Project Towards No Tobacco Use:

1-Year Behavior Outcomes

Steve Sussman, PhD, Clyde W. Dent, PhD, Alan W. Stacy, PhD, Ping Sun,MS, Sande Craig, Thomas R Simon, BA, Dee Burton, PhD, and Brian RFlay, DPhil

IntroductionThis paper presents the 1-year behav-

ioral outcomes data from the preventionstudy of Project Towards No TobaccoUse (Project TNT), a 5-year school-basedtobacco use prevention and cessationproject.' The prevention study involves abetween-groups experiment with 48southern Califomia junior high schools.The main project objective is to determinewhich of the three most common compo-nents of social influence programs is themost effective in preventing the use of to-bacco-containing products. The two to-bacco products on which this project fo-cused are cigarettes and smokelesstobacco.

Since the passage of the Comprehen-sive Smokeless Tobacco Health Educa-tion Act of 1986 (Public Law 99-252), thedangers of using smokeless tobacco havebecome widely known to the general pub-lic. Yet use of this product has become arapidly developing health problem amongteenage males,2,3 who are now tending tofavor it over the use of cigarettes.4 Thus,it is imperative for us to understand how toprevent the use of smokeless tobacco.Only very recently have prevention pro-grams been developed to counter use ofboth smokeless tobacco and cigarettes;however, the generalizability of effectsfrom one tobacco product to the other isuncertain.' More research is needed todiscem whether different prevention pro-gram components exert unique effects indeterring the use of these two tobaccoproducts.

Social Influences and AdolescentTobacco Use

Social influences are among the mostimportant determinants of adolescent cig-

arette and smokeless tobacco use, andmost such use begins in a peer groupcontext.5-9 There are two main types ofsocial influence that may facilitate adoles-cent tobacco use: normative and informa-tional.'0 Nornative social influence refersto pressure applied by the peer group tomake youth act in ways to achieve groupacceptance. Often, this type of influence ispresent in those situations in which youngpeople are confronted with offers to usetobacco products. If these youth yield tosuch offers, theymaybe or perceive them-selves being accepted by the group. Iftheydo not yield to such offers, they may berejected in some way by the group or theymay perceive themselves as being re-jected.

Informational social influence refersto more covert pressure applied to makeyoung people adopt social values favor-able to tobacco use. Such values may beacquired from statements made by thepeer group, from tobacco advertising, orfrom other social sources (e.g., parents,movies, music videos). These sources ofinformation suggest that use of tobaccoproducts is widespread or will help thetobacco user achieve a desired social im-age, such as looking older or rebellious."IMeasures of these two types of social in-

Steve Sussman, Clyde W. Dent, Alan W.Stacy, Ping Sun, Sande Craig, and Thomas R.Simon are with the Institute for Health Promo-tion and Disease Prevention Research, Univer-sity of Southern California-Alhambra. DeeBurton and Brian R. Flay are with the Preven-tion Research Center, University of Illinois-Chicago.

Requests for reprints should be sent toSteve Sussman, PhD, IPR-USC, 1000 S Fre-mont Ave, Suite 641, Alhambra, CA 91803-1358.

This paper was accepted May 25, 1993.

American Journal of Public Health 1245

Susama de aL

fluence are empirically separable in basicsocial psychology research and in appliedresearch on the prediction of adolescenttobacco use.9,10"12,13 (A detailed discus-sion of these two types of social influencein the context oftobacco use prevention isprovided elsewhere.9)

Comprehensive Social InfluencePrevention Programs

The social influence prevention ap-proach posits that resistance to using to-bacco will be greater if one has developedan awareness of, and skills that counter-act, social pressures to use tobacco.14-17Cigarette smoking prevention programsthat focus on teaching strategies to coun-teract social influences have successfullyreduced the onset of adolescent smokingby as much as 50% 3 years postpro-gram.'15"8 But although these findings areencouraging, it is not known which ypesof activities exert the greatest preventiveeffects.'9

