The State of Norm-Based Antismoking Research: Conceptual Frameworks, Research Designs, and Implications for Interventions

ABSTRACT

This synthesis review examined 189 qualified studies on norms and smoking in terms of conceptual frameworks, types of social norms, research designs, dependent variables, independent variables and covariates, and findings related to norms. Results show that 7.9% were experimental, and the remaining were cross-sectional. By far, the reasoned action approach (RAA) was the most-cited theory, but RAA was not used to guide experimental designs. The social norms approach, norm focus theory, social cognitive theory guided the intervention experiments. Harmful norms were more frequently examined than healthful norms. Pro-smoking norms positively predicted smoking intentions and behaviors, whereas antismoking norms positively predicted antismoking intentions and behaviors. The over-application of RAA in cross-sectional antismoking research has yielded repetitive findings. Norm-based experiments can adopt other theoretical perspectives to offer insights into antismoking interventions. The RAA constructs are still applicable and can be integrated into intervention designs.

Social norms refer to implied or informal rules that guide individuals’ behavior in groups and societies (Geertz, 1973). Volumes of research have examined the role of norms in harmful behavior and health interventions (Chung & Rimal, 2016), particularly in the areas of smoking and drinking (e.g., Mollen et al., 2010). Social norms have served as important predictors of smoking and smoking cessation (East et al., 2019; Mao et al., 2015). Interventions based on normative influences have been found to be effective for inducing behavioral changes, such as reduced drinking and drug use (Perkins, 2004). Although norm-related constructs (e.g., injunctive/descriptive norms and peer pressure) are frequently applied in antismoking research, the selection of and methods for operationalizing these constructs have seldom been reported in the literature (Dolcini et al., 2013). Systematic reviews have yet to address the extent to which and how norm-related theories have been used to design antismoking research. To inform researchers and practitioners of the state of norm-based antismoking research, the current review systematically selected qualified studies, and synthesized and evaluated theoretical/conceptual frameworks of norms, types of social norms, research designs, dependent variables, and independent variables and covariates. Implications for antismoking interventions are discussed.

Theories of norms

An initial search of norms examined in antismoking research yielded six oft-cited theories. These theories are overviewed, and norm constructs and intervention implications, when applicable, are explained.

Reasoned action approach

Reasoned action approach (RAA) refers to the same conceptual framework in three versions: theory of reasoned action (TRA) (Fishbein & Ajzen, 1975), theory of planned behavior (TPB) (Ajzen, 1985), and the integrative reasoned action model (Fishbein & Ajzen, 2010). The main thesis states that an individual’s behavior is predicted by behavioral intention, which is predicted by (a) behavioral beliefs and attitudes, (b) normative beliefs and perceived norms, and (c) control beliefs and perceived behavioral control (e.g., self-efficacy).

Norm constructs. Subjective norm (SN) was the exclusive norm construct of the TRA and the TPB until the recent integrative model that incorporated descriptive norms. SN, equated with injunctive norm (IN) in the integrative model, is defined as an individual’s assessment of “their” important people’s attitudes regarding whether the individual should perform a specific behavior. Descriptive norm (DN) refers to an individual’s perceptions of the extent to which “their” important others exhibit this behavior. The most salient reference group here is an individual’s important others. Both SN/IN and DN, as well as other constructs in the model, influence an individual’s intention to perform a behavior and the actual exhibition of the behavior. Fishbein and Ajzen (2010) proposed that norms not to perform the behavior should also be considered.

Intervention implications. To change an individual’s behavior, Fishbein and Ajzen (2010) detailed ten intervention steps, with the first six steps as formative research to identify relevant factors, including norms, and the remaining four steps aimed at removing barriers to actions. Norm intervention involves changing the antecedents to norms whereby influencing individuals to form normative perceptions conducive for behavioral change. Thus far, no publications or websites have documented behavioral interventions that have used some or all of the ten steps.

Social cognitive theory/social learning theory

Social learning theory (Bandura, 1977), which was later renamed social cognitive theory (SCT) (Bandura, 1986), posits that individuals can acquire new behaviors by observing and imitating others in various social contexts. In addition, individuals vicariously learn and repeat or reinforce behaviors that are rewarded, and avoid behaviors that receive punishment. A full social learning/cognitive process involves observing (the behavior and/or its consequences), retaining, and replicating the behavior observed in others. Individuals are more likely to replicate socially rewarded behaviors than socially disliked behaviors.

Norm constructs. Although SCT contains no explicitly defined norm constructs, cognitive learning processes are directly relevant to descriptive norms, which are defined as the perceived prevalence of a behavior exhibited within a social network or group. In a social network, normative behaviors are well accepted and positively evaluated, and individuals are motivated to model after these behaviors. Studies grounded in SCT mention descriptive norms as a key construct. In SCT, influential social groups are not limited to important others; instead, an individual can be voluntarily and involuntarily exposed to behaviors of multiple groups. Bandura (1999) classified an individual’s social environments into (a) the imposed – environments that individuals cannot choose (e.g., schools or workplaces); (b) the selected – environments with which individuals voluntarily associate themselves (e.g., smaller groups of friends and peers); and (c) the constructed – environments that individuals perceive as a group practicing specific behaviors. Moreover, Bandura (1999) stated that individuals from the imposed and/or selected environments could form a constructed group.

