Since A. Sauvageot has requested his critics to give their explanation of «coincidence,» and since I am responsible for the dismissal of some of these coincidences, I will touch once more on this side of the problem. As a matter of fact, I am not competent to give an answer in reference to all languages discussed; but as to the Tungus language, the coincidences are not numerous enough to occupy the attention of A. Sauvageot. Practically two «coincidences» are not a serious evidence for showing the common origin of the Tungus language with the Ural-Altaic languages. However, this question can be discussed even without establishing Tungus parallels.
There is a great number of words in Tungus met with also more than in two different languages, such as Mongol, Turk, Ugrian, Indo-European, and Chinese, without speaking of the Paleasiatic languages. The number of common words is so great, that it will constitute an ungrateful task for the one who will attempt to establish the original linguistical complexes for different Asiatic languages, and particularly the Tungus language. But what is interesting for us is the linguistic meaning of the common words and the relation between the common words and the common origin of languages. In this way, A. Sauvageot's warning to the critics in charging them with the duty of explaining to him the fact of the existence of common words otherwise than through the common origin of languages will lose its seeming roughness and gravity for the future of comparative linguistics.
In one of the previous sections, I have already pointed out that there are different sources of origin of common words, so that one has to find out with reference to each word discussed by which way the word has happened to be similar in two or more languages. The number of common words due to statistical convergence is very great in case one deals with monosyllabic stems. It is much less in the case of disyllabic stems. A coincidence of similar words in two languages is greater than in three; in three languages, greater than in four; and so on. However, it is possible to calculate, as I have shown, the percentage of possible coincidences in two, three, and more languages if the stems are grouped according to the types of consonants; e.g., labial, dental, glottal, and others. The point of difficulty in this case will be the establishment of the fact whether it is a statistical phenomenon or not. In some cases it is absolutely impossible to establish it, indeed; for the source of borrowing may be lost altogether. The only way is mathematically, to establish the probable frequency of occurrences and to compare the number of cases showing similarity and at the same time not bearing any traces or documents of their alien origin. In case the number of such words greatly exceeds the number expected by calculation of probablity, then two things may be suggested: (1) the words were received from a source which is now unknown; and (2) the words were preserved from the time of the existence of a common language from which other languages originated by differentiation. The last suggestion is one which can be done only after a minute analysis of word origin. If one postulates the existence of such a pra-language and explains the common words through their common origin, — the transmission from a pra-language, — one may easily be misled by the statistical convergence. Although A. Sauvageot or his colleagues have never calculated these probabilities, for they do not want to face the problem and reject this possibility, owing to their conviction that the languages have originated from common ancestors; yet the issue of investigation is already known, for the answer already resides in the question.
The difficulty of understanding the origin of the similarity of languages otherwise than through the direct origin from the parent language is not that residing in the nature of language, but in the fact of accepting the original axioms of origin from the ancestor language, just as the difficulty of non-Euclidean geometry for a person unfamiliar with some other systems resides not in the difficulty of other systems, but in the fact that the original Euclidean axioms are not postulated.
Other sources of common words, as cultural phenomena, local phenomena, and imitation, are productive of a great number of common words. Unfortunately, only a minor part can be traced back to their sources, for the spreading of words might have taken place in such a remote time that there may be no question as to the finding of sources. There has been a great inclination amongst some authors for finding common words originating from certain centers, so that even a slight phonetic similarity was considered to be sufficient evidence for establishing the source of origin. In this case a great source of the mistakes made resides in the method too. First of all, it is presumed that such and such a class of phenomena are «cultural phenomena» and as such they ought to come from a certain definite source. Let us take instances. Of course, it is evident that the Greek word χύβερναω [originally «to stir,» whence χύβερνητηρ], as P. P. Schmidt shows, is responsible for the existence of various European terms, and it may be established when this word made its appearance amongst different groups, and in what particular «meaning.» It has even reached the Yakuts in the form of krimidatal (and kubarnatar) adapted from Russian gubernator to the Yakut phonetic complex, in which form it could not be connected with the Greek original word without the competence of E. K. Pekarskii, who knew how it could happen to change in the Yakut soil. But a word such as «war» may be accepted as cultural phenomenon connected with Chinese only on the acceptation of several hypotheses, the principal of which is that «enemy,» «war,» etc., are cultural phenomena of Chinese origin, supported by linguistical hypotheses [154]. Indeed, it is not improbable that this word is of Chinese origin, but there is a methodological danger, namely, a reasoning by analogy and reference to the cultural character of «war,» which is a universal ethnographical phenomenon, the term for which may also be borrowed. The question is thus confined to the phonetic evidences only. We have seen that xuse and vase are now used by the Manchus in forms very close to their Chinese equivalents, besides other Manchu words, so the supposition of borrowing is confirmed by the facts, while in the case of dain it is supported by the hypotheses. It is true that the terms for «war» in Tungus are borrowed; e.g., bulov (Ner. Barg.), bulen (Bir., war amongst the shamans), bilan (Ur., Castr.), which can be compared with bulija (Mong., Rud.)