¹ This paper was read, in abridged form, at the meeting of the Linguistic Society in New York, 31 December 1948.

² Some of the important sources in these subjects are: V. G. Childe, The dawn of European civilization (New York, 1939); Conway, Johnson, and Whatmough, The Prae-Italic dialects of Italy (Cambridge, Mass., 1933); Friedrich von Duhn, Italische Gräberkunde (Heidelberg, 1924); C. F. C. Hawkes, The prehistoric foundations of Europe (London, 1940); F. Messerschmidt, Bronzezeit und frühe Eisenzeit in Italien (Berlin, 1935); E. Pais, Storia dell'Italia antica (Torino, 1933); T. E. Peet, The stone and bronze ages in Italy and Sicily (Oxford, 1909); D. Randall-MacIver, The iron age in Italy (Oxford, 1927), Italy before the Romans (Oxford, 1928), Villanovans and early Etruscans (Oxford, 1924); J. Whatmough, The foundations of Roman Italy (London, 1937).

³ See below 251.

⁴ Vittorio Bertoldi, Problèmes de substrat, BSL 32.118 (1931): ‘Que la langue latine victorieuse ait effacé, par exemple, partout sur le domaine où elle s'est superposée, toutes les habitudes articulatoires des indigènes, qui étaient le résultat d'une hérédité séculaire, nul ne saurait l'affirmer.’ But some do so affirm. (The term hérédité will be discussed below, fn. 11.) See on this basic principle especially Antoine Meillet, Sur les effets des changements de langue, Scientia 51.91–8 (1932). However, his use of the term substratum, as meaning ‘des changements que subirait, postérieurement au changement de langue, la langue nouvellement adoptée’ (95), is wider than we want it to be (see below, fn. 11); Meillet himself calls this atavistic theory an ‘hypothèse hardie’ in any event.

⁵ A paper by George Hempl, Language rivalry and speech differentiation in the case of race mixture, TAPA 39.31–47 (1898), offers valuable suggestions, apart from the unfortunate term ‘race mixture’. Too little attention has been paid to it. (Jespersen discusses the article in his Language, chapter XI.)

⁶ See above, fn. 4.

⁷ Romanic Review 32.445–9 (1941).

⁸ It may be noteworthy that Pei, in spite of his anti-substratum position, calls Piedmontese, Lombard, Emilian, and Ligurian ‘Gallo-Italian’ dialects (146 and passim). Perhaps the term is geographical; but there is no Gallo-Italy.

⁹ A student of mine declared that his attempts to learn to pronounce French correctly were doomed in advance, since he, being of Anglo-Saxon ‘extraction’, could not possibly cope with this Gallic way of talking. More frequent are the unfounded hopes of some that because their grandfather spoke Italian, French, or German, the languages were ‘in their blood’. Often such reliance upon ancestral virtues determines a student's choice of language study in college.

¹⁰ Perhaps there is a connection between this motor-automatism of audible speech and the non-motorized automatism of the neural paths of inaudible speech, i.e. thought, which consists in thalamo-cortical circulation. Cp. G. C. Campion and G. Elliot Smith, The neural basis of thought 32, fig. 3 (London, 1934).