Comprehensive prevention pro-grams often are described as being com-posed of "components": sets of activitiesdesigned to counteract the effects of a hy-pothesized cause of an unhealthy behav-ior.20,21 Current comprehensive tobaCcouse social influence programs generallyare composed of three prevention compo-nents. One component consists of activi-ties that counteract normative social in-fluence to use tobacco. The most widelyimplemented activity within this compo-nent involves provision of refusal asser-tion skills. Although this is generallythought tobe the most effective single pre-vention activity, small sample sizes ofstudies evaluating its effects limit the po-tential generalizability of those effects.19Recent studies suggest that thesuccessfulcounteraction of normative social influ-ence is due to manipulation of peer disap-proval oftobacco use in the classroom andnot to instruction in refusal assertionskills.22n23

A second prevention componentconsists of activities that counteract infor-mational social influence to use tobacco.Such activities include provision of infor-mation about modeling and advertisingtactics, and correction of inflated tobaccouse prevalence estimates.19'24 In addition,instruction in effective communicationskills has been used to facilitate acquisi-tion of accurate social information.25 Thiscomponent attempts to help students eval-uate and use effectively various socialsources of information.

Finally, a third prevention compo-nent consists of activities that counteract

misperceptions or a lack of knowledge re-garding the physical consequences of to-bacco use. Although not a social influ-ence-type component, information aboutrisk and the severity of long- and short-term physical consequences of tobaccouse is often provided by social influenceprograms, and researchers report that pro-viding information about short-term con-sequences may prevent tobacco use.15,26'27

Development ofEmpincally DistinctCurricula

One way to compare the effective-ness of these three components of socialinfluence prevention programs is to en-gage in a priori development of curriculathat address them separately.20'2'

Four curricula were developed inProject Towards No Tobacco Use. Threewere designed to counteract the effects ofthe separate (single) componentsjust men-tioned, whereas a fourth curriculum wasdesigned to counteract the effects of com-bined social and physical consequences-related influences. The process of derivingthe current curricula from these three pre-vention components is descrnibed in detailelsewhere.21'28329 In brief, the curriculawere formed through use of an empiricallybased building-block method. After sev-eral preliminary tests of single activitieswere completed, sets of 10 lessons wereconstructed and then combined into cur-ricula so that earlier lessons would moti-vate involvement in the curricula and laterlessons would motivate a commitment notto use tobacco products. Each curriculumwas piloted to maximize its feasibility, per-ceived efficacy, and effects on behavioralintentions to use tobacco products. Knowl-edge items were developed and examinedas manipulation checks ofeffects ofthe dif-ferent activities.

One-Year Outcomes Evaluation

An outcomes evaluation considersthe effectiveness of a program in achiev-ing immediate and long-term goals.3-32The main long-term goal of an evaluationin the present context is to assess pre-ventive effects on tobacco use. Behav-ioral effects among adolescents are ex-amined at least 1 year after a program isimplemented because detectable changesin behavior usually demand this time lagor longer.'5 The present study examinedthe effects of the four different curriculaon changes in cigarette and smokeless to-bacco use behavior over a 1-year post-program interval.

MethodExperimental Design

A five-group randomized block de-sign was used.33'34 Schools were ran-domly assigned within blocks defined byregion (urban, rural), school type (middleschool with sixth through eighth grades,junior high school with only sevenththrough eighth grades), and a compositevariable. The composite variable wascomposed of a linear composite of schoolsize, socioeconomic status (Aid to Fami-lies with Dependent Children percentilerank, English as a second language per-centile rank, median income in zip code),academic status (California AssessmentProgram reading, writing, and math per-centile ranks), demographic variables(percentage White, percentage growth ofpopulation in zip code, county name, me-dian age in zip code), and estimate of to-bacco use prevalence (based on schoolstaff estimates and pilot data collected atthe school).

Forty-eight junior high schools from27 southern California school districtswere recruited and randomly assigned toparticipate in one offive conditions. Therewere 8 schools assigned to each ofthe fourprogram conditions (i.e., curricula), and16 schools assigned to a "standard" cur-riculum control condition. All seventh-grade students at the program schools re-ceived tobacco programming. Studentswere evaluatedbytwo sampling methods.In cohort 1, all seventh-grade students at20 of the schools were surveyed and fol-lowed as an individual-level collection. Incohort 2, students from the remaining 28schools were surveyed as repeated cross-sectional partial samples (approximatelythree randomly selected classes perschool).32 In each ofthe fourprogram con-ditions, 4 schools were urban, 4 were ru-ral; also, 4 involved individual-level col-lection whereas the other 4 schools wereinvolved in a repeated cross-sectional col-lection. Region by collection variableswas also fully crossed and balanced: in thecontrol condition, 8 schoolswere urban, 8were rural, 4 involved individual-level col-lection (half rural/half urban), and 12 in-volved a repeated cross-sectional collec-tion (half rural/half urban).

