Objectives: The recall period specified for patient-reported outcomes (PROs) has been somewhat arbitrary, ranging from current symptoms to 4 weeks. In this study, accuracy of recall of pain and fatigue PRO items was evaluated for each of four time frames (1-day, 3-day, 7-day, 28-day) by comparing it with the averages of real-time electronic diary measures taken during the same recall period. Methods: Rheumatology patients (N=100) participated in a 30-day study utilizing electronic diaries (ED) and interactive voice recording (IVR) collection of PRO item responses. The ED gathered 6 randomly scheduled assessments each day for 30 days, and IVR assessments time- and date-stamped two 1-day, two 3-day, one 7-day, and one 28-day recall of pain and fatigue PRO items that were semi-randomly scheduled across 28 days. ED ratings were aggregated to correspond to each of the recall assessment timeframes and compared with PRO item ratings. This permitted an estimation of the relative accuracy of memory for pain and fatigue for each recall period. Conclusions: 1. patients can provide ratings across the entire range of time frames that correspond well to aggregated non-retrospective ratings (EMA) 2. the correspondence between recall and EMA for 28-day recall was not significantly worse than for 1-day recall 3. as the time frame increases from 1 to 28 days, recall shows a linear trend of higher pain and fatigue ratings; whereas, EMA remains stable These results support the validity and reliability of recall PRO items when used in a between-person methodology. It also replicates the finding that recall items consistently yield higher ratings of pain and fatigue than EMA for any given time frame, and that as the recall period increases, the discrepancy also increases, suggesting that PROs using different time frames within or across subjects will not yield an equivalent metric and should not be compared.