¹¹ So eminent a linguist as Jakob van Ginneken maintained that the two varieties of the Dutch r and l are due to racial differences among speakers of Dutch (Proceedings of the Second International Congress of Phonetic Sciences at London 65–6 [1935])—that ‘il y a vraiment des races [N.B.] qui aiment la fermeture du larynx et il y a des races qui n'en veulent rien’ (67). He then passes on to a description of dolichocephalic and brachycephalic speech organs and sound production habits (67–8). (One is reminded of Max Müller's scathing words about ‘a dolichocephalic dictionary or a brachycephalic grammar’, Biographies of words 120 [London, 1888]) Another paper on the same subject by the same author is cryptically entitled La tendance labiale de la race méditerranéenne et la tendance laryngale de la race alpine, Archives néerlandaises de phonétique experimentale 8/9.76–92 (1933). See also M. J. Sirks, Klang, Stimme und Laut im Geltungsbereich der Vererbungslehre, Mélanges van Ginneken 1–13 (Paris, 1937), with bibliography 14–5: ‘Die typischen laryngalen Menschen aus Leuven, Aalst und Hasselt in Belgien, welche die Lippen fast immer geschlossen halten [which is supposedly a racial characteristic, not merely a habit] und auch Lippen ohne jeden Wulst besitzen, sind ein ausgezeichnetes Beispiel …, da solche Menschen den ö-Laut gar nicht aussprechen können’ (13); note the word ‘können'! It is hardly necessary to observe that if a Belgian child, however tight-lipped and thin-lipped his ‘racial heredity’ might be, were brought up in an area where ö was a phoneme, he would naturally have no difficulty in mastering it. It was again Meillet who answered the problem most succinctly (Scientia 51.97 [1932]) ‘Il ne s'agirait pas de l'hérédité de caractères anatomiques acquis, mais d'une chose bien différente, d'hérédité d'habitudes acquises.’ Instead of the second ‘hérédite', however, one had better read, ‘continuation de père en fils.‘ No doubt physical properties of the vocal organs may be inherited. But to make a linguistic shibboleth of Mendel's discoveries is absurd.

Similarly, F. L. Pullè, Italia: genti e favelle (Torino, 1927), shows how, in spite of the ‘diversità di forme craniche, di composizione di plasma [N.B.]’ (564), all Italians really form one people, and how there is basic agreement of the ‘tipo somatico annalizzato ne’ suoi principali caratteri, col tipo glottologico e col tipo psicologico dei rispettivi idiomi' (560).

¹² Even so comprehensive a change as, say, the dissimilation of aspirates (Grassmann's law), operates in Sanskrit and in Greek independently. Palatalization of consonants before yod in French can hardly have a causal connection with the same phenomenon in Umbrian. Colloquial American innerest for interest and unnerstand for understand have no relation with the Oscan change from -nd- to -nn-. But if in Naples, i.e. in former Oscan territory, they say today unnici for undici, quanno for quando, a linguist ought to take another look.

¹³ So, for example, syntactical affinities between Egyptian and Keltic—see Julius Pokorny, Zeitschrift für celtische Philologie 16.95–144, 231–66, 363–94 (1926); 17.373–88 (1928); 18.233–48 (1930); William Lansberg, Lg. 16.95–9 (1940)—cannot be said to have any more connection with one another than ποταμός and Potomac, both meaning ‘river,‘ unless someone first proves the reality of contact and exchange of cultural goods between the Egyptians and the Kelts. Similarly, the theory of substratum influence of Oscan in northern Spain, claimed by Menéndez Pidal (Origenes del español §§53 ff.) and supported by the local name Huesca, allegedly derived from Osca, remains debatable until it has been proved that a comparatively large number of Oscans, legionnaires or colonists or both, actually lived in that area, spoke Oscan, and were influential enough to impose at least one of their speech habits over a wide district. And even if that were proved, we should be dealing with a superstratum, like Frankish in Gaul. Besides, Rohlfs finds a town named Huescar in the south of Spain, and thinks that both names are of pre-Roman origin. See also W. D. Elcock, De quelques affinités phonétiques entre l'aragonais et le béarnais (Paris, 1938), and Gerhard Rohlfs, Le gascon: Études de philologie pyrénéenne, Zeitschrift für romanische Philologie, Beiheft 85 (Halle, 1935).

¹⁴ Bertoldi, op.cit. 117 n. 1: ‘Que les linguistes les plus éminents se tiennent en droit de désavouer les conclusions par trop hâtives de ceux qui d'un sain principe de recherche ont fait une clef pour ouvrir toutes les portes, qu'ils réclament de plus en plus de rigueurs de la méthode, … tout cela est compréhensible, c'est surtout salutaire pour le prestige de notre discipline. Mais il s'agit toujours de réserves relatives à la méthode de nature à ne préjuger guère l'attitude à l'égard de la théorie. En général, on ne saurait trop insister sur la nécessité que dans ce genre de recherches, où forcément une si large place est réservée à l'hypothèse, la clarté de la forme soit en raison directe de l'obscurité de la matière.'

