Influence of Friends on Children’s Physical Activity: A Review

Abstract

We examined evidence for friendship influences on children's physical activity (PA) through systematic searches of online databases in May 2012. We identified 106 studies (25 qualitative) published in English since 2000 that analyzed indicators of friendship influences (e.g., communication about PA, friends’ PA, and PA with friends) among persons younger than 19 years.

Children's PA was positively associated with encouragement from friends (43 of 55 studies indicating a positive relationship), friends' own PA (30/35), and engagement with friends in PA (9/10). These findings are consistent with friends influencing PA, but most studies did not isolate influence from other factors that could explain similarity.

Understanding friendship influences in childhood can facilitate the promotion of lifelong healthy habits. PA with friends should be considered in health promotion programs.

Extensive research has linked physical activity to health and inactivity to poor health, especially to obesity and other cardiometabolic problems such as diabetes and cardiovascular disease.^1–3 Increases in overweight, obesity, diabetes, and metabolic syndrome among children and adolescents over the past few decades^4–7 have stimulated a surge of research into possible contributing factors, including insufficient physical activity.^8–10

In spite of its health benefits, physical activity is declining around the world.¹¹ In the United States, levels of physical activity in children and adolescents are not sufficient to promote optimal health.¹² The Centers for Disease Control and Prevention recommends that children and adolescents be active for at least 60 minutes per day, but in 2009, only 18.4% of adolescents reported this much physical activity and 23.1% of adolescents reported not participating in 60 minutes of physical activity on any day.^13,14

A factor that may affect health behaviors that has received increasing attention in recent years is social influence. Research has suggested that family, friends, and peers can guide behaviors such as smoking, sexual initiation, and eating behaviors.^15–19 Influences from these close contacts can occur through social pressure, social modeling and imitation, social comparison, and behavior approximation.²⁰ However, not all social relations are equally important: the closer and stronger the connection, the broader and stronger the possibilities for influence.^21,22 Information transmitted from friends is generally more clearly understood and more likely to be internalized than is information from others, meaning that friends may be more influential than other social contacts,^23–26 such as cliques and peer groups.^24,27,28

Although social influences on physical activity can occur throughout life, they are particularly important to study in children and adolescents, for several reasons. First, childhood and adolescence are a formative period when friends are a primary point of reference^29,30 in deciding which behaviors, values, and attitudes are desirable and which activities warrant effort.^31–34 Friends have been shown to affect timing of first sexual intercourse, drug use, delinquency, and educational attainment among adolescents.^{31,32,35–42} Second, physical activity tracks across life: activity levels during adolescence predict adult levels, and active children and adolescents are more likely to become active adults.⁴³ Third, physical activity often decreases during adolescence, so children and adolescents may be the most important age groups to target for interventions to establish lifelong activity levels.⁴⁴ Finally, the development of successful interventions to increase physical activity in children and adolescents requires an understanding of the unique determinants of physical activity at these ages.

Quantifying the extent to which friends can promote or discourage physical activity is critical for understanding why physical activity levels are decreasing and for developing strategies to promote active lifestyles. We reviewed the literature to determine the role of friends in shaping physical activity among children and adolescents.

METHODS

We conducted a systematic search of peer-reviewed journals in May 2012 in the PubMed, Sociological Abstracts, EconLit, and Web of Science databases. Because patterns of physical activity have changed over time, we focused on studies published since 2000 to identify the most relevant information for today’s youths. Searches used combinations of keywords that included a friendship term (“friend*” or “peer*”) and a physical activity term (“sport*” or “physical activity” or “exercise”). We included the word “peer” in searches because it is sometimes used to mean friend, but only retained the articles where peer referred explicitly to friends rather than to general peer groups or schoolmates. We did not include schoolmates and peer groups not specifically identified as friends because our goal was to quantify the importance of friends, who are more influential than other social contacts, according to the literature.^23–26

Study Selection

We included articles if they (1) were published in English, (2) were published in or after 2000, (3) presented results from analyses of primary or secondary data, (4) used a child or adolescent sample (aged < 19 years or attending secondary school or lower), (5) included a measure of physical activity as the outcome variable, and (6) included as a predictor variable an explicit physical activity measure (communication with friends about physical activity, such as social support, encouragement, or advice; behavior modeling by friends, such as a friend’s own physical activity; or participation with friends in physical activity, such as being teammates in a sport or exercising together). We required that friend variables be analyzed separately from other social contacts, such as family or classmates.

We defined friends as individuals identified as friends by the participant, regardless of whether the friendship was confirmed or reciprocated. We excluded studies that measured only the quality or number of friendships an individual had or the amount of time spent with friends but lacked friendship indicators specific to physical activity. We included studies that collected information from respondents about their friends (e.g., respondent-reported activity levels of friends or attitudes of friends toward physical activity) and that collected data from friends directly as predictor variables. For measures of physical activity, we accepted both self-reported and objective measures (e.g., accelerometer data). Physical activity measures had to pertain to behaviors, such as intensity of activity or activity done in a specific context, not to intentions, such as motivation to engage in physical activity or preoccupation with physical activity. We placed no restrictions on study design.

Searches identified 5357 articles: 2796 in Web of Science, 2424 in PubMed, 79 in Sociological Abstracts, and 58 in EconLit. A review of titles and abstracts narrowed the results to 271 articles, which each author then reviewed independently. The process is illustrated in Figure 1. We identified an additional 5 articles through hand searches of the references of retrieved articles. After all reviews, our analytic sample comprised 106 studies, of which 25 were qualitative studies discussing friends’ influence on physical activity; we reviewed these separately.

FIGURE 1—

Literature search for research on influence of friends on children’s physical activity, 2000–2012.

Data Collection and Analysis

We used a data extraction spreadsheet to collect information about each study, determine whether it met inclusion criteria, and compare across studies. We extracted the following information for each study: authors and year of publication, study design, sample size and participant age range, physical activity outcome variables, friendship exposure variables, control variables, and reported estimated effect sizes and significance levels.

We grouped studies by outcome and exposure measures and analyzed for patterns. We assessed studies’ design, population, and method of data analysis. We considered the control variables used, the size and generalizability of the sample, whether the study used a longitudinal or cross-sectional design, whether physical activity indicators were self-reported or directly measured, and whether information on friends was collected from the index respondent or directly from the friend. These characteristics of the studies meeting inclusion criteria are summarized in Table 1.

