Social Science & Medicine 62 (2006) 2513–2525 Could the peer group explain school differences in pupil smoking rates? An exploratory study Katrina Turner a, , Patrick West b , Jacki Gordon c , Robert Young b , Helen Sweeting b a Community-based Medicine, University of Bristol, Cotham House, Cotham Hill, Bristol, England BS6 6JL, UK b MRC Social and Public Health Sciences Unit, UK c Choose Life, UK Available online 20 December 2005 Abstract Schools differ in the proportion of their pupils who smoke. Such differences transcend pupil intake characteristics and relate to the internal life of the school. Although adolescents’ smoking behaviour has been associated with that of their peers, little consideration has been given to whether peer structures and processes contribute to school differences in pupil smoking rates. In two relatively deprived Scottish schools, one with a higher and one with a lower rate of pupil smoking, 13 and 15 year-olds were surveyed. Sociometric data and information on pupils’ smoking behaviour and views were gathered. Twenty-five single-sex discussion groups were then held with a sub-sample of the 13 year-olds in order to explore in detail their views on smoking, smokers and fellow pupils. Findings showed that in the higher smoking school, pupils were more often in groups, smokers were identified as popular, and attitudes (especially among non-smoking females) were more pro- smoking. In the lower smoking school, by contrast, there were more isolates and dyads, there were no popular smokers and attitudes (especially among non-smoking females) were much less pro-smoking. Thus, evidence suggests peer group structures and related influences could be one explanation for school differences in smoking, and that the popularity of smokers together with the views of non-smoking females may be particularly important in creating such differences. r 2005 Elsevier Ltd. All rights reserved. Keywords: School differences; Pupil smoking; Peer groups; Sociometry; Scotland Introduction Schools differ in relation to their pupils’ health- related behaviours. These differences are indepen- dent of school intake and relate to the internal life of the school. Aveyard et al. (2004), for example, in their study of pupil smoking in secondary schools in England, found that after controlling for pupil characteristics, schools characterised as ‘authorita- tive’ had significantly lower smoking rates than ‘laissez-faire’ schools. School-level differences in smoking were also found in another study of secondary school pupils in Wales, those with lower rates being more likely to have a strong written policy on smoking and to enforce it (Moore, Roberts, & Tudor-Smith, 2001), a conclusion echoed by Pinilla, Gonzalez, Barber, and Santana (2002) in their study of secondary school pupils in ARTICLE IN PRESS www.elsevier.com/locate/socscimed 0277-9536/$ - see front matter r 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.socscimed.2005.11.017 Corresponding author. Tel.: +44 117 331 0946. E-mail addresses: [email protected] (K. Turner), [email protected] (P. West), [email protected] (J. Gordon), [email protected] (R. Young), [email protected] (H. Sweeting).
Could the peer group explain school differences in pupil smokingrates? An exploratory study
Katrina Turnera,�, Patrick Westb, Jacki Gordonc, Robert Youngb, Helen Sweetingb
aCommunity-based Medicine, University of Bristol, Cotham House, Cotham Hill, Bristol, England BS6 6JL, UKbMRC Social and Public Health Sciences Unit, UK
cChoose Life, UK
Available online 20 December 2005
Abstract
Schools differ in the proportion of their pupils who smoke. Such differences transcend pupil intake characteristics and
relate to the internal life of the school. Although adolescents’ smoking behaviour has been associated with that of their
peers, little consideration has been given to whether peer structures and processes contribute to school differences in pupil
smoking rates. In two relatively deprived Scottish schools, one with a higher and one with a lower rate of pupil smoking, 13
and 15 year-olds were surveyed. Sociometric data and information on pupils’ smoking behaviour and views were gathered.
Twenty-five single-sex discussion groups were then held with a sub-sample of the 13 year-olds in order to explore in detail
their views on smoking, smokers and fellow pupils. Findings showed that in the higher smoking school, pupils were more
often in groups, smokers were identified as popular, and attitudes (especially among non-smoking females) were more pro-
smoking. In the lower smoking school, by contrast, there were more isolates and dyads, there were no popular smokers and
attitudes (especially among non-smoking females) were much less pro-smoking. Thus, evidence suggests peer group
structures and related influences could be one explanation for school differences in smoking, and that the popularity of
smokers together with the views of non-smoking females may be particularly important in creating such differences.
r 2005 Elsevier Ltd. All rights reserved.
Keywords: School differences; Pupil smoking; Peer groups; Sociometry; Scotland
Introduction
Schools differ in relation to their pupils’ health-related behaviours. These differences are indepen-dent of school intake and relate to the internal lifeof the school. Aveyard et al. (2004), for example, in
e front matter r 2005 Elsevier Ltd. All rights reserved
their study of pupil smoking in secondary schools inEngland, found that after controlling for pupilcharacteristics, schools characterised as ‘authorita-tive’ had significantly lower smoking rates than‘laissez-faire’ schools. School-level differences insmoking were also found in another study ofsecondary school pupils in Wales, those withlower rates being more likely to have a strongwritten policy on smoking and to enforce it (Moore,Roberts, & Tudor-Smith, 2001), a conclusionechoed by Pinilla, Gonzalez, Barber, and Santana(2002) in their study of secondary school pupils in
ARTICLE IN PRESSK. Turner et al. / Social Science & Medicine 62 (2006) 2513–25252514
Spain. In addition, and of direct relevance to thestudy reported here, the longitudinal ‘West ofScotland 11–16’ study found large differences insmoking, drinking and experience of illicit drugs ina cohort of some 2500 pupils in 43 secondaryschools, both at ages 13 (1996) and 15 (1999) (West,Sweeting, & Leyland, 2004). These differences werenot explained by socioeconomic or other pupilcharacteristics but were related to school size andpupils’ views about school life.