It has been said that the substratum theory never proved anything that could not be proved by other methods. In some instances that may be so, whereas in others only the corroborative evidence of substratum can settle an argument.

¹⁵ Romanic Review 32.113 (1941), in the review of Walther von Wartburg, La posizione della lingua italiana (Firenze, 1940).

¹⁶ Germanisch-romanische Monatsschrift 18.37–56 (1930).

¹⁷ La struttura linguistica dell'Italia 21 (Leipzig, 1937).

¹⁸ So already Hermann Hirt, Die Verwandtschaftsverhältnisse der Indogermanen, IF 4.38 f. (1894); see also Hirt, ed. H. Arntz, Die Hauptprobleme der indogermanischen Sprachwissenschaft 47, 209 (Halle, 1939).

¹⁹ The general ethnological principles, applicable to all areas, may be found in the paper by George Hempl already referred to (fn. 5).

²⁰ The boundary which, according to Wartburg and other Romanists, separates Eastern from Western Romance, runs approximately along the same course. Whatever the merits of this division of Romance territory, the real existence of an important linguistic frontier cannot be denied.

²¹ See Joshua Whatmough, Quemadmodum Pollio reprehendit in Liuio Patauinitatem, Harvard Studies in Classical Philology 44.95–130 (1933). Cf. also the older works by Hugo Schuchardt, Der Vocalismus des Vulgärlateins (Leipzig 1866–8), and Karl Sittl, Die lokalen Verschiedenheiten der lateinischen Sprache (Erlangen, 1882).

²² H. Pedersen, Linguistic science in the nineteenth century 54 (Cambridge, Mass., 1931).

²³ G. E. F. Chilver, Cisalpine Gaul: Social and economic history from 49 b.c. to the death of Trajan 7 (Oxford, 1941).

²⁴ 2.35.4.

²⁵ See Joshua Whatmough, keʌtika, Harvard Studies in Classical Philology 55.82–5 (1944). For Keltic relics especially in northern districts (Switzerland) see J. Hubschmied, Sprachliche Zeugen für das späte Aussterben des Gallischen, Vox Romanica 3.48–155 (1938).

²⁶ Loc.cit. 48–53, 56.

²⁷ See above, 242.

²⁸ Loc.cit. 49.

²⁹ It should be added that not all unvoiced intervocalic stops are always aspirated. For a syntactical explanation of the variations see Enrico Bianchi, La lingua italiana 46 ff. (Firenze, 1943).

³⁰ Heinrich Nissen, Italische Landeskunde 1.494 ff. (Berlin 1883–1902).

³¹ For the last see Italia dialettale 3.84 ff. (1927). On the other hand, I have seen it suggested, also by Merlo, that the substratum in Tuscany may be not Etruscan but Latin, whereas that of Rome is Samnite, i.e. Oscan: Lazio sannita ed Etruria latina?, Studi etruschi 1.303–11 (1927). But before that, Francesco Ribezzo had demonstrated how, in the wake of colonial expansion, there reappear in Pugliese, Salentino, certain features of Oscan which do not survive in the Campano-samnite area: Reliquie italiche nei dialetti delPItalia meridionale, Atti della Reale Accademia di archeologia, lettere e belle arti, N.S. 1, Part II, 151–69 (1910).

³² I.XIII.2.

³³ ‘Sed quamquam fere omnes Tusci in suo turpiloquio sint obtusi …‘ (De vulg. el. I.XIII.3).

³⁴ This is an argument ex silentio, hence debatable. But hardly any room for debate is left if one also disputes the validity of the most objective set of data for modern Italian dialects at our disposal, namely Jaberg–Jud's linguistic atlas, as does Carlo Battisti, Aspirazione etrusca e gorgia toscana, Studi etruschi 4.249–54 (1930), who terms some of the atlas data ‘molto discutibili’ (254). In the matter of Etruscan substratum he offers exactly what we do not want, viz. ‘la ripetizione a distanza di tempo di tendenze insite in una predisposizione organica’ (253–4), which is a piece of unsupported racial predeterminism. Current usage with its vagaries indicates that the gorgia is a rather recent development which is still in progress and therefore not easily codified and pressed into rules. After all, linguistic investigation, especially dialectology, should reserve a space for certain unfinished, labile phonological phenomena which have not established themselves so thoroughly as to dislodge the older forms completely among all classes of speakers within an area. One should not proceed on the assumption that all change lies in the past and that, at the time of one's research, a rule or even a ‘law’ must be found for every phonemic divergence among dialects.

