OBJECTIVE: To determine if the length of the recall period is associated with differences in self-reported work loss and to determine the optimal length for the recall period. METHODS: Three different phone interviews were developed to quantify illness-related work loss. Two different recall periods, one-week and four-weeks, were used for each interview. A convenience sample (n = 20,088) of adult residents from the Baltimore, MD and Chicago, IL areas was contacted by phone, of whom 7,691 met occupation eligibility criteria. Respondents were randomized to interviews that used a one-week and four-week recall period. Reference measures were derived from follow-up interviews conducted one and four weeks later, respectively, using a bounded recall method (n = 615). RESULTS: A significantly lower estimate of lost productive work time was observed for interviews that used a four-week recall period compared to interviews that used a one-week recall period. Using the four-week recall data, work loss estimates for weeks 3 and 4 were significantly lower than estimates for weeks 1 and 2. This difference does not appear to be due to forward telescoping. Estimates of weekly work loss from the one-week recall interviews did not differ from estimates derived using data from weeks 1 and 2 of the four-week interview. Bounded recall data was used to confirm these results. If over-reporting were occurring, estimates of lost work time would be significantly lower using bounded recall compared to unbounded recall. We observed no differences by data collection method (i.e., bounded versus unbounded recall). CONCLUSION: We recommend a two-week recall period to minimize under-reporting of health-related lost productive work time and to maximize statistical power (i.e., capture of number of work-loss episodes).