Smoking behavior of Dutch students: validation of a contextualized
assessment
Ilva Grond
Supervisor: Dr. Helle Larsen
Developmental Psychology, University of Amsterdam
ABSTRACT. The aim of this research is to validate an audio simulation of social contexts in order to
optimize the assessment of smoking behavior. In total, 81 Dutch students (34.6% men) participated in this research. In this research we will investigate a) to validate audio simulation in order to optimize the assessment and prediction of the smoking behavior, and b) to measure and compare the predictive value of implicit and explicit smoking associations in a smoking context. Participants listened to audio simulations which contained questions about their willingness to accept a smoking in a smoking con-‐ text, and completed measures about their implicit and explicit smoking attitudes and nicotine depen-‐ dence. The willingness to accept a smoking offer obtained by the audio simulations predicted high-‐risk smoking at one month follow up above and beyond baseline behavior. Especially the dinner party and the festival scenario predicted above and beyond baseline. It can be concluded that the social context plays an important role in the smoking-‐related decision making and this should be included in steps into understanding context-‐related decision making.
Keywords: cigarette smoking, audio simulation, students, willingness, social context, validation.
A major public health problem among young adults is cigarette smoking. According to statistics of the CBS from 2012 about one out of four young Dutch adults (between the age of 16 and 20) is considered a smoker. College stu-‐ dents that see themselves as a social smoker, don’t perceive themselves as “smokers” and don’t think that they are at the risk of becoming nicotine dependent (Majchrzak et al., 2002). By describing themselves as a social smoker, stu-‐ dents mean that their smoking behavior is more a part of their social activities rather than a nicotine dependent behavior (Moran et al., 2004). Majchrzak et al. (2002) discovered that college students believe that after their gradua-‐ tion they will stop smoking, however, research has indicated that this is not the case (Scho[ield et al., 1998). In the last few years, social smo-‐ king behavior of students has been frequently studied. Moran et al. (2004) used cross-‐sectio-‐ nal survey to investigate the social smoking among US college students. They found that the social smoking pattern among occasional smo-‐ kers, was associated with a lower likelihood to quit attempts in the last year and also less wil-‐ lingness to change the smoking behavior.
Most studies investigating smoking be-‐ havior, were cross-‐sectional and focused name-‐ ly on the lifestyle and demographic factors that can be related to this behavior (e.g. Rigotti et al., 2000). However, there are two possible downsides about the cross-‐sectional method (Clapp et al., 2000). Clapp et al. (2000) looked at studies about alcohol consumption, but his
comments are also applicable to previous cross-‐sectional studies about smoking behavi-‐ or. The problem is that cross-‐sectional studies mostly use a standard list of situations to link the consumption-‐decision making to that mo-‐ ment. These situations are for some people en-‐ ough to remember a speci[ic moment and they are able to link their consumption to that situa-‐ tion, while for others this list might be confu-‐ sing and therefore are not able to relate certain consumption-‐decision making to that moment. This might result in a high variation between different subjects without the possibility to correct for that. Because of the downsides of cross-‐sectional studies, this study will have its focus on the longitudinal aspect.
Longitudinal research of previous stu-‐ dies showed that smoking behavior changes in the period between adolescence and young adulthood (Everett et al., 1999; Zhu et al., 1999). It is well known that a lot of different factors can in[luence smoking behavior, and one of the most important factors is the social in[luence (Chassin, Presson & Sherman 1984). Anderson, Duncan, Buras, Packard, and Kenne-‐ dy (2013) investigated the social contexts that are related to heavy alcohol consumption, with the focus on underage college students. The authors developed realistic audio simulations of alcohol-‐related decision making and evalua-‐ te the predictive validity of this model. These simulations consisted of [ive common drinking situations e.g. a drinking-‐game, pre-‐drinking event, and a party at a dorm room. The results demonstrated that the willingness to drink, obtained by the simulations, predicted high-‐ risk alcohol consumption at 8 months follow-‐ up above and beyond baseline consumption
(Anderson et al., 2013). This result showed that the social context plays an important role in the assessment of alcohol consumption.
A lot of studies in the past looked at the explicit smoking association in youth adults and also conducted different methods to mea-‐ sure these associations, for example the Smo-‐ king Consequences Questionnaire (Brandon & Baker, 1991). Measures that are conscious or explicit are direct and are dependent on the willingness of the participant to report the atti-‐ tude as well as the ability of the person to pre-‐ cisely assess this attitude (Sherman et al., 2003). The SCQ was the [irst measurement to examine the subjective expected utility (SEU) of smoking (Brandon & Baker, 1991). The SCQ consists of an 80-‐item questionnaire and Myers et al. (2003) shortened this questionnaire into an 21-‐items list, called the Shortened Smoking Consequences Questionnaire (S-‐SCQ). The items in the S-‐SCQ (as well as the SCQ) can be divided into four different categories namely: Negative Consequences, Positive Reinforce-‐ ment, Negative Reinforcement, and Appetite-‐ Weight Control (Myers et al., 2003). The items regarding to negative consequences, are state-‐ ments about the awareness of the risks of smo-‐ king. A stimulus is called a negative reinforce-‐ ment if the removal of this stimulus is resulting in the increase of a speci[ic behavior (Flora, 2004). If we look at the negative reinforcement for smoking behavior, the stimulus is your he-‐ alth. The knowledge that smoking cigarettes is damaging your health might in[luence the choi-‐ ce of smoking a cigarette. For example: ‘By smoking I risk heart disease and lung cancer’ (Myers et al., 2003). Positive reinforce-‐ ment focusses on the stimulating aspects of smoking, for example: ‘I enjoy the taste sensa-‐ tions while smoking’ (Myers et al., 2003). Nega-‐ tive reinforcement emphasis that the behavior will be strengthened by avoiding a negative outcome. In other words; by performing a cer-‐ tain behavior (smoking), the negative feeling (a bad emotion) you had will go away or become less. An example of this is: ‘Cigarettes help me deal with anger’ (Myers et al., 2003). The appe-‐ tite-‐weight control stresses the aspect of redu-‐ cing the appetite smoking. For example: ‘Smo-‐ king helps me control my weight’ (Myers et al., 2003).
