Ecological Momentary Assessments in Sociology

Abstract

The proliferation of smartphone technology has afforded exciting new methodological opportunities within the social sciences. Ecological momentary assessments (EMAs) leverage this recent technological advancement by tracking the behaviors and perceptions of study participants as they are experienced in real time via smartphone devices in natural environments. Despite their longstanding theoretical interest in how the social environment influences a variety of personal outcomes, sociologists have been slower than many related disciplines to embrace EMAs as a viable methodology. This article promotes the use of EMAs by providing an historical overview of the methodology, highlighting several recent developments within sociology, and exploring future directions while clearly explicating inherent limitations to the EMA approach.

INTRODUCTION

From classical theories of modernization (Durkheim 1893; Weber 1978) to the Chicago School’s study of neighborhoods (Faris and Dunham 1939; Sampson, Morenoff, and Gannon-Rowley 2002) to emerging social network research (Cornwell and Ho 2021; Papachristos, Hureau, and Braga 2013; Perry and Pescosolido 2015), sociology consistently stresses the importance of the social environment in which people reside, work, and play. Empirical research in this tradition has historically relied on administrative records, surveys, and interviews to assess how social environmental factors (e.g., social relationships, concentrated disadvantage, public gathering places) are linked to various outcomes such as health, discrimination, and criminal activity (e.g., Aneshensel et al. 2011; Burt, Simons, and Gibbons 2012; Klinenberg 2002). Collectively, these methodological approaches have provided invaluable sociological insights. Yet one emerging critique of these approaches is their inability to provide real-time exposure to natural social settings (Cagney et al. 2020)—a methodological shortcoming that has limited sociologists from rigorously assessing many prominent theories.

The ecological momentary assessment (EMA) approach attempts to reconcile this limitation by prompting study participants to record their behaviors and perceptions as they are experienced in real time via smartphone technology in natural environments (Keusch and Conrad 2022; Shiffman, Stone, and Hufford 2008; de Vries, Baselmans, and Bartels 2021). This approach provides a rich account of the routine behaviors and real-time exposures to social environments that have long been theorized to influence sociological issues (e.g., drug use, mental health, discrimination). Indeed, one of the key strengths of the EMA approach lies in its ability to link longstanding sociological theories with methodologically innovative data. Although the increasing profusion of smartphone technology has dramatically increased the feasibility of conducting EMA studies, this approach has rarely been adopted by sociologists due perhaps to the financial and logistical burdens that are associated with the methodology (Keusch and Conrad 2022; Markowski et al. 2021a). In this article, I argue that these barriers to entry undersell the relevance of ecological momentary assessments for sociological study. I begin with a historical overview of the methodology and its adoption into academic research. I follow by highlighting recent sociological applications and exploring future directions for sociologists while clearly explicating inherent limitations to the EMA approach.

HISTORICAL OVERVIEW

The primary purpose of ecological momentary assessments—as the name suggests—is to capture momentary assessments of participants as they perform their daily routines (Keusch and Conrad 2022; Shiffman et al. 2008; de Vries et al. 2021). Although the use of smartphone technology to collect these data is a relatively recent advancement, there are several methodological traditions that served as antecedents to EMAs. Time use studies—which have a presence within sociology dating back to the 1930s (Lundberg, Komarovsky, and McInerny 1934; Sorokin and Berger 1939)—were initially introduced as an attempt to understand how people’s daily lives were socially structured (Cornwell, Gershuny, and Sullivan 2019). These studies ask participants to describe their previous day from morning until night, including the sequence of activities they performed throughout the day, the duration and location of each activity, and the people who accompanied them during these activities. Kahneman et al. (2004) refined this approach by introducing the day reconstruction method in which participants are prompted to recall their feelings during each activity. This latter approach has been used to study outcomes such as acute episodes of stress, loneliness, and pain (e.g., Musick, Meier, and Flood 2016).

