Ecological Momentary Assessment

Ecological momentary assessment (EMA) allows the study of behavior, psychological states, and physiological functions in their natural contexts. EMA and its predecessors (e.g., the experience sampling method) were developed with several purposes in mind. First, there was concern that retrospective autobiographical memory was fallible, due primarily to the use of cognitive heuristics during recall. EMA reduces these biases by generally limiting the period over which information is recalled. Second, although laboratory research offers the benefit of experimental control, it is not clear if processes observed in the laboratory are similar to what occurs in the “real” world. EMA often has greater ecological validity and generalizability because assessments can be collected in everyday settings. Third, EMA enables a closer examination of dynamic and temporal processes. EMA designs usually incorporate a large number of repeated measures, which provide a movielike view of processes over time. Such data not only allow examination of temporal patterns but also provide considerable information about (although not confirming) causal associations among variables.

EMA involves participants reporting on current or recent psychological states, behaviors, and/or environmental conditions, typically multiple times each day, for days or even weeks. Responses are collected in several ways, of which some are selfinitiated by research participants and others request responses after some signal (e.g., a pager, a handheld computer alarm). The three most commonly used approaches are interval-contingent, eventcontingent, and signal-contingent responding. Interval-contingent recording involves completing assessments at regular times (e.g., every hour on the hour, before bed). Event-contingent schedules entail completing assessments in response to specific events (e.g., smoking a cigarette, argument with a spouse). Signal-contingent schedules require individuals to report on experiences in response to random or semirandom signals across the day. Recent technological advances, most notably palmtop computers, provide a number of advantages to EMA data capture over paper-and-pencil approaches. As participants can respond directly on a handheld computer, portability is optimized, compliance can be automatically tracked (reports are date- and time-stamped), data can be transferred directly to statistical software, and researchers have greater control over the format and order of assessment items.

Despite the advantages of EMA approaches, they are not without limitations. First, implementation of EMA designs requires considerable time and expertise. There are many logistical issues: the design of the sampling scheme, thoughtful consideration of questionnaire design, training and motivating participants to follow the protocol, and dealing with the technical difficulties inherent in the use of technological devices (e.g., programming the devices). Second, momentary data collection techniques yield masses of complex, time-dependent data. Although such data are a strength of the approach, considerable statistical and data management acumen are necessary to manipulate and appropriately analyze these data sets. Third, given the intensive nature of data collection (e.g., five times each day for 2 weeks), the majority of participants are likely to have some missing data. This presents a problem for EMA research and must be accounted for in the statistical/analytic approach and interpretation of the data (e.g., are missing data random or reflective of an altered environmental state?).