OBJECTIVE. Some authors have proposed a national mammography registry to improve and monitor breast diagnostic practices. However, issues such as confidentiality, accuracy, and direct and indirect costs are practical barriers to implementing such as registry. This paper describes the development and design of a population-based mammography registry in New Hampshire. The project's objective are to assess the accuracy of mammography by comparing interpretive results with pathology and tumor-registry reports and to improve mammographic performance by reporting findings to facilities, radiologists, and pathologists statewide. MATERIALS AND METHODS. We recruited radiologists and pathologists through professional associations and facilities through site visits. Data used to develop and design the registry were collected during site visits, using structured face-to-face interview methods. Only one site refused to provide site-specific information. RESULTS. Facilities in New Hampshire estimated the annual mammographic volume to be approximately 148,000. We have noted a great deal of variability in mammography practices. Their principal methods for determining screening versus diagnostic mammograms were by patient self-reports (44% of practices), referring physicians' reports (38%), and radiologists' reports (18%). Although 71% of practices have computers, only 16% have radiology information systems or hospital information systems that offer computerized patient- tracking capabilities. More than 90% of New Hampshire radiologists exclusively use freehand dictation for reporting, and although almost 50% codify reports, only 11% use the American College of Radiology lexicon. These data and concerns expressed by radiologists, pathologists, technologists, and administrators helped shape the New Hampshire registry. CONCLUSION. Heterogeneity of radiologic practices poses major challenges for implementing a population-based mammography registry. Issues such as confidentiality, the difficulty of assessing diagnostic acumen, and the time involved in providing data to a registry must be adequately addressed. For the registry to succeed in such diverse settings, researchers, radiologists, pathologists, technologists, and administrative staff must collaborate and cooperate.