Naturalistic Observation of Health-Relevant Social Processes: The Electronically Activated Recorder (EAR) Methodology in Psychosomatics

Abstract

This article introduces a novel, observational ambulatory monitoring method called the Electronically Activated Recorder or EAR. The EAR is a digital audio recorder that runs on a handheld computer and periodically and unobtrusively records snippets of ambient sounds from participants’ momentary environments. In tracking moment-to-moment ambient sounds, it yields acoustic logs of people’s days as they naturally unfold. In sampling only a fraction of the time, it protects participants’ privacy and makes large observational studies feasible. As a naturalistic observation method, it provides an observer’s account of daily life and is optimized for the objective assessment of audible aspects of social environments, behaviors, and interactions (e.g., habitual preferences for social settings, idiosyncratic interaction styles, and subtle emotional expressions). The article discusses the EAR method conceptually and methodologically, reviews prior research with it, and identifies three concrete ways in which it can enrich psychosomatic research. Specifically, it can (a) calibrate psychosocial effects on health against frequencies of real-world behavior, (b) provide ecological, observational measures of health-related social processes that are independent of self-report, and (c) help with the assessment of subtle and habitual social behaviors that evade self-report but have important health implications. An important avenue for future research lies in merging traditional, self-report based ambulatory monitoring methods with observational approaches such as the EAR to allow for the simultaneous yet methodologically independent assessment of inner, experiential (e.g., loneliness) and outer, observable aspects (e.g., social isolation) of real-world social processes to reveal their unique effects on health.

Social environments play a pivotal role in the development of disease and maintenance of health (1). This scientific fact was recently reaffirmed by the National Institute of Health when it launched the Basic Behavioral and Social Science Opportunity Network (OppNet) in November, 2009. OppNet is a trans-NIH initiative to “pursue opportunities for strengthening basic behavioral and social science research” and to promote “activities and initiatives that focus on basic mechanisms of behavior and social processes” (OppNet Mission Statement). The launch of OppNet was followed by a concerted release of unique funding opportunities specifically aimed at accelerating progress in the understanding of how social and behavioral processes influence health. The OppNet initiative showcases NIH’s strong commitment to improving health by identifying, understanding, and addressing psychosocial risk factors.

This emphasis on psychosocial factors brings with it the need for their high-fidelity assessment. Because of their phenomenological complexity, this is usually best attained through systematic multimethod measurement (2). Interestingly, it is exactly in this regard, though, that the field of psychosomatic research is subject to a subtle but important imbalance. Whereas the measurement of the “somatic” side of the social-health equation has developed rapidly with progress in noninvasive peripheral and central physiological, endocrinological and immunological assessment (3, 4), the measurement of the psychosocial context continues to rely largely on participants’ self-reports.

Since the early 1980s, psychological ambulatory monitoring methods–also known as ecological momentary assessment (5), experience sampling (6) or ambulatory assessment (7)–have been developed to bypass methodological concerns around global and retrospective self-reports. By now, they have effectively become the gold standard for assessing psychosocial factors in daily life, and their impact on psychosomatic research is hard to overstate. In essence, they have made it possible to study social context in vivo and in real-time (8–10).

Despite this advance, the psychosomatic researcher’s “tool kit” still lacks a method to directly observe “the social context in which [patients] live,” which is central to the biopsychosocial model of medicine (p. 132; (11)). Lab-based observational (as opposed to self-report-based) methods such as videotaping couple (12) or family interactions (13) are available, however, methods for the direct observation of behavior in the natural environment are missing from the existing array of research methods. This is likely because it is not straightforward how one would go about collecting truly naturalistic behavioral data (14). Therefore, self-report-based momentary assessments are generally considered the best available proxy to behavioral observation in the field (15, 16).

