It will be seen that in classes 1 and 3 the speech of the natives prevails, while in the two classes comprised under 2 it is that of the conqueror which eventually triumphs. Further, that, in all cases except type 2b, that language prevails which is spoken by what is at the time the majority.
Sound substitution is found in class 3 in the case of foreigners who come to America after they have learnt to speak, and of the children of foreigners who keep up their original language at home. If, however, while they are still young, they are chiefly thrown with English-speaking people, they usually gain a thorough mastery of the English language; thus most of the children, and practically all of the grandchildren, of immigrants, by the time they are grown-up, speak English without foreign taint. Their origin has thus no permanent influence on their adopted language. The same thing is true when a small ruling minority drops its foreign speech and learns that of the majority (class 1), and practically also (class 2a) when a native minority succumbs to a foreign majority, though here the ultimate language may be slightly influenced by the native dialect.
It is different with class 2b: when a whole population comes in the course of centuries to surrender its natural speech for that of a ruling minority, sound substitution plays an important part, and to a great extent determines the character and future of the language. Hempl here agrees with Hirt in seeing in this fact the explanation of much (N.B. not all!) of the difference between the Romanic languages and of the difference between natural High German and High German spoken in Low German territory, and he is therefore not surprised when he is told by Nissen that the dialects of modern Italy correspond geographically pretty closely to the non-Latin languages once spoken in the Peninsula. But he severely criticizes Hirt for going so far as to explain the differentiation of Aryan speech by the theory of sound substitution. Hirt assumes conditions like those in class 1, and yet thinks that the results would be like those of class 2a. “It is essential to Hirt’s theory that the conquering bodies of Indo-Europeans should be small compared with the number of the people they conquered.... If we wish to prove that the differentiation of Indo-European speech was like the differentiation of Romance speech, we must be able to show that the conditions under which the differentiations took place were alike or equivalent. But even a cursory examination of the manner in which the Romance countries were Romanized ... will make it clear that no parallel could possibly be drawn between the conditions under which the Romance languages arose and those that we can suppose to have existed while the Indo-European languages took shape.” Hempl also criticizes the way in which the Germanic consonant-shift is supposed by Hirt to be due to sound-substitution: when instead of the original
t th d dh
Germanic has
þ þ t ð,
these latter sounds, on Hirt’s theory, must be either the native sounds that the conquered people substituted for the original sounds, or else they have developed out of such sounds as the natives substituted. If the first be true, we ask ourselves why the conquered people did not use their t for the Indo-European t, instead of substituting it for d, and then substituting þ for the Indo-European t. If the second supposition be true, the native population introduced into the language sounds very similar to the original t, th, d, dh, and all the change from that slightly variant form to the one that we find in Germanic was of subsequent development—and must be explained by the usual methods after all.
I have dwelt so long on Hempl’s paper because, in spite of its (to my mind) fundamental importance, it has been generally overlooked by supporters of the substratum theory. To construct a true theory, it will be necessary to examine the largest possible number of facts with regard to race-mixture capable of being tested by scientific methods. In this connexion the observations of Lenz in South America and of Pușcariu in Rumania are especially valuable. The former found that the Spanish spoken in Chile was greatly influenced in its sounds by the speech of the native Araucanians (see Zeitschr. f. roman. Philologie, 17. 188 ff., 1893). Now, what were the facts in regard to the population speaking this language? The immigrants were chiefly men, who in many cases necessarily married native women and left the care of their children to a great extent in the hands of Indian servants. As the natives were more warlike than in many other parts of South America, there was for a very long time a continuous influx of Spanish soldiers, many of whom, after a short time, settled down peacefully in the country. More Spanish soldiers, indeed, arrived in Chile in the course of the sixteenth and seventeenth centuries than in the whole of the rest of South America. Accordingly, by the beginning of the eighteenth century the Indians had been either driven back or else assimilated, and at the beginning of the War of Liberation early in the nineteenth century Chile was the only State in which there was a uniform Spanish-speaking population. In the greater part of Chile the population is denser than anywhere else in South America, and this population speaks nothing but Spanish, while in Peru and Bolivia nearly the whole rural population still speaks more or less exclusively Keshua or Aimará, and these languages are also used occasionally, or at any rate understood, by the whites. Chile is thus the only country in which a real Spanish people’s dialect could develop. (In Hempl’s classification this would be a typical case of class 2a.) In the other Spanish-American countries the Spanish-speakers are confined to the upper ruling class, there being practically no lower class with Spanish as its mother-tongue, except in a couple of big cities. Thus we understand that the Peruvian who has learnt his Spanish at school has a purer Castilian pronunciation than the Chilean; yet, apart from pronunciation, the educated Chilean’s Spanish is much more correct and fluent than that of the other South Americans, whose language is stiff and vocabulary scanty, because they have first learnt some Indian language in childhood. Lenz’s Chileans, who have often been invoked by the adherents of the unlimited substratum theory, thus really serve to show that sound substitution takes place only under certain well-defined conditions.