These approaches led to the experience sampling method (ESM)—a method in which participants are randomly prompted to complete multiple short surveys throughout each day. The primary goal of ESM is to “capture daily life as it is directly perceived from one moment to the next” rather than being asked to recall how they felt during a specific time the day before (Hektner, Schmidt, and Csikszentmihalyi 2007:6). Pioneering ESM studies—which date back to the 1970s (Csikszentmihalyi, Larson, and Prescott 1977)—provided participants with electronic pagers that alerted them when it was time to self-administer a survey using a paper and pencil. As technology progressed, ESM studies eventually transitioned to smartphone applications.

Ecological momentary assessments emerged out of the field of behavioral medicine during the 1990s (Stone and Shiffman 1994). Although EMAs share many methodological similarities with ESMs, a key difference is that the former pays special attention to the ecological surroundings of participants. Recent smartphone technology, in particular, has dramatically changed the landscape in its ability to record fine-grained measures of study participants’ location via GPS (Chaix et al. 2013). This latter component enables researchers to incorporate secondary measures of place that hold theoretical relevance (e.g., crime statistics at the census tract level) as well as traditional measures of self-perception of one’s surroundings. The ability to overlay ecological measures on the momentary behaviors of individuals (hence the term ‘ecological momentary assessments’) marks a unique contribution to this emerging methodology. Despite these advances, sociology has been relatively slow to adopt the EMA approach compared to related disciplines such as psychology (Shiffman et al. 2008), public health (Trull and Ebner-Priemer 2009), and criminology (Solymosi, Bowers, and Fujiyama 2015).

WHY (NOT) ECOLOGICAL MOMENTARY ASSESSMENTS?

Ecological momentary assessments appeal to sociologists who are interested in the momentary actions of individuals as well as the ecological settings in which these actions occur. From a theoretical perspective, mircosociologists have long understood the importance of viewing social life as a chain of momentary events rather than an abstract social experience (Collins 2004; Goffman 1967). A rich history of research on neighborhood effects—and burgeoning literature on activity spaces—complements this perspective by noting the relevance of the social and physical locations in which people spend their time (e.g., home, workplace, street corners, recreational locations) (Cagney et al. 2020; Sampson et al. 2002). Social network theory, meanwhile, emphasizes the connections between people and how these connections influence personal outcomes (Small et al. 2021).¹ Given that these theoretical perspectives collectively highlight the importance of analyzing what people are doing (activity), where they are doing it (location), and who they are doing it with (social interaction), the EMA approach emerges as a powerful methodological tool that sociologists could use to implement many of their long-standing theories of human behavior.

EMAs are especially well suited to study actions, behaviors, and feelings that fluctuate or vary throughout a given day. For example, a researcher may be interested in studying how happiness is linked to the place a person lives. A traditional survey approach to this topic might ask participants something along the lines of “Taken things all together, how would you say things are these days—would you say that you are very happy, pretty happy, or not too happy?”² This global assessment of happiness might then be correlated with secondary data gathered on each participant’s residential neighborhood (e.g., median income, crime rates). Although such an approach can reveal meaningful findings, it is important to note that happiness is not a stagnant emotion (Dambrun, Desprès, and Lac 2012) nor do people only spend time in their own neighborhoods (Cagney et al. 2020). By capturing repeated momentary assessments on each participant that ask about their current levels of happiness and geocoding their exact location, the researcher will be able to determine whether participants are happier in certain locations (e.g., home, park, bar) and whether other contextual factors (e.g., who they are with, what they are doing, time of day) further augment their happiness.

There are, however, several noteworthy barriers and limitations that one must consider before implementing an EMA study. First—and perhaps most obvious—is the associated cost with such an endeavor. EMA studies require that all participants have prolonged access to a smartphone. Although smartphone ownership has become increasingly common in recent years, restricting study participation on this basis is not advisable as it may systematically omit people of a certain demographic. Prospective EMA researchers should instead consider purchasing loaner phones for those participants who do not already own a smartphone. The implementation of the momentary assessments, meanwhile, requires the use of software that is tailored towards collecting EMAs. There are currently many smartphone apps readily available for academic researchers, but most require payment for use. Consequently, the costs can quickly escalate depending on the desired sample size.