Intervention implications. SCT provides conceptual guidance for enacting instruction/learning programs that engage individuals in a cognitive learning process to help them acquire healthful behaviors and reduce risky behaviors. These learning programs typically expose target audiences to reference group members’ normative, healthy behaviors that receive rewards or otherwise risky behaviors that are negatively evaluated.

Social norms approach

The social norms approach (SNA) was developed to provide health interventions based on correcting misconceived social norms (Berkowitz, 2003; Perkins & Berkowitz, 1986). SNA maintains that individuals, especially young people, tend to overestimate the prevalence of problem behaviors (e.g., drinking and smoking) exhibited by reference group members (i.e., descriptive norms) and the members’ favorable attitudes toward these behaviors (i.e., injunctive norms). These overestimations encourage problem behaviors, whereas underestimations or reductions in overestimates can discourage problem behaviors.

Norm constructs. Perceptions of norms tend to exert greater influences than the actual norms. Misconceptions, the differences between perceived and actual norms, include pluralistic ignorance – underestimates of the majority’s healthful behavior as minority behavior; false consensus – overestimates of minority problem behavior as majority behavior; and false uniqueness – false individual perceptions that “their” behavior is more unique when such behavior is not. SNA posits that providing normative feedback to correct these misconceived norms can effectively promote healthful behaviors and reduce risky actions. SNA acknowledges that reference groups often differ in size and membership (e.g., peers in the school and people of the same age in the country).

Intervention implications. A generic SNA-based intervention process comprises three stages: baseline, intervention, and assessment. In the baseline stage, relevant actual and perceived norms are identified to estimate the extent of misperceptions. In the intervention stage, individualized and/or collective messages regarding actual norms are disseminated to the target members. In the assessment stage, researchers evaluate the extent to which misperceptions of norms have been corrected, harmful behavior has been reduced, and healthful behavior has increased.

Norm focus theory

This theory, alternatively titled “focal theory of norms” or “theory of normative conduct,” was developed by Cialdini et al. (1991). Norm focus theory (NFT) posits that when an individual focuses on or is primed/made to focus on one norm, that norm becomes more salient/accessible than others, thereby guiding individual behavior. This central thesis also explains that individuals may act based on the focal norm without being affected by other incompatible norms. Norm accessibility, the extent to which a norm can be cognitively accessed, plays a key role in predicting behavior.

Norm constructs. Three norms – two social norms and one personal norm – are said to influence individual behaviors. Descriptive norms refer to perceptions regarding the actions of majority individuals, whereas injunctive norms refer to the perception of what most individuals in a social group or network approve or disapprove of. Personal norms denote self-based standards or expectations for behavior due to the internalized values of an individual. Most discussions in NFT are centered on DNs and INs, whereas personal norms are used to contrast the two social norms. When all three norms are consistent, focusing individuals toward any one of the norms can lead to the primed normative behavior. When these norms conflict with one another, focusing on the norm applicable to the current situation can lead individuals to exhibit the respective behavior. Existing strong norms are accessible salient norms, whereas weak norms are less accessible.

Intervention implications. Researchers and practitioners need to identify norms conducive to promoting healthful behavior (or reducing risky behavior) and the external environments or situations in which the targeted behavior is likely to occur. They can thus prime audiences with healthful norms for a likely or actual environment (e.g., expressed disapproval of smoking in a restaurant for IN, or scenes of individuals not smoking in a restaurant for DN). Alternatively, audiences can be made to focus on or access internal cognitions and/or personal norms consistent with the intervention goals. Thus, focal norms induce desirable behaviors. Strong healthful or socially beneficial norms are easily primed, positive but weak norms, nevertheless, can be activated through carefully designed priming techniques to induce the target behaviors.

Theory of normative social behavior

Contributing to the NFT (Cialdini et al., 1991), Rimal and Real (2005) proposed that moderators can explicate the conditions under which norms predict behaviors, and named their work the theory of normative social behavior (TNSB). Injunctive norms, perceived social distance, outcome expectations, group involvement, and ego involvement are included as moderators that influence the relationship between descriptive norms and behavior. When INs are strong (e.g., when individuals perceive strong social sanctions if they do not comply with the norm), they are more likely to conform to the DNs they have observed. TNSB did not propose new norm constructs. The implied intervention strategies refer to changing moderators to maximize behavioral conformity to positive norms, beyond which TNSB does not provides further discussions on interventions.

Self-characterization theory/social identity theory

Self-categorization theory (ST) (Turner et al., 1987) is an extended version of social identity theory (Tajfel & Turner, 1979). Primarily, ST posits that social groups are a product of cognitive classification. All social groups have their own collection of norms that help to form group identities. In the process of identification, individuals cognitively identify with the group/category before formally becoming members. Gradually, they learn the set of group norms, internalize those norms as their own, exhibit normative behaviors, and consequently strengthen group identities. Therefore, ST accounts for similarities in the behaviors of group members.

Norm constructs. ST has not proposed any norm constructs that substantially differ from those reviewed thus far. Observed behavioral norms should predict an individual’s behavior. The strength of an individual’s identification with a group moderates the influence of the group norms because group norms tend to influence behavior and behavior-related cognitions more strongly among those who highly identify with the social group than among those who experience weaker group identification.

