Séminaire du Groupe Transversal Prosodie : Carlos Gussenhoven

Limburgish is spoken in the north-western 12% of the area in which the Franconian tone is spoken, where Cologne was the medieval centre of prestige. It was there, after the beginning of the construction of the cathedral in 1248, that a remarkable lexical tone contrast arose from the contact between the hundreds of highly qualified immigrant craftsmen and thousands of local workmen whose employment was related to the building enterprise. With hindsight, that tonogenesis can be seen as an experiment in pushing the conditions for L1 language acquisition to its limits: the tonogenesis occurred in declarative intonations, yet in principle the tone contrast needed to be available under all intonational conditions.

Unsurprisingly, the embedding of the tone contrast in the intonation grammar led to a wide variety of solutions, causing the dialects to vary in the number of intonation melodies (from 1 to 4) as well as in the pitch shapes of the tones. When focusing on individual dialects, a unique typological feature becomes apparent:

It is not possible to specify the pitch shapes of the tones: These vary with intonational conditions.

Importantly, despite the high pressure that must have been placed on infants to create grammars for what must be seen as implausibly varied set of phonetic forms, the grammars of the dialects known to date all conform to the model of Pierrehumbert (1980) and the assumptions of OT. The Limburgish data offer three confirmations of these models:

1. The star in T* is an instruction to associate with a *-marked TBU.
2. Edge-aligned tones align their (R/L) edge with the (R/L) edge of some other phonological constituent.
3. OT constraints are ranked.

With no basis in those models, other work added the stipulations for (a) that T* must always associate, for (b) that alignment implies association, and for (c) that some rankings are ungrammatical. The Limburgish data prove those stipulations wrong.

Today, the survival of the tone contrast may depend on the function of Acc2 as a local identity marker. In fact, this may be one route for tonoexodus. neutralizations would appear to occur through the generalization of Acc2. Another route lies in reducing the salience of the contrast: one case will be presented in which the tone contrast does exist, but is phonetically too subtle for outsiders to hear as anything than repetitions of the same unremarkable standard Dutch intonation contour. A third way out is through reinterpretation of the tone contrast in terms of consonants, vowels or vowel quantity. This will be illustrated on the basis of the Weert dialect.