is particularly the case when the numerous modal forms become further complicated by incorporating the direct pronomial object, as in the Magyar varjak = they await him, and the Mordvinian palasa = I embrace him. Thus arise endless verbal combinations, reckoned in Turki at nearly 30,000, and past counting in the Urian group.
Another marked peculiarity of the Ural-Altaic, at least as compared with the inflecting orders of speech, is weak subjectivity, the subject or agent being slightly, the object of the action strongly accentuated, so that "it was done by him" becomes "it was done with him, through him, or in his place" (apud eum). From this feature, which seems to be characteristic of all the branches, there follow some important consequences, such as a great preponderance of locative forms in the declension,—the nominative, and often even the possessive, being expressed by no special suffix. Hence also the object normally precedes the subject, while the idea of possession (to have) is also everywhere replaced by that of being (to be), so that, even in the highly developed Osmanli, "I have no money" becomes "money-to-me not-is" (Akchehím yok-dür"). In fact the verb is not clearly differentiated from the noun, so that the conjugation is mainly participial, being affected by agglutinating pronomial, modal, temporal, negative, passive, causative, reciprocal, reflexive, and other suffixes to nominal roots or gerunds: I write = writing-to-me-is. Owing to this confusion of noun and verb, the same suffixes are readily attached indifferently to both, as in the Osmanli ján = soul, ján-ler = souls, yázár = he will write, yázár-ler = they will write. So also, by assimilation, the Yakut word kötördör kötöllör = the birds fly (from root köt, where kötöl stands for kötör, and dör for lör, the Osmali ler, or a suffix of plurality.
But notwithstanding this wealth of nominal or verbal forms, there is a great dearth of general relational elements, such as the relative pronoun, grammatical gender, degrees of comparison, conjunctions, and even postpositions. Byrne's remark, made in reference to Tungus, that "there is a great scarcity of elements of relation, very few conjunctions, and no true postpositions, except those which are given in the declension of the noun,"1 is mainly true of the whole family, in which nouns constantly do duty for formative suffixes. Thus nearly all the Ostiak post-positions are nouns which take the possessive suffix and govern other nouns in the genitive, precisely as in the Hindi: ādmī-kī-țărăf, being a feminine noun = direction, requires the preceding possesive particle to also be feminine (kī for kē).
As there are thus only two classes of words,—the roots, which always remain roots, and the suffixes, which always remain suffixes,—it follows that there can be no true composition or word-building, but only derivation. Even the numerous Magyar nominal and adjectival compounds are not true compounds, but merely two words in juxtaposition, unconnected by vowel harmony and liable to be separated in construction by intervening particles. Thus in aran-sinü = gold-colour = golden, the first part aran receives the particle of comparison, the second remaining unchanged, as if we were to say "gold-er-colour" for "more golden"; and ata-fi - relative becomes ata-m-fi-a = my relative, with the intrusion of the pronomial m = my.
But while these salient features are more common, or nearly common, to all, it is not supposed to be that the various groups otherwise present any close uniformity of structure or vocabulary. Excluding the doubtful members, the relationship between several branches is far less intimate than between the various divisions of the Semitic and even of the Aryan family, so that, great as is, for instance, the gap between English and Sanskrit, that between Lapp and Manchu is still greater.