Intervention implications. Cognitive interventions based on individual group incompatibility are often used. Therefore, cueing or presenting an incompatibility between audience members’ own behavior and a positive group norm can motivate individuals to conform to the group norm. Using the same method, individuals can be primed to the incompatibility between their risky behaviors and the group norm not exhibiting such behavior, which leads to reduced risky behavior.

To summarize, the social norms mentioned in the aforementioned six theories primarily consist of injunctive norms (which include subjective norms) and descriptive norms. Personal norms, which are determined on the basis of an individual’s own values or reasons, are not social. In addition, several other theories/models have been mentioned, albeit infrequently. These theories include the theory of triadic influence, theory of collective action, risk information seeking and processing model, and expectancy violation theory.

Research purposes

This study aims to inform researchers and practitioners of the current state of research in terms of types of social norms examined, conceptual frameworks, research designs, independent and dependent variables, and general patterns between norms and the dependent variables.

Methods

Selection of articles

Using the keywords, “smoking and norm,” a search of PsycInfo, Communication and Mass Media Complete, Medline, Cochrane, Eric, ABI/INFORM Complete, Academic Research Complete, and ProQuest Nursing & Allied Health Journals for the period of January 1, 2000, to May 5, 2017, generated 1227 unique references. A total of 575 abstracts were selected when they reported empirical articles published in English language and included explicit mentions of smoking and norms. The researcher read these abstracts and selected 301 abstracts after exercising further inclusion and exclusion criteria. To be included in the current review, a study must report at least a measured or manipulated norm, and norms were predictors (i.e., not dependent variables) of smoking-related attitudes, behavioral intentions, and behaviors although in some studies, norms were also predicted by other variables. A study was excluded if it (a) used a sample of pregnant women, people with mental illness, terminal illness, or other severe health problems, or children aged below 10 years, as norms among such respondents are likely to deviate from the types of norms covered in the theories reviewed, and/or (b) included only norms unrelated to smoking (e.g., feminine norms, reproductive norms, norms related to climate change policy, food consumption, using mental health services, sexual practices, and eating). The final review sample consisted of 189 studies, and all full reports were obtained.

Information classification

First, a research assistant (RA) was hired and trained to extract information by cutting and pasting data, statistics, and relevant text from each article for these general categories (required by the research purposes): social norms examined, conceptual bases, sample characteristics, research design, dependent variables, and independent variables. Then, the researcher randomly selected 20% of the articles, read the cut-and-pasted information from these articles, referred to the original articles when necessary, and created subcategories for each general category. These subcategories were further improved after the researcher did a trial of coding the cut-and-pasted information from the first 30 articles. These categories are reported in the tables of results. The RA was subsequently trained to code/classify the subcategories. Due to errors discovered in the RA’s coding work, the researcher herself read and coded the cut-and-pasted information for all articles. To ensure classification rigor, the researcher commissioned a postdoctoral researcher to code 20% of the randomly selected articles (i.e., cut-and-pasted information); agreements between the two ranged from 92.5% to 100%. The results are reported in the following section.

Results

Characteristics of the reviewed study samples

The total review sample included 393,093 respondents (Mean = 2,080, SD = 4,091, Median = 818, Min = 32, and Max = 38,099 per study). The respondents consisted of adolescents and adults of all ages, males, females, and people from more than 20 countries (see Table 1).

Table 1. Sample characteristics of studies that measured social norms (K = 189).

Sample Characteristics			k (%)
	Age	Adolescents (10–19 years old)	97 (51.3%)
		Adults	88 (46.6%)
		All ages	4 (2.3%)
	Gender	Mixed gender	182 (96.3%)
		Females only	5 (2.6%)
		Males only	2 (1.1%)
	Smoking status	Mixed	164 (86.21%)
		Smoker	22 (11.6%)
		Non-smoker	3 (1.6%)
	Country	One country	177(93.7%)
		U.S.	89
		Netherlands	20
		Norway	12
		UK	11
		China	8
		Greece	6
		Australia/Denmark/ South Korea	4
		Portugal	3
		Egypt/ South Africa/Mexico/ Switzerland	2
		Botswana/Cyprus/France/Germany, Iran/Malaysia/Singapore	1
		Two countries (involving 2 of US, UK, Australia, South Korea, and Malta)	5 (2.6%)
		Three countries (Australia, Mexico, and France)	1 (.5%)
		Four countries (US, UK, Australia, and Canada; Hungary, Slovakia, Czech, and Romania)	2 (1.1%)
		Six countries (involving 6 of UK, Netherlands, Norway, Demark, Portugal, Spain, Germany, France, Italy, Belgium, Finland, and Balkan Countries)	3 (1.6%)
		Seven countries (UK, Demark, Belgium, Canada, Germany, Slovakia, and Turkey)	1 (.5%)

For the types of smoking examined, 157 (83.1%) studies investigated cigarette smoking, 12 (6.3%) examined unspecified tobacco use, 3 (1.6%) analyzed e-cigarette smoking, and the remaining 17 studies (9.0%) probed more than one smoking type by including alternative products (e.g., chewing tobacco, snuff, dip, cigars, pipes, shisha, marijuana, and smokeless tobacco).

