Seeking to improve criterion and construct validity of retrospective self-reports of physical activity with brief...

Self-reported information is central to many scientific fields and achieving high levels of validity for these data is paramount. Concerns have been voiced about the validity of retrospective self-reports because of possible distortion of recall by memory failures and cognitive heuristics. In this study we test two techniques with the potential to improve self-report validity: the use of brief reporting periods (near-momentary reports) and of quantitative response scales (duration in minutes). To evaluate validity, a set of self-reported physical activity outcomes were compared with objectively assessed physical activities. A total of 258 community-dwelling adults self-reported their momentary physical activity 5 times per day and wore accelerometers throughout 7 days; they also provided end-of-week assessments of physical activity. Regarding criterion validity, all self-reports indicated lower duration of sedentary behavior, however, momentary reports were closer in magnitude to accelerometry than were 1-week recall reports. All self-reports indicated greater duration of physical activity than accelerometry. Correlations between self-reports and accelerometry data were low to modest for momentary and retrospective reports; they were also low for reporting with quantitative versus relativistic, Verbal Response Scales, with the exception of momentary reports of sedentary duration (r = .61). Regarding construct validity, associations between most demographic variables and self-reports did not replicate those observed for accelerometry-based measures; in some cases, these associations showed opposite directions, indicating poor validity. In sum, apart from two criterion validity analyses showing higher validity for momentary reporting (compared with recall), the results did not support the hypotheses that the use of brief reporting periods and quantitative response scales improved validity over 7-day reporting and Verbal Response Scales, respectively. Particularly concerning was the poor performance of all methods in tests of construct validity.