Theory, measurement, and methods in the study of family influences on adolescent smoking

REVIEW

copy 2003 Society for the Study of Addiction to Alcohol and Other Drugs

Addiction

98

(Suppl 1) 21ndash36

Blackwell Science Ltd

Oxford UK

ADDAddiction

1359-6357copy 2003 Society for the Study of Addiction to Alcohol and Other Drugs

98

Supplement 12136

Original Article

Theory measurement and methodsNancy

Darling amp Patricio Cumsille

Correspondence to

Nancy DarlingProgram in PsychologyBard CollegeAnnandale-on-Hudson NY 12504USA

E-mail darlingbardedu

RESEARCH REPORT

Theory measurement and methods in the study of family influences on adolescent smoking

Nancy Darling

1

amp Patricio Cumsille

2

1

Program in Psychology Bard College NY USA

2

Escuela de Psicologiacutea Universidad Catoacutelica de Chile Chile

ABSTRACT

This paper discusses three dilemmas faced by researchers interested in familyinfluences in substance use the transitional nature of adolescent smoking thecomplexity and multi-dimensionality of family forms and influences and theinter-relationship of family influences with other key developmental contextsMethodological and conceptual issues stemming from these dilemmas are dis-cussed with regard to understanding why previous reviews have found the cor-relations between family predictors and adolescent smoking to be relatively lowIn particular the importance of understanding time the transitional nature ofthe phenomenon and within- and between- family processes are emphasizedMore appropriate conceptual and statistical models for analyzing family influ-ences are suggested including both mixed models and person-centeredapproaches

KEYWORDS

Adolescence family methodology parental monitoring

parenting substance use smoking tobacco use

INTRODUCTION

Although researchers interested in the development ofadolescent smoking have long been interested in familyinfluences the literature has been less than clear aboutthe strength of relationships (described as lsquounexpect-edly lowrsquo by Conrad Flay amp Hill (1992)) or the key pro-cesses through which they operate Researchersinterested in studying family influences face three keydilemmas The first is the nature of adolescent smokingMost individuals enter adolescence as non-smokersMany begin smoking during their adolescent yearsSome of those who begin to smoke quit never to smokeagain while others go on to become committed long-term smokers In other words smoking during adoles-cence is best characterized by change (Petraitis

et al

1995) On the other hand many of the family charac-teristics used to predict adolescent tobacco use (egfamily bonding parent smoking parent attitudestoward adolescent smoking parental monitoring) arerelatively stable characteristics Thus the dilemma howdoes one conceptualize the processes that relate stabil-ity to change

The second dilemma researchers interested in familyinfluences on adolescent smoking must grapple with isthe nature of the family itself Although we speak com-monly of lsquothe family unitrsquo within-family influences can-not accurately be conceptualized as singular Insteadmothers fathers siblings and other family memberseach influence adolescents in different ways Systemicproperties of the family operate as well (Whitchurch ampConstantine 1993) For example it is possible that an ado-lescent might have a greater tendency to begin smoking ina household where her mother father and only siblingsmoke than would be predicted from the simple additiveinfluences of each family member The confluence of influ-ences may make conformity to family norms more salient(Bronfenbrenner amp Morris 1998) Alternatively a thresh-old effect may operate such that a single smoking familymember increases adolescent risk to the same extent asthree In other words modeling family influences on ado-lescent smoking is complex both because of the number ofpotential sources of influence within the family andbecause families have systemic properties over and abovethe influence of the individuals that compose them Mod-eling family influences is complicated further by two


Addiction

98

(Suppl 1) 21ndash36

22

Nancy Darling amp Patricio Cumsille

additional characteristics of families First families comein many diverse forms Obviously we would not wantmodeling maternal and paternal smoking as separatevariables to result in all adolescents living in single parentfamilies eliminated from the analyses because of missingdata How should this problem be best addressed The sec-ond problem is more subtle Family member characteris-tics are inherently interdependent posing problems fortraditional statistical techniques For example if weassume that parent and offspring tobacco use are corre-lated how appropriate is it to model sibling influence onadolescent smoking by controlling parent tobacco useThese latter two concerns make use of traditional statis-tical techniques problematic

The third dilemma that researchers interested in theinfluence of the family on adolescent smoking must grap-ple with is the nature of the family

during adolescence

Although we have moved beyond thinking primarily ofthe family as a place from which adolescents move out(Steinberg 1990) it is still true that a key characteristic ofadolescence is the expansion of the social world Adoles-cents spend more time in the company of peers and lesstime with family members than do children (Larson ampRichards 1991 Larson

et al

1996) In additionalthough parental influence remains strong especially inareas involving values and long-term goals (Steinberg1990) peer influence peaks during mid-adolescenceespecially in day-to-day activities (Berndt 1979) Thusthe relationship between the family and other key con-texts should be of special interest to researchers interestedin adolescent tobacco use How do families influence thenature of other contexts to which adolescents areexposed How do they influence the processes that occurin those other contexts How are they influenced by out-side contexts How given the complex nature of the fam-ily can such interrelationships be modeled Similarlynormative changes in both contexts and behavioralexpectations allow greater expression of individual differ-ences in temperament genetic predispositions andlearned attitudes and behaviors in adolescence than inchildhood (Scarr amp McCartney 1983) How do familycharacteristics inhibit or facilitate the expression of suchindividual differences

This paper has two goals to examine each of theseissues as it relates to developmental patterns of adolescentsmoking trajectories and to discuss potential strategiesfor modeling family processes

STATUS STAGE TRANSITION AND TRAJECTORY

Recent reviews of the literature have highlighted the dis-tinction between examining adolescent smoking

status

and examining the development of smoking

trajectories

(Petraitis

et al

1995 Mayhew

et al

2000) Such distinc-tions are critical in understanding and modeling accu-rately the relationship between family characteristics andadolescent tobacco use Status refers to the current cate-gorization of an individual according to the variable ofinterest in this case tobacco use Such statuses may be assimple as lsquocurrent smokerrsquo versus lsquocurrent non-smokerrsquoor they may include an explicit time element for examplelsquonever tried tobaccorsquo versus lsquotried tobaccorsquo Other statuscategorization schemes include an implicit time elementFor example Mayhew

et al

(2000) explicate four lsquostagesrsquoof tobacco use Individuals are labeled lsquotriersrsquo lsquoexperi-mentersrsquo lsquoregular usersrsquo and lsquodependent usersrsquo Implicitin the stage model is the idea that a snapshot of an indi-vidual at any given time tells us something about theirhistorical tobacco use Just as we assume that the middle-aged individual pictured in a photograph was once achild the explicated stages imply that a person who iscaptured in the status of lsquoregular userrsquo has gone throughthe stages of lsquotrierrsquo and lsquoexperimenterrsquo(Petraitis

et al

1995)

1

Current tobacco use status is limited in what it tells usabout time however because it captures only a snapshotof a particular temporal moment In contrast the conceptof lsquotransitionrsquo incorporates time explicitly but incom-pletely A transition is defined by change over time forexample by a move from status 1 (eg lsquonever triedrsquo) to sta-tus 2 (lsquoexperimenterrsquo) Transitions are not identical totime however because the unit of analysis is

change

rather than

time

For example a person who is a non-smoker at both time 1 and time 2 cannot be said to havemade a transition even though time has passed Simi-larly two people both of whom have moved from status 1to status 2 may have existed in the statuses for differentperiods of time For example person 1 may have been instatus 1 at measurement points 1 2 and 3 and moved tostatus 2 at measurement point 4 while person 2 mayhave been at status 1 at measurement point 1 but status2 at measurement point 2 Although we can look at thepredictors of a transition without regard to time (forexample we can try to differentiate people who never

1

A lsquonever usedrsquo category is not included in these stages Also fre-quently neglected are statuses that capture current levels thatare lower than past use For example an individual who hadbeen a heavy smoker but now smokes only occasionally wouldpresumably be at a different stage of tobacco use than someonewho currently smokes occasionally but had never shown anyother pattern This distinction highlights the limited way inwhich time is reflected in categorization schemes based on cur-rent status Most stage models of substance use appear to bebased on lsquostrong stagersquo assumptions including unidirectionalityand irreversibility (Wohlwill 1973) Tobacco use histories how-ever are not consistent with these assumptions

Theory measurement and methods

23


Addiction

98

(Suppl 1) 21ndash36

smoke from those who make the transition however fastthat transition occurs) the comparison group of peoplewho have not made the transition always implies the ele-ment

within a given time frame

Individuals included in thenon-smoking status may make a transition to smoking atsome time in the future

Transitions then are quite different from

trajectories

which are defined by

time

rather than by

change

2

Becausethe unit is time change can take a value of 0 so stabilityis one potential trajectory The minimum number of time-points needed to capture a trajectory is two

3

thus the dif-ference between smoking statuses at adjacent time pointsis change over time (ie speed of transition) With morethan two points of measurement trajectories also caninclude both measures of acceleration (changes in thespeed of transition) if change is smooth and measures ofdeflection where comparisons in trajectories are madebefore and after the temporal occurrence of a particularpredictor (divorce or the cessation of parental smokingfor example) The concept of deflection also implies thatchange can either be toward higher levels of smoking ormark a change to lower levels

Before going on to discuss trajectories two additionalpoints should be made about the interface of time withtransitions and trajectories First like transitions trajec-tories implicitly assume lsquowithin a given time framersquo Takethe example where an individual is measured at time 1time 2 and time 3 and during that time moves fromlsquoexperimenterrsquo to lsquoregular userrsquo to lsquodependent userrsquo Wecan calculate a trajectory for this individual of a move-ment of two stages within the time frame of time 1 to time3 A second individual may be observed at time 1 andtime 2 as an lsquoexperimenterrsquo and time 3 as a lsquoregular userrsquoWithin the time frame from time 1 to time 3 the secondindividualrsquos trajectory appears to show a slower progres-sion This is not necessarily the case however For exam-ple the first individual may have been an experimenter attime

-

1 (ie at a hypothetical measurement point prior tothe start of observation) but was not observed The sec-ond individual may move to lsquodependent userrsquo at time 4 Inother words both individuals may have identical trajec-tories of tobacco use but be observed at different points intheir trajectories This observation introduces additionalproblems (and error) into the modeling of tobacco use

trajectories a point to be returned to in the discussion ofsibling influences on tobacco use