Research designs and data collection methods

Of the 189 studies, 15 (7.9%) were experimental, 133 (70.3%) reported cross-sectional or survey data, and 41 (21.72%) were longitudinal surveys. To qualify as an experiment/trial, a study must contain a manipulation/treatment/intervention, and its effects must be measured. The remaining studies were surveys that did not involve manipulation/treatment/intervention. A longitudinal study must contain at least two waves of data collection, but without treatment. A cross-sectional survey design analyzed data collected in a one-shot survey or only one of a series of longitudinal surveys.

Social norms examined

A total of 347 social norms (146 DNs and 201 INs) were measured, excluding five moral norms and one personal norm, because they are reflections of an individual’s conviction of morality and internalized values and criteria, instead of reflecting direct social influence (Mackie et al., 2015). The vast majority (79.5%) examined general and/or specific smoking and related norms, whereas 20.5% examined antismoking norms (Table 2).

Table 2. Social norms examined (N = 347).

	DN (n = 146)	IN (n = 201)
Smoking norms (n = 276, 79.5%)
Smoking in general (“smoking” with no other description in measures)	81	106
Smoking specified with frequency, timeframe, venue, and/or context	44	31
Smoking certain tobacco products	3	3
Offering cigarettes to others	1
Smoking intake	1
Miscellaneous attitudes labeled “norms” (e.g., smoking is cool)		6
Antismoking norms (n = 71, 20.5%)
Not smoking		13
Antismoking behavior	3	3
Smoking resistance	4
Antismoking policy		6
Quit smoking	8	26
Disapproval of various smoking behaviors		8

Conceptual bases of surveys and experimental studies

Of the total review sample, 106 (56.1%) mentioned at least one theory – with 48.2% using one theory, 4.2% naming two theories, 2.7% citing three theories, and .5% quoting four and five theories respectively. The remaining 43.9% or 83 studies did not invoke any theory; that is, 39.1% provided a literature review without citing a theory and 14.8% measured norms without referring to any conceptual framework. For the 106 studies that cited at least one theory, the RAA was the most cited theory (in 75.5% of the studies), followed by social cognitive theory (10.4%) and norm focus theory (9.4%). Norms measures largely followed the operationalization guidelines stated in the RAA. Refer to Table 3.

Table 3. Conceptual bases used (k = 106).

Theories Mentioned/Invoked		Cited as the only theory	Cited as one of the theories	Total (%) of studies citing the theory (k = 106)	Basis for Experimental Treatment (k = 16)
	Reasoned action approach (TRA, TPB, Extended model)	63	17	80 (75.5%)
	Social cognitive/learning theory/model (SCT)	6	5	11 (10.4%)	1
	Social norms approach/theory (SNA)	8	2	10 (9.4%)	2
	Norm focus theory (NFT)	5	3	8 (7.6%)	3
	Social identity theory/self-categorization theory	2	2	4 (3.8%)	1
	Theory of normative social behavior	2	1	3 (2.8%)
	The theory of triadic influence	1	2	3 (2.8%)
	The attitudes-social influence-efficacy model	2	1	3 (2.8%)
	Theory of collective action	1	0	1 (1%)	1
	The RISP model	1	0	1 (1%)
	Expectancy violation theory	1		1 (1%)	1*
	5 theories not closely related to norm (diffusion of innovation, the Integrated Model of Behavior Change, health belief model, theory of subjective culture, self-regulation theory)		1	1(1%)**

*Expectancy of violation theory was used to compare with social norms approach in one study

**Each of the 5 theories listed in the left column was cited only once.

The conceptual bases for the 15 experimental studies were specifically analyzed to examine the effects of norm-related constructs. Experimental designs included pre-and post- tests, treatment and control groups, pre-and posttests with control group, and one-shot between-subjects factorials. Three studies based on norm focus theory employed the key construct of norm accessibility. Two studies grounded in the social norms approach invoked the concept of correcting misconceived pro-smoking norms as a key element in the antismoking intervention designs. One study applied modeling elements from social cognitive theory in a 4-staged intervention. Another study used social identity theory and the effects of norm and personal (versus group-based) self-efficacy. Four studies used constructs (e.g., message ambivalence) derived from a deductive literature review but without a theoretical framework. Four studies reported the effects of complex, multi-staged intervention programs. TPB was not used for designing interventions, but TPB variables were often measured as the effects of interventions. Refer to Table 4.

Table 4. Norm-related conceptual basis, treatment, and findings for experimental studies (k = 15).