These studies highlight a number of school-basedpolicies and practices that may impact on pupils’health behaviours. Within the literature, referencehas also been made to the possible role of peers andbest friends as an explanation for school differencesin smoking and other health behaviours (e.g. Westet al., 2004). To our knowledge, this has receivedremarkably little attention. This is surprising giventhat schools are quintessentially social places (Ryan,2000) and it has been argued that the peer groupconstitutes the most pervasive influence on adoles-cents’ attitudes and behaviours (e.g. Harris, 1998).The power of the peer group therefore should not beignored as one candidate explanation for ‘schooleffects’.
Consensus on the influence of peers on theadoption and maintenance of health-related beha-viours is reflected in the substantial body ofliterature on social relationships and friendships(e.g. Conrad, Flay, & Hill, 1992; Lloyd & Lucas,1998; West & Michell, 1999). Most of this workconsiders social relationships, focussing on indivi-duals or dyads, particularly ‘best friends’. Suchfriendships, however, only give a partial picture ofpeer relationships in childhood and adolescence andfail to account for the wider relations (e.g. cliques orwider crowds within which friendships exist) thatalso characterise young people’s social milieu. Theserelations merit attention since peer influence ismultidimensional and can operate via subtle pro-cesses such as modelling and normative pressures(Brown, 1990).
Research into the role of friends and peer groupson smoking has been considerably enhanced bystudies using sociometric methods, a quantitativetechnique that involves mapping the inter-relation-ships between individuals in a social system such asa school (e.g. Benenson, 1990; Michell & Amos,1997). Sociometry not only enables the identifica-tion of peer groups but also common (intra-group)attributes such as gender (Michell, 1997), statuswithin the peer group (Benenson, 1990), and health
behaviours, e.g. smoking (Pearson &Michell, 2000).In addition, sociometric techniques within a long-itudinal research design have been used to identifyprocesses producing peer group homogeneity (e.g.Ennett & Bauman, 1994), principally by reference toselection (i.e. individuals who have similar char-acteristics select each other as friends) and/orinfluence (i.e. individuals who join a group areinfluenced by other group members and throughthis process become similar).
In considering the possible role of peer networksas one explanation for school effects on healthbehaviours like smoking, attention is directed topupil culture(s) and in particular attitudes towardsthe behaviour in question. While it is well knownthat adolescent smokers hold more positive viewsabout smoking than their non-smoking peers (Lloyd& Lucas, 1998), much less is known about varia-tions in pupils’ attitudes between schools. Variationin attitudes towards smoking may also occurbetween pupils occupying different social networkpositions in those schools, e.g. among those ingroups or among those least integrated within a peergroup. Given the gendered nature of peer groups(Heaven, 1994), it is also possible that the attitudesof males and females in similar sociometric posi-tions may vary. Thus, in addition to the identifica-tion of social networks, and the position of smokersin those networks, it becomes important to under-stand how pupils in schools, which differ in theirsmoking profiles, view smoking (e.g. ‘mature’,‘cool’) or use this behaviour (e.g. as a badge ofbeing in the right crowd). Such data may inform usabout the nature of the relationship betweensmoking and friendship groups, and to the reasonsfor variation in the prevalence of smoking in theseschools.
While quantitative methods can gather attitudinaldata to characterise school and peer group differ-ences, qualitative techniques collect data that directattention to underlying processes of selection andinfluence. The identification of a group of popularand attractive ‘top girl’ smokers in Michell’s (1997)qualitative study is one such example. Anothermight be the identification of individuals or groupsof individuals perceived as wielding influencethrough coercion (e.g. bullying) rather than popu-larity. Such qualitative data can be used to assesshow peers may influence one another, and toconsider in detail which particular individuals areregarded as popular and/or powerful, and as suchpotential influences on the smoking behaviour of
ARTICLE IN PRESSK. Turner et al. / Social Science & Medicine 62 (2006) 2513–2525 2515
pupils in a specific network. In cross-sectionalstudies, these data compensate to some extent forthe lack of a longitudinal design.
This paper uses both quantitative and qualitativedata in an exploratory study to examine whetherpeer structures and influences might contributeto school differences in pupil smoking rates.We do this by comparing the sociometricstructure and attitudes to smoking of pupils intwo secondary schools, together with their accountsof peer processes, to address the following specificquestions:
1.
Do the schools differ in the peer group structureof their pupils and does this vary by gender andbetween smokers and non-smokers?
2.
Do pupils in the schools hold different attitudestowards smoking and does this vary by genderand smoking status?
3.
Do pupils in different peer group structures holddifferent attitudes towards smoking and if somight they account for differences between theschools’ smoking rates?
4.
Are there different pupil views about mechanismsof influence to smoke in the two schools?
Methods
Background and school selection
Following on from the identification of schooleffects on smoking in the West of Scotland 11–16study (West et al., 2004), nine of the 43 schoolsinvolved were selected as candidate schools for the‘Teenage Health in Schools Study’ (THiS) (Gordon& Turner, 2003; Turner, Gordon, & Young, 2004).These schools were known to differ in their pupilsmoking rates at ages 13 (in 1996) and 15 (1999). In2000, surveys were repeated in each of these nineschools as part of THiS, focussing on the same agegroups (13 and 15) as before, the aim being toidentify a pair of schools with higher and lowersmoking rates but similar pupil intake in terms ofdeprivation. Two schools with significantly differentpupil smoking profiles at both ages 13 and 15 wereselected for further investigation. These schools arereferred to as Highacres and Lowlands. Highacreshad a relatively high rate of current smoking andLowlands a relatively low rate (Table 1). Theprincipal investigators (KT and JG) were blind tothe schools’ smoking status. According to areadeprivation scores (Carstairs & Morris, 1991), both
schools served disadvantaged populations. Bothwere large urban schools, Highacres with a schoolroll of 1100 and Lowlands with a roll of 1400.Although neither school had a written policy onsmoking, interviews with senior management in-dicated that both endorsed an informal policybanning pupil smoking in school (Gordon &Turner, 2003).
