Journal of Behavioral Medicine, Vol. 29, No. 5, October 2006 ( C© 2006)DOI: 10.1007/s10865-006-9065-4

Adolescent Smoking Continuation: Reduction andProgression in Smoking after Experimentation andRecent Onset

Rinka M. P. Van Zundert,1,3 Rutger C. M. E. Engels,1

and Regina J. J. M. Van Den Eijnden2

Accepted for publication: June 15, 2006Published online: July 20, 2006

In the present study, the role of cognitive concepts derived from the Theory of PlannedBehavior in adolescent smoking reduction, continuation, and progression was investigated.These concepts include pro-smoking attitudes, perceived social norms regarding smoking,and self-efficacy to resist smoking. Logistic regression analyses were performed on data from397 Dutch adolescents aged 11–15 years, who had at least once tried smoking. Attitudes, per-ceived social norms, and self-efficacy, including significant interactions between these threeconcepts, explained up to 41% of variance in smoking behavior cross-sectionally. Longitudi-nally, an interaction between pro-smoking attitudes and low self-efficacy increased the chanceof reduction in smoking, and all three cognitions inclusive of two interactions between pro-smoking perceived soc ial norms and low self-efficacy or positive attitudes towards smokingpredicted progression of smoking. Cognitions may play relatively small roles in adolescentsmoking reduction, but do seem to be relevant in progression in smoking after experimenta-tion or recent onset. Interactions between positive attitudes towards smoking and prosmok-ing perceived social norms provide cumulative risks for adolescents to increase their levels ofsmoking, whereas interactions between less favorable attitudes and high self-efficacy to resistsmoking may provide a protective effect for adolescents to reduce or to quit their smoking.

KEY WORDS: adolescent smoking continuation; theory of planned behavior; longitudinal study.

INTRODUCTION

Despite the fact that the highly hazardous con-sequences of smoking have become common knowl-edge, representative figures demonstrate that thereis still a large proportion of adolescent smokers.Prevalence rates in the Netherlands show that 23.6%of 10–19 year old students smoke at least once amonth, and 46.3% have ever experimented with

1Behavioural Science Institute, Radboud University Nijmegen,Nijmegen, the Netherlands.

2IVO, Addiction Research Institute, Rotterdam, the Netherlands.3To whom correspondence should be addressed at Be-havioural Science Institute, Radboud University Nijmegen,P.O. Box 9104, 6500 HE, Nijmegen, the Netherlands; e-mail:r.vanzundert@pwo.ru.nl.

smoking (Dutch Foundation for National Healthand Smoking (STIVORO), 2004). Also, most ex-perimentation of smoking occurs among Dutch ado-lescents between 12 and 14 years (Dutch Founda-tion for National Health and Smoking (DEFACTO),2002). These prevalence rates are in accordance withthe general figures in most European countries,where country-specific prevalences of smoking dur-ing the last 30 days range between 22% and 56%, andwhere 50% to 80% of 15–16 year old students haveever tried smoking (Hibell et al., 2003).

At present, ample studies have been conductedto examine which factors constitute the motivationof young people to initiate smoking. One of themost frequently studied theoretical frameworks ofsmoking initiation involves the Theory of PlannedBehavior (TPB) (e.g. De Vries et al., 1995; Godin

435

0160-7715/06/1000-0435/0 C© 2006 Springer Science+Business Media, Inc.

436 Van Zundert, Engels, and Van Den Eijnden

et al., 1992; Hanson, 1997; Harakeh et al., 2004).The TPB aims to predict motivational influences ondeliberate behavior, such as experimentation and ini-tation of smoking, through consideration of attitudes,normative beliefs, and self-efficacy (Ajzen, 1991).Despite the fact that the motivational processes ofthe TPB have been proved to be related to initiationof regular smoking, little is known about the motiva-tional processes that are involved in continuation ofsmoking after experimentation with or recent initia-tion of smoking. It is crucial, however, to determinewhy some adolescents do not continue to smoke afterexperimentation or recent initiation whereas otheradolescents do. Of those who do continue to smoke,it is useful to know which factors can predict eitherreduction or progression in their levels of smoking.

The few studies that consider smoking continu-ation have mainly focused on adult populations andon demographics, such as education (Droomers et al.,2002), on tobacco availability (Pokorny et al., 2003),on nicotine dependency and amount of cigarettessmoked at baseline (Nordstrom et al., 2000; Pierceet al., 1996), and on parental smoking (Bauman et al.,2001; Fergusson et al., 1995). Also, smoking behav-ior of friends has been found to predict progres-sion from experimental smoking to regular smoking(Pierce et al., 1996). Cognitive predictors of smok-ing continuation, such as attitudes, self-efficacy andperceived social norms, however, have hardly beentaken into account so far, neither among adolescents,nor among adults. However, since the concepts of theTheory of Planned Behavior have been abundantlyand successfully linked to other stages of smoking be-fore, such as intention to smoke (De Vries et al., 1995;Godin et al., 1992; Kremers et al., 2004), and initiationof smoking (Harakeh et al., 2004), the TPB may pro-vide an appropriate framework within which to studycontinuation of smoking after experimentation or re-cent onset .

Attitudes, Perceived Social Norms, and Self-Efficacy

Applied to smoking behavior, pro-smoking atti-tudes reflect a positive and favorable view on smok-ing (Harakeh et al., 2004). When adolescents valuesmoking as advantageous, it is plausible that theywill see no reasons to discontinue smoking. In fact,the experience of smoking may influence the atti-tudes towards smoking. When adolescents have pos-itive experiences with smoking, it is possible thatthey will come to view smoking as more positive,

which in turn may make them more susceptible toincrease their levels of smoking. Furthermore, moststudies have focused on peer smoking as a measureof peer influence, yet passive peer pressure throughperceived normative beliefs appears to have been ne-glected (Perrine and Aloise-Young, 2004). However,it is understood that adolescents are strongly suscep-tible to their friends’ attitudes and values (Berndt,1996; Cohen, 1977). Therefore, it is likely that adoles-cents take their friends’ norms and approval of theirbehavior as a reference in making their decision to ei-ther quit, to continue, or to even increase their levelsof smoking after experimentation or recent initiation.The effect of self-efficacy may also be applicable tosmoking continuation in the sense that, when adoles-cents have experimented with or have recently takenup smoking, and expect that they will find it difficultto resist smoking in subsequent tempting situations,they are more likely to continue their recently ac-quired habit. Low self-efficacy may also make themmore vulnerable to progression of smoking, whereashigh self-efficacy to resist smoking may make adoles-cents more prone to smoking reduction. In conclu-sion, previous studies have indicated that attitudestowards smoking, perceived social norms, and self-efficacy concerning smoking resistance may explainand predict smoking initation and are likely to play arole in smoking continuation as well.