TABLE 1—

Summary of Studies of Friends’ Influences on Children's Physical Activity, 2000–2012

Friend Influence Indicator	Study	PA Indicator (Outcome)	PA Objectively Measured	Friend Indicator Objectively Measured	Longitudinal Design	Sample Size	Age Group, Range or Mean	Gender	Location	Covariates	Findings Consistent With Influence
Friends' communication about PA
Friends' support for PAa	Anderson et al.114	VPA				408	13–15 y	Both	US
Friends' support for PAa	Kelly et al.92	VPA	X			1180	Grade 6	Female	US	X	Xb
Friends' support for PAa	Sallis et al.103	VPA				781	Grades 1–12	Both	US	X	Xc
Friends' support for PAa	Sallis et al.103	VPA	X			200	Grades 1–12	Both	US		Xc
Friends' support for PAa	Anderson et al.115	MPA				408	13–15 y	Both	US
Friends' support for PAa	Bergh et al.46	MVPA	X			1129	11 y	Both	Norway	X	X
Friends' support for PAa	Heitzler et al.51	MVPA	X			63	10–17 y	Both	US	X	X
Friends' support for PAa	Hsu et al.118	MVPA				666	Middle school	Both	US	X
Friends' support for PAa	Huang et al.52	MVPA				303	9–14 y	Both	China		X
Friends' support for PAa	Hamilton and White117	MVPA				423	12–16 y	Both	US	X
Friends' support for PAa	Kitzman-Ulrich et al.119	MVPA	X			669	Grade 6	Both	US
Friends' support for PAa	Kelly et al.92	MVPA	X			1180	Grade 6	Female	US	X	Xb
Friends' support for PAa	Wenthe et al.107	MVPA	X			205	12–14 y	Both	US	X
Friends' support for PAa	Patnode et al.106	MVPA	X			294	10–17 y	Both	US	X	Xd
Friends' support for PAa	Davison and Schmalz47	Overall PA	X			202	Grades 6–8	Both	US	X	X
Friends' support for PAa	Duncan et al.49	Overall PA	X			371	12–17 y	Both	US	X	Xd
Friends' support for PAa	Lubans and Morgan120	Overall PA	X			119	14–15 y	Both	Australia	X
Friends' support for PAa	Loucaides58	Overall PA				652	Grades 7–9	Both	Cyprus		X
Friends' support for PAa	Kirby et al.109	Overall PA			X	641	11–15 y	Both	Scotland	X	Xe
Friends' support for PAa	De Bourdeaudhuij et al.48	Overall PA	X			6078	11–19 y	Both	Belgium	X	X
Friends' support for PAa	Raudsepp and Viira62	Overall PA			X	193	Grade 6	Female	Estonia	X	X
Friends' support for PAa	Roesch et al.63	Overall PA			X	878	11–15 y	Both	US		Xf
Friends' support for PAa	Taymoori et al.122	Overall PA				558	12–17 y	Female	Iran
Friends' support for PAa	Wu et al.72	Overall PA				832	12–15 y	Both	Taiwan		X
Friends' support for PAa	Wenthe et al.107	Overall PA				205	12–14 y	Both	US	X	Xd
Friends' support for PAa	Duncan et al.87	Overall PA	X		X	371	12–17 y	Both	US	X
Friends' support for PAa	Zhang et al.78	Overall PA				285	12–15 y	Both	US	X	X
Friends' support for PAa	Ievers-Landis et al.53	WBPA				354	8–11 y	Female	US	X	X
Friends' support for PAa	Leslie et al.94	Active travel to/from school				2961	10–14 y	Both	Australia	X	Xgd
Friends' support for PAa	Ommundsen et al.55	Active commute to school				760	9–15 y	Both	Norway	X	X
Friends' support for PAa	Ommundsen et al.55	Active commute to school				760	9–15 y	Both	Norway	X	X
Friends' support for PAa	Ommundsen et al.55	Participates in active games at school				760	9–15 y	Both	Norway	X	X
Friends' support for PAa	Jago et al.91	MVPA after school, follow-up	X		X	657	10–11 y	Both	UK	X	X
Friends' support for PAa	Jago et al.91	MVPA on weekend, follow-up	X		X	657	10–11 y	Both	UK	X	Xd
Friends' support for PAa	Anderson et al.115	Participates in sports team				1139	8–15 y	Both	US
Friends' support for PAa	Wenthe et al.107	Bouts of MVPA	X			205	12–14 y	Both	US	X
Friends' support for PAa	Lytle et al.111	MVPA before school, mean min				3378	Grade 8	Female	US	X	X
Friends' support for PAa	Lytle et al.111	MVPA after school, mean min	X			3378	Grade 9	Female	US	X	X
Friends' support for PAa	Lytle et al.111	Weekend MVPA, min				3378	Grade 10	Female	US	X	X
Friends' support for PAa	Lytle et al.111	Total MVPA out of school, min	X			3378	Grade 11	Female	US	X	X
Friends' support for PAa	Dunton et al.124	Change in VPA			X	63	Grades 10–11	Female	US	X
Friends' support for PAa	Dunton et al.124	Change in MVPA			X	63	Grades 10–11	Female	US	X
Friends' support for PAa	Prochaska et al.102	Overall PA				138	Grades 6–8	Both	US	X	X
Friends' support for PAa	Prochaska et al.102	Overall PA	X			138	Grades 6–8	Both	US	X
Friends' support for PA	Ha et al.79	Overall PA				1538	11–16 y	Both	Hong Kong, Australia
Friends' support for PA	Wu and Pender82	Overall PA				832	12–15 y	Both	Taiwan
Friends' support for PA	Hohepa et al.110	After school PA				3471	12–18 y	Both	New Zealand	X	X
Friends' support for PA	Hohepa et al.110	Lunchtime PA				3471	12–18 y	Both	New Zealand	X	X
