The goal of this study was to describe the magnitude, direction and sources of error of the American Medical Association's (AMA) masterfile (MF) in estimating physician supply in small towns. A random sample of nonmetropolitan towns in the United States was selected, and physicians with AMA MF (MFMDs) addresses in these towns were listed. Local pharmacists were asked to confirm or disconfirm the identities and locations of practice for the listed physicians and to add any unlisted physicians who were there. We took pharmacist confirmed or identified local source physicians (LSMDs) to be the "gold standard." The sample of 57 towns yielded 1,341 potential physician names. In these towns, there were 377 physician listings only from the MF, 188 only from local pharmacists, and 776 from both sources. About 80 percent of physicians identified by local informants were also listed on the MF; only 67 percent of physicians listed on the MF were identified by local informants as currently practicing in the town where they were listed. The error in these measures declined with increasing town size. The aggregate ratio of MFMDs to LSMDs was 1.20, ranging from 1.10 to 1.28 across size classes of towns. Given the persistence of local shortages of physicians, despite a national oversupply, accurate measurement of physician supply should be a priority of rural health care planners and advocates. Although the MF is the most comprehensive available national physician database, reliance on it alone to make local estimates of physician supply might lead one to believe that there are 20 percent more physicians in small rural communities than are actually there. Local pharmacists can be valuable informants about rural physician availability and their in- and out-migration.