The Present Study

Pro-smoking attitudes, perceived social normswhich reflect friends’ approval of smoking, and lowself-efficacy to resist smoking in tempting situationswere hypothesized to be positively related to smok-ing behavior cross-sectionally. From a longitudinalperspective, pro-smoking attitudes, perceived socialnorms in favor of smoking, and low self-efficacy toresist smoking were expected to predict smokingcontinuation, and progression. Negative attitudes to-wards smoking, perceived social norms disapprovingof smoking, and high self-efficacy to resist smokingwere expected to be positively related to smokingreduction. Furthermore, Ajzen and Madden (1986)recommend that interactions between the cognitiveconcepts be included in testing the theory. They pro-pose that direct effects of attitudes, norms, and self-efficacy on behavior need not be additive in nature.Each of these predictors may be necessary, but notsufficient conditions for the formation of intentionsto perform behavior. Regarding behavioral control,

Adolescent Smoking Continuation 437

for example, one must also be inclined to perform thebehavior for other motivations than merely believingthat one could perform the behavior. This line of rea-soning implies the possibility that perceived behav-ioral control affects behavior in interaction with atti-tudes and self-efficacy. Also, testing interactions mayilluminate which combinations of predictors placeyoung smokers in a high risk group of adolescentswho are most likely to continue to smoke or evento progress their levels of smoking. Despite Ajzenand Madden’s valuable recommendation, no studyhas hitherto implemented this strategy with regardto adolescent smoking behavior. Accordingly, thepresent study is unique in including interactions be-tween smoking attitudes, perceived social norms andself-efficacy to test if particular combinations of theseconcepts have cumulative effects, and will thus posean intensified threat for starting smokers to continueor to increase their levels of smoking. These interac-tions were tested in a three-wave longitudinal studyamong 397 adolescents who had at least once triedsmoking.

METHOD

Sample and Procedure

During the 1999–2000 school year, 1969 first-grade students4 from ten secondary schools acrossthe Netherlands were recruited for a three-wavestudy which examined smoking behavior, drinkingbehavior, and delinquency. During November andDecember 2000, when the students were at thebeginning of the first grade, the first wave (T1) hadbeen conducted, the second wave (T2) took placesix months later in May-June 2001, and the third andfinal wave (T3) in November-December 2001, whenthe students were at the beginning of the secondgrade. The parents or guardians had been informedabout the aims of the study in advance and had beengiven the opportunity to respond if they had anyobjections to their child’s participation. A few par-ents contacted the research institute for additionalinformation, yet none of the parents disapprovedof their child taking part in the study. The students

4Please note that the term ‘grade’ has dissimilar meanings acrosscountries. In the Netherlands, students usually leave elementaryschool when they are aged 12, varying between 11 and 14, andthereupon continue their educational careers as first graders insecondary school, of which the first and second year will be re-ferred to as ‘first and second grade.’

themselves agreed to participate as well, and inciden-tal missings were solely due to truancy and sickness.Questionnaires had been administered during schoolhours in the presence of an instructed teacher. Allstudents were assured of strict confidentiality, thatany information given would not be revealed to anyother person than the primary researchers. Studentswere notified that participation included them in alottery through which CD-vouchers could be won,to motivate them to complete the questionnairesconscientiously at all three measurements.

Of the eligible 1969 students, 1595 (81%) par-ticipated in all three measurements. Sample attri-tion over the several measurements was mainly dueto students who changed school, dropped out, orwho repeated the same curriculum for a second year.Also, two schools had not been able to administerthe questionnaires during one of the three measure-ments. Previous attrition analyses on the present datahave demonstrated that there were no substantial dif-ferences between the students who dropped out ofthe sample and those who were included (De Kempet al., 2004). Of these 1595 respondents, only those re-spondents were included in the analyses who had re-ported that they had at least smoked once, thus con-stituting the final sample (N = 397).

At baseline, the average age of the respondentswas 12.4 years (SD = .52; range: 11–15 years). At thesecond measurement most students were 13 years old(72.0%), and at the third measurement, most ado-lescents were aged 13 (51.6%), and 14 (45.6%).Thesample comprised 215 (54.2%) boys and 182 girls.The vast majority of respondents (94.7%) were ofDutch descent. Types of education that were in-volved in the study were a) lower education (14.9%),b) middle education (35.5%), and c) pre-universityeducation (38.8%). The remaining 2.3% included nototherwise specified types of education. With regardto domiciliary situation, 85.0% lived with both theirparents, 12.2% lived in a single parent household,and 2.8% resided in other arrangements (e.g., withother family members, institutions, or foster homes).