A commonly used method for predict-‐ ing unique variance in measures of alcohol consumption is the Implicit Association Test (IAT; Greenwald et al., 1998). The IAT is based on the associations between two concepts and it measures the strength of this association. The test is executed on a computer and it records the response latencies (RTs) of the par-‐ ticipants when they categorize different stimuli
in the correct concept (Lindgren et al., 2013). When a participant has a shorter average la-‐ tency in for example the alcohol-‐tasty/water-‐ nasty task than for the water-‐tasty/alcohol-‐
nasty task, this can be interpreted as that the
participant has a stronger association of alco-‐
hol with tasty than nasty. People might have
limited control over memory-‐associations which are thought to be re[lected by the effects of the IAT (Osta[in, Palfai, & Wechsle 2003). There are different kind of Implicit Association Tests, based on the variety of associations that can be measured. The outcomes of the IAT vary per test, depending on the study method (Lindgren et al., 2013). Lindgren et al. (2013) compared [ive alcohol-‐related variants of the IAT to develop an overview of their validity and reliability. The most reliable predictor of the consumption of alcohol was the Drinking Iden-‐ tity IAT (Lindgren et al., 2013). This IAT was specially designed for Lindgren’s research and it measured associations of “drinker” with “me”. This Drinking Identity IAT is based on the [indings of Fekadu & Kraft (2001), demon-‐ strating that the predictability of a model im-‐ proves when the measures of how strongly one identi[ies with the behavior (e.g. drinking) is included. By showing words that represent the self vs. others, Lindgren et al. were able to mea-‐ sure the identity of the association of alcohol-‐ related stimuli. This Identity IAT is relatively new and has only been conducted with alcohol consumption and is never been used for other research.
In the current study, we investigated whether the social context plays an important role in the smoking-‐related decision making and if we can get similar result as the results mentioned above from Anderson et al. (2013) research. The main aim is to validate audio si-‐ mulations for smoking in order to optimize the assessment and prediction of the smoking be-‐ havior, with particular interest in students. We are also interested in whether the Smoker Identity IAT (specially designed IAT for this research)(IAT; Greenwald et al., 1998) and the behavioral willingness obtained with audio si-‐ mulations are good predictors for smoking be-‐ havior at one month follow up.
The main question of this research is if there is a correlation between the behavioral willingness to accept a smoke offer with the self-‐reported smoking at baseline and at fol-‐ low-‐up? And also whether willingness to smo-‐ ke (assessed with the audio simulation) is a better predictor of smoking behavior at follow-‐ up than self-‐reported smoking? To be able to answer these questions a few sub questions will be posted. The [irst one is; do implicit smoking associations, assessed with the IAT in
a smoking context, correlate with the self-‐re-‐ ported smoking behavior at baseline and at follow-‐up? The second question is whether the willingness to smoke a cigarette in social events is in[luenced by different environments (different scenes e.g., public versus private set-‐ tings)?
If there is a correlation between the behavioral willingness to smoke a cigarette and the self-‐reported smoking behavior at baseline and at follow-‐up, then this might lead to a bet-‐ ter understanding of smoking behavior of Dut-‐ ch students what might result in a optimizing the validation of the smoking simulation. The same applies to the IAT, if there is a correlation between implicit smoking associations and the self-‐reported use then this will give a better understanding of the social smoking behavior and its consequences. Besides that, if we [ind that a different environment (public versus private settings) has a different in[luence on the willingness to smoke a cigarette, this could mean that in an environment with a lot of sti-‐ muli for smoking cigarettes and less restricti-‐ ons (for example being outdoor), there could be a higher willingness to smoke than in an en-‐ vironment with less stimuli and more restricti-‐ ons. This could be due because the threshold for smoking a cigarette will be lower.
These [indings can be the beginning of the generation of ecologically valid models of smoking consumption. If we will be able to de-‐ velop these models then it will be beginning of the development of prevention programs that can target motivational factors and the con-‐ texts, that put young adults at increased risk for illicit and harmful cigarette use. To improve the strategies to prevent young adult cigarette smoking, the understanding of the underlying process in smoking-‐decision making is very important.