Secondly tobacco use trajectories which are definedby historical or objective time interface with life coursetimemdashie with age (Elder 1998) For example one wouldpredict a different distribution of normative trajectories oftobacco use at different ages During the pre-adolescentyears one might expect that the normative trajectorywould be a flat pattern of non-use with a secondary pat-tern of non-use with a singular status change to lsquoever-triedrsquo One would also expect a small group of individualsat this age to make a rapid transition from non-user intodependence For example Jessor amp Jessor (1977) Moffitt(1993) Patterson

et al

(1989) and Pulkkinen (1990) allsuggest that a certain subset of individuals evince astrong and consistent pattern of highly intercorrelatedproblem behaviors that appear early in their lives but takesomewhat different forms depending on the developmen-tal stage of the individual Based upon their work onemight expect that individuals who during late childhoodor early adolescence show a trajectory of rapid move-ment from tobacco non-use to use would also be moreprone toward sensation-seeking and impulse controlproblems be more likely to experience dysfunction withinthe family system be associated with a deviant peergroup be socially rejected by normal peers and beengaged in a wide variety of other deviant behaviors

4

However there is no theoretical or empirical basis fromwhich to predict that the same trajectory of rapid changewould be associated with the same set of predictors at adifferent developmental periodmdashduring late adolescencefor example (Moffitt 1993) For late adolescents onemight hypothesize that particular social processes andcontext (entrance into the military for example) mightpredict a steep trajectory from non-use to use (see forexample Schei amp Sogaard 1994) Thus at different pointsin the life course we might predict different distributionsof tobacco use trajectories and we might also expect to seedifferent correlates of trajectory types These two differentissues of the interface of time and trajectory (implicit timeframe and age) taken together might help to bring orderto the sometimes divergent findings of studies relatingfamily predictors with adolescent tobacco use

4

This hypothesis about differences in the predictors of similartrajectories (and implicitly of the processes underlying thesetrajectories) at different points in the life course is consistentwith Graham

et al

rsquos (1991) finding that adolescents who beginthe transition to substance use with tobacco showed a morerapid trajectory to high levels of use than those who began withalcohol It is also consistent with Brook

et al

rsquos (1996) reportthat only predelinquency predicted smoking among youngeradolescents but that a wider range of contextual variables pre-dicted smoking among older adolescents

2

Methodologically the terms lsquotransitionrsquo and lsquotrajectoryrsquo areused respectively in latent transition analysis (LTA) to describequalitative shifts in status and in growth curve models todescribe the quantitative shifts evident in the time component ofgrowth models however it is not the intent of this paper toimply a distinction between qualitative and quantitative shiftsby this usage

3

However see Collins amp Sayer 2000) for an excellent discussionof why more and more frequent collection of longitudinal dataprovide better estimates of developmental trajectories


Addiction

98

(Suppl 1) 21ndash36

24


THE NATURE OF ADOLESCENT SMOKING PREDICTING CHANGE FROM STABILITY

Family characteristics have been used to predict adoles-centsrsquo tobacco use statuses transitions and trajectoriesDemographically adolescents from single-parent andlower SES families whose families grow tobacco andadolescents whose parents are less educated are morelikely to smoke than their peers (Tyas amp Pederson 1998)At the proximal level strong positive parentndashchild bondsauthoritative parenting higher levels of parental moni-toring and parentndashchild communication parent disap-proval of smoking and parent and sibling non-smokinghave all been shown to decrease risk of tobacco use (forreviews see Moncher

et al

1991 Conrad

et al

1992Hawkins

et al

1992 Centers for Disease Control andPrevention 1994 Avenevoli amp Merikangas 2003)

As in most areas of developmental research themajority of studies present cross-sectional findings anddemonstrate associations between family characteristicsand adolescent tobacco use status Longitudinal researchhas provided researchers with the opportunity to exam-ine both transitions and trajectories Contrasting the twotypes of studies has provided some interesting insight intothe importance of the distinction between statuses tran-sitions and trajectories Longitudinal research has alsohighlighted the differences in processes (or at least differ-ences in predictors) that are involved in moving frombeing a non-user to an experimenter and those involvedin moving from experimentation into regular use Forexample Flay

et al

(1998) found that peer factors weremore important in moving adolescents from trial toexperimental stages but that family processesmdashspecifi-cally parental smoking and family conflictsmdashpredictedtransitions from experimental to regular use In contrastChassin

et al

(1984) found that parentsrsquo smoking-specificsocialization predicted smoking cessation amongyounger adolescents but that peer processes predictedcessation among older adolescents Sibling smoking hasbeen found to predict smoking status (Avenevoli amp Meri-kangas 2003) but not transitions or changes in smoking(Spielbeger

et al

1983 Hanson

et al

1985 Ary amp Biglan1988)

Because much of the longitudinal work has focused onthe prediction of smoking status from family characteris-tics it has obscured the fact that such studies are attempt-ing to predict change from characteristics that are stableAlthough such analyses are common a more carefulexamination of their underlying assumptions is in orderWhen a stable family characteristic such as parent smok-ing is used to predict smoking status the dependent vari-able is either the end point of a trajectory (for examplewhen parent smoking is used to predict smoking status in

late adolescence) or it is capturing adolescents lsquomid-flightrsquomdashie at a particular time point within a trajectoryBecause all smoking trajectories start at the same status(non-smoker) the ability to detect differences in adoles-centsrsquo trajectories based on estimates at a particular pointdepends upon where you are in the diverging trajectoriesIf we start sufficiently early enough statuses of futuresmokers and non-smokers will be identical Thus the abil-ity of a stable predictor to differentiate adolescent trajec-tories based on smoking status depends upon howappropriate the age of the participants and the time frameof the study is as well as on statistical power and measure-ment error

In addition adolescence is a time when long-termsmoking patterns are in flux Because most individualsenter adolescence as non-smokers but most people whobecome smokers start smoking during adolescence stud-ies of adolescent smoking are inherently studies ofchange even when such change is captured as smokingstatus Although stable family characteristics may be pre-dictors of change they cannot be said to be the proximal

cause

of change To use an analogy when a rock rollsdown a hillside its movement is not initiated by the hillbut by the person who kicked it Taking the analogy onestep further one might predict that a rock on a hill willtumble down at some point in the future but its positionon the hill tells you neither when that will happen norwhat the particular event that will trigger the change willbe Knowing the rock is on a hillside however does allowyou to predict that the effect of a kick on it may be differ-ent from the effect of a kick on a rock in a field Focusingon proximal processes operating at points of transitionstable characteristics (such as parental smoking) can pre-dict change only in the presence of a triggering event(such as the offer of a cigarette by a peer)

Triggering events can be internal to the child or theycan be environmental Stable characteristics of the familycan make it more likely that a triggering event occursthey can change the likelihood that a triggering event willbe sufficient to cause change or both For example socialcognitive theory (Bandura 1989) postulates that theknowledge and ability to enact a behavior is a necessarybut not sufficient condition for the behavior to becomemanifest With regard to smoking growing up in familieswhere parents are observed smoking provides the oppor-tunity for children to learn the rudiments of smokingmdashhow to light a cigarette how to inhale what to do withashes etcmdashand virtually guarantees that smokingbecomes part of the repertoire of behaviors that the childis capable of enacting It also provides knowledge of whereand when it is appropriate to smokemdashwhile talking aftera meal drinking alcohol driving or relaxing for exampleNote that the family is not the only context in whichknowledge of smoking behaviors can be acquired (peers


25


Addiction

98

(Suppl 1) 21ndash36

and the media are obvious other sources) however it ispossible that the length and breadth of exposure to smok-ing in the family make knowledge of smoking acquired inthis setting both quantitatively and qualitatively differentfrom that acquired through other sources

The likelihood that the acquired behavior will beenacted depends upon many other factors For exampleobservers are more likely to enact learned behaviors mod-eled by attractive others who are similar to the observerand behaviors that are observed to have positive conse-quences One would expect then that smoking by same-gender parents would increase the likelihood that chil-dren smoke more than smoking by cross-gender parentsand that parents who are positive about the smokingexperience are more likely to be imitated than those whoare negative There is evidence to support each of thesehypotheses although tests have been relatively infre-quent and results have been somewhat mixed (Avenevoliamp Merikangas 2003) Learned behaviors are also morelikely to be enacted when models are attractive Withregard to smoking one would expect to see the linkbetween parent and child smoking to be strongest in thepresence of a positive parentndashchild bond Few researchershave investigated this prediction

5

Interestingly the pre-ponderance of evidence suggests that poor relationshipswithin the family and low levels of perceived parental sup-port are associated with an increased risk for tobacco use(for review see Avenevoli amp Merikangas 2003) and agreater likelihood of moving from experimental to estab-lished smoking (Distefan

et al

1998) Several theorists(Jessor amp Jessor 1977 Patterson

et al

1989) have sug-gested that adolescents who have poor relationships withparents are at increased risk for tobacco use because suchrelationships increase the likelihood that they will associ-ate with deviant peers and there is strong evidence tosupport this contention (Melby

et al

1993 Flay

et al al

1994 Biglan

et al

1995 Cohen amp Rice 1997 Frauen-glass

et al

1997 Chassin

et al

1998) Taken togetherthese theories suggest that (a) parental smoking will bemost influential when adolescents enjoy a positive rela-tionship with their parents and (b) that adolescents are

more likely to be part of a peer group that encouragessmoking when they experience poor relationships withparents In other words good relationships with parentswho smoke may make adolescents more vulnerable to atriggering event such as being offered a cigarette but lesslikely that they will be in a setting where that situationwill occur It is possible that the inconsistency in the asso-ciation between parent and child tobacco use (describedas lsquounexpectedly lowrsquo by Conrad

et al

(1992 p 1711))may be accounted for partially by characteristics of theparentndashchild relationship or observed consequences ofparental smoking that moderate the relationship betweenparent and adolescent smoking For example it seemscounterintuitive that observing a parent struggle to quitsmoking or suffer from a smoking-related illness wouldincrease the likelihood of tobacco use

In addition to illustrating the importance of examin-ing moderating effects in understanding family processes(a point brought out again later in this paper) the previ-ous discussion of social cognitive theory also illustratesthree mechanisms through which stable family charac-teristics can predict change in behavior directly throughchild characteristics that (1) increase their exposure (2)increase their vulnerability to triggering events or (3)change the childrsquos environment in such a way that itincreases the likelihood that the child will be exposed totriggering events Each is discussed in turn

Stable family characteristics can have a direct effect on child characteristics that change the likelihood that children will be exposed to triggering events

