Promoting Exercise Adherence Through Groups

A Self-Categorization Theory Perspective

Mark R. Beauchamp


Exerc Sport Sci Rev. 2019;47(1):54-61. 

In This Article

Abstract and Introduction


The hypothesis presented in this paper is that adherence to exercise programs can be understood, and fostered through intervention, by considering how social identities form within group settings. This paper explains these processes from a self-categorization theory perspective.


Across the age spectrum, involvement in regular physical activity consistently is associated with a number of physical and psychological health benefits. Indeed, there is now compelling evidence to support the use of physical activity as both a means of prevention for, as well as treatment of, a number of chronic diseases such as depression, obesity, diabetes, hypertension, chronic obstructive pulmonary disease, osteoarthritis, and several forms of cancer.[1,2] In addition to the prevention and treatment of chronic disease, physical activity participation at moderate-to-vigorous intensities also bolsters cardiorespiratory fitness, supports the development of bone health and muscular function, aids in the maintenance of a healthy weight status, and supports cognitive function, as well as overall quality of life.[3–5] Interestingly, high levels of moderate-intensity physical activity seem to eliminate the detrimental effects of sedentary behavior (i.e., sitting time).[6] There also is clear evidence of a dose-response relation, such that greater amounts of moderate-to-vigorous physical activity (MVPA) tend to be beneficial, but also that people can benefit from relatively modest increases in physical activity (i.e., moving from inactive to slightly active).[7]

Despite these widely recognized benefits, the results of extensive global surveillance research point to the fact that most people do not meet international guidelines for engaging in health-enhancing physical activity (i.e., at moderate-to-vigorous intensity levels).[8] In North America, for example, less than 20% of adults engage in 150 min or more of MVPA per week when objective assessments are operationalized rather than self-report assessments.[9,10] In light of the prevalence of inactivity, as well as the associated consequences of such inactive lifestyles, there have been considerable attempts to understand the various personal, social, environmental, and policy factors that might be implicated in a) understanding the determinants of sustained physical activity behavior, as well as b) developing evidence-informed interventions that might be scalable on a broad basis. This paper focuses on the social context, and in particular the capacity for groups to foster strong social identities that can be harnessed (through intervention) to foster sustained health-enhancing physical activity behavior. Specifically, this paper draws from the tenets of self-categorization theory,[11–13] to examine how people develop social identities in exercise group settings and how shared social identities and affiliative ties within exercise groups are related to sustained adherence behavior. With a view to supporting health-enhancing physical activity, a discussion also is provided in relation to how social identities can be targeted through intervention, as well as the extant empirical evidence that points to the efficacy/effectiveness of these intervention approaches.

The Group as a Social Platform to Enable and Enhance Physical Activity Participation. Broadly conceived, physical activity represents an umbrella term that subsumes a range of pursuits that include but are not limited to active transportation, occupational activities (work activities that require notable energy expenditure), and exercise. Although people pursue some forms of physical activity (e.g., sports) for social and esteem reasons (among others), a defining feature of "exercise" is that it involves the pursuit of purposeful and planned activities with a view to optimize physical fitness. Although physical activity and exercise often are used interchangeably, the latter is considered a subcategory of the former. This paper is concerned primarily with efforts to promote exercise adherence, although the principles also hold in other settings such as physical activity pursued in the workplace, through active commuting, and so forth.

With a view to understanding some of the conditions likely to support sustained exercise behaviors, a number of systematic and meta-analytic reviews over the past three decades have sought to examine a basic question concerning the extent to which exercising alone or with others in groups is most advantageous for sustaining physical activity behavior.[14–16] These reviews consistently point to the superior capacity of groups to promote exercise adherence. Nevertheless, there seem to be some important caveats to understanding the utility of groups, particularly in exercise settings. That is "not all exercise groups are equal," and that some groups are inherently better at encouraging people to adopt and stick with exercise than others. For example, the psychological climate that exists within groups seems to be particularly salient, whereby groups that are cohesive around the group's task (i.e., instrumental activities and exercises) and social (i.e., forming social bonds) activities tend to produce superior adherence than those groups that are not/less cohesive.[14] Similarly, if members do not feel connected in some meaningful sense to those other members of a specific group, they are less likely to want to stick with that group.[17]

