Project Summary (2nd funding phase 2009-2012)

The projects goal is to advance our understanding of the role that prosody plays in online sentence comprehension (parsing) in German and Hindi.

The first phase of the project revealed that German and Hindi differ significantly as regards sentence prosody: unlike German, prosodic phrasing does not appear to reflect syntactic structure in Hindi. This and other related findings raise three questions that we will address:

(i) do prosodic correlates other than phrasing reflect syntactic structure,

(ii) what role do phrasing, accentuation, constituent length and rhythm play in parsing decisions in German versus Hindi, and

(iii) how can we develop a model of parsing that considers language-independent prosodic cues but allows us to model language-specific use of prosody?

The experimental component will involve online comprehension and production methods. The broader significance of this project is that it is the first systematic study of Hindi prosody which provides the first systematic comparison of Hindi and German prosody in parsing.

Project Summary (1st funding phase 2006-2009)

Prosody in Parsing

In this project, we will study the processing of prosody in its relationship with phonetics, syntax and psycholinguistics. In the past, research on speech processing has concentrated mainly on word order, argument structure, syntactic attachment and segmental phonology. The time is ripe to investigate less understood elements of grammar in more detail, and prosody is an ideal candidate for such an undertaking.

Our general assumption is that prosody feeds parsing, in the same way that other components of grammar do. If syntax and/or semantics is ambiguous, the prosodic structure may provide crucial cues for the decision between equally plausible options. For this reason, and given that spoken speech is replete with locally or globally ambiguous structures, prosodic phrasing can be considered the immediate underlying cause of highly efficient speech processing.

The ultimate goal of the project is to either integrate prosody into existing models of speech processing or, if this goal turns out to be unrealistic, to elaborate a model of speech processing with a strong prosodic component.

Examining the perception of prosody requires a good understanding of its production. Thus we will complement our studies on processing by an examination of production in areas which are relevant for our concerns:
Formation of prosodic domains, pitch excursions, realizations of phrasal boundaries, etc. These studies will address the role of acoustically realized prosody, like changes of phrasing and displacements of pitch accents.

We will conduct a series of experiments to study the interaction of prosodic phrasing with syntax and other grammatical components. Some of them will target 'silent' prosody.
We hypothesize that a default prosodic structure is projected onto silently read utterances, except when the context provides cues for a more marked one.
Garden-path sentences arise in silent reading because unmarked prosody and syntax are first projected on the beginning of a sentence which turn out to be incompatible with the rest of the sentence. This results in a cognitively costly reanalysis of prosody and syntax.
The same holds for other kinds of sentences which have been shown to receive an increased reading time or reduced acceptability.

In order to disentangle the role of language dependent factors and universal factors, we will ground our project on a comparison between at least German and Hindi, two languages with different word orders and different prosodic structures. Ideally, our research will extend to other languages in the second phase of the project.