The effect of respiratory motion on pulmonary activity determinations by positron emission tomography (PET) was studied in dogs with experimentally created pulmonary emboli (PE). The location of the PE was evaluated by planar 99mTc lung imaging to determine the appropriate sites for transaxial PET scans. PET scans of the lung then were obtained after i.v. injection of 68Ga-labeled microspheres. PET scans were acquired during slow (15 breaths/min) and fast (30 breaths/min) breathing with the same minute ventilation and then postmortem. Lung perfusion patterns were documented by i.v. injection of India ink before sacrifice. Cross sections of the excised lungs were made at the same levels as the PET scans, and eight sections containing 14 perfusion defects were analyzed. The scans obtained during slow breathing consistently showed edge blurring and demonstrated defects less well than scans obtained during fast breathing or postmortem. The normal-to-defect activity ratios during fast breathing and on the postmortem studies were similar and approximately 17% higher (P less than .01) than in scans obtained in the same animals during slow breathing. The results demonstrate the need for motion correction during quantitative analysis of regional lung activity by positron tomography, and suggest that high-frequency respiration at small tidal volumes may be one means for obtaining this correction.