Conceptual Basis and Key Constructs	Treatment/Intervention	Key Findings
Norm focus theory (k = 3) Construct: norm accessibility	Comparisons between accessible versus less accessible normative messages	– Norm accessibility accounted for smoking behavior in addition to TPB (Rhodes & Ewoldsen, 2009) – Norm accessibility influenced information process such that smokers with an accessible pro-smoking norm tended to use the peripheral route. Norm accessibility was not related to intention to quit smoking (Rhodes et al., 2008) – Accessible peer antismoking norms reduced resistance to antismoking messages. Accessible pro-smoking norms increased readiness to engage in smoking (Rhodes et al., 2016)
Social norms approach (k = 2) Construct: overestimated pro- smoking DNs	Provision of information of accurate smoking prevalence to treatment groups, coupled with discussions and/or other forms of interventions	– Misconceptions about smoking prevalence were not reduced six weeks or one year later. Smoking prevalence at parties was reduced from 8% to 5% (Balvig & Holmberg, 2011) – The intervention changed judgment accuracy of DN and INs but not attitudes to smoking (Campo et al., 2004)
Social cognitive theory (k = 1) Construct: learning normative behavior	4-staged learning intervention to resist pro-smoking normative/conformity pressure	– Successful. More of those receiving interventions changed from smoking to not smoking, fewer changed from not smoking to smoke, and smoking prevalence dropped by 1.2% (Buller et al., 2008)
Social identity theory (k = 1) Constructs: personal self-efficacy and group-based self-efficacy	A 2 (PSE: high vs. low) x 2 (antismoking norm: weak vs. strong) design with group-based self-efficacy as a covariate	– Antismoking norm can produce paradoxical effects via either increasing or decreasing smokers’ motivation to quit depending on whether the strategy was individual or collective (Falomir-Pichastor et al., 2013)
Simple comparison or factorial design with constructs extracted from a deductive literature review (k = 4)	Interventions typically included norm messages that were often compared with fear appeal, ambivalent messages, and/or counter-tobacco-marketing ads	– Nicotine dependence was not mediated by subjective norm (Guilamo-Ramos et al., 2011) – Information on anti-tobacco use IN reduced message ambivalence and decreased pro-tobacco attitudes and intention to use tobacco (Hohman et al., 2016) – Compared to health consequence and counter-tobacco marketing messages, social norm messages were least effective in persuading viewers to intend to quit smoking (Murphy-Hoefer et al., 2008) – Both DN and IN anti-secondhand-smoking messages largely triggered DN and IN perceptions which positively predicted intention to engage in anti-secondhand-smoking behaviors (Paek et al., 2014)
Intervention program with multiple theoretical perspectives not clearly identified, no main theory (k = 4) Some or all of TPB variables often measured	Interventions often included Web feedback, health information provision, coaching, skills training, use of media, and/or social media interaction.	– Intervention was minimally effective as DN and IN were improved without change in behavior (Bühler et al., 2016) – Social norm predicted intentions to smoke (Cremers et al., 2015) – Family norm was significantly correlated with prevalence of abstinence (J. A. Andrews et al., 2008) – Interactive social media communication led to information search, changed attitudes, changed perceived social norms about smoking behaviors, and reduced smoking intention (Namkoong et al., 2017)

Dependent variables

Variables that were predicted by norms, either based on the invoked theoretical frameworks or the research purposes, were treated as outcome/dependent variables (DVs), which were classified into four categories: attitudes and cognitions, self-efficacy, behavioral intentions, and behaviors. Smoking-related behaviors were the most frequently examined dependent variables, followed by behavioral intentions. Attitudes and cognitions, and self-efficacy were third and fourth respectively (Table 5).

Table 5. Dependent/outcome variables predicted by social norms.

Dependent Variables	Frequency
Attitudes and Cognitions	(n = 29)
Perceived smoking harms/risks	9
Attitudes toward antismoking policies	5
Favorable attitudes toward smoking (smoking enjoyment, smoking acceptance)	4
Attitudes toward smoking	7
Cognitive response	4
Self-efficacy	(n = 10)
For smoking cessation	2
For smoking resistance	8
Behavioral intentions	(n = 99)
Intention to smoke (sometime with a timeframe)	60
Intention to accept cigarettes	1
Intention to engage in antismoking actions (e.g., antismoking persuasion, support for antismoking policies)	7
Intention not to smoke	4
Intention to quit smoking	26
Behaviors	(n = 109)
Smoking in general	50
Smoking frequency	26
Smoking intake	21
Social smoking behaviors	3
Compliance with antismoking policies	2
Anti-secondhand smoking behavior	1
Addictive behavior	1
Cigarette purchase	1
Smoking cessation attempts and incidents	4

Independent variables/covariates

Excluding the typical demographics, all other variables examined with norms to predict dependent variables were classified as independent variables/covariates. Of all 25 categories identified, the most frequently examined ones included self-efficacy, smoking intention/prevalence/behavior, information variables, and family influence. Interestingly, approximately 50 variables were measured in a way similar to descriptive norms, and 14 variables were similar to injunctive norms; but they were treated as covariates and were not labeled or analyzed as norms (Table 6).

Table 6. Independent variables and covariates (excluding demographics and norms).

