Inclusion and Exclusion Criteria

Articles that met the following criteria were included in the review: (1) human-participant research, (2) available in English, (3) quantitative data available for at least one association between a social cognitive determinant (ie, the independent variable) and movement-related behavior (ie, the dependent variable), and (4) within-day or daily EMA study design. Social cognitive determinants were defined as constructs specified within Social Cognitive Theory and the Theory of Planned Behavior, 2 popular social cognitive frameworks in the movement-related behavior literature [7-9,22,39]. Therefore, social cognitive determinants of interest for this systematic review included intentions, attitudes, subjective norms, perceived behavioral control, self-efficacy, outcome expectations, risk perceptions, barriers, facilitators, goals, and plans regarding movement-related behaviors. To focus on naturally occurring associations between social cognitive determinants and movement-related behaviors, articles were excluded if they used an active experimental design. In addition, articles were excluded if they were not published in a peer-reviewed scholarly journal.

Search Process

Literature searches were conducted on May 25, 2022, in the PubMed, SPORTDiscus, and PsycINFO databases to identify relevant articles that used EMA methods to examine associations between social cognitive determinants and movement-related behaviors. No date restrictions were applied in the searches. This process is shown in the PRISMA flow diagram in Figure 1. In total, 3 sets of search terms were used to identify potentially relevant articles. The first set of search terms used general terms, including the following: (“physical activity” OR “exercise” OR “sedentary behavior” OR “movement behavior” OR “physical exercise” OR “sitting”) AND (“ecological momentary assessment” OR “EMA” OR “daily diary” OR “experience sampling”). The second set of search terms included the general terms (from the first search) along with specific terms related to psychological determinants: (“social cognitive” OR “motivation” OR “psychosocial” OR “behavioral cognitions”). The third set of search terms included the general terms from the first search along with specific terms related to social cognitive determinants: (“self-efficacy” OR “outcome expectation” OR “intention” OR “attitude” OR “subjective norm” OR “control” OR “risk perception” OR “barriers” OR “facilitators” OR “goal” OR “plan”). See Multimedia Appendix 1 for the complete search term queries in each database. These specific social cognitive terms were selected by identifying constructs outlined within the Theory of Planned Behavior and Social Cognitive Theory, 2 of the most prominent social cognitive frameworks [6]. The first search using general terms was completed so as to not miss any EMA movement-related behavior articles that may have assessed social cognitive determinants but did not have them as their focus. All articles were collected in Zotero (Corporation for Digital Scholarship) and uploaded to Rayyan (Rayyan Systems Inc), a web tool used to provide support for systematic reviews [40].

Open in a separate window

Figure 1

PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) flow diagram for the literature search on social cognitive determinants, movement-related behaviors, and ecological momentary assessment (EMA) methods.

Article Screening and Coding

A total of 9 reviewers were involved in the process of screening and extracting information from the eligible articles. All reviewers received training on EMA studies of movement-​related behaviors and social cognitive determinants from an expert in this content area (JPM). Following training, all reviewers screened the same 50 abstracts to determine whether the full text should be examined for eligibility. All reviewers then met with the content area expert to discuss and resolve discrepancies. Following this, reviewers independently screened assigned abstracts. One reviewer (KMB) screened all abstracts, and one additional reviewer (KYO, RMR, LB, CM, JY, JMS, BLH, or DJH) was assigned to each abstract. Following screening of all abstracts, discrepancies were discussed and resolved between each pair of reviewers with the content area expert.

The articles retained after title and abstract screening were gathered as full texts. In total, 2 reviewers (KMB and JPM) reviewed the full articles independently to determine eligibility and came together to discuss and resolve any discrepancies. A flowchart using the PRISMA 2020 guidelines shows the screening process for eligible articles (Figure 1) [41].

After identifying the articles to be included in this systematic review, 4 reviewers (KMB, RMR, KYO, and BLH) independently extracted relevant information from 8 articles, met with the content area expert (JPM) to discuss and resolve any discrepancies, and then independently coded the remaining articles. One reviewer (KMB) extracted relevant information from all the articles, and one additional reviewer (RMR, KYO, or BLH) was assigned to each article. Finally, the reviewers and content area expert met to discuss and resolve any discrepancies regarding article information extraction.