Second, the repeated collection of momentary assessments places considerable burden on participants (Shiffman et al. 2008). Unlike a traditional survey that is completed in a single sitting, EMA studies require participants to continually self-report their data into a smartphone throughout the course of multiple days. The automated alerts that signal that it is time to provide a momentary assessment are typically programmed to be pseudo-random with the intent to catch the participant as they engage in different activities. Although these assessments typically require only a few minutes to complete, participants may nevertheless be unlikely to complete the assessments if they are occupied by a demanding task or activity. Indeed, missing data is a real concern with EMA studies (Markowski et al. 2021a). Prospective EMA researchers should therefore carefully consider how many assessments per day they need from each participant. Underestimating the number of assessments may miss certain moments of interest throughout the day (as well as causing the subsequent analysis to be statistically underpowered) whereas overestimating the number of assessments may result in excess missing data and unethically waste each participant’s time by interrupting their daily routine.

Third, not everyone is equally equipped to use EMA technology (e.g., older adults with limited dexterity, people who are visually or cognitively impaired). This is particularly concerning because unlike a computer-assisted personal interview, EMA participants are effectively on their own once the EMA phase of the study begins. Therefore, it is important for researchers to consider potential pitfalls and explore mitigating actions prior to the start of data collection.

Lastly, it is worth noting that EMAs are not always a logical methodological choice. For instance, EMAs do not provide a full account of daily life. Although it has roots in the time-use tradition, the EMA approach is designed to capture only a select few moments as they are experienced in natural settings. If the researcher is interested in studying the entire course of one’s day—as done in sequence analysis—they would be better off relying on time use diaries (Vagni and Cornwell 2018). Moreover, the EMA approach is unsuitable for studying social processes that unfold over long periods of time. EMA studies typically track participants for approximately a week as anything beyond this leads to participant burnout (Broderick et al. 2003). For instance, one is unlikely to notice any accumulating health declines that may occur as a result of long-term exposure to harmful environments. Researchers interested in these types of social processes should consider alternative methodologies (e.g., longitudinal surveys). Finally, researchers who are interested in observing social behaviors within a specific location (e.g., a crowded marketplace) may be better suited to implement an ethnographic methodology in which they rely on systematic social observations (e.g., Anderson 2011; Grazian 2008). The downside to this latter approach, however, is that the observer cannot readily gather data on the feelings, thoughts, and perceptions of their subjects—a limitation that may or may not affect the overall purpose of the study in question.

RECENT SOCIOLOGICAL APPLICATIONS

As previously mentioned, sociology has long acknowledged the theoretical importance of the social environment. Indeed, various elements of the social environment—such as the places individuals live and visit (Cagney et al. 2020; Sampson et al. 2002), the people with whom they share social connections (Roth 2020; Small et al. 2021), and the social groups and organizations to which they belong (Small 2009; Wharton, Rotolo, and Bird 2000)—all have profound impact on a person’s life chances. In this section, I highlight three exemplary sociological studies that use EMAs to build on these theoretical traditions. These include the Adolescent Health and Development in Context study, the Real-Time Neighborhoods and Social Life Study, and the Chicago Health and Activity in Real Time study.

The Adolescent Health and Development in Context (AHDC) was among the first sociological studies to employ an EMA approach. Fielded in 2014–2016, the AHDC longitudinally sampled over 1400 adolescents (age 11–17) residing in the Columbus, Ohio metropolitan area using an EMA approach to assess how social context (e.g., neighborhoods, social relationships) affects mental health and behavioral development. A recent analysis of these data found that higher levels of local violence were negatively associated with time spent in one’s own residential neighborhood (Browning, Calder, et al. 2021). Another study analyzed data on a subsample of AHDC participants who provided nightly saliva samples during the EMA study period and found that recent exposure to local police-related deaths was associated with a significant increase in cortisol levels among black adolescent boys (Browning, Tarrence, et al. 2021). The AHDC was particularly innovative as it collected biometric data that could be analyzed in conjunction with the EMA data.

