An administrative record check of program participation reporting in the first two SIPP interviews shows that, while response errors are rare, they have important biasing effects on estimates of means and correlations. Our search for models of the causes of the errors includes classical hypotheses about forgetting, memory decay, confusion about name attributes, telescoping, learning to underreport, interviewers, and proxy responses. While these hypotheses give occasional insights into isolated error problems, they do not provide a fundamental understanding of the error dynamics. We mention some exploratory cognitive research that may provide the broader understanding and may be useful in devising better measurement procedures.