Fifteen-four studies of 201T1 brain tumor SPECT were independently interpreted by 9 nuclear medicine physicians with and without reference magnetic resonance images in 2 separate sessions to define an effect of referring images, and inter-observer variations. The physicians were requested to detect foci of abnormal deposits, and to discriminate whether they were malignant or not according to 5-grade scaling of subjective diagnostic confidence. Receiver-operating characteristics (ROC) analysis was performed. Mean sensitivity for presence of lesions (SEP), and sensitivity and specificity for malignancy of 201T1 SPECT were 84, and 53 and 55%, which were changed to 94, and 74 and 55% after referring to the MR images. The SFP was significantly improved (p < 0.05), but sensitivity and specificity for malignancy, assessed by areas under the ROC curves, were not significantly improved. Inter-observer variation of the SFP was significantly reduced with addition of the referring MR images. Cerebral lobar localization of lesions by SPECT showed great inter-observer variations (true localization ranged from 30 to 68.4%). It is concluded that 201T1 brain tumor SPECT has moderate sensitivity and specificity for malignancy, which is not improved by addition of anatomical reference images, that additional MR images reduce inter-observer variation of confidence on lesion presence, and that SPECT localization of lesions has great inter-observer variations.