SPECT images are degraded by the inclusion of Compton-scattered photons within the pulse-height window. Phantom and patient studies with Tc-99m were used to evaluate a compensation method that consists of subtracting a fraction of the image reconstructed using events recorded within a secondary pulse-height window (92-125 keV) from that derived from the photopeak pulse-height window (127-153 keV). Images of line sources in air and in a water-filled phantom were stored. The compensated line spread functions (LSFs) were evaluated. In water, the absolute counting rates for the SPECT LSFs were within 10% of the rates measured in air. The phantom consisted of six solid acrylic spheres (diameters 10, 13, 16, 19, 25, 32 mm) placed within a cylindrical (22 cm diam) distribution of Tc-99m. For sphere diameters greater than 25 mm, the measured image contrasts were within 8% of the true uptake ratios. Our results have shown that high-quality, camera-based SPECT systems can reconstruct artifact-free images by making additional use of projection data acquired in a pulse-height window other than that over the primary photopeak. This compensation method results in qualitative and quantitative improvements for the limited source geometries investigated. Further studies are required to optimize this heuristic approach for other source geometries.