Researchers from the Oregon Social Learning Centerhave presented compelling evidence that coercive parent-ing practices experienced in early childhood lead to asequence of events mediated through child behaviorthat increases the likelihood that individuals will beexposed to deviant peer groups where tobacco use is com-mon and delinquent acts are reinforced (see for examplePatterson

et al

1989 Dishion

et al

1991 1996)Although these researchers focus on family reinforce-ment of aversive child behaviors and childrenrsquos failure toacquire prosocial skills other mechanisms are possible aswell For example genetic characteristics or prenatalexposure to alcohol are biological mechanisms throughwhich families can influence child sensation seeking ordecrease impulse control each of which may increase thelikelihood that children are exposed to triggering events(Rutter

et al

1995) Girls whose mothers smoked duringpregnancy are more likely to smoke during adolescence(Kandel

et al

1994) Although part of this effect appearsto be mediated through girlsrsquo behavior problems (a resultconsistent with Patterson DeBaryshe amp Ramseyrsquos (1989)model of the etiology of deviance) prenatal tobacco

5

Exceptions to this general trend is the finding by Bauman

et al

(1984) that maternal disapproval of smoking is associated onlywith lower levels of cigarette use when the adolescent is close tothe mother (although no such association was found withregard to fathers) and by Andrews Hops amp Duncan (1997) thatmodeling of parent alcohol and marijuana was highest whenadolescents had relatively good relationships with their parentsIn contrast Doherty amp Allen (1994) found that parent smokingcombined with

low

family cohesion was the best predictor ofsmoking Although inconsistent with what would be predictedfrom social cognitive theory these findings are consistent withthe contradictory processes that often operate within familiesdiscussed in the next section of the paper


Addiction

98

(Suppl 1) 21ndash36

26


exposure appears to have a direct effect as well (Griesler

et al

1998) It is possible that such exposure increasesgirlsrsquo preferences for or responsiveness to tobacco a pointdeveloped in the next section

Stable family characteristics can have a direct effect on child characteristics that increase the likelihood that triggering events will cause a change in behavior

One way of conceptualizing family characteristics as riskfactors is by using the analogy of action potential innerves Nerves have a particular action potential thatthey must reach before firing Repeated stimulation ofthe nerve increases its excitation state The greater theexcitation state the less additional stimulation the nerveneeds before it gets above a critical level and the nervefires Familial risk factors can be thought of as raisingthe excitation state of the system Peer processes are thetriggering events that bring the system over the criticallevel With higher excitation states less stimulation isneeded to trigger the release of the action potential Forexample exposure to positive smoking models (such asparents) may increase the likelihood that adolescentswill accept a cigarette when it is offered Jessor amp Jessor(1977) describe adolescents who are easily influenced toengage in deviant behaviors as lsquotransition pronersquo andhypothesize that rejection of mainstream values andrejection of parents increases adolescentsrsquo transitionproneness by making them more vulnerable to negativepeer influences Although Jessor amp Jessor focus on familysocialization of dominant norms and the impact offamily processes on child personality characteristicsbiological processes may be seen to increase transitionproneness as well To the extent that susceptibility to theeffects of nicotine or to tobacco addiction or to theextent that personality characteristics such as opennessto peer influence have a genetic component familycharacteristics may be said to predict susceptibility totriggering events through biological mechanisms Inaddition family characteristics can change biologicalpredisposition to smoke if exposure to nicotine ortobacco smoke (either prenatally or postnatally)increases preferences for or susceptibility to smokingFor example prenatal exposure to tobacco predicts ado-lescent girlsrsquo current smoking status but not their life-time smoking history once current maternal smoking iscontrolled (Griesler

et al

1998) Life-time smoking his-tory is predicted by childhood problem behaviors butcurrent smoking status is not These results are consis-tent with the hypothesis that prenatal tobacco exposurecreates a lsquolatent addictionrsquo in girls that is activated bytobacco exposure (Griesler

et al

1998) The likelihoodthat postnatal tobacco exposure occurs varies depend-ing upon characteristics of the girlsrsquo proximal

environment (including her peer group) which in turnis dependent partially upon her own characteristicsincluding childhood problem behavior

Stable family characteristics can have a direct effect on the likelihood that children will be exposed to triggering events

It is likely that stable family characteristics will influencechildrenrsquos exposure to triggering events through theirinfluence on childrenrsquos social network (for examplethrough choice of neighborhood or friends) Regionalneighborhood and social class variations in the preva-lence of smoking result in variability in the likelihood thatchildren will be exposed to smoking models and offers Inone study of substance use (excluding tobacco) it wasfound that living in neighborhoods where adolescents aremore involved in substance use increases the likelihoodthat adolescents themselves will become involved evencontrolling for parent characteristics (Darling amp Stein-berg 1997) Residential location is chosen normally byparents and is influenced by numerous family character-istics (see Furstenberg

et al

1999 for in-depth treatmentof this topic) In addition Darling amp Steinberg also reportthat the relationship between substance use amongneighborhood peers and individual substance use variedwith the extent to which families were integrated into theneighborhood Adolescents whose families were inte-grated into their neighborhoods were more vulnerable tothe negative influence of high levels of peer substance useIn this study families had both a direct influence on ado-lescentsrsquo exposure to substance-using peers throughneighborhood selection and within neighborhood influ-enced their adolescentsrsquo vulnerability to neighborhoodinfluences through their social integration

Family processes influence exposure to risk factors inother ways as well One process through which parentalmonitoring parenting style and parental supportivenessmay influence adolescent smoking is by changing thelikelihood that adolescents associate with deviant peers(Patterson

et al

1989 Steinberg

et al

1995) In additionto changing the exposure of adolescents to substance-using peers family characteristics may also increase thelikelihood that adolescents are exposed to triggeringevents such as stressors that may move adolescents morequickly along the trajectory to dependent tobacco useChildhood exposure to family-related stressors (includingmarital disruption divorce physical sexual or emotionalabuse growing up with substance-abusing mentally illor incarcerated family members or an abused mother) isrelated to earlier initiation of smoking greater likelihoodof ever smoking current smoking status and currentsmoking level during adolescence and adulthood (Anda

et al

1999) Anda

et al

argue that exposure to stressful


27


Addiction

98

(Suppl 1) 21ndash36

events increases the salience of the stress-reduction prop-erties of nicotine increasing the benefit of tobacco tousers during early stages and making it more difficult toquit at later stages Thus family stressors may bothincrease adolescentsrsquo exposure to triggering eventsthrough impact on the peer group and also increase ado-lescentsrsquo vulnerability to triggering events

Note that each of these three potential classes of pro-cesses linking stable family characteristics with change inadolescent smoking is mediated through triggeringevents such as exposure to tobacco-using peers Theinter-relationship of these processes can be seen in astudy of contextual influences on substance use (Caldwellamp Darling 1999) Low parental monitoring and greateramount of time spent in unsupervised peer settings pre-dicted more time spent in settings where it was likely thatadolescents would be exposed to substance use (lsquoparty-ingrsquo) However unsupervised time spent with peers wasassociated with time partying only when the peers withwhom adolescents associated thought partying wasimportant In addition although time spent partying wasassociated with higher levels of substance use (excludingtobacco) this was true only for adolescents with lowresistance to peer pressure In other words the linkbetween low parental monitoring and substance usecould be interrupted at several points both by character-istics of adolescentsrsquo peer groups and by characteristics ofthe adolescents themselves This complexity is almost cer-tainly one underlying reason for the weak associationobserved between stable family processes and changes inadolescent tobacco use

One final point that is important in thinking about theissue of predicting change in behaviors from stabilityreturns to the issue of age As several authors have notedthe same parent behaviors may take on different charac-teristics when the child is at different ages as these behav-iors become more or less developmentally appropriate (egDarling amp Steinberg 1993 Dishion amp McMahon 1998)For example strict curfews and parental supervision thatare developmentally appropriate for a younger child mightbe met with hostility and rebellion by an adolescent Poorparental monitoring which might have had only modestnegative effects during middle childhood may becomeincreasingly problematic as deviant behaviors becomemore normative and adolescents spend more time inunsupervised settings In other words stability in parentbehavior coupled with change in the developing childmay combine to form a triggering event for change It isexactly this type of lag between changes in parent behav-ior and child expectations that Collins amp Gunnar (1990)Collins amp Luebker (1994) and Smetana (1994) credit withthe increase in conflict observed in early adolescence

Similarly if parents communicate that smoking is anunacceptable behavior

for children

the older adolescent

may not perceive parent disapproval as applying to themAs adolescents grow older they increasingly differentiatebetween their belief that parents have a right to set rulesand their belief that they are not obligated to obey paren-tal rules that they do not agree with (Smetana 1994)Although alcohol drug and tobacco use are defined aslsquoprudentialrsquo (ie safety-related) issues about which ado-lescents expect parents to set rules and believe parentshave a right to set rules about they are also the issues onwhich adolescents are most likely to differentiate betweenparentsrsquo right to set rules and their own obligation to obey(Nucci

et al

1991) Again here the change from which abehavioral transition might be predicted is not a charac-teristic of the parents which remains constant but ofadolescentsrsquo thinking about the issues involved

THE NATURE OF THE FAMILY THE FAMILY AS A COMPLEX SYSTEM

In their 1957 study of the influence of mothers on infantdevelopment Sears Maccoby amp Levin (1957 1976)summarized the relationships between hundreds ofmaternal characteristics and dozens of child outcomesTheir main conclusion was that families are complicatedsystems and that it is impossible to fully disentangle theinfluence of the many different processes that operatewithin them Similar conclusions had been reached bythe Child Study Association in 1936 (Greenbergerquoted by Symonds 1939) It is this inherent complexitythat has motivated much of the research on parentingstyle (Darling amp Steinberg 1993) Although SearsMaccoby and Levin were writing about the influence ofmothers on young children and despite the decades thathave passed and the striking advances in statisticalmodeling the same conclusions might be drawn aboutthe influence of the family on adolescent smoking

Multiple processes the case of style

There are three major difficulties in modeling familyinfluences First individuals within families engage inmany different exchanges and interactions across abroad range of issues As was seen in the discussion ofsocial cognitive theory these exchanges may push theadolescent in contradictory directions as when having agood relationship with a mother who smokes may predis-pose an adolescent to look at smoking positively butmake it less likely that he spends time with friends whosmoke Although the influence of different processes maybe additive it is also possible that they will moderate eachothermdasheither acting together synergistically or bufferingone another For example authoritative parenting aparenting style that combines strictness and warmth is


Addiction

98

(Suppl 1) 21ndash36

28


associated with lower levels of tobacco use (Baumrind1991 Radziszewska

et al

1996 Cohen amp Rice 1997)Darling amp Steinberg (1993) have argued that parentingstyle might best be thought of as a context that moder-ates the influence of other parenting behaviors byincreasing the effectiveness of those behaviors and bychanging adolescentsrsquo openness to socialization Consis-tent with this model Bauman

et al

(1984) report thatmaternal disapproval of smoking is only associated withlower levels of cigarette use when the adolescent is closeto the mother (although no moderating effect was foundwith regard to fathers) and Andrews

et al

(1997) reportthat modeling of parent alcohol and marijuana washighest when adolescents had relatively good relation-ships with their parents

