This paper (which went online at the journal site a year ago) has now officially appeared:
Michael Wagner (2012). A givenness illusion. Language and Cognitive Processes 27 (10). 1433–1458
*Constituents that encode information that is salient in the discourse or “given” are often prosodically reduced and remain unaccented. What is given and new is usually defined at the level of meaning: given expressions are those that refer to salient referents or predicates that have been made salient by the previous discourse. This paper presents evidence from two production studies that sometimes, a constituent that semantically should be contrastive, and hence accentable, is treated prosodically as if it was given, and placing an accent on it is consistently avoided—an illusory case of givenness. This effect can be explained by assuming that givenness is not only evaluated in terms of semantic content, but also at the phonological level. Prosodically marking a semantic contrast requires the presence of a phonological contrast. This effect thus provides evidence that the notion of “antecedent” relevant for prosodic givenness-marking needs to include reference to linguistic form, and not just to referential meaning.**