Traditionally, reinterviews have been designed for one (or more) of the following four purposes:

• to detect whether interviewers have deliberately falsified data,
• to evaluate interviewer performance,
• to estimate response variance, or
• to estimate response bias (Forsman and Schreiner, 1991).

Many reinterviews performed by the Census Bureau focus on estimating response variance. Although measuring response variance exposes inconsistencies in respondents' answers between interviews, it does little to explain why the inconsistencies occur.

Consequently, the 1991-92 Teacher Follow-up Survey (TFS) Reinterview and Extensive Reconciliation was designed with a new objective in mind. Primarily, it focused on determining the reasons for respondent and instrument errors.

In this paper, we briefly describe the methods that were used to conduct this reinterview, followed by a discussion of both the methodology's benefits and limitations.