6

A further complication is that processes within fami-lies operate over a long period of time and their influencemay differ depending upon the age of the child For exam-ple several researchers have reported that parentsrsquoformer cigarette use is associated with adolescent smok-ing (Bailey

et al

1993 Bauman

et al

1990 Jackson ampHenriksen 1997 Farkas

et al

1999) Farkas

et al

(1999)found that parental smoking reduction is most beneficialif it occurs before the child reaches age 9 ie before mostchildren encounter frequent cigarette offers Nonethe-less children whose parents quit when they were rela-tively young were only one-third less likely to smoke thanthose whose parents were currently smoking Theseresults suggest that children absorb messages abouttobacco use across their life-time not just during theperiod when they are likely to begin to smoke Thus fam-ilies are unique in the breadth of areas they touch uponin the length of time they interact and in the fact that theytouch the lives of the children with them at all develop-mental stages

Reciprocal processes the case of monitoring

The second major difficulty posed by the complexity offamily systems is that patterns of family interactions andinfluence develop reciprocally over time and are mutuallyinterdependent Parental monitoring is a case in pointEffective monitoring of adolescent behavior is associatedconsistently with lower levels of substance use (see forexample Dishion

et al

1991 Metzler

et al

1994 1998

Duncan

et al

1998 Ary

et al

1999a 1999b Mott

et al

1999) Although monitoring has often been interpretedas a parenting practice involving active attempts on thepart of parents to watch over children as a means of con-trol (for reviews see Kerr amp Stattin 2000 Stattin amp Kerr2000) recent theoretical and empirical work has under-mined this interpretation (Dishion amp McMahon 1998Kerr amp Stattin 2000 Stattin amp Kerr 2000) For examplealthough parental monitoring predicted a wide range ofpositive adolescent outcomes most of this relationshipwas explained by adolescentsrsquo spontaneous disclosure ofinformation to parents and not by parentsrsquo attempts tosecure information (Kerr amp Stattin 2000)

Preliminary results from a newly developing line ofresearch suggests that both adolescentsrsquo decisions aboutthe extent of information they disclose to parents andtheir motivations for doing so depend upon many factorsincluding their perception that the issue can be kept suc-cessfully from parents their belief that their parent has alegitimate right to set rules about the behavior in ques-tion and their perception that the issue is important(Darling

et al

2000 Darling 1998 pp 138 523 Darlingamp Koehle 2000 p 524 Cumsille 2002 p 1233) Adoles-cents are more likely to disclose disagreement with par-ents when they think the issue is important than whenthey do not which suggests that most adolescents areinterested in maintaining open communication with par-ents even when so doing might have negative conse-quences for themselves

7

This new research suggests thata critical factor in determining both disclosure strategyand motivations appears to be adolescentsrsquo beliefs aboutthe legitimacy of their parentsrsquo authority Stattin amp Kerrinterpret non-disclosure as a sign of dysfunction withinthe family system (Kerr amp Stattin 2000 Stattin amp Kerr2000) In contrast Elkind (1980) suggests that increas-ing privacy concerns and decisions not to share somekinds of information with parents represent a normativeshift in parentndashadolescent relationships that is part of thehealthy development of autonomy an interpretationmore consistent with literature on strategic communica-tions and interactions (see for example Buhrmester ampPrager 1995 other chapters in the same volume) It ispossible however that although normative shifts in

6

It should be noted that Bailey

et al

(1993) concluded in astudy of 6th and 7th graders that parent smoking and familycontextual variables (monitoring time spent with child disap-proval of smoking family conflict and family disunion) that anadditive rather than interactive model best fit the data How-ever the young age and high functioning of this sample limits itsability to inform this issue in that as discussed previously differ-ent processes may be involved in taking up smoking at differentdevelopment stages

7

On average high school students in this dataset ranked smok-ing drinking and drug use as among the least important tothem of 36 potentially conflictual issues (31st 32nd and 36threspectively) Smoking had a mean importance score of 217with 1 representing lsquonot importantrsquo and 4 lsquoimportantrsquoalthough 50 of 121 students rated smoking as either lsquoveryimportantrsquo or lsquoimportantrsquo Virtually equal numbers of adoles-cents agreed and disagreed with parents about smoking andagreement with parents did not differentiate between those whoreporting smoking was lsquovery importantrsquo or lsquoimportantrsquo to themfrom those who reported that it was lsquoa littlersquo or lsquonotrsquo important


29


Addiction

98

(Suppl 1) 21ndash36

disclosure patterns are a healthy part of the evolution ofthe parentndashadolescent relationship premature shifts tonon-disclosure may represent problematic developmen-tal trajectories If true this pattern would be parallel bothto Jessor amp Jessorrsquos (1977) work on problem behavior syn-drome and Steinberg amp Silverbergrsquos (1987) research onthe development of adolescent autonomy In both thosestudies early evidence of pseudoadult behavior was indic-ative of developmental problems This newly developingliterature on the adolescentsrsquo role in parental knowledgeof adolescent activities makes it clear that what has beencalled parental monitoring is a function of both parentand child characteristics and reflects current character-istics of the relationship as well as past history

As children enter adolescence they spend increasingtime in unsupervised settings outside the home In addi-tion to increasing parentsrsquo reliance on adolescents forinformation about their activities this shift also meansthat the effectiveness of parental socialization effortsdepends upon adolescentsrsquo willingness to obey ruleswhen adolescents are away from direct parental supervi-sion Chyung amp Darling (1998 2003) report that adoles-centsrsquo beliefs in the legitimacy of their parentsrsquo authoritybuffers them from the negative influence of substance-using peers and that monitoring predicts adolescent sub-stance use only when adolescents do not believe that theirparents have the right to set rules Taken together thesefindings underscore the reciprocity of parentndashchild influ-ence and the reliance of adolescentsrsquo parents on theirchildrenrsquos willingness to be socialized (Darling ampSteinberg 1993) as well as the importance of under-standing both adolescentsrsquo and parentsrsquo roles in processesunderlying constructs such as monitoring (Kerr

et al

1999 Stattin amp Kerr 2000)

Diverse family forms

Finally the third major difficulty posed by the complexityof family systems is that lsquothersquo family is not a singularsource of influence but is usually made up of several dif-ferent sources each of whom may influence and be influ-enced by the developing individual in many differentways Traditional techniques for modeling multiplesources of influence run into several problems First fam-ilies take many forms For example in a study of a smallrural town in upstate New York in which the first authorwas involved it was necessary to code 34 different basiccategories of family types including a lsquojoint custodyrsquo cod-ing that did not specify who was living in each householdin which the child spent significant time In this categoryalone we found families in which children were spendingtime with two now divorced parents in separate homes afamily in which both parents and their new spouses livedwith the children of the first marriage in one large home

a family in which the child moved between living with herbiological father and a stepmother and her mother andher lesbian partner etc More diverse communitiesexpand the number of observed living situations Forexample lsquoWoodlawnrsquo a poor predominantly African-American neighborhood in Chicago was described ashaving more than 80 different family types (Pearson

et al

1990) This diversity poses a challenge for researchers Ifcurrent smoking status of mothers fathers and siblingsare coded separately decisions must be made about howto code adolescents who live only with a mother or onlywith a father Is residential status important or is smok-ing by non-residential parents important as well Shouldadolescents without complete sets of data (ie those livingin non-intact families) be dropped from the analysisWhat of step-parents and other adults living in the homeSensitivity to the definition of lsquofamilyrsquo is especially impor-tant in understanding social influences on adolescentsmoking among immigrant and ethnic minority groupswhose kinship networks include larger numbers of kinwho may have neither biological nor legal ties to the ado-lescent (Spencer amp Dornbusch 1990) Substantial num-bers of adolescents change their residential mobilitydepending upon opportunities for schooling recreationand the availability of supervising adults In a countrywhere more than one-third of children are born to unwedmothers and more than one-quarter of adolescents spend5 or more years living with a single parent this issue willbecome increasingly important

Siblings add additional complexity First number typeand residential status of siblings varies even more widelythan number type and residential status of parent figuresRelative age and gender are important considerations insibling research (Dunn amp Plomin 1990) If one collapsessibling smoking information by coding number of smok-ing siblings should adolescents who have no siblings anonly sibling who is still a toddler and three older non-smoking siblings all be coded in the same way Obviouslythe answers to these questions should vary dependingupon theory-driven hypotheses about developmental pro-cesses To date many of the analyses have been simplisticonly rarely examining the differential impact of mothersand fathers on their same and cross-gender offspringResearch on sibling smoking could benefit especially fromcareful theory-driven research in that it is both compli-cated and appears to have powerful effects on adolescentsmoking status (Avenevoli amp Merikangas 2003)

The standard method of modeling multiple familyinfluences is to use techniques such as multiple regres-sion entering mother father and sibling smoking as sep-arate variables that are independent of one anotherstatistically Potential moderating effects are modeled asstatistical interactions using multiplicative terms Severalproblems with this standard practice are worth


Addiction

98

(Suppl 1) 21ndash36

30


discussing however These problems are usually classedunder the categories of co-linearity and power If weassume that the smoking behaviors of family membersare associated with adolescent smoking (this is the reasonfor testing the model after all) we have to assume thatsmoking behaviors of other family members are corre-lated as well This problem of co-linearity makes it difficultto disentangle the separate influence of each variable (ieit reduces the power of the tests associated with each indi-vidual variable) and makes differences in variability animportant determinant of the results of the analyses Forexample take the case where mother and father smokingare more highly correlated with one another than theyare with sibling smoking When mother father and sib-ling smoking are entered into a regression equationsimultaneously co-linearity may eliminate the indepen-dent effects of mother and father smoking as individualvariables even though

treated as a block

these variablesmay still predict adolescent smoking In addition parentsmoking probably has a very different distribution thandoes sibling smoking If most adults either smoke regu-larly or do not smoke at all but adolescents exhibit awider range of smoking patterns (experimenting smok-ing regularly etc) then the variability of sibling smokingwill be much greater than the variability in parent smok-ing This would enhance the power of sibling smoking asa predictor In this situation modeling interactionsbetween variables may be particularly problematic notonly because the predictors are not independent but alsobecause the distribution of the component variables willnot approximate normal exacerbating the problem ofoutliers (Wilcox 1997)

Traditional techniques for modeling family influence the example of sibling smoking

