In this paper we examine different approaches to obtain judgments of perceptual prominence. We discuss different prominence scales, the influence of the linguistic unit on which prominence is rated and the normalisation of prominence judgments. We propose the use of a multilevel scale for obtaining prominence judgments. It seems that naive listeners can rate word prominence better than syllable prominence, resulting in better correlations to acoustics. It is shown that normalisation should be applied to the obtained ratings.
Index Terms: prosody, prominence, methods, normalisation, acoustic correlates