A perceptual magnet effect in pitch accents: parametric modelling
Summary
This thesis presents the results of two experiments that suggest the presence of a Perceptual Magnet Effect (PME, Kuhl 1991) in pitch accents (linguistically significant fundamental frequency (F0) movements). The perceived goodness and discriminability of re-synthesised Dutch L*H rise contours were evaluated by naive participants. The presence of a PME is strong evidence of category-internal structure, implying that perception and mental representation of the feature in question may be exemplar-theoretic. PME has not previously been conclusively found in pitch accents, but has been found in boundary tones, a related intonational event (Schneider 2012).
“Parametric” modelling of speech melody, such as the CoPaSul model (Reichel 2011), involves using mathematical techniques to stylise the movements in the fundamental frequency of speech, extracting parameters describing the shape of the contour. I quantified pitch accent variation in terms of these polynomial parameters, in place of the more traditional “measurement” approach where, for instance, excursion size, peak alignment and pitch register are used to quantify pitch variation. Using this polynomial-parametric approach to calculate the perceptual distance between different stimuli, PME was detected: (1) rated “goodness” decreased as perceptual distance increased, and (2) equally spaced items far from the prototype were less frequently generalised than equally spaced items in the neighbourhood of the prototype. The dominant Autosegmental Metrical framework of intonational phonology (Pierrehumbert 1980) makes strong predictions (the Free Gradient Variability hypothesis) about the kind of speech melody variation that should be linguistically salient, specifically that pitch register and excursion size are paralinguistically salient, rather than linguistically salient. I found evidence that variation in these dimensions does in fact result in differences in goodness ratings, suggesting that this variation is perceived as linguistically salient by Dutch listeners.