Although the co-linearity issue inherent to modelingfamily processes is a methodological problem it has astrong conceptual component as well If we believe thatfamily smoking-related behaviors are interrelated does itmake sense to examine the influence of sibling smokingon target adolescent smoking for example but controlfor the effect of parents Several researchers havereported that when parent and sibling smoking areentered into a regression equation simultaneously theparent effect is diminished (Avenevoli amp Merikangas2003) Normally this would be interpreted as a media-tional effect the association between parent characteris-tic and adolescent outcome is caused by the parentsrsquoeffect on the sibling which in turn influences the targetadolescent directly This is consistent with empiricalresearch suggesting that adolescents sometimes beginsmoking in response to offers by older siblings (Presti

et al

1992) Logically however this explanation lacks

substance One cannot posit that parents influence onesibling directly but the other only indirectly

A second model consistent with this pattern of find-ings is that both siblings are influenced by a third(unmeasured) process that influences both of their smok-ing behaviors In this model parent behavior is associ-ated with adolescent smoking only because it iscorrelated with this third factor ie parent behavior canbe treated as a confound For example if the family lives ina neighborhood that has been targeted heavily by ciga-rette advertising aimed at children one can imagine thatparent smoking behaviors might be correlated withamount of neighborhood advertising but both siblingsmight be more strongly influenced by the advertisingbecause they were part of the targeted population Oncesibling smoking behavior is controlled for the parenteffect would drop out of the model but the sibling variablewould not because it captured the advertising effect Inother words predicting the behavior of one sibling fromthat of another taps the shared environment experiencedby the dyad

A third model consistent with a measured sibling vari-able diminishing the relationship between a measuredparent variable and adolescent smoking is similar butbuilds upon the idea that parent behavior is much morecomplex than measured parent variables Measured sib-ling smoking has the potential to capture all of the pro-cesses operating in the family that would influencesmoking These are presumably the same processesaffecting the target adolescent

8

If parental smoking werethe measured parent variable its effect on the target ado-lescent as well as all other unmeasured characteristics ofthe family neighborhood school and shared environ-ment that influence smoking would be captured in themeasure of sibling smoking When sibling smoking wascontrolled the relationship between parent smoking andadolescent smoking would be diminished

Within and between family differences

A better strategy for modeling family influences on smok-ing is to acknowledge the nested structure of the data anduse analytical methods that take this structure intoaccount Such methods can be modeled using structuralequation models or through specialized programs such asPROC MIXED (SAS) or HLM (hierarchical linear model-ing) (Bryk amp Raudenbush 1992) Taking a simple casedata could be collected on smoking status of two

8

Obviously sibling smoking would also include non-shared envi-ronmental effects as well as individual differences in the effectsof these processes on adolescent smoking that might not becommon to the sibling plus random error (Rowe amp Plomin1981)


31


Addiction

98

(Suppl 1) 21ndash36

adolescent siblings and two parent predictors in a sampleof families In this example adolescent smoking could bepredicted by the parent predictors sibling characteristics(smoking status for example) and individual character-istics (age and birth order for example) In a standardHLM analysis the first step would be to estimate the intra-class correlation between sibling smoking This is an esti-mate of the amount of variance in the smoking outcomethat can be attributed to being a member of a particularfamily (ie the similarity of siblings on smoking) In thesecond step of the analysis predictors would be used topredict both between-family and within-family varianceThis latter distinction between between-family andwithin-family models is critical Parent smoking is com-mon to both siblings but differs between families and canbe used to predict between-family differences in adoles-cent smoking Within-family differences would be pre-dicted by the differential experiences of the two siblingsFor example in a family in which one sibling smokes andthe other does not having a smoking sibling differenti-ates between the experience of the two siblings Note thatin these two models the parameter of the relationshipbetween shared family characteristic such as parentsmoking and adolescent smoking is estimated betweenfamilies but the parameter of the relationship betweennon-shared family characteristics such as sibling smok-ing is estimated within-families

Models that use two siblings from the same family toestimate sibling influence have several advantages overstandard techniques One is the differentiation ofbetween- and within-family parameter estimates Bymodeling sibling effects within families common sourcesof between family variance are removed from the siblingestimate In addition individual differences in siblingcharacteristics can be used to further specify the modelmdashfor example age This is particularly important in tryingto understand sibling influences Most researchers inter-ested in sibling influence on tobacco use have estimatedcurrent tobacco use status However because exposure tocontexts in which smoking is likely to be initiated is oftenage-graded the age differences between siblings becomesproblematic For example would having a 16-year-oldsibling who began smoking at 15 better predict the smok-ing status of a 13-year-old the age at which the 13-year-old would make the transition to smoking or thetrajectory of smoking In other words adolescentsrsquo devel-opmental trajectories might be more similar than theirstatuses at different ages In addition it may be that hav-ing an older sibling already smoking may speed theyounger sibling along their own trajectory Within-familygrowth curve analysis in which sibling trajectories arecompared with one another may be the most promisingmethod of understanding sibling influences (for a discus-sion of such models in the context of modeling reciprocal

influences in marital dyads see Maguire 1999) Differen-tial influence of family characteristics on siblings with dif-ferent characteristics can be modeled as well Rose

et al

(1999) used HLM to model family and peer influences onadolescent smoking using a sample of sibling dyads andfound strong between-family effects Unfortunately sib-ling influence was not modeled This paper was particu-larly interesting in that it noted strong sharedenvironmental effects within the family (both in that theintraclass correlation between siblings was high and theeffects of parent smoking were relatively powerful) aswell as strong non-shared environmental influences frompeers

Contexts communities and niches

An alternative way of conceptualizing the family as a con-text for adolescent smoking is to take the word lsquocontextrsquoliterally rather than simply using it as a synonym for theword lsquoinfluencersquo Personndashcontext interaction and ecolog-ical systems theories (Bronfenbrenner amp Morris 1998Magnusson amp Stattin 1998) both emphasize the inter-related nature of the processes and characteristics ofnaturally occurring systems such as families and thepotential for distorting our understanding of how thesesystems work when we ignore their systemic nature Oneway of respecting the systemic properties of families is toapply an ecological or community systems approach to itsstudy In the biological science of ecology the concept of

community

is used to describe a set of interrelated speciesthat co-occur commonly (Wilson amp Bossert 1971) Forexample a high desert supports a community that is quitedifferent from a bog community Although the individualspecies found in each community type vary in differentparts of the world species occupying a community sharecommon characteristics due to the processes underlyingcommunity formation Species in communities co-occurbecause they share overlapping niches A

niche

in thistechnical sense of the word is the set of environmentalconditions necessary for species survival (Wilson ampBossert 1971) For example temperature range rainfallsun exposure and soil pH are characteristics that help todifferentiate niches and thus separate species into differ-ent plant communities Species within communitiesshare a suite of correlated features due to the selection andadaptation processes that operate in their shared environ-mental conditions Although different community typesvary across the environmental features that define them(soil pH for example differentiates high desert and bogenvironments) it is important to note that the differencesbetween bog and high desert communities are not cap-tured along those single dimensions

Applying this kind of community systems perspectiveto the family provides an alternative strategy for studying


Addiction

98

(Suppl 1) 21ndash36

32


adolescent tobacco use and opens several lines of investi-gation not explored previously For example this modelsuggests that particular constellations of interrelatedcharacteristics arise in response to particular ecologicalprocesses (Ricklefs 1976) Cataloging and describingcommunity types leads to greater understanding of theprevalence and operation of etiological processes Appliedto adolescent tobacco use this suggests that the within-family patterning of tobacco use and a careful study of theconditions under which particular patterns arise mayprove helpful in gaining insight into how tobacco usebegins and is maintained For example if families inwhich only fathers smoke and both parents smoke arecommon but it is less common to find families in whichonly the mother smokes what are the processes under-lying this lsquoholersquo in the distribution How common is it forthe adolescents in a family to smoke when neither parentdoes and under what conditions does this pattern occurAlthough data collected to answer questions about thepredictors of individual smoking can be analyzed in sucha way as to answer these questions framing questionsfrom a community systems perspective that focuses onthe patterning of features changes the emphasis and gen-erates a different set of hypotheses (see Magnusson ampStattin 1998 for a discussion of this point)

Grouping families according to patterning and pre-dicting group membership also implies very differentmethodological approaches than individually centeredtechniques Focusing on individuals in particular roles(eg oldest child) within an ecological community frame-work embeds the child in a particular constellation ofprocesses Because typological analyses capture a greatdeal of information about environmental conditionsthrough the classification they increase the power of theanalyses markedly For example Darling amp Koehle(2000) presented an unpublished analysis of 516 siblingdyads who were a subset of a diverse sample of studentsfrom six high schools (see Steinberg

et al

1994 for a fulldescription of the complete sample and measures) Sib-ling dyads were classified according to the similarity oftheir peer groups and their perception of parental moni-toring Darling amp Koehle examined the relationship ofparental monitoring sibling substance use (excludingtobacco) and peer substance use with adolescent sub-stance use Based on hypotheses about buffering pro-cesses they predicted that parental monitoring would bemost powerful and peer substance use least powerful aspredictors in the condition where siblings experiencedsimilar levels of parental monitoring but had quite differ-ent types of peers Their hypotheses were confirmed Amore traditional non-system-based analysis would haverequired testing these hypotheses using two three-wayinteraction terms a daunting task Two things about thisexample are important The first is the issue of power

Complex interactions require large samples to model ade-quately The second is that dyads were classified notaccording to the patterning of their values on monitoringand peer substance use but on the

confluence

of influ-ences operating within the family In other words fami-lies were not grouped according to whether they hadhigh or low monitoring and on whether peers were highor low on substance use The studyrsquos major hypothesesconcerned the way that parent sibling and peer pro-cesses operated when family sibling and peer processeswere all working together compared to how they oper-ated when family and peer influences were working atcross-purposes Because of this dyads were groupedaccording to the confluence or disjunction of influences(maintaining the full range of variance in each group)Using this strategy direct comparison of the relationshipbetween variables across dyad types yields informationon the full range of adolescent outcomes in very differentcircumstances Other hypotheses might lead to differentsystemic properties becoming more important in deter-mining group membership

Systemic approaches may be especially useful whenexamining longitudinal data For example comparingthe smoking trajectories of adolescents who have similarcharacteristics but who are embedded within differentfamily types yields information about the influence ofmultiple factors (including moderating effects) with onlylimited loss of power

