Adolescent girls' constructions of smoking identities: implications for health promotion

Journal of Adolescence 1997, 20, 43–56

Adolescent girls’ constructions of smoking identities:implications for health promotion

BARBARA LLOYD, KEVIN LUCAS AND MADELINE FERNBACH

A model of social identities constructed in terms of group membership is used toexplore the incidence of cigarette smoking among adolescent girls. A sharpincrease in girls’ smoking at about 14 years of age related both to age and topuberty is demonstrated in a discriminant function analysis. The model is success-ful in identifying membership in never smoked and regular smoker groups but doesnot identify occasional smokers. Girls’ ratings of 13 identity descriptors are sum-marised in similar, three scale nonsmoker and smokers identities. The nonsmokeridentity is rated higher on the “mature” and “sensible” scales while the smokeridentity is rated higher on the “fun-loving” scale and a fourteenth descriptor,breaks rules. The descriptors loading on the “mature” scale are contested suggestingthat members of both smoking and nonsmoking groups value cool, popular, grown-up and makes up own mind. Material from focus group discussions demonstratesthat the particular meanings of some descriptors may vary with group membershipand values. Both the prevalence analysis and the focus group material support thehypothesis that interventions are best targeted at specific groups and that theoccasional smoker identity needs to be targeted specifically in interventionsdesigned to reduce smoking among adolescent girls.

1997 The Association for Professionals in Services for Adolescents

Introduction

This paper reports initial findings from a large quantitative and behavioural study ofadolescent cigarette smoking. The research was motivated by a Department of Healthconcern about the high incidence of smoking among 13- to 15-year-old girls (Diamond andGoddard, 1995). An exploration of young people’s views about the meanings of smokingwas undertaken in order to provide new approaches in the development of interventionstrategies. This report focuses primarily on the responses of girls, since it is their cigarettesmoking uptake that has been a matter of particular concern.

Moscovici’s (1981) concept of “social representations” was used to explore the meaningswhich adolescents construct about smoking. Social representations are defined as theproducts or features of social groups which form organised systems of “values, ideas andpractices” (Moscovici, 1973, p. xiii). It is through access to shared social representationsthat individuals are able to understand the structure of social life and to communicate withothers. The interdependence between social representations and the collectives for whichthey function means that social life is always viewed as a construction, rather than beingtaken as a given. Social representations of smoking are used as a reference point for

Reprint requests and correspondence should be addressed to B. Lloyd, School of Social Sciences, University ofSussex, Falmer, Brighton, East Sussex BN1 9QN, U.K.

0140-1971/97/010043+14/$25.00/0/ad960063 1997 The Association for Professionals in Services for Adolescents

44 B. Lloyd, K. Lucas and M. Fernbach

analysing the social identities of young adolescent girls who smoke cigarettes and those whodo not smoke.

The concept of identity was introduced into the social sciences by Erickson (1946, 1950)who used the psychoanalytic term ego identity. Highlighted by its pivotal position betweenchildhood and adulthood, Erickson saw issues of identity as particularly salient inadolescence. He postulated that the major developmental issue in adolescence was theestablishment of a sense of identity. The decision to become a smoker or not to smoke isimplicated significantly in adolescent identities.

In this study, the concept of social identity was employed to focus on the sharing ofsignificant characteristics with others. This approach helps to maintain an awareness thatindividuals are constructed in terms of the groups of which they are members (Duveen andLloyd, 1986). This use of the term social identity has origins in the social psychologicaltheories of Moscovici (1973, 1976, 1981, 1984, 1988) and Tajfel (1981, 1982). Individuals’social identities are held to reflect the social representations of the groups in their societythat are significant for them. The development of social identities draws upon resourcesmade available through the reconstruction and internalization of these socialrepresentations. Our basic premise is that individuals are inextricably interwoven in thefabric of social relations within which their lives are lived. A representation of the“individual” divorced from the “social” is therefore theoretically and empiricallyinadequate. For adolescents in particular, there is no pure “individuality” that can beapprehended independently of social relations and their associated social representations.

Duveen and Lloyd (1990) observed that some social representations impose animperative obligation on individuals to adopt a particular social identity. This is the casewith representations of age, gender, or ethnicity, where individuals are generallyconstrained to construct prescribed social identities. In other instances, the influence ofsocial representations is exercised through a contractual obligation rather than animperative one. In these cases an individual joining a social group contracts to adopt aparticular social identity. Social representations of the law provide an example of acontractual obligation. For example, when a person chooses to train as a solicitor, they mustabide by the regulations of The Law Society. In a less formal but no less restrictive manner,adolescent peer groups shape the behaviour of their members. The development of thecompetence to participate in society relies upon adolescents’ capacity to re-construct thesocial representations of their community and create various social identities. In earlyadolescence an important new identity is that of the smoker. Tajfel’s stress on theemotional and evaluative significance of group membership is particularly important whenconsidering smoking.

