Descriptive Normative Nutrition Messages to Maximize Effect in a Videogame: Narrative Review

Abstract

Messages containing descriptive norms have been manipulated and tested for effects on dietary intake in a number of studies. The effects have been mostly small or nonexistent. This narrative review analyzed the structural elements of the nutrition messages most likely related to effectiveness. Elements examined included the following: source, similarity of source to the participant, reference group, identification with or evaluative significance of the reference group, delivery, salience, background, tailoring, gain-loss framing, and whether an assessment of the group norm was conducted before designing the descriptive norm message. Most studies incorporated few of the reviewed structural elements, and message manipulations appeared weak. Controlled, systematic studies are needed to examine the effect of the structural elements, singly and in combination, on behavior and its mediators. Videogames are designed to be delivered in the same way to players; therefore, they offer a controlled environment for extensively and systematically manipulating structural elements of messages and examining the resulting behavioral effects. How this might be accomplished was described. Experimental research was identified as being necessary to delineate the individual and combined element effects, thereby contributing to the body of literature on videogame design and effectiveness.

Introduction

Dietary intake has been related to many and diverse health outcomes.¹ Descriptive norm messages have been employed to help change dietary intake.² Most descriptive norm messages concerning dietary intake attempted to increase the purchase or intake of fruit and vegetables (FV) due to their health benefits.³ Descriptive normative nutrition messages have also targeted sweets/chocolate, snacks, junk food, pizza, and “healthy” foods.² “Healthy food” is not an accepted specific category of food, but the name carries the implication that their consumption should be increased. Sweets/chocolates are often grouped with snacks and as a category is a major contributor to energy intake.⁴ “Junk food” is not a clearly defined category, but the title carries the implication that it should not be eaten. Pizza is considered a mixed dish⁴; is one of the three main contributors to energy intake and sodium⁴; thus, its intake is usually discouraged.

Videogames have been used to influence dietary intake.^5,6 However, they often emphasize increased knowledge and awareness, resulting in modest to no effect on intake.⁷ More theoretically informed behavioral procedures are needed to address this important health behavior outcome.⁶ Social factors, such as descriptive norms,² have been shown to influence dietary intake. Descriptive norms are a person's beliefs about what other people do (e.g., “most other people eat few or no vegetables”), which have predicted diverse behaviors^8–10 and influenced children's self-reported FV intakes.¹¹

Descriptive norms have correlated substantially with vegetable intake in both cross-sectional and longitudinal research.¹² Peer FV norms were related to concurrent and FV intake 1-year later.¹³ A conceptual model of how descriptive norms influence behavior has been suggested.¹⁴ It hypothesized that a descriptive norm message initiates a social comparison process of the self in regard to a reference group (i.e., me compared to what I believe my friends do). Systematic reviews examining the relationship of descriptive norms to dietary behavior have shown statistically significant, but generally small, effects.²

Serious videogames are likely to be more effective than other media, for example, posters, in delivering health messages. Videogames have incorporated diverse behavior change procedures¹⁵ and can employ story and animation, which can be engaging and immersive, which would increase the attention to the message.¹⁶ This article reviews the literature on descriptive norms and analyzes 10 characteristics likely influencing behavioral outcomes with an emphasis on how to design videogames to optimally deliver these messages.

Theory

Conventional descriptive norm-based interventions attempt to manipulate an individual's perceived descriptive norm to come closer to the actual or desired descriptive norm. Descriptive norms have been relatively easily manipulated but had effects in only some members of a target group under some circumstances.² Some of the descriptive norm messages may not have worked because the true reference group norm was not known or was misspecified. Figure 1 presents a graphic hypothesized representation of the theory-specified relationships among personal preferences (behaviors), perceived descriptive norm, reference group's actual descriptive norm, and injunctive norm (i.e., either the expectation(s) others may have for the participant's behavior or a government or professionally specified prescription for level of intake).

FIG. 1.

Likely relationships among preference, perceived social norm (in a reference group), objective social norm (in a reference group), and some nationally specified guidelines or expectations of a reference group (generally called injunctive norms) for a generally liked food (sweetened beverages) and a generally disliked food (vegetables).

In Figure 1, levels of descriptive norms are specified concerning a person's preferences for vegetables (i.e., a less-preferred behavior that health professionals would like to increase, e.g., vegetables) and sweetened beverage intake (i.e., a more preferred behavior that nutritionists would like to decrease, e.g., pizza). This hypothesized participant's preference for vegetables is low (left side of Fig. 1), and their perception of the reference group's (e.g., best friends and kids your age) descriptive norm for eating vegetable intake is low (line F). However, the reference group's objectively assessed descriptive norm for eating vegetables is higher than the perceived descriptive norm (line E), but is lower than national dietary guidelines (line D).¹⁷ The same kinds of relationships can be specified for a highly preferred behavior such as pizza (right half of Fig. 1).

Participants are motivated to conform to perceived reference group norms.¹⁷ Manipulating the person's perception of intake by a reference group will only have the desired effect if the person's preference (or behavior) is more extreme from guideline-specified healthy alternatives than the person's perception of the reference group's descriptive norm (line F). For example, the vegetable recommendation for teens is approximately 2–3 cup equivalents per day; if an individual rarely eats vegetables, and they perceive their reference group eats ∼1.5 cup equivalents per day, they may decide they need to eat more vegetables. A message stating the objectively assessed reference group descriptive norm (line C) is a common way to manipulate descriptive norms in an intervention. If the person's preference or behavior is less extreme than the reference group's descriptive norm, then a manipulation of the reference group's descriptive norm may only serve to decrease the vegetable intake (or increase the sweetened beverage intake, called a boomerang effect). Thus, descriptive norm message interventions should assess reference group behavior to specify the norm. If the participant's preferences or behaviors are further from the healthy level (line A) than the reference group's intake (lines B or D), the normative message should have an effect. If there is equal distribution of the participants' preference above and below the reference group, a study risks zero effects (increasing half and decreasing the other half).¹⁸

Structural Elements

Health messages are complex.¹⁹ While methods for designing and testing procedures for use in community programs have been developed,²⁰ little attention has been given to analysis of the many elements constituting the structure of these messages. Ten structural elements of messages can be identified from descriptive norm messages: source, identification with source, personal relationship to source, reference group, identification with or evaluative significance of the reference group, method of delivery (direct, surreptitious, or story), salience, environment, tailoring, gain-loss framing, and group norm assessment. (See Table 1 for a description of each element.) Some of these elements have been shown to influence outcomes.

Table 1.

Structural Elements of a Descriptive Norm Message

Structural elements	Definition of structural elements
Source	The person(s), group, or institution identified as specifying the message.
Identification with the source	The extent to which the participant believes they have some rapport or fellowship with the source often based on similarity of characteristics to the participant (e.g., age, sex, and ethnicity).
Personal relationship to source	Whether the source has a relationship to the participant (e.g., friend, sibling, and avatar for which a story develops identification or some other relationship).
Reference group	The group which the specified norm characterizes.
Identification with the reference group	The extent to which the participant believes they have some rapport or fellowship with the reference group often based on similarity of characteristics to the participant (e.g., age, sex, and ethnicity).
Evaluative significance of the reference group	The extent to which the participant believes the reference group shares their values.
Method of message delivery
Direct	Whether the message was clearly stated, often using statistics or other numerical information.
Surreptitious	Whether the participant needed to see something and interpret it as a norm.
Story	Whether the message was delivered as part of a story manipulating multiple structural characteristics of the message to specify a descriptive norm.
Salience	The extent to which the participant was aware of the message at the time of performing the behavior (usually manipulated to draw attention to the descriptive norm message at the time of decision/choice).
Background	The background for delivering the message (e.g., images of appealing FV on a poster stating an FV norm).
Tailoring	Whether the message was individualized for the participant on some relevant characteristic (usually based on a response to a questionnaire at baseline or early in the trial).
Framing
Gain framing	Whether the message was stated as the behavior being a benefit to the participant.
Loss framing	Whether the message was stated as the behavior being the avoidance of a loss.
Assessment of group norm	Whether the stated norm message reflected prior assessment of the actual norm.

FV, fruit and vegetables.