METHOD
Phase 1: Development of the method Focus groups
Seventeen undergraduate students (29.4% men; Mage = 21.5 years; SD=1.81) from the University of Amsterdam, the Netherlands, were recruited for two focus groups (7 and 10 participants) via [lyers that were spread out through the different faculties. All participants were Dutch and they all smoked at least one cigarette in the past month. The focus groups were held to achieve a better understanding about the smoking behavior of students at dif-‐ ferent social events (e.g. when do they smoke, why do they smoke, what they smoke) The stu-‐ dents also answered these questions about
their eating behavior, with the focus on eating snacks (e.g. when do they eat, why do they eat, what do they eat). After every question partici-‐ pants had the opportunity to discuss their answers with the group. Participating in the focus group was rewarded with school credits. Script evaluation. Students listened to [ive
conversations that were told out loud by the instructor of the focus group. Each conversati-‐ on described a different social scene that could be related to smoking. The [ive scenes were: (1) drinking at a terrace after an exam, (2) wat-‐ ching football in the pub, (3) a birthday party at someone’s place, (4) having dinner at a friend’s house and (5) visiting a music festival. As an example, the music festival can be described as: you are at a music festival, everywhere around you is music and it is very crowded. You and your friends are waiting for two others to arri-‐ ve and to kill the time you are talking about the line-‐up and the high price of the coins to buy drinks and food. While you are waiting a friend asks you: “I want a cigarette while we are wai-‐ ting for the other to arrive, do you want one as well?” Students answered questions regarding likability and how realistic the scenes were and what they thought that should be changed per scene. These answers were discussed in the group and resulted in improvements in the scripts.
Actor ratings. Eleven actors sent in their
audition tape containing a part of a scene read out loud. The participants of the focus group wrote down their reactions to each of the ac-‐ tors. Every actor was scored on their realism, believability and attractiveness. Based on these results the voice actors for the audio simulati-‐ ons were selected.
Audio simulation production
A professional scriptwriter developed the [ive different scripts and adapted the script ba-‐ sed on the commentary from participants of the focus groups. The [ive scripts could be divi-‐ ded into 3 different event sizes, small (drinking at a terrace after an exam, a birthday party at someone’s place), medium (watching football in the pub, having dinner at a friend’s house), and large (visiting a music festival). Another distinction that can be made between the diffe-‐ rent scenes is the distinction between public and private situations. The birthday party at someone’s place and having dinner at a friend’s house are the two private events. Drinking at a terrace after an exam, watching football in the pub, and visiting a music festival are classi[ied as public events. Included within each scenario were two types of offers recorded by the voice actors: smoking offers and neutral food offers.
These offers were [ixed within every different scenes.
The audio simulations were produced at a recording studio at the University of Amster-‐ dam. Also the background sounds (e.g. foot-‐ steps, clinging glass, opening the door), actor recordings and sounds for the party were re-‐ corded at the University of Amsterdam. Each audio simulation contains four different offers (two smoking offers and two food offers) fol-‐ lowed by a beep, allowing the participants to indicate how willing they were to accept the offer. Soundtrack Pro was used to edit all the background sounds, the scenes from the voice actors and the offers into complete audio frag-‐ ments that were usable for the research. These audio fragment were cut into four parts, with every part always ending with an offer. Eprime software was used to add the willingness scale for the offers in between the audio fragments.
Phase 2: Validation of the audio simulation
Participants
Students from the University of Amsterdam participated in this study. They were recruited via online promotion on the site of the Univer-‐ sity of Amsterdam and via [lyers spread out through different buildings from the university. Only students that had smoked at least one ci-‐ garette in the past month were allowed to par-‐ ticipate in the research. Twenty students with an average of 0 cigarettes smoked in the past two weeks were excluded from the analysis. A total of 81 students (34.6% men; Mage=22.8 years; SD=6.48) fully [inished the study.
Design and procedure
Audio simulations. Participants took place
in soundproof cubicles. Every student had to read and sign the informed consent form befo-‐ re they were allowed to participate in the expe-‐ riment. The protocol that we used was appro-‐ ved by the ethical committee of the University of Amsterdam. Before the audio simulations started, the participants could read the proce-‐ dure of the study and were told to put up the headphones. The students were asked to close their eyes and to visualize themselves as good as possible in the upcoming [ive situations. Right after that, the participants listened to the [ive audio simulations (around the 2 minutes) that were presented in random order on a Dell computer. Every 20-‐30 seconds the scene con-‐ tained a smoke or food offer followed by a break to answer the question how willing they were to accept the offer. The order of these of-‐ fers were [ixed for every scene and every scene contained two smoke offers and two neutral food offers. After answering the willingness
question, the audio simulation continued au-‐ tomatically. After the [ive scenes, participants answered a set of questionnaires and did a computer task.
Time 1 measures: May 2014
In May 2014 (T1), the participants comple-‐ ted the audio simulation assessments of wil-‐ lingness in a computer cubicle at the University of Amsterdam. First, the students answered some demographic and institutional questions regarding their age, sex, study, and [irst langua-‐ ge, followed by the audio fragments. After the audio simulations the data on nicotine depen-‐ dence, implicit and explicit smoking cognitions and smoking behavior were collected. This or-‐ der was maintained to avoid potential priming effects on what? Of what?.
