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

by A. H. Sayce

A morphological review of the languages of the world reveals one curious and significant fact. Particular types of language belong to particular localities. In other words, a morphological classification of speech is also a geographical one. The polysynthetic idioms are characteristic of America, the isolating dialects of the extreme east of Asia. So, too, the leading inflectional families of speech, the Aryan and the Semitic, have both proceeded, it would seem, from Western Asia, like the Alarodian family, also inflectional, and best represented by the modern Georgian. The prefix-pronominal languages are confined to Southern Africa, as the incorporating Basque to the Pyrenees and the verbless Malayo-Polynesian to the islands of the Pacific. This fact would go to show that the distant emigration of languages, like the distant emigration of races, is very exceptional and chiefly characteristic of the higher species with their greater energy and expansiveness. The wanderings of savage tribes are circumscribed by the climatic and other conditions to which they are peculiarly subject. Without canoes voyages cannot be taken, and mountains, rivers, deserts, or stronger neighbours are all obstacles to movement more or less insurmountable. The fact would also go to show that it is only within the area peculiar to a certain class of languages that we may look for their progress and development. It is only in Eastern Asia or in America that we can hope to discover the highest development of which an isolating or a polysynthetic language is capable, and so regard Chinese and Mexican not as “arrested growths,” but as instinct with the progressive intelligence and cultivated life of the peoples that speak them. Where no traces of a type of speech different from the prevailing one are to be found, we are justified in concluding that it never existed there. And finally the fact will correct that tendency we all have to assume a unity upon insufficient evidence. Types of language, like types of race, are as strongly marked off from one another as the countries to which they belong. Polysynthetism is as much characteristic of America as the hatchet face and red skin of the aboriginal; isolation of Eastern Asia as the yellow skin and oblique eyes of the Chinaman or the Burman. Modern discoveries are gradually producing a conviction that the civilizations of China, of Babylonia, and of Egypt were all independent and self-evolved. Such at all events is the case with their modes of writing, the best product of any civilization, and no one can study the character of these three civilizations without perceiving that they are radically distinct. Egypt, when the monuments first cast light upon her some 6,000 years ago, is in the height of her culture and advancement; but she comes before us as a pharos of light in the midst of utter darkness, self-contained and self-sufficient, but surrounded on all sides by tribes and nations even more barbarous than the untaught Negro of to-day. And such as was the civilization, such too was the language; the civilizations of the Nile, of the Euphrates, and of the Hoang-ho, were not more isolated and peculiar than the languages which embodied them. It is difficult for us with our steamers and railways and telegraphs to realize the separation and practical immobility of the ancient world. Geographical barriers cut off tribe from tribe, race from race, language from language, and war instead of peace was the sole means that existed of overcoming them. It is to these barriers, however, that we owe the persistency of racial and linguistic type which we may still note in so many parts of the world. It has often been remarked that the fauna and flora of America take us back to a geological rather than a historical age; the same may also emphatically be said of the American type of speech. The Eskimaux may or may not be the survivor of the man of the reindeer age; his grammar, at all events, is a relic of a bygone era of speech.

The morphology of speech, then, deals with the relation of the parts of the sentence one to another. This relation is expressed by what we term grammatical forms. Position, it is true, as well as accent, frequently takes the place of grammatical forms, especially in languages like Burman or English, but in this case both position and accent will have to be considered as belonging to the province of morphology. The rule which in Burman makes the first of two substantives a genitive or in English a substantive which follows a transitive verb an accusative is itself a grammatical form. Even in those tongues in which the expression of grammatical relations is fullest and most exact, there is much that can never be expressed by outward means, but only hinted at and understood. “The rudest of men,” says Chaignet,[242] “are yet sages; ils s’entendent à demi-mot; ils parlent par sous-entendus.” “It is,” as he goes on to observe, “the gesture, the tone, the connection of the sense or its abrupt breaking off, the undefinable and speaking expression of the face, that supply and complete our thought, marking its relations, or more truly its formal side, its most spiritual element, whereby language raises itself above mere sensation and matter.” The structure of a language is determined not only by the general type, isolating, agglutinative, or otherwise, to which it conforms, but also by the mode in which its words are linked together, by the way in which its grammatical forms are used and connected, and by the greater or less extent to which the quickness of the hearer in understanding what is not expressed is called upon. Structurally, Coptic belongs to the inflectional class of tongues, but among these it is distinguished by its prefixing its grammatical forms instead of affixing them, as was the case with its parent the Old Egyptian.