For each eligible article, the extracted information focused on study participant characteristics, study design characteristics, study main findings, and methodological quality assessments according to 2 established instruments. First, the Methodological Quality Questionnaire (MQQ) [42] assessed overall study quality based on 9 dimensions (ie, theoretical or conceptual definition, research design, sampling design, sample, evidence of reliability and validity, data analysis, implications for practice, and implications for policy). MQQ scores can range from 0 to 27. Second, the Checklist for Reporting Ecological Momentary Assessment Studies (CREMAS) [43] assessed quality with regard to reporting EMA methodology on 5 dimensions that outline specific criteria that have to be included in the title, introduction, methods, results, and discussion. Across these 5 dimensions, there were 16 items (ie, title and keywords, rationale, training, technology, wave duration, monitoring period, prompting design, prompt frequency, design features, attrition, prompt delivery, latency, compliance rate, missing data, limitations, and conclusions).

CREMAS scores can range from 0 to 16, with 1 point per item addressed. For both the MQQ and CREMAS, higher scores indicate better methodological quality.

Study Sample Characteristics

The analytic sample size of each study ranged from 7 to 198 participants, with a mean sample size of 90.05 (SD 50.70). All articles except for the one by Reifsteck et al [52] (23/24, 96%) reported on studies with samples in which at least 50% of the participants identified as women. The samples reported in 21% (5/24) of the articles were college or university students [52,55,62,63,66]. The samples reported in 50% (12/24) of the articles focused on adults [45-49,53,54,58-61,67]. The samples reported in 29% (7/24) of the articles were older adults (aged ≥50 years) [50,51,56,57,64,65,68]. The mean age of the participants from the 21 unique studies was 47.65 (SD 17.84) years. None of the studies included samples of children or adolescents. Of the 18 articles that reported on participants’ racial or ethnic identities, 4 (22%; n=3, 75% using data from the same study) [47-49] indicated that participants identifying as White or non-Hispanic White did not make up most of the sample [47-49,53].

EMA Protocol

Of the 24 articles, 21 (88%) reported on studies that collected data over 1 wave. A total of 12% (3/24) of the articles (using data from the same study) reported that data were collected across 3 measurement waves spaced over 1 year [47-49]. Regardless of the number of waves, the monitoring period in each study ranged from 4 to 30 days, except for 4% (1/24) of the articles, which sampled participants daily for 2 to 7 months [57]. Regarding EMA prompting design, of the 24 articles, 12 (50%) reported on studies that had 1 prompt per day [46,54,55,58-64,66,67], 3 (12%) reported on studies that had 2 prompts per day (ie, one in the morning and one in the evening) [57,65,68], and the remaining articles (9/24, 38%) reported on studies that delivered between 4 and 8 prompts per day [45,47-53,56].

In total, 38% (9/24) of the articles indicated that EMA prompting was a signal-contingent design (ie, occurring at randomly prompted times) [45-53]. A total of 17% (4/24) of the articles reported on studies that used interval-contingent design (ie, occurring at fixed times) [52,54,56,57], whereas 46% (11/24) of the articles reported on studies that used event-contingent design (eg, self-initiated EMA questionnaire after a specific event) [55,58-65,67,68].

Regarding design considerations to reduce participant burden, the studies reported in 17% (4/24) of the articles did not ask about social cognitive determinants at every EMA prompt to limit the number of items in each prompt [47-49,56]. The studies reported in 8% (2/24) of the articles customized the time of the prompts based on each participant’s wake and sleep schedules [45,57], whereas another study had participants select their own start day to fit their work schedule [61].

EMA protocols were delivered via smartphone (8/24, 33%), PDA (6/24, 25%), or website (10/24, 42%). Among the articles that reported on studies that used a smartphone to deliver EMA protocols, commercially available apps including Personal Analytics Companion (1/24, 4%) [66], movisensXS (2/24, 8%) [50,51], and MyExperience (3/24, 12%) [47-49] were used. A total of 8% (2/24) of the articles reported on studies that had participants complete the study protocol on a smartphone but did not specify the platform used to deliver the EMA prompts [59,67]. PDA devices comprised handheld computers (3/24, 12%) [56,60,68], tablets (2/24, 8%) [58,65], and a wrist-worn Patient-Reported Outcomes Diary device (1/24, 4%) [57]. Articles on studies that used websites either did not specify the website [45,52,53,55,62-64] or did specify it using Prolific [54], Qualtrics [61], or REDCap (Research Electronic Data Capture; Vanderbilt University) [46].