Our empirical work is an exploration of the nature of adolescent girls’ socialrepresentations of cigarette smoking, and the construction of their social identities inrelation to smoking. Our data analyses have three objectives. First, we explore theprevalence of cigarette smoking among girls and boys. Secondly, we show how girlsconstruct two distinct social identities: a nonsmoking identity and a smoking identity.Finally, we relate these identities to girls’ own membership in groups of nonsmokers,occasional smokers and regular smokers.

Method

The quantitative study is based upon a questionnaire survey of all pupils in Years 7 to 11

45Smoking identities in adolescent girls

inclusive from six East Sussex secondary schools. It was undertaken in the Autumn term1994 and again in Summer term 1995. The original questionnaire comprised 246 variables,including conventional demographic information, and items on smoking behaviour, bodyimage and eating, stress and coping, home environment, peer relations and social identities.The second wave questionnaire was essentially the same although three questions wereremoved, three additional descriptors added to the six identity questions and an item addedto indicate whether this was an individual’s first or second testing. In the Autumn 4771useable questionnaires were collected and 4487 in the Summer. Systematic proceduresproduced 3521 matched questionnaires.

Following Charlton et al. (1985), anonymity and confidentiality were ensured byemploying a self-created personal code number and providing each pupil with a sealableenvelope. It is argued that this method promotes accurate reporting of smoking prevalenceand produces results as valid as research which employs cotinine assays.

Smoking group membership was measured using a conventional five-response question:

(i) I have never smoked.(ii) I have smoked (if only for a few puffs).

(iii) I have smoked one cigarette in the past 4 weeks.(iv) I usually smoke one or more times a week.(v) I usually smoke at least once a day.

These responses were combined into three categories in order to represent meaningfulsocial groups. Individuals who had never smoked (i) were differentiated from those who hadexperienced cigarette smoking and labelled Never Smoked. In order to identify the group ofindividuals who had experimented with cigarettes, (ii) and (iii) were combined, and thisgroup labelled Occasional Smokers. Finally, (iv) and (v) were combined to identify RegularSmokers.

Measures of social identity were derived from a review of research findings concerningimages of adolescent smokers and nonsmokers (Bewley and Bland, 1978; Barton et al., 1982;Kannas, 1985). The aim was to select from these studies descriptors that girls might usethemselves when considering nonsmokers and smokers. The terms functioned as girls’“social representations” of smoking. Fourteen such descriptors were consistently employedto measure six identities, two concerned with self identity, and four concerned with boy andgirl smoking and nonsmoking identities. Each of these descriptors was rated on a six pointscale from 0 to 5. Table 1 shows the girl nonsmoker item with 14 descriptors. Thebehavioural phase of the research comprised an impression formation task (IFT) followedby focus group discussions. Girls were recruited from volunteer friendship groups containingat least two girls. Four of the six schools that participated in the quantitative study took partin this phase. In each school, four groups were drawn from Year 7 and four from Year 9,making a total of eight per school. Group size varied from two to six girls.

Two videos were produced for use in the IFT. Both portrayed a 15-year-old girl describinga typical weekend with her friends. The videos were identical except that in one the girlwas a smoker, and in the other she was a nonsmoker. A short questionnaire wasadministered to elicit girls’ own descriptions and evaluations of the main character. The 14descriptors which formed the social identity items in the full questionnaire were used toelicit further ratings of the character. IFT data were collected in 30 groups; 125 girlscompleted the IFT identity questionnaire.

The subsequent focus group discussions were guided to ensure that smoking biographies,


social representations of smoking, social identities, health-related beliefs and values, andperceptions of anti-smoking education were all considered. All discussions were audio-tapedand were subsequently transcribed. These data were indexed using the NUD*IST (Non-numerical Unstructured Data Indexing, Searching and Theorising) program.

Results

Questionnaire results are presented in three sections. The first is an analysis of smokingprevalence for girls and boys in Years 7 through 11. The second is an examination of thenature of girls’ perceptions of nonsmoking and smoking identities. These identities areexamined in terms of the smoking group membership of the girls themselves in the thirdsection. Data from the IFT are included where relevant.

Girls’ and boys membership in smoking groupsTable 2 shows the proportions of girls and boys by chronological age in the three smokinggroups—never, occasionally and regular.

