There is yet another reason for thinking that the majority of Indo-European roots-that is of the types which underlay the oldest Aryan vocabulary of which we know -must be regarded as polysyllabic. Prof. Max Müller' draws attention to the fact that the existence of parallel roots of similar meaning, but different terminations, like mardh, marg, mark, marp, mard, smar, and mar, can be better explained by elimination than by composition. The so-called determinatives or final letters cannot be classificatory, as they convey no modification of meaning, and are to be found in words belonging to all the parts of speech. There is at all events no à priori argument against treating the simplest roots as the latest, rather than the earliest products of language." 'It would be perfectly intelligible that such roots as mark, marg, mard, mardh, expressing different kinds of crushing, became fixed side by side, that by a process of elimination their distinguishing features were gradually removed, and the root mar left as the simplest form expressive of the most general meaning." In other words, the vocables

"

not belonging to the root. In the second part of the third volume of the same periodical (1879), Fick points out that, “through the influence of the accent the vowels e and o can be reduced to ĕ and o, mere voice-checks, for which loss of the vowel can be substituted, especially in open syllables. This original ě and ŏ, which I call sh'wa for the sake of brevity, generally appears in Sanskrit as i or i (also as u or u before and after liquids), in Zend as e and i, in Greek mostly as a, in German as o (Gothic u)." "In the aorist the final sh'wa is retained in the a of the Greek aorist Exevas," &c. "Hence it is clear that stems like ed, bhid, or ruk had no existence whatever originally, but first arose out of edě, bhidě, and rukě; the Indian theory of roots, with all its perverse consequences (thematic and union vowels, Guna, &c.), must be definitely set aside."

that embodied these roots underwent the wear and tear of phonetic decay, many of them passed out of the living speech and were replaced by others, and there was left at last a whole family of nouns and verbs, whose sole common possession was the syllable mar. That alone had resisted the attacks of time and change. We indeed have some difficulty in realizing the variability of savage and barbarous languages, or of the readiness with which new words are coined and old ones forgotten. Mr. Theal, illustrating the Kafir rule that a woman may not mention the names of any of her husband's male relations in the ascending line,1 states that "a woman who sang the song of Tangalimlibo for me used the word angoca instead of amanzi for water, because this last contained the syllable nzi, which she would not on any account pronounce. She had therefore manufactured another word, the meaning of which had to be judged of from the context, as standing alone it is meaningless." It will be noticed that the word is trisyllabic, and not a monosyllable, as the Indianist theory would require, and if other words came to be framed after its model, it would originate a root, which would certainly be of more than one syllable. Phonetic decay alone could reduce it to the orthodox monosyllabic form.

The existence of compound roots has already been alluded to, implying a division of roots into simple and compound, the first class consisting of those which were really simple from the first, as well as of those which our ignorance prevents us from decomposing. Compound roots form part of the class of "secondary" roots as disCape Monthly Magazine," xiv. 36 (June, 1877), p. 349.

1 66

tinguished from "primary ;” yu and mar being examples of primary roots, yudh and yug, mardh, mard, marg, mark, and smar of secondary ones. A primary root, therefore, is the simplest element of sound and meaning which can be extracted from a group of words; it constitutes their characteristic mark and sign of relationship, and indicates where the line of division must be drawn between them and other unallied words. A secondary root determines a species within the larger genus; words containing the root mardh, for instance, form a specific class within the wider class of those which contain the root mar. The Latin ju-s, “right” or "bond," is an example of the genus of which jung-ere, “to join," is an example of the species; but whereas in natural history a species is posterior to the genus, the converse is the case with the roots of the philologist. The reason of this is plain enough; the genera and species of zoology and botany answer to actually existing forms of life, whereas the roots of language are due to the reflective analysis of the grammarian. At the same time, some of the secondary roots are undoubtedly compounds, that is to say, are extracted from compound words, and wherever this is the case, the species or secondary root will necessarily be later than the primary or generic one.

One of the first attempts to decompose the secondary roots was made by Professor Pott. He started the view that a large number of them were compounded with prepositions; thus pinj, “painting,” is derived from api or inì and anj, "anointing." But such a view is no longer tenable. The loss of the initial vowel in a word like api is peculiar to Sanskrit, and not a characteristic of the

parent-Aryan; the origin of the Latin ping-ere would therefore be inexplicable. Moreover, the preposition was a late growth in Aryan speech, and in early times there was no close amalgamation of it with the verb. Even in Greek and Sanskrit the prepositions are still so independent that the augment and reduplication are inserted between them and the verbal form, and we all remember how loosely attached they are to the verb throughout the larger part of Homer. Pott's theory must therefore be given up, and another be proposed in its place. This has been done by Professor G. Curtius, who suggests that many of the compound roots were similar to such Latin tenses of a later day as amav-eram (for amavi-eram) and amav-ero (for amavi-ero), where we have two verbal forms agglutinated one to another. Hence in a secondary root like yudh, we may see an amalgamation of the two primary roots yu and dha, the first with the sense of "mingling," and the second with that of "placing." It is very possible that the Greek passive aorist -TÚQ-ON-V and optative Tuao-in-v may contain the roots dha, "placing," and ya, "going," which we find in the Latin ven-eo, venum-ire; at all events, the existence of such compounds in the parent-Aryan is shown to be more than a mere conjecture by the Latin credo which appears under the form of 'srad-dadhâmi in Sanskrit. Sanskrit and Latin alike throw light on one another, and show us that credere, "to believe," is really a compound of cor(d), “the heart," the Greek nagdia, and the root dha, "to place," which elsewhere appears in the Latin ab-dere, con-dere, e-dere. "To believe" was therefore originally "to

1

1 Darmesteter: "Mémoires de la Société de Linguistique," iii. p. 52.

place" or "set the heart" towards another object. How old the compound is may be gathered from the form it has assumed in Sanskrit. The ordinary word for "heart" in both Sanskrit and Zend presupposes a root ghard; 'srad alone in this curious old compound has the same root, kard, as the words which signify "heart" in the European branch of the Aryan family. The parentAryan had its dialects like all spoken languages, and these dialects possessed slightly differing forms of the same word. One form finally triumphed in Western Aryan, another form in Eastern Aryan, but before this happened the compound crêdo, 'srad-dadhâmi, was already in existence, testifying to a time when the WestAryan form was employed in East Aryan itself.

A very common secondary root is one formed by reduplication. Originally the whole root was probably repeated; but in course of time broken reduplication became prevalent, consisting in the repetition of only a part of the whole root. Thus by the side of uáquap-os and furfur we find me-mor and in(e)T-w, tu-tud-i and Tέ-TUπ-a. The loss of the second consonant might be compensated for; in the Greek haina and dadanos, for instance, a diphthong marks the existence of a former consonant. On the other hand, the vowel of the second syllable might be lengthened or intensified, as in the Greek ȧywyn and έT-Tuμos, and when the second syllable was thus strengthened the vowel of the first syllable was very liable to become correspondingly weak. So in Latin we have ci-conia and ci-catrix, and in Greek di-daona for di-danσκω, and ἵστημι for σιστημίο When the variation of vowel had once been introduced, the changes that could be