Measures

Smoking Continuation

A frequently used self-report instrument tomeasure smoking status was employed (cf. Engelset al., 2004; Kremers et al., 2001; Harakeh et al.,2004). Smoking status was assessed through oneitem asking the respondents which statement applied

438 Van Zundert, Engels, and Van Den Eijnden

best to them. Percentages will be given for the first,second, and third measurement respectively. (1) “Ihave tried smoking once, but I no longer smoke”(68.6%; 63.4%; 53.1%), 2) “I try smoking once ina while” (15.1%; 13.1%; 14.5%), 3) “I smoke lessthan once a month” (2.5%; 2.5%; 7.1%), 4) “I donot smoke weekly, but at least once a month” (2.5%;4.5%; 3.0%), 5) “I do not smoke daily, but at leastonce a week” (6.8%; 7.6%; 6.1%), and 6) “I smokeat least once a day” (4.5%; 8.8%; 16.2%). Thus, thepreponderance of students at baseline concernedadolescents who had at least tried smoking oncebut who no longer smoked (68.6%), a percentagewhich had decreased to 63.4% at T2, and to 53.1%at T3. Moreover, considering its skewed distribution,the outcome variable in the cross-sectional analyses‘smoking status’ had been dichotomised at all threemeasurements. A score of ‘1’ thus included respon-dents who reported that they had tried smokingonce, but who had not continued smoking, whereasscore ‘2’ included the respondents who at that timereported to smoke, regardless of frequency. Toexplore possible effects of the cognitions on smokingcontinuation, a distinction had been made betweenreduction in smoking, and progression in smoking.To this end, the original smoking status variable with6 categories of smoking frequency had been recodedinto 3 categories: ‘1’ represented those who had triedsmoking but who no longer smoked, ‘2’ respresentedthe ‘occasional smokers’, who smoked once in awhile or at least once a month, and ‘3’ representedthe weekly and daily smokers. This was performedfor all three measurements. Reduction was estab-lished when a respondent showed that frequency ofsmoking had decreased, for instance, when he or shehad reported to smoke weekly or daily at the firstmeasurement, and reported to be smoking once in awhile or once a month at the second measurement.Progression was similarly defined, with the change insmoking frequency indicating an increase in smokingfrequency. The reduction and progression variableswere computed as dichotomous variables of which‘1’ represented the group with those respondentswho had maintained the same level of smokingfrequency, and of which ‘2’ represented the group ofeither reducing or progressing respondents. The re-spondents who reported to have tried smoking once,but who at baseline had already quit experimentingwith smoking were excluded from the referencegroup in the longitudinal ‘reduction’ analyses, sincereduction in smoking could not be accomplishedwithin this group. Thus, the reference group in the

reduction analyses contained only respondents whoat baseline had been occasional or regular smokers,and who had maintained this level of smoking at T2and/or at T3. Similarly, the regular smokers at base-line were excluded from the reference group in thelongitudinal ‘progression’ analyses, as progressionin smoking could not be reached within this groupas well. Thus, 6 longitudinal logistical regressionanalyses were performed: continuation of the samesmoking frequency level versus smoking reduction orsmoking progression between T1 and T2 (six monthinterval), between T1 and T3 (one year interval),and between T2 and T3 (six month interval).

Attitudes Towards Smoking

Attitudes towards smoking reflect to which ex-tent adolescents appraise or have a positive regardfor smoking. Attitudes towards daily smoking weremeasured on a bipolar scale of which the seven itemsrepresented negative and positive attitudes (Harakehet al., 2004). The negative words were: ‘unpleas-ant,’ ‘harmful,’ ‘useless,’ ‘boring,’ ‘hazardous,’ ‘un-healthy,’ and ‘bad’. The positive words on the bipolarscale were respectively: ‘pleasant,’ ‘innocuous,’ ‘use-ful,’ ‘exciting,’ ‘harmless,’ ‘healthy,’ and ‘good.’ Ona scale of 1 to 7, respondents could rate their pref-erence regarding each of the attitudes. High scoresimply strong pro-smoking attitudes. Internal consis-tencies for T1, T2, and T3 respectively were .83, .85,and .86, as assessed through Cronbach’s alpha.

Perceived Social Norms

To examine to which degree adolescents subjec-tively feel that their best friends and friends in gen-eral either approve or disapprove of the respondent’ssmoking, the respondents were asked to indicate towhich extent they thought that (a.) their best friend,and (b.) their friends would approve that he or shesmoked or would smoke (Harakeh et al., 2004). Re-sponse choices ranged from 1 ‘certainly not’ to 5 ‘cer-tainly yes.’ The mean of the two items was used inthe analyses. High scores imply strong approval. ThePearson correlation between the two items was .70.

Self-Efficacy to Resist Smoking

A self-efficacy instrument was employed to ex-plore the expected self-efficacy not to smoke in

Adolescent Smoking Continuation 439

smoking-specific and tempting situations (De Vrieset al., 1988; Engels et al., 1999). A few examples of the6 items are: “For me, not to smoke, while my friendsare smoking is . . .”, “For me, to think of a reasonto refuse a cigarette is . . .”, and “For me, becominga non-smoker is . . .”. The expected difficulty not tosmoke in the described situations ranged from 1 ‘veryhard’ to 6 ‘very easy’. High scores imply high self-efficacy to refrain from smoking. Internal consisten-cies for T1, T2, and T3 respectively were .83, .86, and.89, as assessed through Cronbach’s alpha.

RESULTS

To test the relatedness and predictive power ofattitudes, perceived social norms, and self-efficacyin relation to concurrent smoking, and to smok-ing reduction or progression, both Pearson andSpearman correlations and logistic regression anal-yses were conducted cross-sectionally as well aslongitudinally.

Descriptive Analyses

Table I lists the cross-sectional and longitu-dinal Pearson and Spearman correlations betweenthe cognitive factors and smoking for all threemeasurements. Both cross-sectionally and longitu-dinally, attitudes were moderately correlated withperceived social norms, with correlations rangingbetween .10, and, .32, and more strongly corre-lated with self-efficacy, with correlations between−.24, and −.51. Perceived social norms and self-efficacy were moderately correlated as well, withcorrelations between −.13, and, −.28. Furthermore,from a cross-sectional and longitudinal perspective,all cognitive determinants were significantly corre-lated with smoking at all three measurements, withcorrelations ranging between .14, and −.46. Cross-sectionally, the cognitive determinants and smokingseemed to be increasingly correlated in the courseof the year. Lastly, the longitudinal correlations be-tween the cognitive determinants at T1, and smok-ing at the subsequent measurements T2 and T3, gen-erally showed a decreasing trend. Thus, the longerthe interval between the waves, the weaker becamethe correlations between cognitive determinants andsmoking.