Students in the control condition re-ceived routine prevention activities pro-vided directly by their school. Controlschool activities were generally limited toassemblies that presented values clarifica-tion material, long-term physical conse-quences information, or simple "just sayno" to drug use messages. Aside from two

1246 American Journal of Public Health September 1993, Vol. 83, No. 9

Project Towards No Tobco Use

lessons in the health education classes,which provided information about to-bacco products and long-term conse-quences, control schools did not provideprogramming specifically for tobacco useprevention. However, 5 of the 16 controlschools did provide in-class lessons ongeneral drug use prevention.

SubjectsStudent posttest data were collected

the next school day after completion of a10-day curriculum. Data were collectedfrom 6716 seventh-grade students, 50% ofwhom were male. Regarding ethnic com-position, 60%o were White, 27% were His-panic, 7%were Black, and6% were Asianor "Other." School districts selected forparticipation in this research were re-stricted to be majority White by design.One-year follow-up data were collectedfrom 7052 students. Ninety-three percentof the students reported attending thesame school 1 year earlier. Data were ag-gregated to the school level at each timepoint using the total school sample.

Questionnaires and Data CollectionOn the days immediately prior to and

following the 2 weeks of curriculum de-livery, students in the program conditionswere administered an in-class, 20-pageself-report questionnaire. This same ques-tionnaire was used in the control schools,collected 2 weeks apart. The question-naire included a core section at the front,which contained items that assessed de-mographic and behavioral information.The same questionnaire was administeredagain 1 year later.

Students from the individual-levelcohort had breath or saliva samples col-lected, and theywere read a script inform-ing them that their data were confidential.Students from the grade-level cohort didnot provide biological samples, and theywere read a script informing them that thedata collected were anonymous. Both ofthese procedures have been shown to in-crease the accuracy of self-reported to-bacco use.35 3 The biological sampleswere not biochemically analyzed becauseof cost and incomplete collection.

Four behavioral outcome measureswere examined: "Haveyou ever tried cig-arettes?" (yes/no); "Have you ever triedsmokeless tobacco?" (yes/no); "How of-ten do you smoke cigarettes?" (weeklyuse = a few times each month or more);and "How often do you use smokelesstobacco?" (weekly use = a few timeseach month or more). Fonns ofsmokelesstobaccowere descnbed on the cover page

of the questionnaire, which the data col-lector reviewed for the class. Studentswere instructed to code chewing tobaccoand snuff as forms of smokeless tobacco.

Implemnentaton, ProcessEvaluation, and Posttest Knowledge

To maximize implementation,trained project health educators deliveredthe four curricula over 10 consecutiveschool days. Previous analysis has shownthat implementation data and process rat-ings were favorable and similar across thefour program conditions, and knowledgescores indicated discriminant validityamong conditions.29 This pattem of re-sults permitted future behavioral outcomedifferences among conditions to be attnb-uted to differences pfimarily in the contentof material provided rather than to differ-ences in the quality of delivery3O or in cur-riculum credibility.31 A detailed descrip-tion ofthe curricula contents canbe foundelsewhere.29 The types of material pro-vided reflect the components of compre-hensive social influence programs.

AnaysisOf principle interest in this study is

the change in prevalence of tobacco useover the 1-year study interval. We con-structed four behavioral change outcomemeasures: two measured the change inprevalence of initial trial of cigarettes andsmokeless tobacco, and two indicated thechange in prevalence of weekly use ofthese substances. Measures were con-structed at the school level by first calcu-lating the prevalence estimates of tobaccouse at each school at immediate postpro-gram and 1-year follow-up time points. Anaverage of the survey results before andimmediately following the survey wasused as the initial time-point estimate.(lTere was little change in behavior overthat 2-week interval.)A change score wasthen calculated by subtracting initial prev-alence from follow-up. The number ofsubjects measured at each school wasused as aweighing factor in all subsequentcalculations (i.e., small schools contnb-uted less weight).

