In this paper, a technique is described for obtaining the information needed to perform attenuation correction in the thorax entirely from an emission study. This technique is based on the observation that the variation in soft tissue and lung attenuation coefficients is small among individuals. Thus only the outer contours of the body and lungs need be determined for obtaining the attenuation map. The contours are determined by using 99Tcm-macroaggregated albumin (MAA) to locate the lungs and an external source wrapped about the chest to locate the body boundary. Simulation studies were performed to investigate how errors in the presumed tissue attenuation coefficients affect the accuracy of the correction. Body and lung attenuation coefficients were varied from 20% less to 20% more than the coefficients used in the corrections. Over this range, there was less than a 15% alteration in the relative distribution of counts in the left ventricle. To test clinical feasibility, seven patients referred for clinical myocardial perfusion studies were scanned before and after the placement of the body source and the administration of 99Tcm-MAA. Reconstructed images from these studies showed clear demarcation of all body and lung boundaries. The presence of 99Tcm-MAA in the lungs had no significant effect on the clinical interpretation of the single photon emission computed tomographic (SPECT) studies. It is concluded that this technique is feasible for clinical application and that it offers important advantages over other current methods.