From a multimethod perspective, however, momentary and global or retrospective self-reports share important method variance because both derive their data from participants’ reports of their introspections, perceptions, and cognitions. Thus, some concerns raised for global/retrospective self-reports potentially also apply to momentary self-reports (e.g. impression management, self-deceptive enhancement, limitations to what participants are aware) (16). Therefore, to complete the psychosomatic researcher’s tool kit, it would be desirable to complement momentary self-report with momentary observational data. In this article, we review and discuss a novel ambulatory monitoring method for the unobtrusive sampling of naturalistic observations.

A Method for the Naturalistic Observation of Social Processes in Daily Life: The Electronically Activated Recorder (EAR)

Over the last ten years, we have developed and validated the Electronically Activated Recorder or EAR (17), a method that unobtrusively samples acoustic observations of participants’ momentary environments within the natural flow of their lives.

The EAR system

The EAR is a portable audio recorder that periodically records brief snippets of ambient sounds. Participants wear it on their waistline while going about their daily lives. In tracking moment-to-moment ambient sounds around the participants, it yields acoustic logs of their days as they naturally unfold. In sampling only a fraction of the time instead of recording continuously, it makes large naturalistic observation studies feasible.

While the EAR started out as a system based on analog micro-cassette recorders, recent versions of it have run on handheld computers. The handheld-computer based EAR system has critical advantages: (a) it is software-based and runs on regular, commercial devices; (b) it is available at an affordable price (the cost of a handheld computer); (c) it allows for freely programmed recording schedules as well as blackout periods with no recordings (e.g., overnight). Finally, because now traditional, self-report-based ambulatory monitoring methods and the EAR use the same platforms (18), it is possible to merge both methodologies.

Since 2005, we have used a version of the EAR that runs on Windows Mobile devices (e.g., HP iPAQ 110). It was successfully employed in several research projects (19, 20) and continues to be used in ongoing studies. Recently, Apple certified the iEAR app which runs on both the iPhone and the iPod touch and is now available for free download on iTunes. The development of the iEAR app constitutes a significant methodological advance because the iOS (a) is currently among the most robust and refined mobile computing platforms, (b) has an enormous user base and is expected to have long-term technology support and (c) is an operating system that maximizes compatibility across regions (with iPhones and iPod touches being sold in various countries).

A comparison of self-report-based ambulatory monitoring methods and the EAR method

As a real-time data capture method, the EAR method compares most directly to self-report-based ambulatory monitoring methods (6–8, 21). Table 1 summarizes important similarities and differences between the two methods.

Table 1.

A Comparison of self-report-based Ambulatory Monitoring Methods and the EAR Method

	Self-Report-Based Ambulatory Monitoring	EAR-Method
Approach	Naturalistic	Naturalistic
Medium	Paper & Pencil, Electronic (handheld computer or mobile phone)	Electronic (handheld computer or mobile phone)
Mode	Active (data provided through voluntary response)	Passive (data collected through automatic recording)
Method	Self-report	Behavioral Observation
Perspective	Self (Agent)	Other (Observer)
Awareness of Assessment	High	Low after Habituation
Burden for Participant	Practical (Interruption of Daily Life)	Psychological (Intrusion of Privacy)
Burden for Researcher	Preparing Participants (Instruction and Training)	Preparing the Sound Data (Coding and Transcribing)
Data Collection Limited By	Response Burden	Privacy Considerations, Lab Capacity for Data Coding
Optimized for Assessment of	Subjective Experiences and Perceptions	Objective Social Environments and Interactions

Note: Adapted from Mehl and Conner (2012). Copyright 2012 by Guilford Press. Adapted by permission.

The most important difference between the two methods lies in the fact that traditional, psychological ambulatory monitoring methods are based on momentary self-reports, whereas the EAR is based on momentary behavioral observation. The two types of methods hence adopt different assessment perspectives: the self, with the corresponding subjective account, versus the bystander, or observer, with the corresponding objective account (12, 26, 27).