Initial examination of these data suggest that a simple comparison of smoking prevalenceby age and gender is not adequate as smoking is related to age and physical maturity. Girlsreach puberty earlier than boys (Falkner and Tanner, 1978) and their smoking behaviour isstrongly influenced by both age and maturity. Girls who have not reached puberty and areless than 14-years-old are less likely to smoke than are boys of similar age but girls in thisage range who have achieved puberty are more likely to smoke cigarettes than are boys.

Statistical analysis is limited by levels of measurement and distributions of values.Membership in smoking groups is ordinal but the intervals are not equal and the values arenot normally distributed. Both gender and reaching puberty are nominal. Despite a strongcorrelation between age and puberty, the age range 11 to 17 years results in few of theyoungest individuals having reaching puberty and few of the oldest not having reached it.

Figure 1 is derived from a multiple discriminant analysis using a model developed frominsights gained from a LOWESS plot (Cleveland, 1979). A variable was created to take

Table 1 Girl Nonsmoker Identity Item: “Think carefully about a girl who doesn’t smoke. Whatis she like?”

uncool 0 1 2 3 4 5 coolwimpish 0 1 2 3 4 5 toughsticks to rules 0 1 2 3 4 5 breaks ruleschildish 0 1 2 3 4 5 grown-upunhealthy 0 1 2 3 4 5 healthyunhappy 0 1 2 3 4 5 happyfollows others 0 1 2 3 4 5 makes up own minddull 0 1 2 3 4 5 excitingunpopular 0 1 2 3 4 5 popularunattractive to opposite sex 0 1 2 3 4 5 attractive to opposite sexthick 0 1 2 3 4 5 cleverdoesn’t care about environment 0 1 2 3 4 5 cares about environmentdoesn’t like partying 0 1 2 3 4 5 likes partyingdoesn’t like the opposite sex 0 1 2 3 4 5 likes the opposite sex


17.0

Regular

Never10.5

Age (years)

13.0

Occasional

11.0 11.5 12.0 12.5 13.5 14.0 14.5 15.0 15.5 16.0 16.5

Figure 1. Smoothed estimates of smoking prevalence by gender, age and physical maturity. Genderby maturity: (- - -)=girl post-puberty; (—)=girl pre-puberty; (. . . .)=boy post-puberty; (- . - .)=boy pre-puberty.

account of the influence of the interaction of age and maturity upon girls’ smokingbehaviour. This variable, along with age, gender, maturity, and an age by gender interactionserved as discriminators. In order to verify responses about puberty, data from bothassessments were sometimes required. These analyses are based upon the matched samplebut report smoking behaviour in the Autumn term. All of these variables were significant atthe p<0·001 level (F statistics with df=3392,2).

The success of the multiple discriminant analysis is assessed by its ability to identify groupmembership. Inspection of Table 3 shows that 51% of the “grouped” cases were correctlyclassified. Adolescents who never smoke and who are regular smokers are effectivelyidentified, 64% and 71%, respectively, but membership in the occasional smokers group isnot well predicted, 20% by this analysis.

Table 2 Girls’ and boys’ smoking group membership: percentages by age

Age (years) 11 12 13 14 15 16 Total

Girls nNever % 88 69 58 37 31 23 968Occasional % 11 27 35 40 38 35 526Regular % 1 4 8 23 31 42 229

Total n 289 386 389 326 290 43 1723

Boys nNever % 81 70 58 51 46 55 1096Occasional % 18 25 36 34 39 16 533Regular % 1 5 6 15 15 29 157

Total n 320 409 358 351 304 44 1786


Girls’ constructions of nonsmoking and smoking identitiesTwo types of evidence for girls’ constructions of nonsmoking and smoking identities arepresented here. The first is girls’ responses to the “girl nonsmoker” and “girl smoker”identity items from the questionnaire survey in the Autumn term (n=2262). The secondcomes from the impression formation task (IFT) undertaken as part of the focus groupsessions. It is based upon girls’ ratings of the video actor on the same descriptors whichcomprised identity items in the questionnaire. Two analyses of the questionnaire data arepresented:

1. The factor structure for the girl nonsmoker and girl smoker identities.2. Differences between scale scores for these identities derived from the factor analyses.