All articles (24/24, 100%) reported participants’ compliance or response rates within the EMA protocol. These rates ranged from 56.9% to 95%, indicating general compliance with the EMA methodology.

Social Cognitive Determinants

Intentions (15/24, 62%) [45,47-51,55,57-60,62,63,65,66] and self-efficacy (14/24, 58%) [46,47,49-51,56-59,64-68] were the most frequently assessed social cognitive determinants. Of the 14 articles categorized as reporting on studies that assessed self-efficacy [57,58], 2 (14%) operationalized their construct of interest as perceived behavioral control; however, perceived behavioral control is generally considered synonymous with self-efficacy [9,69], so the findings were combined into 1 category for this review. Other social cognitive determinants assessed included planning (4/24, 17%) [52,54,59,65], outcome expectations (3/24, 12%) [47,49,66], barriers (2/24, 8%) [53,61], and explicit attitudes (1/24, 4%) [58].

Intentions and self-efficacy were primarily assessed using 1 to 2 items, but 4% (1/24) of the articles reported on studies that used 4 items to assess self-efficacy [46]. The studies reported in 29% (7/24) of the articles adapted social cognitive determinant items from validated scales [46,54,55,58,59,61,65], but almost all (5/7, 71%) reduced the number of items or adapted the time frame of the items for delivery as part of their EMA protocol. Common behavioral targets assessed with social cognitive determinant items included engaging in PA [46,53-56,58-62,64,66-68], engaging in a specified duration of PA [45,47-52,57], or limiting SB to a specified total amount of time [53,63,65,66].

Movement-Related Behaviors

Methods for assessing movement-related behaviors included device-based (15/24, 62%) and self-reported (8/24, 33%) assessments or both (1/24, 4%) [65]. Common devices used included the ActiGraph GT3x or GT3x+ (7/24, 29%) [45,55,58,59,63,66,67], ActiGraph GT2M (3/24, 12%) [47-49], ActiGraph GT1M (1/24, 4%) [53], Fitbit Alta HR (1/24, 4%) [46], and activPAL 3 (3/24, 12%) [50,51,65]. A total of 4% (1/24) of the articles reported on studies that provided participants with either the ActiGraph GT1M or GT3x+ device [68], and 4% (1/24) used the Patient-Reported Outcomes Diary device [57]. In total, 75% (18/24) of the articles reported on studies that required a minimum threshold of valid wear time for data to be included in the analysis [45-53,56,58,​59,61,63,65-68]. Minutes of moderate- to vigorous-intensity PA (MVPA) were the most common operationalization of PA (12/24, 50%) [45,47-49,53,55,56,58,59,63,66,67], and minutes of sedentary time were the most common operationalization of SB (5/24, 21%) [50,51,53,55,65] using device-based measures.

Among the self-reported measures of movement-related behaviors, the studies reported in 25% (6/24) of the articles adapted items from existing validated measures, including the International Physical Activity Questionnaire–Short Form by Sjöström et al [63,65], Godin Leisure-Time Exercise Questionnaire by Godin and Shephard [52,54,61], and a measure of sedentary time in older adults developed by Gardiner et al [65]. Other assessments of daily movement-related behaviors included checklists in which participants indicated the activities they had participated in that day and for how long [56,64]. Bond et al [60] and Rebar et al [66] assessed the daily duration of MVPA using items created for the study.

Methodological Quality

The average MQQ score was 21.38 (SD 4.31; range 14-27), with 21% (5/24) of the articles scoring 27 (the highest possible score). The interrater reliability for the MQQ scores was 98.3%, which is an acceptable level of agreement [55]. On the basis of the 9 criteria of the MQQ, the most frequently omitted information pertained to evidence of reliability and validity provided for the data collected (12/24, 50%) [45,46,53,56-58,60-63,66,67] followed by implications for policy (11/24, 46%) [45,46,53,55-59,61,64,67].

