Despite evidence of coverage errors and net undercounts in surveys due to listing errors and omissions from household rosters, there has been little systematic attention devoted to the measurement properties of household rosters. This paper draws on data from the Living Situation Survey, a pilot study which experimented with expanded roster cues and probes to identify individuals with marginal or tenuous attachments to households. The data are analyzed to test hypotheses that disagreements among individuals may cause omissions, and to evaluate the traditional reliance on household respondents as the source of information about who lives in a household. Evidence from the Living Situation Survey are used to identify categories of persons most at risk of omission due to disagreements, and to suggest further directions for research to improve the rostering process in household surveys.