Our purpose was to determine whether changes in myocardial capillarity underlie the exercise training-induced increases in coronary transport capacity previously observed in dogs (J. Appl. Physiol. 58: 468-476, 1985). The approach was to measure capillary diffusion capacity (PS) in working hearts and then measure capillary numerical density (CD), capillary surface area density (CSA), and capillary volume density (CV) in specimens from perfused-fixed hearts. Eight dogs (20-30 kg) were exercise trained (ET) for 12-18 wk and compared with a group of seven control dogs. PS for 51Cr-labeled ethylenediaminetetraacetic acid was determined during maximal adenosine coronary vasodilation with perfusion pressures equal to 100 mmHg in both groups. The trained dogs' maximal PS averaged 58 +/- 10 ml.min-1.100 g-1, which was significantly greater than the control value (31 +/- 6). Maximal PS was linearly related to CV (r = 0.61) and CSA (r = 0.78) in the ET group. However, there was no difference between control and trained average left ventricular CD, CSA, CV, or intercapillary distance. The data indicate that although coronary blood flow capacity and capillary transport capacity may be improved in exercise-trained dog hearts, these changes are not the result of an increase in myocardial capillarity. Rather, the increased maximal PS appears to be due to changes in the determinants of capillary blood flow and/or the relationship between capillary area available for exchange and capillary perfusion.