Using positron emission tomography (PET) and intravenously injected (13)N(2), we assessed the topographical distribution of pulmonary perfusion (Q) and ventilation (V) in six healthy, spontaneously breathing subjects in the supine and prone position. In this technique, the intrapulmonary distribution of (13)N(2), measured during a short apnea, is proportional to regional Q. After resumption of breathing, regional specific alveolar V (sVA, ventilation per unit of alveolar gas volume) can be calculated from the tracer washout rate. The PET scanner imaged 15 contiguous, 6-mm-thick, slices of lung. Vertical gradients of Q and sVA were computed by linear regression, and spatial heterogeneity was assessed from the squared coefficient of variation (CV(2)). Both CV and CV were corrected for the estimated contribution of random imaging noise. We found that 1) both Q and V had vertical gradients favoring dependent lung regions, 2) vertical gradients were similar in the supine and prone position and explained, on average, 24% of Q heterogeneity and 8% of V heterogeneity, 3) CV was similar in the supine and prone position, and 4) CV was lower in the prone position. We conclude that, in recumbent, spontaneously breathing humans, 1) vertical gradients favoring dependent lung regions explain a significant fraction of heterogeneity, especially of Q, and 2) although Q does not seem to be systematically more homogeneous in the prone position, differences in individual behaviors may make the prone position advantageous, in terms of V-to-Q matching, in selected subjects.