After the labours of Castrén, Csink, Gabelentz, Schmidt, Böhtlingk, Zenker, Amqvist, Radloff, Munkacsi-Berat, and especially Winkler, their genetic affinity can no longer be seriously doubted. But the order of their genetic descent from a presumed common organic Ural-Altaic language is a question presenting far greater difficulties than the analogous Aryan problem. The reason is, not only because these groups are spread over a far wider range, but because the dispersion from a common centre took place at a time when the organic speech was still in a very low state of development. Hence the various groups, starting with little more than a common first germ, sufficient, however, to give a uniform direction to their subsequent evolution, have largely diverged from each other during their independent development during the remotest prehistoric times. Hence also, while the Aryan as now known to us represents a descending line of evolution from the synthetic to the analytic state, the Ural-Altaic represents on the contrary an upward growth, ranging from the crudest syntactical arrangements in Manchu to a highly agglutinating but not true inflecting state in Finnish.2 No doubt Manchu also, like its congeners, had formerly possessive affixes and personal elements, lost probably through Chinese influences; but it can never have possessed the surprisingly rich and even superabundant relational forms so characteristic of Magyar, Finn, Osmanli, and other western branches. As regards the mutual relations of all groups little more can now be said than that they fall naturally into two main divisions—Mongolo-Turkic and Finno-Ugro-Samoyedo-Tunguisic—according to the several methods of employing the auxiliary elements. Certainly Turkic lies much closer to Mongolic than it does t Samoyedic and Tungusic, while Finno-Ugric seems to occupy an intermediate position between Turkic and Samoyedic, agreeing chiefly in its roots with the former, in its suffixes with the latter. Finno-Ugric must have separated much earlier, Mongolic much later, from the common connexion, and the latter, which has still more than half its roots and numerous forms in common with Turkic, appears on the whole to be the most typical member of the family. Hence many Turkic forms and words can be explained only by reference to Mongolic, which has at the same time numerous relations to Finno-Ugric and Samoyedic, that have been lost in Turkic and Tungusic. It may therefore be concluded that the Finno-Ugric migrations to the north and west and the Tungusic to the east had been completed while the Turkic and Mongolic tribes were still dwelling side by side on the Altai steppes, the probable cradle of the Ural-Altaic peoples.
How profoundly the several groups differ one from the other even in their structure is evident from the fact that assumed universal features as unchangeable roots and vowel harmony are subject to numerous expressions, often spread over wide areas. NOt only is assimilation of final consonants very common, as in the Osmali bulun-mak for the Uigur bulul-mak, but the root of the vowel itself is frequently subject to umlaut through the influence of suffixed vowels, as in the Aryan family. THus in the Surgut dialect of Ostiak the long vowels of nominal stems become modified before the possessive suffix, ā to ē to ī to ō to ū (Castrén). It is still more remarkable to find that the eastern (Yenisei) Ostiak has even developed verbal forms analogous to the Teutonic strong conjugation, the presents tabāq', abbatag'an, and datpaq' respectively; so also taig, tōrg, and tārg, present, past, and imperative, are highly suggestive of Teutonic inflexion, but more probably due to Tibetan influences. In the same dialects many nouns form their plurals either by modifying the root vowel, in combination with a suffixed element, or by modification alone, the suffix having disappeared, as in the English foot—feet, goose—geese. So also vowel harmony, highly developed in Finnish, Magyar, and Osmanli, and all of which two distinct forms occur in Yakutic, scarcely exists at all in Tehere-missian, Votyak, and the Revel dialect of Esthonian, while in Mordvinian Siryenian, not the whole word, but the final vowels alone are harmonized. The unassimilated Uiguric kilur-im answers to the Osmanli kilur-um, while in Manchu the concordance is neglected, especially when two consonants intervene between the root and the suffixed vowels. But too much weight should not be attached to the phenomenon of vowel harmony, which is of comparatively recent origin, as shown in the oldest Magyar texts of the 12th century which abound in discordances such s halál-nek,
1 Gen. Prin. of Struct. of Lang., i. p. 391, London, 1885.
2 "Meine Ansichten werden sich im Fortgange ergeben, so nametlich dass ict nicht entfernt die finnischen Sprachen für flexivische halten kann" (H. Winkler, Uralataische Völker, i, p. 54). Yet even true inflexion can scarcely be denied at least to some of the so-called Yenisei Ostiak dialects, such as Kotta and others still surviving about the middle Yenisei and on its affluents, the Agul and Kan (Castrén, Yen., Ostjak, und Kott. Sprachlehre, 1858, Preface, pp. v-viii.) These, however nay be regarded as aberrant members of the family, and on the whole, it is true that the Ural-Altaic system nowhere quite reaches the stage of true inflexion.