Hirt goes so far as to think it possible by the help of existing dialect boundaries to determine the extensions of aboriginal languages (Idg 19).
There is certainly something very plausible in this manner of viewing linguistic changes, for we all know from practical everyday experience that the average foreigner is apt to betray his nationality as soon as he opens his mouth: the Italian’s or the German’s English is just as different from the ‘real thing’ as, inversely, the Englishman’s Italian or German is different from the Italian or German of a native: the place of articulation, especially that of the tongue-tip consonants, the aspiration or want of aspiration of p, t, k, the voicing or non-voicing of b, d, g, the diphthongization or monophthongization of long vowels, the syllabification, various peculiarities in quantity and in tone-movements—all such things are apt to colour the whole acoustic impression of a foreigner’s speech in an acquired language, and it is, of course, a natural supposition that the aboriginal inhabitants of Europe and Asia were just as liable to transfer their speech habits to new languages as their descendants are nowadays. There is thus a priori a strong probability that linguistic substrata have exercised some influence on the development of conquering languages. But when we proceed to apply this natural inference to concrete examples of linguistic history, we shall see that the theory does not perhaps suffice to explain everything that its advocates would have it explain, and that there are certain difficulties which have not always been faced or appraised according to their real value. A consideration of these concrete examples will naturally lead up to a discussion of the general principles involved in the substratum theory.
XI.—§ 2. French u and Spanish h.
First I shall mention Ascoli’s famous theory that French [y·] for Latin u, as in dur, etc., is due to Gallic influence, cf. Welsh i in din from dun, which presupposes a transition from u to [y]. Ascoli found a proof in the fact that Dutch also has the pronunciation [y·], e.g. in duur, on the old Keltic soil of the Belgæ, to which Schuchardt (SlD 126) added his observation of [y] in dialectal South German (Breisgau), in a district in which there had formerly been a strong Keltic element. This looks very convincing at first blush. On closer inspection, doubts arise on many points. The French transition cannot with certainty be dated very early, for then c in cure would have been palatalized and changed as c before i (Lenz, KZ 39. 46); also the treatment of the vowel in French words taken over into English, where it is not identified with the native [y], but becomes [iu], is best explained on the assumption that about 1200 A.D. the sound had not advanced farther on its march towards the front position than, say, the Swedish ‘mixed-round’ sound in hus. The district in which [y] is found for u is not coextensive with the Keltic possessions; there were very few Kelts in what is now Holland, and inversely South German [y] for u does not cover the whole Keltic domain; [y] is found outside the French territory proper, namely, in Franco-Provençal (where the substratum was Ligurian) and in Provençal (where there were very few Galli; cf. Wechssler, L 113). Thus the province of [y] is here too small and there too large to make the argument conclusive. Even more fatal is the objection that the Gallic transition from u to y is very uncertain (Pedersen, GKS 1. § 353). So much is certain, that the fronting of u was not a common Keltic transition, for it is not found in the Gaelic (Goidelic) branch.[42] On the other hand, the transition from to [y] occurs elsewhere, independent of Keltic influence, as in Old Greek (cf. also the Swedish sound in hus): why cannot it, then, be independent in French?
Another case adduced by Ascoli is initial h instead of Latin f in the country anciently occupied by the Iberians. Now, Basque has no f sound at all in any connexion; if the same aversion to f had been the cause of the Spanish substitution of h for f, we should expect the substitution to have been made from the moment when Latin was first spoken in Hispania, and we should expect it to be found in all positions and connexions. But what do we find instead? First, that Old Spanish had f in many cases where modern Spanish has h (i.e. really no sound at all), and this cannot be altogether ascribed to ‘Latinizing scribes.’ On the contrary, the transition f > h seems to have taken place many centuries after the Roman invasion, since the Spanish-speaking Jews of Salonika, who emigrated from Spain about 1500, have to this day preserved the f sound among other archaic traits (see F. Hanssen, Span. Gramm. 45; Wiener, Modern Philology, June 1903, p. 205). And secondly, that f has been kept in certain connexions; thus, before [w], as in fuí, fuiste, fué, etc., before r and l, as in fruto, flor, etc. This certainly is inexplicable if the cause of f > h had been the want of power on the part of the aborigines to produce the f sound at all, while it is simple enough if we assume a later transition, taking place possibly at first between two vowels, with a subsequent generalization of the f-less forms. Diez is here, as not infrequently, more sensible than some of his successors (see Gramm. d. roman. spr., 4th ed., 1. 283 f., 373 f.).