The factor structure of the girl nonsmoker identityThe descriptor “breaks rules” was excluded from the factor analyses when preliminaryanalyses of the identity items revealed inconsistencies in the positive and negative valuingof the descriptor. The remaining 13 descriptors employed in rating the nonsmoking identitywere factor analysed using the principal components procedure from SPSS, and threefactors, each with eigenvalues greater than 1, were retained. These accounted for 59·8% ofthe variance. These factors were subjected to a varimax rotation, which allowed each of thedescriptors to be identified with the factor on which it had the highest loading. This yieldedthree distinct sets of descriptors, each of which was summed to create a scale. Thedescriptors identified with each of the three scales were analysed for reliability. Sevendescriptors loaded on the first scale. These were: “cool”, “tough”, “grown-up”, “happy”,“makes up own mind”, “exciting” and “popular”. We labelled this scale “mature”. It had amean value of 26·0 (highest possible mean=35, Cronbach’s alpha=0·86). The three itemswhich loaded on the second scale were “attractive to the opposite sex”, “likes the oppositesex”, and “likes partying” and it was labelled “fun-loving”. This scale had a mean value of11·4 (highest possible mean=15, Cronbach’s alpha=0·78). The descriptors “healthy”, “caresabout the environment”, and “clever” loaded on the third scale which was labelled“sensible”. It had a mean score of 12·4 (highest possible mean=15, Cronbach’s alpha=0·60).

The factor structure of the girl smoker identityThe procedures used to factor analyse the ratings of the girl nonsmoker were repeated forthose of the girl smoker. Three factors, each with eigenvalues greater than 1, were retainedand these accounted for 62·3% of the variance. The three factors were subjected to avarimax rotation, which allowed each of the descriptors to be identified with the scale on

Table 3 Discriminant function classification of smoking group membership

Actual group No. of cases Predicted group membership

Never Occasional Regular

Never 1999 n 1275 303 421% 64 15 21

Occasional 1025 n 408 200 417% 40 20 41

Regular 371 n 52 57 262% 14 15 71


which it had the highest loading and the descriptors identified with each scale wereanalysed for reliability.

The seven descriptors that loaded on the “mature” scale for the nonsmoker identity itemloaded on this scale. It had a mean score of 21·6 (highest possible mean=35, Cronbach’salpha=0·87). The descriptors which loaded on the second scale were the same as thosewhich loaded on the “fun-loving” scale for the nonsmoking identity. The mean score was9·2 (highest possible mean=15, Cronbach’s alpha=0·64). Three items, “healthy”, “caresabout the environment”, and “clever” loaded on the third, “sensible” scale. It had a meanscore of 6·62 (highest possible mean=15, Cronbach’s alpha=0·78).

Differences between nonsmoker and smoker identitiesThe similarity of the factor structures and resulting scales allows comparisons between thetwo identities (see Table 4). Analyses yielded significant differences between them on allthree scales. Girls rated the “mature” scale descriptors higher for the nonsmoker identitythan for the smoker identity (p<0·0001). Similarly, the nonsmoker identity was ratedhigher than the smoker identity on descriptors loading on the “sensible” scale (p<0·0001).Only on the descriptors which loaded on the “fun-loving” scale was the smoking identityrated higher than the nonsmoking identity (p<0·0001).

Table 4 Varimax factor loadings for girl nonsmoker and smoker identity

“Mature” scale “Fun-loving” scale “Sensible” scale

Nonsmokercool 0·73334tough 0·78315grown-up 0·58015healthy 0·61478happy 0·50041makes up own mind 0·62112exciting 0·70498popular 0·66431attractive to opposite sex 0·73788clever 0·65828cares about the environment 0·68785likes partying 0·70005likes opposite sex 0·89148

Smokercool 0·77432tough 0·74810grown-up 0·71880healthy 0·50104happy 0·66267makes up own mind 0·69176exciting 0·75660popular 0·62913attractive to opposite sex 0·51316clever 0·58611cares about the environment 0·67112likes partying 0·81912likes opposite sex 0·85556


Mean scores for the descriptor “breaks rules” were also significantly higher for the smokeridentity than for the nonsmoker identity (3·9 and 1·7, respectively, p<0·0001).

Nonsmoking and smoking identities constructed with IFT dataRatings of the video actor on each of the 14 identity descriptors were analysed in 14separate ANOVAs with three main effects: (1) Year in school (7 or 9); (2) smoking identity ofthe video actor; and (3) a simplified smoking group membership classification (never vs.ever smoked). Neither the two-way nor the three-way interactions were significant in anyof these analyses. In 9 of the 14 analyses there was a significant main effect for actor’ssmoking identity, i.e. nonsmoking identity (nsi) and smoking identity (si). As well asfinding a significant difference for the smoking identity of the actor, “clever” wassignificantly different in terms of the subjects’ own smoking group membership. “Attractiveto the opposite sex” was the only descriptor which differed significantly by year in school.