The Social Identity Approach and Self-categorization Theory. Originating from the field of social psychology, the social identity approach subsumes two related, but discernable, psychological theories that correspond to social identity theory[18] and self-categorization theory.[11–13] Although social identity theory preceded self-categorization theory in their respective temporal development, both theories are concerned with the way in which people develop social identities, and the implications associated with the development of these social identities. As highlighted hereinafter, social identity theory focuses to a greater extent on between-group processes,[18] whereas self-categorization theory explicates to a greater extent the psychological mechanisms at play within groups that lead people to develop a sense of "us" and to identify with other group members.[13] Although a considerable body of research over the past four decades has taken a social identity approach, its application to health and exercise psychology[19,20] is quite recent, with Haslam et al.[21] referring to the social identity approach as a "new psychology of health." Haslam et al.[21] specify that a social identity approach to health emphasizes the "importance of social groups for health…. and the importance of people's psychological identification with those groups" (p.10) that collectively provide invaluable directions for intervention that are not addressed by other biomedical, psychological, or social approaches.

Identity can be considered along a continuum. At the one end exists a person's personal identity that reflects the various individual qualities and characteristics that make him or her unique,[13] and at the other end of this continuum exists one's social identity that reflects the extent to which a person's sense of "self" is reflected in the various social groups within which one lives, works, socializes, and psychologically aligns. Specifically, social identity refers to "that part of an individual's self-concept which derives from [her or] his knowledge of [her or] his membership of a social group (or groups) together with the value and emotional significance attached to that membership"[22] (p. 255). According to Tajfel,[22] the extent to which people identify with some social groups, rather than others, shapes their emotional responses, cognitions, and behavioral engagement with those groups.

With its primary focus on intergroup processes, much of the early work using social identity theory[18] examined phenomena such as intergroup conflict, discrimination, and prejudice.[22] The foundational basis for social identity theory[18] derived from Tajfel's[23,24] seminal work on what became known as the minimal group paradigm in which he observed that people tend to act more favorably toward members of their own social groups (also known as "ingroups") and sought to disassociate or discriminate against members of other social groups (also known as "outgroups"). The reference to "minimal" groups is consistent with the notion, and finding, that people tended to derive such discriminatory biases on the basis of minimal, even arbitrary, criteria. For example, within a series of laboratory studies, Tajfel et al.[24] found that people held such discriminatory attitudes even when group membership was derived by virtue of completely random means. Such discriminatory attitudes toward one's own social groups and against other social groups are now widely recognized as a common (if not universal) human response in the context of everyday events. These include people's attitudes and actions related to: a) sports fandom, whereby people tend to align with those who support one's "own" team and disavowal of rival fans[25] (study 1); b) ethnicity, whereby those of minority ethnic groups can become economically and socially disadvantaged relative to the ethnic majority;[26] c) religion, whereby people hold positive views of those who share the same religious beliefs versus those that do not;[27] and politics, whereby people often endorse political decisions derived from one's self-identified political party even if those actions are contrary to one's own self-interests.[28]

With a view to further explaining the manner by which people develop their own social identities in relation to others within various social groups, Turner et al.[11–13] developed self-categorization theory. As a central tenet of self-categorization theory, Turner et al. contended that people categorize themselves and others in relation to a set of core attributes that are salient and observable (e.g., age, gender, ethnicity, socioeconomic status) and that these categorizations shape a) the inferences they make about those others, b) perceptions of others' normative behaviors, and c) their initial attraction to, and engagement toward, those others. According to self-categorization theory, people process the various attributes of people within their (and other) social groups and consider the extent to which they consider themselves as reflecting (or not) the prototypical qualities of those groups. If they consider members of a particular social group to be similar (i.e., "like me") to themselves, they are more likely to feel a greater sense of belonging and social connectivity toward those others. However, if they perceive themselves to be dissimilar (i.e., "not like me"), they are more likely to feel disconnected and disassociated.

