Critical discussion of social–cognitive factors in smoking initiation among adolescents

Abstract

Social–cognitive models have often been used in research on prevention in adolescent populations, even though the models were designed to describe adult behavior. The aim of the study reported here was to examine critically and constructively the five social–cognitive factors in the ‘attitude, social influence, self-efficacy’ (ASE) model. Methods. The examination draws on the results of a qualitative follow-up study of smoking initiation based on semi-structured interviews and observations of 12 adolescents in two Danish school classes, grades 7 and 8. The qualitative study was conducted in connection with and sampled from a large quantitative study and the results of both studies are discussed. In the analyses, we explored the ASE constructs according to how they are described in the ASE theory. Furthermore, we examined contradictions and aspects which are not explained in the model and if relevant discussed these aspects using other theoretical frameworks. Results. The results showed that aspects other than those in the ASE model are also important. Smoking initiation was often situational and unplanned and was sometimes used in negotiating social relationships and identity. Furthermore, the social–cognitive models are based on the assumption that adolescents talk about smoking norms and have a high degree of individual reflexivity, which is not always characteristic of adolescent behavior. Conclusion. Applying theoretical models in health research should be a continuous process of both applying the model and discussing the theoretical assumptions of the model when applied to a specific sample. The results of the qualitative study provide some support for use of the ASE model, but the results also suggest that further studies are needed to explore how social–cognitive models can be expanded to be more comprehensive behavioral models.

Social–cognitive models have been applied in the field of health promotion in order to understand and ultimately change unhealthy behavior, with the aim of preventing disease [1]. The advantage of this type of multivariate theoretical models is that they point to the social–cognitive factors which should be investigated and to the psychosocial and biological processes in which they are grounded. Still, quantitative studies rarely reflect on the theoretical assumptions of the applied model and the possible consequences of these assumptions. The premises are important because they directly influence the results the study will produce. Thus, applying theoretical models in health research should be a continuous process of both applying the model and discussing the theoretical assumptions of the model when applied to a specific sample. A social–cognitive model commonly used in research on prevention of adolescent smoking is the ASE model (Figure 1). The ASE model was designed by Hein de Vries and colleagues [2], who combined theoretical constructs from the ‘theory of planned behavior’ (attitude, subjective norms, perceived control (similar to self-efficacy) and ‘social cognitive learning theory’ (behavior of others and self-efficacy) [3]. This model suggests that attitude, social influence and self-efficacy influence the decision to experiment with cigarettes [2,4].

Figure 1. Revised ‘attitude, social influence, self-efficacy’ (ASE) model.

1. Attitude is conceptualized as positive and negative evaluations of a type of health-related behavior. According to Ajzen and Fishbein, substance-specific attitudes are based on a mathematical function of the personal consequences of smoking and the value of those consequences [5].

2. Social influence is conceptualized as “the processes whereby people directly or indirectly influence the thoughts, feelings and actions of others” [6] and is measured in three dimensions: social norms, perceived smoking behavior and direct pressure. These dimensions may overlap and may work in different directions.

3. Self-efficacy is defined as an individual's expectations of his or her capability to perform a certain action [7]. The focus in this study is on self-efficacy in social refusal (i.e. the expected ability to refrain from smoking in social situations). Adolescents may not have plans about smoking but may lack the skills necessary to refuse when they are offered a cigarette by a friend [5].”

We previously extended the ASE model by adding three school factors and thus taking into account effects of the school environment on adolescents [8] (Figure 1). Five reviews and theoretical studies (the latest was in 2002) suggest that norms of friends, smoking behavior of friends, attitude and self-efficacy for refusal are the ASE factors most closely associated with smoking [3,9–13], while the roles of parental smoking and norms are of questionable relevance [10]. Qualitative studies have indirectly addressed the role of ASE factors in adolescent smoking, most focusing on peer influence from an explorative perspective with no explicit theoretical framework. A review published in 2007 of 78 qualitative studies on adolescent tobacco use underlines the importance of peer influence in smoking initiation [14].

In this article, which is based on several previous studies for a PhD degree, we critically and constructively examined the theoretical and methodological implications of using the ASE model in studying adolescent smoking initiation, using the results of a qualitative study. The qualitative study was conducted in connection with and sampled from a large quantitative study which was previously reported [8,15,16]. As illustrated in Figure 2, both were follow-up studies using data from the 2004–2006. In this article, the results of the qualitative study are discussed while taking into account the results of the quantitative study and other studies (quantitative and qualitative studies published after 2001 as well as review papers).

