Table 1.

Characteristics of included studies

Study (country)	Study design	Sample	Setting	Description/Aim	Outcome measurement tools	Results
Physical activity
Barkley et al. 2014 (USA) [44]	Cross-over controlled trial	20 children (50 % girls) 3–6 years	Laboratory	Children’s PA was assessed during 30 minute sessions under two social conditions: while playing alone and with a friend. Aim: To assess the effect of the presence of a friend on amount and intensity of PA	PA: Accelerometer Sedentary activity: Direct observation by research personnel	Children had 54 % greater (P < .02) average accelerometer counts during the friend condition (μ = 2629, SD = 1080 or 5.7 METs) than during the alone condition (mean = 1707, SD = 1009 or 4.5 METs).
Brown et al., 2009 (USA) [45]	Cross-sectional	476 children (51 % boys) 3–5 years; 55 % African Americans	Preschool	Children’ PA and context was observed and recorded during indoor and outdoor activities at preschool. Aim: To determine which contextual conditions were predictors of PA of children during outdoor play.	PA and social environment: Observational System for Recording Activity in children (OSRAC-P)	Compared to outdoor activities with an adult present, MVPA was 3.55 times more likely if children were alone, 2.29 times more likely when one-to-one with another peer, and 2.04 times more likely when in a group of peers. Non-sedentary PA was 2.77 times more likely when children were alone, 1.53 times more likely when one-to-one with a peer and 1.48 times more likely when with two or more peers without adults, compared to activities with an adult present.
Eaton & Keats, 1982 (Canada) [46]	Cross-over clustered-RCT	69 children (27 girls) Mean age = 51.1 months (4.3 years);	Mobile research trailer	Children were randomly assigned to same-sex triads and visited the play setting twice, once alone and the other with peers. Aim: To test whether girls’ and boys’ PA is influenced by same-sex peers and if they are influenced differently	PA: Accelerometer	Girls and boys were more active in triads than alone (girls: μ = 3.43, SD = .19 vs μ = 2.80, SD = .24; boys: μ3.56, SD = .24 vs μ = 3.15, SD = .23). Peer presence did not influence boys and girls differently.
Gubbels et al., 2011 (Netherlands) [47]	Cross-sectional	175 children 2–3 years; mean age = 2.6 years	Childcare centre	Children’s PA was observed at childcare centres and aspects of the environment were assessed. Aim : To examine the association between the social and physical childcare environment and PA of children	PA and social environment: OSRAC-P	Indoor prompts by peers were not associated with PA (P = .966). Negative prompts by peers had a positive effect in boys (P < .05) but not in girls (P > .05). Positive peer prompts had a stronger effect in boys than girls (both P’s < .01). Larger group size of peers was associated with lower PA both indoors (P < .001) and outdoors (P = .015). 3 year-olds’ PA was negatively associated by one (P < .05) or more (P < .001) peers present. 2 year-olds’ PA was not associated with group size (P > .05).
Lehto et al., 2012 (Finland) [48]	Cross-sectional	892 children (51 % boys) Mean age = 4.7 years (SD = 1.313)	Childcare centre	Children’s PA level and nearest peer contact was observed during childcare hours. Aim: To investigate the association between peer relations and PA	PA: Direct observation by research personnel Personality and skills: Evaluated by teachers using a 5 point scale	Children who were physically active sought each other’s company (P < .001). When 3–5 year old children interacted with a group of children, their percentage of high PA was the highest (18.8 % of the time), while interaction with one child was second highest (12.4 %). When children were more withdrawn from other children, their PA tended to be lower (P = .006). Children tended to be less physically active when in the presence of a more independent and self-directed peer (P < .001).
Schwarz, 1972 (USA) [49]	RCT	57 children (22 boys) 4 years	Laboratory	Children’s mobility was videotaped for 5 minutes in an unfamiliar room under one of three conditions: with a close friend, with an unfamiliar peer or alone. Aim: To examine the influence of a peer in an unfamiliar situation on distress of preschoolers.	Children’s mobility: Direct observation with videotape Distress: rated by teachers from videotape recordings	Mobility was greater in the friend condition than in the stranger or alone conditions (P < .05). No significant differences were shown for motility between stranger and alone conditions.
Eating behaviors
Birch, 1980 (USA) [50]	Pre-, post study	39 children (20 girls); 87 % Caucasian; middle-class 2.11–4.10 years; Median age = 3.10	Nursery	Target children who preferred vegetable A to B were seated with 3 or 4 peers with opposite preferences. Children were served their preferred and non-preferred vegetable and asked to choose one. Aim: To investigate peers’ short and long-term influence on preschoolers’ food choices and eating behaviors.	Food preference rating: Assessment of taste preference of nine vegetables using a “Faces” Likert scale depicting a food as good, bad or ok. Amount of food consumed: Observer recorded the number of tablespoons served and plate waste was recorded	Target children who chose their preferred food on day 1 chose their non-preferred food on day 4 (P < .05). Target children made significantly more choices of their non-preferred food than their peers (P < 0.001). Younger children were more influenced by their peer than the older children (P < .05). Post-influence assessment found that 12 of the 17 target children increased their preference for the non-preferred food (median increase of 2.5 positions), while less than half of the peers did so.
