This study compares the sensitivity of electrocardiographic R-R methods, used alone and with respiration, for diagnosis of cardiac disautonomy in diabetic patients. In 12 diabetic subjects with laboratory signs of neuropathy (45 +/- 10.4 years) (group A), 24 diabetic subjects without neuropathy (45 +/- 11.9 years) (group B) and 16 normal subjects (43 +/- 12.6 years) (group C), electrocardiogram and respiration were on-line digitized during spontaneous breathing and hyperpnea at 3-18 breaths/minute. Heart rate, R-R standard deviation (SD), R-R range (RG) and cross-correlation function (CC) were computed. Reproducibility was tested in 10 normal volunteers (age 22-28 years, mean 25.2). During spontaneous breathing, heart rate was 87.5 +/- 11.8 beats/min in group A, 77.3 +/- 10.8 in group B, and 71.6 +/- 14.9 in group C (p less than 0.05); RG was 102 +/- 51 msec in group A, 166 +/- 78 in group B and 272 +/- 168 in group C (p less than 0.005); SD was 18.2 +/- 10.1 msec in group A, 29.3 +/- 13.7 in group B and 49.2 +/- 26.5 in group C (p less than 0.001); CC was 2.38 +/- 0.66 units in group A, 3.00 +/- 0.72 in group B and 3.85 +/- 0.59 in group C (p less than 0.0001). During hyperpnea the difference between A and B increased for SD, RG (p less than 0.01) and CC (p less than 0.001). CC better discriminates between normals and diabetics without neuropathy during normal breathing; hyperpnea enhances the relationship heart rate-respiration in normal and, at a lesser degree, in diabetic subjects, thus slightly improving the method. The CC method had better reproducibility and lower intersubject variability than traditional methods of R-R variability.