We are conducting a collaborative, population-based case-control study in Maine, New Hampshire, and Vermont to investigate the reasons for the elevated bladder cancer mortality in northern New England. Arsenic in drinking water is one of the primary exposures under investigation. To estimate subjects' lifetime exposure to waterborne arsenic, it will be necessary to obtain water samples from private wells that subjects used in the past. We conducted a methodologic study to assess the feasibility of locating and sampling from private wells at subjects' past residences. Ninety-eight New Hampshire residents (mean age 67 years) completed a questionnaire requesting the complete address, dates of occupancy, and drinking water sources for each home lived in since birth. An interviewer then asked subjects for more detailed information about each home to assist in a field search of past homes in the three-state study area of Maine, New Hampshire, and Vermont. Fifty-eight of the 98 subjects indicated that they had used a total of 103 private wells in 95 previous homes located in these three states. We conducted a field search to locate these 95 homes, visited town offices to find the properties on tax maps and obtain the current owners' names and addresses, attempted to obtain permission from the current owners to sample the wells, and collected water samples. In all, 48 (47%) of the 103 past wells in the study area were sampled successfully. The remaining wells were not sampled because the homes were not located (22%) or had been demolished (2%), permission to sample the wells was not obtained (17%), the wells had been destroyed (7%) or could not be found on the grounds of the residence (3%), or for other reasons (2%). Various approaches for improving the success rates for sampling water from private wells are discussed, as is the use of predictive modeling to impute exposures when sampling is not feasible.