This paper describes a research project exploring respondents' use of visual imagery in the American Housing Survey (AHS) to report the rooms in their homes. Previous qualitative research has demonstrated that the current AHS strategy results in errors -- both omissions and double reporting. In addition, formal reinterview studies have shown moderate levels of inconsistency with the current rooms inventory. We find these results surprising, given the non-sensitive, familiar, and well- learned nature of the information to be retrieved. The current AHS approach relies on a category-by- category tally, with an "all other rooms" catchall follow-up. We suspect that, rather than guiding respondents to an easy and accurate response, this approach actually inhibits a full accounting of rooms, because it fails to exploit information that is easily accessible in memory, forces a retrieval structure that does not correspond to cognitive organization, and fails to follow a spatially logical sequence. Our proposed solution encourages respondents to engage in a floor-by-floor "visual tour" of their homes. We offer evidence from exploratory cognitive research which suggests that the approach has promise for improved reporting of rooms in the AHS. We also describe the results of a small-scale field experiment to test the revised approach against the standard AHS questions.