Temporal distribution of interrogativity markers in Dutch : A perceptual study

A set of 128 different intonation patterns were synthesized on a single Dutch sentence with (potential) pitch accents on the subject (sentence-initial position) and on the object (sentencemedial position). The sizes of the accents were 0, 4, 8 and 12 semitones (st) in all 16 logical combinations. The accents were superposed on four different baseline slopes (B3, B1.5, 0 and +1.5 st/s), and utterances could or could not end in a terminal rise H% (0, 8 st). The 128 versions were presented to two groups of 20 native Dutch listeners. One group heard the 128 complete utterances and identified each version as either a statement or a question. The second group performed a gating task: they first heard (16) different versions that were truncated after the subject accent (gate 1), then (the same 16) with the truncation point delayed to the onset of the object accent (gate 2), next the (64) different versions with the truncation point at the end of the object accent (gate 3), and finally the (64) versions with truncation point at the onset of the (potential) terminal rise (gate 4). On hearing each gated stimulus listeners guessed whether they had heard the beginning of a statement or of a question. The results of both experiments were analysed in order to reveal the temporal development of cues that signal statement versus declarative question in Dutch. Sentence type is clearly signalled well before the terminal rise is heard. In terms of the experimental variables, the overall F0 trend (slope of baseline) is the most important early cue, followed by the size of the object accent. All early cues were overruled by the information in the terminal boundary. However, the percept of statement vs. question suffered significantly when early cues were contradictory to the absence versus presence of a terminal rise. The results bear on the issue whether the signalling of sentence type can still be modelled by a sequential, linear tone model or whether some global tone shape feature needs to be involved.