We must not forget, however, that whether in Coptic or Old Egyptian, or any other language, the grammatical form, the relation to be expressed, the idea to be developed and formulated, lay quite as much in the mere act of prefixing or affixing as in the sounds which were prefixed or affixed. The Sanskrit ad-mi means “I eat,” not only because it is a compound of a verbal stem or root signifying “eating,” and the personal pronoun mi, but because the pronoun is attached to the stem in such a way as to convey the conception of the relation intended to exist between the two ideas “eating” and “I.” We may therefore lay down that one of the modes adopted by language for denoting the relations of grammar is (1) the attachment of prefixes or affixes which may or may not be significant when used alone. (2) A second is the insertion of what are called infixes, as in Dayak, where from kan, “to eat,” the stem k-um-an comes, or in Malay, where by the side of ka-kan and ma-kan we have also k-um-akan. So, too, in Tagala we find b-in-atin for in-batin, just as in the secondary conjugations of the Semitic verb, iphteal, iphtael, istaphal, the suffix ta is inserted between the first and second consonants of the root instead of being prefixed as elsewhere. No doubt, metathesis aided by analogy was the primary cause of this curious phænomenon, as it is in the Sanskrit yu-na-j-mi, “I join,” instead of yuj-na-mi corresponding with the Greek ζεύγ-νυ-μι. The incorporating and polysynthetic languages are examples of the principle on a large scale. (3) A third mode of expressing the relations of grammar is by a change of vowel. The vowel may either pass into another or receive a different quantity or accent. Professor Pott refers to the use of vṛiddhi in Sanskrit patronymics by way of illustration as well as to change of accent in Greek proper names or vocatives. A difference of vowel which was originally purely phonetic has been adapted to distinguish between singular and plural in the English man and men, between transitive and intransitive in Greek verbs in -όω and -έω. Among the less cultivated languages of the world extended use has been made of this method of indicating the forms of grammar. In Javanese, for instance, iki is “this,” ika, “that,” iku, “that there;” in Japanese ko is “here,” ka, “there;” in Carib, ne is “thou,” ni, “he;” in Brazilian Botocudo ati is “I,” oti, “thou.”[243] In African Tumali ngi is “I,” ngo, “thou,” and ngu, “her.” Even differences of signification may be denoted by the same means; the Carib baba, “father,” is contrasted with bibi, “mother,” just as the Mantschu chacha, “man,” and ama, “father,” stand over against cheche, “woman,” and eme, “mother,” or the Finnic ukko, “old man,” and African Ibo, nna, “father,” over against akka, “old woman,” and nne, “mother.” The numerals have not escaped being distinguished in a similar manner; tizi is “one” in Lushu, and tazi, “two;” “three” and “four” are ngroka and ngraka in Koriak, niyokh and niyakh in Kolyma, gnasog and gnasag in Karaga, and tsúk and tsaak in Kamschatkan, while in Japanese fitó, mi(tsu), and yo, are “one,” “three,” and “four,” fŭtá, mu(tsu), and yá, “two,” “six,” and “eight.”[244] The Grebo of West Africa can distinguish between “I” and “thou,” “we” and “you,” solely by the intonation of the voice, mâ di being equally “I eat” and “thou eatest,” a di, “you” and “we eat,” and in Bâ-ntu Mpongwe tŏnda means “to love,” tōnda, “not to love.”[245] (4) An internal change of consonant will be the next mode adopted by language of marking a grammatical idea. Thus in Burman the active is distinguished from the passive or neuter by aspirating an unaspirated consonant, kya, for instance, being “to fall,” but khya, “to throw,” pri, “to be full,” phri, “to fill.”[246] (5) Fifthly, position may be the determining mark of relations of grammar, as is so pre-eminently the case in Chinese and the Taic languages. It makes a good deal of difference in English whether we say, “The man killed the dog,” or “The dog killed the man.” (6) Another determining mark is reduplication, which is common to all the languages of the world though used to express very different grammatical ideas. Sometimes it may denote a past tense, as in Aryan (δέδωκα, cecidi, did, &c.); sometimes a plural, as in the Bushman tu-tu, “mouths,” the Sonorian qui-qui, “houses,” or the Malay raja-raja, “princes;” sometimes a collective, as in the Canarese nîru gîru, “water and the like;” sometimes a superlative, as in the Accadian gal-gal, “very great,” the Mandingo ding-ding, “a very little child,” or the French beaucoup-beaucoup, “very much;” sometimes continuous action, as in the Dayak kaká-kaka, “to go on laughing loud,” or the Tamil muru-muru, “to murmur;” sometimes intensity, as in the Sanskrit upary-upari, “higher and higher,” the Greek παμ-φαίνω, “to shine brightly,” or the Dayak ku lyang ku lyang, “to think deeply;” sometimes emphasis and asseveration, as in the Dayak kwai kwai, “very strange!” shi shi, “yes, yes;” sometimes frequentative or repeated action, as in the Brazilian acêm, “I go out,” ace-acêm, “I go out frequently,” oce-cem, “they go out one after the other.”[247] The reduplication is often a broken one, that is, only the first syllable or part of a syllable is reduplicated, as in the Latin mo-mordi for mor-mordi. Broken reduplication is very common in the Aryan languages, but Brugman[248] has shown reason for believing that it has arisen out of an earlier complete reduplication through the action of phonetic decay. Now and then the reduplication takes place in the middle of a word, as in the Sonorian Tepeguana where some plurals are formed by repeating the second syllable, as in aliguguli, “boys,” from alguli, “boy,” or a medial syllable, as in hiim, “gourds,” and googosi, “dogs,” from the singulars him and gogosi.[249] Instead of the first syllable, only the initial vowel of a word may undergo reduplication; thus in Tepeguana ali, “child,” is a-ali in the plural, ogga, “father,” is o-ogga, ubi, “woman,” is u-ubi. On the other hand, a word may be lengthened by the repetition of the vowel at the end, as well as in the middle; the Botocudos of Brazil, for instance, turn uatu, “a stream,” into uatu-u-u-u, “ocean;” with the Aponegricans “six” is itawuna, “seven,” itawu-ú-una, while the Madagascar ratchi, “bad,” becomes ra-a-atchi, “very bad.”[250] When whole words are reduplicated a change may be made in the initial consonant of the second part of the reduplication; thus in Canarese the initial consonant becomes the guttural g, as in the example quoted above, and the French pêle-mêle and English hurdy-gurdy are familiar instances of the same fact. Sir John Lubbock[251] has made an interesting calculation of the proportion of reduplicated words found in English, French, German, and Greek on the one side, and some of the barbarous languages of Africa, America, and the Pacific on the other, the result being that whereas “in the four European languages we get about two reduplications in about 1,000 words, in the savage ones the number varies from 38 to 170, being from twenty to eighty times as many in proportion.” Reduplication, in fact, is one of the oldest contrivances of speech. It is largely employed by children in their first attempts to speak, and we need not, therefore, be surprised at finding it so persistently holding its ground both in the nursery and among barbarous tribes. The Polynesians seem to have a special affection for it, though on the other hand, Mr. Matthews tells us that in North America while reduplication is a prominent feature of the Dakota verb it occurs in only one verb in the closely allied Hidacha dialect.[252] Reduplication, however, is one of the most important modes adopted by language for denoting the relations of grammar; it is, in fact, one of the most obvious and natural of its outward means of expressing those inward forms and grammatical conceptions which the human intelligence has painfully struggled to realize.[253]