Friends' support for PA	Hohepa et al.110	Active commute to school				3471	12–18 y	Both	New Zealand	X	X
Friends' support for PA	Johnson et al.54	Sports team participation				145	Grades 9–12	Both	US	X	X
Change in friends' support for PA	Raudsepp and Viira62	Change in PA			X	193	Grade 6	Female	Estonia	X	X
Change in friends' support for PA	Jago et al.91	Change in MVPA after school	X		X	657	10–11 y	Both	UK	X	Xh
Change in friends' support for PA	Jago et al.91	Change in MVPA on weekend	X		X	657	10–11 y	Both	UK	X	Xh
Best friend's support for PA	Sabiston and Crocker64	Overall PA				857	15–18 y	Both	Canada		X
Best friend's support for PA	Sabiston and Crocker64	VPA				857	15–18 y	Both	Canada		X
Best friend's support for PA	Sabiston and Crocker64	MVPA				857	15–18 y	Both	Canada		X
Friend influences sport participation	Seabra et al.68	Sport team participation				3352	10–18 y	Both	Portugal	X	X
Friend influences PA	Seabra et al.76	High-intensity PA				3352	10–18 y	Both	Portugal	X	X
Friend influences PA	Seabra et al.76	Moderate-intensity PA				3352	10–18 y	Both	Portugal	X	X
Best friend encourages PA	Wilson and Dollman108	Overall PA				180	Grades 6–9	Male	Australia	X
Best friend encourages PA	Wilson and Dollman123	Overall PA				113	Grades 8–9	Male	Australia	X
Best friend encourages PA	Wilson and Dollman108	MVPA				180	Grades 6–9	Male	Australia	X
Best friend encourages PA	Wilson and Dollman123	MVPA				113	Grades 8–9	Male	Australia	X
Best friend encourages PA	Wilson and Dollman108	VPA				180	Grades 6–9	Male	Australia	X
Best friend encourages PA	Wilson and Dollman123	VPA				113	Grades 8–9	Male	Australia	X
Friend encourages PA	Sharma et al.69	Overall PA				718	Grade 6	Female	US	X	X
Friend encourages PA	Sharma et al.70	Overall PA				718	Grade 6	Female	US		X
Friend encourages PA	Sharma et al.69	Jumping activities, d/wk				718	Grade 6	Female	US	X	X
Friend encourages PA	Sharma et al.70	Jumping activities, d/wk				718	Grade 6	Female	US		X
Friend encourages PA	Springer et al.71	MVPA				718	10–14 y	Female	US	X	X
Friend encourages PA	Anderson et al.115	MVPA	X			1139	8–15 y	Both	US		X
Friend encourages PA	Springer et al.71	VPA				718	10–14 y	Female	US	X	X
Friend encourages PA/sports	Lemstra et al.113	≥ 60 min MVPA 7 d/wk				4197	9–15 y	Both	Canada	X	X
Friend encourages PA/sports	Prochaska et al.102	Overall PA				138	Grades 6–8	Both	US		X
Friend encourages PA/sports	Prochaska et al.102	Overall PA	X			138	Grades 6–8	Both	US
Friend's feedback on health behavior	Stanford and McCabe104	Frequency of changing exercise to decrease weight				362	12–14 y	Male	Australia	X	Xi
Best friend encourages PA	Martín-Matillas et al.96	MPA				3007	12.5–17.5 y	Both	Europe	X	X
Best friend encourages PA	Martín-Matillas et al.96	VPA				3007	12.5–17.5 y	Both	Europe	X	X
Best friend encourages PA	Martín-Matillas et al.96	MVPA				3007	12.5–17.5 y	Both	Europe	X	X
Best friend encourages PA	Martín-Matillas et al.96	Overall PA				3007	12.5–17.5 y	Both	Europe	X	X
Friend discourages PA	Kitzman-Ulrich et al.119	MVPA	X			669	Grade 6	Both	US
Friends talk about PA positively	Coppinger et al.86	Overall PA	X		X	106	9–11 y	Both	UK		X
Friends remind to be active	Coppinger et al.86	Overall PA	X		X	106	9–11 y	Both	UK		X
Discuss PA with friends	Finnerty et al.50	Overall PA	X			315	10–13 y	Both	UK		X
Friends do not tease for not being good at sports	Lemstra et al.113	≥ 60 min MVPA 7 d/wk				4197	9–15 y	Both	Canada	X	X
Friends praise sports abilities	Lemstra et al.113	≥ 60 min MVPA 7 d/wk				4197	9–15 y	Both	Canada	X
Being thin (girls) or muscular (boys) is important to friends	Kahn et al.97	Change in MVPA			X	12 812	10–18 y	Both	US	X
Pressure from friends	de la Haye et al.88	Change in MVPA			X	378	12–15 y	Both	Australia		X
Friends not supportive of PA (perceived)	Coppinger et al.86	Overall PA	X		X	106	9–11 y	Both	UK
Friends encourage walking or biking to school	Panter et al.100	Bike to school				2012	9–10 y	Both	UK	X	Xj
Friends encourage walking or biking to school	Panter et al.100	Walk to school				2012	9–10 y	Both	UK	X
Friends criticize PA	Price et al.101	WBPA				1000	9–12 y	Female	US	X	X
Friends praise PA/sports	Prochaska et al.102	Overall PA				138	Grades 6–8	Both	US		X
Friends praise PA/sports	Prochaska et al.102	Overall PA	X			138	Grades 6–8	Both	US
Friends praise PA/sports	Viira and Raudsepp84	MVPA			X	197	Grades 7–8	Both	Estonia	X	X
Friends' modeling of PA
Mean overall PA of friends	Macdonald-Wallis et al.60	Overall PA	X	X		559	10–11 y	Both	UK	X	X
Mean MVPA of friends	Macdonald-Wallis et al.60	MVPA	X	X		559	10–11 y	Both	UK	X	X