The first element, source, is the person, group, or institution from which the message was communicated. When “adolescents” were the source of a message, adolescent participants reported increased beliefs about the adverse effects of and negative attitudes toward smoking.²¹ Greater similarity of the participant to the source²² can lead to identification with source²³ and having a personal relationship with the source (e.g., friend) influenced the participant's behavior.²⁴ Reference group is the group about which the relevant norm is communicated. Identifying with the reference group (e.g., sharing similar characteristics)^23,25,26 or perceived similarity with the reference group's values (called evaluative significance), both, influenced behavior.^17,18 Comparing typical students, friends, and parents as reference groups revealed that descriptive norms among typical students and friends predicted adolescent alcohol intake.²⁷ Identification with the referent group was measured in one study and had a marginally significant interaction effect with the norm manipulation and vegetable intake.⁸

Method of delivery of the descriptive norm message can be direct, surreptitious, or by story. Direct delivery concerns whether the message was just stated, often using quantitative representations, for example, “most students.”²⁸ Surreptitious delivery concerns whether the participant needed to see something and interpret it as a statement of norm (e.g., notice the empty wrappers next to a container of wrapped candies, and infer how much others were eating). Finally, story/narrative concerns whether the message was communicated using others as examples, for example, animated characters doing the behavior as part of a story that reinforces the importance of the behavior.²⁹ We know of no study that varied the method of delivery, but we believe story holds the biggest promise since stories can be immersive and thereby can better engage participants to attend to the messages.^16,29

A message may not influence a participant unless they are aware of it.¹⁸ Salience concerns the extent to which the participant was aware of the message at the time of performing the behavior.²² One study may have manipulated the salience of the message by having the participant perform the experimental manipulation right after seeing the poster with the message. This manipulation had no effect on behavior.¹⁷ In another study, the experimenter specifically pointed out the message just before the assessment, but this was confounded with other design factors.²⁰

Background is the context in which the message is communicated (e.g., poster and school). For example, a poster promoting FV intake may have a typed message in the midst of an appealing colorful background of FV. While we know of no study that systematically varied background, we hypothesize a background consistent with the message (e.g., pleasing FV in the background of a FV message) and appealing (e.g., colorful and audience pleasing images) would have greater impact. Tailoring, that is, whether the message was individualized to some characteristic of the target person (normally based on the collection of data about that person) can effectively change behavior.^30,31 One study tailored the message by providing feedback in relationship to the participant's prior consumption. Both individualized feedback groups with incentive, and without, increased healthier purchases than a control group at the end of the intervention, but were not different from the control group after 3 months.¹⁸ A group getting individualized feedback consumed more FV during the ensuing 8 weeks.²²

Framing, that is, whether the message was expressed to the participant as a benefit (positive framing), or avoidance of loss (negative framing), has also been an effective communication strategy.^30,32 A gain-framed message would state the benefits of eating fruit, while the loss-framed message would state the costs of not eating fruit. The gain-framed message resulted in significantly greater fruit intake 1 week after receiving the message, but did not affect intention to eat fruit either right after the manipulation or 1 week later.³³ Group norm assessment concerns whether the message framer actually did research to identify the prevalence of the behavior, from which the group norm in the message might be expressed. This should avoid the problem of selecting a norm level that was not realistic, and thereby not being able to influence behavior (Fig. 1).

Limited research exists on each of the descriptive normative structural elements. While there is variability across studies on how each of the elements was employed, few studies manipulated a specific element in the study.

This article reviews how descriptive norm messages have varied on each of the 10 structural elements and proposes innovative procedures that may influence dietary intake within a videogame.

Methods

Since we were not conducting a meta-analysis of behavior outcomes, but simply qualitatively examining the structure of the messages, we needed a representative sampling of articles containing descriptive norm messages, not a definitive list of all relevant studies. We identified experimental studies that tested descriptive norms in regard to dietary intake. Our inclusionary criteria were that a descriptive norm for dietary intake was tested, either in a randomized trial or a field experiment, and the target behavior (e.g., FV) was measured. Since we were interested in the diet-related effects of normative messages, we inserted the terms “norms and diet” and “norms and food” into PubMed. In addition, the reference lists from published articles were reviewed to identify candidate articles. As a result, we used all articles identified in our search that met the inclusionary criteria.

The first author conducted a preliminary review of articles for meeting inclusionary criteria, and then abstracted the article author's name, date of publication, academic home, research design, sample size, method of dietary intake assessment, method of manipulating descriptive norm, and results primarily concerning the effects of the descriptive norm statement (Table 2). Table 2 describes how each study in sequence of appearance in the literature manipulated descriptive norms plus other information to understand the results.^26,33–51 Table 3 presents the descriptive norm messages found in these articles. Several studies used the same, or highly similar, message as prior studies, which accounts for the number of statements being smaller than of studies.

Table 2.