As an explanation for why people engage in social categorization, Oakes et al.[29,30] contended that this process enables people to make meaningful representations of diverse stimuli in their social environments that are relevant to action in that setting. They further suggested that this process derives from people's naive knowledge (and lay theories of the world) in which "context" is critical, such that a given category can have distinct implications in different settings. As Oakes et al.[29] state, the purpose of characterization "is to bring together stored knowledge and current input in a form which both makes sense of the world and facilitates our goals within it" (p. 125). As a complement to these ideas, Hogg[31] presented a motivational explanation for self-categorization, such that people engage in self-categorization to reduce uncertainty about things that are perceived to be important.

In terms of how self-categorization takes place and shapes people's behavior within group settings, self-categorization theory highlights a set of key psychological mechanistic processes (see the Figure). The first corresponds to the extent to which an attribute is salient, which is determined by category accessibility and fit.[29,32] Accessibility (also referred to as perceived readiness) reflects those attributes that are readily recalled (i.e., held in memory) and that are valued and considered to be important.[29,32] An attribute may emerge as salient as a result of any number of factors that include early socialization processes (i.e., the values and attributes communicated by significant others such as family, peers, and educational influences), as well as exposure to current norms, expectations, and experiences. For instance, if one considers the self-categorization (or identity) of "being athletic," the extent to which this attribute is salient for a person might depend on those expectations and norms conveyed and made frequently accessible at both home and in school during childhood. As that person progresses through adolescence as well as adulthood, he or she might be exposed to various social agents and motivational conditions that act to further reinforce or even undermine the salience of "being athletic."


Self-categorization processes within exercise groups and adherence behavior.

Fit, on the other hand, is composed of two concepts, namely comparative fit and normative fit.[29,32] Comparative fit takes place as people (perceptually) make sense of differences within and between various social groups. Through a process that is otherwise known as the meta-contrast principle, attributes become highlighted or emphasized as prototypical qualities of a given social group to the extent that similarities within a given social group (i.e., between group members) are evidentially different to those qualities that exist between that group and other social groups. For example, an ex-smoker in an exercise group made up of former smokers might well identify with the attribute of "ex-smoker" to a greater extent if/when exercising in that group than if he or she were to participate in regular exercise group that is not made up of ex-smokers. Prototypical (i.e., characteristic) qualities of groups will vary from one setting to another; however, if/when an individual believes that his or her own attributes or personal qualities align, or fit, with the prevailing attributes of the group (when compared with other groups), then comparative fit takes place.

Normative fit, on the other hand, reflects the extent to which a certain attribute accounts for particular behavioral patterns of group members and one's own perceived alignment with those patterns (i.e., group norms).[29,32] For example, if an older adult believes that the perceived physical competencies of members of a given exercise class are comparable with those of the older adult in question, that person may perceive him or herself to fit well with this group. Conversely, if the members of this exercise class are perceived to be considerably more physically competent, that person may not see him or herself to (normatively) "fit" within that context.

Turner et al.[13] contended that the process of self-categorizing reflects a set of interactions between fit and accessibility, which subsequently influence the emergence of salient social identities. Interestingly, although perceptions of intragroup similarity contribute to the development of social identities (via perceptual processes of "fit"), it is noteworthy that some evidence also exists in support of perceptual similarity deriving from the process of self-categorization and social identity formation (see the Figure). For example, in a series of experimental studies, Haslam and Turner[33] revealed that when people perceive themselves to share identities with other in-group members (in this case via experimental manipulation), they started to assimilate those in-group members' perspectives as their own. Haslam et al.[21] refer to this as a connection hypothesis, such that when people share social identities, they will be more likely to consider themselves as similar to, and feel socially connected with, others. When taken together, the process of social categorization (of self and others) and perceptions of intragroup similarity are inextricably interlinked, with perceptions of intragroup similarity both contributing toward the development of social categorization processes (and salient social identities) and also deriving from social categorization. In sum, the extent to which certain attributes are salient (i.e., highly accessible and a strong sense of comparative and normative fit) and shared (via perceptions of intragroup similarity) shapes one's attraction to and behavioral integration within particular social groups (see the Figure).