Cross-Sectional Analyses

Table II presents the cross-sectional findings forattitudes, perceived social norms and self-efficacyin relation to adolescent smoking. Age, gender, andethnicity were included in the analyses as controlvariables. The enter-method was used in the lo-gistic regression analyses, and both Table II andTable III represent the findings per step. Neitherone of the external variables age, gender, ethnicity,and educational attainment level appeared to beassociated with adolescent smoking. Furthermore,the odds’ ratios proved to be significant at all threemeasurements for attitudes, and for self-efficacy.This indicates that holding pro-smoking attitudesand anticipating low self-efficacy to resist smokingincrease the odds of young adolescents’ smoking.Perceived social norms only appeared to be sig-nificantly related to smoking at T2, which showsthat adolescents are more likely to smoke at T2when they perceive the perceived social norms tobe in favor of their smoking habit. To conclude,Nagelkerke’s explained variance from T1 to T3increased from 22% to 41%. Apparently, over aninterval of a year, the TPB derived variables areincreasingly powerful to explain largely more thanone third of the variance in smoking behavior. Theseresults thus support the idea that smoking attitudes,perceived social norms, and non-smoking self-efficacy are related to smoking behavior in a sampleof early adolescents who had reported life timesmoking.

Longitudinal Analyses

Table III shows the longitudinal associations be-tween the cognitive factors and reduction in smok-ing. The background variables did not seem tobe relevant, neither in reduction nor progressionin smoking, with exception of educational attain-ment at T1. A high educational attainment levelserved a protective function in that adolescentswho received high secondary education were themost likely to have reduced or to have quit smok-ing at T2, relative to the first measurement. Al-though cognitive factors were generally significantlyrelated to smoking cross-sectionally, smoking re-duction did not appear to be affected by the in-cluded cognitions. Concerning progression in smok-ing (Table IV), the cognitive concepts seemed toplay a more important role. When adolescents who

440 Van Zundert, Engels, and Van Den Eijnden

Tab

leI.

Pea

rson

and

Spea

rman

Cor

rela

tion

sB

etw

een

Mod

elV

aria

bles

Mea

sure

s1

23

45

67

89

1011

12

1A

ttit

udes

T1

—2

Att

itud

esT

2.3

4∗∗

—3

Att

itud

esT

3.2

5∗∗

.39∗

∗—

4So

cial

norm

sT

1.2

0∗∗

.17∗

∗.1

4∗∗

—5

Soci

alno

rms

T2

.10

.26∗

∗.1

6∗∗

.52∗

∗—

6So

cial

norm

sT

3.1

0.2

4∗∗

.32∗

∗.4

0∗∗

.42∗

∗—

7Se

lf-e

ffica

cyT

1−.

43∗∗

−.24

∗∗−.

24∗∗

−.28

∗∗−.

13∗∗

−.18

∗∗—

8Se

lf-e

ffica

cyT

2−.

26∗∗

−.42

∗∗−.

32∗∗

−.17

∗∗−.

19∗∗

−.16

∗∗.5

6∗∗

—9

Self

-effi

cacy

T3

−.27

∗∗−.

35∗∗

−.51

∗∗−.

14∗∗

−.14

∗∗−.

25∗∗

.42∗

∗.5

8∗∗

—10

Smok

ing

stat

usT

1.2

7∗∗

.28∗

∗.2

2∗∗

.12∗

.11∗

.15∗

∗−.

29∗∗

−.23

∗∗−.

23∗∗

—11

Smok

ing

stat

usT

2.2

9∗∗

.45∗

∗.3

2∗∗

.18∗

∗.2

5∗∗

.18∗

∗−.

28∗∗

−.38

∗∗−.

35∗∗

.40∗

∗—

12Sm

okin

gst

atus

T3

.21∗

∗.3

3∗∗

.49∗

∗.2

0∗∗

.20∗

∗.3

1∗∗

−.17

∗∗−.

26∗∗

−.44

∗∗.3

2∗∗

.46∗

∗—

Not

e.Sp

earm

anco

rrel

atio

nsw

ere

calc

ulat

edfo

rth

eco

rrel

atio

nsw

ith

smok

ing

stat

usva

riab

les.

∗ p<

0.05

;∗∗ p

<0.

01;∗

∗∗p

<0.

001.

Adolescent Smoking Continuation 441

Table II. Cross-Sectional Associations Between Attitudes, Perceived Social Norms and Self-Efficacy, andAdolescent Smoking Behavior

Smoking T1 N = 377 Smoking T2 N = 378 Smoking T3 N = 371

OR 95% CI OR 95%CI OR 95%CI

Step 1Age 1.18 .77–1.81 1.30 .87–1.96 .77 .77–1.76Gender .86 .55–1.33 1.07 .70–1.62 .65 .65–1.48Ethnicity 1.00 .94–1.07 .98 .92–1.04 .93 .86–1.00Education .85 .66–1.11 .84 .66–1.08 .87 .68–1.11

Step 2Attitudes (A) 1.37∗∗ 1.11–1.68 1.67∗∗∗ 1.35–2.08 1.86∗∗∗ 1.47–2.37Social norms (N) 1.25 .99–1.57 1.53∗∗∗ 1.22–1.92 1.21 .97–1.51Self-efficacy (S) .57∗∗∗ .44–.74 .56∗∗∗ .44–.72 .56∗∗∗ .43–.75

Step 3AxN 1.36∗ 1.04–1.79 1.20 .90–1.60 .64∗ .45–.91AxS 1.08 .84–1.39 .85 .64–1.13 1.06 .76–1.48NxS 1.16 .88–1.54 1.27 .97–1.68 .73 .51–1.03

Step 4AxNxS 1.23 .93–1.62 1.22 .89–1.68 2.07∗∗∗ 1.41–3.03Nagelkerke R2 .22 .35 .41

Note. The figures in this table represent the results per step. N = Number of adolescents.∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001.

had experimented with smoking, or who had re-cently taken up smoking, perceived the social normregarding smoking to be in favor of smoking atthe first measurement, they were significantly more

likely to have increased their levels of smoking sixmonths later (T2) than to have continued to experi-ment or to have maintained the same smoking level.Positive attitudes towards smoking, as well as low

Table III. Longitudinal Associations and Interactions Between Attitudes, Perceived Social Norms and Self-Efficacy, and Reduction in Smoking

Reduction in smokingT1– T2 (N = 100)