Manipulation of Descriptive Social Norms

Author(s) (year of publication); study location	Design/sample size	Targeted dietary behavior and measure	Description of alternative norm manipulation	Results of norm manipulation
(1) Croker et al.³⁴ (2009); University College London, United Kingdom	Randomized experimental home interview/n = 1083	Rating of intended FV consumption (number of portions/day) over next 6 months	During interview were asked the following: “Did you know that 80% of people in Britain try to eat at least 5 portions of fruit and vegetables a day?”	Significant sex + condition effect: separate analyses by sex. No effect in women. Men: 0.6 portion increase among normative intervention versus control.
	Randomized experimental home interview/n = 1083		Control: no specific information
(2) Hermans et al.³⁵ (2012); Radboud University Nijmegen, The Netherlands	(a) Study 1: RCT, two groups; n = 63 female undergraduates	Whether to eat, or not, a sweet (M&Ms with peanuts); actual weighed intake from a bowl	During video watching:	Study 1: participants in the experimental condition did not eat more sweets.
			Confederate: ate four winegums
			Control: did not eat any
	(b) Study 2: three groups; n = 51 young females	Whether to eat, or not, a sweet (milk chocolate M&Ms); actual weighed intake from a bowl	During video watching:	Study 2: participants in the experimental condition did not eat more sweets.
			Confederate:
			(a) ate 8 milk chocolate M&Ms
			(b) ate 20 milk chocolate M&Ms
			Control: did not eat any
(3) Mollen et al.³⁶ (2013); Maastricht University, The Netherlands	Randomized field experiment/n = 220, and 18–22-year-old undergraduates	Food choice on school food court categorized into “salad choice” or “hamburger choice”	Four posters with descriptive norm messages at entrances and at the salad/or/grill bar “Every day more than 150 [university] students have tossed salad/or/burger for lunch here.”	Unhealthy norm message did not predict hamburger selection. Healthy norm message predicted salad selection.
			Control: no signs posted
			Messages counterbalanced and rotated by day over 4 consecutive weeks
(4) Prinsen et al.³⁷ (2013); Utrecht University, The Netherlands	(a) Study 1: two arm counterbalanced (not randomized) in the back of a bakery/n = 144 customers	Number of chocolates taken from bowl	Bowl of 200 individually wrapped chocolates on a countertop in the bakery, a second bowl next to the first bowl with 20 empty chocolate wrappers/or/no empty wrappers	Nineteen chocolates were taken when wrappers present with 65 customers, 11 chocolates were taken when wrappers not present with 79 customers.
	(b) Study 2: two arm (not randomized)/n = 65 undergraduates	Number of chocolates taken from bowl	Bowl with 20 chocolates in a laboratory for 10 minutes before a reaction task with a second bowl next to the first bowl with 10 empty chocolate wrappers versus no empty wrappers	Around 72% of participants took one or more chocolates when empty wrappers present, 45% of participants took one or more chocolates when no empty wrappers present.
	(c) Study 3: RCT: factorial design: 2 cue (healthy vs. unhealthy snack) × 2 goal prime: healthy versus hedonic eating/n = 90 undergraduates	Three healthy and three unhealthy snack choices	Food wrappers by “previous” participants in a bowl, 10-minute-laboratory exposure to healthy eating intentions, a bin with six snacks: three healthy and three not; a second bin with three empty healthy wrappers/or/three empty unhealthy wrappers; two magazines present healthy versus hedonic eating	The interaction effect of goal prime by environmental cue was not significant, environmental cue main effect was statistically significant.
(5) Robinson et al.³⁸ (2013); University of Liverpool, University of Birmingham, United Kingdom	RCT, three norm groups: descriptive norm, healthy norm, control × usual junk food intake/n = 129 young adults	Immediate postintervention consumption of high calorie “junk” food (three items) and FV (three items) from a snack buffet	On a poster, with pictures:	Descriptive and health norm conditions consumed fewer high-calorie snacks than the control condition (no difference between social and injunctive norm groups) and no difference in FV intake. Moderation effects for hunger, but not usual junk food intake.
			Descriptive Norm: “Most students limit how much junk food they are eating …”
			Health norm: “Limiting junk food to 1 or less than 1 serving a day is part of a healthy diet.”
(6) Stok et al.²⁶ (2014); Utrecht University, Wageningen University, The Netherlands	(a) Study 1: RCT, two groups: majority norm versus minority norm/n = 57 undergraduates	Self report of “sufficient” vegetable intake in the past 7 days	On a sheet of paper during psychology class:	Strength of identification related to vegetable intake, norm manipulation marginally related (majority norm higher than minority norm). Strong identifiers ate more vegetables than moderate or weak identifiers in the majority norm group.
			Majority norm: “… a full 73% of … University students eat sufficient vegetables …”
			Minority norm: “27% of … University students eat sufficient vegetables”
	(b) Study 2: RCT, two groups: majority norm versus minority norm/n = 52 undergraduates, strong identifiers with the university	Same	Same	Majority norm participants reported higher intention. Effect on intake not reported. The effect on intention was mediated by self-identification, attitude, and self-efficacy.
(7) Stok et al.³⁹ (2014); Utrecht University, The Netherlands; University of New South Wales, Australia	RCT, three groups: descriptive versus injunctive versus control/n = 80 high school students (14–17 years old)	Self reported fruit intake over following 2 days (not assessed at baseline); response scale not reported; intention to do so	On a sheet:	Descriptive norm increased consumption by 0.6 fruit pieces above control and 0.8 fruit pieces above injunctive norm. Injunctive norm appeared to decrease intention in comparison to both descriptive norm and control. No other significant difference. Moderation analyses not reported.
			Descriptive: “a majority of high school students try to eat sufficient fruit …”
			Injunctive: “a majority of high school students think other high school students should eat sufficient fruit.”
			Control: no additional statement
(8) Robinson et al.⁴⁰ (2014); University of Liverpool, University of Birmingham, United Kingdom	(a) Study 1: RCT, two groups: descriptive norm versus HM/n = 71 undergraduates	Weighed selections of vegetables from a buffet lunch: four vegetarian items and six nonvegetarian items	Poster and flyer: Descriptive: “… the typical student eats over three servings of vegetables each day …” Health: “… (health) … can be improved by eating over three servings per day.”	No effect for condition on vegetable intake. Significant condition by usual intake interaction. Low usual consumers consumed more vegetables in norm condition. No difference for high consumers.
	(b) Study 2: RCT, three groups: descriptive norm versus injunctive norm versus HM/n = 70 undergraduates	Percent of snacks from FV from a snack buffet: two FV items and two “high-calorie junk food items”	Poster and flyer: Descriptive: “… typical student eats their five servings of fruits and vegetables each day.” Injunctive: “… the typical student thinks their peers should eat five servings of fruits and vegetables each day.” Health: “Eating five servings of fruit and vegetables a day can improve your health.”	Significant condition by usual FV intake interaction. Low usual consumers selected more FV in the descriptive norm condition versus both injunctive norm and HM group and less high-calorie snacks and total calories in the descriptive norm group versus the HM, but not the injunctive norm. No statistical significant difference among high FV consumers.
(9) Payne et al.⁴¹ (2015); New Mexico State University, Cornell University, Paso del Norte Foundation, United States	(a) Study 1: two grocery stores in Texas: one selected for intervention and one for control (not randomized), both for 14 days/n = 396,017 person transactions	Purchases of produce (FV) during the shopping visit. In-store FV purchases per day—percent increase in average spending/person	Intervention: sign on the grocery cart (English and Spanish) saying “In this store most people choose at least × produce items” ( × reflecting the average for that store)	Pilot: significant group by time interaction with intervention store having 12% greater increase in produce spending than control without a significant increase in total spending per person.
			Control: no sign on cart
	(b) Study 2: two grocery stores in New Mexico: alternated intervention and control days for 4 weeks/n = 575,689 person transactions	Same	Same	Increases in produce spending for treatment and control.
				No treatment by time interaction.
				No increase in total spending.
(10) Sharps and Robinson⁴⁷ (2015); University of Liverpool, United Kingdom	RCT: four groups: high intake norm, low intake norm, no norm, control/n = 143, 6–11-year-old children	Weighed carrots taken from bowl; 24-hour dietary recall, liking for FV (two items), perception of how many FV other children eat every day	Seven-minute-exposure to fictitious participant information sheet about the number of carrots eaten by six prior participants (including date of birth and sex): carrots (all), carrots (one), carrots (empty)	Significant one-way ANOVA; children in the high intake norm condition ate more carrots than the other groups.
			Control: no carrot column; researcher pointed out carrot sheet; bowl full of carrots versus bowl with only one carrot
(11) Sharps and Robinson⁴² (2015); University of Liverpool, United Kingdom	(a) Study 1: RCT factorial design 2 (peer acceptance vs. acceptance ambiguity) × 2 (norm: high vs. low)/n = 100, 6–11-year-old children	Bowl of carrots left for 7 minutes in front of child	Fictitious information sheet about 6 prior participants: all carrots eaten, no carrots eaten, and bowl of an apparently previous participant with 1 remaining carrot or full	No significant interaction term. High norm group ate more carrots than low norm group. No significant effect for peer acceptance. Moderating variables did not influence the V intake.
		Bowl of carrots left for 7 minutes in front of child	Acceptance versus ambiguity: child told especially liked versus child told would be told about liking after a short break
	(b) Study 2: 2 sessions 1 day apart; RCT: 2 norm (high vs. low) × 2 familiarity of environment (Eating norm shown at first session vs. second session) × 2 eating sessions (1, 2)/n = 127, 6–11-year-old children	Same	Same norm manipulation; familiarity: unfamiliar: shown norm information in session 1; familiar: show norm information in session 2	Three-way interaction effect for norm: familiarity, effects, and sessions.
(12) de Bruijn et al.³³ (2015); University of Amsterdam, The Netherlands	RCT, four groups (factorial design): 2 message framing groups (gain vs. loss) × 2 majority descriptive norm (low vs. high intake)/n = 177 adults (39.8% dropout rate)	Self-report of usual fruit intake/week in past week in frequency (per week) and intake (per day) after message exposure, self-reported intentions before and after 1 week after message exposure	Method of delivery not clear: Positive frame: “When you eat at least two cups of fruit per day, the vitamins, minerals, and fibers in fruit will ensure that (a) your immune system is boosted, …”	No significant main or interaction effect for message framing or descriptive norm for any variable.
			Negative frame: “When you do not eat at least two cups of fruit per day … (a) your immune system is weakened …”
			Unhealthy norm: “… around 75% of Dutch adults do not eat at least two cups of fruit per day.”
			Healthy norm: “… around 75% … eat at least two cups of fruit per day.”
(13) Sharps and Robinson⁴³ (2016); University of Liverpool, United Kingdom	(a) Study 1: RCT: three groups: received descriptive SN versus HM versus no message/n = 143, 6–11-year-old children	Weighed selection of snack foods at an unobtrusive after exposure to message (1 F, 1 V, two high-calorie snacks)	Received cards as part of a board game, and landed on three spaces and obtained a card for each	Children in the HM conditions ate more FV at snack than control, but no difference between two experimental groups and no difference in high-calorie food intake. No moderation effects. No effect on children's belief about amount of FV intake of other children.
			Descriptive norm: board with images of FV and a message card: “other children eat lots of FV every day and like them,” “other children eat FV every day as snacks,” and “other children eat FV at break time”
			Health: “FV are really good for you,” “FV have lots of vitamins,” “FV make you strong and healthy”
			Control: “All polar bears are left handed,” “Snails can sleep for up to three years,” “Penguins can jump really high in the air”
	(b) Study 2: RCT: four groups: same as Study 1/n = 164, 6–11-year-old children	Same	Same board game:	No group difference. No moderation effect. Children in the Descriptive SN condition believed other children ate more FV than the HM, but not the control conditions.
			Experimental: “Children like you eat lots of fruit and vegetables every day and like them; children like you eat fruit and vegetables every day as snacks; children like you eat fruit and vegetables at break time.”
			Exposure: “Strawberries have seeds on the outside; carrots help you to see in the dark; grapes are actually a berry.”
			Health and Control messages: same
			Similar, plus a scale where the researcher indicated other children ate five or more FV/day
(14) Thomas et al.⁴⁴ (2016); University of Liverpool, University of Birmingham, University of Oxford, United Kingdom; University of Toronto, Canada	RCT: 2 × 5 × 3 design delay (immediate vs. 24-hour later intake), message (liking norm, descriptive norm, HM, vegetable variety, and neutral control), and habitual consumption (low vs. high)/n = 353 young adults (mostly undergraduates)	Weighed buffet selection from three vegetables, three energy-dense foods, and two dips	Posters with messages:	Main effect for vegetables and for habitual vegetable consumption. No effect of message for high or low consumers. No difference between immediate and delayed conditions.
			Neutral control: images of University of Birmingham
			Veg Variety: “… there are over 1000 different types of vegetables”
			Health: “… (Health) … can be improved by eating over three servings of vegetables per day.”
			Descriptive norm: “… the typical student eats over three servings of vegetables each day.” Liking norm: “… 80% of students actually like vegetables a lot.”
			All completed a poster rating after seeing poster
(15) Thorndike et al.⁴⁵ (2016); Harvard Medical and Business Schools, Department of Medicine—Massachusetts General Hospital, United States	RCT, three groups: monthly letter + SN feedback; monthly letter + SN feedback + incentive for increasing green purchases; control/n = 2672 employees	Proportion of selections of green (labeled “healthy”) from the cafeteria at Mass General	Feedback only: four letters over 3 months with proportion of employee selection in the prior month labeled green (healthy), yellow or red; average proportion of green, yellow, red for all employees, and average proportion of purchases labeled green, yellow, or red and the top quintile of green purchases	Slight increase in healthy selections (2.1%) in the feedback + incentive group in comparison to control, but not to feedback only. No significant differences at end of 3-month washout period.
			Feedback incentive: same plus incentive ($10) for achieving proportion of green for all cafeteria purchases of 40%, 60%, or 80% based on baseline intake
			Control: no further contact
(16) Feeney et al.⁴⁶ (2017); University of Western Ontario, Canada; University of Toronto, Canada	RCT: factorial design: 3 persistence of (low norm, high norm, control) × 4 resistance to (selection of two, three, or four sessions of eating alone) SN/n = 108 female undergraduates	Number of 15 g pieces of pizza eaten after a 3-hour fast in a laboratory	Persistence: a fictitious participant information sheet about 10 prior participants: high norm: 10 ate 18 bite-size pieces of pizza; low norm: 3 bite-size pieces	Interaction terms not significant. Significant main effect for eating norm: high norm ate more than low norm, neither significant difference from control which was in between. Significant effect for persistence: participants ate more in session 3 than in session 1. Significant main effect for resistance norms had most effect when participant had fewer alone sessions.
			Resistance: number of sessions of eating alone (1–4) before exposure to norm message
(17) Wengreen et al.⁴⁸ (2017); Utah State University, United States	RCT: three groups: Individualized Norm, Manipulated Norm, Control/n = 185 undergraduates	Skin carotenoid measure at baseline versus 8 weeks later	All groups received “Eating more vegetables and fruit that are red, …, will help increase your skin carotenoid score”; (1) individual group received individualized printed message that gave their carotenoid value; (2) manipulated individualized message “Your score is at the 20th percentile of students at this university, 80% have higher values”; (3) control: “You will receive your score in 7 weeks”	Significant main effect: manipulated norm carotenoid level was significantly higher than control; no other differences.
(18) Thomas et al.⁴⁹ (2017); University of Birmingham, Aston University, University of Liverpool, Oxford Brooks University, University of Oxford, United Kingdom; University of Toronto, Canada	Three phases: quasi-experimental; preintervention (2 weeks), intervention (2 weeks), and postintervention (2 weeks); three workplace restaurants/n = 9445 meal purchases	Percentage of meals purchased with vegetables from cash register receipts (not including potatoes or vegetable garnishes)	Five posters +10 tabletops in restaurant: “Most people here choose to eat vegetables with their lunch.”	Significant increase from 60% to 64% of meals from preintervention to intervention phase. Significant increase from 64% to 67% from intervention to postintervention phase.
(19) Collins et al.⁵⁰ (2019); University of Birmingham, University of Bath, Aston University, University of Liverpool, University of Oxford, United Kingdom; University of Toronto, Canada	(a) Study 1: randomized two undergraduate canteens: descriptive norm versus healthy; three phases (1-week long each): baseline, intervention, immediate postintervention /n = 7568	Observed percentages of meals with a vegetable 3 days of observation in each of three phases: baseline, mid-intervention, post intervention	Messages on posters in the school lunchroom: SN: “Did you know that most students here choose to eat vegetables with their meal?”	SN group increased vegetable purchases (63%–68%) which was sustained post intervention (67%): no change from HM (75%–74%), decline to 69% at post. Most students (66.1%) in SN condition reported not seeing the poster versus 39.5% in HM condition.
			HM: “Did you know that students who choose to eat vegetables have a lower risk of heart disease?”
	(b) Study 2: same design as Study 1/n = 4052 meal purchases	Number with side orders of vegetables (excluding potatoes), 5 days of assessment at each of three phases	Messages on posters:	Both the SN and HM groups increased (SN: 22.9%–32.5%; HM: 43.8%–52.8%), but SN group effect was not sustained at postintervention (22.1%), but HM group (48.1%) was. Around 17.5% reported not seeing poster.
			SN: “Most University of Birmingham students eat more than 3 servings of vegetables each day according to a 2014 survey.”
			HM: “Eating over 3 servings of vegetables each day can reduce cancer risk, according to a 2014 survey.”
(20) Coates et al.⁵¹ (2019); University of Liverpool, United Kingdom	RCT: three groups: unhealthy snacks, healthy snacks, nonfood products/n = 176, 9–11-year-old children	Child selection from two unhealthy and two healthy snacks (weighed)	Two popular social media bloggers (Instagram) were shown with two mock profiles: six images	Significant main effect: unhealthy exposure group ate fewer healthy snacks than healthy group and more calories.
			Unhealthy: three of six images were unhealthy snacks
			Healthy: three of six images were healthy (e.g., banana)
			Control: six branded nonfood items