As a complement to the concepts of accessibility and fit, a key mechanism of behavioral engagement within social settings corresponds to the role of group cohesion that emerges when people perceive themselves to share category membership.[34] Building on the work of Byrne,[35] a central tenet of self-categorization theory corresponds to the similarity-attraction hypothesis whereby perceived shared category membership acts as the glue that binds members together and contributes to sustained involvement within groups.[34] This is reflected in the common expression, "birds of a feather flock together." A complementary tenet espoused by self-categorization theory corresponds to the norm enactment hypothesis,[21] which purports that if and when people define themselves in terms of a given social identity they will tend to enact those norms associated with that identity. For example, if people self-identify with being a member of a physical activity group, and observe that the norms of that group are to be regularly active, they will try to behave in ways that are consistent with those norms. Why these processes matter from a population health perspective, is that, with knowledge of self-categorization processes, the social environment can be targeted (through intervention) to optimize shared identities related to health-enhancing physical activity that can thereafter result in improved adherence behaviors within social settings.

Observational Evidence From the Exercise Domain. In recent years, a growing body of observational and experimental work has begun to accumulate (by our research group and others) that points to the utility of self-categorization theory in helping us to understand some of the social determinants of adopting and sustaining health-enhancing physical activity. When people's sense of self is highly integrated with the physical activity groups with which they are involved, this bolsters their affiliative ties as well as sustained involvement in those groups.[20,36] In a recent study that nicely illustrates this phenomenon, Stevens et al.[37] examined the extent to which people involved in an exercise setting (parkrun) closely identified with being "a parkrunner," and when they did so they subsequently engaged in behaviors that were highly normative to that group (i.e., sustained participation). Specifically, they found that when people highly identified with being a parkrunner, they perceived their groups to be more cohesive, displayed better objective measures of participation, and were more satisfied (including displays of life satisfaction) than those with lower levels of group identification.

Some evidence has accumulated in recent years that points to certain factors that might be pertinent in fostering social identities associated with exercise group participation. One particularly salient social categorization factor within the exercise domain corresponds to perceptions of intragroup age.[38–40] In one observational study, Beauchamp et al.[38] queried people across the adult age spectrum (from ages 30 to 91 yr) about their preferences to exercise alone as well as in groups made up of people in five age-based categories [aged 30–39, 40–49, 50–59, 60–69, and 70+ yr]. Although there were no differences across the adult age spectrum for preferences to exercise alone, across all five age cohorts, people consistently displayed positive preferences for exercising with people of the same age and a relative dislike of exercising with people either younger or older than themselves (i.e., within different age cohorts). From a theoretical perspective, these findings are consistent with the concept of comparative fit, whereby people ascertain the extent to which they align with one social group relative to other groups (i.e., made up of people markedly different to themselves), and that these relational conceptions shape people's attitudes and preferences. These findings also are important from an empirical and practical perspective as they challenged previous evidence that older adults ostensibly reported a dislike of physical activity within groups,[41,42] which parenthetically resulted in a call to move away from group-based programs involving older adults.[42] In those previous exercise preference studies,[41,42] the researchers did not consider older adults' preferences to exercise with similar others (vis-a-vis their own age) and instead queried older adults about "typical" exercise classes (which tend to be populated by younger adults). In sum, the study by Beauchamp et al.[38] pointed to the importance of considering contextual congruence, and that adults (across the age spectrum) do report positive preferences for group-based activity, provided they tend to be made up of people their own age.