The first thing to be examined wasthe prevalence of tobacco use by genderand region. For each time point and for thedifference, two-group t tests between re-gions (urban vs rural) and paired t testsbetween genders (male vs female) werecalculated, with school as the unit of anal-ysis. In the analysis of gender compari-sons, each school contributed a separatemean for its males and females. For urban

versus rural comparisons, each schoolcontnrbuted a single mean.

Examined next were the effects ofthe four curricula on each of the tobaccooutcome measures. This was done bycomputing a five-group, one-way analysisof covariance (ANCOVA) model withschool as the unit of analysis. Covariatesincluded gender, the blocking assignmentvariables (composite, school type, re-gion), and a measure of school "turn-over." Turnover was defined as the pro-portion of new students at the school at1-year follow-up (average value = 7%),and itwas added to the model to adjust forpossible differences in student composi-tion at follow-up.

Use of the ANCOVA model at theschool level entailed our assuming that theeffects of the intervention are additive inthe scale ofchanges in school proportions(i.e., that each school in a group respondsthe same way, on average, to treatment)and that rate of change without any sys-tematic intervention can be characterizedas a constant that is estimated by the con-trol schools. To test this assumption, in-teraction terms between covariates andtreatment were calculated. None of theseterms was significant, and they were sub-sequently dropped from the model. Thisset of results suggests that our assump-tions are tenable and that any discerniblepattem of treatment effects may be con-sidered to be the same across levels de-fined by the covariates, regardless of thepresence of main effects of gender and re-gion. Therefore, our planned mean differ-ence comparisons between the four treat-ment group means and the control groupmean collapsing across gender and regionwere calculated using the ANCOVAmodel error term. Additionally, usingTukeys least significant difference criteriafor multiple comparisons, all possiblepairs oftreatment group meanswere com-pared post hoc to determine if treatmentsdominated each other.

ResuWithin-Tine Prevalence by Genderand Region

Asshown inTable 1, immediate post-test and 1-year follow-up cigarette use didnot differ by gender; however, trial andweekly use of smokeless tobacco use diddiffer. Use of smokeless tobacco washigher among males. Regarding region,trial of cigarette smoking and smokelesstobaccousewas higherin the rural schoolsthan in the urban schools at both time

Amencan Journal of Public Health 1247September 1993, Vol. 83, No. 9

Sun et aL

points. Weekly use ofthe tobacco productdid not differ by region.

Change in Prevalence by Genderand Region

From immediate posttest to the1-year follow-up, trial and weekly useprevalence of cigarettes rose about 8%and 3%, respectively-an equal percent-age among males and females. On theother hand, trial and weekly use preva-lence of smokeless tobacco rose only formales (5% change for trial use and 1%change for weekly use).

Cigarette use prevalence rose equallyacross urban and rural regions (about 8%and 3% change regarding trial and regularuse, respectively). Trial of smokeless to-bacco use rose about 3% at urban schoolsand about 1% at rural schools, whereas

weekly use increased only about 0.5% atboth regions, which did not differ fromeach other.

Change in Prevalence by ConditionChanges in prevalence data compar-

ingthe prgram and control conditions areshown in Table 2. For both trial andweeklyuse ofcigarettes, the informationalsocial influence, physical consequences,and combined conditions (which did notdiffer from each other) were superior tothe normative social influence and controlconditions (which did not differ from eachother). For trial of smokeless tobacco, allprogram conditions except for the infor-mational social influence condition weresuperior to the control condition. Finally,for weekly use of smokeless tobacco, thecombined condition was superior to all

other program conditions and the controlcondition (which did not differ from eachother).

DisrussionThese data suggest that (1) physical

consequences information can be used tocompose a curriculum that is as successfulas a social influence program; (2) a nor-mative social influence program is not asefficacious for tobacco use prevention asare other types of programs, except forprevention of smokeless tobacco onset;and (3) the same tobacco use preventionprogramming can be successful when tar-geted to use ofboth cigarettes and smoke-less tobacco in the school-based context.