The common division of speech into formal and material is at once defective and misleading. The articulate sounds of which words are composed may indeed be called their matter, but they do not become words, do not constitute a part of speech until they have thought and significancy breathed into them like the breath of life into man. This significancy is a relative one, that is to say, the meaning of a word depends upon its relation to some other. But this relation may be of two kinds, it may exist either between the ideas denoted by the words or between the words when coupled together in some particular sentence. In the first case we have to do with sematology, in the second with grammar. We can understand what is meant by the word tree only by comparing and contrasting the idea of tree with other cognate ideas; but the relation between tree and sheds in such a sentence as “the tree sheds its leaves,” is of a totally different nature. The idea of tree remains the same whatever be the outward symbol by which it is expressed, whether tree, or arbor, or baum, or anything else; the relation between tree and sheds is one that can be discovered only by a historical and comparative investigation of English grammar. It is to this grammatical relation alone that the term formal is strictly applicable; it has to do with the forms, or, as in the instance before us, the want of forms, whereby the relations of grammar, the relations, that is, of words in a sentence, are denoted. Going back to the primitive sentence-word, we shall have to distinguish between the material sounds of which it was composed, the meaning it always possessed whenever and however used, and the form (or position) that it assumed according to the occasion on which it was used. The child who says “Up!” always attaches the same signification to the general idea contained in the word, but whether it is to be regarded as an imperative, a hortative, an optative, or any other particular grammatical form is left to the context, the tone and gesture, or the intelligence of the bearer. Language consists of the material, the significant, and the formal, and it is only the latter, that part of language, in fact, the origin of which we have elsewhere traced to gesture, that properly concerns morphology.

