¹ Language 302 (New York, 1933).

² Linguistica 104 (Firenze, 1946). On this point, cf. also the reviews by T. A. Sebeok (American Speech 22.137–8 [1947]) and the present writer (Lg. 22.259–61 [1946]).

³ Studies in Philology 43.463 (1946).

⁴ I use this term to refer specifically to Brugmann, Leskien, and the other Indo-Europeanists of the 1870's, 80's, and 90's who first developed the explicit formulation of the principles of comparative reconstruction. For that group of scholars—far more extensive in number and in time—who have accepted the basic postulate of regular sound-change, I would suggest using the term ‘regularist’; cf. my Terminological Notes, Studies in Linguistics 7.60–2 (1949).

⁵ Leipzig, 1890–1900.

⁶ E.g. Italienische Grammatik (Leipzig, 1890); Historische Grammatik der französischen Sprache (Heidelberg, 1908–21).

⁷ Cf. the present writer's discussions of the ‘idealistic’ approach and its effects on linguistics, in Italica 20.239–43 (1938); Lg. 17.263–9 (1941); Italica 23.30–4 (1946); Lg. 22.273–83 (1946); and SIL 6.27–35 (1948).

⁸ Cf. such criticisms of the comparative method as those of Bàrtoli, Introduzione alla Neolinguistica (Geneve, 1925; Biblioteca dell'Archivum Romanicum II.12); B. A. Terracini, ¿Qué es la Lingüística? (Tucumán, 1942).

It must be emphasized that our reconstruction does not lead us to set up a completely ‘unified’ or ‘unitary’ Proto-Romance, as is often assumed (cf. most recently Y. Malkiel, StP 46.512 [1949]). We do not have to suppose absolute uniformity for proto-languages, any more than for any actually observed language (cf. B. Bloch, Lg. 24.194 fn. 1 [1948]). Our Proto-Romance was undoubtedly a composite of several dialects of the Latin spoken at the end of the Republican period.

⁹ Cf. also Y. Malkiel, Lg. 21.149 (1945).

¹⁰ E.g. Meyer-Lübke's Grammatik der romanischen Sprachen; Bourciez's Elements de linguistique romane⁴ (Paris, 1946); Grandgent's Introduction to Vulgar Latin (Boston, 1907); and the historical grammars of individual languages, e.g., for Italian: Meyer-Lübke's Italienische Grammatik and its various Italian reworkings; Grandgent's From Latin to Italian (Cambridge, 1927); and Pei's The Italian Language (New York, 1941).

¹¹ ZRPh. 23.313–20 (1899).

¹² ZRPh. 22.521–5(1898).

¹³ For instance, the evidence afforded by Upper Aragonese dialects for preservation of intervocalic unvoiced stops, or by Asturian (Cabranes) dialect for preservation of the distinction between final -u and -o (cf. Y. Malkiel, Lg. 23.63 [1947], reviewing Josef a Maria Canellada, El bable de Cabranes [Madrid, 1944; RFE Anejo 31]); the similar evidence afforded by Upper Bearnese for preservation of intervocalic unvoiced stops (cf. the discussion in Part 2 of this article) and their ascription to Proto-Gallo-Romance; or the evidence afforded by Central and South Italian dialects (in the so-called ‘metafonesi centromeridionale‘) for the distinction between final -u and -o, which we must, therefore, likewise ascribe to Proto-Italo-Romance.

¹⁴ Die Mundarten Südlukaniens (Halle, 1939; ZRPh. Beiheft 90), especially pp. 84–6.

¹⁵ Examples will be given in phonemic transcription (roman type) and in the conventional orthography of the language (italics), with English glosses (enclosed in quotes). The ONFr. conventional orthography is the normalized style used, for instance, by Schwan-Behrens-Bloch (Grammaire de l'ancien français⁴ [Leipzig, 1932]) and by Jenkins in his edition of the Chanson de Roland (revised ed., Boston, 1929). The transcription is based on that of the IPA, except that c stands for [t^a], 5 for [d^z], and a prime after a consonant letter indicates palatalization.

¹⁶ The symbol suggested by me (StP 43.579 [1947]) for the phonemic entity in ONFr. which normally corresponds to /á/ of OSFr. and the other Romance languages, which assonated only with itself in the earliest stage of literary Old French (including the earlier stratum of the Roland), and which was later merged with /ε/ and spelled e. The exact phonetic nature of this phoneme is irrelevant, whether it was [æ] (as seems to me most likely, at least for the first stage of its development), [ε·], [e·], or what not.