The average CREMAS score was 11.92 (SD 2.21; range 8-15). The interrater reliability for the CREMAS scores was 93.75%. Commonly omitted elements of the CREMAS among applicable articles included participant training procedures, latency, and missing data analysis. Of the 20 articles in which latency was applicable (eg, interval- and signal-contingent designs), none reported latency. A total of 54% (13/24) of the articles did not report any missing data analyses [45,46,52-54,58-61,64-67]. In total, 50% (12/24) of the articles did not report any training to familiarize participants with the EMA protocol [46,49,53-56,58,61,64,66-68]. Except for the studies reported in 12% (3/24) of the articles [47-49], wave duration (ie, the number of data collection waves in a study) was not applicable as all other studies collected data over a single wave.

PA and Social Cognitive Determinants

This systematic review identified studies examining associations between specific social cognitive determinants (ie, intentions, self-efficacy, outcome expectations, planning, perceived barriers, and attitudes) and PA; however, the availability of data regarding these associations differed at the momentary and daily levels. The findings are summarized in the following sections.

Limitations of This Review

Although studies using EMA methods can uncover more nuanced associations between social cognitive determinants and movement-related behaviors by collecting ecologically valid and intensive longitudinal data, the limitations of the studies included in this review should be noted. First, the findings synthesized in this review may not apply to all developmental periods across the life span or individuals of racially or ethnically minoritized backgrounds. No studies included in this review focused on child or adolescent samples, and almost all studies (12/14, 86% that reported race and ethnicity) featured samples in which most participants identified as White or non-Hispanic White. Given that PA levels decline rapidly throughout late childhood and adolescence and racial and ethnic minorities experience physical inactivity–related health disparities [72], EMA may be a critical methodological tool to understand relationships between motivation and behavior in these vulnerable populations. In addition, the studies featured in this review comprised insufficiently active or sedentary individuals as well as those who were sufficiently active; therefore, it is likely that the results are aggregated across individuals at various stages of the behavior change continuum. There is evidence suggesting that social cognitive determinants may differentially regulate behavior across the behavior change continuum, which is an important direction to explore in EMA work to understand movement-related behaviors [24,73]. Further influencing the generalizability of the findings is the lack of missing data analysis in several studies included in this review (12/24, 50%). Such an analysis is essential in EMA studies to determine if data are missing at random or in systematic patterns, which may influence the extent to which documented associations translate to all occasions or all people [43,74].

Most EMA studies included in this review (16/24, 67%) created measures to assess social cognitive determinants or behavior. Although such an approach is common in EMA research as fewer items are used to limit participant burden and fatigue that may result from repeated, intensive assessments [29], the items used in the studies in this review rarely presented psychometric data to establish the validity or reliability of the measures, and the studies often appeared to create items based on face validity. More rigorous and systematic approaches are necessary to develop EMA items to assess social cognitive determinants and movement-related behaviors (for a more in-depth discussion, see the study by Reichert et al [75]).

Furthermore, the limitations of the systematic review itself should be addressed. Although this review assessed the quality of the studies and study reporting through the MQQ and CREMAS, respectively, the information was not used in the interpretation of the results. Neither quality assessment tool specifies thresholds for low-, medium-, or high-quality articles. Therefore, we chose to present the raw scores on each assessment tool. On the basis of those scores, of the 24 articles included in the review, only 5 (21%) scored <18 on the MQQ, and 7 (29%) scored <11 on the CREMAS, which would indicate that those articles satisfied less than two-thirds of the criteria specified within the respective quality assessment tools.