THE NATURE OF THE FAMILY DURING ADOLESCENCE THE INTERFACE OF FAMILY AND NON-FAMILY PROCESSES

The changing nature of adolescent social relationshipsand the previous discussion of processes linking stablefamily characteristics with changes in smoking behaviorboth highlight the importance of examining the interfacebetween family processes and processes operating inother contexts For example how do peer processes influ-ence adolescent smoking trajectories in families thatdiffer in parent smoking One approach to modeling mod-erating effects is using multiplicative interaction terms instandard regression models However this method haslimitations many of which have already been discussedReturning to the analogy of the plant communities adrought will have a very different effect on a high desertcommunity than on a bog community These differencesare captured to only a limited extent by examining thestatistical interaction of observed and average rainfall Ina small sample they would probably not be captured intests of the interaction of observed rainfall and soil pH atall This example illustrates two related points First andmost broadly the ability of an interaction term to capture


33


Addiction

98

(Suppl 1) 21ndash36

differences in process depends upon the ability to mea-sure the processes accurately In this particular examplesoil pH can be measured quite accurately but measuringrainfall is problematic in both settings The infrequenttorrential rains typical of high deserts can overwhelmmeasuring instruments while measurement of rainfall inbogs does not capture well constant low-level precipita-tion and dew and may not reflect deposition of watersheddrainage that results from distant rainfall Althoughmeasurement error is always problematic poor measure-ment is especially problematic in modeling interactionswhere error is compounded and deviations from normal-ity can have strong and unexpected effects on the results(Wilcox 1997) Secondly the differences between thehigh desert and bog communities are not captured wellby any single dimension along which they vary Despitethis limitation average rainfall probably captures the dif-ferences between the environments better

in this situation

because it is more strongly tied to the process of interestdrought In situations characterized by highly correlatedfeatures such as families single indicators may best bethought of as marker variables ie variables that do notmeasure the construct of interest but rather are standingin as proxies for them In this situation the more highlycorrelated the marker variable is with the latent unmea-sured construct the more successful it will be in captur-ing those differences in an interaction term

Structural equation models are designed to capturelatent processes more fully and have the additionaladvantage of enabling modeling of error terms Unfortu-nately most current applications are limited in that theydo not lend themselves to testing interactions betweenlinear measures Because the most important effects offamilies on adolescent smoking may be specifically intheir ability to moderate the influence of other contextsthis limits SEM modelsrsquo applicability to studying familyinfluences in this domain One promising strategy may beto combine the typological approach discussed in the pre-vious section with an SEM approach to modeling processFor example one might classify families according to apatterning of potential processes Comparison of SEMmodels of non-familial processes across family typeswould test hypotheses about the differential influence ofnon-familial processes on adolescent smoking in differentfamily contexts

CONCLUSION

The goal of this paper was to discuss how the nature ofadolescent smoking the nature of the family and thenature of the family during adolescence present seriousconceptual and methodological challenges to research-ers interested in studying the family as a context for the

development of adolescent smoking First although thestudy of smoking during adolescence is fundamentallythe study of change smoking is often predicted from rel-atively stable family characteristics such as demo-graphics parent smoking status or relationship qualityFocusing on adolescent smoking lsquostatusrsquo rather than ontrajectories introduces considerable error into models offamily process In addition it is likely that the influenceof families on smoking trajectories is indirect throughchanging adolescentsrsquo exposure and vulnerability totriggering events Greater sensitivity to the transitionalnature of adolescent smoking may help to bring thissometimes contradictory literature into clearer focusSecondly the complexity of the family system includ-ing multiple simultaneous processes the reciprocalnature of processes and the diversity of family formsmakes it imperative that research be grounded theoreti-cally Approaches that recognize the nested nature offamily data or take a typologicalcommunity systemsapproach to modeling family influences offer severaladvantages over more commonly employed statisticaltechniques in capturing this complexity These tech-niques also offer promise in modeling the interface offamily processes with processes operating in other socialcontexts Because adolescents spend increasing timeoutside the family the connections between the familyand the larger social world may be a critical area ofresearch

Although the many excellent prospective studies cur-rently in the literature provide a wealth of data about thecorrelates of adolescent tobacco use our understandingof the developmental processes underlying the initiationof tobacco use remains limited One factor contributing tothis problem may be disciplinary stemming from interestin predicting risk Although traditional statistical tech-niques provide excellent means of identifying risk factorsthat increase the likelihood of the development of a par-ticular behavior (smoking for example) they are morelimited in modeling developmental processes in complexinterrelated systems These limitations become especiallyapparent when many questions of major theoreticalinterest center around how the functioning of these com-plex systems varies in different circumstances as well asacross time Several less traditional approaches to model-ing adolescent smoking trajectories including growthcurve analyses use of techniques that take the nestednature of within-family processes into account typologi-cal approaches and the combination of these techniqueswere discussed Stronger links between the sophisticatedtheoretical discussions in the growing literature on ado-lescents tobacco use trajectories with the methodologicaltechniques used to test and model those processes willmove the field toward a clearer understanding of develop-mental processes


Addiction

98

(Suppl 1) 21ndash36

34 Nancy Darling amp Patricio Cumsille

REFERENCES

Anda R F Croft J B Felitti V J Nordenberg D Giles WH Williamson D F amp Giovino G A (1999) Adverse child-hood experiences and smoking during adolescence andadulthood Journal of the American Medical Association 2821652ndash1658

Andrews J A Hops H amp Duncan S C (1997) Adolescentmodeling of parent substance use the moderating effect of therelationship with the parent Journal of Family Psychology 11259ndash270

Ary D V amp Biglan A (1988) Longitudinal changes in adoles-cent cigarette smoking behavior onset and cessation Journalof Behavioral Medicine 11 361ndash382

Ary D V Duncan T E Biglan A Metzler C W Noell J Wamp Smolkowski K (1999a) Development of adolescent prob-lem behavior Journal of Abnormal Child Psychology 27 141ndash150

Ary D V Duncan T E Duncan S C amp Hops H (1999b) Ado-lescent problem behavior the influence of parents and peersBehaviour Research and Therapy 37 217ndash230

Avenevoli S amp Merikangas K R (2003) Family influences onadolescent smoking Addiction 98 (Supplement 1) S1ndashS20

Bailey S L Ennett S T amp Ringwalt C L (1993) Potentialmediators moderators or independent effects in the relation-ship between parentsrsquo former and current cigarette use andtheir childrenrsquos cigarette use Addictive Behaviors 18 601ndash621

Bandura A (1989) Human agency in social cognitive theoryAmerican Psychologist 44 1175ndash1184

Bauman K Fisher L Bryan E amp Chenoweth R (1984) Ante-cedents subjective expected utility and behavior a panelstudy of adolescent cigarette smoking Addictive Behavior 9121ndash136

Bauman K E Foshee V A Linzer M A amp Koch G G (1990)Effect of parental smoking classification on the associationbetween parental and adolescent smoking Addictive Behav-iors 17 413ndash422

Baumrind D (1991) The influence of parenting style of adoles-cent competence and problem behavior Journal of Early Ado-lescence 11 56ndash95

Berndt T (1979) Developmental changes in conformity to peersand parents Developmental Psychology 15 608ndash616

Biglan A Duncan T E Ary D V amp Smolkowski K (1995)Peer and parental influences on adolescent tobacco use Jour-nal of Behavioral Medicine 18 315ndash330

Bronfenbrenner U amp Morris P (1998) The ecology of develop-mental processes In Lerner R M ed Handbook of Child Psy-chology Theoretical Models of Human Development 5th edn Vol1 pp 993ndash1028 New York John Wiley

Brook J S Whiteman M Czeisler L J Shapiro J amp Cohen P(1996) Cigarette smoking in young adults childhood andadolescent personality familial and peer antecedents Journalof Genetic Psychology 158 172ndash188

Bryk A S amp Raudenbush S W (1992) Hierarchical Linear Mod-els Applications and Data Analysis Methods Newbury Park CASage Publications

Buhrmester D amp Prager K (1995) Patterns and functionsof self-disclosure during childhood and adolescence InRotenberg K J ed Disclosure Processes in Children and Adoles-cents pp 10ndash56 Cambridge Cambridge University Press

Caldwell L L amp Darling N (1999) Leisure context parentalcontrol and resistance to peer pressure as predictors of

adolescent partying and substance use an ecological perspec-tive Journal of Leisure Research 31 57ndash77

Centers for Disease Control and Prevention (1994) PreventingTobacco Use Among Young People A report of the Surgeon Gen-eral Atlanta GA US Department of Health and HumanServices

Chassin L Peterson C C Todd M Rose J S amp Sherman S J(1998) Maternal socialization of adolescent smoking theintergenerational transmission of parenting and smokingDevelopmental Psychology 34 1189ndash1201

Chassin L Presson C C amp Sherman S J (1984) Cognitive andsocial influence factors in adolescent smoking cessationAddictive Behaviors 9 383ndash390

Chyung Y amp Darling N (1998) Parental Monitoring and Legiti-macy of Parental Authority Flip Sides of the Same Coin Posterpresented at the 1998 Biennial Meeting of the Society forResearch on Adolescence San Diego CA Retrieved fromthe World Wide Web httpinsidebardeduacademicspecialprojdarlinglabsramon98pdf

Chyung Y amp Darling N (2003) Varying association betweenpeer problem behavior and adolescent problem behavioras a function of parental rule obedience Journal of the KoreanHome Economics Association Available through httpinsidebardeduacademicspecialprojdarlinglabchyungpdf

Cohen D A amp Rice J (1997) Parenting styles adolescentsubstance use and academic achievement Journal of DrugEducation 27 199ndash211

Collins W A amp Gunnar M R (1990) Social and personalitydevelopment Annual Review of Psychology 41 387ndash416

Collins W A amp Luebker C (1994) Parent and adolescentexpectancies individual and relational significance InSmetana J G ed Beliefs About Parenting Origins and Develop-mental Implications New Directions for Child Developmentno 66 pp 65ndash80 San Francisco CA Jossey-Bass Inc

Collins L M amp Sayer A G (2000) Modeling growth and changeprocesses design measurement and analysis for research insocial psychology In Reis H T amp Judd C M eds Handbookof Research Methods in Social Psychology pp 478ndash495Cambridge Cambridge University Press

Conrad K M Flay B R amp Hill D (1992) Why children startsmoking cigarettes predictors of onset British Journal of Addic-tion 87 1711ndash1724

Cumsille P Darling N amp PeZa-Alampay L (2002) Legitimacybeliefs and parent-adolescent conflict and adjustment inadolescence a Chilean and Filipino comparison Posterpresented at the Society for Research on Adolescent Develop-ment Biennial Meeting New Orleans LA Availablefrom httpinsidebardeduacademicspecialprojdarlinglabsra01cpupdf