Survey administration and quantitative measures
The surveys were conducted on school premises,in school time and under exam-type conditions. Allpupils aged 13 and 15 were invited to take part.Parents were notified about THiS and both theyand their child(ren) were given the opportunity toopt out. Only eight parents refused their child’sparticipation and no pupils refused to take part. Thefinal numbers of participants, together with re-sponse rates, are shown in Table 1.
The questionnaire principally focussed on pupils’own smoking behaviour, their views about smokingand smokers, experiences of and attitudes to school,and identification of friends and their character-istics. The measures of relevance here are:
Own smoking status: Using a standard (ONS)question (Goddard & Higgins, 1999), pupils wereasked to indicate if they had never smoked (not evena puff), tried smoking once, used to smoke but hadgiven up, smoked occasionally (sometimes) orregularly (one or more cigarettes a week). Followingprevious practice (West et al., 2004), here we refer tocurrent smokers as those who smoked occasionallyor regularly.
Perceived level of smoking in school: Pupilsestimated the proportion of pupils smoking (a) intheir school (b) in their year group, the responsecategories being ‘most’, ‘about half’, ‘a few’ and‘none’. Here, the last two categories are combined,the data being used descriptively to supplement thefindings on pupils’ own smoking behaviour.
Attitudes to smoking: Pupils were asked toindicate how much they agreed (strongly agree,agree, disagree, strongly disagree) with 11 itemsrelating to smoking: viz [smoking] calms yournerves, makes you unfit, is fashionable, helps youmake friends, makes you look more mature, onlydamages your health if you’re old, makes you smellhorrible, gives you confidence, keeps you slim, and‘others make fun of you if you don’t smoke’ and ‘Iwould rather kiss a non-smoker than a smoker’. Thepoles of these items were aligned so that a higher
ARTICLE IN PRESS
Table 1
Response rates (number of participants), gender composition and rates of pupil smoking and proportion of pupils smoking in school and
school year in Highacres and Lowlands
Highacres Lowlands P
Response rate Age 13 91.8 (213) 92.0 (276) 0.98
Age 15 87.1 (189) 79.9 (218) o0.05
Gender Males 47.3 48.2 0.79
Females 52.7 51.8
Current smokers Age 13 19.0 11.7 0.02
Age 15 26.5 10.6 o0.001
Males 16.4 7.2 0.003
Females 28.0 15.0 0.001
Estimated proportion of pupils in school smoking Most 39.2 18.2 o0.001
Half 49.4 49.8
Few/none 11.5 40.2
Estimated proportion of pupils in year group smoking Most 30.6 13.2 o0.001
Half 46.7 38.6
Few/none 22.7 48.2
K. Turner et al. / Social Science & Medicine 62 (2006) 2513–25252516
(mean) score on each indicates a more favourableattitude to smoking. Following factor analysisof these items (details on request), one dominantfactor was extracted, the highest loading items(maturity, helps make friends, confidence, fashion-able, calms nerves, keeps slim) suggesting itcaptured a general attitude towards smoking andsmokers. The resulting scale, distinguishes betweenthose who are pro- (higher scores) and anti-smoking(lower scores).
Acceptability of smoking: Pupils were also askedabout the extent to which they approved of (a) girls,and (b) boys ‘of [their] age smoking’ (with responsecategories ‘totally OK’, ‘fairly OK’, ‘don’t like itmuch’, ‘hate it’). As with the above items, higher(mean) scores indicate a more favourableattitude (greater acceptability) to smoking andsmokers.
Social network position: The sociometric sectionof the questionnaire asked pupils to list up to sixfriends and gathered information subsequently usedto assess the strength of a friendship. The strengthof a relationship was increased if the individualnamed was a best friend, went to the same primaryschool, was seen both in and out of school, and wasa ‘good laugh’. Although much sociometric researchis based only on the identification of friends, theinclusion of data relating to type (‘best’ friend),duration (primary school) and location (in/out ofschool) of friendship, together with joint activities
(having a ‘good laugh’) increases the probabilitythat pupils’ nominations refer to stronger relation-ships with the potential for mutual influence.
The sociometric data were analysed via NEGO-PY 4.50 (Richards, 1995). Only reciprocated nomi-nations of a certain strength were noted asfriendships, i.e. X and Y were only classed asfriends if X said Y was a friend, and Y said X was afriend, and the strength of the relationship was 1 ormore. Based on their linkages with others, NEGO-PY classified pupils as:
�
‘isolates’ (individuals who either did not nomi-nate a friend within the peer group under studyor whose nomination was not reciprocated) or asbelonging to � ‘dyad’ (isolated friendship pair) � ‘peripheral’ and ‘tree’ (connected to someone
within a group but not part of the group)
� ‘group’ (consisting of three or more individuals.
Each group member had more than 50% of his/her linkage with other group members, and a linkto at least two other members).
NEGOPY was also used to calculate a ‘popular-ity’ score: pupils who had received six or morenominations (i.e. top 20%) were defined as ‘popu-lar’. The quantitative data were analysed usingSPSS, the w2 test and F-ratio being used asappropriate to establish the significance of
ARTICLE IN PRESSK. Turner et al. / Social Science & Medicine 62 (2006) 2513–2525 2517
differences found between Highacres’ and Low-lands’ pupils. Because the study is essentiallyexploratory, and numbers in some sub-groups arevery small, multivariate analysis is not used.