³⁵ See Wilhelm Schulze, Zur Geschichte lateinischer Eigennamen, Abhandlungen der königlichen Gesellschaft der Wissenschaften zu Göttingen, phil.-hist. Klasse, N.F. 2 (Berlin, 1904).

³⁶ See Pulgram, The origin of the Roman nomen gentilicium, Harvard Studies in Classical Philology 58/59.163–87 (1948). It has also been claimed that the strong stress on the initial syllable in archaic Latin, which led to the weakening of the stem vowel in compounds (facere but conficere), rests on Etruscan habits of accentuation. (These habits we glean from the form in which non-Etruscan names appear on Etruscan inscriptions: powerful syncope after the strongly stressed first syllable produces such telescoped forms as Klutmnsta for Klytaemnestra.) But it is also thought that Greek pitch accent modified Latin pronunciation (see C. D. Buck, Comparative grammar of Greek and Latin 161 ff. [Chicago, 1923]), and here there can be no question of a substratum, nor of an extensive popular superstratum. Perhaps we are dealing, in both cases, with a fashion introduced by, and prevalent in, the upper classes, modeled on a foreign speech of cultural superiority.

³⁷ Concerning the phenomenon most frequently connected with Umbrian, the change from d to r, see the exhaustive study by Ernst G. Wahlgren, Un problème de phonétique romane (Skrifter utgivna av K. Humanistiska Vetenskaps-Samfundet i Uppsala, 26:4; Uppsala, 1930).

³⁸ For the development of this theory see Joshua Whatmough, Foundations of Roman Italy 194 ff. (London, 1937).

³⁹ See Ettore Pais, Storia dell'Italia antica 1.51 ff. (Torino, 1933).

⁴⁰ Although the ‘Latin’ heard in Italian, French, Spanish, and Rumanian churches is not the same, still it is not Italian, French, Spanish, or Rumanian.

⁴¹ 5.4.8.

⁴² See the novel by Carlo Levi, Cristo si è fermato a Eboli (1945). An English translation was published in 1946.

⁴³ On the doubtful value of an Oscan substratum theory if divorced from non-linguistic continuity see above, fn. 13. As for the phonetic faithfulness of our records, Oscan spelling is quite reliably phonetic (which Umbrian is not), so that, for example, the distinction between V and V̂ I and Í, in spelling (in the Oscan alphabet) and pronunciation, is relevant to the explanation of southern Italian i for Latin ē, ĭ (elsewhere in the Romania, except Sardinia, e), and of u for Latin ō, ŭ (elsewhere in the Romania, except Sardinia and Rumania, o).

⁴⁴ See Gerhard Rohlfs, Etymologisches Wörterbuch der unteritalienischen Gräzitä ^t (Halle, 1930); Die Quellen des unteritalienischen Wortschatzes, Zeitschrift für romanische Philologie 46.135–64 (1926); Scavi linguistici nella Magna Grecia (Roma–Halle, 1933); and other publications.

⁴⁵ A good case for a Greek substratum for a number of dialectal phonemes has been made by Vittore Pisani, Neolinguistica e indoeuropeo 174 ff. (Roma, 1940).

⁴⁶ Vorlateinische Einflüsse 38–41, 56.

⁴⁷ See above, fn. 45.

⁴⁸ keʌtika 10.

⁴⁹ Meillet, Scientia 51.98 (1932): ‘… certaines concordances sont trop frappantes pour être attribuées au hasard et il faut reconnaître que dans les évolutions linguistiques consécutives à un changement de langue, certains traits résultent … d'une influence exercée par la langue qui a été éliminée.‘

⁵⁰ For Italy, going one step beyond earliest documentation, see Vittorio Bertoldi, Plurale mediterraneo in residue fossili, Mélanges van Ginneken 157–69, on a Mediterranean substratum. Names are probably the only clues one may properly expect.