Results for the actor’s smoking identity provide another measure of girl nonsmoker andsmoker identities. The identity items in the questionnaire required girls to construct theirown images of a nonsmoker and a smoker while this measure is influenced by specificcharacteristics of the girl performing in the two videos. Results for the descriptor “breaksrules” replicate the findings reported in the previous section. There was a highly significantmain effect for actor’s smoking identity (p<0·001; xnsi=2·4, xsi=3·5). Higher ratings for thesmoker identity were significant for two other descriptors. These were “likes partying” (p<0·0001; xnsi=4·1, xsi=4·6) and “popular” (p<0·03; xnsi=4·1, xsi=4·6). The former term, “likespartying” loaded on the “fun-loving” scale on which the smoking identity was rated higher.The descriptor “popular” loaded on the “mature” scale on which the nonsmoking identityhad been rated higher.

Results for the remaining six significant actor identity effects are all in line with earlierfindings. The nonsmoking identity is rated higher on all of them. Three descriptors loadedon the “mature” scale. These are “makes up own mind” (p<0·02; xnsi=3·6, xsi=2·6), “cool” (p<0·01; xnsi=3·5, xsi=2·6) and “grown-up” (p<0·03; xnsi=4·0; xsi=3·2). The remaining maineffects for actor’s identity were the three dimensions which loaded on the “sensible” scale.These were “health” (p<0·0001; xnsi=3·9, xsi=1·6), “clever” (p<0·02; xnsi=3·6, xsi=2·7) and“cares about the environment” (p<0·0001; xnsi=3·3, xsi=2·0).

Smoking group membership and ratings of the nonsmoker and smokeridentities

These analyses were undertaken to assess the relationship between girls’ membership indifferent smoking groups and their ratings of the two identities.

“Mature” scaleAn ANOVA design with repeated measures was undertaken on ratings of descriptors thatloaded on the “mature” scale for the nonsmoker and smoker identities by year and smokinggroup membership (see Table 5). There was a significant effect for identity, the withinsubject factor (p<0·0001). There were no significant effects on the between subjects factors,year and smoking group membership. There was, however, an interaction between thewithin factor and smoking group membership (p<0·0001).


“Fun-loving” scaleAnother ANOVA with repeated measures was employed to compare ratings of the descriptorswhich loaded on the “fun-loving” scale for the nonsmoker and smoker identities by yearand membership in the smoking groups. These are shown in Table 6 below. There was asignificant effect for identity, the within subject factor (p<0·0001). There was also asignificant effect for smoking group membership (p<0·0001) but these factors did notinteract.

“Sensible” scaleA further ANOVA with repeated measures was carried out on ratings of the descriptors whichloaded on the “sensible” scale for the nonsmoker and smoker identities by year and smokinggroup membership (see Table 7). There was a significant effect for identity, the withinsubject factor (p<0·0001). There was also a significant effect for smoking group membership(p<0·0001). Again there was no interaction. There was also a significant differencebetween the occasional and regular smokers (Bonferroni’s post hoc test p<0·05).

“Breaks Rules” descriptor and smoking identitiesAlthough the descriptor “breaks rules” was excluded from the factor analyses as the result ofinconsistent valuing, it is included here to illustrate the influence of group membership.Ratings of the nonsmoking identity on the “break rules” descriptor varied significantly withsmoking group membership (p<0·0001, never smoked mean=1·6, occasional smokers=1·9,

Table 5 Mean “Mature” scale ratings of the nonsmoking and smoking identities by smokinggroup membership

Smoking group Nonsmoker identity Smoker identity

Never smoked 27·5 19·9Occasional smokers 25·2 22·9Regular smokers 22·4 25·2

Table 6 Mean “Fun-loving” scale ratings of the nonsmoking and smoking identities by smokinggroup membership


Never smoked 8·1 9·1Occasional smokers 8·1 9·4Regular smokers 7·3 9·3

Table 7 Mean “Sensible” scale ratings of the nonsmoking and smoking identities by smokinggroup membership


Never smokes 12·7 5·7Occasional smokers 12·2 7·2Regular smokers 11·7 8·4


regular smokers=1·8, potential maximum=5). Girls who never smoked rated a girl who didnot smoke as less likely to break the rules than did girls in either of the other groups(Bonferroni’s post hoc test p<0·05).

Ratings of a girl who smoked on the “breaks rules” descriptor also varied significantlywith smoking group membership (p<0·0001, never smoked mean=4·0, occasional smokers=3·8, regular smokers=3·7, potential maximum=5). Girls who never smoked viewed a girlwho smoked as more likely to break the rules than did girls in either of the other groups(Bonferroni’s post hoc test p<0·05).