The practical and conceptual differences between traditional self-report-based ambulatory monitoring and the EAR as outlined in Table 1 suggest that the two methodologies are best suited for different assessments. In capturing the agent’s “insider” perspective, self-report-based ambulatory monitoring is optimized for the assessment of participants’ subjective experiences and perceptions (e.g., thoughts, feelings). In contrast, in capturing the observer’s “outsider” perspective, the EAR is optimized for assessing audible aspects of participants’ objective social environments and interactions (e.g., social settings, communication behaviors, emotional expressions).

Psychosocial information contained in the EAR recordings

Although other approaches to extracting information from the EAR recordings exist (29), we have primarily used a behavior coding or counting strategy. Specifically, in our research, trained coders listen to all of a participant’s EAR recordings and code each sound file for the presence (‘1’) or absence (‘0’) of behaviors using a standardized coding system. Over the years, we have developed the Social Environment Coding of Sound Inventory (SECSI) (30, 31) to capture acoustically-detectible aspects of participants’ social environments and interactions. In its basic form, the SECSI comprises four category clusters: (a) the person’s current location (e.g., at home, outdoors; inferred from ambient cues, such as sounds of TV and housework at home or wind blowing outside), (b) activity (e.g., watching TV, eating), (c) interaction (e.g., alone, talking, on the phone) and (d) emotional expression (e.g., laughing, crying, sighing).

Conceptually, it captures information about how individuals (a) select themselves into social environments (e.g., displaying a preference for spending time in one-on-one versus group settings) and (b) interact with their social environments (e.g., laughing or arguing a lot; see Figure 1). Adding to the basic SECSI system, we have then developed more specific coding systems such as one for capturing aspects of patients’ interactions with their support networks (e.g., disclosure, positive or negative support received), and other coping-relevant behaviors (e.g. complaining, expressing optimism, showing humor). In our lab, coders also routinely transcribe participants’ utterances captured by the EAR. We then analyze the transcripts to obtain information about participants’ linguistic styles (32).

Figure 1.

EAR assessment of participants’ daily social environments and interactions.

Note: Adapted from Mehl and Conner (2012). Copyright 2012 by Guilford Press. Adapted by permission.

Psychometric properties of the EAR data

The psychometric properties of EAR data can be determined through estimation of (a) intercoder reliability, (b) test-retest reliability, and (c) parallel-test reliability. We routinely obtain estimates of intercoder agreement by having all coders code a standard set of training EAR recordings. Consistent with the specific, concrete, and behavioral nature of the codings (e.g., ‘talking’ or ‘laughing’), intercoder reliabilities tend to be high (30,31,33). It is an advantage of the behavior counting approach that behavior codings of specific, narrowly defined behaviors (e.g., ‘talking’) are less susceptible to interpretational ambiguity (and thus coder unreliability) than ratings of broader psychological constructs (e.g., ‘relationship satisfaction’).

Further, Mehl and Pennebaker (31) analyzed the four-week stability of different daily behaviors. Participants’ activities and language use emerged as remarkably stable over this period supporting the test-retest reliability of the EAR data. For example, the correlation between the two two-day assessments (spaced four weeks apart) was .64 for how much time participants spent alone, .60 for how much they listened to music, and .77 for how much time they spent in transit. A similar picture emerged for aspects of their habitual language use. The degree to which participants used swear words (r_tt = .86), non-fluencies (r_tt = .62, and filler words (r_tt = .59) was highly stable over the four-week period.

Finally, estimates of parallel-test reliability can be obtained by simulating sub-samples or “parallel tests” that are based on 50% of the collected data (i.e. aggregating across every other recording). Employing this strategy on data from two published studies (based on the default sampling pattern of 30 sec every 12.5 min) (30, 38), we found that the “test halves” and the full data yielded essentially identical descriptive statistics (e.g., central tendency, dispersion). The measures based on the test halves and the full data also evidenced a very high degree of correspondence (test-half—total correlations: M = .97, Min = .66, Max = 1.00). This was especially true for high base rate (being on the phone, watching TV) but also applied to low base rate (laughing, singing) behaviors.