Duncker, 1938 (England) [51]	Pre, post-design	Study 1: 2.8–5.2 years; urban, low-income Study 2: 31 children Mean age = 4.5 years	Nursery	Children were selected as either a predecessor or a successor and had opposite preferences for six food items. Pairs of the food items were presented and both children chose the food they preferred. The predecessor made his choice in front of the successor. A story was told to children about how the heroes liked Maple sugar than Hemlock bark. Modified sugars were used to represent the foods in the story and children were asked to pick which one they preferred. Aim: To examine the influence of peers and age of peers on food choices of children	Food choices: Direct observation of the food item chosen	Children made more identical choices (81 %) in the experimental condition than in the control condition (25.6 %). Younger children made more identical choices when the predecessor was older (26/28) than when the predecessor was younger (14/24). When the predecessor has a high degree of prestige or friendship with the successor, the successor made identical choices for all food items (100 %).
Greenhalgh et al., 2009 (Wales) [52]	RCT	49 children 3–4 years	Nursery	Children were randomized into either Group A, B or C and received a novel food on four snack occasions. Group A received positive modeling of the novel food on the first and third occasions, and were alone on the second and fourth occasions. Group B received negative modeling on the first occasion, positive modeling on the third, and were alone on the second and fourth occasions. Group C was alone at all four occasions. Aim: To determine the influence of peer modeling on young children’s consumption or rejection of a novel food.	Amount of food consumed: Visual estimation of plate waste using a five-point scale	Children ate more of the target food when exposed to positive peer modeling than when exposed to negative modeling (Phase 1: P < .001; Phase 2: P < .001), and ate less of the target food when exposed to negative modeling than when peers were absent (Phase 1: P = .001; Phase 2: P = .009) . The mean difference between the negative modeling group and the control group (μ = 43.75 %) was greater than the mean difference between the positive modeling group and the control group (μ =16.25 %).
Hendy & Raudenbush, 2000 (USA) [53]	Controlled trial	14 children (6 boys) Mean age = 51.4 months (SD = 11.0)	Childcare centre	Children’s number of bites of new food was videotaped across five meals. Three new foods were presented with enthusiastic teacher modeling, enthusiastic peer modeling, or simple exposure (no modeling). Delayed food acceptance was gathered one month later. Aim: To compare the effectiveness of teacher modeling and peer modeling on acceptance of new food and whether peer modeling modified the effects of teacher modeling.	Amount of food eaten: Direct observation of number of bites eaten, recorded by researchers Food acceptance: preference ratings were obtained with a “Faces” Likert scale depicting food as good, bad or ok	Boys accepted new foods equally under all three modeling conditions (P < .43), while girls accepted new foods most when modeled by peers P < .03). With trained peer models, girls’ number of bites increased across the meals. Immediate acceptance and delayed acceptance of peer modeled foods was greater for girls (P < .04) than boys (P < .002). Enthusiastic teacher modeling was ineffective if competing peer models were present.
Hendy, 2002 (USA) [54]	Controlled trial	38 children (50 % boys) 3–6 years; mean age = 54.7 months (SD = 7.9); 86.8 % Caucasian; rural, low-income	Preschool	Peer models were trained by preschool teachers. Three novel foods were presented to children during five lunch meals (3 baseline meals, 2 modeling meals). Each food was assigned to either no model, girl model or boy model conditions. Delayed food preference was assessed one month later. Aim: To examine the effectiveness of trained peer models to encourage food acceptance in children during preschool lunch, and one month later.	Amount of food consumed: Direct observation of number of bites taken, recorded by research assistants Food preference rating: Assessment of taste preference of the three novel foods using a “Faces” Likert scale depicting a food as good, bad or ok.	Same-gender models were no more effective than opposite-gender models in increasing food acceptance (P = .768). Girl models were more effective than boy models to increase food acceptance of children of either gender from baseline to modeled meals (P = 0.014). For target children, no significant differences were found for delayed food preference ratings (P = .731) or number of bites (P = .557) from the modeling condition to the one month assessment.
Lumeng & Hillman, 2007 (USA) [55]	Pre-, post study	54 children (68 % boys) 2.5–6.5 years; mean age = 4.2 years (SD = 1.1); 74 % Caucasian	Preschool	Children ate a standardized snack in a group of three and nine children. Consumption was videotaped. Aim: To determine the effect of group size on children’s food consumption	Amount of food eaten: Number of crackers eaten recorded on videotape	Children ate slightly more when eating in larger groups, than when eating in smaller groups (P = .03). During short snacks, there was no effect of group size on amount eaten (P = .42). During long snacks, large group size increased the amount eaten by 30 %.
Marinho, 1942 (Brazil) [56]	Controlled trial	66 children 4–6 years	Kindergarten	Children were divided into groups according to their food preference (predominant and indefinite taste) and subdivided into experimental and control groups. A peer was chosen as the leader and chose the food that the target child disliked. The target child was then asked to choose one of the foods. After eliminating peer influence, children’s isolated choices were assessed over 5 weeks and 2 weeks one year later. Aim: To determine if a leader causes lasting modifications of a child’s original taste preference.	Food choice and type of leadership: Direct observation by researcher	50 % of children with predominant taste modified their original taste. After-effects were observed for 48.9 % in the first four choices after the experiment and 16.7 % showed after-effects one year later. 100 % of children with originally indefinite taste modified their choice during the experiment. All but four showed after-effects one year later. Children modified their choice when the leader was socially agreeable but not when the leader was domineering.

RCT randomized-controlled trial