Rate dependent spectral reduction for voiceless fricatives

The relationship between local speaking rate and spectral reduction of voiceless fricatives is examined based on a corpus of spontaneous speech. The spectral balance of about 18k items is statistically analyzed with regard to consonant duration and local speaking rate. Both measures show comparable effects: For most of the treated allophones, significant within-speaker reduction occurs with higher rates. The speaking style of the material is constant and stress, part of speech, and position in the syllable was controlled. However, as the observed effects are small compared to differences between allophones, perceptional relevance is considered to be implausible.