Daniel B. Kopans, Dean McKnight, Jack E. Meyer, Myron Moskowitz, Paul Stomper, and Clifford J. Turner. Drs. Moskowitz and Bartrum very kindly provided mammograms and associated case data from their studies at the University of Cincinnati and Dartmouth-Hitchcock Medical Centers. The authors are also indebted to Allison Hagerman for serving as reading-test administrator and technical assistant throughout the project; to Barbara Freeman for writing the computer programs required for data collection; and to A. W. F. Huggins for assisting with data analyses. This paper benefited from discussions following verbal presentations at a Conference on Visual Search sponsored by the Committee on Vision of the National Research Council, a Radiology Grand Rounds of the University of Massa- chusetts Medical Center, and a seminar and a workshop arranged by the Depart- ment of Health Care Policy of the Harvard Medical School. Though not responsible for the outcome, Colin B. Begg and James A. Hanley made helpful comments on a draft manuscript and Charles E. Metz advised the authors throughout the analysis.
The 23 features listed below were determined to be of potential importance to diagnosis in a consensus meeting that followed interviews and also perceptualsimilarity tests in which data were subjected to multidimensional scaling analysis. Data obtained from subsequent scaling of these features for proven cases were subjected to discriminant analysis, which yielded the importance rankings in the righthand column. The 12 features with ranks listed were incorporated in the checklists for the enhanced readings.
Certain of the 23 features were not retained for the checklist, not because they