Beat-to-beat heart rate (HR) and blood pressure were measured by the Finapres system in 44 healthy and 64 diabetic subjects in the at-rest condition. Autonomic control in diabetic subjects was assessed by the Ewing test. HR variability was explored by both linear and nonlinear methods. Linear methods used HR standard deviation and power spectrum. The percentage of the spectrum in the low frequencies was used to assess the sympathetic tone of the autonomic control. The nonlinear method used the "recurrence plot." This method explored long parallel subsequences in the HR time series. These sequences characterize the dependence of the HR dynamics on initial values. The HR standard deviation was reduced in the diabetic subjects compared with the healthy subjects (2.80 +/- 1.17 vs. 3.64 +/- 1.45 beats/min; P < 0.001). In the diabetic subjects, the HR standard deviation and the percentage of the spectrum in the low frequencies showed no correlations with the Ewing score (P > 0.10). In contrast, the longest length index was very strongly correlated to the Ewing score (r = -0.60; P < 0.0001). The results suggest that nonlinear methods might be powerful to explore the autonomic dysfunction in diabetic subjects.