Around the same time, the Real-Time Neighborhoods and Social Life Study (RNSLS) was launched to study the opposite end of the life course. This study—fielded in 2014—provided smartphones to 61 older adults residing in New York City to assess their activity spaces within the city over the course of a week (York Cornwell and Cagney 2017). Using geocoded momentary assessments, York Cornwell and Goldman (2020) found that immediate exposure to perceived neighborhood disorder—but not objective measures of concentrated disadvantage—was associated with momentary spikes in pain and fatigue among their sample. These findings remained after adjusting for participants’ cumulative exposure to disorder throughout the study period. The RNSLS, which was a small-scale study of a non-probability sample, demonstrated the feasibility of using EMAs to study older adults—a population that has been questioned in their ability to adequately use smartphone technology (Bartels et al. 2020; Maher, Rebar, and Dunton 2018).

The Chicago Health and Activity in Real Time (CHART) study builds on the RNSLS by gathering data on a probability sample of 450 older adults residing in Chicago to understand how the social and spatial environments in which older adults spend their time affect and are affected by their mental physical and emotional health. A recent cross-sectional analysis found that when older adults were outside their homes and or accompanied by other people they were significantly less likely to experience momentary feelings of loneliness compared to when they were alone at home (Compernolle et al. 2021).³ Further analyses of the CHART—which leveraged data on participant’s personal social networks—revealed that social accompaniment is more protective of momentary loneliness for older adults who have larger personal networks than it is for those with smaller personal networks (Goldman and Compernolle 2023).

FUTURE DIRECTION

Like any emerging methodology, EMAs present many exciting opportunities for future research. Researchers interested in studying processes that operate through social interactions and relationships (e.g., discrimination, violence, loneliness) may use EMAs to link participants together via a series of different approaches. Drawing on the ecological network tradition, researchers can indirectly establish connections between participants who occupy the same physical location (assessed via GPS) at the same time (Browning et al. 2017). One downside to this approach is that it is limited to studying connections between study participants.⁴ Researchers who are concerned with this latter issue may alternatively target interactions that participants have with any given person who they encounter during the study period. Taking this one step further, researchers may collect personal network data (i.e., interconnected webs of social relationships surrounding each study participant) prior to the start of the EMA phase. This enables researchers to ask about the presence of these pre-specified network members during each momentary assessment (e.g., Ng et al. 2021).

Following a recent trend in medical sociology (and related sub-fields), prospective researchers may consider incorporating biometrics markers into their studies that extend beyond traditional self-reported health statuses. As seen in the AHDC sub-study, this may come in the form of participant-supplied samples (e.g., nightly saliva) that are periodically collected throughout the EMA phase of data collection rather than during the study baseline. EMA scholars interested in cognitive aging, meanwhile, have leveraged clinical adaptations of mobile cognitive tests to assess how daily social interactions are associated with in-the-moment variations in cognitive performance such as processing speed and attention (Zhaoyang et al. 2021). Implementation of these types of data could serve to further our understanding of how immediate exposures to the social environment are linked with biological markers of health and well-being.

Finally, EMA research should expand to ensure proper representation across various sub-populations. For instance, EMA studies tend to neglect vulnerable populations that have historically been hard to reach (e.g., homeless populations), perhaps due to the extended financial and logistical difficulties that arise when using EMA technology (see Markowski et al. 2021b; Tyler and Olson 2018; Tyler, Olson, and Ray 2018 for noteable exceptions). Moreover, extant EMA studies—including the three highlighted above—have almost exclusively focused on urban samples. Expansion into rural and other non-metropolitan areas could provide valuable insight into the well documented social disparities occurring across geographic classifications (Jensen et al. 2020; Lichter and Brown 2011). Moving forward, the implementation of a large, nationally representative EMA study—akin to the American Time Use Survey (Bureau of Labor Statistics 2021)—would allow researchers to assess whether social environmental factors matter more in certain geographic regions and ensure generalizability to the broader population.

CONCLUSION

Sociology has a historical interest in how the social environment influences a host of personal outcomes. Although it is not universally applicable to all issues concerning sociologists, the EMA approach offers an innovative perspective on numerous longstanding theories by dynamically tracking individuals as they navigate their social environments over the course of daily life. Yet sociology has been relatively slow to adopt such a method compared to its sibling disciplines (e.g., psychology, criminology, public health). As evidenced by several recent studies, EMAs enable researchers to study issues as widespread as drug use, loneliness, stress, neighborhood exposure, social interactions, and violence. Moving forward, sociologists should consider employing the EMA approach when it appropriately matches their theoretical motivations.