Experimental studies have been made of the lesion detectability in myocardial perfusion studies using thallium-201. A series of images (AP-view) was generated using a convolution of a mathematical model of the left ventricular myocardium and an experimentally determined point spread function. Background was added. Images were simulated with 100 k, 200 k, and 300 k counts for the complete image. Each image contained a lesion with either 0% or 50% of the normal tracer concentration. All images were interpreted by five experienced observers, independently of each other. Their interpretations were analysed using the Kolmogorov-Smirnov two sample test. The true positive fraction (TPF) was hardly affected by changes in count density. The TPF decreased significantly if the lesion tracer concentration changed from 0% to 50% of the normal myocardial tracer concentration. The decrease was independent of the count density. The false positive fraction (FPF) decreased significantly if the count density increased; no difference in FPF was found for a change in lesion tracer concentration. In addition, plots were generated with the TPF versus lesion volume. The TPF was lowest for locations far from the camera; FPF was high for these segments. Increase of the count density mainly improved the FPF in these segments. Small lesion were more difficult to detect.