Friends ever do PA	Martín-Matillas et al.95	Overall PA				2260	13–18.5 y	Both	Spain	X
Friends do PA now	Martín-Matillas et al.95	Overall PA				2260	13–18.5 y	Both	Spain	X	Xh
Close friends doing PA, No.	Loucaides et al.59	MVPA				2688	15.62 y	Both	Canada	X	X
Best friend's PA	Martín-Matillas et al.96	MPA				3007	12.5–17.5 y	Both	Europe	X	X
Best friend's PA	Martín-Matillas et al.96	VPA				3007	12.5–17.5 y	Both	Europe	X	X
Best friend's PA	Martín-Matillas et al.96	MVPA				3007	12.5–17.5 y	Both	Europe	X	X
Best friend's PA	Loucaides et al.112	MVPA				256	Grade 6	Both	Cyprus	X
Best friend's PA	Martín-Matillas et al.96	Overall PA				3007	12.5–17.5 y	Both	Europe	X	X
Friends’ PA	Ali et al.45	Exercised ≥ 3 times in past 7 d		X	X	3898	12–18 y	Both	US	X	X
Friends' PA	Ali et al.45	Plays active sports		X	X	3898	12–18 y	Both	US	X	X
Friends' PA	de la Haye et al.75	Participation in organized PA		X		385	Grades 8–9	Both	Australia	X
Friends' PA	de la Haye et al.75	Participation in non-organized PA		Xk		385	Grades 8–9	Both	Australia	X	X
Friends' PA	de la Haye et al.75	Change in MVPA		Xk	X	378	12–15 y	Both	Australia	X	X
Friends' PA	de la Haye et al.75	Change in MVPA			X	378	12–15 y	Both	Australia		X
Friends' PA	Duncan et al.87	Change in PA			X	371	12–17 y	Both	US	X	Xd
Friends' PA	Gesell et al.74	Change in percentage of after school play that is MVPA	X	X	X	81	5–12 y	Both	US	X	X
Friends' PA	Duncan et al.87	Overall PA	X			371	12–17 y	Both	US	X	Xh
Friends' PA	Schofield et al.26	Overall PA	X	X		318	15–18 y	Female	Australia	X	Xk
Friends' PA	Sharma et al.70	Overall PA				718	Grade 6	Female	US	X	X
Friends' PA	Voorhees et al.105	Overall PA				488	Grades 6 and 8	Female	US	X
Friends' PA	Luszczynska et al.81	Overall PA				2387	13–20 y	Both	Hungary, Poland, Turkey, US		X
Friends' PA	Bramoullé et al.83	Overall PA			X	not listed	Grades 6–12	Both	US	X
Friends' PA	Plotnikoff et al.77	Overall PA (energy expenditure)				2688	Grades 9–12	Both	Canada		X
Friends' PA	McDermott et al.61	VPA				1407	9–13 y	Both	US		X
Friends' PA	King et al.56	VPA				535	Grades 9–12	Both	US		X
Friends' PA	King et al.56	MPA				535	Grades 9–12	Both	US		X
Friends' PA	Springer et al.71	MVPA				718	10–14 y	Female	US	X	X
Friends' PA	Springer et al.71	MVPA				718	10–14 y	Female	US	X	X
Friends' PA	Price et al.101	WBPA				1000	9–12 y	Female	US	X	X
Friends' PA	Sharma et al.70	Jumping activities, d/wk				718	Grade 6	Female	US	X	X
≥ one third of best friends engage in school sports or hard PA	Gilmer et al.116	Overall PA				113	11–14 y	Both	US
Friends' MVPA	Jago et al.90	MVPA min/d	X	X		472	10–11 y	Both	UK	X	Xdk
Friends' overall PA	Jago et al.90	Overall PA in CPM	X			472	10–11 y	Both	UK	X
Change in friends' PA	Duncan et al.87	Change in PA			X	371	12–17 y	Both	US	X	Xh
≥ 3 friends are active	Leatherdale et al.57	Moderately active				2379	Grades 5–8	Both	Canada	X	X
≥ 3 friends are active	Leatherdale et al.57	Highly active				2379	Grades 5–8	Both	Canada	X	X
≥ 3 friends are active	Schofield et al.26	Overall PA	X			318	15–18 y	Female	Australia		X
All friends are physically active	Page et al.98	Does muscle-strengthening exercise				1886	16.5 y	Both	Hungary, Romania, Czech Republic	X	Xg
All friends are physically active	Page et al.98	Participates in team sports				1886	16.6 y	Both	Hungary, Romania, Czech Republic	X	Xgd
All friends are physically active	Page et al.98	Is active outside of school				1886	16.7 y	Both	Hungary, Romania, Czech Republic	X	Xg
All friends are physically active	Page et al.98	Overall PA				1886	16.8 y	Both	Hungary, Romania, Czech Republic	X	Xg
≥ 3 friends do sports	Keresztes et al.93	Leisure-time sports				548	10–15 y	Both	Hungary		X
All/most friends are active	Page et al.99	VPA				2519	Grades 7–12	Both	Thailand	X
All/most friends are active	Page et al.99	Does muscle-strengthening exercise				2519	Grades 7–12	Both	Thailand	X	Xjd
All/most friends are active	Page et al.99	Participates in team sports				2519	Grades 7–12	Both	Thailand	X	Xj
All/most friends are active	Page et al.99	Is active outside of school				2519	Grades 7–12	Both	Thailand	X	Xj
All/most friends are active	Page et al.99	Overall PA				2519	Grades 7–12	Both	Thailand	X	Xj
All/most friends are active	Page et al.99	VPA				2519	Grades 7–12	Both	Thailand	X
Best friend engages in active play	Wilson and Dollman108	Overall PA				180	Grades 6–9	Male	Australia	X
Best friend engages in active play	Wilson and Dollman123	Overall PA				113	Grades 8–9	Male	Australia	X
Best friend engages in active play	Wilson and Dollman108	MVPA				180	Grades 6–9	Male	Australia	X
