The subtraction method for radioimmunodetection (RID) of cancer is showing potential as a diagnostic tool. An objective assessment of this technique is needed in order to determine its clinical usefulness. A numerical analysis of the RID scintigrams was developed and applied to 21 RID studies on 20 patients for whom a final diagnosis had been established. The subtraction technique was shown to give rise to artifactual positive regions, which could be avoided by experienced clinicians. When artifact-producing regions were excluded, 97 regions were examined (80 nontumor, 17 tumor-containing). The "background" scintigrams were shown to be good models for nontumor regions of the antibody scintigrams, and tumor and nontumor regions were shown to form statistically distinct groups.