Accentazione, deaccentazione e struttura informativa

The commonly held view that deaccenting (the absence of a pitch accent on a word that might otherwise be expected to be accented) may be used to signal ''given''information or information that can otherwise be expected in the discourse has been recently challenged. Despite plenty of linguistic and psycholinguistic evidence for a close correspondence between information status and (de-)accentuation in Germanic languages (e.g [19], [12], [13], [1], [6], [9], [15]), few studies have shown that a previous mention in the discourse does not suffice to trigger deaccentuation nor in controlled experimental settings (cfr. [17]) nor in spontaneous dialogue ([3]).
Stronger evidence against deaccentuation comes from Romance languages ([4], [5]). Italian has been shown to deaccent clauses and simple NPs ([2], [11], [7], [8]) but to avoid deaccenting within NPs ([16]). On the other hand, Avesani [1]) showed that in samples of broadcast speech, items that have been previously mentioned in the same discourse segment can be accented again irrespective of grammatical function and surface position in the sentence.
In this study we examined word tokens repetedly mentioning the same entity within a task-oriented dialogue. Each co-referential repetition (294 items) was coded according to its syntactic category, grammatical function, position in the utterance and in the prosodic phrase, focus status (contrastive vs. non contrastive) and information status. Each mention was scored as given or new with respect to the hearer and to the discourse ([14]) and with respect to its belonging to the same or to a different discourse segment ([10]).
Each utterance containing a co-referential repetition was then perceptually analysed to determine the location of pitch accents, ToBI transcribed, and acoustically measured.
Results show that the vast majority of the repeted mentions are accented irrespective of their being hearer-, discourse- or discourse segment old. Change of grammatical function with respect to the previous mention, or utterance position do not show any effect on deaccenting.
In the same vein, preliminary results for one of the speakers show and that the duration of the accented words is not affected by their syntactic category, their functional role, by the change of their functional role, nor by their informational status.
Deaccenting occurs in less than 5% of the given items, and it appears to affect only postfocal items or items occurring initially in a complex NP.

Publication type: 
Contributo in volume
Author or Creator: 
Avesani C.
Vayra M
Banelli D.
Lovisolo M.G
Misura di Parametri. Aspetti tecnologici ed implicazioni nei modelli linguistici, edited by Cosi P., pp. 287–312, 2005
Resource Identifier:
ISTC Author: 
Cinzia Avesani's picture
Real name: