An examination of reader variability and bias among x-ray readers used in surveillance programs for underground coal miners was performed. These surveillance programs, which were mandated by the Coal Mine Health and Safety Act of 1969, were directed toward the prevention of pneumoconiosis in coal miners. An analysis was conducted using 14,883 x-ray records from miners with 10 or more years of underground mining experience. Two types of x-ray readers, A-readers and B-readers, were considered in this study. The difference between them was the qualifying exam that they passed to become an x-ray reader. A series of two readings, first and second round, have been examined. In either round, the reader could have been an A-reader or a B-reader. Overall results demonstrated that agreement between first and second round reader interpretations occurred 78 percent of the time for the 14,883 x-rays analyzed. A more detailed data analysis revealed that although 78 percent agreement was achieved overall, most agreements occurred for the preemployment screenings or when no abnormality was detected. It was observed that the first round readers reported more abnormalities. Comparisons were also made based upon location (including six states), reader classification, and or number of films read. Only reader classification demonstrated any notable differences and these were observed between A-readers and B-readers in the first round. Inter reader variability was large in both first and second round groups considering either group of certified readers. The authors suggest that information regarding reader level history become available to researchers since many researchers use the surveillance program x-rays, and their study results are dependent upon reader interpretations.