Krank et al. (2005) designed a study to test if the predictive value is in[luenced by conditi-‐ ons that are designed to enhance implicit me-‐ mory associations. They created two different settings, a neutral and a alcohol context. By placing the IAT in front of any questions about drugs or alcohol, the neutral context was crea-‐ ted. The alcohol context was created by placing the IAT after the questions about alcohol and drug use. In the alcohol related context, the IAT was placed either directly after the questions, or delayed and the IAT was placed later in the survey, to determine if the context effect was time limited. Their results showed that the ma-‐ nipulation of the alcohol context improved the prospective predictive value of the IAT. Krank et al., (2005) suggest that placing the IAT di-‐ rectly behind the questions about alcohol might be more effective then later in the survey. Based on these results, we choose to do the IAT after the audio simulations because this might have a positive effect on the prospective predic-‐ tive value of the IAT.
Measures
Smoking behavior. Nicotine dependence
was assessed with the modi[ied Fagerström Test (mFTQ; Heatherton, Kozlowski, Frecker, & Fagerstrom, 1991). The amount of smoking from the past two weeks was assessed with the Time Line Follow Back (TLFB; Sobell & Sobell, 1990).
Behavioral willingness. During the audio
scenes, participants responded to a set of ques-‐ tions on a 1 (not willing to accept the offer) to 5 (very willing to accept the offer) scale asses-‐ sing willingness to accept or reject smoking offers or neutral food offers. The [ive different scenes were divided into smaller and larger private contexts: (1) drinking at a terrace after an exam, (2) watching a soccer game in a bar, (3) a birthday party at a friend’s house, (4) ha-‐
ving dinner at a friend’s house, and (5) a festi-‐ val. The participants will rate every offer made within each scene separately (e.g., ”in this sce-‐ ne, how willing were you to accept the food/ smoke offer?”).
Self-‐reported willingness. The self-‐reported
willingness (general willingness) was obtained by asking three different questions. Partici-‐ pants had to remember the scene they just lis-‐ tened to and for every scene they had to ans-‐ wer the following statements on a 1 (complete-‐ ly disagree with the statement) to 7 (complete-‐ ly agree with the statement) scale: (1) I would accept the smoke offer, (2) I would say “No, thank you”, and (3) I would leave the room. Explicit smoking cognitions. Explicit smo-‐
king cognitions were measured with the Short Smoking Consequence Questionnaire (S-‐SCQ, Myers et al., 2003). The S-‐SCQ consists of 21 items on a 10-‐point Likert scale (0: completely unlikely to 9: completely likely). Cigarette smo-‐ king outcome expectancies give inside in the motivations of students to smoke. The outcome of this questionnaire can be divided into four different categories: (1) Negative Consequen-‐ ces (4 items), (2) Positive Reinforcement (5 items), (3) Negative Reinforcement (7 items), and (4) Appetite-‐Weight Control (5 items). Implicit smoking cognitions. The implicit
smoking cognitions were measures with an adapted version of the Implicit Association Task (IAT; Greenwald et al., 1998), namely the Smoker Identity IAT, based on the Drinking Identity IAT (Lindgren et al., 2013). The IAT design that was used (based on Lindgren et al., 2013), consisted of seven blocks (see Table 1 in the appendix). The critical blocks involve sor-‐ ting stimuli items that represent the four con-‐ cepts in each IAT (e.g., smoke, non-‐smoker, me, not me) with two response options (left or right). For example, stimuli belonging to the ”non-‐smoker” or ”not me” categories are sor-‐ ted using the key on the left; stimuli belonging to the ”smoker” or ”me” categories are sorted using the key on the right. After two blocks, each containing 20 trials, the pairings are swit-‐ ched: stimuli belonging to the ”non-‐smoker” or ”not me” categories will be sorted using the key on the right; stimuli belonging to the ”smoker” or ”me” categories are sorted using the key on the left. The order of the pairings is counterba-‐ lanced (see Appendix B for more details).
Time 2 measures: June 2014
One month after the participants did the audio simulations, they received an e-‐mail with a link to a secure survey site to complete the smoking measures from that moment until two weeks before (TLFB; Sobell & Sobell, 1990).
After they did the online survey, the student were credited with 1 class credit or €10,-‐.
Results
All 81 students that were allowed to participated in the research smoked at least one cigarette in the past month, with an avera-‐ ge of 6.15 cigarettes per day (SD=9.07). No dif-‐ ferences in sex were found in behavioral wil-‐ lingness on the audio simulation or in the self-‐ report smoking behavior from the Time Line Follow Back at baseline or follow up. Table 1 provides the descriptive statistics for behavio-‐ ral willingness to accept a smoke offer and to accept a food offer across the [ive different sce-‐ nes. A one-‐way repeated measures ANOVA was conducted to compare scores on
the willingness to accept a smoke or a food of-‐ fer obtained with the audio simulations at [ive different environments; the terrace, the pub, a birthday party, a dinner party and a music fes-‐ tival. There was a signi[icant difference in the willingness to accept a smoke offer between different environments, Wilk’s Lambda = .683, F(4, 77) = 8.947, p<0.0005, multivariate ƞ2 = 0.317. There was also a signi[icant difference in the willingness to accept a food offer between different environments, Wilk’s Lambda = 0.537, F(4, 77) = 16.592, p<0.0005, multivariate ƞ2= 0.463.