The physical consequences condi-tion was as efficacious as the social influ-ence programs in most comparisons. Thispattern of results may contradict previousresearch, which found that social influ-ence programming ismore successful thanphysical consequences programmng.18,37On the other hand, the present physicalconsequences curriculum included sev-eral novel features such as correctingmyths about tobacco experimentation andaddiction,26 role-playing diseaseS,31'M andpresenting probabilities of consequencesinformation in ways more personally rel-evant to youth.39 Previous physical Con-sequences curricula focused more on thedidactic presentation of long-term conse-quences. Although generally not mea-sured in previous studies, social influenceprevention programming may have beenmore successful in earlier studies becauseit was better received by students thantraditional approaches.31

The data regarding the normative so-cial influence condition suggest that teach-ing refusal assertion skills and facts aboutclassmate peer disapproval oftobacco usewere ineffective in this sample (aside fromprevention of trial of smokeless tobacco).A previous analysis has revealed that pro-cess ratingswere slightly lower in this con-dition.29 Anecdotally, health educatorsand students reported that the "flooding"of their school systems and homes withnormative social influence-type programs(e.g., red ibbonweek, "just sayno" cam-paign public service announcements) mayhave led to a lack of excitement over ad-ditional normative social influence-typeinformation, which may have affected theresults obtained. Nonetheless, knowledgeitems in the survey indicated that studentsin the normative social influence conditionleamed more about this type of informa-tion than did students in the other condi-

1248 American Journal of Public Health September 1993, Vol. 83, No. 9

tions. Thus, there is no ceiling effect ontheir learning of this type of information.

Another possibility is that the appar-ent effectiveness of refusal assertion skillstraining, the main normative social influ-ence component, in previous smokingprevention programs may have been dueto the confounding of training in refusalassertion skllls with other social influenceprogram components. Most social influ-ence programs that have taught refusal as-sertion skills also have taught some as-pects of informational social influence. Inthe present project, the informational so-cial influence condition had stronger ef-fects than the normative social influencecondition on smoking, consistentwith thisinterpretation. This findingmay imply thatteaching students how to interpret socialsources of information more effectively ismore important than teaching specific re-fusal assertion skldls (or facts about class-room levels of peer disapproval for usingtobacco products).

Another way of interpreting differ-ences in effects obtained between the nor-mative and informational social influenceconditions is based on the observation thatthe informational social influence condi-tion exerted an effect on cigarette smokingbut not on smokeless tobacco use, andthat the normative social influence condi-tion exerted an effect on smokeless to-bacco use (trial behavior) but not on cig-arette smoking. This finding convergeswith some pilot study data, which had in-dicated effects ofrefusal assertion trainingon intention to use smokeless tobacco butnot on intention to smoke cigarettes.40One may conjecture that teaching refusalassertion skills is effective for preventingadolescents from trying smokeless to-bacco because normative social influencepressures are more prevalent for use ofthis substance (e.g., related to dares to usea new substance), whereas informationalsocial influence pressures (e.g., magazineadvertising) are more prevalent for ciga-rettes. Alternatively, perhaps simplylearning that one should sayno to offers ofsmokeless tobacco is perceived as novelto adolescents, whereas the same instruc-tion is perceived as being "nothing new"regarding cigarettes. Future researchshould be conducted to examine these al-ternative possibilities.

Generally, the combined conditionwas the most efficacious. This conditionwas the onlyone that showed a preventiveeffect onweekly use of smokeless tobaccooas well as on the other three behavioraloutcomes. We did not expect to achieve arelative superiority of the combined con-

dition. As we reported previously,29greater teaching effortwas required to im-part this thematically diverse material.Also, it was fairly likely that curriculummaterial would be modified during imple-mentation, possibly owing to attempts toprovide a theoretical integration ofthe dif-ferent perspectives. Finally, knowledgelearned from the combined curriculumwas relatively likely to be diluted; in otherwords, students were likely to learn lessabout more material.29 Although each ofthese effects of a heterogeneous curricu-lum conceivably could have detrimentaleffects on program success, the heteroge-neity of the combined curriculum insteadled to the strongest overall effects.

The overall predictive superiority ofthe combined condition might imply thatdifferent causes of tobacco use need to becounteracted simultaneously because thebehavior is determined by multiplecauses. Also, a heterogeneous programmay reach a wider variety of youth, whomay differ in risk factors that influenceuse. One might conclude that currentcomprehensive social influence program-ming should be continued and emphasizedover the development of isolated programcomponents. On the other hand, asidefrom effects onweekly smokeless tobaccouse, single-component program condi-tionswere found to be superior to the con-trol condition and not to differ in effectfrom the combined condition. An empha-sis on effective deliveiy of any of thesecurricula is warranted. Implementation,as well as content, is an appropriate focusof future research?3 0

AcknowledgmentsSupport for this research was provided bygrants from the National Cancer Institute(CA44907) and National Institute on DrugAbuse (DA07601).

