrescence of speech which originated in childish habits of thought and has now lost all sense and reason for its existence. A mere tax upon the memory and an embarrassment to free literary expression, it is no wonder that German genders are a sore trial to the children, who are sometimes several years before they learn to use them correctly. In this respect they resemble the Swedish peasantry, who are said to find an equal difficulty with the genders of their own tongue.

The origin of gender is one of the questions belonging to what some German scholars have termed “the metaphysics of language.” The metaphysics of language deals with the source and nature of grammatical ideas as distinct from the phonetic machinery by which they are expressed; it seeks by a comparison, firstly of cognate dialects and then of families of speech, to discover the conception which lay at the bottom of such grammatical facts as gender, number, and the like. We want to know not merely how the relations between the several parts of the sentence are expressed, but what those relations actually are. The idea must exist before phonetic means are adapted to represent it, and in order to reach it we must scientifically trace the history of the phonetic means. The metaphysics of speech, therefore, is but the second branch and division of its morphology, bearing the same relation to the inquiry into the growth and origin of stems and suffixes and suchlike phonetic forms of grammar that sematology does to phonology. The morphology of language is as much concerned with grammatical ideas as with the external form in which they are embodied. It is these grammatical ideas more than their phonetic embodiment that constitute the structure of a tongue.

Let us see, for example, whether we can track the conception of number back to its first starting-point. Strange as it may seem there are some uncivilized languages which make as little distinction between the singular and the plural as we do ourselves when we use words like sheep. Thus Mr. Matthews states that “Hidatsa nouns suffer no change of form to indicate the difference between singular and plural,”[265] and in the Sonorian tongues, according to Buschmann,[266] “the simple word in the singular serves also for the plural,” while the monosyllabic Othomi can distinguish between singular and plural only by the prefixed article na and ya,[267] and the Amara of Africa can only say fŭrŭsn ayŭhu, “I have seen horse,” leaving the hearer to decide whether the horse is one or many. In spite of the vast length of time during which these languages have been shaping and perfecting themselves, the conception of number is still so far from being consciously realized that no phonetic means have yet been adapted or devised to express it. If we turn to the Tumali of Africa we find in the case of the personal pronouns ngi, “I,” ngo, “thou,” and ngu, “he,” a slight advance upon this poverty of thought. Here the plural is denoted by the postposition da, “with,” so that ngi-n-da, “we,” is literally “(some one) with me.” The mind has come to distinguish between itself and that which is outside itself, to realize, in fact, that it has an individual existence distinct from that of some one else, and so the conception of duality is attained. At this conception mankind stopped for a long while; indeed, there are many races and tribes who have not even yet passed beyond it. Wherever the so-called plural is formed by means of reduplication—that is to say, wherever the doubling of a thing is the furthest point of multiplicity to which the mind can reach, there we have not yet a true plural, but only a dual. All over the world reduplication seems to have been the earliest contrivance for denoting something beyond the singular, and to this day in Bushman, as in many other savage jargons, it serves for a plural.[268] The same evidence that is borne by the so-called reduplicated plural is borne also by the numerals. The aborigines of Victoria, according to Mr. Stanbridge, “have no name for numerals above two;”[269] the Puris of South America call “three” prica or “many,” which is also the original meaning of the same numeral in Bushman, and “the New Hollanders,” says Mr. Oldfield of the western tribes, “have no names for numbers beyond two.” It is even possible, as has been already noticed, that our own Aryan tri, three, goes back to the same root as that of the Sanskrit tar-ô-mi, “I pass beyond,” and once signified nothing more than that which is “beyond” two. The fact that the conception of duality preceded the conception of plurality, explains how it is that the seemingly useless dual has been preserved in so many languages by the side of the plural. It is a relic of a bygone epoch, a survival, as Mr. Tylor would call it, which tends to be more and more restricted in use until it disappears altogether. In both Aryan and Semitic the dual appears only as an archaic and perishing form. The Æolic, in this as in the throwing back of the accent, the least conservative of the Greek dialects, has lost it entirely; the Latin keeps it merely in duo, octo, and ambo, and if we pass to the Semitic idioms, the dual of the noun is preserved only in words which denote natural pairs like “the eyes” or “the ears,” while in the verb it has been maintained by Arabic alone, and in some exceptional cases by Assyrian. Language, however, did not always proceed at once from the dual to the plural, from the conception, that is, of limited plurality to the conception of unlimited plurality. Many languages possess a trinal number, or what are called inclusive and exclusive forms of the personal pronouns, and in one of the Melanesian idioms, as well as in Vitian or Fijian, we even find a quadruple number formed by the attachment of tavatz or tovatz, “four,” to the pronouns na, “you,” and dra, “we.”[270] In Cheroki the dual of the first person has one form when one of two persons speaks to the other, another form when the one speaks of the other to a third, inaluiha being “we two (i.e. thou and I) are tying it;” awstaluiha, “we two (i.e. he and I) are tying it.” In Annatom, again, aniyak is “I,” akaijan, “you two + I,” ajumrau, “you two - I,” akataij, “you three + I,” aijumtaij, “you three - I.” More usually the reduplicated dual led to a plural without the intervention of a trinal number, or the plural was denoted by some word like “multitude” or “heap,” which in course of time came to be a plural sign, just as in other instances it came to signify the numeral “three.” In the Aryan languages M. Bergaigne has shown[271] that the plural of the weak cases (nominative, accusative, and vocative) was identical with the singular of abstract nouns, and their formatives, -as or -âs, -i or -î, -â or -yâ, and -an, continued to the last to mark abstracts like the Sanskrit áhan, “the day,” lipi, “writing,” vrajyâ, “the act of travelling,” or mudâ, “joy.” So in Semitic Assyrian, where an abstract is generally regarded as feminine, the feminine plural in -utu has become the termination of singular nouns like śarrutu, “a kingdom,” and then by a curious change of function been appropriated to a certain class of masculine plurals. There are reasons for thinking that the Semitic plural has been based on the dual; however this may be, the suffixes of the Aryan plural, so far at least as the weak cases are concerned, are suffixes which we find elsewhere used as secondary and not classificatory ones.

