The Reduction and Fall of Gallo-Romance vowels : synchronic and diachronic functioning of the Merovingian Charters (7th - 8th centuries)

In my thesis I attempt to mend a paradigmatic fissure between historical phonologists and historians of the French language. While generations of historical linguists have painstakingly described the transition of Latin to French with neogrammarian-type sound-change rules, data from the Merovingian period has rarely been employed in a systematic way.

The key issue here is the reconfiguration of the vowel system. Stressed vowels are strengthened in diachrony, even undergoing diphthongisation while most unstressed vowels end up disappearing. The question is how? Here we argue that vowel loss was preceded by a typologically well attested form of vowel reduction: once length ceased to be contrastive in Late Latin further reduction occurred through loss of aperture contrasts resulting in three contrastive weak vowels which we can reconstruct as /ɪ/, /ʊ/ and /ɐ/. While /ɐ/ remained quite stable, giving the schwa of Old French, /ɪ/ and /ʊ/ were much less stable, alternating with ∅. After a period of synchronic variation, the syncopates/apocopated forms were lexicalised leading to vowel loss in diachrony and the proto-typical shape of most Gallo-Romance words. In certain support environments, the reduced vowels were structurally preserved but further neutralised, merging as /ə/.

Furthermore, the dating of syncope and apocope have implications for the relative and absolute chronology of other phenomenon such as lenition and diphthongisation, and because apocope obliterated most of the final case and gender-bearing vowels, both phenomena and their dating are critical in understanding how spoken Gallo-Romance stands in relation to written Latin and the rest of the Romance diasystem.

The central part of this thesis comprises a philological survey of 48 original charters dating from the 7th and early 8th centuries. We observe the type and calculate the frequency of graphic variation on a lexemic and morphological basis. Having established clear graphophonemic correspondences we employ element theory to demonstrate that the reduced number of vowel phonemes found in unstressed syllables can be explained by both typologically attested and theory-predicted forms of vowel reduction. Aperture contrasts are limited, then lost entirely leading to three reduced vowels |I|, |U|, |A| which may in turn undergo colouring loss resulting in a neutral schwa as is the case in Old French. Synchronic vowel alternations, which in auto-segmental CVCV phonology are parameterized as the extension of government from a phonologically more prominent position, are then modelized for Gallo-Romance. We argue that representationally weak “reduced” vowels are synchronically targeted for delinking. Diachronic syncope and apocope are then modelized as an inter-generational change after a period of representational variation and underspecification.

Counter to most deductive analyses of this diachronic problem, our inductive analyses imply that both vowel reduction and alternation were synchronic phenomena of the Merovingian Period, and that Merovingian Latin does in fact represent an acrolectal written variety of the spoken language.

These conclusions throw significant doubt on the traditional dating of major Gallo-Romance sound changes but open news paths to understanding the break-up of Latin into its daughter languages.