A paired-‐sampled t-‐test was conducted to evaluate the difference in behavioral wil-‐ lingness to accept a smoking or a food offer in private and public environment (Table 1). The-‐ re was a signi[icant decrease in the behavioral willingness to accept a food offer in a public environment (M=5.72, SD=.99) compared to the acceptance in a private environment (M=4.80, SD=1.21), t(80)=6.95, p˂.001. There was no signi[icant difference in the willingness to accept a smoking offer between a private (M=4.03, SD=1.44) and public environment (M=3.99, SD=1.50), t(80)=.33, p˃.05. This indi-‐ cated a very small effect size. No signi[icant dif-‐ ference was measured between men and wo-‐ men. Varia ble Terra ce M (SD) Pub M (SD) Birth day party M (SD) Dinne r party M (SD) Music Festiv al M (SD) Smok ing 3.96 (1.93 ) 3.85 (1.82 ) 4.64 (1.64 ) 3.43 (1.78 ) 4.17 (1.68) Food 4.94 (1.81 ) 4.65 (1.61 ) 5.37 (1.47 ) 6.06 (1.06 ) 4.80 (1.41)
Table 2 shows the descriptive statistics and correlations between TLFB T1 and T2, ge-‐ neral willingness items, explicit associations, implicit associations, and behavioral willing-‐ ness to smoke and to eat. This table shows a very high positive correlation between the self-‐ report at baseline (TLFB T1) and the self-‐re-‐ port one month follow-‐up (TLFB T2). The TLFB T1 (at baseline) and the TLFB T2 (month follow up) correlate positively with the negati-‐ ve reinforcement, general willingness item 1 (I would accept the offer) and 3 (I would leave the room), the behavioral willingness to accept a smoke offer obtained with audio simulations and the implicit associations. The more cigaret-‐ tes the participants smoked according to the TLFB, the more they would accept a smoking offer and the more they imply themselves with being a smoker. There is a negative correlation between the TLFB T1 and TLFB T2 and the ge-‐ neral willingness item 2 (I would say: “No, thank you”). This means that the more cigaret-‐ tes the participants smoke regarding the TLFB, the less they would reject an offer. An indepen-‐ dent t-‐test was conducted to compare the self-‐ reported smoking behavior at baseline and fol-‐ low up for men and women. At baseline there was no signi[icant difference in scores for men (M=3.80, SD=5.66) and women (M=2.87,
SD=4.01; t(79)=.86, p=.39). Also for the follow
up measurement there was no signi[icant diffe-‐ rence between men (M=3.59, SD=5.24) and women (M=2.62, SD=4.11; t(79)=.91, p=.36).
Validation of audio smoking
The behavioral willingness to accept an offer (obtained with audio simulations) to predict the smoking behavior at one month follow up is obtained by performing a hierarchical multi-‐ ple regression. This regression was controlled
for the use of calming drugs and self-‐reported willingness (general willingness) to accept an offer. In Table 2 the correlations between the variables are presented. All correlations were moderate to moderately strong ranging bet-‐ ween r = .26, p<.05 to r = .49, p<.01. Except for general willingness item 3 (I would leave the room), all the variables were correlated with the TLFB T2 one month follow up. The correla-‐ tions between the variables and the dependent variable (TLFB T2) were moderate ranging between r = .28, p<.05 to r = .36, p<.01. Table 3 shows the hierarchical regression mo-‐ del of general willingness (self-‐report) to smo-‐ ke a cigarette and the behavioral willingness obtained with the audio simulation. The [irst step of the regression contains the four predic-‐ tors; calming drugs, general willingness item 1 (I would accept the offer), general willingness item 2 (I would say: “No, thank you”), and ge-‐ neral willingness item 3 (I would leave the room). This model was signi[icant F(3, 76)=3.27; p<.05 and explained 22.9% of the variance in smoking at one month follow up. After adding the behavioral willingness obtai-‐ ned from the audio simulation, the total varian-‐ ce explained by the model as a whole was 30.1% (F(2, 74)=3.80; p<.05). There was an addition of 7.2% in explanation of the model. This shows that the behavioral willingness to smoke (obtained from the audio simulation) predicts the self-‐report TLFB at one month fol-‐ low up above and beyond the general willing-‐ ness to smoke. The willingness to accept a food offer is not signi[icant, showing that there is no protective effect.
Table 2: Descriptive statistics and correlations between TLFB, general willingness (self-‐report), explicit associations, implicit associations, and behavioral willingness to smoke and to eat (N=81)
Notes: TLFB T1 = TLFB at baseline; TLFB T2 = TLFB at one month follow up; S-‐SCQ Neg. reinf. = S-‐SCQ negative reinforce-‐
ment; Gen. will. 1 = General willingness (self-‐report) item 1 (I would accept the smoke offer); Gen. will. 2 = General willing-‐ ness (self-‐report) item 2 (I would say “No, thank you”); Gen. will. 3 = General willingness (self-‐report) item 3 (I would leave the room); S-‐SCQ Neg. Reinf. = S-‐SCQ negative reinforcement; Impl. Ass. = implicit associations from identity IAT; Will. Smoke = behavioral willingness to accept a smoke offer measured with the audio simulations; Will. Food = behavioral willingness to accept a food offer measured with the audio simulations.