Pușcariu (in Prinzipienfragen der romanischen Sprachwissenschaft, Beihefte zur Zschr. f. rom. Phil., 1910) says that in a Saxon village which had been almost completely Rumanianized he had once talked for hours with a peasant without noticing that he was not a native Rumanian: he was, however, a Saxon, who spoke Saxon with his wife, but Rumanian with his son, because the latter language was easier to him, as he had acquired the Rumanian basis of articulation. Here, then, there was no sound substitution, and in general we may say that the less related two languages are, the fewer will be the traces of the original language left on the new language (p. 49). The reason must be that people who naturally speak a closely related language are easily understood even when their acquired speech has a tinge of dialect: there is thus no inducement for them to give up their pronunciation. Pușcariu also found that it was much more difficult for him to rid himself of his dialectal traits in Rumanian than to acquire a correct pronunciation of German or French. He therefore disbelieves in a direct influence exerted by the indigenous languages on the formation of the Romanic languages (and thus goes much further than Hempl). All these languages, and particularly Rumanian, during the first centuries of the Middle Ages underwent radical transformations not paralleled in the thousand years ensuing. This may have been partly due to an influence exerted by ethnic mixture on the whole character of the young nations and through that also on their language. But other factors have certainly also played an important rôle, especially the grouping round new centres with other political aims than those of ancient Rome, and consequent isolation from the rest of the Romanic peoples. Add to this the very important emancipation of the ordinary conversational language from the yoke of Latin. In the first Christian centuries the influence of Latin was so overpowering in official life and in the schools that it obstructed a natural development. But soon after the third century the educational level rapidly sank, and political events broke the power not only of Rome, but also of its language. The speech of the masses, which had been held in fetters for so long, now asserted itself in full freedom and with elemental violence, the result being those far-reaching changes by which the Romanic languages are marked off from Latin. Language and nation or race must not be confounded: witness Rumania, whose language shows very few dialectal variations, though the populations of its different provinces are ethnically quite distinct (ib. p. 51).
XI.—§ 9. Summary.
The general impression gathered from the preceding investigation must be that it is impossible to ascribe to an ethnic substratum all the changes and dialectal differentiations which some linguists explain as due to this sole cause. Many other influences must have been at work, among which an interruption of intercourse created by natural obstacles or social conditions of various kinds would be of prime importance. If we take ethnic substrata as the main or sole source of dialectal differentiation, it will be hard to account for the differences between Icelandic and Norwegian, for Iceland was very sparsely inhabited when the ‘land-taking’ took place, and still harder to account for the very great divergences that we witness between the dialects spoken in the Faroe Islands. A mere turning over the leaves of Bennike and Kristensen’s maps of Danish dialects (or the corresponding maps of France) will show the impossibility of explaining the crisscross of boundaries of various phonetic phenomena as entirely due to ethnical differences in the aborigines. On the other hand, the speech of Russian peasants is said to be remarkably free from dialectal divergences, in spite of the fact that it has spread in comparatively recent times over districts inhabited by populations with languages of totally different types (Finnic, Turkish, Tataric). I thus incline to think that sound substitution cannot have produced radical changes, but has only played a minor part in the development of languages. There are, perhaps, also interesting things to be learnt from conditions in Finland. Here Swedish has for many centuries been the language of the ruling minority, and it was only in the course of the nineteenth century that Finnish attained to the dignity of a literary language. The sound systems of Swedish and Finnish are extremely unlike: Finnish lacks many of the Swedish sounds, such as b, d (what is written d is either mute or else a kind of weak r), g and f. No word can begin with more than one consonant, consequently Swedish strand and skräddare, ‘tailor,’ are represented in the form of the loan-words ranta and räätäli. Now, in spite of the fact that most Swedish-speaking people have probably spoken Finnish as children and have had Finnish servants and playfellows to teach them the language, none of these peculiarities have influenced their Swedish: what makes them recognizable as hailing from Finland (‘finska brytningen’) is not simplification of consonant groups or substitution of p for b, etc., but such small things as the omission of the ‘compound tone,’ the tendency to lengthen the second consonant in groups like ns, and European (‘back’) u instead of the Swedish mixed vowel.
But if sound substitution as a result of race-mixture and of conquest cannot have played any very considerable part in the differentiation of languages as wholes, there is another domain in which sound substitution is very important, that is, in the shape which loan-words take in the languages into which they are introduced. However good the pronunciation of the first introducer of a word may have been, it is clear that when a word is extensively used by people with no intimate and first-hand knowledge of the language from which it was taken, most of them will tend to pronounce it with the only sounds with which they are familiar, those of their own language. Thus we see that the English and Russians, who have no [y] in their own speech, substitute for it the combination [ju, iu] in recent loans from French. Scandinavians have no voiced [z] and [ʒ] and therefore, in such loans from French or English as kusine, budget, jockey, etc., substitute the voiceless and [ʃj], or [sj]. The English will make a diphthong of the final vowels of such words as bouquet, beau [bu·kei, bou], and will slur the r of such French words as boulevard, etc. The same transference of speech habits from one’s native language also affects such important things as quantity, stress and tone: the English have no final short stressed vowels, such as are found in bouquet, beau; hence their tendency to lengthen as well as diphthongize these sounds, while the French will stress the final syllable of recent loans, such as jury, reporter. These phenomena are so universal and so well known that they need no further illustration.