Reduction in smokingT1–T3 (N = 95)

Reduction in smokingT2–T3 (N = 95)

OR 95% CI OR 95%CI OR 95%CI

Step 1Age .61 .26–1.42 .50 .21–1.19 .37 .14–1.02Gender 1.44 .62–3.34 2.01 .84–4.79 1.16 .44–3.03Ethnicity 1.01 .88–1.14 1.08 .93–1.25 .99 .86–1.14Education 1.86∗ 1.15–3.00 1.53 .95–2.47 1.34 .78–2.30

Step 2Attitudes (A) .706 .44–1.11 1.18 .77–1.82 1.21 .80–1.83Social norms (N) .77 .51–1.16 .93 .60–1.44 .96 .60–1.54Self–efficacy (S) .84 .51–1.38 .69 .42–1.13 1.02 .64–1.63

Step 3AxN .69 .34–1.40 .86 .48–1.57 .86 .50–1.49AxS .39∗∗ .21–.74 .65 .38–1.11 1.65 .89–3.06NxS .93 .49–1.74 1.23 .74–2.07 1.37 .73–2.57

Step 4AxNxS .54 .27–1.04 .88 .50–1.54 .86 .46–1.60Nagelkerke R2 .36 .24 .17

Note. The figures in this table represent the results per step. N = Number of adolescents. The independent vari-ables are from the first measurement of the various intervals as indicated above, e.g., from the second measurementwhen reduction in smoking was analyzed for the T2–T3 interval.∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001.

442 Van Zundert, Engels, and Van Den Eijnden

Table IV. Longitudinal Associations and Interactions between Attitudes, Perceived Social Norms andSelf-Efficacy, and Progression in Smoking

Progression in SmokingT1–T2 (N = 305)

Progression in SmokingT1– T3 (N = 304)

Progression in SmokingT2–T3 (N = 293)

OR 95% CI OR 95%CI OR 95%CI

Step 1Age 1.12 .67–1.87 .86 .53–1.40 .71 .42–1.22Gender 1.42 .84–2.37 1.25 .78–2.00 1.13 .68–1.88Ethnicity 1.00 .92–1.08 .60 .23–1.56 .72 .34–1.51Education 1.07 .78–1.47 .94 .71–1.25 1.07 .79–1.46

Step 2Attitudes (A) 1.05 .82–1.35 1.27∗ 1.01–1.58 1.16 .91–1.48Social norms (N) 1.51∗∗ 1.15–2.00 1.23 .96–1.56 .94 .75–1.19Self–efficacy (S) .76 .57–1.02 .71∗ .54–.93 .84 .65–1.10

Step 3AxN 1.08 .80–1.46 1.36∗ 1.01–1.83 .87 .65–1.16AxS 1.21 .91–1.61 .98 .72–1.32 1.24 .90–1.71NxS 1.42∗ 1.01–2.00 1.07 .80–1.48 .97 .73–1.28

Step 4AxNxS 1.02 .72–1.46 .94 .79–1.46 1.00 .71–1.39Nagelkerke R2 .12 .15 .08

Note. The figures in this table represent the results per step. N = Number of adolescents. The independentvariables are from the first measurement of the various intervals as indicated above, e.g., from the secondmeasurement when progression in smoking was analyzed for the T2–T3 interval.∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001.

self-efficacy to resist smoking at the first measure-ment were predictive of progression in smoking oneyear later (T3).

Interaction Analyses

Three two-way and one three-way interactionterms between the cognitive factors had been com-puted as the products of the centered main effectsvariables (Aiken and West, 1991). Interactionsbetween attitudes and perceived social norms weresignificantly related to adolescent smoking cross-sectionally at T1, and T3. Thus, when adolescentexperimenters or starting smokers hold positiveattitudes towards smoking in conjunction with aperception of high perceived social approval of theirsmoking, they are at elevated risk to smoke (seeFig. 1a and Fig. 1b). The cross-sectional three-wayinteraction at T3 also demonstrated that adoles-cents who concurrently experienced pro-smokingattitudes, pro-smoking perceived social norms, andwho additionally experienced low self-efficacy toresist smoking, were twice as likely to smoke, thanadolescents who were not influenced by all threecognitions simultaneously (see Fig. 1c). Longitu-dinally, the interaction between negative attitudestowards smoking and high self-efficacy to resist

smoking at T1 significantly predicted reduction insmoking six months later (see Table III, and Fig. 2a).This implies that adolescents who consider smokingto be unfavorable and who concurrently feel capableof resisting smoking, are more likely to reduce theirsmoking levels. That is, compared to adolescents whoeither view smoking as unfavorable, but who havelow self-efficacy to refrain from smoking, or com-pared to adolescents who anticipate to be capable ofrefraining from smoking, but who concurrently retainor develop a positive regard for smoking. Also, an in-teraction between positive attitudes and pro-smokingperceived social norms at T1 was significantly relatedto progression in smoking at T3 (see Table IV, andFig. 2b). Apparently, adolescents who view smokingas advantageous, and who at the same time perceivethe perceived social norms to be in support of theirsmoking habit, are more likely to increase theirlevels of smoking after experimentation or recentinitiation. That is, compared to adolescents whoeither have pro-smoking attitudes but who do notexperience social approval, or adolescents who doexperience social approval, but who themselves donot endorse positive attitudes towards smoking. Alsoof influence on progression, yet after six months,was an interaction between pro-smoking perceivedsocial norms and low self-efficacy to resist smoking(see Fig. 2c).

Adolescent Smoking Continuation 443

Fig. 1. Profile plots of the cross-sectional relationship of the two-way interaction between attitudesand norms at T1 with smoking at T1 (a), the cross-sectional relationship of the two-way interactionbetween attitudes and norms at T3 with smoking at T3 (b), and the cross-sectional relationship ofthe three-way interaction between attitudes, norms, and self-efficacy at T3 with smoking at T3 (c).Note. ‘High’ indicates scores one whole standard deviation above the mean score, ‘low’ indicatesscores one whole standard deviation below the mean score.