**p<0.01; *p<0.05. Variable 1 2 3 4 5 6 7 8 9 10 1. TLFB T1 1.00 2. TLFB T2 .96** 1.00 3. Calming drugs .35** .36** 1.00 4. Gen. will. 1 .25* .28* .15 1.00 5. Gen. will. 2 -.29* -.29* -.20 -.50** 1.00 6. Gen. will. 3 .08 .14 -.12 .09 .09 1.00 7. S-SCQ Neg. Reinf. .34** .32** .22* .25* -.03 .09 1.00 8. Impl. Ass. .24 .27* .10 .11 -.03 .04 .13 1.00 9. Will. Smoke .37* .27* .08 .37* -.26* .06 .20 .16 1.00 10. Will. Food -.07 -.13 -.17 .01 -.03 -.02 -.21 -.01 .08 1.00 Means 3.19 2.96 1.17 3.42 4.54 1.61 3.66 0.51 4.01 5.17 St. Dev. 4.63 4.52 .69 1.67 1.70 1.33 1.97 .36 1.37 .96 Min. .00 .00 1.00 1.00 1.00 1.00 .00 -.70 1.00 3.10 Max. 20.00 20.00 6.00 6.00 7.00 7.00 7.86 1.31 6.80 7.00
Table 3: Hierarchical regression model of general willingness (self-‐report) to smoke a cigarette and the behavioral willingness obtained with the audio simulation.
Notes: Gen. will. 1 = General willingness (self-‐report) item 1 (I would accept the smoke offer); Gen. will. 2 = General willing-‐
ness (self-‐report) item 2 (I would say “No, thank you”); Gen. will. 3 = General willingness (self-‐report) item 3 (I would leave the room); Will. Smoke = behavioral willingness to accept a smoke offer measured with audio simulations; Will. Food = beha-‐ vioral willingness to accept a food offer measured with audio simulations.
*p<.05; **p<.005.
Table 4. Prediction of the past two weeks average cigarettes per scene at the one month follow up by behavioral will-‐ ingness obtained by audio simulation (N=81)
Notes: Gen. will. 1 = General willingness (self-‐report) item 1 (I would accept the smoke offer); Gen. will. 2 = General willing-‐
ness (self-‐report) item 2 (I would say “No, thank you”); Gen. will. 3 = General willingness (self-‐report) item 3 (I would leave
Variable R R2 ΔR2 B SE β t Step 1 Calming drugs Gen. will. 1 Gen. will. 2 Gen. will. 3 .48 .23* 2.17 .32 -.48 .64 .68 .32 .31 .35 .33** .12 -.18 .19 3.17 1.01 -1.52 1.81 Step 2 Calming drugs Gen. will. 1 Gen. will. 2 Gen. will. 3 Will. Smoke Will. Food .55 .30* .07* 2.02 .10 -.41 .58 .92 -.46 .67 .32 .31 .34 .35 .47 .31** .04 -.15 .17 .28* -.10 3.01 .30 -1.34 1.70 2.65 -.99
Terrace Pub Birthday party Dinner party Festival
Variable B SE ΔR2 B SE ΔR2 B SE ΔR2 B SE ΔR2 B SE ΔR2 Step 1 Calming drugs Gen. will. 1 Gen. will. 2 Gen. will. 3 2.17** .32 -.48 .64 .68 .32 .31 .35 .23** 2.17** .32 -.48 .64 .68 .32 .31 .35 .23** 2.17** .32 -.48 .64 .68 .32 .31 .35 .23** 2.17** .32 -.48 .64 .68 .32 .31 .35 .23** 2.17** .32 -.48 .64 .68 .32 .31 .35 .23** Step 2 Calming drugs Gen. will. 1 Gen. will. 2 Gen. will. 3 Will. Smoke Will. Food 2.12** .19 -.34 .63 .50* -.20 .67 .32 .32 .35 .25 .25 .05 2.16** .30 -.50 .58 .14 -.20 .69 .32 .32 .36 .26 .29 .01 2.16** .25 -.37 .64 .51 -.12 .70 .32 .32 .35 .29 .32 .03 2.10** -.03 -.46 .58 .85** .21 .66 .32 .30 .33 .27 .43 .10* 2.11** .14 -.46 .61 .64* -.36 .68 .32 .31 .34 .27 .32 .06* Overall model F(2, 74) = 2.27, p>.05 F(2, 74) = 0.40, p>.05 F(2, 74) = 1.54, p>.05 F(2, 74) = 5.39, p<.05 F(2, 74) = 3.36, p<.05
the room); Will. Smoke = behavioral willingness to accept a smoke offer measured with audio simulations; Will. Food = beha-‐ vioral willingness to accept a food offer measured with audio simulations.
