Autoregressive method for the spectral analysis of turbulent flow signal is discussed in this paper. When small samples are available owing to the record length or he limited memory she of experimental device or inherent characteristics of flow signal the AR method is more useful than the FFT method for spectral analysis, because its resolution is independent of the size of data samples. Especially, low frequency characteristic is much better analyzed by the AR method. To illustrate the advantages of die AR method, some typical signals often encountered in turbulent flows are analyzed using the AR method.