this paper is just the within households. beginning. It is critical that we continue the cycle of Evidence suggests that household respondents tend to analysis, qualitative exploration, and experimentation, err on the side of excluding marginal or peripheral that recently has begun to yield both knowledge and household residents. This may be due to flaws in the ways practical improvements in roster methods. As we the census roster questions were asked and in their continue to gain insight into the roster process, potential premise that each person can be assigned to one and only causes of coverage error emerge. We have a clearer one "usual residence." The variety of residence terms used picture than we did even a year ago of the sources of in the traditional census roster may also be a source of errors in household rosters, and of potential underreporting by household respondents. Gerber and methodological improvements which may reduce or Bates (1994) hypothesize that respondents assimilate the eliminate many of these errors. One issue we have not various contrary meanings and multiple rules by yet confronted is the mode of data collection and how it interpreting the roster as a request for the permanent or affects rostering using this questionnaire. We hope to official residents of the household. Possibly, the screening report back to you in a year additional improvements in questions developed in the more recent tests to determine that and other aspects of our roster research.