Independent Variables/Covariates	n
Personality trait (i.e., self-esteem – 7*, sensation seeking/risk taking – 5, depression – 4, anxiety – 2, extroversion, optimism, psychological wellness, self-identity, social ability, social hostility, self-control, smoking cognition, originality)	25
Cognition (i.e., self-identity, smoker identity, parent identity, ethnic identity)	4
Affect (i.e., negative emotions about smoking such as guilt and regret, guilt – 4, anticipated or actual affect after quit – 6)	10
Attitude accessibility	3
Attitudes toward smoking and related matters (i.e., general attitude toward smoking – 6, attitude toward smoking/tobacco company – 2, attitudes toward smoking policies – 19)	27
Attitudes toward secondhand smoking	5
Beliefs and perceptions about smoking benefits and risks (i.e., health risks – 30, social risks – 9, mixed health and social risks – 3, social benefits – 8, physical and psychological benefits – 5)	55
Beliefs and perceptions about quitting smoking (i.e. benefits of a quitting smoking – 11, risks about quitting smoking – 4, overall risks and benefits)	16
Beliefs and perceptions about health risks of secondhand smoking	6
Perceptions about e-cigarette smoking (i.e., health risks – 3, social risks, benefits – 2, gratification, innovativeness)	6
Expectancy of smoking/quit smoking outcomes	10
Self-efficacy (i.e., control over smoking timing, intake, and not smoking – 14, refraining from smoking and smoking situations – 19, stopping others’ smoking – 2, resisting cigarettes – 23, making home smoke free – 3, quitting smoking – 13, control of life in general – 2, managing recovery, discussing cessation strategies, coping with stress, decision-making, and social skills)	81
Cigarette accessibility	8
Smoking intention/prevalence/behavior (i.e., intention to smoke – 7, smoking prevalence – 3 smoking status – 22, pastor cigarette use 22, cigarette intake/frequency – 6, smoking behaviors – 4, addictive smoking behaviors – 6, use of different tobacco products – 3, pro- smoking action)	72
Cessation intention/behavior (i.e., physician history/attempts – 10, quit intention – 4)	14
Antismoking/anti-secondhand smoking behavior	4
Other risky behaviors (i.e., alcohol consumption – 10, deviant/risky behaviors – 3)	13
Community characteristics (i.e., Community support for Tobacco control, school connectedness, Neighborhood SES, neighborhood influence, community affirmation, community lunch voucher urban versus rural, city characteristics)	8
Culture (i.e., acculturation – 4, individualism/collectivism – 3, cultural environment)	8
Information/media/observation variables (i.e., knowledge/awareness of smoking consequences – 8, media exposure – 15, Internet/media use – 3, interventions/education – 10, smoking modeling – 7, exposure to antismoking messages – 3, exposure to pro-smoking messages)	49
Social support (i.e., general social support – 6, collective efficacy, professional support, social capital)	9
Active Family influence (i.e., communication about smoking such as risks, challenges, and guidance – 8, setting house rules for smoking – 9, parental monitoring/supervision/warning/sanctioning/psychological control – 16, parental support and parenting style – 6, relationships within the family – 4, pressure to smoke from parents – 2)	41
Friend influence (i.e., offering cigarettes – 3, peer pressure to smoke – 2, peer selection and friend socialization – 6, popularity among peers – 3, time spent with friends, isolation, interpersonal discussion, friend encouragement to quit smoking – 2, peer group identity – 8)	27
Other influences (i.e., teacher influence – 2, workplace influence – 2)	4
Variables similar to but not labeled descriptive norms and treated as covariates (i.e., family smoking – 22 [generalized 6, parent/s 16, siblings 2], friend/peer smoking – 18, friend not smoking – 1, important others smoking – 3, adult smoking – 4, peer exposure to smoking – 2)	50
Variables similar to but not labeled injunctive norms and treated as covariates (i.e., peer attitudes/approval/pressure/sanctioning about smoking – 9, parent attitude to smoking – 1, sibling/student approval/disapproval of smoking – 2, generalized others approval/acceptance of smoking – 2)	14
Religious influence	3
Other variables (e.g., norm typicality, adult risk behavior, literary role model, asthma, trouble with teachers, liking for school, woman’s role in the family)	10

*Frequency number is indicated for each variable; when no number is given, the frequency is 1.

Relationships between social norms and dependent variables

Several patterns emerged from cross-sectional studies. INs and DNs, when both were measured, exhibited moderate to high correlations, particularly when the reference groups were peers, friends, young individuals of the same age, and fellow students. By and large, pro-smoking DNs/INs positively predicted pro-smoking attitudes, intention to engage in smoking-related behaviors, and various indicators of smoking behaviors, such as cigarette intake, frequency, and social smoking behaviors. Antismoking DNs/INs largely positively predicted antismoking attitudes, intention not to smoke, and antismoking behaviors, such as persuading others not to smoke. Norms exhibited direct and indirect statistical impacts on smoking-related behavioral intentions and behaviors. Social norms were correlated with the examined dependent variables that included attitudes and cognitions, self-efficacy, behavioral intentions, and behaviors. Most of the studies generated correlational evidence supporting relationships among some or all of the TPB variables. See Table 7.

Table 7. Relationships between social norms and dependent variables in surveys.

Dependent Variables	Subtotal n	Pro-smoking Norms			Antismoking Norms
		+*	-	0	+	-	0
Pro-smoking attitudes and cognitions	10	7	1			1	1
Pro-smoking behavior intentions	53	41		2		7	3
Pro-smoking behaviors	93	76		7		9	2
Antismoking attitudes and cognitions	16		3	2	11
Antismoking Self-efficacy	8				7		1
Antismoking behavior intentions	37		9	3	25
Antismoking behaviors	6				5		1

*± indicates a positive/negative relationship, and 0 indicates no relationship.