Best friend engages in active play	Wilson and Dollman123	MVPA				113	Grades 8–9	Male	Australia	X
Best friend engages in active play	Wilson and Dollman108	VPA				180	Grades 6–9	Male	Australia	X
Best friend engages in active play	Wilson and Dollman123	VPA				113	Grades 8–9	Male	Australia	X
All friends are active	Page et al.98	VPA				1886	16.5 y	Both	Hungary, Romania, Czech Republic	X	X
All friends are active	Page et al.98	Does muscle-strengthening exercise				1886	16.5 y	Both	Hungary, Romania, Czech Republic	X	X
All friends are active	Page et al.98	Participates in team sports				1886	16.5 y	Both	Hungary, Romania, Czech Republic	Xd	X
All friends are active	Page et al.98	Is active outside school				1886	16.5 y	Both	Hungary, Romania, Czech Republic	X	X
All friends are active	Page et al.98	Overall PA				1886	16.5 y	Both	Hungary, Romania, Czech Republic	X	X
Friends' use of facilities for PA	Ries et al.121	MVPA	X			327	Grades 9–12	Both	US	X
Best friend does PA	Sabiston and Crocker64	Overall PA				857	15–18 y	Both	Canada		X
Best friend does PA	Sabiston and Crocker64	VPA				857	15–18 y	Both	Canada		X
Best friend does PA	Sabiston and Crocker64	MVPA				857	15–18 y	Both	Canada		X
All/most friends exercise	Saxena et al.67	VPA				305	12–21 y	Female	US	X	X
In sports team or class with friend	Voorhees et al.105	Overall PA				488	Grades 6 and 8	Female	US	X
Invited friend to do PA	Voorhees et al.105	Overall PA				488	Grades 6 and 8	Female	US	X
Friends invite to be active together	Voorhees et al.105	Overall PA				488	Grades 6 and 8	Female	US	X
Friends' use of facilities for PA	Ries et al.121	MVPA	X			327	Grades 9–12	Both	US	X
Number of active friends	Leggett et al.73	Moderately active				31 202	Grades 9–12	Both	Canada	X	X
Number of active friends	Leggett et al.73	Active				31 202	9–12 Grades	Both	Canada	X	X
Friends' overall sport participation	Denault and Poulin80	Overall sport participation			X	272	Grade 6	Both	Canada	X	X
Participation in PA with friends
PA with friends	Lemstra et al.113	≥ 60 min MVPA 7 d/wk				4197	9–15 y	Both	Canada	X
PA with friends	Voorhees et al.105	Overall PA				488	Grades 6 and 8	Female	US	X	X
PA with friends	Fermino et al.89	MVPA (≥ 60 min ≥ 1 d/wk)				1518	14–18 y	Both	Brazil	X	X
PA with best friend	Wilson and Dollman108	Overall PA				180	Grades 6–9	Male	Australia	X	X
PA with best friend	Wilson and Dollman123	Overall PA				113	Grades 8–9	Male	Australia	X
PA with best friend	Wilson and Dollman108	MVPA				180	Grades 6–9	Male	Australia	X
PA with best friend	Wilson and Dollman123	MVPA				113	Grades 8–9	Male	Australia	X
PA with best friend	Wilson and Dollman108	VPA				180	Grades 6–9	Male	Australia	X	X
PA with best friend	Wilson and Dollman123	VPA				113	Grades 8–9	Male	Australia	X
Friends invite to be active together	Lemstra et al.113	≥ 60 min MVPA 7 d/wk				4197	9–15 y	Both	Canada	X
Friend present during PA	Salvy et al.65	PA intensity				20	12–14 y	Both	US	X	X
Friend present during PA	Salvy et al.66	Distance biked	X			88	12–14 y	Both	US	X	X
Friends change schedule to exercise/do PA together	Coppinger et al.86	Overall PA	X		X	106	9–11 y	Both	UK		X
PA/sports with friend	Prochaska et al.102	Overall PA				138	Grades 6–8	Both	US		X
PA/sports with friend	Prochaska et al.102	Overall PA	X			138	Grades 6–8	Both	US
PA with friends	Fermino et al.89	MVPA (≥ 60 min ≥ 5 d/wk)				1518	14–18 y	Both	Brazil		X
Exercise with friends	Finnerty et al.50	Overall PA	X			315	10–13 y	Both	UK		X
Friends plan PA together	Coppinger et al.86	Overall PA	X	X		106	9–11 y	Both	UK		X
Weekly PA with friend	Jago et al.90	MVPA min/d	X			472	10–11 y	Both	UK	X	Xhl
Weekly PA with friend	Jago et al.90	Overall PA in CPM	X			472	10–11 y	Both	UK	X	Xhm
On sports team/in PA class with friend	Jago et al.90	MVPA min/d	X			472	10–11 y	Both	UK	X
On sports team/in PA class with friend	Jago et al.90	Overall PA in CPM	X			472	10–11 y	Both	UK	X
Social affiliation, do PA to be with friends	Jago et al.85	Light PA	X			131	10–11 y	Both	Finland
Social affiliation, do PA to be with friends	Jago et al.85	MVPA	X			131	10–11 y	Both	Finland
Social affiliation, do PA to be with friends	Jago et al.85	Overall PA	X			131	10–11 y	Both	Finland
Neighborhood affiliation, do PA to be with friends	Jago et al.85	Light PA	X			131	10–11 y	Both	Finland
Neighborhood affiliation, do PA to be with friends	Jago et al.85	MVPA	X			131	10–11 y	Both	Finland
Neighborhood affiliation, do PA to be with friends	Jago et al.85	Overall PA	X			131	10–11 y	Both	Finland

