All chest radiographs of 107 proven lung cancer patients who received consecutive biennial chest radiography were reviewed to elucidate problems detecting their cancers, and diagnosing them when initially radiographically detected. Subjects, members ofa fixed population sample, originally numbered 20000 persons, 17000 of whom consistently received consecutive biennial chest radiography during examinations for late effects of atomic-bomb radiation. Among the 107 subjects, 64 had radiographic manifestations of cancer; 47 were initially correctly diagnosed; 17 were not. Eleven of the 17 were initially equivocal, diagnosable only after subsequent radiography and retrospective review of serial radiographs. Diagnostic problems consisted of I) six detection errors with cancer images superimposed on musculoskeletal and cardiovascular structures, reducible by stereoscopic p.a. instead of single p.a. radiography; immediate tentative interpretations; and by comparing earlier with current radiographs. 2) Eight decision errors, wherein cancers mimicked other diseases, were reducible by greater index of suspicion and scrutiny during interpretations.