Figure 2. Overview of sub-studies.

Materials and methods

This qualitative study was based on baseline and follow-up semi-structured interviews and observations with adolescents at two schools that participated in the Danish Youth Cohort (DYC). The DYC study is a randomized study (schools assigned to either intervention or control condition) of the effect of a school prevention programme called ‘Tackling’, which was inspired by the Life Skills Training prevention program found to be effective among adolescents in the USA [17]. The Danish program was implemented by the National Board of Health at 79 Danish schools, beginning in 2004, and involved educational sessions in grades 7, 8 and 9. The program focused on prevention of alcohol, tobacco and illegal drug use by informing adolescents about these substances and by improving their social skills, to enable adolescents to handle group pressure and feelings such as anger and anxiety. Smoking is initiated in adolescence [18], and, as we were interested in information about adolescents before they started smoking, we decided to focus on adolescents in 7th grade and follow them over time.

School classes

We decided to draw a convenience sample of two schools classes from Danish Youth Cohort study, which we had access to: one intervention and one control school class. The two schools, the Earth School and the Ocean School, were both located in the Copenhagen area but were quite different in terms of representation of ethnic groups, the socioeconomic status of the area and intervention vs. control school. The differences between the two school classes were considered factors, which may potentially influence smoking and the ASE factors, but not the relationship between smoking and the ASE factors.

The Earth School was an intervention school located in an area of Copenhagen, Denmark, with a population of 46 000, 9.7% persons with foreign citizenship and a mean yearly income of 167 000 DKK [19]. There were 26 students in the class, seven of whom had an ethnic background other than Danish. The Ocean School was a control school located in a Copenhagen area with a population of 15 349, with a higher percentage of immigrants compared to the other school class area (13.7% with foreign citizenship) and the lowest mean income (146 200 DKK) of the 15 local areas of Copenhagen [19]. There were 12 students in the class, 10 of whom had an ethnic background other than Danish.

Interviews

A structured interview guide was designed (Figures 3 and 4) on the basis of the ASE model and with inspiration from Wengraf (2001) [20] and pilot tested with five adolescents in June 2004. We wished to interview eight adolescents in each class, and we aimed at including an equal number of boys and girls and lifetime smokers and non-smokers. Teachers were asked to assist in selecting students who represented these criteria and who were expected to agree to participate. Adolescents were informed about the study as a group and were then approached individually and invited to participate. If the adolescent provided written consent to participate from their parents, they were enrolled in the study; if they did not provide written consent to participate, the teacher was asked to identify another student. Fifteen adolescents (eight at the Earth School and seven at the Ocean School) participated in the first interview round in September 2004–January 2005 (grade 7), and 12 of them in the follow-up interview in March–May 2006 (grade 8). Of those participating in both rounds, seven adolescents (three boys and four girls) were at the Earth School (of a total of 26 in the class) and five (three boys and two girls) were at the Ocean School (of a total of 12 in the class). Of the 15 participants in the first interview round, three did not participate in the second round. One had moved to a new school and two did not want to participate: one gave no reasons for not wanting to participate and one reported personal reasons. Two of the participants at the Earth School had non-Danish ethnic background: one was Asian, and the other was from Pakistan. At the Ocean School, all of the participants had non-Danish ethnic background: the parents of two were from Turkey, one was born in Pakistan, the parents of one were from Pakistan, and one was born in Iran. The interviews took place at school, were conducted in Danish, were tape-recorded and lasted 20–40 min.

Figure 3. Qualitative interview guide: structure of themes. The central research question was: how do social–cognitive factors influence smoking initiation among adolescents?

Figure 4. Sample of interview questions.

Observations

Field observations lasting one to two weeks were made twice at an interval of six months at the two schools in order to establish relationships with the participants before the interviews and also to gather information on their social context [21]. Although the focus was on smoking, shorter observations were made of: the physical environment, the actors, the overall events taking place, concrete actions and the actors’ goals and emotions. Also, two students at each school were asked to give the author a guided tour of the school.

