Objectives: To establish a method for the development of a mathematical model of autonomic activity in gastrointestinal movements and to basically evaluate of the application of the nonlinear analysis method to electrogastrography, we performed feature extraction of electrogastrographic changes in healthy elderly and gastrectomized subjects. Methods: The subjects consisted of 9 healthy elderly males and 3 elderly males without constipation who had undergone resection of 2/3 of the stomach. Electrogastrograms were obtained in a sitting position for 30 minutes and in a supine position for 90 minutes. Spectrum analyses of electrographic time series, the maximum Lyapunov exponent for the evaluation of the chaos of dynamic systems forming time series, and translation error for the evaluation of the smoothness of the attractor orbit were performed. Results: The maximum Lyapunov exponent was a positive number in all analysis intervals in all subjects. This suggests the irregularity of electrogastrograms in gastrectomized subjects. The translation error in the gastrectomized subjects was higher than that in the healthy elderly subjects, showing irregularity. However, as a result of spectrum analysis, gastric electric activity was predominant on electrogastrograms of the healthy elderly subjects, but intestinal electric activity was predominant in the gastrectomized subjects. Conclusions: Differentiation between healthy and gastrectomized elderly people is difficult using only one of the spectrum analysis methods, the maximum Lyapunov exponent, or the translation error. However, since the extracted features differed among the 3 analysis methods, differentiation and diagnosis may be possible using a combination of these methods.