XI.—§ 3. Gothonic and Keltic.
Feist (KI 480 ff.: cf. PBB 36. 307 ff., 37. 112 ff.) applies the substratum theory to the Gothonic (Germanic) languages. The Gothons are autochthonous in northern Europe, and very little mixed with other races; they must have immigrated just after the close of the glacial period. But the arrival of Aryan (Indogermanic) tribes cannot be placed earlier than about 2000 B.C.; they made the original inhabitants give up their own language. The nation that thus Aryanized the Gothons cannot have been other than the Kelts; their supremacy over the Gothons is proved by several loan-words for cultural ideas or state offices, such as Gothic reiks ‘king,’ andbahts ‘servant.’ The Aryan language which the Kelts taught the Gothons was subjected in the process to considerable changes, the old North Europeans pronouncing the new language in accordance with their previous speech habits; instead of taking over the free Aryan accent, they invariably stressed the initial syllable, and they made sad havoc of the Aryan flexion.
The theory does not bear close inspection. The number of Keltic loan-words is not great enough for us to infer such an overpowering ascendancy on the part of the Kelts as would force the subjected population to make a complete surrender of their own tongue. Neither in number nor in intrinsic significance can these loans be compared with the French loans in English: and yet the Normans did not succeed in substituting their own language for English. Besides, if the theory were true, we should not merely see a certain number of Keltic loan-words, but the whole speech, the complete vocabulary as well as the entire grammar, would be Keltic; yet as a matter of fact there is a wide gulf between Keltic and Gothonic, and many details, lexical and grammatical, in the latter group resemble other Aryan languages rather than Keltic. The stressing of the first syllable is said to be due to the aboriginal language. If that were so, it would mean that this population, in adopting the new speech, had at once transferred its own habit of stressing the first syllable to all the new words, very much as Icelanders are apt to do nowadays. But this is not in accordance with well established facts in the Gothonic languages: we know that when the consonant shift took place, it found the stress on the same syllables as in Sanskrit, and that it was this stress on many middle or final syllables that afterwards changed many of the shifted consonants from voiceless to voiced (Verner’s law).[43] This fact in itself suffices to prove that the consonant shift and the stress shift cannot have taken place simultaneously, and thus cannot be due to one and the same cause, as supposed by Feist. Nor can the havoc wrought in the old flexions be due to the inability of a new people to grasp the minute nuances and intricate system of another language than its own; for in that case too we should have something like the formless ‘Pidgin English’ from the very beginning, whereas the oldest Gothonic languages still preserve a great many old flexions and subtle syntactical rules which have since disappeared. As a matter of fact, many of the flexions of primitive Aryan were much better preserved in Gothonic languages than in Keltic.
XI.—§ 4. Etruscan and Indian Consonants.
In another place in the same work (KI 373) Feist speaks of the Etruscan language, and says that this had only one kind of stop consonants, represented by the letters k (c), t, p, besides the aspirated stops kh, th, ph, which in some instances correspond to Latin and Greek tenues. This, he says, reminds one very strongly of the sound system of High German (oberdeutschen) dialects, and more particularly of those spoken in the Alps. Feist here (and in PBB 36. 340 ff.) maintains that these sounds go back to a Pre-Gothonic Alpine population, which he identifies with the ancient Rhætians; and he sees in this a strong support of a linguistic connexion between the Rhætians and Etruscans. He finds further striking analogies between the Gothonic and the Armenian sound systems; the predilection for voiceless stops and aspirated sounds in Etruscan, in the domain of the ancient Rhætians and in Asia Minor is accordingly ascribed to the speech habits of one and the same aboriginal race.