Data analysis

Smoking was defined as lifetime smoking, thus including all experience with smoking cigarettes, from a single puff to daily smoking. In the analyses, we explored the ASE constructs according to how they are described in the ASE theory. Furthermore, we examined contradictions and aspects which are not explained in the model and if relevant discussed these aspects using other theoretical frameworks. The guide to qualitative data analysis of Miles and Huberman [22] was found to be practical for organizing data because it allows use of the strict theoretical model guiding the analysis [22]. The steps in the qualitative analysis are illustrated in Figure 5. Data were coded and managed with Nvivo 2.0 software for qualitative data management. We also drew on Burr's approach to the construct of discourse in social constructionism [23], which is a critical approach which may challenge but also supplement social–cognitive theory, especially when contradictions or phenomena not defined in the theory are identified. To protect the identity of the participants, all names and references to schools have been changed.

Figure 5. Steps in qualitative data analysis.

Results and discussion

Three of seven participants at the Earth School and four of five participants at the Ocean School had tried smoking. Four persons had tried smoking by the time of the first interview, and six had tried smoking by the second interview; one smoked daily. One boy said at the first interview that he had tried smoking, but at the second interview he said that he had never tried smoking. There are many potential reasons for this change. He might have redefined the smoking experience as ‘not a real one’, indicating the importance of time in investigations of ASE factors and smoking initiation. It is possible that the categorization into lifetime smoker and non-smoker may not be as solid as is often assumed: adolescents may use the category to define themselves in certain situations. As this was based on a single observation it needs to be further explored in future studies.

Social–cognitive factors

Attitude: Impact of discourse. Previous findings. Our previous quantitative results suggested that attitude could play a role in smoking initiation, but mainly among ‘early’ starters [16]. Other quantitative studies have obtained divergent results: of four studies (two cross-sectional [24,25] and two longitudinal [26,27]), only one cross-sectional study provided support for this hypothesis [24].

Current findings. The attitude construct was challenged in the current qualitative study. According to the social constructionist perspective presented by Burr, what people say when asked about their attitude represents discourses that can be seen as a story or a way of structuring language, to interpret things in a certain way [23]. All the participants in the qualitative analyses could describe the disadvantages of smoking, including health effects, effects on looks and smell, risk for addiction and effect on physical shape. Some said that it was ‘stupid and disgusting’. Interestingly, none of the participants mentioned any advantages of smoking. In a British qualitative study, 13- and 15-year-olds cited positive aspects of smoking, such as making them appear older, having a calming effect, keeping them thin, liking it and finding it useful for socializing [28]. There might be several reasons for the lack of a positive attitude in the current study. The adolescents may not yet have developed stable attitudes towards smoking. Alternatively, in the social constructionist perspective of Burr, the participants used negative discourses about smoking and used information on the negative health effects of smoking from the public health arena to position themselves in the social world. To our knowledge, no studies have investigated the changes in discourses on smoking in Denmark over time. A study was conducted recently in Britain, however, which hints at a development also taking place in Denmark [29]. The authors reported that the public health discourse in the 1950s changed a focus from acute infectious disease to chronic disease. In the 1950s, the public health tactic was characterized by ‘systematic gradualism’, which implies reducing harm by treatment and prevention, often with industry connections. From the 1960s to the 1980s, the public health tactic was characterized by ‘coercive permissiveness’, which argued for individual self-determination in a framework defined by the state and thus regulation of public space [29]. In the 1990s, the public health tactic was characterized by ‘pharmaceutical public health’, with a focus on smoking as an addictive drug [29]. A recent article from Australia also discussed developments regarding smoking in public health as a denormalization process in which smokers may be stigmatized [30]. In Denmark, the discourse related to smoking also seems to have changed radically over the past few decades. In the 1950s, more than 75% of Danish men smoked [25] and being an adult smoker was broadly acceptable. Since then, the smoking rates as well as the public attitude towards smoking have changed and smoking was banned in public places in 2007. Negative discourses on smoking were also seen at local level in the current study, at the two schools. Participants at the Earth School, for example, reported that that they had told two new students in the class who were daily smokers about the negative effects of smoking.

Amir, interview 2, Earth School

A: And we tell them. My teacher Inge, she has, she also criticises them a lot, right? […]

A: I tell them personally: “It is not so good”, right?, “It is your life, ‘you’, right?” […] But I say “it is up to you”. But I just want to. I just want to help them, right? They say “We know that it is not good, but can't stop and we won't stop”.

Both the teacher and the students tried to make these students stop smoking by using negative smoking discourses. The discourses may have been used because they told a story about the students and teacher, e.g. that they care about others and about health issues.