Darling N Cumsille P amp Dowdy B (1998) Parenting stylelegitimacy of parental authority and adolescentsrsquo willingnessto share information with their parents Why do adolescentslie Paper presented at the International Society for the Studyof Personal Relationships Biennial Meeting Saratoga NYAvailable from httpinsidebardeduacademicspecialprojdarlinglabissprsdpdf

Darling N Hames K amp Cumsille P (2000) When Parents andAdolescents Disagree Disclosure Strategies and MotivationsPoster presented at the Society for Research in AdolescenceBiennial Meetings Chicago IL Retrieved from the World WideWeb httpwwwpersonalpsuedunxd10stratdispdf

Darling N amp Koehle L (2000) Variation in developmental pro-cess as a function of siblingsrsquo consensus on parental monitor-ing and peer group characteristics Poster Presented at the

Theory measurement and methods 35

copy 2003 Society for the Study of Addiction to Alcohol and Other Drugs Addiction 98 (Suppl 1) 21ndash36

Society for Research on Adolescence Biennial MeetingsChicago IL Retrieved from the World Wide Web httpwwwpersonalpsuedunxd10siblingspdf

Darling N amp Steinberg L (1993) Parenting style as context anintegrative model Psychological Bulletin 113 487ndash496

Darling N amp Steinberg L (1997) Community influences onadolescent achievement and deviance In Brooks-Gunn JDuncan G J amp Aber J L eds Neighborhood Poverty Vol 2pp 120ndash131 New York Russell Sage Foundation

Dishion T J amp McMahon R J (1998) Parental monitoring andthe prevention of child and adolescent problem behavior aconceptual and empirical formulation Clinical Child and Fam-ily Psychology Review 1 61ndash75

Dishion T J Patterson G R Stoolmiller M amp Skinner M L(1991) Family school and behavioral antecedents to earlyadolescent involvement with antisocial peers DevelopmentalPsychology 27 172ndash180

Dishion T J Spracklen K M Andrews D W amp Patterson GR (1996) Deviancy training in male adolescents friendshipsBehavior Therapy 27 373ndash390

Distefan J M Gilipin E A Choi W amp Pierce J P (1998)Parental influences predict adolescent smoking in theUnited States 1989ndash93 Journal of Adolescent Health 22466ndash474

Doherty W J amp Allen W (1994) Family functioning andparental smoking as predictors of adolescent scigarette use asix-year prospective study Journal of Family Psychology 8347ndash353

Duncan S C Duncan T E Biglan A amp Ary D (1998) Con-tributions of the social context to the development of adoles-cent substance use a multivariate latent growth modelingapproach Drug and Alcohol Dependence 50 57ndash71

Dunn J amp Plomin R (1990) Separate Lives Why Siblings Are SoDifferent New York Basic Books Inc

Elder G (1998) The life course and human development InLerner R M ed Handbook of Child Psychology TheoreticalModels of Human Development 5th edn Vol 1 pp 939ndash991New York John Wiley

Elkind D (1980) Strategic interactions in early adolescence InAdelson J ed Handbook of Adolescent Psychology pp 432ndash446 New York John Wiley amp Sons

Farkas A J Distefan J M Choi W S Gilpin E A amp Pierce JP (1999) Does parental smoking cessation discourage adoles-cent smoking Preventive Medicine an International Devoted toPractice and Theory 28 213ndash218

Flay B R Hu F B amp Richardson J (1998) Psychosocial pre-dictors of different stages of cigarette smoking among highschool students Preventive Medicine 27 A9ndashA18

Flay B R Hu F B Siddiqui O Day L E Petraitis JRichardson J amp Sussman S (1994) Differential influence ofparental smoking and friendsrsquo smoking on adolescent initia-tion and escalation of smoking Journal of Health and SocialBehavior 35 248ndash265

Frauenglass S Routh D K Pantin H M amp Mason C A(1997) Family support decreases influence of deviant peers onHispanic adolescentsrsquo substance use Journal of Clinical ChildPsychology 26 15ndash23

Furstenberg F F Cook T D Eccles J Elder G H amp SameroffA J eds (1999) Managing to Make It Urban Families and Ado-lescent Success Chicago University of Chicago Press

Graham J W Collins L M Wugalter S E Chung N K ampHansen W B (1991) Modeling transition in latent stage-sequential processes a substance use prevention exampleJournal of Consulting and Clinical Psychology 59 48ndash57

Griesler P C Kandel D B amp Davies M (1998) Maternal smok-ing in pregnancy child behavior problems and adolescentsmoking Journal of Research on Adolescence 8 159ndash185

Hanson W B Collins L M Johnson C A amp Graham J W(1985) Self-initiated smoking cessation among high schoolstudents Addictive Behaviors 10 265ndash271

Hawkins J Catalano R amp Miller J (1992) Risk and protectivefactors for alcohol and other drug porlbmes in adolescenceand early adulthood Implications for substance abuse preven-tion Psychological Bulletin 112 64ndash105

Jackson C amp Henriksen L (1997) Do as I say parent smokingantismoking socialization and smoking onset among chil-dren Addictive Behaviors 22 107ndash114

Jessor R amp Jessor S (1977) Problem Behavior and PsychosocialDevelopment a Longitudinal Study of Youth New YorkAcademic Press

Kandel D B Wu P amp Davies M (1994) Maternal smokingduring pregnancy and smoking by adolescent daughtersAmerican Journal of Public Health 84 1407ndash1413

Kerr M amp Stattin H (2000) What parents know how theyknow it and several forms of adolescent adjustment furthersupport for a reinterpretation of monitoring DevelopmentalPsychology 36 366ndash380

Kerr M Stattin H amp Trost K (1999) To know you is to trustyou parentsrsquo trust is rooted in child disclosure of informationJournal of Adolescence 22 737ndash752

Larson R amp Richards M (1991) Daily companionship in latechildhood and early adolescence changing developmentalcontexts Child Development 62 284ndash300

Larson R Richards M Moneta G Holmbeck G amp Duckett E(1996) Changes in adolescentsrsquo daily interactions with theirfamilies from ages 10ndash18 disengagement and transforma-tion Developmental Psychology 32 744ndash754

Magnusson D amp Stattin H (1998) Personndashcontext interactiontheories In Lerner R M ed Handbook of Child PsychologyTheoretical Models of Human Development Vol 1 pp 685ndash759New York John Wiley

Maguire M C (1999) Treating the dyad as the unit of analysisA primer on three analytic approaches Journal of Marriage andthe Family 61 213ndash223

Mayhew K P Flay B R amp Mott J A (2000) Stages in thedevelopment of adolescent smoking Drug and Alcohol Depen-dence 59 S61ndashS81

Melby J N Conger R D Conger K J amp Lorenz F O (1993)Effects of parental behavior on tobacco use by young maleadolescents Journal of Marriage and the Family 55 439ndash454

Metzler C W Biglan A Ary D V amp Li F (1998) The stabilityand validity of early adolescentsrsquo reports of parenting con-structs Journal of Family Psychology 12 600ndash619

Metzler C W Noell J Biglan A Ary D amp Smolkowski K(1994) The social context for risky sexual behavior amongadolescents Journal of Behavioral Medicine 17 419ndash438

Moffitt T E (1993) Adolescent-limited and life-course-persis-tent antisocial behavior a developmental taxonomy Psycho-logical Review 100 674ndash701

Moncher M S Holden G W amp Schinke S P (1991) Psycho-social correlates of adolescent substance use a review of cur-rent etiological constructs International Journal of theAddictions 26 377ndash414

Mott J A Crowe P A Richardson J amp Flay B (1999) After-school supervision and adolescent cigarette smoking contri-butions of the setting and intensity of after-school self-careJournal of Behavioral Medicine 22 35ndash58



Nucci L Guerra N amp Lee J (1991) Adolescents judgements ofthe personal prudential and normative aspects of drugusage Developmental Psychology 27 841ndash848

Patterson G R DeBaryshe B D amp Ramsey E (1989) A devel-opmental perspective on antisocial behavior American Psy-chologist 44 329ndash335

Pearson J L Hunter A G Ensminger M E amp Kellam S G(1990) Black grandmothers in multigenerational house-holds diversity in family structure and parenting involve-ment in the Woodlawn community Child Development 61434ndash442

Petraitis J Flay B R amp Miller T Q (1995) Reviewing theoriesof adolescent substance use organizing pieces in the puzzlePsychological Bulletin 117 67ndash86

Presti D E Ary D V amp Lichtenstein E (1992) The context ofsmoking initiation and maintenance findings from inter-views with youths Journal of Substance Abuse 4 35ndash45

Pulkkinen L (1990) Young adultsrsquo health and its antecedentsin evolving life-styles In Hurrelmann K amp Loesel F edsPrevention and Intervention in Childhood and Adolescence Vol 8pp 67ndash90 Berlin FRG Walter De Gruyter

Radziszewska B Richardson J L Dent C W amp Flay B R(1996) Parenting style and adolescent depressive symp-toms smoking and academic achievement ethnic genderand SES differences Journal of Behavioral Medicine 19 289ndash305

Ricklefs R E (1976) The Economy of Nature a Textbook in BasicEcology Portland OR Chiron Press

Rose J S Chassin L Presson C C amp Sherman S J (1999)Peer influences on adolescent cigarette smoking a prospectivesibling analysis Merrill-Palmer Quarterly 45 62ndash84

Rowe D C amp Plomin R (1981) The importance of nonsharedenvironmental (E1) influences in behavioral developmentDevelopmental Psychology 17 517ndash531

Rutter M Champion L Quinton D Maugham B amp PicklesA (1995) Understanding individual differences in environ-mental-risk exposure In Moen P Elder G H amp Luumlscher Keds Examining Lives in Context Perspectives on the Ecology ofHuman Development pp 61ndash96 Washington DC AmericanPsychological Association

Scarr S amp McCartney K (1983) How people make their ownenvironments a theory of genotypendashenvironment effectsChild Development 54 424ndash435

Schei E amp Sogaard A J (1994) The impact of military serviceon young menrsquos smoking behavior Preventive Medicine 23242ndash248

Sears R R Maccoby E E amp Levin H (1957 1976) Patterns ofChildrearing Stanford CA Stanford University Press

Smetana J G ed (1994) Beliefs About Parenting Origins andDevelopmental Implications San Francisco CA Jossey-BassInc

Spencer M B amp Dornbusch S M (1990) Challenges in study-ing minority youth In Feldman S S amp Elliott G R eds Atthe Threshold pp 123ndash146 Cambridge MA Harvard Univer-sity Press

