Response rates to various drug regimens are determined in multi-center trials by radiographic measurement of tumor dimensions before and after treatment. The way in which lesions are measured may affect the reported results. Lesions due to primary lung cancer may be more difficult to assess than rounded peripheral metastases. No previous study has assessed the variability of tumor measurements in patients with primary lung cancer. Nineteen radiographs from 12 patients with non-small-cell lung cancer (NSCLC) were shown to 25 experienced readers on two separate occasions. Pre- and post-treatment (combined chemotherapy and radiotherapy) radiographs were first shown separately, unidentified and in random order. Three months later, they were shown side by side with the facility for comparison. A decrease in tumor size of greater than or equal to 50% was considered as a "response." Viewing paired films together increased the viewers ability to measure tumors compared with reading each film of the radiographic pair individually. Fifty-one percent of unmeasurable pairs became measurable when read together, whereas 13% of previously measurable pairs now became unmeasurable. The interobserver variation in measurements was 10% to 265%, while the intraobserver variation was 4% to 10%. The correlation between responses calculated on the first and second readings often was no better than chance alone. Subjective impressions of degree of improvement showed no good correlation with response rates measured objectively. Because of the large variation in measurements and consequently in response rate, we recommend that all radiographs be reviewed by a single observer for any particular study, and pre- and post-treatment radiographs be evaluated together.