The Accreditation Council for Graduate Medical Education Resident-Fellow Survey measurement of compliance with duty hours uses remote retrospective resident report, the accuracy of which has not been studied. We investigated residents' remote recall of 16-hour call-shift compliance and workload characteristics at 1 institution. We sent daily surveys to second- and third-year internal medicine residents immediately after call shifts from July 2011 to June 2012 to assess compliance with 16-hour shift length and workload characteristics. In June 2012, we sent a survey with identical items to assess residents' retrospective perceptions of their call-shift compliance and workload characteristics over the preceding year. We used linear models to compare on-call data to residents' retrospective data. We received a survey response from residents after 497 of 648 call-shifts (77% response). The end-of-year perceptions survey was completed by 87 of 95 residents (92%). Compared with on-call data, the recollections of 5 (6%) residents were accurate; however, 48 (56%) underestimated and 33 (38%) overestimated compliance with the 16-hour shift length requirement. The average magnitude of under- and overestimation was 18% (95% confidence interval  =  13-23). Using a greater than 10% absolute difference to define under- and overestimation, 39 (45%) respondents were found to be accurate, 27 (31%) underestimated compliance, and 20 (23%) overestimated compliance. Residents overestimated census size, long call admissions, and admissions after 5 pm. Internal medicine residents' remote retrospective reporting of compliance with the 16-hour limit on continuous duty and workload characteristics was inaccurate compared with their immediate recall and included errors of underestimation and overestimation.