Discussion

Analyses of the prevalence of cigarette smoking among 11- to 16-year-olds reinforce theconcerns of the Department of Health about its increasing incidence among adolescentgirls. One important question raised by these analyses is the differential influence of gender,age, puberty and the interactions of age and maturity and age and gender on occasionalsmokers when compared with adolescents who have never smoked or those who claim tosmoke regularly. The failure of these variables to predict membership in the occasionalsmokers group has implication for intervention. Social psychological factors related togroup membership may have a particular importance in influencing these individuals.

Analyses of the factors employed in the study of smoker and nonsmoker identities yieldedsome clear results. Regardless of smoking group membership there was consensus among allgirls about the “sensible” and “fun-loving” scales which make up smoking identities. Thenonsmoker identity was rated more “sensible” (based upon the descriptors “healthy”,“clever” and “caring about the environment”) than was the smoker identity. Results fromthe IFT also linked these descriptors to the actor as a nonsmoker providing further supportfor the hypothesis that the social representations which comprise the “sensible” factor arenot contested by nonsmoking and smoking group members. The smoker identity was ratedas “liking partying” more in the IFT. This was one of the descriptors which loaded on the“fun-loving” scale that characterised the smoker identity for both nonsmokers and smokersin the factor comparisons. Again this provides further support for the hypothesis that thesocial representations which comprise the “fun-loving” scale are not contested. Similarlythe smoker identity was rated higher on the descriptor “breaks rules” in both thequestionnaire and in the IFT.

The nonsmoker and smoker identities were contentious only on the “mature” scale. Inthe ANOVA for the “mature” scale, the significant interaction between smoking identitiesand smoking group membership makes the interpretation of IFT results for “popular”,“makes up own mind”, “cool” and “grown-up” more complex. The differential meanings ofthe term “grown-up” employed by nonsmokers and smokers in focus group discussionselaborates issues about this term. Closer examination of Table 5, which shows the meanratings on the “mature” scale, indicates that nonsmokers rate the nonsmoker identity higherthan they do the smoker identity, while regular smokers rate the smoker identity higherthan they do the nonsmoker identity. For neither the “sensible” nor the “fun-loving” scaleswas there a significant interaction of smoker identity by own smoking group membership.The highly significant interaction for the “mature” scale indicates that the seven descriptorswhich load on this scale (“cool”, “tough”, “grown-up”, “happy”, “makes up own mind”,“exciting” and “popular”) may be construed differently by the two groups but are viewed as


desirable by girls in both groups; regular smokers and girls who never smoke claim them fortheir own group identity. Indeed, in the IFT, the smoker identity was rated higher on the“popular” descriptor, while the nonsmoker identity was rated higher on “makes up ownmind”, “cool” and “grown-up”. In focus group discussions, both girls who never smoked andregular smokers used the descriptors “cool” and “popular” to describe themselves. Theinfluence of group membership on the ratings of different social identities may beinterpreted as evidence that valued attributes are claimed for one’s own group.

Conversely, the lack of significant interactions between smoking identities and smokinggroup membership may be interpreted as follows. The willingness of nonsmokers to rate thesmoker identity higher than their own group’s identity on descriptors which form the “fun-loving” scale may suggest that nonsmokers do not value “likes partying” and “likes theopposite sex” as highly as do smokers. The willingness of smokers to rate the nonsmokeridentity higher than the smoker identity on the “sensible” scale descriptors “healthy”,“clever” and “caring about the environment” suggests that girls who smoke do not valuethese dimensions as highly as do nonsmokers. Evidence from focus group discussionssupported such an interpretation of the quantitative analysis.

Focus group data provide an opportunity to explore the meanings of the terms used in theidentity items as descriptors or social representations. The terms “grown-up” and “makes upown mind” were elaborated within the groups of nonsmokers and smokers and revealedshifts in meaning. A lower rating of the smoker identity on the “grown-up” descriptor wasechoed by never smokers’ descriptions of smokers as childish, and of themselves as “grown-up”. “Grown-up” had an ambiguous meaning for smokers. When they applied it tononsmokers “grown-up” was used to imply “being like an adult” i.e. being responsible, butwhen it applied to themselves, “grown up” represented being sophisticated and worldlywise. Smokers believed that smoking gave them such an image. There is little pressure tomodify the meaning of descriptors where there is agreement on the value assigned to socialrepresentation by different groups. By contrast, the meaning of the descriptor “makes upown mind” was refined by nonsmokers. Nonsmokers were proud that they wereindependent whereas they described smokers as “sheep”. The term “makes up own mind”did not enter the conversations of the smokers.

Focus group discussions produced two descriptors, “hard” and “keen about school work”that had not been included as terms in the questionnaire identity items. The moreimportant and more frequently cited social representation was “hard”. This term was used todescribe those who smoked, but “hard” was also linked to overtly aggressive behaviour,particularly by Year 7 girls. Girls who were perceived as being “hard” enjoyed a very highstatus that was connected to their appeal to boys.

