In 11 anesthetized pigs, the left anterior descending coronary artery (LAD) was cannulated and pump perfused with blood before and during maximum adenosine vasodilation. For LAD plasma flows (F) ranging from 0.42 to 3.6 ml.min-1.g perfused tissue-1, we injected radiolabeled microspheres to measure heterogeneity and used the multiple indicator-dilution method to measure permeability-surface area product (PS) for EDTA. Heterogeneity of flow from the LAD was expressed as relative dispersion (RD) = standard deviation of flow/mean flow. Values of RD, corrected for tissue sample size using fractal theory, ranged from 13 to 87%, approaching 16-35% at high F. We developed a "variable-recruitment model" of regional heterogeneous capillary transport to correct PS for flow heterogeneity and capillary surface area recruitment. Values of PS ranged from 0.14 to 0.96 ml.min-1.g-1. Accounting for heterogeneity increased PS values by 0-18% compared with homogeneous values. Results revealed PS to be proportional to flow up to F = 1.5-2.1 ml.min-1.g-1 and then was constant at higher flows. The initial increase of PS with F may be due to capillary recruitment. When full recruitment is reached, PS becomes independent of F. We conclude that flow heterogeneity is significant but not readily predictable in the pig myocardium and that the use of microspheres to correct indicator-dilution data for flow heterogeneity improves the interpretation of multiple-tracer studies, particularly when tracers are used to study interventions that may alter flow distribution.