DISCUSSION

The present study aimed to identify the influ-ence of cognitive concepts derived from the Theoryof Planned Behavior on adolescent smoking continu-ation in terms of reduction and progression in lev-els of smoking. As hypothesized, pro-smoking at-titudes, perceived social norms endorsing smoking,and low self-efficacy to resist smoking were relatedto smoking behavior cross-sectionally, which is in linewith previous studies (De Vries et al., 1995; Harakehet al., 2004). Moreover, interactions between atti-tudes and norms were found to be related to adoles-cent smoking at T1 and T3. Also, adolescents whoconcurrently held positive attitudes towards smok-ing, who scored low on self-efficacy to resist smok-

ing, and who perceived the social norms to be infavor of their smoking, were twice as likely to besmoking at T3. Cross-sectionally, the explained vari-ance appeared to increase from T1 to T3. Froma statistical viewpoint, this is plausible since thereis increasing variance on smoking status as adoles-cents are older. Moreover, adolescence is known tobe the period in which (meta)cognitions are devel-oped, and in which the ability to self-reflect is en-hanced (Finkenauer et al., 2002; O’Mahony, 1989).As cognitive skills are more developed, they aremore likely to be applied in adolescent decisionmaking.

Judging from the longitudinal results, the pre-dictive validity of the cognitive concepts appearedto be very limited in terms of reduction of smoking.

444 Van Zundert, Engels, and Van Den Eijnden

a.)

c.)

low

1,00

1,10

1,20

1,30

1,40

1,50

high

low

Self-efficacy T1

1,30

1,40

1,50

1,60

1,70

1,80

1,90

2,00Self-efficacy T1

high

low

low high

2T

noitc

ude

R g

niko

mS

Attitudes T1low high

1,20

1,30

1,40

1,50

low

high

3T

noisser

gor

P g

niko

mS

Norms T1

Attitudes T1

Norms T1

high

2T

noiss er

go r

P g

ni ko

mS

b)

Fig. 2. Profile plots of the longitudinal effect of the two-way interaction between attitudes and self-efficacy at T1 on smoking reduction at T2 (a), the longitudinal effect of the two-way interactionbetween attitudes and perceived social norms at T1 on smoking progression at T3 (b), and thelongitudinal effect of the two-way interaction between self-efficacy and perceived social norms atT1 on smoking progression at T2 (c). Note. ‘High’ indicates scores one whole standard deviationabove the mean score, ‘low’ indicates scores one whole standard deviation below the mean score.

However, in the present study, the application ofconcepts derived from the Theory of Planned Behav-ior has proven successful in predicting progressionin smoking. It is therefore interesting and importantto explore alternative explanations as to why thesecognitions did not predict reduction of smokingamong adolescents, whereas they do seem to play arole in progression in smoking.

Perhaps, reduction in smoking does not involvea typical cognitive process. The group of reducingadolescents mainly consisted of occasional smok-ers at baseline who had quit smoking after six ortwelve months. It is conceivable that when ado-lescents smoke only once in a while, it does notrequire an active rational decision to quit the occa-

sional smoking. It may also be the case that otherfactors are relevant in smoking reduction than insmoking progression. It is possible that the needto experiment with smoking is fulfilled after a fewmonths, and adolescents may no longer feel the needto continue their smoking. However, this contentiondoes not hold for all experimenters, as the reduc-ers’ reference group in the analyses had maintainedthe same level of smoking at baseline one year later.Considering the intervention opportunities in thisspecific smoking trajectory, it is important for futureresearch to examine which factors, other than cog-nitions, stimulate some adolescents to reduce or toquit smoking after experimentation or recent onset,whereas others maintain the same levels of smoking.

Adolescent Smoking Continuation 445

Alternatively, a combination of cognitions maybe more significant in explaining reduction in smok-ing than each of the cognitions individually. Thepresent results evidence of a considerable negativeimpact of the interaction between pro-smoking atti-tudes and low self-efficacy to resist smoking at base-line on reduction in smoking six months later (T2).

This means that adolescents who score high onpositive attitudes towards smoking and who concur-rently score low on self-efficacy to resist smoking arethe least likely to reduce their smoking. Interventionprograms aimed to make experimenters or recentsmokers quit or reduce their smoking should there-fore target the smoking-related attitudes and self-efficacy simultaneously rather than separately. Edu-cational attainment level should also be accountedfor, as the present data demonstrate that adoles-cents who received high secondary education werethe most likely to have reduced or quit smoking aftersix months. Students at a lower educational attain-ment level may require relatively more attention inencouraging reduction of smoking.

Regarding progression in smoking, the involvedsmoking-related cognitions seem to play a marginalyet significant role. Perceived social norms emergedas a significant predictor of smoking progressionamong students after six months of the first grade.Over the course of one year (T1-T3), both positiveattitudes towards smoking, and low self-efficacyto resist smoking displayed marginal effects onprogression on smoking. The fact that perceivedsocial norms at the beginning of the first grade (T1)appeared to be relevant for progression in smokingsix months later (T2), and that similar effects hadnot been found for attitudes and self-efficacy, mightindicate that when adolescents enter secondaryschool, they may adhere to the prevailing normsregarding smoking in the friendship group ratherthan that they contemplate why they should stop,continue, or increase their smoking. Adolescentsmay also be most vulnerable and susceptible to theprevailing perceived social norms when they entersecondary school, since this is a distinct period inwhich new friendships are manifested. Aloise-Youngand colleagues (1994), for example, have suggestedthat teenagers may regard smoking as a way to enterdesired friendship groups. When adolescents con-sider smoking as a way of making new friends, theymay be vulnerable to passive peer pressure (Perrineand Aloise-Young, 2004). This may especially applyto the first six months after entering a new social andeducational environment. We found perceived social

norms at T1 to significantly predict smoking continu-ation at T2, but not at T3. It is possible that in the firstcourse of the second grade, adolescents have alreadyestablished a more stable friendship group. The needto comply in order to make friends may be lessenedat that time, or friends’ perceived social norms mayhave become the adolescents’ own norms for smok-ing. Still, the influence of positive attitudes towardssmoking, and low self-efficacy to resist smoking maynot be disregarded as these cognitions at baselineresulted in progression in levels of smoking oneyear later (T3). As such, all the cognitions derivedfrom the Theory of Planned Behavior could providerisk for young adolescents to increase their levelsof smoking after experimentation or recent onset.This was also reflected by the negative impact theconcurrent combination of pro-smoking perceivedsocial norms with low self-efficacy at T1, and thecombination of positive attitudes with a pro-smokingperceived social norm at T1 appeared to have onsmoking progression at T2, and T3 respectively. Allin all, both in stimulating teenagers to reduce theirlevels of smoking after experimentation or recentonset, and in discouraging teenagers to increasetheir levels of smoking, prevention and interventionprogrammes could find useful targets in the smoking-related cognitions presented in the present study.