¹⁷ Dialectal differentiation within OSFr.

¹⁸ With /r/ occurring in word-final position and involving diphthongization of a preceding /ε/ or /ɔ/.

¹⁹ For further discussion of this procedure of reanalysis and rephonemicization, cf. Zellig S. Harris, Methods in Descriptive Linguistics (to be published by the Linguistic Society of America).

²⁰ For the significance of such limitations on occurrence and their implications for earlier stages of the language, cf. H. M. Hoenigswald, Internal Reconstruction, SIL 2.78–87 (1944), and Sound Change and Linguistic Structure, Lg. 22.138–43 (1946); and for a discussion of the relation between phonemic change and the loss of factors which condition positional variants, cf. most recently W. F. Twaddell, Lg. 24.151 (1948).

²¹ Cf. M. Swadesh, The Phonemic Interpretation of Long Consonants, Lg. 13.1–10(1937).

²² Cf. the material gathered by W. D. Elcock, De quelques affinités phonétiques entre le béarnais et l'aragonais (Paris, 1938), especially the sections entitled Versant béarnais in Chapters 1–3 and the Conclusion in Chapter 4. The examples given here are from Elcock and the ALF.

²³ Cf. Elcock, op.cit. 121–2.

²⁴ Cf. G. L. Trager, The Phonemes of Russian, Lg. 10.334–44 (1934).

²⁵ Cf. Denzel R. Carr, Notes on Marshallese Consonant Phonemes, Lg. 21.268–9 (1945).

²⁶ Cf. E. H. Sturtevant, The Indo-Hittite Laryngeals (Baltimore, 1942), and An Introduction to Linguistic Science 158–63 (New Haven, 1947) and references given in fn. 7, p. 160, particularly the criticisms of Pedersen and Bonfante.

²⁷ Cf. the present writer's discussion of such objections in Lg. 22.273–83 (1946).

²⁸ In this table and in following discussions of PRom. vowel phonemes, the symbol ˆ indicates relatively high and tense tongue position, and ^v indicates relatively low and lax position; a raised dot · following a vowel indicates length.

²⁹ We could, of course, operating on an abstract plane, equally well decide to extract lowness of tongue position as a separate phonemic feature and symbolize it by //, leaving the high vowels unmarked. Our decision to extract height of tongue position and leave low vowels unmarked is admittedly determined by ulterior considerations, namely the ease of equating PRom. /ˆ/ with Latin /·/.

³⁰ Cf. the present writer's discussion of this word, SIL 5.65–8 (1947).

³¹ L'origine des langues romanes, Renaissance 1.573–88 (1943).

³² M. A. Pei, Symposium 1:3.118 (1947).

³³ Cf. the discussion in Bloomfield's Language, Chapter 20.

³⁴ Cf. Bloomfield, Language §18.2.

³⁵ Cf. the penetrating and illuminating discussion of the comparative method and of the assumption of phonetic change by C. F. Hockett, Implications of Bloomfield's Algonquian Studies, Lg. 24.117–31 (1948), especially 125–7.

³⁶ H. F. Muller, The Chronology of Vulgar Latin (Halle, 1929; ZRPh. Beiheft 78); H. F· Muller and P. Taylor, A Chrestomathy of Vulgar Latin (New York, 1932); H. F. Muller, L'Epoque mérovingienne (New York, 1945), and cf. the review of the latter book by A. H· Krappe, Philological Quarterly 26.92–5 (1947).

^36a In this connection, the Romance words belonging to the family of It. pensare, Fr. penser, Sp. pensar ‘to think’ seem at first to contradict this statement, and to give evidence for a cluster /ns/. The answer is that the evidence thus afforded is valid only for Italo-Western Romance. In Roumanian and Sardinian, only forms without /n/, meaning ‘weigh’ or ‘press, worry’, are present; cf. Meyer-Lübke, REW³ §6391. For Proto-Romance, therefore, we have the right to set up only /peˆsáre/ ‘to weigh, press down’; the learned word /peˆnsáre/ is to be ascribed only to the PItWRom. stage. Naturally, as soon as /peˆnsáre/ was introduced from Classical Latin, it brought the cluster /ns/ back again, but evidently only into that part of Romance speech which was continued in Italy and the West. A similar argument applies in the case of the Greek loan-word κάμπτ∊ιν ‘bend, turn, double around, bow down’ > Lat. campsāre > It. cansare ‘set aside, avoid’ and Sp. cansar ‘weary’, and other Romance words showing the cluster /ns/. (Cf. Lg. 14.205–6 and 19.154–6.) Late Latin spellings such as thensaurus for thesaurus show simply that there was a dialectal difference at the time, and do not necessarily prove that we must assume the cluster /ns/ for PRom.

