The use of a Maximum Entropy method for the analysis of short‐time biomredical rhythms is described using examples taken from gastro‐intestinal electrical slow‐waves, human digit blood flow, rat locomotor activity and psychological mood data. The method is similar to auto‐regressive modelling in that it estimates coefficients in a difference equation model which are related to spectral components. Fine spectral resolution of short time‐series is obtained in the Maximum Entropy method by predictive extrapolation of the time‐series, unlike the methods of Fast Fourier Transforms and Periodogram via Autocorrelation Functions. It is shown that Maximum Entropy is superior to normal auto‐regressive analysis for signals having a higher noise content and that problems caused by initial phase‐shift are no more severe than for the zero‐extended FFT analysis and normal auto‐regressive techniques.