In addition, the inclusion and exclusion criteria of this review did not place any restrictions on the participant sample size necessary for inclusion. A small participant sample size can affect the likelihood of obtaining significant results as well as the magnitude of the association documented. Only 17% (4/24) of the articles included in this review reported on studies that had <50 participants—a threshold determined through simulation to provide adequate power for variance, SE, and fixed-effects estimates at both the between- and within-person levels [76,77]. However, the findings presented in this review primarily focus on within-person findings (or the extent to which social cognitive determinants predict subsequent behavior on a given day or moment). Therefore, power for the analysis is derived from the number of occasions in the analytic sample, and because of the intensive assessments that are a hallmark of EMA, studies typically generate many observations per person. For instance, the study by McDonald et al [57] had the smallest participant sample size of the studies included in the review (N=7) but collected daily assessments from 2 to 7 months, resulting in between 87 and 196 observations per participant. Furthermore, the size of the associations was not considered to fully interpret the relationships documented across the studies as few (7/24, 29%) reported effect sizes. As the volume of studies in this area increases, a meta-analysis should be conducted that accounts for the sample size and effect sizes of relevant studies to further characterize momentary and daily associations between social cognitive determinants and movement-related behaviors.

Future Directions

PA and SB are considered independent health behaviors with different processes regulating each behavior and different health consequences associated with each [78]; however, most of the research investigating relationships between social cognitive determinants and movement-related behaviors focuses on PA. Given the prevalence of excessive SB as well as recent calls to design movement-related behavior interventions to initially focus on reducing SB and over time build to engaging in MVPA [79,80], EMA studies specifically addressing social cognitive determinant–SB relationships are necessary to develop and refine theoretical frameworks to explain and predict SB.

Across both behaviors, intentions and self-efficacy were the most investigated social cognitive determinants. This is not surprising as the Theory of Planned Behavior and Social Cognitive Theory posit that intentions and self-efficacy are proximal determinants of behavior, respectively [7,9,10,39]. However, based on preliminary evidence from this review, momentary self-efficacy and daily planning may also be important motivational determinants of behavior, but this is based on a limited number of studies. Interestingly, the relationships between planning and PA seemed to depend on the type of planning used (eg, action planning vs coping planning), but it is unclear why these differences might exist across microtimescales given the consistent finding across macrotimescales [81]. Perhaps associations between coping planning and movement-related behavior on microtimescales depend on whether an anticipated barrier is encountered in one’s daily experiences. Similarly, although many studies (12/24, 50%) indicated that self-efficacy was assessed, further inspection of items revealed subtle differences in the type of self-efficacy assessed across the studies (eg, task self-efficacy and barrier self-efficacy), which may be differentially related to behavior and are likely important to consider in future EMA research [82,83].

Findings from this review provide evidence that both time-varying and time-invariant factors can moderate associations between social cognitive determinants and subsequent movement-related behaviors at the momentary and daily levels. A recent systematic review of studies investigating the PA intention-behavior coupling across macrotimescales (eg, weeks and months) found that consistent moderators of the intention-behavior coupling included motivational factors such as intention stability, intention commitment, low goal conflict, affective attitude, anticipated regret, perceived behavioral control or self-efficacy, and exercise identity [69]. Although some EMA studies have investigated some of these motivational factors as moderators [49,55], more work is needed to establish whether between-person moderators across macrotimescales also serve as within-person moderators across microtimescales. In addition, to date, only a handful of studies have investigated the moderating role of affective states and physical and social contexts on social cognitive determinant–behavior relationships [48]. There is recent evidence suggesting that these time-varying contextual factors can influence behavior, and it may be that this influence on behavior is through motivational processes [36,37]. Future research should continue to investigate context-sensitive moderators as this work is an essential first step in theory refinement that is sensitive to the contexts that people encounter in their daily lived experiences [84].

This systematic review focused on observational EMA studies to better understand the naturally occurring relationships between social cognitive determinants and subsequent behavior. However, using intensive assessment methods such as EMA and accelerometry to understand how social cognitive determinant–movement-related behavior relationships change over the course of an intervention may be important for developing more effective behavioral interventions. For instance, Basen-Engquist et al [85] had endometrial cancer survivors complete 3- , 10- , and 12-day EMA protocols spaced over the course of a 6-month PA intervention. Although the authors aggregated data across these 3 time points to determine momentary associations among self-efficacy, outcome expectations, and PA, such data could reveal the extent to which associations between social cognitive determinants and subsequent behavior change over the course of an intervention. Such data could also reveal the extent to which behavior change content delivered at a specific point in the intervention is able to affect social cognitive determinant–behavior relationships. Collecting EMA data regarding social cognitive determinant–behavior relationships may help identify potent intervention content in the context of everyday life.