As part of a general study of the statistical characteristics of speech spectra, a number of characteristics relevant to talker identification are being investigated. The use of some statistical measures for this purpose has been described, in which mean values of filter outputs [P. B. Denes and M. V. Mathews, J. Acoust. Soc. Amer. 32, 4150 (1960)] and statistical manipulations of spectral features [S. Pruzansky, M. V. Mathews, and P. B. Britner, J. Acoust. Soc. Amer. 35, 1877 (1963)] were applied to word and talker recognition. Extensions of these methods are possible, such as evaluation of the temporal stability of the correlation matrix; characterization of patterns in the correlation matrix, particularly in the form of isocorrelation contours [K-P. Li, G. W. Hughes, and A. S. House, J. Acoust. Soc. Amer. (to be published)]; and extraction of eigenvector correlations and patterns. Results of some preliminary experiments utilizing these techniques are described and potential applications to speech-analyzing and talker-identifying schemes are discussed. [This work was supported in part by the Air Force Cambridge Research Laboratories, Office of Aerospace Research.]