The specificity of aerobic training for upper-body exercise requiring differing amounts of muscle mass was evaluated in 25 college-aged male recreational swimmers who were randomly assigned to either a non-training control group (N = 9), a 10-wk swim(S)-training group (N = 9), or a group that trained with a standard swim-bench pulley system (SB; N = 7). For all subjects prior to training, tethered-swimming peak VO2 averaged 19% below treadmill values (P less than 0.01), while SB-ergometry peak VO2 was 50% and 39% below running and swimming values, respectively (P less than 0.01). Significant (P less than 0.01) increases of peak VO2 in tethered swimming (11%) and SB (21%) were observed for the SB-trained group, while the S-trained group improved (P less than 0.01) 18% and 19% on the tethered swimming and SB tests, respectively. No changes were observed during treadmill running, and the control subjects remained unchanged on all measures. Comparisons between training groups indicated that although both groups improved to a similar extent when measured on the swim bench, the 0.53 l X min-1 improvement in tethered-swimming peak VO2 for the S-trained group was greater (P less than 0.05) than the 0.32 l X min-1 increase noted for the SB-trained group. The comparisons between SB and S exercise vs treadmill exercise support the specificity of aerobic improvement with training and suggest that local adaptations contribute significantly to improvements in peak VO2. Furthermore, the present data indicate that SB exercise activates a considerable portion of the musculature involved in swimming, and that aerobic improvements with SB training are directly transferred to swimming.