Parental smoking: Implications for social positioning. Previous findings. Our previous quantitative baseline and the first follow-up (six months) analyses showed that father's smoking was significantly associated with smoking initiation, but no significant association was seen at the second follow-up (18 months) [16]. A significant association with mother's smoking was found only at baseline. Other quantitative studies had mixed results on the effect of parental smoking. Of 11 studies, six longitudinal [26,31–35] and three cross-sectional studies [36–38] supported an association, and two cross-sectional studies did not [25,39]. This indicates that different factors are important for adolescents who initiate smoking ‘early’ and ‘late’ [27,40,41].

Current findings. In the ASE model, parental smoking is seen as working through social modelling, in that adolescents infer that smoking is acceptable when they see their parents smoke. In the current qualitative study, the smoking habits of the parents of classmates were brought up by several participants at the Earth School. Adolescents used negative smoking discourses to describe classmates whose parents were smokers, perhaps as a way of influencing social position.

Mette, interview 1, Earth School

M: For example, the classmate I sit next to. I get sick because she really stinks of smoke, her clothes. I feel sorry for her because I get sick from sitting next to her because she stinks of smoke.

I: It is her parents who smoke or what?

M: Yes, her dad smokes. I am not sure if her mom smokes but I think so. And it is really unpleasant, right? Because I can't just say “You smell - I get sick”, right?

Parental smoking may thus involve more than social modelling, as implied in the ASE construct. Even if parents smoke, the child may not model the parents. Interpretation of parental smoking by friends may play an important role in whether the behavior will be modelled.

Friends’ smoking: Fluidity of friendships. Previous findings. Our previous quantitative study showed that best friend's smoking was mainly important for ‘early’ but not ‘late’ initiators [16]. Of seven other quantitative studies, three longitudinal [35,42,43] and three cross-sectional studies [24,37,39] supported an association and one longitudinal did not [44].

Current findings. One challenge to examining the influence of friends on smoking over time is that friendships are fluid: adolescents may participate in several friendship groups and enter and leave different groups over time [45]. This is reflected in the qualitative data in the current study, in which different definitions of friends (close friends versus someone at school) are used. Nevertheless, all of those who had tried smoking had friends who smoked, and all reported that friends or someone they knew were present when they first tried smoking. A review of qualitative studies on the importance of friends’ smoking [14] concluded that the opportunity to try smoking arises through friends.

Norms of friends and teachers. Do adolescents talk about smoking? Previous findings. Our previous quantitative analyses showed only a marginally significant association between friends’ and teachers’ norms and smoking initiation in the first follow-up [16]. Only two other studies (both longitudinal) examined the effect of friends’ norms, one supporting an effect [42] and the other not [26]. Although teachers’ norms on smoking have been suggested to affect smoking initiation, only two studies (both cross-sectional) were identified, neither of which supported an association [46,47].

Current findings. Several participants in the current qualitative study said they did not know what their friends thought about smoking because they had not discussed it with them. Instead, knowledge about friends’ norms appeared to come from their actions.

Omar, interview 1, Ocean School

I: Do you know what your different groups of friends think of smoking? Is that something you have talked about?

O: No, we don't talk about things like that, because we are playing [football] and walking around.

The ASE construct for friends’ norms thus attempts to capture a construct that might not always exist in this age group in the form proposed.

The participants in the current qualitative study were not asked directly about their teachers’ smoking norms, but some of those at the Earth School brought up the issue: the norms of the teacher were clear to all the students. Thus, the qualitative analysis suggested that friends’ norms might not be as explicit as those of adults; it is therefore possible that separate constructs should be used for teachers and friends. Overall, both findings of the current as well as previous studies suggest that the combination of friends’ and teachers’ norms does not play a role in lifetime smoking initiation.

Pressure from friends: A way of defining friendships. Previous findings. Our previous quantitative results suggested that peer pressure may play a role among the adolescents who start ‘late’ and who have been able to withstand other influences [16]. Only one cross-sectional study found an association [24]. A broad range of pressures to smoke by friends [48] have been identified, making comparisons difficult.

Current findings. The participants in the current qualitative study were asked if they had felt pressured to smoke by friends. Some of the participants changed their answers between the two interview rounds. All three participants who said that they had felt pressured were boys; three girls said that they had been offered a cigarette by their friends but had not felt pressured. One interviewee who said that she was not pressured was Mette:

Mette, interview 2, Earth School

I: OK. So it was them [some friends in the 7th grade] who asked you? It wasn't something you wanted and had asked if you could try?