***p<.001, **p<.005, *p<.05.
To [ind out which of the [ive scenes had the most impact on the model, the regressions were done separately per scene as can be seen in Table 4. In the [irst step of the regression, four predictors were entered: calming drugs, general willingness item 1 (I would accept the smoke offer), general willingness item 2 (I would say “No, thank you”), and general wil-‐ lingness item 3 (I would leave the room). This model was for all the [ive different environ-‐ ments signi[icant F(3, 76) = 3.27, p<.05 and explained 22.9% of variance in willingness to accept a smoking offer at one month follow up self-‐report (Table 3). After entry of the behavi-‐ oral willingness to accept a smoking offer ob-‐ tained from the audio simulations at the se-‐ cond step, it was not signi[icant for the terrace, pub, and birthday party. The dinner party was signi[icant after entry of the behavioral wil-‐ lingness, namely F(2, 74) = 5.39, p<.05 and ex-‐ plained 32.7% of the variance. The addition of the behavioral willingness explained 9.8% va-‐ riance in the self-‐report one month follow up, after controlling for calming drugs and general willingness. Also the festival scene was signi[i-‐ cant F(2, 74) = 3.36, p<.05 and explained 29.4% of the variance. The introduction of the behavioral willingness explained 6.4% variance in the self-‐report TLFB one month follow up. Two out of [ive environments were signi[icant, with the dinner party recording a higher Beta value (β = .34, p<.005) than the festival (β = . 24, p<0.05). So a greater willingness to accept a smoke offer in the dinner party and festival scenarios, above and beyond baseline use, pre-‐ dicted increased smoking behavior one month later. There was no signi[icant difference if we looked at greater willingness as a predictor to accept food offers scene by scene.
Implicit and explicit associations
Table 5 shows the correlations between self-‐ report TLFB one month follow up, explicit as-‐ sociations (S-‐SCQ), behavioral willingness to smoke and to eat, and implicit associations. The self-‐report at one month follow up has a very strong positive correlation with the cal-‐ ming drugs, negative reinforcements obtained from the S-‐SCQ, and the behavioral willingness to smoke. The positive correlation of the TLFB one month follow up and the behavioral wil-‐ lingness indicates that the higher the amount
of reported cigarettes that have been smoked, the bigger the willingness to accept a smoke offer. Also a higher TLFB T2 report, the higher the negative reinforcement. This would mean that the more cigarettes you smoke, the chance that you smoke a cigarette to cope with negati-‐ ve emotions is bigger. In other words, heavier smokers are more likely to smoke a cigarette to deal with their negative situation. Besides that, TLFB T2 correlates positive with the calming drug report, meaning that heavier smokers would more easily smoke to calm down in a stressful situation. This explains the positive correlation between the calming drug use and the negative reinforcement. The self-‐report one month follow up also positively correlates with the measures from the implicit association task, indicating that the implicit associations that a person has with “being a smoker” are stronger when you are a heavier smoker. Table 5 shows a strong correlation between the be-‐ havioral willingness to accept a food offer and the S-‐SCQ Appetite Weight Control. However, this is not related to this research and will not be examined further.
Our results in Table 6 show that behavioral wil-‐ lingness to accept a smoking offer is predicting the self-‐report at one month follow up (TLFB 2) above and beyond the explicit smoking out-‐ come expectancies of the S-‐SCQ. Table 6 also shows that the negative reinforcement in the [irst step of the regression is signi[icant. After adding the behavioral willingness to smoke (obtained by the audio simulation), the associa-‐ tions with the negative reinforcement disappe-‐ ars. To see if the associations between the ne-‐ gative reinforcement and explicit smoking as-‐ sociations also disappears after the addition of implicit associations instead of behavioral wil-‐ lingness, we did another regression analysis. However, the implicit associations did not in-‐ [luence the negative reinforcement (Table 7).
Table 5: Descriptive statistics and correlations between self-‐report, explicit associations, behavioral willingness to smoke and to eat, and implicit associations (N=81)
Notes: TLFB T2 = TLFB at one month follow up; S-‐SCQ Neg. Con. = S-‐SCQ negative consequences; S-‐SCQ Pos. Reinf. = S-‐SCQ
positive reinforcement; S-‐SCQ Neg. Reinf. = S-‐SCQ negative reinforcement; S-‐SCQ App. W. Cont. = S-‐SCQ Appetite Weight Con-‐ trol; Will. Smoke = behavioral willingness to accept a smoke offer measured with audio simulations; Will. Food = behavioral willingness to accept a food offer measured with audio simulations, Impl. Ass. = implicit associations from identity IAT. **p<0.01, * p<0.05.
Table 6: Hierarchical regression model of self-‐report one month follow up and explicit associations.
