The problem-solving strategies used by 10 radiologists and 10 radiology residents during the interpretation of difficult mammograms were examined under two experimental conditions. In the "authentic" condition, standard unmarked mammograms were used. On a second set of the same cases for the "augmented" condition, mammographic findings were highlighted. Verbal protocols revealed that mammography interpretation predominantly used data-driven or mixed-strategies but depended on case typicality and clinical experience. Radiologists scanned the cases significantly faster than residents, but no group differences were found in the number of findings, observations, and diagnoses. The groups (1) used the same types of operators, control processes, and diagnostic plans; (2) committed the same number of errors; and (3) both groups committed case-dependent errors. Overall, the results indicated that mammogram interpretation is a well-constrained visual cognitive task. Results were applied to the design of a computer-based tutor for mammogram interpretation. (Author, DB)