of the phenomenon of interest. Whereas random reports are essential for the study of rapidly changing psychological states, they are not the only sampling strategy available for use with EMA. Depending on the research question at hand, event- contingent (e.g., complete report every time a stressful situation occurs) and time-contingent (e.g., fill out report every 2 hours) strategies can be employed. These are appropriate for the study of relatively rare or infrequent events.
In addition to avoiding recall biases and providing a representative view of the phenomenon of interest, the intensive sampling strategy of EMA allows for a detailed analysis of within-subjects effects. Diurnal cycles and the impact of environmental and situational factors on participants' experiences can be explored with EMA data. For example, using an EMA design, diurnal cycles in rheumatoid arthritis patients' pain and fatigue levels have been examined, providing new insights into disease experience (Stone et al., 1997). Variations in outcomes due to environmental and social factors may prove to be an important factor to consider when assessing disease severity and treatment response. Although extremely useful for detailed within-subjects analyses, the multiple reports per person generated through use of an EMA design can also be aggregated across time to examine between- subjects effects. These momentary-based “trai” measures provide summaries of participants' experiences that are not subject to retrospective recall biases.
Although there are many advantages to using EMA, there are also a few notable limitations. Given the intensive nature of the assessment strategy, participant burden is a legitimate concern. Early indicators suggest that completion rates in EMA studies are quite high and missing data are not a significant problem (Ockenfels et al., 1995; Smyth et al., 1996; Stone et al., 1997), but assessments should be brief so as to not add to participants' work load. There also is some concern about possible reactive effects of momentary assessments on participants' daily experiences. By requiring participants to attend closely to their daily experiences, their perception of their own experiences may inadvertently change. Initial work in the area of pain indicates that EMA monitoring does not change participants' perceptions (Cruise, Broderick, Porter, Kaell, & Stone, 1996), however, more work needs to be done to understand the effects monitoring has on participants' daily lives.
In addition to challenging participants, EMA designs pose some problems for researchers. Statistically, the multiple reports per person generated through EMA assessment strategies violate the assumptions of traditional linear modeling. To side step this issue, alternative, and at times complex, methods of analyzing the data need to be employed (see Schwartz & Stone, 1997). EMA can also be financially taxing. Special equipment, including signaling devices (e.g., watches or palmtop computers) and computer programs, are necessary to complete EMA projects.
In sum, although EMA requires extra effort on both the part of researchers and participants, it improves on traditional retrospective self-reports measures in many respects. Most notably, recall biases and cognitive errors are eliminated. In addition, EMA data are suitable for the examination of individual differences and complex interactions between environmental and psychosocial factors. Given these strengths, EMA is a viable research tool for gathering accurate autobiographical facts.
This chapter reviews recall biases and cognitive errors that affect retrospective self-reports. The studies examined suggest that the magnitude of retrospective reporting errors are significant and can threaten the validity of information obtained from research participants. Within the past two decades, researchers have begun to develop means of improving participants' reports. These efforts have taken two general forms: improve the accuracy of retrospective reporting through methods such as memory aids and bounded recall, and eliminate the possibility of retrospective recall biases by moving to momentary-based assessment strategies. Both approaches offer ways to improve the accuracy of information gained from self-reports, however, momentary assessments have the additional advantages of maximizing ecological validity and allowing for examination of within-subjects effects such as diurnal cycles and associations between environmental and psychosocial factors. Although ideal for the study of rapidly changing states and frequently occurring events, momentary assessments are not well-suited for all research questions (e.g., studies of extremely rare events). Thus, in addition to the continued development of momentary-based assessment strategies, further research examining autobiographical memory processes contributing to retrospective recall biases is needed. Of particular use would be studies identifying instances when reporting errors are most likely to occur. More work is needed to fully understand the cognitive processes involved in both momentary and retrospective self-reports, but the implications of the existing research are clear: Momentary assessment strategies should be employed whenever possible, and in cases in which the use of EMA is not feasible, the impact of recall biases and cognitive errors on retrospective self-reports needs to be acknowledged.
Aseltine, R. H., Carlson, K. J., Fowler, F. J., & Barry, M. J. (1995).
Comparing prospective and retrospective measures of treatment outcomes. Medical Care, 33(4), AS670–AS76.
Babor, T. F., Brown, J., & Del Boca, F. K. (1990). Validity of self-reports in applied research on addictive behaviors: Fact or fiction? Behavioral Assessment, 12, 5–31.
Baker, T. B., & Brandon, T. H. (1990). Validity of self-reports in basic research. Behavioral Assessment, 12, 33-C I.
Barclay, C. R., & Subramariam, G. (1987). Autobiographical memories and self-schemata. Applied Cognitive Psychology, I, 169–482.
Barclay, C. R., & Wellman, H. M. (1986). Accuracies and inaccuracies in autobiographical memories. Journal ofMemory and Language, 25, 93–103.
Blair, E., & Burton, S. (1987). Processes used in the formulation of behavioral frequency reports in surveys. Proceedings ofthe Sec