2 × 5 × 3, A systematic factorial design where all combinations of the three conditions are tested; ANOVA, analysis of variance; arm, number of experimental conditions; HM, health message;; M&Ms, a form of chocolate candy; n, sample size; RCT, randomized clinical trial; SN, social norm.

Table 3.

Nonduplicative Listing of the Social Norm Messages or Procedure Manipulations

Verbal/text messages
(1)	… 80% of people in Britain try to eat at least 5 portions of fruit and vegetables a day?³⁴
(2)	… 78% of respondents to a national survey were aware they should be eating 5+ portions of FV/day, and 90% reported healthy eating was important/or/no specific information³⁴
(3)	Every day more than 150 [university] students have tossed salad/or/burger for lunch here³⁶
(4)	Most students limit how much junk food they are eating … /or/no message³⁸
(5)	… a full 73%/or/27% of [university] students eat sufficient vegetables …²⁶
(6)	… a majority of high school student try to eat sufficient fruit … /or/no message³⁹
(7)	… the typical student eats over three servings of vegetables each day/or/healthy message: … (health) … can be improved by eating over three servings per day⁴⁰
(8)	Most people at [store name] choose at least x produce items (x reflecting average at that store)/or/no message⁴¹
(9)	… around 75%/or/65% of Dutch adults … eat/or/do not at least two cups of fruit per day³³
(10)	(1) … other children eat lots of FV every day and like them. (2) … other children eat FV at break time. (3) … other children eat FV as snacks./or/(1) All polar bears are left handed. (2) Snails can sleep for up to three years. (3) Penguins can jump really high in the air.⁴³
(11)	(1) … children like you eat lots of FV every day and like them. (2) … other children eat FV at break time. (3) … other children eat FV as snacks./or/(1) All polar bears are left handed. (2) Snails can sleep for up to three years. (3) Penguins can jump really high in the air./or/(1) Strawberries have seeds on the outside. (2) Carrots help you see in the dark. (3) Grapes are actually a berry.⁴³
(12)	Four letters over three months with employee selection in the prior month labeled green (healthy), yellow, or red/or/no such communication⁴⁵
(13)	… your (carotenoid) score is × (reflecting own score)/or/… your (carotenoid) score is at the 20th percentile of students at this university/or/you will receive your score in 7 weeks⁴⁸
Nonverbal/procedures
(1)	Bowl with individually wrapped chocolates accompanied by a bowl with 20 empty wrappers or an empty bowl.³⁷
(2)	A confederate in a video ate 4 winegums/or/did not eat any.³⁵
(3)	A confederate in a video ate 20 chocolates, 8 chocolates, or none.³⁵
(4)	Information sheet supposedly surreptitiously indicating the behavior of six prior participants (none/or/all carrots), + children were presented with a bowl of a previous participant, which contained one carrot vs. full to corroborate information sheet.⁴⁷
(5)	Video images of popular social media bloggers shown with six images: 3/6 images were unhealthy snacks, 3/6 images were healthy snacks, all six images were branded with nonfood items.⁵¹

Related messages for different conditions are separated by /or/.