—«to rob,» «to take by force,» etc., also carik (Mank.) chirik (Khin.) which are connected with cherig (Mpng., Rud.) — «the troops,» «soldier,» «military»; also shiri (Turn.) connected with sari (Yakut, Pek.) — «the war.» From the stem discussed above there has originated only dajan (Bir.) in the sense of war (cf. the previously particularized meaning of an older term bulen), recently borrowed. However, all Tungus dialects know the verb va — «to kill,» and valdi — «to kill one another,» «to fight a war,» valdinki — «the battlefield,» etc. The danger resides in the fact that this method may become of common practice and simple statistical convergences may be mistaken for borrowings. In my «Social Organization of the Northern Tungus,» I have made an attempt to show how the borrowing of such an elementary term as «son» might have happened and what was the reason of such a borrowing. It has required, not only phonetic parallels, but also a thorough analysis of social institutions and their possible variations. These instances show us that the grouping of phenomena into «cultural» and «non-cultural» phenomena may become a source of mistakes. It is easier to operate with the complexes. For instance, in the complex of local institutions covering the marriage and wedding, we have seen that most of the terms are Mongol, Manchu, and Yakut. This may be understood from the fact of the change of these institutions at a certain historic moment. On the other hand, there are cases which are more difficult than this one. Here I have in view, for instance, the complex of the reindeer breeding unknown amongst the present Mongols and Manchus. The term for reindeer in Tungus is oro~oron, but in Chukchi it is horana, and in Finn poro (cf. P. P. Schmidt, «Etymologische Beitrage»; although P. P. Schmidt supposes it to be Western Paleasiatic poro it may also be Tungus oro, aspirated and bilabialized), the origin of which is unknown. But the Tungus terms connected with breeding are very often borrowed from the Mongols; e.g., kure, etc., for the reindeer fences. So the complex of reindeer breeding, although one of the most characteristic elements of the Northern Tungus complex, is built up of phenomena borrowed with the terms from various sources. It is simple with the complex of horse and cattle breeding which is recently borrowed by the Tungus from Mongol- (Buriat-) speaking people together with the terms, so no doubt may be raised up as to the origin of words. It is thus evident that in every particular case of borrowing or hypotheses as to borrowing, the complexes and the possibility of their total or partial borrowing must be shown. Yet if the borrowing of the most «elementary» complexes is a fact which has been established, we cannot reject such a possibility for any phenomenon whether it appears to the European mind to be a «primitive» or a «cultural» one.
On different occasions we have already seen that there are some local phenomena, e.g., buran, purga, which are characteristic of the regions with snowstorms and these stems are met with in various languages with nearly the same sense. It is quite natural that every new comer meeting with a new phenomenon, when possible, borrows the name for it from the antecedent experience of the local people, so that if we do not postulate that the people were always living in the same region (the history of human migration does not allow us for a moment to soothe our mind with this idea), the borrowing of these terms is very likely. However, it is difficult in some cases to show documentary evidences for it. The names of local animals, plants, meteorological and other local phenomena ought to be included in this class.
It is evident that the class of words of onomatopoetic origin is also a source of common words, but they are not very numerous. It may be noted that the onomatopoetic words may become subject to borrowing as well. This is perhaps the case of the word «cuckoo,» which might have been invented through the imitation of the bird's cry only once or twice, and afterwards it might receive a greater geographical distribution. Some languages, as pointed out (for example, the Manchu), show a quite definite tendency to form words from these sources, while others may borrow these stems as stems and not as onomatopoetic complexes. In this way a local reference to an onomatopoetic origin of some words is neither convincing nor historically correct. The Tungus words for some birds, as, for example, gaki (Bir.) (Neg. Sch.) (Gold., Orochi Sch.), gaxa (Manchu Writ.) — «the crow,» is explained by the Birarchen as an imitation of gak! gak! supposed to be produced by this bird; saksaxa (Manchu Writ.) sazhiga, sazhia (Bir.) — «the magpie,» is explained by the Birarchen as an imitation(?) of sak! sak! etc. However, there are also other words in other dialects for the same birds.
It is evident that to use common words as an evidence of the common origin of languages requires the most careful analysis, which is well known from the previous experience of comparatists. In spite of it all, possible considerations are rejected and the only one is adopted: the common origin of words due to the common origin of language. There is thus something in the idea of language which implies this attitude. This is the idea that the language is an organical entity existing in time and transmitted as such.
A. Sauvageot charges his critics who would not accept his idea that common words quoted by him are of Ural-Altaic origin to show a great number of coincidences between Uralian and Keshua or Algonquian, and he says that they «veuillent bien se donner la peine d'en determiner un aussi grand nombre, et de la meme qualite,» as that presented by himself. Nobody can accept this challenge, for if the comparative material is cf the same quality as his Tungus material, the work will not be convincing — it must be better done. Second, before we proceed to such an experiment, the probability of meeting common words must be calculated, which has not been naturally done by A. Sauvageot, who does not know how great the chance of meeting common words is. Third, the experiment must be carried out with two languages, the phonetic systems of which do not differ very much; so, for instance, the presence of some rare consonants may greatly reduce the chance of finding coincidences. I do not know Keshua and Algonquian and I have no dictionaries at hand, but if the phonetic difference is not very great and if the experimentator allows himself to have the same liberty of comparing phonetic and semantic groups, it may be supposed that the number of common words will be also great. In the course of our analysis of cases taken from A. Sauvageot's work we have seen many instances of coincidence with various Indo-European languages. Such a work, at least in reference to the Finno-Ugrian languages, has already been undertaken by several authors and the attempts at comparing Chinese and Indo-European is still fresh, and we have some other recent instances of this kind.
154. The Chinese etymology of the Manchu word dain was originally proposed by L. Zaxarov. P. P. Schmidt has maintained the Chinese etymology by the help of the hypothetical restoration of old Chinese (cf. the discussion between P. Pelliot, op. cit.. p. 256, and P. P. Schmidt, «The Language of the Samagirs,» p. 4) and supported with reference to an analogy with other «cultural phenomena.»