Abstract:
This paper presents production data testing the analysis of Arvaniti and Baltazani (2005) and Arvaniti and Ladd (2005) according to which the default melody used with Greek wh-questions is L*+H L- !H% (showing a delayed accentual peak on the utterance-initial wh-word, a low stretch, and a final curtailed rise), with !H% sometimes being replaced by L%. Here it was hypothesized that the melodies also differ in pitch accent and are used in different contexts. Four speakers, two male and two female, took part in reading a varied corpus of questions in contexts that lead to the use of a wh-question either in order to seek information or in order to politely register disagreement (a function of wh-questions peculiar to Greek). Our results confirmed that there are two different melodies: L*+H L- !H%, with a delayed accentual peak and a final rise, and L+H* L- L%, with an early peak and no final rise. The former is used for requesting information and the latter when questions function as dissenting statements. In addition to leading to a revision of the existing analysis, these results show that distinctions such as statement versus question are too coarse-grained for the analysis of intonational meaning and function.