The qualitative study
Following completion of the surveys, single-sexdiscussion groups were held with a sub-sample ofthe 13 year-olds. Only 13 year-old pupils wereinterviewed as analysis of West of Scotland 11–16data had shown school effects for smoking werestronger at age 13 than 15 (West et al., 2004), andbecause the schools were reluctant for older pupilsto be involved due to academic pressures.
Predicting pupils would talk more freely if groupswere friendship-based (Lees, 1986), we recruitedparticipants by asking pupils to volunteer alongwith friends from their class. The groups weresingle-sex to encourage participants to discussfriendships and smoking, and to explore gender-specific issues (Michell & Amos, 1997). Groups werenot divided by smoking status. Twenty-five discus-sion groups were conducted (13 in Highacres (sevenmale, six female) and 12 in Lowlands (six male, sixfemale). The groups ranged in size from three toeight individuals. One hundred and thirty-sixindividuals took part in total, and 19 participantsdescribed themselves as smokers. Most of thediscussions lasted for over an hour. With partici-pant consent, each was audio taped and transcribedverbatim.
As THiS aimed to identify both school and peer-related explanations for school differences in pupilsmoking rates, in each school half the groupsfocussed on the former (e.g. the school’s ethos andHealth Education curriculum), the other half onpupils’ views of smoking, peer structures andfriendships. Participants taking part in these lattergroups were invited to discuss their views onsmoking and smoking peers, and asked ‘why dopupils/you smoke?’, ‘who are the popular pupils?’,and ‘what sort of boys/girls are popular andadmired by pupils in this school?’. Although suchquestions were not posed during the ‘school-related’discussions, participants in these groups commentedon smoking and fellow pupils. For the purposes ofthis paper, analysis focusses on all data relating tosmoking, smokers, popular pupils, named indivi-duals, and reasons for smoking, whether thesearose through direct questioning or spontaneouslythroughout the discussions.
Analysis of the discussion group data involvedone of us (KT) thoroughly reading each interviewtranscript and developing a coding frame thatreflected areas covered during the interviews andemergent themes. Transcripts were imported intoNVivo version 1.2 (NVivo, 2000) and fully codedusing this software package. Data coded underspecific themes (e.g. views on smoking, popularpupils) were then selected and analysed in detail.This latter stage involved KT writing descriptivesummaries detailing what each group had saidunder each theme, and noting the nature of anydivergent views and group consensus. Within andcross-school comparisons were made using thesesummaries.
In this paper, in order to clarify whether we arereferring to individuals who were surveyed or thosewho were interviewed, the former will be referred toas pupils, the latter as participants. To maintainconfidentiality, all names have been replaced withpseudonyms.
Findings
Quantitative
School differences in smoking
Table 1 details response rates to the survey,gender composition of the sample together withpupil smoking rates and estimates of the proportionof pupils smoking.
Response rates were generally good, and particu-larly so at age 13 when over 90% of all target pupilsin both schools took part. However, among 15 year-olds, the response rate in Lowlands (80%) wassignificantly lower than in Highacres (87%). Nodifferences were observed in the gender compositionof the two schools, either overall or in the two agegroups (not shown). As indicated earlier, signifi-cantly higher rates of current smoking were found inHighacres than in Lowlands, both at ages 13 (19%vs. 12%) and 15 (27% vs. 11%), a differenceextending to both sexes.
Since absentees are more likely to smoke (Conradet al., 1992), it is probable that the lower responserate among 15 year-olds in Lowlands underesti-mated the true smoking rate in that year group. It iscertainly implausible that a smaller proportion ofolder than younger pupils smoked. While this is animportant issue, for two reasons we consider it veryunlikely to be more than a partial explanationfor the differences in smoking rates between the
ARTICLE IN PRESSK. Turner et al. / Social Science & Medicine 62 (2006) 2513–25252518
schools. Firstly, at age 13 response rates weresimilar in each school but smoking rates weresignificantly higher in Highacres than in Lowlands.Secondly, pupils in the two schools had verydifferent perceptions of levels of pupil smoking intheir schools, those in Highacres estimating sig-nificantly higher proportions of smokers both in theschool as a whole and within year groups, a patternapplying to 13 and 15 year-olds (not shown). Thelower response rate in Lowlands might, however,mean that the 15 year-olds surveyed in this schooldid not fully represent the peer group structure andculture of that year group in the school. Again,particularly because an 80% response rate is stillvery high, we think it unlikely this has more than aminor effect on the results.
Differences in peer group structure
Data on the sociometric status of pupils inHighacres and Lowlands are shown in Table 2,both in the total sample and by gender. Also shown
Table 2
Sociometric status and popularity of pupils (male and female) in High
Socio-status Males Females
Highacres Lowlands Highacres
Group 43.4 31.9��� 69.3
Periphs 35.4 30.2 16.5
Dyads 2.1 9.4 4.7
Isolates 19.0 28.5 9.4
Popular 16.4 16.0 24.0
��po0:01,���po0:001 significance of differences in overall sociometric status
Table 3
Sociometric status and popularity of pupils (male and female) in High
Socio status Males Females
Highacres Lowlands Highacres
Smk Nsmk Smk Nsmk Smk Nsm
Group 36.7 44.4 11.8 33.6 59.3 73.7
Periphs 36.7 36.6 29.4 30.4 13.6 17.8
Dyads 3.3 2.0 11.8 9.2 10.2 2.6
Isolates 23.3 17.0 47.1 26.7 16.9 5.9
Popular 10.7 16.1 0.0 17.3� 32.7 21.1
�po0:05,��po0:01,���po0:001 significance of difference between smokers and non-sm
is the percentage of pupils in each of these categoriesregarded as popular.