Whatever, therefore, belongs to grammar belongs also to morphology. Not only general form and structure, but also grammar in the narrower sense of the word, as well as composition, and what our German neighbours term “word-building” must be included under it. Composition, indeed, is but a species of declension and conjugation. Parricida and patris (oc)cisor, φερέοικος and οἶκον φέρει, have exactly the same force and meaning. The only difference between good-for-nothing as a compound and “he is good for nothing” in a complete sentence, is that the first can be used as an attribute. The ordinary genitive of the Semitic tongues, the so-called “construct state,” is really an instance of composition, the first noun—that which “governs” the second—being pronounced in a single breath with the other, and accordingly losing the case-terminations. This did not happen originally, as may be seen from the occasional occurrence of these terminations even in Assyrian, which is more strict in following out the rule than any other of the cognate idioms. The power of composition is greater in some languages than in others. The polysynthetic sentences of an American dialect present the appearance of gigantic compounds, with this difference, however, that in a true compound the language has put together two words that have already been used independently, or at all events are capable of being used independently, whereas in the less advanced American languages the several members of the sentence have never attained the rank of independent words which can be set apart and employed by themselves. Even in some of the compounds of the Aryan family, where the flectionless “stem” shows itself, it may be questioned whether we have not before us the relics of that earliest stage of speech when the flections had not yet been evolved, and when the relations of grammar were expressed by the close amalgamation of flectionless stems in a single sentence-word. However that may be, the power of forming compounds possessed by the Aryan group of languages stands in marked contrast to the repugnance felt by the Semitic tongues in this respect. Composition is as rare in Semitic as it is common in Aryan, and this contrast between the two families of speech is one of the many that demonstrate the radical difference existing between them. Perhaps the extended use made by the Semitic languages of denoting the relations of grammar by internal vowel-change had much to do with their objection to the employment of compounds. They are less agglutinative in character than the Aryan dialects, truer, in fact, to the principle of flection, and the same instinct that makes them represent the ideas of “killing” and “a killing” by kodhêl and kidhl (kedhel), rather than by trucida-n-s and trucida-ti-o(n), makes them also use two unallied roots like hâlach and yâtsâ where the Aryan would have said ire and exire. Even within the Aryan family itself we find the Greek with compounds like the comic λεπαδο-τεμαχο-σελαχο-γαλεο-κρανιο-λειψανο-δριμ-υπο-τριμματο-σιλφιο-παραο-μελιτο-κατα-κεχυμενο-κιχλ-επι-κοσσυφο-φαττο-περι-στερ-αλεκτρυον-οπτ-εγ-κεφαλο-κιγκλο-πελειο-λαγῳο-σιραιο-βαλη-τραγανο-πτερύγων,[254] and the Latin comparatively poor in them, while modern English, in spite of the loss of its flections, lags but little behind German. Russian can form such specimens of agglutination as bezbozhnichestvovat, “to be in the condition of being a godless person,” from bez Boga, “without God,” and classical Sanskrit almost dispenses with syntax by its superabundant use of composition. Where syntax is highly developed, as it was in Latin, the growth of composition is checked and limited.

Composition has been a fruitful source of grammatical flection, and a still more fruitful source of what is meant by “word-building.” It is highly probable that the person-endings of the Aryan verb as-mi, a(s)-si, as-ti, or ἐσ-μι, ἐσ-σι, ἐσ-τι, are but the personal pronouns closely compounded with the verbal stem. Such, certainly, has been the case with the so-called tempus durans of Aramaic, where kâdhêlnâ, “I am killing,” is resolvable into kâdhêl + ’anâ, “killing + I,” and kâdhlath, “thou art killing,” into kâdhêl + at’, “killing + thou.”[255] The Latin imperfect and future in -bam and -bo seem to be compounds of the verbal stem with the verb fuo, “to exist,”[256] like the perfect in -ui or -vi (fui), while the pluperfect scripseram is a combination of eram or esam and the perfect scripsi (itself formed from the verbal stem scrib- and the old perfect esi of the substantive verb “sum”). So, too, the form amavissem is just as much a compound of amavi (ama + fui) and essem (es + siem) as is amatus sum of the passive participle and the substantive verb. If we turn to our own language we can trace our perfects in -ed back to the Gothic amalgamation of the verb with dide, the reduplicated perfect of the verb do, while the origin of the French aimerai in the infinitive aimer (amare) and the auxiliary ai (habeo) is as plain as that of the Italian dármelo (“to give it to me”) or fáteglielo (“do it for him”). The real character of the compound has come to be forgotten in course of time, and its final part has gradually lost all semblance of independence and been assimilated to the terminations which simply denote grammatical relations. The general analogy of the language has been too strong for it, and the agglutinated word has become a flection.

