This study attempts to model the perceptual similarity data of natural English voiceless fricative syllables in terms of the auditory distance metrics using multidimensional scaling technique (MDS). First, it was proved that the perceptual configuration of nonspeech sounds is adequately modelled by Euclidean distance space. Next, the three distance metrics were analysed to show which metric is most efficient in modelling the perception of speech sounds. Finally, it was shown that the perceptual and physical configurations of five voiceless English fricatives were highly correlated. This result seems to support a model of speech perception based mainly on the general physical characteristics of speech.