More importantly, the measures based on the two test halves also showed highly comparable criterion validity. Specifically, the analyses based on 50% of the data successfully “recovered” the published effects based on the full data. For example, Mehl and his colleagues (38) reported that small talk correlated −.33 and substantive conversations .28 with well-being. The correlations based on the two test halves were −.33/−.32 and .29/.26, respectively. Similarly, Mehl and colleagues (30) found that talking correlated .30 with extraversion and class attendance .42 with conscientiousness. Again, the correlations based on the test halves were highly comparable (.30/.27 and .41/.42).

Taken together, these analyses suggest that (a) psychological information can be reliably decoded from the sampled ambient sounds, (b) behavioral aggregates based on two days of monitoring evidence good temporal stability, and (c) the default sampling pattern (30 sec every 12.5 min) yields generalizable estimates of a person’s daily behavior. We therefore suggest that—although longer monitoring periods are always preferable to shorter ones—a 48-h assessment window represents a viable compromise between optimizing data quality and ensuring study feasibility and yields reliable estimates of habitual aspects of participants’ daily social environments, behaviors, and interactions. Future research should test for potential day-of-the-week effects (39).

Ethical considerations around conducting EAR research

Recording ambient sounds raises ethical and legal questions. EAR studies conducted in our lab routinely implement a series of safeguards to protect participants’ privacy and to ensure the confidentiality of the data (25). We have found these safeguards to be highly effective at alleviating concerns that participants may have about the method. First, the EAR is programmed to record only a small fraction of a person’s day. For example, sampling 30 sec every 12.5 min, leaves 95% of the time unrecorded. Second, the recordings are kept short; 30 sec recordings are long enough to reliably extract basic behavioral information, yet they are short enough to capture only little contextualized personal information. Finally, and most importantly, before the investigators access the data, all participants can listen to their EAR recordings and delete parts they do not want on record. In one study, 19 out of 96 participants (19.8%) reviewed their recordings but only erased a total of 10 sound files (30). In another study with 13 rheumatoid arthritis patients, only 1 out of 2,948 waking sound files was erased (19). This suggests that participants feel comfortable with sharing the sounds of their daily lives under our safeguards.

However, more serious concerns revolve not around the participants themselves, but around bystanders whose behaviors are captured by the EAR. In the United States, there are few restrictions about recording people’s utterances in public places. The situation concerning recording private conversations is more ambiguous. In most parts of the U.S., recordings can be made legally if at least one of the interactants (e.g., the participant who is wearing the EAR) has knowledge of the recording device. A small number of states only allow recordings if all interactants have knowledge of the recording device. But even in the most legally restrictive states, unauthorized recordings are only problematic if they are personally identifiable.

In our lab, we encourage participants to wear the microphone visibly and to readily mention the EAR in conversations with others. Irrespective of such notification, anonymity of other people’s utterances is of critical importance, because their behavior is collected without explicit informed consent. As mentioned above, the brief recordings minimize the chance that personally identifying information about others is captured in the first place. Further, in the coding process, any personally identifying information is omitted from the transcripts. It is thus highly unlikely that the EAR paradigm as we have established it violates privacy rights of people who are inadvertently recorded.

Obtrusiveness of and participants’ compliance with wearing EAR device

The EAR method requires participants to tolerate being intermittently recorded without exactly knowing when. This can create evaluation apprehension and result in reactance (i.e. censored or artificial behavior) or non-compliance (i.e. not wearing the EAR). Thus, it is critical to estimate how obtrusive the method is in daily life and how well participants comply with it.

