The heterogeneity of pulmonary blood flow was examined using a fractal analytic procedure, and the results were compared with the traditional gravitational model of flow distribution. 99mTc-labeled macroaggregate was injected intravenously at functional residual capacity in six supine anesthetized dogs. The lungs were fixed in situ and sliced in transverse sections. The slices were imaged on a planar gamma camera, and a three-dimensional array of blood flow measurements was reconstructed for each lung. Fractal analysis was used to examine the spatial heterogeneity or RDs (relative dispersion = SD/mean) as a function of the number of pieces into which the flow array was subdivided. RDs was fractal and could be characterized by a fractal dimension (Ds) of 1.09 +/- 0.02, where a Ds of 1.0 reflects homogeneous flow and 1.5 indicates a random flow distribution. The data fit the fractal model exceptionally well with an average r = 0.98. RDs was examined in gravitational and isogravitational planes and as expected was greatest in the gravitational direction. However, the difference was small, suggesting that gravitation plays a secondary role to an underlying process producing heterogeneity. Within the limits of resolution attained by this study (piece volumes greater than 0.25 cm3), the heterogeneity of pulmonary blood flow is well characterized by a fractal model.