Another particularly salient and observable variable by which people use to categorize one's self and others corresponds to "gender." A number of programs have emerged over the years that provide opportunities for men and women to exercise with those of the same gender, with some qualitative work also pointing to the fact that men and women may display preferences for same-gender settings, providing them with an environment in which they feel most comfortable.[43,44] With this in mind, Dunlop and Beauchamp[45] examined adults' preference for exercising within same-gender (i.e., gender segregated) versus mixed-gender (i.e., gender integrated) groups. Consistent with the similarity-attraction hypothesis, the results revealed that both women (d = 0.76) and men (d = 0.30) reported a stronger preference for same-gender rather than mixed-gender exercise group settings. They also examined weight status as a moderator, on the basis (and consistent with self-categorization theory) that when dissimilarity in physique is coupled with intragroup gender diversity (with diversity considered diametrically opposite to similarity), then one would expect that people who are overweight would feel particularly uncomfortable. For both men and women, adults who were overweight (with a body mass index (BMI) of ≥25) reported a particularly strong preference, relative to adults in the normal-weight range (BMI of 18.5 to 24.99), for exercising within same-gender groups relative to mixed-gender groups, but these effects were particularly pronounced for women (d = 0.42) when compared with men (d = 0.22). Balanced against these findings, a separate study from Australia by Burton et al.[46] involving older adults did not find any evidence for stronger preferences to exercise with others of the same gender. A potential methodological explanation for the differences in these findings is that in the Dunlop and Beauchamp[45] study, participants were queried about their preferences to exercise within group-based settings, whereas Burton et al. study[46] did not. Specifically, they asked participants about their preferences to exercise "with people of my own sex"[46] (p. 4), which could have included preferences to exercise within contexts other than exercise groups (e.g., with a spouse).

Although the previously noted findings shed light on people's preferences for exercising within different social, or group-based, settings, they do not shed any light on the relations between intragroup similarity on these variables and people's behavioral responses (e.g., adherence). A series of studies that examined both perceptions of intragroup similarity and actual intragroup similarity in relation to some of these variables addresses this question. One study involving new mothers in a group-based postnatal exercise program (entitled Fit4Two) sought to examine which variables, including those directly observable as well as those less observable, might be related to subsequent group cohesion and adherence behavior.[39] The study drew from the conceptualization of Harrison et al.,[47] which differentiated between similarity with respect to a) surface level qualities, which reflect "overt, biological characteristics that are typically reflected in physical features" (p. 97), and b) deep-level qualities, which consist of characteristics that require extended personal interactions before they can be ascertained by others (e.g., attitudes, values beliefs). Surface-level similarity was assessed using self-report measures of intragroup similarity with regard to age, ethnicity, and physical condition along with a composite self-report measure of deep-level similarity that included items related to women's "values," "attitudes," and "beliefs." In this study, perceptions of intragroup similarity by new mothers were assessed shortly after group formation (second class) and examined in relation to attendance behavior over the course of 6- to 8-wk exercise programs. The results revealed that perceptions of intragroup similarity with regard to women's respective ages as assessed early in the program prospectively predicted subsequent adherence.[39] Interestingly perceptions of intragroup similarity with regard to ethnicity, physical condition, and the deep-level composite score were not significant predictors of participant attendance.

In another study involving 46 community-based exercise programs (from nine community centers), group members' perceptions of surface level similarity (assessed through items related to age, physical condition, and physical appearance) and deep-level similarity (assessed through items related to values, attitudes, and beliefs) were examined in relation to group cohesion and program attendance.[48] The results revealed that when people perceived themselves to be similar to other group members in terms of their respective physical characteristics (such as age and physical condition), they tended to report higher levels of social cohesion toward the end of their respective exercise program and also displayed improved adherence behaviors when compared with those participants who perceived themselves to be dissimilar to other members of their exercise programs.[48]

Although these studies provide some indication of the extent to which perceptions of intragroup similarity around demographic variables are related to adherence, one of the limitations of using perceptual measures is that individuals who are demographically dissimilar to the rest of the group are less likely to take part in group-based research by completing the requisite questionnaires. To address this problem, Dunlop and Beauchamp[40] used archival measures of intact/complete group data (rather than perceptual measures) based on demographic similarity and exercise class adherence. The researchers found that by calculating Euclidean distance scores (i.e., the absolute average difference between each participant and every other group member), and through use of multilevel modeling, intragroup similarity with regard to participants' ages was related to subsequent adherence behavior.