³⁷ In an inscription of the Christian era from Rome, reprinted in Muller and Taylor. A Chrestomathy of Vulgar Latin 108, without further indication of source.

³⁸ For this reason, old documents and relic forms in modern speech acquire a value, in this connection, seemingly quite disproportionate to their usefulness in other connections (literary or esthetic, or even in characterizing modern dialects). The scholar reconstructing a proto-language must, of necessity, be something of an anticuario verbal (as one Aragonese termed Elcock; cf. Elcock, op.cit. 19). Opponents of the comparative method have made a reproach of this fact, and have characterized comparatists and Neo-Grammarians as ‘seekers after dead fossils’ and the like; cf. M. G. Bàrtoli, Introduzione alla Neolinguistica; B. A. Terracini, ¿Qué es la Lingüística? 34; G. Bonfante, Lg. 23.360, 367 (1947). There is, of course, as much justification for ‘fossil-seeking’ in this connection as there is in any other historical study, such as geology or comparative anatomy. Far from being out of touch with the process of growth and change in biological or social life, the good ‘fossil-seeker’ derives an understanding of life from his work.

³⁹ Cf. E. Bovet, Ancora il problema andare, in Scritti varî di filologia (A Ernesto Monaci) 243–62 (Roma, 1901); and the criticisms of his procedure by E. Gorra, Rassegna bibliografica della letteratura italiana 10.103 (1902), and by C. Salvioni, Archivio glottologico italiano 16.209–10.

⁴⁰ G. Bonfante, Lg. 23.374 (1947).

⁴¹ E.g. Bàrtoli, Terracini, Bonfante.

⁴² As pointed out most recently by A. Goetze, Lg. 17.168 (1941); cf. also the well-balanced discussion of the relation between comparative method and linguistic geography by Bloomfield, Language, Chapters 18–20.

⁴³ It is through exaggeration of the factors of geographical position that M. G. Bàrtoli and his followers have been led into setting up ad-hoc rules (norme) by which all the evidence is judged. Cf. Bàrtoli's unsuccessful attempts to explain important exceptions to his rules, such as the conservative character of Sardinian (Introduzione alla Neolinguistica) or of Italian in the center of the Romance-speaking territory (Per la storia della lingua d'ltalia, AG1B 21.72–94 [1927]).

⁴⁴ Cf. H. Lausberg, Die Mundarten Südlukaniens; and M. L. Wagner's works on Sardinian, especially his Historische Lautlehre des Sardischen (Halle, 1941; ZRPh. Beiheft 93) and Flessione nominale e verbale del sardo antico e moderno, Italia Dialettale 14.93–170 (1938) and 15.1–29 (1939).

⁴⁵ Certainly it is not conducive to clarity to use the term ‘Vulgar Latin’ to apply indiscriminately to all material written in Latin since Classical times, since the degree to which popular speech is reflected in such documents varies greatly and is anything but trustworthy. ‘Late Latin’ is a much better term for this type of material, since it implies no judgment as to the accuracy with which the writing reflects everyday usage.

⁴⁶ That is to say, we might assume that in a certain dialect of PRom., initial /l/ and /i̯/ were merged, either in /l/ or in /i̯/ or in a third development (such as /l'/), so that PRom. lí^ˆliu and i̯ú^ˆli̯u came to be identical in their initial sound. Then at a later stage, these two forms were subjected to 'false regression', with *i̯í ^Aliu and *lú^ˆliu arising as over-corrected forms and surviving in Italian. Cf. J. Babad, ZRPh. 19.270 (1895); also C. H. Grandgent, From Latin to Italian 70; Meyer-Lübke (tr. Bàrtoli and Braun), Grammatica storica della lingua italiana e dei dialetti toscani 91–2 (Torino, 1931); Pei, The Italian Language 50–1 (New York, 1941)—all of which either leave the problem unsolved or assume some kind of dissimilatory process.

⁴⁷ Cf. the discussion of the relation between philology, field method, and reconstruction, by C. F. Hockett, Lg. 24.118 ff. (1948).

⁴⁸ Cf., for instance, L. H. Gray, Foundations of Language 460 (New York, 1939).