Overall, the sociometric status of pupils in thetwo schools is significantly different, Highacrescontaining a higher proportion of pupils in groups(57%) than Lowlands (38%), Lowlands containingmore pupils in dyads (10% vs. 3%) and isolates(24% vs. 14%). This pattern is similar at both ages(not shown) and in both genders, females in eachschool being more involved in groups (particularlyHighacres), males more likely to be isolates. Giventhe close correspondence between group member-ship and popularity, it is not surprising that theproportion of popular pupils was higher in High-acres (20%) than Lowlands (13%). This differencecan be entirely attributed to females (24% vs. 9%),among males popularity is not differentiated byschool.
Table 3 shows the sociometric patternsfurther broken down by the smoking status ofpupils. The general tendency for pupils in Highacres
acres and Lowlands (column %)
Total
Lowlands Highacres Lowlands
43.9��� 57.1 38.2���
25.5 25.4 27.8
11.0 3.5 10.2
19.6 14.0 23.9
9.0��� 20.4 12.5��
and popularity between schools.
acres and Lowlands according to smoking status (column %)
Total
Lowlands Highacres Lowlands
k Smk Nsmk Smk Mk Smk Nsmk
� 21.1 47.9�� 51.7 59.0� 18.2 40.7���
21.1 26.5 21.3 27.2 23.6 28.5
23.7 8.8 7.9 2.3 20.0 9.0
34.2 16.7 19.1 11.5 38.2 21.8
0.0 10.9� 25.0 18.6 0.0 14.1��
okers within schools.
ARTICLE IN PRESSK. Turner et al. / Social Science & Medicine 62 (2006) 2513–2525 2519
to be more often in groups and for those inLowlands to be in dyads and isolates persists amongsmokers and non-smokers of both genders, thoughthis is not significant among male smokers inHighacres. Within-school comparisons (right handcolumns) also reveal that in both schools smokerswere less likely to be in groups and more likely to bein dyads or isolates than non-smokers. However,this difference between smokers and non-smokers ismuch more marked in Lowlands (e.g. 18% ofsmokers, 41% of non-smokers in groups) thanHighacres (52% vs. 59%). Furthermore, the differ-ence in patterns between the schools is strikinglyillustrated by the results for popularity. In High-acres, there is no significant difference in popularitybetween smokers and non-smokers. In Lowlands,by contrast, popular pupils were entirely confined tonon-smokers; there were no popular smokers.
Further inspection of gender patterns in Table 3(left hand columns) also reveals that the sociometricdifferences between smokers and non-smokers inthe two schools were largely confined to females.The differences for males are not statisticallysignificant in either school. However, this is muchless evident in Lowlands than Highacres and mayexplain why the difference in popularity betweensmokers and non-smokers extends to both sexes inLowlands. In Highacres, male and female smokerswere just as likely to be popular as male and femalenon-smokers; indeed there is a suggestion thatfemale smokers in Highacres were particularly
Table 4
Pupils’ attitudes towards smoking (mean pro-smoking scores) and acce
item scores) in Highacres and Lowlands according to gender and smok
Smokers Non-smoke
Highacres Lowlands Highacres
Attitude to smoking (pro-smoking score)
Males 0.68 1.10 �0.11
Females 0.84 0.63 0.03
Acceptability of girls smoking
Males 2.62 2.71 1.72
Females 3.04 2.95 1.85
Acceptability of boys smoking
Males 3.11 3.00 1.88
Females 3.09 2.84 1.88
�po0:05,��po0:01,���po0:001 significance of difference between schools.
popular (33%). These findings indicate that thetwo schools were characterised by very differentpeer group structures both with respect to pupils ingeneral and with respect to differences betweensmokers and non-smokers, especially those offemale gender. They also suggest (particularlythe findings on popularity) that the peer cultureof the two schools with respect to smokingdiffered, Highacres being more pro-smoking thanLowlands.
Differences in attitudes to smoking
When assessing attitudes to smoking, the 2-yeargroups were combined as the patterns were similarin each. Mean scores for each of the 11 attitudeitems and the pro-smoking (factor) scale in the twoschools were explored in relation to gender andsmoking status.
Among males there were very few differences inattitudes between Highacres and Lowlands. Therewere no significant differences in the pro-smoking(factor) score (Table 4), though on two items(smoking helps you make friends and smokingkeeps you slim), Highacres’ males were more pro-smoking (po0:05 in both cases). By contrast,females in Highacres were much more pro-smokingthan those in Lowlands. They were significantlymore likely to endorse the view that smoking isfashionable (po0:05), calms nerves (po0:01), be-stows confidence (po0:01), keeps you slim(po0:001) and that kissing a smoker is OK
ptability of girls and boys smoking among same age peers (mean
ing status
rs All
Lowlands Highacres Lowlands
�0.20 0.03 �0.11
�0.24�� 0.25 �0.12���
1.58 1.86 1.65�
1.55�� 2.17 1.75���
1.79 2.07 1.87�
1.55��� 2.21 1.74���
ARTICLE IN PRESSK. Turner et al. / Social Science & Medicine 62 (2006) 2513–25252520
(po0:01), differences reflected in the significantlyhigher pro-smoking score (Table 4). Broken downfurther by smoking status, in addition to predictable(and large) differences in attitudes between smokersand non-smokers irrespective of school, it wasapparent the difference between pupils in Highacresand Lowlands was much greater among non-smoking females. With a few exceptions on indivi-dual items, there were no significant differencesbetween the schools among smokers of either sex, oramong male non-smokers. Female non-smokers inHighacres were however significantly more pro-smoking overall (po0:01), particularly endorsingthe view that smoking calms nerves (po0:05) andkeeps you slim (po0:001).
