In the general population, the use of stamps rather than business reply postage significantly improves response rates in mail surveys. Among physicians, however, a smaller effect might be anticipated due to their greater sophistication. An experiment was conducted to test the hypothesis that stamps would improve response rates and lower costs in a physician survey that included intensive follow up. In 1989, 380 physicians who reported providing primary care were surveyed. The protocol included two mailings, a postcard reminder, and two telephone reminders. Physicians were randomly assigned to receive a return envelope with a first-class stamp or an envelope that had been preprinted "business reply mail" in the first and second mailings. Response rates, calculated as completed surveys divided by eligible physicians, were 83.8 and 72.1% for stamps and business reply respectively, a difference of 11.7 percentage points (p < 0.01). Moreover, the total cost per completed survey was $11.18 for the physicians receiving stamps and $14.25 for the physicians receiving business reply. As in mail surveys of the general public, the use of first-class stamps on return envelopes both improves response rates and reduces cost in surveys of physicians.