5+, 5 or more; FV/day: fruit and vegetables per day.

Table 4 characterizes each study's descriptive norm message in regard to the 10 elements, in the same sequence of studies as in Table 2. The second author reviewed the abstracted information for accuracy by comparing the tables to the articles. Differences, for example, about how to categorize aspects of each study according to each element, were discussed and resolved. A meta-regression analysis relating descriptive norm message characteristics to dietary intake was not conducted since (1) the focus of our review was analysis of the descriptive norm elements, (2) there were inadequate numbers of articles in some subcategories of the characteristics, and (3) the presence of some characteristics were likely interrelated leading to likely confounding, which makes a meta-regression analysis difficult to interpret.⁵²

Table 4.

Characteristics of the Descriptive Norm Messages on 10 Structural Elements in Sequence of the Studies in Table 1

Source	Source: similarity to or relationship with participant	Reference group	Reference group: identification with or evaluative significance of	Direct (fact)/surreptitious/or story-narrative	Salience at time of decision (manipulated or measured	Tailored (individualized)	Background of message	Gain/loss framing	Assessment before experiment to identify group norm
(1) Interviewer stated as a question³⁴	None	People in Britain	Not clear	Direct	Not clear	No	Conducted in home interview	No	Baseline assessment of personal servings/day
(2a) Experimenter³⁵	Both experimenter and participants were female undergraduates	25-Year-old average-weight woman confederate	Not clear	Surreptitious observation of confederate in video eating some/none winegums, while doing work-related tasks in a university office	Not clear	No	Laboratory	No	No
(2b) Experimenter³⁵	Both experimenter and participants were female undergraduates	25-Year-old average-weight woman confederate	Not clear	Surreptitious observation of confederate in video eating some/none M&Ms, while doing personal tasks in home living room	Not clear	No	Laboratory	No	No
(3) Anonymous³⁶ (poster)	None	[Maastricht University] students	Reference group and participants, both ate at food court	Direct	Not clear	No	University logo and photos of the salad/or/burger bar in school lunch rooms	No	Number of students eating salads and burgers sold per hour on 2 days
(4a) Anonymous³⁷	Other customers—not clear	Other customers at the store	Reference group and participants, both customers at a bakery	Surreptitious observation of empty candy wrappers	Free candy was likely salient for some	No	In bakery/lunch room	No	No
(4b) Anonymous³⁷	Other participants in experiment—not clear	Other participants in a lab experiment	Reference group and participants, both customers at a bakery	Surreptitious observation of empty candy wrappers	Free candy was likely salient for some	No	Laboratory testing room for an experiment	No	No
(4c) Experimenter³⁷	None	Other research participants	Reference group and participants, both participants in research	Surreptitious observation of empty candy wrappers	Free food may be salient for some	No	Healthy or hedonic magazines available for viewing	No	No
(5) Experimenter³⁸ (poster)	None	Students	Reference group and participants are both students	Direct	Free food may be salient for some	No	Images of junk food on poster with message	No	Yes, n = 40
(6a) Experimenter²⁶ (short information sheet)	None	[U] students	Measured: moderate identification; significant interaction for identification	Direct	No	Not clear	Part of a questionnaire	No	No
(6b) Experimenter²⁶	None	[U] students	Selected for identification with [U] students	Direct results of an experiment completed	No	Not clear	Part of a questionnaire	No	No
(7) Experimenter³⁹	None	High School students like yourself	Not clear	Direct	No	No	Booklet containing information and questions	No	No
(8a) Experimenter⁴⁰ (poster)	None	“Most students …”	Not clear	Direct	No	No	Images of vegetables	No	No
(8b) Experimenter⁴⁰ (poster)	None	“… typical students …”	Not clear	Direct	No	No	Images of FV	No	No
(9a) Grocery chain⁴¹	Reference group and participants were shoppers at that store	Spending on produce items purchased at this store	Not clear	Direct	No	No	Top 10 FV purchased in store on grocery cart	No	No
(9b) Grocery chain⁴¹	Reference group and participants were shoppers at that store	Spending on produce items purchased at this store	Not clear	Direct	No	No	List of top 10 FV purchased in store	No	No
(10) Experimenter⁴⁷	None	Supposed six previous participants in this research	Not clear	Surreptitious: saw previous participant info sheet	No	No	Bowl on table in laboratory	No	No
(11a) Experimenter⁴²	None	Fictitious prior six participants	Reference group and participants are both participants in this research	Surreptitious	Not clear	No	Fictitious participation sheet	No	No
(11b) Experimenter⁴²	None	Fictitious prior six participants	Reference group and participants are both participants in this research	Surreptitious	Experimenter pointed out data column on form	No	Fictitious participation sheet	No	No
(12) Experimenter³³	None	“Dutch people aged 18 to 65”	Not clear	Unknown	No	No	Not specified	Yes—benefits of eating versus costs of not	No
(13a) Board game⁴³	None	“Other children …”	Not clear	Direct	Not clear	No	Images of FV on board game	No	Assessed perceptions of intake in “other children”
(13b) Board game⁴³	None	“Children like you …”	Not clear	Direct	No	No	Images of FV on board game	No	No
(14) Experimenter⁴⁴	None	“Most students” [U] named at bottom	Not clear	Direct	Participant was asked to read poster	No	Images of vegetables on poster	No	Yes (study not reported)
(15) Employer⁴⁵	None	All employees eating at this cafeteria	Reference group and participants, both eat at cafeteria	Direct	Not clear	Yes, proportion of cafeteria purchases from prior months	Letters with colored columns for intake of “green yellow, and red foods”	No	No
(16) Experimenter⁴⁶	None	Fictitious 10 prior participants	Reference group and participants are both participants in this research	Surreptitious	Not clear	No	Sheet of paper taped to table	No	No
(17) Class teacher⁴⁸	None	“College students at [U]”	Reference group and participants were both students at that college	Direct	Measured personal skin carotenoid value	Measured personal skin carotenoid value	Visual graphic of the distribution of scores	No	Yes, distribution of scores at [U]
(18) Workplace restaurant⁴⁹	None	“Most people here …”	Reference group and participants are both customers at restaurant	Direct	Not clear, large number of posters	No	Attractive colorful posters across workplace restaurant	No	Yes, based on assessment of intake over 2 weeks at restaurant
(19a) U cafeteria⁵⁰	None	“… most students here …”	Not clear	Direct	Not clear	No	Posters in student canteen	No	No
(19b) U cafeteria⁵⁰	None	“… most students here …”	Not clear	Direct	Not clear	No	Posters in student canteen	No	No
(20) Experimenter⁵¹	None	Two of the top 10 vloggers in the United Kingdom	Apparently, popular vloggers	Surreptitious holding foods	Popular vlogger holding pictures	No	Instagram banner in school	No	No

[U], a specific university, not specified.

