On the Role of Palatalization in the Vulgar Latin Sound Change /w/ > /β/

Our paper focuses on two convergent Vulgar Latin sound changes, i.e. the intervocalic fricativization of both the bilabial voiced stop phoneme /b/ and the labial velar semivowel phoneme /w/ to bilabial voiced fricative [β]. Regarding the latter, i.e. /w/ > [β], Stephens (1988) proposes that palatalization of /w/ promoted the fricative pronunciation [β] and tries to demonstrate, on the basis of spelling variation in Vulgar Latin inscriptions, that in word-internal position the fricative pronunciation was significantly more frequent in palatalizing (i.e. before front vowels) than in non-palatalizing environments (i.e. before back vowels). Stephens (1988: 430) states “there is further, very strong evidence in favor of the palatalization hypothesis”, since as for the V → B substitution he found a predominance of the position before high front vowels, i.e. /i(:)/ over the position before mid-front vowels, i.e. /e(:)/, which is perfectly in accordance with the hierarchy of palatalizing effectiveness (since high front vowels promote palatalization to a greater degree than mid front vowels). A revision of this palatalization hypothesis is, however, reasonable on more than one score. Firstly, in Stephens’ analysis all the rates of V → B substitution were calculated in proportion to the corresponding correct spellings, i.e. all frequency data were calculated according to the method of Barbarino (1978) – which is nowadays regarded outdated, since the involvement of cultural factors distorts and misrepresents the linguistic reality, see Adamik (2012: 128f) and Adams (2007: 626). Secondly, the analyzed predominance of the intervocalic environment before /i(:)/ (both short and long /i/) over the intervocalic environment before /e(:)/ (both short and long /e/) might be proved to be illusory and irrelevant, since Stephens’ data coming from Southern Italian Christian inscriptions reflect a developmental stage of Vulgar Latin vowel system where the originally short /i/, the originally long /e:/, and, in unstressed syllables, the originally short /e/ have already merged into a single phoneme, the closed /e/. Since in the Latin of Southern Italy (just like in that of most Romance areas except for Sardinia) only the originally long /i:/ remained as /i/, only the position before long /i:/ can be taken into account for establishing a hierarchy of palatalizing effectiveness between high and mid front vowels, while “the rarity of /i:/ and /e:/ in the sample precludes reliable estimation of the V → B rates before them” (Stephens 1988: 430). Thirdly, there is no relevant counter-argument against also involving in the survey the word-initial and post-consonantal spelling confusions between B and V (beside the intervocalic ones) and also the substitutions B → V (beside the items of V → B). Accordingly, in our paper we intend to test the palatalization hypothesis of Stephens by a distributional analysis of all types of B/V confusions with regard to the quality of the following vowel, first of all on data sets recorded (by Lupinu 2000 and/or LLDB) from Sardinia, as that is the only area where the phenomenon in question is explorable.