Mehl and Holleran (28) addressed these questions by analyzing measures of both self-reported and behaviorally-assessed EAR obtrusiveness and compliance in two samples: a short-term (two days; (30) and a longer-term (10–11 days; (34)) monitoring. Self-reported obtrusiveness was measured with items such as “To what degree were you generally aware of the EAR?” As a behavioral measure of obtrusiveness, the coders counted in how many sound files participants mentioned the EAR with others. As a self-report measure of compliance, participants reported what percentage of the day they were wearing the EAR. Finally, as a behavioral compliance measure, the coders counted the number of sound files that indicated that participants were not wearing the EAR. This was coded if, over the whole sound file, no ambient sounds at all were recorded—not even sounds of breathing or clothes rubbing against the microphone.

Results revealed that soon after receiving the EAR, participants go through a brief period of heightened self-awareness in which conversations about the EAR are frequent. Within two hours, however, most participants habituate to the method and rarely mention it with others (Panel A of Figure 2). This habituation effect was found for the short-term and longer-term monitoring.

Figure 2.

Behaviorally assessed obtrusiveness of and compliance with the EAR Method.

Note: From Mehl and Holleran (2007). Copyright 2007 by Hogrefe Publishing. Reprinted by Permission.

The study further yielded the following findings about participants’ compliance: In the short-term monitoring, participants’ compliance was very high in the first hours after they had received the EAR. Non-compliance gradually increased over time and leveled off at about 10–12% of the time on the second day of monitoring (Panel B of Figure 2). Compliance in the longer-term monitoring was high for at least 6 days. After that, variability in non-compliance increased suggesting that some participants’ tolerance threshold may have been reached.

The compliance data reported in Mehl and Holleran (28) are based on two studies with student samples. We have since run a series of EAR studies with samples of older adults (e.g., individuals with rheumatoid arthritis (19); couples in which one member was receiving treatment for breast cancer (20); faculty members at a Tier I research institution (35)) and have obtained highly comparable results regarding EAR obtrusiveness and compliance.

Together, this suggests (a) that EAR compliance and obtrusiveness can be reliably assessed, (b) that compliance is generally high and comparable to what has been reported for self-report-based ambulatory monitoring (23, 24) and (c) that after an initial, short habituation period, the method operates fairly unobtrusively and does not interfere much with participants normal activities.

Practical considerations for running an EAR study

Researchers can obtain a copy of the (old) Windows Mobile based EAR software from our laboratory or download the new iEAR app directly and freely from iTunes. Also available from our laboratory is a researcher’s guide with practical information for how to run an EAR study. It contains instructions for installation and use of the software, a sample consent form, a set of standard questionnaires (e.g., EAR compliance and obtrusiveness questionnaire), and a script for how to administer the EAR. Finally, we are always glad to help researchers start and troubleshoot EAR studies.

Potentials of the EAR Method for Psychosomatic Research

As a novel, observational ambulatory monitoring method, the EAR needs to justify its use through bringing unique potential to the assessment of social processes. In this section, we discuss three concrete and important ways in which naturalistic observation of social processes with a method such as the EAR can enrich psychosomatic research.

Calibration of psychosocial effects on health against frequencies of real-world behavior

A persistent problem in research on psychosocial influences on health is the calibration and interpretation of effects. In a seminal article, Sechrest, McKnight, and McKnight (36) pointed out that “very few psychological measures of any kind are expressed in a metric that is intuitively or immediately meaningful” (p. 1065). Instead, the vast majority of standardized scales use arbitrary metrics, such as 5- or 7-point rating scales. Such measures have no clear referents that inform about what scoring at a certain level (e.g., a “5” on optimism) means for how a person fares in important domains of life. For example, how much less time does a person with a “4” on a social integration scale spend alone compared to a person with a “2”? And, how does a person’s daily life change if an intervention reduces his depression score by 7 points? Finding answers to questions like these is crucial for understanding the implications of psychosocial effects on health. Yet, the field has struggled greatly with accomplishing this (37).

