The effect in healthy elderly subjects of cycle ergometer or arm ergometer training on peak oxygen consumption (VO(2peak)) and ventilatory threshold (VT) was studied. The aim was to determine the benefit of each training modality on specific and cross exercise capacity. The cross-effect was also evaluated as an index of the central nature of the adaptive response to training. Twelve non-smoking healthy males (age: 67 +/- 5 year; body mass: 75 +/- 9 kg) were randomly divided in two age-matched groups of six, performing an arm cranking (ARM) or a cycloergometer (CYC) training (12-week, 30 min, 3 times/week), while a third group of 6 subjects (age: 73 +/- 4 year; body mass: 80 +/- 8 kg) performed no training (control, C). At baseline and following the intervention, subjects carried out an incremental test to exhaustion both on the ergometer on which they trained (specific test) and on the other ergometer (cross test). Respiratory variables were measured breath by breath and heart rate (HR) was recorded. Peak oxygen consumption (VO(2peak)), ventilation (VE(peak)), oxygen pulse (O2P(peak)) and heart rate (HR(peak)) were averaged over the last 10 s of exercise. Following training, while HR(peak) remained unchanged, significantly higher W(peak), VO(2peak), VE(peak) and O2P(peak) were obtained in both training groups, on both ergometers. The amplitude of the increase in W(peak), VO(2peak) and O2P(peak) was significantly higher for specific than for cross tests ( approximately 19% vs. approximately 8 % in CYC; approximately 22% vs. approximately 9% in ARM, P < 0.01) while the increase in same test condition was similar. No change was observed in the C group. The results indicate that aerobic training brought about with different muscle masses, produce similar improvements in maximal and submaximal exercise capacity. Roughly half of such improvements are specific to exercise mode, which suggests peripheral adaptations to training. The other half is non-specific since it influences also the alternative exercise modality, and is probably due to central adaptations.