The Integrated Coverage Measurement (ICM) process estimated the population of persons in the 1995 Census Test. The Person Interview component of ICM was conducted using Computer Assisted Personal Interview (CAPI). As part of the Census Bureau's evaluation of the 1995 Census Test, interviewers tape recorded a sample of ICM interviews. The tapes were then systematically analyzed using behavior coding techniques in order to evaluate the question-and-answer process of the ICM interview. Results from this evaluation were successful in identifying problem areas with question wordings, question order and the overall "flow" of the instrument. The study also examined the prevalence of reconciliation bias during the rostering and matching portion of the interview. Among other things, the evaluation uncovered questionnaire design deficiencies in the roster and reconciliation sections of the survey. We also detected interviewer behavior which suggests that interviewers preferred cases that did not encounter ICM/census discrepancies during reinterview. However, there was little direct evidence that interviewers biased the data by suppressing information that might lead to a discrepancy or by systematically favoring certain reconciliation codes. The paper concludes by describing changes that have been made to the ICM instrument as a result of this and other Census Bureau evaluations.