M: No, but I just tried it. I don't know. I wanted to, or I didn't want to try, right, but still. I wanted to try but it wasn't because I would become addicted or anything. It was just to try it.

In contrast, several studies found that peer pressure was associated with smoking initiation only among girls [49–51]. The girls in the current study were friends with the persons who offered them a cigarette, and they may have accepted in order to keep the friendship intact. If they had defined the act as pressure, they would have been questioning the quality of the friendship and the reason for keeping it going, as suggested in a previous article [52]. Several qualitative studies have explored pressure from friends [14,53]. A study of adolescents aged 15–16 questioned in individual interviews (n = 20) and focus groups (n = 123) found that adolescents did not feel pressured but reported that friends and peers had influenced whether they smoked [53]. In the ASE model, pressure is almost an objective event, which either takes place or does not. Instead, the event can be interpreted according to the situational goals. As the interpretation says more about the relationships with the persons present this may make it difficult to measure a true effect of pressure in quantitative studies.

Self-efficacy. When smoking is unplanned and situational. Previous findings. Our previous quantitative baseline analyses showed only a marginally significant association in the second follow-up [16]. Other previous studies on self-efficacy have had mixed results. Only one study (longitudinal) found that self-efficacy had an effect, and that only indirectly, on intention to smoke [26]. The lack of effect of self-efficacy may be due to the complexity of the construct, especially for adolescents.

Current findings. Social–cognitive models are based on an assumption of a rational adolescent, who consciously and systematically uses the available information to make the best decision and who is conscious about his or her future behavior and intention to behave in certain ways [54]. Some of the participants in the current study who were coded as having strong self-efficacy (said that they would be able to refuse smoking), however, had not planned to try smoking but had tried it anyway, perhaps because smoking was unplanned, unmotivated and driven by curiosity, as noted in other quantitative [55,56] and qualitative studies [57]. Three persons changed from strong self-efficacy to weak self-efficacy over time, suggesting that self-efficacy is not a stable trait. One girl explained that the ability to say ‘No’ to smoking depended on the social opportunities in a specific situation.

Mette, interview 1, Earth School

I: […]Do you also think that you will be able to say “No” to cigarettes later on?

M: I am not sure because it depends a lot on who you are with and who you are and those kinds of things. And I am not able to say beforehand who I will be with at a certain time. Perhaps I will be with people I shouldn't really be with. I am not sure, but right now my attitude is that I don't want to smoke.[…]

Mette indicated that smoking was part of who you are—your identity—and that it could be used as a social tool to become part of a group. A similar observation was made in other qualitative studies [58]. Friendships are not necessarily formed before smoking: adolescents might try to become adopted by a community of smokers in order to improve their social position (‘social selection’). Some studies have emphasized the role of peer selection as opposed to peer pressure [43,52], but most conclude that both processes are at play [36,47,59,60]. These nuances in decision-making are not covered by the ASE model.

School class and factors influencing smoking through friendship structures. Current findings. The school class may play an important role in the selection of friends, because persons in the same class are likely to have more in common than with those in other classes [61], and the school culture may affect adolescents’ lifestyle and health choices [45]. The two school classes in our current study differed in terms of school class culture, intervention status and ethnicity. The observations suggested that participants in the Earth School class usually had friends in their own class, and most considered that the class was cohesive. At the Ocean School, the participants were heterogeneous, with many different cultures and view points represented. There was a strong division between boys and girls, both physically (boys sat on the left-hand side of the classroom and girls on the right) and socially. They were characterized as having few friends in the class and many friends outside the class. The class cultures at the two schools are likely to have played a role in smoking initiation: the strong cohesiveness and the widespread negative discourse on smoking at the Earth School may have made experimenting with smoking less attractive. The quantitative study also indicated a role of school class and gender group in the class in smoking initiation [16]. Furthermore, as the Earth School class was the intervention class and the Ocean School class the control class, students at the Earth School received lessons on prevention between the first and second interviews, while those at the Ocean School did not. This is likely to have made the Earth School adolescents’ more conscious about smoking and about the ASE factors. In our interpretation of the ASE model we expected that an intervention program would affect the ASE factors which would then affect smoking. Furthermore, we did not expect that the ASE factors would work in different ways in the intervention class compared to the control class. Although, this was not the focus of our study, we found no data going against this expectation.

