In this paper, we report the findings that the action potential trains transmitted along the nerve fibers are encoded not only periodically but also chaotically. First, spontaneous action potentials along a single fiber of injured sciatic nerves in the., anesthetized rat were recorded. Then, the data were divided into two groups and analyzed with different methods. Phase space: representation, spectral analysis and the calculation of correlation dimension were used for the first group of data sampled with constant frequency. Due to the serious influence of the measurement noise, no reliable conclusion can be drawn from them. For the second group of data of the interspike intervals (ISI) which seem to convey more rich and important information. nonlinear forecasting method, the surrogate data and the plot of ISI (n+1) vs. ISI (n) were used in the analysis, good results have been obtained which confirm with those from the/3-cell model. The largest Lyapunov exponent (LLE) were calculated not only to further support our findings of chaos but also to quantitatively determine the degree of chaos.