One advantage of the EAR is that its sound-file based behavioral codings can be readily converted into a metric that is non-arbitrary, intuitively meaningful, and inherently real-world relevant. If the EAR captures a person talking in 40 out of 120 recordings, one can estimate that the person spent about a third of her time awake (about 5 hours) talking. Or, if TV sounds are present in 20% of the recordings, one can estimate that the person was (actively or passively) watching TV 20% of the waking day (about 3 ½ hours). By linking EAR-derived frequencies of daily behavior to the metrics of measures, a better understanding of the real-world implications of psychosocial influences on health can be obtained.

In a recent study, for example, Mehl and colleagues (38) found that greater subjective well-being is related to (a) spending less time alone and more time with others and (b) having less small talk and more substantive conversations. The magnitudes of these effects were vividly illustrated by the fact that, compared to the unhappiest participants (-2.0 SD) the happiest ones (+1.5 SD) spent 25% less time alone (58.6% vs. 76.8%) and 70% more time talking (39.7% vs. 23.2%). They also had roughly one third as much small talk (10.2% vs. 28.3%) and twice as many substantive conversations (45.9% vs. 21.8%).

Similarly, testing the myth that women are by a factor more verbose than men, Mehl and colleagues (40) estimated based on six EAR studies that both men and women use about 16,000 words per day. A sex difference of 546 words compared to a range of over 46,000 words between the least and most talkative individual (695 vs. 47,016) rendered significance testing close to meaningless—and vividly illustrated the large magnitude of individual differences in social engagement.

Finally, and directly relevant to the health context, we recently began a collaboration with Charles Raison where we use the EAR to study how proinflammatory cytokines regulate prosocial and affiliative behavior. Prior research has found inflammatory and socioemotional processes to be tightly linked, but how and how much specific, real-world social behaviors are affected by chronic inflammation is largely unknown (41–43). In using the EAR with Hepatitis C patients before and after interferon-alpha (or no) treatment, subtle changes in daily social life (e.g., amount of arguing or swearing) can be objectively traced and quantified in an intuitive, non-arbitrary metric.

Thus, in facilitating an absolute metric in the measurement of daily social processes, the EAR can help “benchmark” and interpret psychosocial influences on health.

Ecological measurement of health-related social processes independent of self-report

Understanding both “the quantity and quality of social relationships and support in society” (p.544) is central to unpacking how social relationships affect health (44). However, how can the effects of objective versus subjective aspects of social processes be studied separately if the two components cannot be assessed independently without shared method variance? Clearly, it is possible to ask participants to report on both aspects and often their answers are meaningful and interpretable (e.g., feelings of loneliness (45); size of social network, (46)). Yet, ultimately, in both cases, the data are fundamentally based on participants’ accounts of their social worlds and thus susceptible to biases that are inherent to self-reports (e.g., social desirability, impression management, lack of self-knowledge, mood effects, cultural relativity). For example, a recent study of the differential impact of loneliness and social isolation on health (47) brought to light the need for independent measures of subjective (loneliness) and objective (social isolation) aspects of the social environment. Specifically, the authors concluded that the complex nature of social environments necessitates a more complex and multimethod measure of social isolation.

In addition, researchers studying the impact of psychosocial factors on health are often interested in constructs that also influence the reporting of health factors which renders self-report as a measure of both predictors and the outcomes problematic. For example, self-focus and rumination are characteristic depressive symptoms that at the same time significantly affect symptom reporting (48). Using a behavioral observation approach to replace self-reports of either audible depressive behaviors or illness symptoms (e.g. sighing, groaning) could eliminate critical shared method variance. Thus, at a basic level, the EAR provides researchers with an opportunity to assess social processes with a method that is independent of participants’ self-reports.

An example of how this independent assessment can matter comes from a recent cross-cultural EAR study. Ramirez-Esparza and her colleagues (49) collected self-report and observational data on how sociable Americans and Mexicans are. Consistent with prior self-report research, they found that American participants reported being more sociable than their Mexican counterparts. At the same time, though, they spent significantly less time with others and had fewer social conversations. Further, whereas Americans rated themselves significantly higher than Mexicans on the item “I see myself as a person who is talkative”, they spent, in fact, almost 10% less time talking.