Table 4 shows mean scores on the two itemsrelating to acceptability of smoking among sameage male and female peers. As with attitudes tosmoking, the views of smokers and non-smokerssharply diverged, the former being significantlymore accepting (higher scores) of smoking amongboys and girls of their age. This applies to males andfemales in both schools. With respect to schooldifferences, it is apparent that among all pupils(right hand columns) those in Highacres weresignificantly more likely to accept boys and girlssmoking, but that this was much more markedamong females than males. Broken down further bysmoking status, it is equally apparent that thedifference in views is almost entirely confined tofemale non-smokers, those in Highacres beingsignificantly more likely to think it was acceptablefor girls and boys to smoke than their counterpartsin Lowlands. This result exactly parallels that inrelation to attitudes and identifies non-smokingfemales as the sub-group that most distinguishes thetwo schools.
Table 5
Pro-smoking scale by sociometric status among pupils in Highacres an
Males Females
Highacres Lowlands Highacres
Group �0.10 �0.37� 0.15��
Periphs 0.14 �0.09 0.26
Dyads �0.23 �0.01 1.30
Isolates 0.06 0.13 0.73
�po0:05,��po0:01,���po0:001 significance of difference between sociometric categorie
Differences in attitudes by sociometric status
In the final quantitative analysis, we examinewhether the different attitudes to smoking expressedby pupils in Highacres and Lowlands are associatedwith particular sociometric positions. The results forthe pro-smoking scale, both overall and by genderare shown in Table 5.
In the sample as a whole, it is evident therelationship between sociometric position and thepro-smoking index differs between the schools. InHighacres, there is no significant difference in meanscores between pupils in any sociometric category.In Lowlands, by contrast, pupils in groups andperipherals held significantly more negative viewstowards smoking than those in dyads or isolates.However, while this difference in patterns betweenschools extends to males, among females variationin attitudes by sociometric position is rather similarin each school, those in dyads and isolates beingmore pro-smoking than those in groups or periph-erals. Excepting isolates, though, Highacres’ femalesremain more pro-smoking than their Lowlands’counterparts.
Table 6 shows the patterns within schools brokendown further by smoking status. Again, the maindifference occurs among non-smoking females inLowlands. In this sub-group, the views of pupils indifferent sociometric categories are sharply differ-entiated, group members in particular holding themost negative views on smoking of all pupils. Thereis no corresponding significant difference amongfemale non-smokers in Highacres, those in groupsbeing most sharply distinguished from groupmembers in Lowlands. Among smokers, there ismuch less difference in attitudes both within andbetween schools, though the views of male smokersin Lowlands vary significantly by sociometric status,
d Lowlands by gender
Total
Lowlands Highacres Lowlands
�0.42��� 0.07 �0.40���
�0.13 0.18 �0.11
0.16 0.69 0.09
0.39 0.30 0.24
s within schools.
ARTICLE IN PRESS
Table 6
Attitude to smoking (pro-smoking scale) by sociometric status among pupils in Highacres and Lowlands by gender and smoking status
�po0:05,��po0:01,���po0:001 significance of difference between sociometric categories within schools.
K. Turner et al. / Social Science & Medicine 62 (2006) 2513–2525 2521
isolates having the most pro-smoking views of allsub-groups.
Overall, the results show that the main differencein attitudes towards smoking occurs among non-smoking females, among whom Lowlands’ groupmembers held particularly anti-smoking views, astriking contrast both with their counterparts inHighacres and with isolates and dyads in their ownschool.
The discussion group data
The participants’ views of smoking, smokers andfellow pupils provide another viewpoint from whichto consider whether or not peer structures andinfluences contributed to Highacres’ and Lowlands’differing smoking rates. They also illuminatepossible explanations for some of the quantitativefindings.
Participants’ views on smoking
Both Highacres’ and Lowlands’ participantsdescribed smoking as ‘disgusting’, ‘stinking’, a wasteof money, and a habit which made you smell, unfitand less attractive. Some participants linked smok-ing with cancer and heart disease, and a fewmentioned the risks associated with passive smok-ing. Interestingly, Highacres’ smokers commentedon the increased risk of cancer, the expense ofsmoking, and the fact that they, themselves, smelledof smoke. Some potential benefits were acknowl-edged, however.
In both schools male and female participantscommented that smoking could make you lookmore mature and, being a sociable habit, couldenable you to make new friends. Some of the girlsalso commented that smoking could have a calming
effect and keep you slim. Additionally, complement-ing the quantitative findings, all those who smokeddescribed smoking as both enjoyable and sociable.
The smokers
In each school, most of the non-smokers de-scribed those who smoked as ‘troublemakers’, and/or as individuals who annoyed them, had no money,smelt of cigarettes, were not interested in theireducation, and tried to act ‘hard’, ‘big’, mature and‘cool’. In Lowlands there was also a sense of thesmokers being individuals who bullied others since afew participants described them as ‘thugs’ and ‘gangmembers’ (in the areas surrounding Lowlands therewere various youth gangs who fought one another).Yet, despite such negative images, when asked if thesmokers were popular, in both schools there was amixed response, some saying no, others yes.Additionally, in Highacres one group of malesargued that all the ‘nice’ (i.e. attractive) girlssmoked and described how this encouraged othersto do so:
P:
Everybody smokes, everybody that’s nice. P: Aye all the nice ones smoke. P: That’s why all the wee ones smoke. P: Everybody says ‘Oh look at her she’s nice, she
smokes’ and that’s what it is like.
P: And see like all the nice ones, they all like tend
to hang around with each other and cause likeone of them smokes, they all smoke.
P:
And that why the wee folk that urny (are not)popular want to smoke because they see thepopular folk smoking and go—‘I want to bepopular to, and they just start smokingy
P:
All the nice lassies smoke. P: That’s how you get to hang aboot with the
popular ones. (Highacres, group 2, boys)
ARTICLE IN PRESSK. Turner et al. / Social Science & Medicine 62 (2006) 2513–25252522
The theme of ‘popular smokers’ was particularly
evident during the Highacres discussions as, corre-sponding to the sociometric data, smokers wereregularly described as popular. Furthermore, duringthe discussions concerning who was popular/influ-ential, most of the individuals identified weresmokers. One girl (Katie) was mentioned regularly.Although Katie was three years older than theparticipants, they knew who her boyfriend was andthat she had two sisters. Both male and femaleparticipants described Katie as pretty and popular,and it was commented that her smoking behaviourencouraged even those who did not personally knowher.