For experimental studies, norm accessibility (from norm focus theory) influenced smoking behavior, accessible antismoking norms reduced resistance to antismoking messages, and accessible pro-smoking norms increased readiness to smoke. Based on social norms approach, the provision of accurate norms as a key element of intervention reduced smoking prevalence and changed perceptual accuracy of INs and DNs. Social learning, as the key element of social cognitive theory resulted in quitting smoking and preventing smoking onset. Other interventions showed mixed results in that some norm-based messages were ineffective in persuading viewers to quit smoking (e.g., Murphy-Hoefer et al., 2008), other norm-based messages predicted intention to engage in antismoking behaviors (e.g., Paek et al., 2014), and some other interventions were minimally effective in changing perceptions of DN and IN, but not behavior or behavioral intention (see Table 4).

Discussion

A total of 189 studies were reviewed; the vast majority of surveys generated similar correlational findings regarding positive relationships between pro-smoking norms and intention to smoke/smoking-related behaviors, whereas antismoking norms predicted antismoking intentions and behaviors. The volumes of the correlational findings seem to have repeatedly reported similar norm–behavior relationships. These survey-based studies appear repetitive and offer similar intervention implications. This begs the questions as to whether more such norm–behavior correlational studies are needed. Experimental studies that tested intervention approaches offer much more interesting results that bear direct relevance to actual interventions.

The intervention utility of theories of norms

Of the 189 studies, 56.1% mentioned at least one theory, whereas the remaining studies did not mention any theory. The most frequently cited theory was the RAA, and social cognitive theory (SCT) and social norms approach (SNA) were the distant second and third respectively. Interestingly, none of the 15 experimental studies tested any RAA-based intervention. In fact, when RAA was cited in several experimental studies, the RAA constructs were measured as dependent/effect variables, which were still correlated with each other. RAA did not play a role in steering the intervention designs, but its norm instrumentation guidelines were widely applied.

N focus theory and its key construct of norm accessibility, social norms approach and its central intervention function of correcting misconceived norms, and social cognitive theory and its central thesis of acquiring social normative behaviors via modeling/cognitive learning, respectively, served as the key frameworks for antismoking intervention programs tested in the experimental studies. The effects of these intervention programs were generally desirable, but mixed to a certain extent (Table 4). On one hand, these interventions resulted in desired changes, such as reduced resistance to antismoking messages and reduced smoking prevalence. A highly successful theory-based intervention was an intensive learning program based on SCT (Buller et al., 2008) for resisting pro-smoking normative pressure. On the other hand, norm accessibility was unrelated to intention to quit smoking; misconceptions about smoking prevalence were not reduced after six weeks or longer (e.g., Rhodes et al., 2008). Social identity theory was the fourth theory tested in an experimental study, which showed that collective identity interacted with the strength of antismoking normative messages (Falomir-Pichastor et al., 2013).

Of the other experimentally tested intervention messages/programs, the results were even more mixed. Anti-tobacco IN and/or DN messages, for example, reduced pro-tobacco attitudes and intention to use tobacco (Hohman et al., 2016) and predicted intention to engage in anti-secondhand-smoking behaviors (Paek et al., 2014). However, social norm messages, compared to other forms of messages, were the least effective in persuading viewers to quit smoking (Murphy-Hoefer et al., 2008). Overall, these intervention programs were minimally effective because no change in behavior was observed. The limited experimental research indicates that the theories useful for intervention designs were the SNA, NFT, and SCT.

Surprisingly, the by-far-most-cited RAA did not result in any experimentally tested antismoking intervention program. RAA appears to be a theoretical approach that predicts correlations among attitudes, norms, behavioral intention, and behavior without providing a method that directly changes norms to change behaviors. Beyond continuing to guide cross-sectional studies using the theoretical constructs, RAA alone does not seem to be capable of generating new research that informs intervention designs; its intervention utility is severely compromised. Future research must transcend the RAA paradigm to focus more on the norm–behavioral change link.

Social norms examined

The 189 studies examined a total of 347 social norms in smoking contexts. Approximately 80% pertained to descriptive/injunctive norms that primarily reflected smoking in general and in certain contexts (e.g., venue and social situation); but rarely concerned smoking different products, offering cigarettes to others, or cigarette intake. These smoking norms were measured mostly to predict smoking intentions and behaviors. The intervention implications of the norms of generalized smoking seem rather limited. Most likely, practitioners can provide accurate information about smoking prevalence to correct misconceived DNs to motivate individuals not to smoke or reduce smoking. Smoking norms tied to certain contexts offer more information for intervention targeting. Although rarely mentioned in the 189 studies examined, inoculation-based programs that carefully forewarn individuals of normatively risky contexts (such as parties) can work.

