Athletic performance generally is thought to improve with increases in training load. However, few data exist showing the quantitative relationship between training load and performance. We followed 56 athletes (16 runners; 40 cyclists/speed skaters) during 12 weeks of training. We recorded index performances (3.2 km time trial or 5 or 10 km bicycle ergometry) after 6 weeks of baseline training and 6 weeks of a self-selected training increases. Training load was quantitated as the product of intensity (global rating of perceived exertion (RPE)) and the duration (time) of each training session. Load was expressed as the weekly average over the 6 weeks preceding each index performance. We also recorded the duration of high intensity training (RPE>5, hard) (inten). From 6 to 12 weeks, performance improved 12.95 +/- 3.83 to 12.66 +/- 3.00 min (p < .01). Training time (345 +/- 282 to 355 +/- 273 min/wk) and inten (61 +/- 88 to 71 +/- 91 min/wk) did not change significantly, although RPE (3.8 +/- 0.7 to 4.0 +/- 0.8) and load (1242 +/- 957 to 1386 +/- 978) increased significantly. No strong correlations existed between changes in performance and changes in any training measure (TIME, r = -0.031, RPE, r = -0.039, LOAD, r = 0.29, INTEN, r = 0.025.) Data suggest that improved performance in events of 7-20 minutes duration in response to intensified training is primarily dependent upon increases in total load and overall RPE during training and; that a 10-fold increase in training load may be associated with an approximately 10% improvement in performance. These data suggest the possibility of understanding the training responses of athletes on a quantitative basis.