Note. CPM = counts per minute; MPA = moderate-intensity PA; MVPA = moderate-to-vigorous PA; PA = physical activity; VPA = vigorous-intensity PA; WBPA = weight-bearing PA.

^aSallis et al. Social support scale.¹²⁵

^bFor Blacks and Whites but not Hispanics.

^cOnly in some grades for girls and for boys.

^dBoys only.

^eOnly medium level of support, compared with low support.

^fIntervention study.

^gCompared with no physically active friends.

^hGirls only.

ⁱFor the purpose of increasing muscle but not for losing weight.

^jCompared with no or some physically active friends.

^kDistance < 1 km.

^lBest friend only.

^mFor ≥ 5 times/week.

RESULTS

Eighty-one quantitative articles met inclusion criteria. Of these studies, 40 reported significant positive associations between a measure of friendship influences and respondents’ physical activity in all populations and models,^45–84 30 reported significant positive associations in some but not all populations or models,^26,85–113 and 11 reported no significant positive associations.^114–124

The types of physical activity considered in this literature can be grouped as overall physical activity, defined as a measure of the respondents’ total activity of all types over a given period (n = 41)^{26,45,47–50,57,58,60,62–64,69,70,72,73,77-79,81–83,85–87,89,90,95,96,98,99,102,105,107–109,111,116,120,122,123}; engagement in specific activities, such as weight-bearing physical activity and sports team participation (n = 11)^{53,54,68–70,80,98,99,101,114,123}; physical activity in specific contexts, such as leisure time or commuting to and from school (n = 10)^{93,94,97–100,68,110,111,123}; physical activity performed at a specific intensity, such as vigorous or moderate (n = 35)^{46,51,52,55,56,59–61,64,65,67,71,74,76,84,85,90–92,96,98,99,103,106–108,112–114,116–118,120,122,124}; and change in any of these over time (n = 7).^{62,74,75,87,91,97,124} Vigorous-intensity physical activities are those that have an assigned metabolic expenditure value greater than 6 (requires a large amount of effort and causes rapid breathing or a substantial increase in heart rate); moderate-intensity physical activities have an assigned metabolic expenditure of 3 to 6 (requires a noticeable amount of effort and raises the heart rate).¹²⁶

We identified a priori and confirmed empirically from the literature 3 broad mechanisms through which friends could influence physical activity: friends’ communication through social norms and conversations about physical activity (n = 55), including positive messaging, such as encouragement and social support (n = 54)^{46–55,58,62–64,68–72,75,76,78,79,82,84,86,87,90–92,94,96,97,100,102-111,114,115,117–120,122,123,113} and negative messaging, such as criticism and teasing (n = 4)^{86,91,101,119}; friends’ modeling of physical activity through their own physical activity, that is, friends’ own behavior that might provide guidance, enticement, or pressure for the respondent (n = 35)^{26,45,56,57,59–61,64,67,69–71,73–75,77,80,81,83,85,87,88,90,93,95,96,98,99,101,105,108,112,116,121,123}; and participation in physical activity with friends, including organized sports and recreational play in the company of friends (n = 10).^{50,65,66,89,90,102,105,108,113,123} We organized our findings according to these proposed pathways of influence.

Communication About Physical Activity

Communication about physical activity most frequently consisted of social support or friends’ encouragement for physical activity. The items used to measure social support generally described 1 or more of the following behaviors: friends reminding the respondent to exercise, encouraging the respondent to engage in physical activity, praising the respondent for engaging in physical activity, or discussing physical activity with the respondent. Other indicators of communication used in studies (but not referred to as social support) included the respondent’s perceptions of friends’ attitudes and beliefs about physical activity (such as the importance of physical activity to the friend), respondent-perceived influence from friends over their participation in physical activity, and invitations to or from a friend to engage in physical activity together. Twenty-nine studies used the Sallis et al. social support scale¹²⁵ to measure friends’ influence.^{46,47,49,51–53,55,58,62,63,72,78,87,91,92,94,102,103,106,107,109,114,117–120,122,124} This scale incorporates measures of peer encouragement and support for physical activity, as well as 2 measures of participation in physical activity together: friends exercising with the respondent and friends changing their schedule to be active with the respondent.¹²⁵ We categorized separate measures of friends exercising with the respondent and of friends changing their schedule to exercise with the respondent as a combined measure of participation in physical activity with a friend.

Of the 55 studies that investigated communication about or social support for physical activity, 25 showed significant positive relationships in all models^{46–55,58,62–64,69–72,75,76,78,79,82,84,96,}; 18 showed significant associations in some models or some groups within the population^{55,86,87,91,92,94,97,100,102–104,106,107,109–111,113,124}; and 12 did not show any significant associations^{90,101,105,108,114,115,117–120,122,123} (including 1 focused on negative messages¹⁰¹). One of the studies with mixed results yielded an unexpected relationship: lower friend social support at the beginning of an intervention was associated with a greater subsequent increase in physical activity.¹²⁴

The most common measures of physical activity in these studies were overall physical activity (n = 27)^{47–50,58,62–64,69,70,72,78,79,82,86,87,90,96,102,105,107–109,111,120,122,123} and moderate-to-vigorous physical activity (n = 19).^{46,51,52,64,76,84,91,92,97,103,106,107,113-115,117-119,124} Overall physical activity was more often associated with friends’ communication than was activity intensity (21/27 vs 14/23). Eleven of the 55 studies measured physical activity directly (with an accelerometer, a pedometer, or observation).^{46,48,50,51,90–92,106,111,118,119} Thirty-six studies used self- or parent reports,^{52–55,58,62–64,68–72,75,76,78,79,82,84,94,97,100,101,103–105,108–110,113,114,117,118,122,123,124} and 6 studies combined indirect and direct measures.^{49,86,87,96,102,103} Studies that used direct measures, or a combination of direct and indirect measures, found significant associations (at least some support in 10/18^{46,48,50,51,68,86,91,92,106,111}) less frequently than did studies that used self- or parent reports of physical activity (at least some support in 28/36^{47,52–55,58,62–64,68,69–72,75,76,78,79,82,84,94,97,100,103,104,109,110,113}).

Four studies investigated negative messaging from friends about physical activity, specifically criticism for being active or exercising, making fun of the respondent for being active, declining an invitation from the respondent to be active together, and the respondent’s belief that friends would not be supportive of being active.^{86,91,101,119} None of these forms of negative messaging were significantly associated with respondent’s physical activity, which was measured most often in these studies with objective measures (e.g., pedometer, accelerometer).^86,91,119

Modeling of Physical Activity

Thirty-five studies focused on friends’ demonstrated or perceived physical activity, which we called friends’ modeling of physical activity. Only 9 of these obtained physical activity data directly from respondents’ friends.^{26,45,60,74,75,80,83,88,90} Nineteen of the 35 studies showed significant positive associations between friends’ physical activity and respondents’ physical activity across specifications,^{45,56,57,59–61,64,67,69–71,73–75,77,80,81,83,101} and 11 reported positive associations only in some specifications.^{26,85,87,88,90,93,95,96,98,99,112} Five studies did not find significant relationships between measures of friends’ physical activity and respondents’ physical activity.^{105,108,116,121,123}

The most commonly used measures of physical activity explored in the context of friendship modeling were overall physical activity (at least some support in 16/21)^{26,45,57,64,69,70,74,77,80,81,83,87,95,96,98,99} and physical activity at a specific intensity (at least some support in 12/16).^{56,59,61,64,67,71,73,85,90,96,98,112} Studies that used objective measures of physical activity yielded significant results less often than studies that used self-reported measures.