The present review discovered that researchers rarely discussed correcting INs as a means of intervention. The theoretical principle of the social norms approach should also be applied to correcting misconceived injunctive norms. For example, when individuals perceive a higher percentage of peers approving of smoking than the actual percentage, misconceived IN can be corrected by providing the lower actual percentage. Further, researchers examined smoking norms by focusing on the act of smoking without adequate attention to other acts that directly or immediately lead to smoking. For example, in one study that examined offering cigarettes as a normative social smoking behavior, passing cigarettes to others was generally met with little resistance, particularly during smoking experimentation among teenagers (Sheer et al., 2018). Interventions targeting the reduction of cigarette passing can reduce social cigarette sharing. Relevant norms can include peer DN and IN of offering cigarettes to others. Paradoxically, focusing on smoking norms can inflate the perception of prevalence of such risky behavior.

To avoid inflating risky behavior incidence, smoking prevention can focus on normative antismoking behaviors for message design. The existing norm-guided research appears largely repetitive without adequate efforts to explore positive behaviors for preventing smoking. Smoking resistance was examined in DNs, but not in INs. Plausibly, if an individual discovers that most of their smoking peers do not feel offended when their cigarette offers are rejected, they may become more motivated to resist cigarettes. Thus, the research focus can be shifted to the under-examined antismoking norms, such as cigarette resistance, persuading others not to smoke in front of children, and simply asking people not to smoke.

Dependent variables and norms

The majority of dependent variables were behavioral intentions and behaviors, with only a small percentage investigating attitudes and cognitions. Unsurprisingly, both pro-smoking INs and DNs positively predicted pro-smoking intentions and behaviors, whereas antismoking INs and DNs positively predicted their antismoking counterparts.

Independent variables

A systematic review of IVs/covariates (Table 6) typically informs commonly examined moderators to the relationship between norms and intention/behavior. In descending order, the most frequently examined IVs/covariates are self-efficacy, information-related variables (e.g., knowledge and exposure to health information), health beliefs, and active family influence (e.g., setting smoking-free house rules). The seldom-measured variables were not good candidates as moderators for future meta-analyses. For example, culture and attitude accessibility were not adequately examined for inclusion in a meta-analysis. Several variables that are similar to measures of norms were treated as covariates but were not analyzed as norms (Table 6). For example, measures of family smoking and peer smoking were nearly identical to DNs of the same reference groups, and peer attitudes/pressure/sanctioning was similar to peer IN. These similar measures could create conceptual confusion and generate findings of meaningless cyclic relationships.

Due to the cross-sectional nature of 92% of the reviewed studies, IVs and DVs were conceptually interchangeable. To illustrate, attitudes and cognitions, which often serve as independent variables, were also measured as dependent variables, whereas behavioral intention and behavior were treated as independent variables in several studies. Thus, to understand the pattern of relationships among those variables, researchers can use a theoretical framework to guide meta-analytical reviews because different theories have their own sets of constructs for independent and dependent variables. However, conducting a meta-analysis of the effects of norms can be challenging for this particular set of studies on norms and smoking.

Challenges for conducting meta-analyses

The researcher hoped that the present systematic synthesis would inform the design of a meta-analysis that assesses the extent to which norms derived from different theoretical perspectives exert relative impacts on smoking-related attitudes, intentions, and/or behaviors. Because of the over-application of the RAA and the under-application of other theoretical perspectives, secondary data based on the other theories are insufficient. The relative theoretical impacts cannot be assessed. For the vast majority of the cross-sectional studies, norms were measured by following the guidelines of the RAA. Possibly, a meta-analysis can be conducted to compare the relative correlational predictive power of DNs and INs. However, correlational predictive power cannot be translated into effect sizes of norm-based intervention experiments. At most, those largely RAA-based norm–behavioral intention/behavior correlations suggest that changing norms can lead to changed behavior, but offer few practical guidelines for designing interventions. A notably meaningful meta-analysis should focus on the effects of the normative elements of interventions/experiments.

Unfortunately, of the total of 189 studies examined, only 15 were experiments. Eight studies utilized five different theories to design the norm elements in the experimental treatments/intervention, whereas the remaining six either compared normative messages with other messages or used complex intervention packages, but none used the RAA for experimental treatment. Consequently, these 15 experiments are too few and too varied for the researcher to feasibly conduct a meta-analysis. However, a future analysis of more available norm-based experimental studies should be conducted to assess the effect sizes of norm elements from different theories.

Future research on norms and smoking for interventions

The present synthesis implies several needs for norm-based research for antismoking interventions. Instead of the RAA, more intervention-relevant norms theories (e.g., NFT, SNA, and SCT) should be applied for designing intervention trials. Effective interventions sometimes require an integration of several theories, including the RAA. For example, interventions can trigger highly accessible norms (NFT) to motivate individuals to engage in cognitive learning (SCT). Although DNs and INs were both examined as predictors of smoking-related attitudes, intentions, and behaviors, correcting misconceived DNs was frequently mentioned as a method for intervention, with little discussion of targeting INs. More research can be conducted to investigate the effects of INs as an intervention. Finally, although antismoking studies used samples from various countries, the research designs predominantly adopted the RAA as the guiding conceptual framework. A few studies that measured collectivism/individualism suggested that norms had a stronger impact on collectivistic cultures. Thus, the social norms approach may be more promising for antismoking interventions in these cultures. Culture-specific intervention strategies will still need solid research that measures specific traits of a given culture, instead of treating all collectivistic/individualistic cultures as one.

Additional information

Funding

This project was supported by [GRF 12633816] University Grants Committee of Hong Kong.