Results

Twenty articles reporting 28 studies were included in this review. Descriptions of each of the studies are included to provide context for the message analysis. Table 2 lists the studies in sequence by year of appearance in the literature. The first article appeared in 2009³⁴ and the last in 2019.⁵¹ Ten of the articles were primarily based in the United Kingdom,^{34,38,40,42–44,47,49–51} six in the Netherlands,^{26,33,35–37,39} three in the United States,^41,45,48 and one in Canada,⁴⁶ while several of these articles involved collaborations across multiple countries.^44,49,50 Sixteen of the individual studies were one-factor randomized clinical trials (two to four groups)^{26,35–40,43,45,47,48,51}; seven were field studies of various designs^37,41,49,50; five were randomized factorial designs^33,42,44,46; and one was a randomized, two-group interview.³⁴

Study sample sizes ranged from 51²⁶ to 2672¹⁸ participants,³⁵ 7568 to 9445 meal purchasers,^23,24 or 396,017 to 575,689 person transactions⁴¹ per study. Eleven studies targeted vegetables alone^{26,36,40,42,44,47,49,50}; one study targeted skin carotenoid values (a proxy for vegetable intake)⁴⁸; five studies targeted FV^34,38,40,41; three studies targeted snacks^43,51; five studies targeted sweets^35,37; two studies targeted fruit alone^33,39; and one apiece targeted small pizzas⁴⁶ and green (GO foods, those to eat more of) foods.⁴⁵

The method of measuring intake varied substantially, including 12 studies with a weighed, or number of, selection of item(s) from a bowl^{35,37,40,42–44,47,51}; 4 with observation of meal selections^45,46,50; 2 each for selections from a buffet,^38,40 food choice on a school lunch line,³⁶ purchase of produce during a shopping visit,⁴¹ self-reported intake of “sufficient vegetable intake,”²⁶ and self-reported intake of fruit^33,39; and one each of number of the items selected,³⁷ cash register receipts,⁴⁹ skin carotenoid value,⁴⁸ and rating of intention to consume.³⁴ The nature of the control groups varied substantially.

Whether the descriptive norm intervention influenced the dietary intake measure varied across studies. Twenty studies reported at least one significant effect for the descriptive norm manipulation.^{26,34,36–43,45,47–51} The patterns of successful change varied substantially across studies. When descriptive norm manipulations worked, the effect tended to be small.

The descriptive norm text messages (Table 3) varied on targeted food, the specified reference group, the level of norm for high and low, and whether the message specifically mentioned the target behavior or negatively mentioned the opposite behavior (hoping that the target behavior would respond positively). The descriptive norm message statements varied across studies with substantial variety in the method of delivery and message, which indicates a more quantitative assessment of effectiveness was not desirable, or even possible. Table 3 is a nonduplicative list of descriptive norm manipulations.

Source: identification and relationship

In regard to the structural elements of descriptive norm messages (Table 4), the source of the message (i.e., from whom it came) in most of these studies was an experimenter,^{26,33–35,38–40,42,44,46,47} with one each for grocery chain,⁴¹ board game,⁴³ employer,⁴⁵ class teacher,⁴⁸ workplace restaurant,⁴⁹ and university cafeteria,⁵⁰ or some anonymous source.^36,37 In no case was there an attempt to manipulate or emphasize any similarity or relationship of the source to the participant.

Reference group: identification and evaluative significance

The reference groups were more varied, including students at their university or high school,^{26,36,38,39,43,44,48} people from their country,^33,34 other research participants,^37,42,46,47 other customers at the store or cafeteria,^{37,41,45,49,50} most or typical students,⁴⁰ other children or children like you,⁴³ and 2 of the top 10 vloggers.⁵¹ The identification with or evaluative significance of these reference groups appears minimal.

Method of message delivery: direct, surreptitious, and story

The descriptive norm was directly delivered to participants,^{26,33,34,36,38–40,44,45,48–50} including providing a sheet of paper with a message on it, posters with the message sometimes with flyers,^4,7,9,14,20 a sign on a grocery shopping cart,⁴¹ player information cards as part of a board game,⁴³ messages displayed at a salad and grill bar,³⁶ four letters sent to customers of a cafeteria,⁴⁵ and video of attractive social media vloggers with images of six foods.⁵¹ Surreptitiously delivered messages in three studies used bowls of empty wrappers (indicating consumption by previous confederate customers or participants),³⁷ two studies each with a video of a confederate engaged in the behavior,³⁵ a sheet of information on fictitious prior participants,^42,46,47 and one unknown method of delivery.³³ In no study was a story used to convey the norm; no study systematically varied the method of message delivery; and no study used a videogame to deliver the message.

Salience

Salience was not specified or manipulated in 10 studies^{26,33,39–41,43,47}; was not clear if it was addressed in 8 studies^{34–36,42,43,45,46,49,50}; and was manipulated in 8 other studies. Salience was clearly manipulated when the experimenter pointed out the data column on the form,⁴² or asked the participant to read the poster.⁴⁴ We inferred salience when free candy or food was offered (free candy³⁷ and food being attention getting³⁸), a popular vlogger held the messages,⁵¹ and personal carotenoid values were presented to the participant (assuming personal biometric data generated attention).⁴⁸ In one case, a large number of posters were present and should have increased attention,⁴⁵ but this was not manipulated, nor was a measure of salience employed.

Background

The backgrounds mostly reflected the environments in which the message was delivered. Deliberately employed specific backgrounds included use of university logo and images of the school lunchrooms on posters,³⁶ images of junk food on the poster with the message, images of V or FV on a poster with the message,^40,43,44 a listing of the top 10 FV purchases on the grocery store cart,⁴¹ letters with colored columns for the types of foods,⁴⁵ and attractive colorful posters.⁴⁹

Tailoring

While most studies did not tailor, two studies tailored the message to the participant's purchase at a cafeteria⁴⁵; one tailored to baseline skin carotenoid value⁴⁸; and one study tailored by providing feedback in relation to the participant's prior consumption.¹⁸ Both individualized feedback groups with incentive, and without, increased healthier purchases than a control group at the end of the intervention, but were not different from the control group after 3 months.¹⁸ Feedback was provided about the participant's skin carotenoid level in comparison to a low norm, versus no individualized feedback. The group getting individualized feedback consumed more FV during the ensuing 8 weeks.²²

Framing: gain or loss

Gain or loss framing of the message was manipulated in one study.³³ The gain-framed message stated the benefits of eating fruit, while the loss-framed message stated the costs of not eating fruit. The gain-framed message resulted in significantly greater fruit intake 1 week after receiving the message but did not affect intention to eat fruit either right after the manipulation or 1 week later.³³

Assessment of group norm

Only six studies conducted an assessment of the usual intake in the reference group to specify the descriptive norm message.^{17,34,36,38,48,49} One study assessed perceived intake in “other children.”⁴³ None used videogames to deliver descriptive norm messages.

Discussion

Consistent with a meta-analysis of the descriptive norm effect literature,² these descriptive norm messages influenced dietary behaviors under some circumstances and/or with some populations, but the significant effects were weak. Part of the limitation on effects may be whether the message was structured and combined to provide maximal effect. We explore how descriptive norms might be optimally structured in videogames to do the same. Videogames may be more advantageous than other media because they permit structuring the environment consistent with, supportive of, and making the intended messages attractive; and can deliver messages with high fidelity to the underlying construct with a consistent presentation across all participants. This ensures strong tests of manipulation.

None of the descriptive norm message studies specifically stated the source of the message, although it could be inferred as the experimenter or grocery store, a fairly anonymous attribution. Within the context of a videogame, the message source could be an avatar or an on-screen character similar to the participants' reference group. Avatars tend to enhance motivation, but need to be carefully considered in terms of one versus several avatars, and whether the avatar should be designed to be self-representational (i.e., created from a digital image of the participant)⁵³ or have similar characteristics (e.g., clothes, age, appearance, color of skin, and hair style),⁵⁴ including participant and avatar body size.⁵⁵ Alternatively, the player could be asked to create an avatar that appeals to them or they feel represents their “ideal self,” thereby possibly enhancing the influence of the source in the videogame.⁵⁶ A story could be designed to generate the perception of a relationship to the avatar. The characteristics of the source that need to be similar to the participant's to enhance the effectiveness of the message need to be addressed in experimental research.¹⁶

The reference group manipulation in the descriptive norm studies appeared to be weak. Social modeling of food intake demonstrated a confederate has more influence on the participant's behavior when represented as a student at the same university versus another university.²⁵ This was interpreted as a manipulation of social identity, which likely applies to reference groups and involves similarity to and/or identification with group.¹⁴ Although the reference group being students from the same university attended by the participant influenced the dietary selection effectiveness of a descriptive norm manipulation,^26,36 videogames offer stronger possibilities. The reference group could be made to appear like the participant to varying degrees. Identifying with, and evaluative significance of, the reference group could be manipulated by story elements that showed how the characters, much like the participant, encountered outcome behavior-related events or situations.