Joint Participation in Physical Activity

Ten studies investigated the importance of being physical active with friends or in the presence of friends for respondents’ physical activity levels.^{50,65,66,89,90,102,105,108,113,123} Nine of these reported significant associations in at least 1 model^{50,65,66,89,90,102,105,108,113} (4 in all models^50,65,66,113). Indicators of joint activity were frequency of physical activity with friends,^{50,89,90,102,105,108,123} friends’ presence during physical activity,^65,66 and whether respondents were on a sports team or in an exercise class with a friend.^90,105 Only 1 study used direct measures (distance biked on a stationary bicycle in a laboratory)⁶⁶; the others relied on self-reports of participation in physical activity with friends. The most commonly used physical activity measure was overall physical activity level, with 6 of 7 studies reporting significant relationships in at least some specifications.^{50,89,90,102,105,108}

Qualitative Studies

Qualitative studies can enhance our understanding of the mechanisms through which friends could influence physical activity or at least illustrate why children believe their friends matter for their own behavior. We separately analyzed the 25 qualitative studies that met inclusion criteria, because the data could not be compared and assessed in the same way as the quantitative studies.^127–151

These studies found that friends were important for children’s and adolescents’ physical activity levels. Nine studies found that children and adolescents perceived fun as an important motivator for engaging in physical activity and that the presence of friends during physical activity was associated with having fun.^{128,130,134,141,142,145,146,149,150} Sixteen identified friends as motivating factors for participating in physical activity.^{127,128,131–142,145,146} In addition, children and adolescents reported that they engaged in physical activity to spend time with their friends (n = 11^{127,131–136,138,139,145,146}) and that their friends’ physical activity guided their own decisions about physical activity (n = 10^{127,129,133,139,141,145–149}). That is, having active friends encouraged physical activity, and having inactive friends discouraged it.^129,147

DISCUSSION

Physical inactivity is a strong predictor of overweight, obesity, and other chronic diseases.¹⁵² Numerous studies have investigated the determinants of physical activity in children and adolescents, and social influences have been proposed as potentially important forces in determining physical activity levels. Diverse analytic approaches have been taken to define, measure, and understand social influences. We synthesized current findings on the ways friends may influence physical activity in childhood and adolescence. Results show a positive association between physical activity and indicators of influences from friends: positive communication about physical activity from friends, friends’ own physical activity levels, and the presence of friends during physical activity.

Our review found substantial evidence for a relationship between positive communication with friends about physical activity and a child’s own physical activity (43/55 studies reported significant associations). The importance of communication is consistent with findings on other health behaviors, such as sexual initiation and eating behaviors.^153,154 We also found substantial evidence that children’s physical activity is associated with the physical activity of their friends (30/35 studies), suggesting that individuals may be guided by their friends’ behavior. Friends’ behaviors have also been shown to predict dieting, body dissatisfaction, and extreme weight control.^155–159 Finally, we found strong evidence that participation with friends and the presence of friends during physical activity is associated with physical activity (9/10 studies), consistent with previous studies reporting that an individual’s eating behavior is affected by the presence of a friend while eating^160–163 and that friends match their food intake more closely than do unfamiliar peers eating together.¹⁶²

There is some indication that children anchor their physical activity on friends’ actions and messages more in some settings than in others. For example, friends may be more important for unstructured physical activity or active play but less so for organized sports involvement.⁹⁰ Also, being active with friends at home or in the neighborhood was more strongly associated with higher levels of physical activity than was being active with friends at school.⁹⁰ Friends’ support for active commuting predicted biking to school for children who lived within a kilometer of school, but not for children who lived farther away,¹⁰⁰ likely because of the importance of other factors, such as age, distance, topography, safety, and parenting styles, in selecting modes of commuting.^164,165 Another study showed that children’s best friends were important for their activity during winter but not summer, perhaps indicating that when the weather and daylight hours are less conducive to physical activity, children are only active if their friends entice them.¹¹² To effectively engage friends in health intervention, it will be necessary to map out more clearly under which circumstances they can alter physical activity behaviors.

An important consideration is whether friendship influences are important only among children who already have a high propensity for active living. Findings suggest that this is not the case. Indeed, 3 studies found that the relationship between friends’ support for physical activity or presence during physical activity and respondents’ physical activity was stronger in children at high risk for physical inactivity, such as overweight children,^47,48,65 children who thought they were not good at sports, and girls.⁴⁷ Thus, the children in most need of physical activity may benefit the most from the involvement of friends.

Also important were friends’ characteristics and the strength of the relationship. One study highlighted that not all friendships were equally influential, with only the physical activity of the best friend significantly predicting respondents’ physical activity.²⁶ In addition, a study distinguishing reciprocal and nonreciprocal friendships found stronger associations in the former, that is, in relationships where a respondent’s nominated friend also nominated the respondent as a friend.²⁶

Age may be another important factor in the relevance of friends for children’s physical activity. The focus of research on friendship influences has been on older children, with only 1 (qualitative) study examining the role of friends in physical activity among preschool-aged children. ¹⁵¹ Three of the 4 studies that focused on children aged 5 to 8 years reported results consistent with friendship influences, as did 39 of 44 studies (89%) among children aged 9 to 11 years, 45 of 53 studies (85%) among participants aged 12 to 13 years, 44 of 54 studies (83%) of adolescents aged 14 to 15 years, and 25 of 29 studies (86%) of young adults aged 18 to 19 years. Younger children tend to be more heavily influenced in all aspects of well-being by family factors,¹⁶⁶ although social influences may commence during the first few years of life.^167,168 A priority for future research is to begin to understand at what ages friends’ involvement and support become important determinants of physical activity and how the process of influence begins.