The interviews and observations showed that the primary friendships of many of the participants were linked to their religious or cultural identity. Religious and ethnic identity was not addressed directly in the interviews, but some of the participants brought up these issues, which were also seen in the choice of language. In the Ocean School, much of the talk during breaks and in group projects was in other languages than Danish, such as Turkish. All the participants who had tried smoking said that the friends they had been with at the time were of the same gender, and some indicated that the friends were of the same cultural or religious group. If friendship groups are structured according to ethnicity, smoking patterns may also be structured according to ethnicity. Danish studies on these issues are needed.

Methodological considerations

The qualitative analyses have several methodological and theoretical advantages. The follow-up design made it possible to study individual development of smoking over time. Strict theoretical models such as the ASE model are rarely used in qualitative studies, because they impose a set of research questions, which may blind the researcher from the field and thus prevent him or her from seeing phenomena outside the pre-defined theoretical constructs. Here, the triangulated approach (discussing results from both the current qualitative and previous quantitative studies) provided a unique opportunity both to examine the influence of social–cognitive factors quantitatively and to discuss and question the theoretical model on which they are based in a qualitative setting. The participants in the qualitative study were not representative of the Danish Youth Cohort or the general Danish population; e.g. a larger percentage were not ethnic Danes and the participants were based on a convenience sample. In addition, a total of seven instead of the anticipated eight adolescents in the Ocean class wished to participate and a total of three adolescents only participated in the first round. It is possible that the non-participating adolescents, and the drop-outs may be different from the rest of the class: they could have more problems and could be more likely to smoke. The results of the qualitative study should thus be viewed only as indicating possible processes. We chose to conduct the interviews and observations at school, because this is considered to be the most important social context for smoking among Danish adolescents. The focus on the school might have imposed greater emphasis on social–cognitive factors related to classmates than those to family members and friends outside school. Further studies of contexts outside the school are needed to explore this possibility. Finally, the observations were conducted in order to establish a relationship with the participants and an understanding of their context as this is relevant in order to create a relaxed and open setting for the interviews. The established relationships have most likely influenced both the behavior and the stories of the participants. In this way, the observations, probably like the qualitative interviews as well as the quantitative questionnaire study, became part of the local discourse and might have affected the adolescents’ lives. A direct example of this was seen regarding the questionnaire study:

Anne, interview 1, Earth School

I: […] Have you talked about it [about what your friends think about smoking] or?

A: It is not something we talk about, but when we answered the questionnaire it also said “What do you think your friends would think if you smoked?” And then I just asked them and they said that they wouldn't like it.

Thus, the exploration of smoking and the ASE factors in scientific studies may influence the adolescents’ perception of the theoretical concepts and thus the answers that they give us.

Conclusions

The current qualitative study partly support use of the ASE model and support previous studies in that friends’ smoking and pressure and parental smoking play roles in smoking initiation. Nevertheless, the qualitative study showed that certain aspects are not captured by the model and some of the processes are unclear. It is possible that models such as the ASE are not ideal for adolescent populations, as the health behavior of this age group is more situational, and social–cognitive constructs are less stable than in adults. The qualitative analysis further suggested that the discourse construct may supplement the attitude construct and underline that stories applied about the world are not stable and may be replaced depending on the situation. Also, although the ASE model takes into account social factors, they are incorporated as distal factors that act on behavior through cognitive processes. It is possible that social factors influence behavior directly, independently of cognitive processes. Two theoretical models have been designed to take this into account: the model of unplanned smoking initiation by children and adolescents [3] and the trans-theoretical model of change integrating behavioral and motivational stages. These models have, however, been challenged [62], as nicotine dependence may start at early stages of cigarette use and thus play an important role in smoking development [63]. It has been reported that smoking prevention programmes for adolescents have no or limited long-term effects [64,65]. Consideration should be given to whether it is possible, as most of these prevention programmes assume, to manipulate social–cognitive factors that have not yet developed into stable characteristics in adolescents. If the potential limitations of the ASE model are confirmed in future studies, the model should perhaps be revised for use in adolescent populations.

Acknowledgements

The study was supported by the Pharmaceutical Foundation of 1991 (Apoterkerfonden af 1991), the Danish Lung Association's Research Foundation (Danmarks Lungeforenings Forskningsfond) and the Foundation of 17.12.1981 (Fonden af 17.12.1981). The authors wish to thank the Danish National Institute of Public Health for access to data.

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.