An effect of clinical information on interpretation of radiographs has been assumed, on the basis of experimental studies. To investigate this effect in a real-life setting, test films were included in the dialy work load of readers who were unaware that a study was being carried out. Eight subtle but unambiguous abnormalities (three lung nodules, lobar collapse, lung cyst, rib destruction, dilated esophagus, congestive heart failure) were included on the test films. For each abnormality there were four readings with a suggestive and four with a nonsuggestive clinical history. The readers were radiology residents and all interpretations were reviewed an sometimes altered by staff radiologists. There was a statistically significant (p less than 0.01) increase in the rate of true-positive readings in the presence of a suggestive as compared to nonsuggestive history: 16%-72% for residents' readings, 38%-84% for combined resident-staff readings. There was some concomitant increase in false positives.