Even the genitive case, necessary as it appears to us to be, once had no existence, as indeed it still has none in groups of languages like the Taic or the Malay. Instead of the genitive, we here have two nouns placed in apposition to one another, two individuals, as it were, set side by side without any effort being made to determine their exact relations beyond the mere fact that one precedes the other, and is therefore thought of first. Which of the two should thus precede depended on the psychological point of view of the primitive speaker. We are all acquainted with the distinction between the objective genitive where the governed word is the object of the other, as in amor Socratis, “love felt for Socrates,” and the subjective genitive where the converse is the case, as in Socratis amor, “love felt by Socrates,” and this distinction has led to two different conceptions of the genitive relation being formed by different races. In the Aryan family, for instance, the genitive must precede its governing noun; Horsetown, equally with horse’s town, means “town of the horse.” In Semitic, on the contrary, the position of the words is reversed; here the genitive has to follow, not precede. Perhaps we may see in the position of the genitive in the two great inflectional families of speech a symbol of the characters of the two races. The Aryan, the inventor of induction and the scientific method, fixes his first attention on the phænomenon and traces it up to its source; the Semite, on the other hand, makes the first cause his starting-point, and derives therefrom with easy assurance all the varying phænomena that surround him.

Now, this apposition of two nouns, which still serves the purpose of the genitive in many languages, might be regarded either as attributive or as predicative. If predicative, then the two contrasted nouns formed a complete sentence, “cup gold,” for instance, being equivalent to “the cup is gold.” If attributive, then one of the two nouns took the place of an adjective, “gold cup” being nothing more than “a golden cup.” The apposition of two substantives is thus the germ out of which no less than three grammatical conceptions have developed—those of the genitive, of the predicate, and of the adjective. It is but another instance of that principle of differentiation which we have found at work upon the phonetic forms whereby the relations of grammar are expressed. Dr. Friedrich Müller has observed[272] that, as a general rule, the attribute and the genitive, or as he terms it the possessive, occupy the same place, and are treated as one and the same relation. In Hottentot, as in Chinese, where the defining noun must precede that which is defined, “right-path” means equally “the right path” and “the path of right,” and our own English language is another example of the same usage. In Malay, on the contrary, as in the Semitic tongues, both adjective and genitive have to follow the noun they define; thus the Malayan ōran ūtan, or “man of the wood,” is literally “man-wood,” and gūmin besar, “a great mountain,” “mountain-great.” On the other hand, the predicative relation is marked off from the attributive and genitival by a converse order of words; in Malay, for instance, the predicate is placed before its subject, as in besar gūmin, “great (is) the mountain,” and the Semitic perfect is formed by affixing the pronouns of the first and second persons to a participle or verbal noun.[273] These primitive contrivances for distinguishing between the predicate, the attribute, and the genitive, when the three ideas had in the course of ages been evolved by the mind of the speaker, gradually gave way to the later and more refined machinery of suffixes, auxiliaries, and the like.