But there are many suffixes which are not flections—that is to say, which do not denote the relations of grammar, or rather the relations that exist between the different parts of the sentence. In I loved for I love-did the grammatical relation which we name a perfect tense, is not really expressed by the suffixed word did, but by the reduplication which that word has undergone. It was the reduplication that gave did (dide) the force of a perfect, and the attachment of did to another verb merely handed on to the latter the perfect force which it already possessed. Strictly speaking the suffix -ed is a flection only because it is the relic of a reduplication, the flection—that is to say, the expression of a grammatical relation—lying in the reduplication or form of the word. So, too, when we find dêv-mã, meaning “in God,” in Gujerati, or andhê-mẽ, meaning “in the blind,” in Hindustani, we must not suppose that the locative sense actually lies in the suffixes mã and mẽ. These suffixes go back to the Sanskrit madhyê, “in the middle,” where the flection is to be sought in the termination i (contained in ê = a + i) not in the stem madhya, “middle.”

When, then, we say that composition may be a fruitful source of flection, what we mean is this. Flection is the means adopted by a certain class of languages for expressing the relations that exist between the members of a sentence, but a perception of these relations must first grow up in the mind before external means are found for embodying them. The idea of past time must be arrived at and realized before the simple process of reduplication can be adopted to denote it. Not only in other languages but also in the Aryan family of speech reduplication serves to represent other relations of grammar than that of past time. When the Frenchman says beaucoup beaucoup—meaning “very much”—he is employing reduplication to express the superlative relation just as much as the old Accadian with his galgal, “very great,” while the very fact that there are Greek presents like δίδωμι and τίθημι, ought to show that there was once a time in the history of Aryan speech when reduplication served other purposes than that of denoting past time. So it is with all the rest of the grammatical machinery which we call flection. First of all the growing intelligence came to have, as it were, an intuition of certain relations between the parts of a sentence, and then sounds and forms already existing were adapted to denote these. And the very same form might at successive periods in the development of a language be adapted to denote different relations, as we have just seen was the case with reduplication. When suffixes were used for a similar purpose, they too had to follow the general analogy. Many of these suffixes seem coeval with the beginnings of Aryan speech, at least so far as we know anything about it, but others of them, like the person-endings of the verb, are really instances of composition, the final part of the compound having become a mere suffix, and so, like many other suffixes, been adapted to the use of flection.

This brings us to those suffixes which have never been applied to a purely flectional purpose. If we turn over the pages of an English dictionary we shall come across the two familiar words knowledge and wedlock, which at first sight seem to have nothing in common. On tracing them back to earlier forms, however, we find that knowledge, Old English know-leche, like wed-lock, Old English wed-lâc, are both compounded with the Anglo-Saxon lâc, “sport” or “gift,” the Old High German leih, the Old Norse leikr, and the Gothic láiks. The word still survives in the north of England under the form of laik, “to play,” and the provincial lake-fellow is merely “play-fellow.”[257] Several abstracts were formed in Anglo-Saxon by the help of it; thus we have feoht-lâc, “fight,” gudh-lâc, “battle,” bryd-lâc, “marriage,” reaf-lâc, “robbery.”

Now what has happened in the case of the English lâc has happened in the case of a good number of other words in all the languages spoken throughout the world. Words originally independent and distinct become so glued together in composition that one of them loses its personal identity, as it were, and comes to be the mere shadow of the other, whose meaning it qualifies and classifies. Thus, for instance, the Greek κατὰ, when compounded with the verb ἄγω, “to lead,” limits the sense of the latter to “leading down,” and our own hood or head, the Anglo-Saxon hâd, “a state,” in words like Godhead or maidenhood, refers the nouns to which it is attached to a new and particular class.