References1. Boyd GM, Glover ED. Smokeless tobacco

use by youth in the U.S. J Sch Health.198859:189-194.

2. Connolly GN, Winn CM, Hecht SS, Hen-ningfield JE, Walker B, Hoffin D. Thereemergence of smokeless tobacco. NewEngiJMed 1986;314:1020-1026.

3. US Surgeon General. The Health Conse-quences ofUsing Smokeless Tobacco. Be-thesda, MD: US Dept of Health and Hu-man Services; 1986.

4. Hunter SM, Croft JB, Burke GL, ParkerFC, Webber LS, Berenson GS. Longitu-dinal patterns of cigarette smoking andsmokeless tobacco use in youth: the BoPa-lusa Heart Study. Am J FPub& Health.1986;76:193-195.

5. Bewley BR, Bland JM, Harrs R. Factorsassociated with the starting of cigarette

PoJet Towards No Tobscw Use

smoking by primary school children. BrJPrev Soc Med 1974;28:37-44.

6. EiserJR. Smoking: the social learningof anaddiction. JSoc Clin PsychoL 1985;3:446457.

7. Friedman IS, Lichtenstein E, Biglan A.Smoldng onset among teens: an empiricalanalysis of initial situations.AddictBehav.1985;10:1-13.

8. Hahn G, Charlin V, Sussman S, et al. Ad-olescents' first and last use situations ofsmokeless tobacco and cigarettes: similar-ities and differences. Addict Behav. 1990;15:439-448.

9. Sussman S. TIwo social influence perspec-tives of tobacco use development and pre-vention. Health Educ Res: Theory andPrct. 1989;4:213-223.

10. Deutsch M, Gerard HB. A study of nor-mative and informational social influencesupon individual judgment. JAbnonn SocPsychoL 1955;51:629-636.

11. Burton D, Sussman S, Hansen WB,Johnson CA, Flay BR. Image attnrbutionsand smoking among seventh grade stu-dents. JAppi Soc PsychoL 1989;19:656-664.

12. GrahamJW, Marks G, HansenWB. Socialinfluence processes affecting adolescentsubstance use. J Appl PsychoL 1991;76:291-298.

13. Stacy AW, Sussman S, Hansen WB,Johnson CA, Flay BR. Social influencevariables as predictors of later cigarettesmoking. Unpublished manuscript, 1991.(Available on request from first author.)

14. Evans RI. Smoking in children: developinga social psychological strategy of deter-rence. Prev Med. 1976;5:122-127.

15. Flay BR. Psychosocial approaches tosmoking prevention: a review of findings.Health PsychoL 1985;4(5):449488.

16. McGuire WJ. Inducing resistance to per-suasion. In: Berkowitz L, ed.Advances inF&pernmental Social Psychology. NewYork, NY: Academic Press; 1964;1:191-229.

17. Silvestri B, Flay BR. Smoking education:comparison ofpractice and state-of-the-art.P)'ev Med. 1989;18:257-266.

18. Tobler NS. Meta-analysis of 143 adoles-cent drug prevention programs: quantita-tive outcomes results of program partici-pants compared to a control or comparisongroup. JDrugIssues. 1986;16:535-567.

19. McCaul KI, Glasgow RE. Preventing ad-olescent smoking: what have we learnedabout treatment construct validity? HealthPsychoL 1985;4:361-387.

20. Chassin LA, Presson CC, Sherman SJ.Stepping backward in order to step for-ward: an acquisition-oriented approach toprimaiy prevention. J Consult Clin Psy-choL 1985;53:612-622.

21. Sussman S. Curriculum development inschool-based prevention research. HealthEduc Res: Theory and Pruct. 1991;6:339-351.

22. MacKinnon DP, Johnson CA, Pentz MA,et al. Mediating mechanisms in a school-based drug prevention program: first yeareffects of the Midwestern Prevention Proj-ect.HealthPsco 1991;10:164-172.