Turning to the limitations of the present study,the study may be considered limited in that self-reports have been used to measure smoking status.Although there has been debate in the past as towhether or not self-reports of substance use maybe viewed as reliable instruments, Barnea and col-leagues (1987) have reported that self-reports of sub-stance use in the adolescent population are stable,and that questionnaires provide highly reliable data.Respondents were also assured of strict confidential-ity of their reports, which should enhance reliabilityas well (Velicer et al., 1992). Moreover, the sampleis restricted in the age range, including respondentsbetween 12 and 15 years old, with a mean age of 12.4years at the first measurement. Generalizability withregard to older adolescents is therefore limited, espe-cially since development of cognitions may be contin-gent on maturity level (Finkenauer et al., 2002). How-ever, at T3, most of the adolescents were in the agesof 13 and 14. Besides, the cognitions may not be com-plex to such a degree that they are likely to be largelyunderdeveloped in this particular age range. More-over, continuation of smoking after experimentationis distinctly different from the initiation phase. Thus,when samples with young adolescents, such as the

446 Van Zundert, Engels, and Van Den Eijnden

present sample, would be solely used to explore initi-ation of smoking, and only samples with older adoles-cents would be used to explore continuation of smok-ing, we would fail to capture the very early phase inwhich teenagers, who have experimented with smok-ing, or who have only recently taken up smoking, de-cide to turn smoking into a habit or decide to refrainfrom smoking.

Despite these limitations, the present study hasseveral important assets. Firstly, to our knowledge,this study is the first to implement interactionsbetween attitudes, perceived social norms, and self-efficacy in examining smoking behavior, more specif-ically, in examining smoking reduction, continua-tion, and progression after experimentation or recentonset. Interactions between attitudes, perceived so-cial norms, and self-efficacy have not been imple-mented and tested so far, while interactions maygive more insight into combinations of factors andmay identify high risk profiles. The present findingsdemonstrate that these interactions certainly maynot be neglected. For instance, the present cross-sectional findings point to significant relatedness be-tween smoking-specific cognitions and adolescentsmoking. In the longitudinal analyses of reductionin smoking, however, attitudes towards smoking andself-efficacy at baseline showed no main effects onsmoking reduction at T2, yet they appeared to berelevant when they were combined as interactingvariables. It is thus possible that attitudes, and self-efficacy exert an influence on smoking reduction, butonly under particular conditions. In addition, the lon-gitudinal design of the study enables us to make in-ferences about the direction of the associations.

In sum, attitudes, perceived social norms, andself-efficacy are not associated with adolescent smok-ing reduction, yet do predict progression in smok-ing among adolescents who have been experiment-ing with or who have recently taken up smoking. Pre-vention programs which focus on a cognitive level onyouth who have only just experimented with smok-ing, may be effective in preventing progression insmoking. However, intervention on cognitions maystill be useful when adolescents have progressed intheir smoking.5 Since perceived social norms do seem

5The reviewers of this paper suggested to control for concurrentcognitions in the longitudinal analyses, to determine whetherprior cognitions predict smoking behavior above and beyond con-current cognitions. If the effects of the prior cognitions would notbe diminished if concurrent cognitions were controlled for, thiswould indicate that once adolescents have taken up smoking, it

to have harmful effects, the present findings also en-dorse previous recommendations that prevention ef-forts should teach teenagers to resist peer pressure(Flay et al., 1983; Leventhal and Cleary, 1980). It isemphasized, however, that there is a possibility thatadolescents come to a point where perceived socialnorms are internalized, regardless of further activepeer pressure. In such cases, it should be aimed tomake adolescents aware of their own cognitive dispo-sitions and to address these cognitions. To conclude,adolescents who conjunctly hold both pro-smokingattitudes and experience either perceived socialnorms in favor of their smoking, or low self-efficacyto resist smoking should be identified as risk groups.

ACKNOWLEDGMENTS

This research was supported by a grant from theDutch Asthma Foundation and a fellowship grant toRutger Engels from the Dutch Organization of Sci-entific Research.

REFERENCES

Aiken, L. S., and West, S. G. (1991). Multiple Regression: Testingand Interpreting Interactions, Sage Publications, Inc., Thou-sand Oaks, CA, US.

Ajzen, I. (1991). The theory of planned behavior. Organ. Behav.Hum. Decis. Process. 50: 179–211.

Ajzen, I., and Madden, T. J. (1986). Prediction of goal-directed be-havior: Attitudes, intentions, and perceived behavioral con-trol. J. Exp. Soc. Psychol. 22: 453–474.

Aloise-Young, P. A., Graham, J. W., and Hansen, W. B. (1994).Peer influence on smoking initiation during early adolescence:A comparison of group members and group outsiders. J.Appl. Psychol. 79: 281–287.

Barnea, Z., Rahav, G., and Teichman, M. (1987). The reliabilityand consistency of self-reports on substance use in a longitu-dinal study. Br. J. Addict. 82: 891–898.

Bauman, K. E., Carver, K., and Gleiter, K. (2001). Trends in par-ent and friend influence during adolescence. The case of ado-lescent cigarette smoking. Addict. Behav. 26: 349–361.

Berndt, T. J. (1996). Transitions in friendships and friends’ influ-ence. In Graber, J. A., Brooks-Gunn, J., and Petersen, A. C.(Eds.), Transitions Through Adolescence, Erlbaum, Mahwah,NJ, pp. 57–84.