The other dimension to emerge in focus group discussions was attitude to school work.Smokers described their attitude as nonchalant and minimalist in their approach to schoolwork, and nonsmokers agreed with this description of smokers. Many nonsmokers describedthemselves as paying a great deal of attention to their work and claimed that they enjoyedworking hard and valued doing well for its own sake. Smokers recognised this portrayal ofnonsmokers but dismissed it as part of nonsmokers’ sycophantic approach to teachers.

Three smoker identities can be drawn from the focus group material. One shares manydescriptors or social representations with the nonsmoker identity yielded by thequestionnaire identity item which described a person who abided by the rules. In additionthe nonsmoker identity is viewed as conscientious about school work and interested inschool life. Such girls were likely to refer to girls who smoke as “stupid”, and cite damage to


health and the repugnant smell as reasons for not smoking. These girls are unlikely to haveany close friends who smoke.

The girl smoker identity was elaborated in focus group discussions to include rebellionagainst authority, lack of interest in school work or activities, and a pursuit of “risky”behaviour. There was some evidence that Year 9 girls who smoked participated infriendship groups that included both smokers and nonsmokers but issues around smokingcaused tension among group members.

A third group (occasional smokers) were characterised by their ambivalent attitudetowards smoking. They experimented with cigarettes and viewed smoking as somethingthey might take up in the future. They were likely to have more nonsmoking than smokingfriends but did not differentiate between smokers and nonsmokers as sharply as nonsmokersdid. In the analyses of smoking prevalence it was difficult to identify predictors of smokingfor this group.

Applications

The evidence that the social representations of smoking that girls adopt is linked to thesocial groups in which they situate themselves has implications for future smokingintervention programmes. Early school-based programmes highlighted the damage to healthcaused by smoking, but it is now appreciated widely that for the majority of adolescents therisk of developing smoking-related diseases in middle-age is of little, if any, consequence.Two influential studies (Evans, 1976; Evans et al., 1978) prompted a shift in the emphasis inintervention programmes towards withstanding social pressure. Subsequently a succession ofprogrammes, employing various techniques, sought to equip adolescents with the socialskills required to resist “peer pressure” to smoke. However, such approaches neglect toconsider the meaning of smoking for adolescents.

The data in this study demonstrate that there are different routes to achieving a desiredidentity depending upon membership in a group of either nonsmokers or smokers. Thefactorial structure of the nonsmoking and smoking identities is very similar, but themeanings attached to their constituent representations depends upon group membership.There is little pressure to modify the meaning of descriptors where there is agreement onthe value assigned to social representation by different groups. When a social representationis not contested its value does not differ between groups of regular smokers and never-smokers. The descriptors loading on the contested scale, “mature”, are valued by bothgroups though their meanings may be modified to be consistent with other group values.Among adolescent girls who smoke, cigarettes are important in establishing a desirableidentity.

In terms of preventing smoking uptake, adolescent girls have conventionally been viewedas an homogeneous group. Analyses in terms of membership of different groups and datafrom different smoking identities suggest that “blanket” approaches to anti-smokinginterventions, whether based upon messages of risk or upon developing refusal skills, areinappropriate. Indeed, two of the most commonly used programmes in the U.K. have beenshown to be ineffective (Nutbeam et al., 1993; Michell, 1994).

Simple interventions are likely to result in differing responses because membership ofdifferent groups results in adolescents attaching different values to social representations ofsmoking. The comparison of groups on the “sensible” scale demonstrated that nonsmokers


value health. Most interventions targeted at these nonsmokers would be successful if theyincluded health-related issues. Conversely, such interventions would have little impact onregular smokers. Whilst regular smokers ascribe the “sensible” descriptors to nonsmokers,they do not actively desire these attributes. The smoker identity places greater value on thedescriptors, “liking, and being attractive to, the opposite sex”, and on having a successfulsocial life. Appeals to being “healthy”, “cool”, “grown-up”, and being “independent of thecrowd” are therefore likely to meet with the lack of success that has been already observedin current programmes. Future intervention programmes aimed at smokers might seek toenhance the desirability of the social representations associated with the nonsmokeridentity, or to reduce the perceived worth of those associated with the smoker identity,although neither of these is a simple task.

The third group, the occasional smokers, provides the greatest scope for intervention.The analyses of smoking prevalence demonstrate that the behaviour of occasional smokersis not as powerfully determined by fixed variables such as age, puberty or gender, as that ofnonsmokers and regular smokers. Occasional smokers may, therefore, be more susceptible toinfluence if programmes can be devised that enhance the attractiveness of the nonsmokeridentity.