23. Hansen WB, Graha JW. Peer pressureresistance trainingversus establishing con-servative social norms to prevent sub-

September 1993, Vol. 83, No. 9 American Journal of Public Health 1249

Suuman et aL

stance use onset. Poster presentation at the12th Annual Scientific Sessions of the So-ciety of Behavioral Medicine; March 20-24, 1991; Washington, DC.

24. Sussman S, Dent CW, Mestel-Rauch J,Johnson CA, Hansen WB, Flay BR. Ado-lescent nonsmokers, trials, and regularsmokers' estimates of cigarette smokingprevalence: when do overestimations oc-cur and by whom? JApp Soc PsychoL1988;18:537-551.

25. Botvin GJ, Eng A. Life SkiJts Tranng. AC hmp ive Smoin Pvnion PFo-gm. New York, NY: Smithfield Press;1979.

26. GlynnK,LevanthalH,H man R.Acog-nitive developmental approach to smokingprevention. In: Bell CS, Battes R, eds. Pre-vention Researck DetmTig Dn4g Abuseamong rhibdnandAdoescents. Rockvlle,Md: National Institute on DrugAbuse; 1985.NIDA Research Monograph 63.

27. Johnson CA. Untested and erroneous as-sumptions underlying anti-smoling pro-grams. In: Coates T, Peterson A, Perry C,eds. Pomotion ofHealth in Youth. NewYork, NY: Academic Press; 1981;149-165.

28. Sussman S, Galaif J, Charin V, et al. TATPJWventionDevelopment. Alhambra, Calif:UniversityofSouthern California, Institutefor Health Promotion and Disease Preven-tion Research; 1989. Technical report 89-TNT403.

29. Sussman S, Dent CW, Stacy AW, Hodg-son CS, Burton D, Flay BR. Implementa-tion, process, and immediate outcomeevaluation ofProjectTowards NoTobaccoUse. Health Educ Res: Theory and Pract.1993;8(1):109-123.

30. Pentz MA, Trebow EA, Hansen WB, et al.Effects of program implementation on ad-olescent drug use behavior. Eval Rev.1990;14:264-289.

31. Sussman S, Dent CW, Brannon BR, et al.The television, school and family smokingprevention/cessation project: IV. control-ling for program success expectanciesacross experimental and control condi-tions. Addict Behav. 1989;14:601-610.

32. Dent CW, Sussman S. A comparison ofcomplete and partial sampling in repeatedlyaccessible populations of naturally clus-tered units. Presented at the Symposium onStatistical Methods for Evaluation of Inter-vention and Prevention Strategies, Decem-ber 4-6, 1990; Atlanta, Ga.

33. Craig S, Dent CW, Sussman S. School Se-lecton and Recnidnent. Phase 1: OtherMajor Tasks, Project Towards No To-bacco Use (7NT). Alhambra, Calif: Uni-versity of Southern California, Institute forHealth Promotion and Disease PreventionResearch; 1988. Technical report 88-TNT-01.

34. Dent CW. RandomAs_ignment in ProjectTNT and Data Colecton Procedwes. Al-hambra, Calif: University of Southern Cal-ifornia, Institute for Health Promotion andDisease Prevention Research; 1989. Tech-nical report 89-TNT-01.

35. Bauman KE, Dent CW. Influence of anobjective measure on self-reports ofbehav-ior. JAppi PsychoL 1982;67:623-628.

36. Murray DM, Perry CL. The measurementof substance use among adolescents: whenis the "bogus pipeline" method needed?Addict Behav. 1987;12:225-233.

37. Thompson EL. Smoking education pro-grams, 1960-1976. Am J Ptublic Health1978;68:250-257.

38. Brannon BR, Dent C, Flay BR, et al. Thetelevision, school, and family project: V.the impact of curriculum delivery formaton program acceptance. Pnrv Med. 1989;18:492-502.

39. Sussman S, Dent CW, Flay BR, et al. Me-dia manipulation of adolescents' personalleveljudginents regarding consequences ofsmokelesstobaccouse.JDn4gEduc. 1989;19:43-57.

40. Turner GE, Burciaga C, Sussman S, et al.Which lesson components mediate refusalassertion skill improvement in school-based adolescent tobacco use prevention?IntJAddict. 1993;28:749-766.

1250 Amencan Joumal of Public Health September 1993, Vol. 83, No. 9