¹ The detailed comparison, of which this article presents the essential conclusions, may be found in my doctoral dissertation, A reconstruction of Proto-Rhaeto-Romance and its implications for the history of French (Cornell University, 1960).

² Nine villages were chosen: Gradisca in SE Friuli, Colloredo di Montealbano in central Friuli, and Pesariis in NW Friuli; Pieve di Livinallongo southeast of the Sella massif (Dolomites), Selva in the Val Gardena west of the massif, and San Vigilio di Marebbe north of the massif; Scuol on the Inn River in E Grisons, Trun on the Rhine in W Grisons, and Cunter halfway between on neither river. I investigated these dialects in 1958–9 under a Fulbright scholarship. A fourth Dolomite dialect, that of Moena south of the Sella massif, was taken from Luigi Heilmann, La parlata di Moena (Bologna, 1955), a study cast in phonemic terms.

To simplify the process of reconstruction the dialects were handled in their traditional groupings, and a Proto-Friulian, a Proto-Dolomite, and a Proto-Grisons were reconstructed first. These substages cannot be taken seriously without further examination. Tracing forward from Proto-Rhaeto-Romance we find the three Friulian dialects quite like each other and quite distinct from the Dolomite and Grisons dialects, and therefore Proto-Friulian may have been the language of a group with a social identity. No generally shared features of phonological innovation, however, characterize the Dolomite or Grisons groups. In many ways similarities in Moena, San Vigilio, and Scuol suggest an intermediate period of common descent, and likewise Gardena and Livinallongo may go together; but apparently Cunter and Trun fit neither with one of the groups nor with each other.

³ Edouard Bourciez, Précis de phonétique française⁸ §14 (Paris, 1950).

⁴ J. Bédier, La Chanson de Roland publiée d'après le manuscrit d'Oxford (Paris, 1922).

⁵ Trun regularly preserves ‘fourth’ etc.

⁶ Read this symbol as ‘umlauted long open o’.

⁷ Comparative work based only on the standard languages does not reveal the need for reconstructing posttonic /u/ for Proto-Italo-Western. Rather, it has been assumed that Proto-Romance posttonic /u/ and /o/ merged to give PItW posttonic /o/. The redefinition that PItW posttonic /o/ < /o/ and /u/ < /u/ were still distinct is the only refinement of Proto-Italo-Western necessary for, or contributed by, the reconstruction of Proto-Rhaeto-Romance.

⁸ There is some suggestion that checked participated in the same split, but the evidence is only in the Grisons dialects, and it is less risky to assume that there was a secondary morphological extension in Grisons than to say there was a phonological innovation in Proto-Rhaeto-Romance that was leveled without trace in Dolomite and Friulian (and French). Cf. such forms as Dissentis (near Trun) ‘body’, pl. etc., in Theodor Gartner, Raetoromanische Grammatik §45 (Heilbronn, 1883).

The few authentic cases of PRR short are not connected with the umlaut mechanism outlined here. They occur before certain PItW palatal clusters (but not before others): PItW ‘eye’ (in which /kl/ checked the stressed vowel without, however, necessarily becoming the geminate found in Italian occhio) > PRR , but PItW ‘thigh’ > PRR .

⁹ Actually for Livinallongo and Gardena it is impossible to tell whether and in final syllables merged as or as before becoming a diphthong. In the discussion that follows, it is assumed that the coalescence was to .

¹⁰ Or is there? What about jeu ‘game’ vs. je joue ‘I play’ or lieu ‘place’ vs. je loue ‘1 rent’ (originally ‘I place’)? Following the Rhaeto-Romance rules, PItW and , the nouns, would have , while and , the verbs, would have . There is also the feminine noun roue ‘wheel’ , in which the umlaut is not to be expected. But the evidence is admittedly fragmentary and may not count for much. For traditional attempts at explanation, see Bourciez, Phonétique §66 III, §83 II.

¹¹ Cf. Charles F. Hockett's front rounded and back unrounded interpretations of Old English biliteral spellings in Lg. 35.575–98 (1959).

¹² Reflexes of PItW free . Later, in the course of attested Old French, some reflexes of PItW free /o/ coalesced with them.

¹³ That is, the indigenous dialects of northern Italy typified by Venetan.

¹⁴ Heilmann, Moena 72.

¹⁵ Stage x must be reconstructed with two different kinds of sibilant which French merged as /s/ or /z/ and Rhaeto-Romance as /š/ or /ž/. PRR /kréšt/ and OFr. ‘grows’ have stage x /š/ as shown by OFr. ; and PRR /kréšta/ and OFr. ‘crest’ have stage x /s/ as shown by OFr. lack of . Stage x /š/ is from PItW /sč č as in /péš/ ‘fish’ < /pésče/, /kró:š/ ‘cross’ < /króče/, /ré:š/ ‘king’ < /réğe/, or /lašá:r/ ‘to leave’ < . Stage x /s/ is from PItW /s/ as in /ná:s/ ‘nose’ < /násu/. Furthermore, there is no trace in any Rhaeto-Romance dialect of the prothetic front vowel developed in Proto-Italo-Western before Proto-Romance initial /s/ plus consonant: PRR /šté:la/ ‘star’ etc. These and a few other slight differences prevent the conclusion that Old French derives directly from Proto-Rhaeto-Romance.

¹⁶ As defined in Carlo Tagliavini, Le origini delle lingue neolatine ² 346 (Bologna, 1952), those Italian dialects above Spezia-Rimini with the exception of Venetan. These generally show vowel length and front rounded reflexes of PItW /u/ and ; reflexes of PItW single intervocalic /n/ are often ; and all final vowels except /a/ have been lost. A reconstruction of Proto-Gallo-Italian based on at least six local dialects should be carried out before a decision can be reached about the relationship between Gallo-Italian and Rhaeto-Romance. If they can be derived from a common source, it would be tempting to conclude that they all represent Bavaro-Romance (Rhaetian), shifted in locus as a result of the Germanic expansion of the fifth century, and not a pre-Roman Celtic community extending on both sides of the Alpine watershed. After its conquest, Rhaetia originally formed part of Gallia Belgica.

¹⁷ In order to derive Provençal from stock x, we must assume four steps.

1. (1)

   (1) After analogical leveling, was quite rare: ‘fire’ > but ‘heart’ > .

2. (2)

   (2) The central rounded /ü/ (long and short) regressed to the back series, and > to avoid coalescing with preexisting , /mü:r/ ‘wall’ > /mú:r/.

3. (3)

   (3) Phonemic length disappeared without trace and long and short vowels coalesced: /múir/ > /múr/, , etc.

4. (4)

   (4) Social factors caused [u] to be replaced by [y] : /múr/ > /mýr/. In time the /ø/-intersection in the new pattern came to be occupied by older plus palatal: 'leaf > mod. Arles . But /k/ was not a palatal, and remained unchanged.

All this may explain Aries only. In any case it is much ado to explain a complete lack of prima facie relationship with stock x. Much evidence from more South French dialects is needed. See Hélène N. Coustenoble, La phonétique du Provençal moderne en terre d'Arles (Hertford, 1945).

¹⁸ One hesitates to give stock x a name before its exact membership is determined. If Provençal should turn out to belong to it, ‘Proto-Gallo-Romance’ is the obvious choice; if not, perhaps ‘Proto-Northern-Romance’.