25 subjects volunteered to document circadian changes in serum magnesium. 4 groups were formed: 7 healthy young males (24.0 years +/- 3.9), 6 elderly males (82.5 years +/- 7.5), 6 elderly females 81.2 years +/- 10.7) and 6 elderly insame subjects of both sexes (80.5 years +/- 8.6). They were socially synchronized with a diurnal activity (07.00 to 21.00 for the old subjects; 07.00 to 23.00 for the young ones) and nocturnal rest. The subjects followed a spontaneous diet. Venous blood was sampled at 4-h intervals and fixed clock hours (07.45, 11.45, 15.45, 19.45, 23.45, 03.45) during 24 h. The single cosinor method was used for the statistical analysis of the time series. A statistically significant circadian rhythm is detected in three of the groups: young males, elderly males and elderly females (no rhythm detection in elderly insane subjects). The 24-h mean is higher in elderly subjects than in the young one. The rhythm amplitude is larger in elderly males than in young ones. The acrophase (peak time) location in the 24-h scale is 10.12 h for elderly females, 11.35 h for elderly males and 16.36 h for young males.