Notes: S-‐SCQ Neg. Con. = S-‐SCQ negative consequences; S-‐SCQ Pos. Reinf. = S-‐SCQ positive reinforcement; S-‐SCQ Neg. Reinf. =
S-‐SCQ negative reinforcement; S-‐SCQ App. W. Cont. = S-‐SCQ Appetite Weight Control; Will. Smoke = behavioral willingness to
Variable 1 2 3 4 5 6 7 8 9 1. TLFB T2 1.00 2. Calming drugs .36** 1.00 3. S-SCQ Neg. Con. .05 -.19 1.00 4. S-SCQ Pos. Reinf. .11 -.11 .23* 1.00 5. S-SCQ Neg. Reinf. .32** .22* .14 -.01 1.00 6. S-SCQ App. W. Cont. .09 .14 .01 -.02 .50** 1.00 7. Will. Smoke .36** .08 -.02 .30** .20 .04 1.00 8. Will. Food -.13 -.17 .01 .02 -.21 -.35** .08 1.00 9. Impl. Ass. .27* .10 -.09 .11 .13 .04 .16 -.01 1.00 Means 2.96 1.17 7.64 5.00 3.66 2.42 4.01 5.17 .51 St. Dev. 4.52 .69 1.31 1.64 1.97 1.94 1.37 0.96 .36 Min. .00 1.00 2.50 .00 .00 .00 1.00 3.10 -.70 Max. 20.00 6.00 9.00 8.00 7.86 8.00 6.80 7.00 1.31 Variable R R2 ΔR2 B SE β t Step 1 Calming drugs S-SCQ Neg. Con. S-SCQ Pos. Reinf. S-SCQ Neg. Reinf. S-SCQ App. W. Cont. .47 .22 2.20 .15 .37 .67 -.24 .71 .37 .29 .28 .28 .33** .05 .13 .29* -‐.10 3.09 .41 1.28 2.39 -.88 Step 2 Calming drugs S-SCQ Neg. Con. S-SCQ Pos. Reinf. S-SCQ Neg. Reinf. S-SCQ App. W. Cont. Will. Smoke Will. Food .54 .30* .08* 2.03 .28 .10 .50 -‐.25 .97 -.43 .69 .36 .30 .28 .28 .35 .50 .31** .08 .04 .22 -‐.11 .30* -.09 2.94 .76 .33 1.79 -‐.91 2.74 -.85
accept a smoke offer measured with audio simulations; Will. Food = behavioral willingness to accept a food offer measured with audio simulations.
**p<.005, *p<.05
Table 7: Hierarchical regression model of implicit associations and S-‐SCQ.
Notes: S-‐SCQ Neg. Con. = S-‐SCQ negative consequences; S-‐SCQ Pos. Reinf. = S-‐SCQ positive reinforcement; S-‐SCQ Neg. Reinf. =
S-‐SCQ negative reinforcement; S-‐SCQ App. W. Cont. = S-‐SCQ Appetite Weight Control; Impl. Ass. = implicit associations from identity IAT.
**p<.005, *p<.05, †p=0.05
Discussion
The goals of this research was to con-‐ vert the mechanism that Anderson et al., 2013 developed for the alcohol-‐related decision ma-‐ king into a mechanism that can be used to in-‐ clude the in[luence of social context in the smoking-‐related decision making. By using the audio simulation the behavioral willingness could by associated with their smoking behavi-‐ or at one month follow up. Besides that we were interested if there is a correlation bet-‐ ween the behavioral willingness to accept a smoke offer with the self-‐reported smoking at baseline and at follow-‐up? And also whether behavioral willingness to smoke is a better predictor of smoking behavior at follow-‐up than self-‐reported smoking?
This research shows that the behavioral willingness to smoke predicts the self-‐report TLFB at one month follow up above and be-‐ yond the general willingness to smoke. This seems to be driven by the dinner party and the festival scenarios. Behavioral willingness is also predicting the self-‐report at one month follow up (TLFB 2) above and beyond the ex-‐ plicit smoking outcome expectancies of the S-‐ SCQ. Besides that, heavier smokers are more likely to smoke a cigarette to deal with their
negative situation. Also the self-‐report at one month follow-‐up correlates positive with the calming drug report, meaning that heavier smokers would more easily smoke to calm down in a stressful situation.
Table 2 shows a very high correlation between the self-‐report at baseline and the self-‐report at one month follow up. This can be due because of the short time period in bet-‐ ween. Anderson et al., 2013 had a time span of 8 months in between the two self-‐reported me-‐ asurements, one measurement at the beginning of the school year and the second measure-‐ ment at the end of the school year. Because the TLFB of this study at baseline and at follow up are measured so closely after each other, it is possible that the this time span of one month was too short to [ind a difference in smoking behavior. Therefor the correlation between TLFB at baseline and at follow up is really high. It is important to do the same experiment, but with a bigger time in between the two self-‐re-‐ ports. By doing this, we will receive a better inside in the changing pattern of smoking ciga-‐ rettes of students. It would also be helpful to have a bigger sample size, to see if the effect will still be the same.
As mentioned earlier, the associations with the negative reinforcement obtained from
Variable R R2 ΔR2 B SE β t Step 1 Calming drugs S-SCQ Neg. Con. S-SCQ Pos. Reinf. S-SCQ Neg. Reinf. S-SCQ App. W. Cont. .47 .22 2.19 .15 .37 .67 -.24 .71 .37 .29 .28 .28 .33** .05 .13 .29* -.10 3.09 .41 1.28 2.39 -.88 Step 2 Calming drugs S-SCQ Neg. Con. S-SCQ Pos. Reinf. S-SCQ Neg. Reinf. S-SCQ App. W. Cont. Impl. Ass. .51 .26† .04† 2.12 .25 .29 .60 -‐.22 2.57 .70 .37 .29 .28 .27 1.30 .32** .07 .10 .26* -‐.10 .20 3.03 .67 1.00 2.15 -‐.83 1.98