An analysis of long-time-average-spectra of twentytwo quality-rated violins

LongTime-Average-Spectra (L TAS:es) were recorded of 22 qualityrated violins. The LTAS:es were analyzed by four different methods: weight functions, factor analysis (FA), multidimensional scaling (MDS), and separate correlation analysis. The average difference between the instruments rated highest and lowest was t r ied a s a function weighting tonal quality. This weight function explained 64 70 of the variance of the tonal quality-ratings. Factor analysis and multidimensional scaling in five factor s/dimensions gave approximately the same solutions. The solutions accounted for 74 70 and 44 ' $0 respec t ive ly of the variance of the LTAS:es and for 74 70 and 69 70 respectively of the variance of the tonal quality-ratings. The variations in certain res t r ic ted frequency regions selected by correlation analysis accounted for 7 1-84 70 of the variance in the tonal quality-ratings. The different methods imply that "strong" f r e quency components a r e favorable in a low frequency region and in a middle high frequency region, while "weak" frequency components a r e favorable in a high frequency region and in a l imited middle frequency region. The resu l t s seem reliable, a t l eas t for the selected instruments.