CNL Lunch

Kriszta Szendroi (Utrecht University)

Theoretical and Empirical Arguments for a Stress-Based Theory of Focus

Thursday October 31st, 12:30pm, 3416 Marie Mount Hall

Reinhart (1995) argued that focus is not unambiguously represented in the syntax, rather the prosodic representation determines the set of possible foci. This position is in contrast to what we may call the standard treatment, where focus is encoded in the syntactic representation by a syntactic [+Focus]-feature and/ or a corresponding function head, Focus.
There are a number of theoretical arguments supporting the stress-based theory. First, syntactic ëfocus constructionsí in languages like Hungarian always involve additional prosodic marking on the focus constituent. Second, languages have different ways to mark noncanonical focus: some allow noncanonical prosody (English), some allow noncanonical syntax (Hungarian). Three, sentences involving nested foci cannot be analysed in a feature-based theory.
We can also find experimental support. In English (or Dutch), when the stressed (focus) constituent is not the one carrying prominence by the nuclear stress rule, the two approaches make different predictions. The stress-based approach argues that these cases can be interpreted as examples of ëreference set economyí in the sense of Reinhart (1995, 1999). Reference set economy cases involve a heavy computational load, requiring keeping track of two alternative derivations and comparing them to context. Whenever such computation is involved, there should be some evidence of processing complexity. The feature-based approach makes the opposite prediction: since focus is determined by a syntactic device, [+F], without reference set computation, there should be no increase in processing load. In the case of young children, the increased processing load should have the effect that they cannot identify the focus correctly. I report two studies carried out on Dutch 4 and 5-year old children, one of which supports the stress-based theory and thus establishes similar results to Halbert et al (1995) and Gualmini et al (2002).