Abstract
Common variability patterns for timing, articulation, and loudness were extracted by means of principal component analysis from music performances intended to express anger, fear, happiness, or sadness. Synthetic performances were generated with either timing, articulation, loudness, or no variability, which were rated by 10 musically experienced listeners on 10 adjectives, including the original emotions. Correlations were found between the ratings and two mathematical properties of the patterns, namely fractal dimension and durational contrast. The results suggest that both these properties describe relevant characteristics of the variability patterns, and that they play a role in emotional expression. Different roles for these properties in timing, articulation, and loudness variability were indicated. These findings may facilitate comparisons between different expressive domains, such as music, dance, speech, and body motion.