Limitations

Because of the variation in indicators and study populations, a meta-analysis of this body of literature was not possible; the large variability would have limited meta-analysis procedures to small subsets of the studies. Our review was limited to friendships rather than broader social networks. This is because we expect, in light of previous studies,^24,28 that the closer and stronger the relationship between 2 individuals, the more possibilities for influence. However, because schools provide a captive audience for interventions and a forum that can be targeted by policy changes, schools may present the most practical avenue for interventions that use social networks to improve health behavior. In addition, previous physical activity interventions delivered in school settings have been shown to be successful and to change long-term physical activity trajectories.¹⁶⁹ Therefore, for future research and applications, we can take friends as an upper bound of possible levels of social influence, but it will be important to consider how other peers, especially within the school context, can encourage healthy behaviors such as promoting physical activity. Intervention experiments are already being carried out to train school-based peer advocates to disseminate information about and promote healthy choices.^170,171

Some weaknesses of this literature remain to be addressed by future studies. Most studies relied on cross-sectional data (n = 68 studies) and on respondents’ reports about their friends (n = 72), approaches that are not suited to studying influence. Another notable weakness was the use of insufficient control variables and statistical methods to understand the nature of the associations: 20 studies only explored correlations and bivariate associations,^{50,52,56,58,61,63,64,70,72,77,79,81,82,85,86,93,115,116,119,122} 11 used only a limited set of control variables,^{26,47,49,62,65,66,95,96,107,120,123} and only 2 were intervention studies.^88,111

Cross-sectional analyses accounting for no or only limited confounding variables do not permit researchers to assert causality. Such study designs also do not make it possible to distinguish influence from homophily; that is, children may behave similarly because they select friends who are similar to them, in their interest in physical activity among other traits, or they may behave similarly to their friends because they live in the same or a similar environment (e.g., neighborhood sidewalks or playgrounds may entail similar physical activity patterns within communities). Thus, significant associations between friends’ behaviors do not necessarily indicate influence, but many studies have not clearly addressed or acknowledged this concern²⁰ and have tended to overestimate influence.^172–174 A large literature addresses the difficulty of distinguishing between influence and selection, and although there is disagreement about the best approach for disentangling them,^{19,45,174–176} only 4 of the studies we reviewed attempted to use such methods to account for homophily or to distinguish influence from homophily.^45,74,75,83 To advance our understanding of the determinants of physical activity, future studies should strive to disentangle the roles of selection, environmental factors, and social influences.

Our findings provide evidence consistent with the existence of friendship influences on children’s physical activity. Specifically, the studies that attempted to distinguish between homophily and influence found that indicators of influence were more strongly associated with physical activity over time than were the indicators of selection.^74,75 Furthermore, the 13 longitudinal studies reported significant associations between indicators of friendship influences and physical activity, consistent with friendship influence.^{45,62,74,75,80,83,84,86,87,91,97,109,124}

Another concern is that reliance on respondents’ reports about their friends’ attitudes, beliefs, and behaviors introduces the possibility of projection bias, or reflecting one’s own attitudes and beliefs onto friends, which would inflate observed associations between respondents and friends. Previous research has found that adolescents tend to misperceive their friends’ behaviors in ways that are partially mediated by their own behaviors.¹⁷⁷ However, all 11 articles that collected data directly from respondents’ friends showed significant associations in some models or population groups,^{26,45,60,65,66,74,75,80,83,88,90} 6 of them in all models.^{45,60,65,66,74,75} A related weakness is that 55 studies used subjective measures of respondents’ own physical activity. These studies more often reported evidence of a relationship between respondents’ physical activity and indicators about friends than did studies that used objective measures. In fact, researchers who used both subjective and objective measures found the relationship between indicators of friend influences and objective physical activity variables to be weaker than the relationship with self-reported physical activity measures.^96,102,103 For example, Martín-Matillas et al. found that friends’ physical activity was associated with all of the subjective measures of physical activity but with none of the objective measures.⁹⁶ These patterns may occur if children exaggerate similarity with their friends to highlight that they are socially connected or engaged with their peers. It may also be that direct measures, such as accelerometers and pedometers, pick up all physical activity and not just voluntary, conscious activity, and respondents may simply not be aware of the differences or may only anchor their voluntary physical activity on their friends.

It may be that some friends, for example, those who are more popular or older, can be more important than other friends in guiding behaviors. The studies in our review did not explore this possibility, but understanding which friends are most likely to be influential will be an important step for the development of interventions.

Conclusions

The value of the findings from the studies in our review comes in part from understanding the social processes that guide health behavior and in part from exploring the possible use of social networks to promote healthy behaviors. We found extensive research about and fairly strong evidence of associations between physical activity and friends’ behaviors. The use of friendships to promote physical activity in experiments or interventions remains largely unexplored: only 1 study attempted to manipulate the influence of friends in a controlled experiment; it found that children biked longer and more intensely in the presence of a friend than in the presence of an unknown peer.⁶⁶ Only 1 study explored the impact of friends’ social support for physical activity in an intervention; it concluded that support from friends predicted change in physical activity over time.¹²⁴ It may be that some interventions have been unsuccessful in manipulating physical activity through friendships and that those results have not been published. However, findings from observational studies suggest that friends have potential to affect physical activity and that friendships may offer opportunities to promote the early development of healthy physical activity habits and preferences that can translate into lifelong habits.⁴³ These opportunities are worth exploring, because a recent review of physical activity interventions found that most interventions until now have had only small effects, amounting to an increase in activity of approximately 4 minutes per day.¹⁷⁸

The challenge to incorporating friendship influences into health interventions is that friendships are complex. Can an intervention create friendships? Can a health professional identify who in a friendship dyad is influential and who is influenced or how mutual influence occurs? Can recommendations alter the type of information transmitted by friends in ways that can overcome habits, tastes, and social norms? If influence operates primarily through individuals’ perceptions of their friends’ behaviors, can children’s perceptions of their friends’ behaviors be altered or overcome? To address these complexities, several observational, experimental, and implementation steps, developed through multidisciplinary collaborations, are needed. These should include (1) data collection efforts including both survey instruments and direct measurements of attitudes, beliefs, and behaviors among sets of friends to understand how, when, and to what extent influence occurs; (2) longitudinal studies with data and methods that permit at least some causal inferences; and (3) experimental studies that test multiple possible mechanisms of friendship influence over the short term (day, week) and over the longer term (months, years).

Our systematic review of the literature revealed strong evidence that physical activity behaviors are positively associated with encouragement and support for physical activity from friends, with friends’ physical activity patterns, and with engagement with friends in physical activity. We found some indication that these associations may result from friends influencing each other’s behaviors, although further research is needed to understand the magnitude of influences and to explore the potential for using friendship to promote a lifetime of healthy physical activity habits.