Numerous symptoms have been associated with the overtraining syndrome (OT), including changes in autonomic function. Heart rate variability (HRV) provides non-invasive data about the autonomic regulation of heart rate in real-life conditions. The aims of the study were to: (i) characterize the HRV profile of seven athletes (OA) diagnosed as suffering of OT, compared with eight healthy sedentary (C) and eight trained (T) subjects during supine rest and 60 degrees upright, and (ii) compare the traditional time- and frequency-domain analysis assessment of HRV with the non-linear Poincare plot analysis. In the latter each R-R interval is plotted as a function of the previous one, and the standard deviations of the instantaneous (SD1) and long-term R-R interval variability are calculated. Total power was higher in T than in C and OA both in supine (1158 +/- 1137, 6092 +/- 3554 and 2970 +/- 2947 ms2 for C, T and OA, respectively) and in upright (640 +/- 499, 1814 +/- 806 and 1092 +/- 712 ms2 for C, T and OA, respectively; P<0.05) positions. In supine position, indicators of parasympathetic activity to the sinus node were higher in T compared with C and OA (high-frequency power: 419.1 +/- 381.2, 1105.3 +/- 781.4 and 463.7 +/- 715.8 ms2 for C, T and OA, respectively; P<0.05; SD1: 29.5 +/- 18.5, 75.2 +/- 17.2 and 37.6 +/- 27.5 for C, T and OA, respectively; P<0.05). OA had a marked predominance of sympathetic activity regardless of the position (LF/HF were 0.47 +/- 0.35, 0.47 +/- 0.50 and 3.96 +/- 5.71 in supine position for C, T and OA, respectively, and 2.09 +/- 2.17, 7.22 +/- 6.82 and 12.04 +/- 10.36 in upright position for C, T and OA, respectively). The changes in HRV indexes induced by the upright posture were greater in T than in OA. The shape of the Poincare plots allowed the distinction between the three groups, with wide and narrow shapes in T and OA, respectively, compared with C. As Poincare plot parameters are easy to compute and associated with the \'width\' of the scatter gram, they corroborate the traditional time- and frequency-domain analysis. We suggest that they could be used to indicate fatigue and/or prevent OT.