In addition to research that has examined the extent to which intragroup similarity (related to factors such as age and gender) might be implicated in supporting sustained physical activity involvement, some evidence also has emerged recently that points to the capacity of exercise leaders to foster a sense of identity within physical activity groups.[49] Reicher et al.[50] refer to those in leadership positions with the capacity to nurture a shared sense of social identity among group members as social identity entrepreneurs. Although the majority of research on social identity leadership had been conducted outside of physical activity settings,[51] some evidence points to the fact that those in leadership positions in sport are able to bolster team members' confidence by virtue of enhancing their social identities,[52] and in exercise groups promote attendance behaviors.[49] In the study by Stevens et al.[49] in exercise settings, when exercise group leaders' engaged in social identity leadership, this was related to improved participant adherence behaviors relative to those that did not, and that exercise group members' identification with their respective groups mediated those effects.

When taken together, the results of these studies point to the potential utility of intragroup similarity (with regard to salient attributes) as well as those in leadership positions to foster a sense of unity (i.e., a sense of "us") as means of promoting adherence behaviors within exercise settings. Nevertheless, it is important to note that given the (nonexperimental) observational nature of the study designs, they do not provide any evidence of causality. Some evidence accumulated recently, however, does provide experimental evidence for such effects. This work is examined in the next section.

Experimental Evidence From the Exercise Domain. In this section, the results of two large-scale randomized controlled trials are examined that sought to foster a sense of social identity among participants by promoting intragroup similarity related to some of the variables highlighted previously. The first study corresponds to the Football Fans in Training randomized controlled trial by Hunt et al..[53] This study explicitly targeted middle-aged men who were overweight or obese (n = 747, age range = 35–65 yr), and thereby grounded their intervention within an environment that sought to foster within-group similarity with regard to both gender (all men) and age (middle-aged). The authors delivered the weight loss and healthy lifestyle intervention program within group settings (up to 30 men per group) by community coaching staff within a traditional male setting (British professional men's soccer clubs, n = 13). The intervention was delivered over 12 wk (once per week) and involved a) the provision of advice about diet and physical activity, b) increased focus on physical activity (and shorter classroom sessions) as participants got fitter, c) availability of coaches to provide social support, and d) teaching self-regulation skills (e.g., self-monitoring, goal setting) related to sustaining a healthy diet and physical activity. After the initial 12 wk, the participants received six postprogram e-mail prompts and a group reunion 6 months after the initial sessions had ended. After 12 months (and after adjusting for baseline weight and the nesting of participants within soccer clubs), those in the intervention condition demonstrated significantly greater weight loss (4.94 kg) and percentage weight loss (4.36%) than those who were randomized to a wait-list control group.[53] When taken together, the results point to the efficacy of a health-promotion intervention that clustered participants on the basis of gender and age, delivered through a gender-sensitized environment that was tailored to prototypical British men's social identities. Based on the success of this trial, this approach to weight loss promotion by delivering the intervention within a same-gender context (that harnesses prototypical notions of masculinity) with overweight and obese men has been applied within Canada using ice hockey rather than soccer clubs,[54] and in Australia using Australian Rules Football clubs,[55] although in each instance, those trials have yet to be completed.