Conclusions

The combination of quantitative and behavioural procedures has identified several of themeanings of smoking for adolescent girls. A greater understanding of the different meaningsof smoking to adolescent girls, and their role in the construction of different socialidentities may enable interventions to focus more effectively than has hitherto been thecase. Never-smokers would probably respond well to most health-related interventions. Bycontrast, regular smokers pose the greatest challenge as they themselves view smoking as auseful means of developing an adult social identity, rather than as the deviant behaviourportrayed by current health education programmes. If ways could be found to make thesocial representations of nonsmokers more highly valued than those of regular smokers,occasional smokers are likely to be amenable to interventions.

We suggest that intervention programmes be adapted to take account of the existence ofthree smoker identities that value smoking behaviour in different ways. There is no easyformula to developing more effective intervention studies, but if their creators are groundedin a thorough understanding of how young people themselves view smoking, repetition ofsome of the dead-ends of earlier programmes may be avoided.

Acknowledgments

The authors are indebted to David Hitchin, Statistical Consultant and Senior Programmerat the University of Sussex for his advice and assistance, and to Kirsti Mitchell and MarnieFreeman for data collection and processing.

The data reported in this paper were collected as part of a study supported by theDepartment of Health. An earlier version of this paper was presented at the BritishPsychological Society Annual Conference, University of Warwick, April 1995.


References

Barton, J., Chassin, L., Presson, C. C. and Sherman, S. J. (1982). Social image factors as motivators ofsmoking initiation in early and middle adolescence. Child Development, 53, 1499–1511.

Bewley, B. R. and Bland, M. (1978). The child’s image of a young smoker. Health Education Journal,37, 236–241.

Charlton, A., Gillies, P. and Ledwith, F. (1985). Variations between schools and regions in smokingprevalence among British schoolchildren—Implications for health education. Public Health, 99,243–249.

Cleveland, W. S. (1979). Robust locally weighted regression and smoothing scatterplots. Journal of theAmerican Statistical Society, 79, 807–822.

Diamond, A. and Goddard, E. (1995). Smoking among secondary school children in 1994. London:OPCS, HMSO.

Duveen, G. and Lloyd, B. (1986). The significance of social identities. British Journal of SocialPsychology, 25, 219–230.

Duveen, G. and Lloyd, B. (1990). Introduction. In Social Representations and the Development ofKnowledge, Duveen, G. and Lloyd, B. (Eds). Cambridge: Cambridge University Press.

Erickson, E. (1946). Ego development and historical change. Psychoanalytic Study of the Child, 2,359–396.

Erickson, E. (1950). Childhood and Society. New York: Norton.Evans, R. I. (1976). Smoking in children: developing a social psychological strategy of deterrence.

Preventive Medicine, 5, 122-127.Evans, R. I., Rozelle, R. M., Mittelmark, M. B., Hansen, W. B., Bane, A. L. and Havis, J. (1978).

Deterring the onset of smoking in children: knowledge of immediate physiological effects andcoping with peer pressure, media pressure and parent modelling. Journal of Applied SocialPsychology, 8, 126–135.

Falkner, F. and Tanner, J. M. (1978). Human Growth: Postnatal Growth. New York: Plenum.Kannas, L. (1985). The image of the smoking and non-smoking young person. Health Education

Journal, 44, 26–30.Michell, L. (1994). Smoking Prevention Programmes for Adolescents: A Literature Review. Oxford:

Anglia and Oxford Regional Health Authority.Moscovici, S. (1973). Forward. In Health and Illness, Herzlich, C. (Ed.). London: Academic Press.Moscovici, S. (1976). La Psychanalyse, son image et son public. Paris: Presses Universitaires de France.Moscovici, S. (1981). On Social Representations. In Social Cognition, Forgas, J. (Ed.). London:

Academic Press.Moscovici, S. (1984). The phenomenon of Social Representations. In Social Representations, Farr, R.

and Moscovici, S. (Eds). Cambridge: Cambridge University Press.Moscovici, S. (1988). Notes Towards a Description of Social Representations. European Journal of

Social Psychology, 18, 211–250.Nutbeam, D., Macaskill, P., Smith, C., Simpson, J. M. and Catford, J. (1993). Evaluation of two

school education programmes under normal classroom conditions. British Medical Journal, 306,102–107.

Tajfel, H. (1981). Human Groups and Social Categories: Studies in Social Categories. Cambridge:Cambridge University Press.

Tajfel, H. (Ed.) (1982). Social Identity and Intergroup Relations. Cambridge: Cambridge UniversityPress.

Documents

Adolescent girls' constructions of smoking identities: implications for health promotion