Directly stating or otherwise presenting the descriptive norm to the participant would appear to be more effective than surreptitious presentation, since surreptitious presentation requires that the participant notices and interprets often subtle information. Surreptitious manipulation of norms can employ cues in the characters' environment (similar to the bowls of empty wrappers or the characters making selections of foods in environments frequented by the player, e.g., cafeterias and convenience stores) without verbalizing this. Augmented reality could be designed to provide direct descriptive norm messages as the player approached a food store (identified by GPS or Google maps). None of the descriptive norm studies used a story/narrative to present the descriptive norm, but story/narrative appears to be a promising approach, which could also enhance identification and evaluative significance.⁵⁷

The salience of the message, when manipulated in the descriptive norm studies, involved the experimenter reading the message or encouraging the participant to read the message. A likeable videogame character could be created with which the child identifies. Having this character believably state the message with conviction may enhance message salience. Other ways might include incorporating and highlighting the descriptive norm message into an immersive storyline.

Two studies tailored the message to the participant.^45,48 Providing the participant's carotenoid value within the context of the distribution of the reference group's values⁴⁸ was creative. However, it may not have effectively manipulated descriptive norms, if the participant's value was at the healthier end of the distribution. Tailoring has been shown to be effective in a number of studies^58,59; and can be easily done in a videogame by collecting relevant data early in a game and providing messages predicated on the initial assessment. However, collecting data near the beginning of game play on a relevant characteristic could interfere with immersion or message processing, both of which have been shown to be associated with behavior change in a videogame.⁶⁰

The environment of the descriptive norm messages in these studies was most commonly manipulated by providing images of the targeted food on a poster or other medium delivering the message. This manipulation similarly appears to be weak. The environment of the message in a videogame could be manipulated by presenting the message in multiple environments, perhaps reflecting the environments in which the player may perform the behavior (e.g., home and school). Virtual crowds could chant the norm; posters, graffiti, or crowds in the virtual world could be programmed to enhance the norm message.

Participant selection of foods in a virtual reality game with a buffet line, after exposure to a descriptive norm message, might be used as the outcome measure as an inexpensive way of studying descriptive norms.⁶¹ However, this method needs to be further validated. Spatial storytelling⁶² might be used to create a three-dimensional (3D) environment in which participants can explore, perhaps interact with other players, which personal meanings are assigned to aspects of the environment over time into which FV (or other food items) are inserted. Correspondence between the fantasy and real worlds could be created by assigning tasks in the real world, including descriptive norm messages reflecting players in the 3D world.

One study manipulated gain-loss framing of messages.³³ The gain frame had a greater effect on behavior. None of the other messages was gain framed. Gain framing would appear to be an easily implemented manipulation of a descriptive norm message.

To induce change, the descriptive norm message needs to be higher than the participant's behavior for desired increased intakes or lower than the participant's behavior for desired decreased intakes (Fig. 1). In part, this might be done by having the game obtain participant baseline information as part of the game player profile, and systematically conveying one or more norm messages in the virtual world that is higher than their specified desired behavior or lower than their specified undesired behavior.

Videogames offer other possibilities to enhance the effect of descriptive norm messages. The message might be a statement of the norm with an added intention. Multiple characters might make similar statements, but there is likely a limit to the number that enhances effectiveness. The characters could state they are engaging in the behavior or indicate who else is doing it. It could be stated in an obvious way, or offhand, surreptitiously, to be less authoritarian. The character could simply state the descriptive norm, or perhaps it would be better to demonstrate the norm by having animated characters engage in the behavior without corresponding verbalizations. Videogames can simulate restaurants or grocery stores in virtual reality to enhance the experience of immersion,⁶³ and thereby increase acceptance of the message.

While cross-sectional and longitudinal research designs can reveal whether social norm variables influence intake in natural settings, experimental studies are needed to isolate effective intervention procedures and components thereof.⁶⁴ A systematic review and meta-analysis of 16 experimental studies concluded that social norms influence what and how much participants selected to eat.² In most of these studies, the norm was presented as the behavior of prior participants in the research (no other personalization).

An important first step in testing these ideas would be to maximize the descriptive norm message effect by manipulating as many of the 10 structural characteristics as possible. If the desired change is obtained, ensuing research could experimentally manipulate one structural element at a time.

The above discussion concerns the structure of descriptive norm messages. The same kinds of issues could be raised for other psychosocial determinants of behavior (e.g., persuasive messages for attitude change, perceived behavioral control, and self-efficacy).

Conclusion

Videogames provide innovative procedures for systematically manipulating many elements in the descriptive norm message structure. Experimental research is needed to determine the optimal structure of messages in regard to influencing dietary intake, both by manipulating individual elements and then by combination of the more effective elements. Research of this type will provide insight that can lead to the design of effective videogames that achieve behavior change goals.

Footnotes

Acknowledgment

This work is a publication of the United States Department of Agriculture (USDA/ARS) Children's Nutrition Research Center, Department of Pediatrics, Baylor College of Medicine, Houston, Texas.

Author Disclosure Statement

No competing financial interests exist.

Funding Information

This work had been funded, in part, with federal funds from the USDA/ARS under Cooperative Agreement number 58-3092-5-001.

References

Katz

, Meller

. Can we say what diet is best for health?. Annu Rev Public Health, 2014; 35:83–103.

Robinson

, Thomas

, Aveyard

, et al. What everyone else is eating: A systematic review and meta-analysis of the effect of informational eating norms on eating behavior. J Acad Nutr Diet, 2014; 114:414–429.

Boeing

, Bechthold

, Bub

, et al. Critical review: Vegetables and fruit in the prevention of chronic diseases. Eur J Nutr, 2012; 51:637–663.

Leme

, Baranowski

, Thompson

, et al. Top food sources of percentage of energy, nutrients to limit and total gram amount consumed among US adolescents: National Health and Nutrition Examination Survey 2011–2014. Public Health Nutr, 2019; 22:661–671.

Hieftje

, Edelman

, Camenga

, et al. Electronic media-based health interventions promoting behavior change in youth: A systematic review. JAMA Pediatr, 2013; 167:574–580.

Baranowski

, Ryan

, Hoyos-Cespedes

, et al. Nutrition education and dietary behavior change games: A scoping review. Games Health J, 2019; 8:153–176.

Rekhy

, McConchie

. Promoting consumption of fruit and vegetables for better health. Have campaigns delivered on the goals?. Appetite, 2014; 79:113–123.

Salvy

, Bowker

. Peers and obesity during childhood and adolescence: A review of the empirical research on peers, eating, and physical activity. J Obes Weight Loss Ther, 2014; 4:207. doi:10.4172/2165-7904.1000207.

McEachan

, Taylor

, Harrison

, et al. Meta-analysis of the reasoned action approach (RAA) to understanding health behaviors. Ann Behav Med, 2016; 50:592–612.

10.

Esposito

, van Bavel

, Baranowski

, et al. Applying the Model of Goal-Directed Behavior, including descriptive norms, to physical activity intentions: A contribution to improving the Theory of Planned Behavior. Psychol Rep, 2016; 119:5–26.

11.

Cullen

, Baranowski

, Rittenberry

, et al. Child-reported family and peer influences on fruit, juice and vegetable consumption: Reliability and validity of measures. Health Educ Res, 2001; 16:187–200.

12.

Berg

, Ritschel

, Swan

, et al. The role of hope in engaging in healthy behaviors among college students. Am J Health Behav, 2011; 35:402–415.

13.

Lehto

, Ray

, Haukkala

, et al. Do descriptive norms related to parents and friends predict fruit and vegetable intake similarly among 11-year-old girls and boys?. Br J Nutr, 2016; 115:168–175.

14.

Higgs

, Thomas

. Social influences on eating. Curr Opin Behav Sci, 2016; 9:1–6.

15.

Baranowski

, Blumberg

, Buday

, et al. Games for health for children—Current status and needed research. Games Health J, 2016; 5:1–12.

16.

Baranowski

, Lyons

, Thompson

. Experimental design to systematically develop a knowledge base for effective games for health. Games Health J, 2019; 8:307–312.

17.

Miller

, Prentice

. Changing norms to change behavior. Annu Rev Psychol, 2016; 67:339–361.

18.

Kallgren

, Reno

, Cialdini

. A focus theory of normative conduct: When norms do and do not affect behavior. Pers Soc Psychol Bull, 2000; 26:1002–1012.

19.

Harrington

NG.