The second study corresponds to the GrOup-based physical Activity for oLder adults (GOAL) randomized controlled trial.[56,57] In this trial, older adults (aged ≥65 yr, range = 65–91 yr) were randomized to one of three physical activity programs, delivered within the context of community exercise centers, that included two intervention conditions that were informed by the tenets of self-categorization theory, and a comparison standard group-based exercise condition (i.e., control condition). Specifically, older adults were randomized to similar age same gender (SASG), similar age mixed gender (SAMG), or mixed age mixed gender (MAMG) conditions, with the first two representing the study's experimental conditions, and the latter reflecting the "typical" group-based exercise program that one would receive in North American community/fitness centers. In this comparison "control" condition, older adults in the MAMG condition exercised with younger adult men and women. As the labels of the conditions would suggest the two experimental conditions were delimited to only including older adults, but with one offering same-gender classes and the other offering mixed-gender classes that allowed the researchers to further examine whether within-group gender similarity represents a salient categorizing variable. By virtue of the intervention conditions being informed by self-categorization theory, and explicitly seeking to bolster member's identities within their respective groups, the researchers sought to further foster a sense of social identity among participants by series of additional means. These included providing them with t-shirts (to further foster a sense of "distinctiveness," via the principle of comparative fit), encouraging postworkout gatherings (e.g., coffee and other refreshments) to enable participants to socially connect, and providing class instructors who were themselves older adults. In addition, in the SASG condition, attention was also given to ensure that those instructors were the same gender as class participants. The trial involved 627 older adults with the respective exercise programs lasting 24 wk. The results revealed that older adults randomized to the SAMG (d = 0.47) and SASG (d =0.29) conditions adhered to a greater extent than those in the standard MAMG condition, although there were no differences between the SAMG and SASG condition.[57] When taken together, a few points are worth highlighting. First, the difference in attendance between the SAMG and MAMG conditions equated to a difference of 9.46 more hours of exercise over the 24-wk program. Second, the patterns of adherence across the three conditions were identical for men and women.[57] Third, although age similarity seemed to be a salient categorizing variable in promoting physical activity adherence, gender similarity was not. This suggests that from the perspective of self-categorization theory, in physical activity groups involving older adults, intragroup gender similarity seems to be a less salient categorizing variable than intragroup age. From an intervention perspective, this suggests that those concerned with developing community group-based exercise programs might look to target age as a grouping variable but not necessarily gender, in constructing group-based programs involving older adults.

Future Research Directions. Although the aforementioned observational and empirical evidence points to the utility of physical activity programs that seek to foster members' social identities within groups, considerable research is still required to fully understand the full range, and potential boundary conditions, related to variables/categories that can be targeted to foster a sense of "us" and thereafter exercise adherence behavior. For example, although the results of the GOAL trial point to the importance of within-group similarity with regard to age, the further delineation of older adult groups on the basis of within-group similarity with regard to gender did not yield additional benefits.[57] As a point of comparison, it would be useful to examine the potential effects of intragroup age and gender similarity in relation to adherence using experimental designs with younger age cohorts. For example, with younger adults, is within-group similarity with regard to gender more salient in predicting group cohesion and adherence than it is with older adults?

Furthermore, although research has sought to examine within-group similarity with regard to age, gender, physical condition, and ethnicity, in relation to people's exercise preferences, cohesion, and adherence behaviors, a number of other grouping variables could and should be examined in relation to supporting people's exercise involvement. For example, a growing body of group-based physical activity interventions point to the efficacy of groups that are tailored to specific populations such as those constructed on the basis of disease (e.g., cancer survivors), weight (tailored for those overweight or obese), and income (i.e., low-income adults) status, as well as specific occupations (e.g., firefighters).[58] In a similar regard, Brawley et al.[59] highlighted the efficacy of group-mediated cognitive behavioral (GMCB) interventions, which use groups as a means to help people develop the cognitive behavioral skills required to self-regulate independent physical activity away from the group setting. In most cases, those interventions were delivered with specific (homogenous) groups such as adults involved in cardiac rehabilitation, obese children, and new mothers.[59] Although these programs sought to foster group cohesion as a key mechanism of change, it is entirely conceivable that those GMCB interventions owe some of their success (at least, in part) to the composition of members being similar around salient shared social identities. In the future, it would be particularly worthwhile to examine the extent to which a sense of identity is developed in such physical activity groups and has downstream effects on cohesion and adherence, especially through the use of experimental designs.

Finally, in light of the observational (nonexperimental) research evidence linking social identity leadership to physical activity participation,[49] it would seem particularly worthwhile to examine the extent to which social identity leadership can be trained and developed through intervention and thereafter result in improvements in health-enhancing physical activity behavior among exercise group members. Intervention research involving managers from allied health profession teams (e.g., occupational therapists, radiographers, social workers, etc.) has found that leaders' abilities to engage in social identity leadership improved after training and enhanced certain team-level outcomes such as improved goal clarity.[51] In the context of physical activity promotion, it would be useful to examine the extent to which exercise leaders can be trained through intervention to become social identity leaders and support sustained physical activity participation among group members, as well as the extent to which participants' social identity development (i.e., increased sense of "us") might mediate those effects.