One of Katie’s sisters (Alice) was in the partici-pants’ year. Alice was also described as popular.Her sister’s popularity was stated as one reasonfor this, but Alice was also liked because she toowas pretty. Alice and her friends were describedas ‘the popular group’ in the participants’ year,and as being known not only to pupils in their yearbut also by those in younger and older years,and as smoking ‘every day’ at smokers’ corner (acorner by the playground on school premises).Interestingly, some of Alice’s friends took part inone of the discussions and described how pupilssmoked in order to socialise with them. Theirpresence within the school is consistent with thesociometric finding that Highacres’ female smokerswere particularly popular and could explain whyfemales in Highacres were more pro-smoking thanthose in Lowlands.
In Lowlands, whilst pupils were named anddiscussed, no particular group or individual wasmentioned regularly. Additionally, when pupilswere referred to, often they were identified becausethey were bullies and not because they wereregarded as attractive and/or as people pupilswanted to know. Some argued that Lowlands’smokers influenced others because they were fearedrather than liked. Furthermore, one group referredto popular girls as the ‘non-smokers’, and anothercommented that smokers were only popular withthose who smoked.
Reasons for smoking
Highacres’ and Lowlands’ smokers stated theysmoked because they were addicted, enjoyed smok-ing and smoking with friends. A few of the maleparticipants in Highacres also mentioned theysmoked in order to get to know older pupils andbecause they felt it would not harm them.
Reasons mentioned by non-smokers and smokersfor their peers’ smoking were being bored, addictedto nicotine, and wanting to look hard, calm theirnerves, be like their smoking friends, and impressothers. In each school it was also stated that theyounger pupils smoked in order to mix with theolder ones, and that pupils smoked to be withpopular individuals and to make new friends. Itseemed smoking enabled pupils to make friendsbecause it provided a basis for interacting withothers at school:
KT:
How does it (smoking) help you make friends? P: See if you smoke, see if people see you
smoking, they’ll say ‘Can I smoke with youbecause I’ve no got any fags the now’.
P:
People that smoke, people that don’t evenknow them can just go up andy
P:
You can say ‘Give us a fag’ or ‘Give us a light’or something like that.
P:
‘Have you got a fag on you’ or somethinglike that. That’s like how they make pals‘Aye, two minutes I’ll go get one, andjust start smoking and that’s how they makepals.
P:
Playtime and lunchtime and all that.(Lowlands, group 2, boys)
In both schools participants commented thatpupils smoked because staff did little to curb pupilsmoking and because pupils were bored and incontact with smokers. In Lowlands it was alsocommented that pupils smoked because they werescared of the smokers and because they wanted tobe in a gang, comments that echoed the descriptionsof smokers as ‘thugs’ and gang members.
Although smokers mentioned they did notencourage others to smoke, non-smokers in bothschools argued that smokers offered others cigar-ettes and put pressure on them to try smoking.Furthermore, participants’ accounts of why theyhad tried/started smoking indicated that smokingfriends/peers had directly or indirectly encouragedthem to experiment. Yet, it was also apparent thatpupils had tried smoking because they wanted to,and that experimenting with peers had providedpupils with the opportunity to discover they did notlike smoking and therefore did not wish to smoke.In addition, there was evidence of pupils resisting oravoiding peer pressure, and accounts of pupilsencouraging friends/peers to stop smoking.
ARTICLE IN PRESSK. Turner et al. / Social Science & Medicine 62 (2006) 2513–2525 2523
Discussion
Cumulative evidence (both quantitative andqualitative), from the comparison of two schoolswith similar pupil intake but different levels ofsmoking, strongly suggests that variations betweenschools regarding peer group structures, associatedpupil cultures and influences could contribute toschool differences in pupil smoking rates.
Results showed the schools differed in the socio-metric structure of their pupils. Highacres containedmore pupils in groups than Lowlands, whichcorrespondingly had more isolates and dyads. Thisdifference persisted among smokers and non-smokers and suggests smokers in Highacres weremore able than those in Lowlands to directlyinfluence the smoking behaviour and views of fellowpupils. With this in mind, it is noteworthy that onlyin Highacres was there a strong sense of smokingbeing a sociable habit, with pupils buying andselling cigarettes from one another and sharingcigarettes with friends (Turner et al., 2004). Yet, inboth schools, and consistent with some previousstudies (e.g. Ennett & Bauman, 1993), smoking wasmore common among dyads and isolates. Thisdemonstrates the difference between the schoolscannot be attributed in any simple way to differentsociometric composition.
Further analysis, however, revealed quite differ-ent relationships between smoking and sociometricposition within the two schools that also varied bygender. Among males there was little variation insociometric structure between smokers and non-smokers. Among females, although isolates andpupils in dyads were more likely to smoke thanthose in groups in both schools, the difference inHighacres was much less marked than in Lowlands.Most notably, female smokers in Highacres weremuch more likely to be in groups, and to bepopular, while those in Lowlands were much lesslikely to be in groups and none were popular. Thispattern of results highlights the importance oftaking gender together with smoking status intoaccount in any analysis of the role of peer groups inexplaining school effects. In statistical terms, itsuggests that any (multivariate) analysis is morelikely to be characterised by complex interactioneffects than simple main effects.