Mehl and colleagues (40) revealed a similar self-report versus observation dissociation in their study of male and female talkativeness. In the absence of any gender difference in observed daily word use, women reliably rated themselves higher than men on the item “I see myself as a person who is talkative”. Together, the two studies provide strong evidence for the need to assess culturally embedded social constructs from both the agent’s and the observer’s viewpoint.

More relevant to the health context, in collaboration with Charles Raison, we are currently employing the EAR to study the potential of compassion meditation to change practitioners’ real-world social lives. There is consensus among meditation researchers that a clear demonstration of the real-world, prosocial effects of meditation is crucial for the field. Importantly, however, retrospective and even momentary self-reports cannot unambiguously distinguish change at the level of the self-concept from change at the objective, behavioral level. If participants report having emerged as calmer and more compassionate after meditation training, it is not clear whether this is a result of meditation transforming self-perceptions or due to practitioners, in fact, engaging in more calm and compassionate acts in their daily lives. Counting objectively (i.e. acoustically) documented prosocial and compassionate acts can circumvent this methodological issue and help provide the direct behavioral evidence the field has been seeking. Interestingly, pilot data based on 24 participants suggest that compassion meditation, compared to a control intervention, increased the amount of time participants spent (a) interacting with others, (b) having substantive conversations, (c) laughing, and (d) expressing empathy, suggesting that meditation may in fact have the potential to transform practitioners’ outer social lives independent of (or in addition to) their inner worlds.

Taken together, in providing ecological, behavioral assessments that are independent of self-reports, the EAR can help disentangle subjective from objective aspects of participants’ social environments and thereby facilitate progress in the scientific understanding of how social processes are implicated in health and disease.

Assessment of subtle and habitual social behaviors with important health implications

Asking participants to accurately report on subtle or habitual social behaviors is a task that often goes beyond what self-report can accomplish. For example, mundane behaviors, like sighing or laughing, tend to be particularly difficult to self-report because occurrences become indistinguishable and irretrievable (50). Momentary or end-of-day event diaries can evade memory problems associated with retrospective self-reports, but even in event diaries participants can only report what is noticed (51). In daily life, many subtle and habitual behaviors do not pass this threshold of conscious recognition. Therefore, the study of how these more elusive—yet psychologically non-trivial—aspects of our social lives influence health necessitates observational approaches.

In a recent study that exemplifies how the EAR can capture subtle social behaviors, we tested the degree to which spontaneous sighing is an objective indicator of depression among rheumatoid arthritis patients (19). Thirteen RA patients wore the EAR for two weekends separated by one month. Depression and physical symptoms were assessed via questionnaires. As an “an obvious exaggerated exhalation of breath” (p. 366; (52)), incidents of sighing could be reliably coded from the sampled ambient sounds. Interestingly, and consistent with the notion that sighing is a depression behavior more so than a pain behavior, the number of sighs per hour was significantly and strongly related to patients’ depressive symptoms and non-significantly and less strongly to their experienced pain and flare days. These findings suggest that sighing is a subtle yet observationally traceable social behavior that has clear implications for mental health and therefore warrants further investigation in psychosomatic research.

Other behaviors are less subtle but highly automatic and thus difficult to report (50). For example, swearing’s habitual nature in everyday conversations makes it much less salient than the interaction as a whole (53). In a recent study of the intra- and interpersonal consequences of swearing, we combined data from two pilot studies (20). Thirteen women with RA and 21 women with breast cancer wore the EAR on weekends to track their daily conversations. All sound files were transcribed and submitted to Linguistic Inquiry and Word Count (LIWC; (54)), a computerized text analysis program, to calculate the frequency of spontaneous swearing. In addition, participants completed self-reported measures of depressive symptoms and emotional support at the start of the EAR weekend and several months later at the follow-up. Consistent with the idea that swearing can repel support at the expense of psychological adjustment, swearing in the presence of others, but not alone, was related to decreases in emotional support and increases in depressive symptoms over the study period. Further, decreases in emotional support mediated the effect of swearing on disease-severity-adjusted changes in depressive symptoms. Similar to the findings regarding the sighing-depression link, they highlight the importance of investigating social behaviors that have potentially important health implications but are often too subtle or habitual for participants to report.