Cohen, J. M. (1977). Sources of peer group homogeneity. Sociol.Educ. 50: 227–241.

De Kemp, R. A. T., Scholte, R. H. J., Overbeek, G., and Engels,R. C. M. E. (2004). Early adolescent delinquency: The role ofparents and best friends. Criminal Justice Behav. (in press).

is too late to intervene on cognition. We found that the effectsof prior cognitions diminished if the concurrent cognitions werecontrolled for in all longitudinal analyses. This suggests that inter-ventions on cognitions may still be effective once teenagers haveinitiated smoking, even if they have increased their smoking rates.

Adolescent Smoking Continuation 447

De Vries, H., Backbier, E., Kok, G., and Dijkstra, M. (1995).The impact of social influences in the context of attitude,self-efficacy, intention, and previous behavior as predictors ofsmoking onset. J. Appl. Soc. Psychol. 25: 237–257.

De Vries, H., Dijkstra, M., and Kuhlman, P. (1988). Self-efficacy:The third factor besides attitude and subjective norm as apredictor of behavioral intentions. Health Educ. Res. 3: 273–282.

Droomers, M., Schrijvers, C. T. M., and Mackenbach, J. P. (2002).Why do lower educated people continue smoking? Explana-tions from the longitudinal GLOBE Study. Health Psychol.21: 263–272.

Dutch Foundation for National Health and Smoking (2004).Roken, de harde feiten: Jeugd 2004. Smoking, the Hard facts:Youth 2004. STIVORO, Den Haag.

Dutch Foundation for National Health and Smoking (2004). DE-FACTO. Annual Report. Den Haag: Defacto, 2002.

Engels, R. C. M. E., Knibbe, R. A., and Drop, M. J. (1999) Pre-dictability of smoking in adolescence: Between optimism andpessimism. Addiction 94: 115–124.

Engels, R. C. M. E., Vitaro, F., Den Exter Blokland, E., De Kemp,R. A. T., and Scholte, R. H. J. (2004). Influence and se-lection processes in friendships and adolescent smoking be-havior: The role of parental smoking. J. Adolesc. 27: 531–544.

Fergusson, D. M., Lynskey, M. T., and Horwood, L. J. (1995). Therole of peer affiliations, social, family, and individual factorsin continuities in cigarette smoking between childhood andadolescence. Addiction 90: 647–659.

Finkenauer, C., Engels, R. C. M. E., Meeus, W., andOosterwegel, A. (2002). Self and identity in early ado-lescence. In Brinthaupt, T. M. and Lipka, R. P. (Eds.),Understanding the Self of the Early Adolescent, State Univer-sity of New York Press.

Flay, B. R., d’Avernas, F. R., Best, J. A., Kersell, M. W., andRyan, K. B. (1983). Cigarette smoking: Why young peo-ple do it and ways of preventing it. In McGrath, P., andFirestone, P. (Eds.), Pediatric and Adolescent BehavioralMedicine, Springer-Verlag, New York, pp. 132–183.

Godin, G., Valois, P., Lepage, L., and Desharnais, R. (1992). Pre-dictors of smoking behavior: An application of Ajzen’s theoryof planned behavior. Br. J. Addict. 87: 1335–1343.

Hanson, M. S. (1997). The theory of planned behavior applied tocigarette smoking in African-American, Puerto Rican, andNon-Hispanic white teenage females. Nurs. Res. 46: 155–162.

Harakeh, Z. M. A., Scholte, R. H. J., Vermulst, A. A., De Vries,H., and Engels, R. C. M. E. (2004). Parental factors and ado-lescents’ smoking behavior: An extension of the theory ofplanned behavior. Prev. Med. 39: 951–961.

Hibell, B., Andersson, B., Bjarnasson, T., Ahlstrom, S.,Balakireva, O., Kokkevi, A., and Morgan, M. (Eds.) (2004).ESPAD Report: Alcohol and Other Drug Use among Stu-dents in 35 European countries (2003). The Swedish Councilfor Information on Alcohol and Other Drugs (CAN), ThePompidou Group at the Council of Europe and the author:Stockholm.

Kremers, S. P. J., Mudde, A. N., and De Vries, H. (2001). Kick-ing the initiation: Do adolescent ex-smokers differ from othergroups within the initiation continuum? Prev. Med. 33: 392–401.

Leventhal, H., and Cleary, P. D. (1980). The smoking problem: Areview of research and theory in behavior risk modification.Psychol. Bull. 88: 370–405.

Nordstrom, B. L., Kinnunen, T., Utman, C. H., Krall, E. A.,Vokonas, P. S., and Garvey, A. J. (2000). Predictors of con-tinued smoking over 25 years of follow-up in the NormativeAging Study. Am. J. Public Health 2000 404–406.

O’Mahony, J. F. (1989). Development of thinking about things andpeople: Social and nonsocial cognition during adolescence. J.Genet. Psychol. 150: 217–224.

Perrine, N. E., and Aloise-Young, P. A. (2004). The role of self-monitoring in adolescents’ susceptibility to passive peer pres-sure. Pers. Individ. Differ. 37: 1701–1716.

Pierce, J. P., Choi, W. S., Gilpin, E. A., Farkas, A. J., andMerritt, R. K. (1996). Validation of susceptibility as a predic-tor of which adolescents take up smoking in the United States.Health Psychol. 15(5): 355–361.

Pokorny, S. B., Jason, L. A., and Schoeny, M. E. (2003). Therelation of retail tobacco availability to initiation and con-tinued smoking. J. Clin. Child Adolesc. Psychol. 32: 193–204.

Pomerleau, O. F., Collins, A., Shiffman, S., and Pomerleau, C.(1993). Why some people smoke and others do not: New per-spectives. J. Consult. Clin. Psychol. 61: 723–731.

Prochaska, J. O., and DiClemente, C. C. (1983). Stages andprocesses of self-change of smoking: Toward an integra-tive model of change. J. Consult. Clin. Psychol. 51: 390–395.

Velicer, W. F., Prochaska, J. O., Rossi, J. S., and Snow, M.G. (1992). Assessing outcome in smoking cessation studies.Psychol. Bull. 111: 23–41.