The loudspeaker is considered most important part in audio systems, because the loudspeaker is the final electro-mechanical device in the chain of sound-reproduction systems. Traditionally, acoustic performance of loudspeakers has been eva1uated by using the listening test by juries. However, this subjective evaluation method has many problems in the viewpoint of reliability and repeatability as well as time and economy, that objective techniques have been tried to evaluate the acoustic performance. In this study, a statistically sound and complete method in extracting the information from the jury test results is pursued and the subjective rating from listening tests is compared with the objective measured data in order to obtain the correlation between two ratings. For the statistical analysis, the analysis of variance (ANOVA) and Tukey's method are employed for dealing with the data from the listening test. For the objective performance evaluation of loudspeaker sounds, Zwicker's loudness that counts for the human hearing characteristics is calculated for the measured sound signal emitted from each loudspeaker. Two mid- and high-frequency related feelings such as fidelity and softness are dealt with and the evaluation method for a single figure of merit for each impression is established in terms of the loudness. Then, the two subjective impressions are compared with the calculated objective performance indices. It is noted that the correlation coefficients relating fidelity and softness impressions with objective indices are about 0.65 and 0.8, that means high correlation between subjective feeling and loudness-based objective data. It is also found that the physical parameters of the loudspeaker diaphragm such as Young's modulus, loss factor, longitudinal wave speed are not correlated very well with aforementioned two subjective impressions. However, this conclusion is under given data range restricted in this study and further research on this matter should be done. The results of this study may be useful in evaluating commercial or prototype loudspeakers without resorting to very time-consuming and expensive subjective testing.