Conclusion and Future Directions

As a naturalistic observation method, the EAR eavesdrops on people’s daily lives and provides highly naturalistic, experientially vivid, and psychologically rich information about their moment-to-moment (acoustic) social worlds. This information can be useful for psychosomatic research to determine how objectively traceable aspects of people’s social worlds influence outcomes such as disease progression and psychological adjustment. Within the array of ambulatory monitoring methods (7, 8, 10), the EAR certainly occupies a methodological niche; it is not for everyone and everything. As an observational method, it is labor-intensive and thus requires careful deliberation as to when it should be used instead of more economic, self-report-based EMA methods (55). However, in providing ecological behavioral measures that are independent of self-report and often beyond what self-report can capture, it can yield insights into findings that are difficult to obtain otherwise, and thereby support the field in the mission to identify “basic mechanisms of behavior and social processes” (OppNet Mission Statement).

With the existing possibility of assessing objective and subjective aspects of these behavioral and social processes truly independently—that is with no shared method variance (2)—the most important future direction will be to begin to merge the EAR method with traditional, self-report-based ambulatory monitoring methods. This would create a uniquely powerful way of studying social influences on health in their natural habitat and with the fine-grained, multimethod approach that is afforded by the complexity of the phenomenon.

Specifically, combining self-report-based and observational ambulatory monitoring methods can yield important progress in three ways. First, having time-synchronized momentary self-reports along with the moment-to-moment EAR codings can help contextualize and disambiguate behavioral measures. For example, when studying the health implications of laughing (56), it could help clarify the degree to which an observed spontaneous laughter is an emotional expression and (thus an indicator of positive affect) or a response to the social context (and thus an indicator of sociability)—implicating two distinct psychosomatic pathways (57).

Second, it can help anchor momentary experiences in or calibrate them against measures of social reality. For example, among coping researchers, the question of “just how much cancer intrudes upon and organizes the lives of couples confronted with the disease” (p. 24; (58)) continues to be a contentious issue. In a recent meta-analysis, Hagedoorn and his colleagues recommended that “direct sampling of their interactions and daily experiences that does not presuppose an answer to that question could prove illuminating in this regard.” (p. 24; (58))—and, clearly, it is the joint sampling of daily interactions and experiences that will provide the most informative answer.

Finally, having parallel access to both the perspective of self and the observer can help with theoretically clarifying those social phenomena where systematic discrepancies between the two accounts are an inherent and important part of the phenomenon. For example, because of the operation of self-protective defense mechanisms, narcissists may rate social interactions (momentarily and retrospectively) as engaging, pleasant, or inspiring whereas observers may judge the same interaction to be rejecting, aversive, and offensive (59). Similarly, the consequences that provided social support that is or is not explicitly perceived as such (i.e. invisible support) has for psychological adjustment also necessitates the simultaneous, event-centered assessment of both the objective support delivery and the subjective support appraisal (60, 61). In short, since every social process has an “inner” and an “outer” part, it is often desirable, and at times essential, to capture the two components independently and with equally high measurement fidelity. Often, traditional self-report based and observational ambulatory monitoring will yield comparable findings; at times, they will reveal important discrepancies – both outcomes are scientifically informative.

Abbreviations

AA

Ambulatory Assessment

EAR

Electronically Activated Recorder

EMA

Ecological Momentary Assessment

SECSI

Social Environment Coding of Sound Inventory

LIWC

Linguistic Inquiry and Word Count