Introduction to the special issue: Message design in health communication research. Health Commun, 2015; 30:103–105.

20.

Lewis

, Watson

, White

. The Step approach to Message Design and Testing (SatMDT): A conceptual framework to guide the development and evaluation of persuasive health messages. Accid Anal Prev, 2016; 97:309–314.

21.

Romer

, Jamieson

, et al. Counteracting the influence of peer smoking on YouTube. J Health Commun, 2017; 22:337–345.

22.

Spink

, Shields

, Chad

, et al. Correlates of structured and unstructured activity among sufficiently active youth and adolescents: A new approach to understanding physical activity. Pediatr Exerc Sci, 2006; 18:203–215.

23.

Buller

, Bettinghaus

, Fluharty

, et al. Improving health communication with photographic images that increase identification in three minority populations. Health Educ Res, 2019; 34:145–158.

24.

Campo

, Brossard

, Frazer

, et al. Are social norms campaigns really magic bullets? Assessing the effects of students' misperceptions on drinking behavior. Health Commun, 2003; 15:481–497.

25.

Cruwys

, Platow

, Angullia

, et al. Modeling of food intake is moderated by salient psychological group membership. Appetite, 2012; 58:754–757.

26.

Stok

, Verkooijen

, de Ridder

, et al. How norms work: Self-identification, attitude, and self-efficacy mediate the relation between descriptive social norms and vegetable intake. Appl Psychol Health Well Being, 2014; 6:230–250.

27.

Lac

, Donaldson

. Testing competing models of injunctive and descriptive norms for proximal and distal reference groups on alcohol attitudes and behavior. Addict Behav, 2018; 78:153–159.

28.

Frank

, Murphy

, Chatterjee

, et al. Telling stories, saving lives: Creating narrative health messages. Health Commun, 2015; 30:154–163.

29.

Baranowski

, Baranowski

, Thompson

, et al. Video game play, child diet, and physical activity behavior change: A randomized clinical trial. Am J Prev Med, 2011; 40:33–38.

30.

Pope

, Pelletier

, Guertin

. Starting off on the best foot: A review of message framing and message tailoring, and recommendations for the comprehensive messaging strategy for sustained behavior change. Health Commun, 2018; 33:1068–1077.

31.

Conway

, Webster

, Smith

, et al. eHealth and the use of individually tailored information: A systematic review. Health Informatics J, 2017; 23:218–233.

32.

Gallagher

, Updegraff

. Health message framing effects on attitudes, intentions, and behavior: A meta-analytic review. Ann Behav Med, 2012; 43:101–116.

33.

de Bruijn

, Visscher

, Mollen

. Effects of previous fruit intake, descriptive majority norms, and message framing on fruit intake intentions and behaviors in Dutch adults across a 1-week period. J Nutr Educ Behav, 2015; 47:234–241.

34.

Croker

, Whitaker

, Cooke

, et al. Do social norms affect intended food choice?. Prev Med, 2009; 49:190–193.

35.

Hermans

, Salvy

, Larsen

, et al. Examining the effects of remote-video confederates on young women's food intake. Eat Behav, 2012; 13:246–251.

36.

Mollen

, Rimal

, Ruiter

, et al. Healthy and unhealthy social norms and food selection. Findings from a field-experiment. Appetite, 2013; 65:83–89.

37.

Prinsen

, de Ridder

, de Vet

. Eating by example. Effects of environmental cues on dietary decisions. Appetite, 2013; 70:1–5.

38.

Robinson

, Harris

, Thomas

, et al. Reducing high calorie snack food in young adults: A role for social norms and health based messages. Int J Behav Nutr Phys Act, 2013; 10:73.

39.

Stok

, de Ridder

, de Vet

, et al. Don't tell me what I should do, but what others do: The influence of descriptive and injunctive peer norms on fruit consumption in adolescents. Br J Health Psychol, 2014; 19:52–64.

40.

Robinson

, Fleming

, Higgs

. Prompting healthier eating: Testing the use of health and social norm based messages. Health Psychol, 2014; 33:1057–1064.

41.

Payne

, Niculescu

, Just

, et al. Shopper marketing nutrition interventions: Social norms on grocery carts increase produce spending without increasing shopper budgets. Prev Med Rep, 2015; 2:287–291.

42.

Sharps

, Robinson

. Perceived eating norms and children's eating behaviour: An informational social influence account. Appetite, 2017; 113:41–50.

43.

Sharps

, Robinson

. Encouraging children to eat more fruit and vegetables: Health vs. descriptive social norm-based messages. Appetite, 2016; 100:18–25.

44.

Thomas

, Liu

, Robinson

, et al. The effects of liking norms and descriptive norms on vegetable consumption: A randomized experiment. Front Psychol, 2016; 7:442.

45.

Thorndike

, Riis

, Levy

. Social norms and financial incentives to promote employees' healthy food choices: A randomized controlled trial. Prev Med, 2016; 86:12–18.

46.

Feeney

, Pliner

, Polivy

, et al. The persistence of and resistance to social norms regarding the appropriate amount to eat: A preliminary investigation. Appetite, 2017; 109:93–99.

47.

Sharps

, Robinson

. Perceived eating norms and vegetable consumption in children. Int J Behav Nutr Phys Act, 2015; 12:135.

48.

Wengreen

, Nix

, Madden

. The effect of social norms messaging regarding skin carotenoid concentrations among college students. Appetite, 2017; 116:39–44.

49.

Thomas

, Ursell

, Robinson

, et al. Using a descriptive social norm to increase vegetable selection in workplace restaurant settings. Health Psychol, 2017; 36:1026–1033.

50.

Collins

EIM

, Thomas

, Robinson

, et al. Two observational studies examining the effect of a social norm and a health message on the purchase of vegetables in student canteen settings. Appetite, 2019; 132:122–130.

51.

Coates

, Hardman

, Halford

JCG

, et al. Social media influencer marketing and children's food intake: A randomized trial. Pediatrics, 2019; 143:e20182554.

52.

Moller

, Merchant

, Conroy

, et al. Applying and advancing behavior change theories and techniques in the context of a digital health revolution: Proposals for more effectively realizing untapped potential. J Behav Med, 2017; 40:85–98.

53.

Thompson

, Cantu

, Callender

, et al. Photorealistic avatar and teen physical activity: Feasibility and preliminary efficacy. Games Health J, 2018; 7:143–150.

54.

Baylor

AL.

Promoting motivation with virtual agents and avatars: Role of visual presence and appearance. Philos Trans R Soc Lond B Biol Sci, 2009; 364:3559–3565.

55.

Peña

, Khan

, Alexopoulos

. I am what I see: How avatar and opponent agent body size affects physical activity among men playing exergames. J Comput Mediat Commun, 2016; 21:195–209.

56.

Kuo

, Lee

, Chiou

. The power of the virtual ideal self in weight control: Weight-reduced avatars can enhance the tendency to delay gratification and regulate dietary practices. Cyberpsychol Behav Soc Netw, 2016; 19:80–85.

57.

Baranowski

, Lu

, Buday

, et al. Stories in games for health: More pros or cons?. Games Health J, 2013; 2:256–263.

58.

Ryan

, Dockray

, Linehan

. A systematic review of tailored eHealth interventions for weight loss. Digit Health, 2019; 5:1–23.

59.

Broekhuizen

, Kroeze

, van Poppel

, et al. A systematic review of randomized controlled trials on the effectiveness of computer-tailored physical activity and dietary behavior promotion programs: An update. Ann Behav Med, 2012; 44:259–286.

60.

, Baranowski

, Thompson

, et al. Story immersion of videogames for youth health promotion: A review of literature. Games Health J, 2012; 1:199–204.

61.

Persky

, Goldring

, Turner

, et al. Validity of assessing child feeding with virtual reality. Appetite, 2018; 123:201–207.

62.

Hameed

, Perkis

Spatial storytelling: Finding interdisciplinary immersion. In: Rouse

, Koenitz

, Haahr

, eds. Interactive Storytelling ICIDS 2018 Lecture Notes in Computer Science. Switzerland: Springer Nature; 2018, pp. 323–332.

63.

Marcum

, Goldring

, McBride

, et al. Modeling dynamic food choice processes to understand dietary intervention effects. Ann Behav Med, 2018; 52:252–261.

64.

Robinson

Perceived social norms and eating behaviour: An evaluation of studies and future directions. Physiol Behav, 2015; 152:397–401.