Second, while the differences in peer groupstructures between the two schools suggest potentialsources of influence, that associated with popularsmoking females being the most obvious example,
they do not inform us about associated pupilculture(s), specifically in relation to smoking.Results of the analysis of pupil attitudes are againinformative in suggesting that the major differenceslie in particular sub-groups within schools. Thus, incontrast to smokers of both genders who heldsimilar pro-smoking views in both schools, andmale non-smokers whose views were also rathersimilar between the schools, those of female non-smokers were sharply differentiated. Female non-smokers in Highacres were much more positiveabout smoking than their counterparts in Lowlands,a difference also seen in the greater acceptability ofsmoking among peers of both genders in this sub-group only. Further analysis also revealed thatamong female non-smokers the biggest differencebetween the schools occurred among those ingroups, pupils in Lowlands having the mostnegative views towards smoking of all sub-groups.In this school, the views of group members andthose in dyads and isolates differed markedly whilein Highacres non-smoking females were not sodifferentiated by sociometric position. Thus, thequantitative analysis points up the importance ofparticular sub-groups within schools, in this casenon-smoking females, whose views may be particu-larly important in shaping pupils’ attitudes tosmoking and, conceivably too, the likelihood thatpupils will smoke.
Third, the possible mechanisms by which indivi-duals are affected by particular pupil cultures tosmoke or not to smoke are illustrated by thequalitative data collected. In many respects, thequalitative findings complement those of the quan-titative analysis of attitudes. Participants in bothschools articulated both negative (e.g. smokingsmells) and positive (e.g. keeps you slim) images ofsmoking, yet there were also clear differencesbetween the schools both in terms of images andmechanisms of influence. In particular, amongHighacres’ participants, smoking was linked tospecific individuals, notably females, who wereperceived as very attractive and highly popular.No such individuals were identified in Lowlands.Furthermore, although in both schools participantsviewed smokers as influencing others to smoke, themechanisms by which this was perceived to occurdiffered. In Highacres, the predominant view wasthat pupils smoked to be like the popular indivi-duals or groups in the school; in Lowlands, wheresmokers were not popular, much greater emphasiswas placed on the behaviour of smokers to make
ARTICLE IN PRESSK. Turner et al. / Social Science & Medicine 62 (2006) 2513–25252524
non-smokers smoke, including specific reference tobullying.
This suggests that two quite different mechanismswere operating in these schools, one (Highacres)emphasising processes of selection (notably model-ling), the other (Lowlands) emphasising processes ofinfluence (notably coercion). The former bearsconsiderable comparison with the school featuredin Michell’s (1997) study where smoking wasassociated with ‘top girls’. Such differences implythat the balance of processes of selection andinfluence are likely to vary between schools, peerselection for smoking conceivably being more likelyin schools with a pro-smoking pupil culture andhigher smoking rate, peer influence (at least of thecoercive variety) being more likely in schools withthe opposite characteristics. In this regard, it shouldbe remembered that selection and influence are notjust processes by which pupils become smokers butalso processes by which pupils remain non-smokers.In Lowlands, where smokers and non-smokers werestrongly distinguished in sociometric position andpopularity, the views of non-smoking females inparticular, may have been especially important incontributing to that school’s lower smoking rate. InHighacres, by contrast, where smokers and non-smokers were not so distinguished, the views of non-smoking females were much less negative and mayhave been outweighed by the views of popularsmoking females, who in turn may have beenparticularly important in contributing to thatschool’s higher smoking rate.
Any suggestion that peer group structures andprocesses might be a candidate explanation forschool effects must acknowledge several limitationsin the study design. First, the extent to whichfindings can be generalised is limited by the fact thatonly two schools were involved and both serveddeprived areas in the west of Scotland. Second, thedata are cross-sectional, thereby limiting any con-clusions about the direction of causality (selection/influence) between sociometric position and smok-ing. Third, as only reciprocated relationships weredefined as friendships, some of those classed asisolates may have been friends with individuals notsurveyed. This is an acknowledged problem with thesociometric method within a school (Wasserman &Faust, 1999) and has important implications when,as in this study, the proportion of isolates variesbetween schools. Fourth, as most participantsdescribed themselves as non-smokers, the discussiongroup data might not have reflected the views of
smokers as fully as those of non-smokers.Furthermore, smokers may have been reluctantto voice their views, as non-smokers were themajority in most of the groups. This situation mayhave been exacerbated by the fact that the groupcontext tends to encourage public rather thanprivate accounts, and can lead to censoring andconformity (Sim, 1998). Finally, the samplingapproach used to recruit participants may haveled to an under-representation of isolates, andtherefore smokers.
Notwithstanding these limitations, this study hasprovided important insights into how the peergroup might be one candidate explanation forschool effects on smoking and other health beha-viours. It suggests this is unlikely to be a simplequestion of differential sociometric composition(and attendant risks for smoking) and more likelyto involve complex interactions between school,sociometric position and smoking in which parti-cular sub-groups are identified as important. If theresults of this comparison are generalisable, it alsosuggests that gender is a major factor to take intoaccount, the sociometric position of non-smokingfemales being of particular interest.
Acknowledgements
We would like to thank Margaret Reilly for herhelp regarding the sociometric coding. Some of thesociometry questions were based on those used byMike Pearson and Lynn Michell. The MedicalResearch Council of Great Britain provided finan-cial support for Katrina Turner and Jacki Gordonduring the course of the study, and continues to doso for Patrick West, Robert Young and HelenSweeting.
References
Aveyard, P., Markham, W. A., Lancashire, E., Bullock, A.,
MacArthur, C., & Cheng, K. K. (2004). The influence of
school culture on smoking among pupils. Social Science &
Medicine, 58(9), 1767–1780.
Benenson, J. F. (1990). Gender differences in social networks.
Journal of Early Adolescence, 10(4), 472–495.
Brown, B. B. (1990). Peer groups and peer cultures. In S. S.
Feldman, & G. R. Elliott (Eds.), At the threshold: The