Spielbeger C D Jacobs G A Crane R S amp Russell S F(1983) On the relation between family smoking habits and thesmoking behavior of college students International Review ofApplied Psychology 32 53ndash69

Stattin H amp Kerr M (2000) Parental monitoring a reinterpre-tation Child Development 71 1072ndash1085

Steinberg L (1990) Autonomy conflict and harmony in thefamily In Feldman S S amp Elliott G R eds At the Thresholdpp 255ndash276 Cambridge MA Harvard University Press

Steinberg L Darling N amp Fletcher A C (1995) Authoritativeparenting and adolescent adjustment an ecological journeyIn Moen P Elder G H amp Luumlscher K eds Examining Lives inContext Perspectives on the Ecology of Human Development pp423ndash466 Washington DC American PsychologicalAssociation

Steinberg L Fletcher A amp Darling N (1994) Parental moni-toring and peer influences on adolescent substance use Pedi-atrics 93 1ndash5

Steinberg L amp Silverberg S (1987) The vicissitudes ofautonomy in early adolescence Child Development 57 841ndash851

Symonds P M (1939) The Psychology of ParentndashChild Relation-ships New York Appleton-Century-Crofts

Tyas S L amp Pederson L L (1998) Psychosocial factors relatedto adolescent smoking a critical review of the literatureTobacco Control 7 409ndash420

Whitchurch G G amp Constantine L L (1993) Family systemstheory In Boss P G Doherty W J LaRossa R SchummW R amp Steinmetz S K eds Sourcebook of Family Theories andMethods a Contextual Approach pp 325ndash355 New YorkPlenum Press

Wilcox R R (1997) Introduction to Robust Estimation andHypothesis Testing San Diego CA Academic Press

Wilson E O amp Bossert W H (1971) A Primer of Population Biol-ogy Sunderland MA Sinauer Associates Inc

Wohlwill J F (1973) The Study of Behavioral Development NewYork Academic Press


Addiction

98

(Suppl 1) 21ndash36

22


additional characteristics of families First families comein many diverse forms Obviously we would not wantmodeling maternal and paternal smoking as separatevariables to result in all adolescents living in single parentfamilies eliminated from the analyses because of missingdata How should this problem be best addressed The sec-ond problem is more subtle Family member characteris-tics are inherently interdependent posing problems fortraditional statistical techniques For example if weassume that parent and offspring tobacco use are corre-lated how appropriate is it to model sibling influence onadolescent smoking by controlling parent tobacco useThese latter two concerns make use of traditional statis-tical techniques problematic

The third dilemma that researchers interested in theinfluence of the family on adolescent smoking must grap-ple with is the nature of the family

during adolescence

Although we have moved beyond thinking primarily ofthe family as a place from which adolescents move out(Steinberg 1990) it is still true that a key characteristic ofadolescence is the expansion of the social world Adoles-cents spend more time in the company of peers and lesstime with family members than do children (Larson ampRichards 1991 Larson

et al

1996) In additionalthough parental influence remains strong especially inareas involving values and long-term goals (Steinberg1990) peer influence peaks during mid-adolescenceespecially in day-to-day activities (Berndt 1979) Thusthe relationship between the family and other key con-texts should be of special interest to researchers interestedin adolescent tobacco use How do families influence thenature of other contexts to which adolescents areexposed How do they influence the processes that occurin those other contexts How are they influenced by out-side contexts How given the complex nature of the fam-ily can such interrelationships be modeled Similarlynormative changes in both contexts and behavioralexpectations allow greater expression of individual differ-ences in temperament genetic predispositions andlearned attitudes and behaviors in adolescence than inchildhood (Scarr amp McCartney 1983) How do familycharacteristics inhibit or facilitate the expression of suchindividual differences

This paper has two goals to examine each of theseissues as it relates to developmental patterns of adolescentsmoking trajectories and to discuss potential strategiesfor modeling family processes

STATUS STAGE TRANSITION AND TRAJECTORY

Recent reviews of the literature have highlighted the dis-tinction between examining adolescent smoking

status

and examining the development of smoking

trajectories

(Petraitis

et al

1995 Mayhew

et al

2000) Such distinc-tions are critical in understanding and modeling accu-rately the relationship between family characteristics andadolescent tobacco use Status refers to the current cate-gorization of an individual according to the variable ofinterest in this case tobacco use Such statuses may be assimple as lsquocurrent smokerrsquo versus lsquocurrent non-smokerrsquoor they may include an explicit time element for examplelsquonever tried tobaccorsquo versus lsquotried tobaccorsquo Other statuscategorization schemes include an implicit time elementFor example Mayhew

et al

(2000) explicate four lsquostagesrsquoof tobacco use Individuals are labeled lsquotriersrsquo lsquoexperi-mentersrsquo lsquoregular usersrsquo and lsquodependent usersrsquo Implicitin the stage model is the idea that a snapshot of an indi-vidual at any given time tells us something about theirhistorical tobacco use Just as we assume that the middle-aged individual pictured in a photograph was once achild the explicated stages imply that a person who iscaptured in the status of lsquoregular userrsquo has gone throughthe stages of lsquotrierrsquo and lsquoexperimenterrsquo(Petraitis

et al

1995)

1

Current tobacco use status is limited in what it tells usabout time however because it captures only a snapshotof a particular temporal moment In contrast the conceptof lsquotransitionrsquo incorporates time explicitly but incom-pletely A transition is defined by change over time forexample by a move from status 1 (eg lsquonever triedrsquo) to sta-tus 2 (lsquoexperimenterrsquo) Transitions are not identical totime however because the unit of analysis is

change

rather than

time

For example a person who is a non-smoker at both time 1 and time 2 cannot be said to havemade a transition even though time has passed Simi-larly two people both of whom have moved from status 1to status 2 may have existed in the statuses for differentperiods of time For example person 1 may have been instatus 1 at measurement points 1 2 and 3 and moved tostatus 2 at measurement point 4 while person 2 mayhave been at status 1 at measurement point 1 but status2 at measurement point 2 Although we can look at thepredictors of a transition without regard to time (forexample we can try to differentiate people who never

1

A lsquonever usedrsquo category is not included in these stages Also fre-quently neglected are statuses that capture current levels thatare lower than past use For example an individual who hadbeen a heavy smoker but now smokes only occasionally wouldpresumably be at a different stage of tobacco use than someonewho currently smokes occasionally but had never shown anyother pattern This distinction highlights the limited way inwhich time is reflected in categorization schemes based on cur-rent status Most stage models of substance use appear to bebased on lsquostrong stagersquo assumptions including unidirectionalityand irreversibility (Wohlwill 1973) Tobacco use histories how-ever are not consistent with these assumptions


23


Addiction

98

(Suppl 1) 21ndash36





trajectories


time

rather than by

change

2


3



-




et al


4


4


et al


et al


2


3



Addiction

98

(Suppl 1) 21ndash36

24




et al

1991 Conrad

et al

1992Hawkins

et al



et al


et al


et al

1983 Hanson

et al





cause




25


Addiction

98

(Suppl 1) 21ndash36



5


et al


et al


et al

1993 Flay

et al al

1994 Biglan

et al


et al

1997 Chassin

et al



et al





et al

1989 Dishion

et al


et al


et al


5


et al


low



Addiction

98

(Suppl 1) 21ndash36

26



et al




et al


et al





et al



et al

1989 Steinberg

et al


et al

1999) Anda

et al



27


Addiction

98

(Suppl 1) 21ndash36





for children



et al







Addiction

98

(Suppl 1) 21ndash36

28



et al


et al


et al


6


et al

1993 Bauman

et al


et al

1999) Farkas

et al




et al

1991 Metzler

et al

1994 1998

Duncan

et al

1998 Ary

et al

1999a 1999b Mott

et al



et al


7


6


et al


7



29


Addiction

98

(Suppl 1) 21ndash36



et al





et al





Addiction

98

(Suppl 1) 21ndash36

30



treated as a block




et al





8




8



31


Addiction

98

(Suppl 1) 21ndash36




et al




community


niche




Addiction

98

(Suppl 1) 21ndash36

32




et al



confluence






33


Addiction

98

(Suppl 1) 21ndash36


in this situation



CONCLUSION





Addiction

98

(Suppl 1) 21ndash36


REFERENCES








































































































23


Addiction

98

(Suppl 1) 21ndash36





trajectories


time

rather than by

change

2


3



-




et al


4


4


et al


et al


2


3



Addiction

98

(Suppl 1) 21ndash36

24




et al

1991 Conrad

et al

1992Hawkins

et al



et al


et al


et al

1983 Hanson

et al





cause




25


Addiction

98

(Suppl 1) 21ndash36



5


et al


et al


et al

1993 Flay

et al al

1994 Biglan

et al


et al

1997 Chassin

et al



et al





et al

1989 Dishion

et al


et al


et al


5


et al


low



Addiction

98

(Suppl 1) 21ndash36

26



et al




et al


et al





et al



et al

1989 Steinberg

et al


et al

1999) Anda

et al



27


Addiction

98

(Suppl 1) 21ndash36





for children



et al







Addiction

98

(Suppl 1) 21ndash36

28



et al


et al


et al


6


et al

1993 Bauman

et al


et al

1999) Farkas

et al




et al

1991 Metzler

et al

1994 1998

Duncan

et al

1998 Ary

et al

1999a 1999b Mott

et al



et al


7


6


et al


7



29


Addiction

98

(Suppl 1) 21ndash36



et al





et al





Addiction

98

(Suppl 1) 21ndash36

30



treated as a block




et al





8




8



31


Addiction

98

(Suppl 1) 21ndash36




et al




community


niche




Addiction

98

(Suppl 1) 21ndash36

32




et al



confluence






33


Addiction

98

(Suppl 1) 21ndash36


in this situation



CONCLUSION





Addiction

98

(Suppl 1) 21ndash36


REFERENCES








































































































Addiction

98

(Suppl 1) 21ndash36

24




et al

1991 Conrad

et al

1992Hawkins

et al



et al


et al


et al

1983 Hanson

et al





cause




25


Addiction

98

(Suppl 1) 21ndash36



5


et al


et al


et al

1993 Flay

et al al

1994 Biglan

et al


et al

1997 Chassin

et al



et al





et al

1989 Dishion

et al


et al


et al


5


et al


low



Addiction

98

(Suppl 1) 21ndash36

26



et al




et al


et al





et al



et al

1989 Steinberg

et al


et al

1999) Anda

et al



27


Addiction

98

(Suppl 1) 21ndash36





for children



et al







Addiction

98

(Suppl 1) 21ndash36

28



et al


et al


et al


6


et al

1993 Bauman

et al


et al

1999) Farkas

et al




et al

1991 Metzler

et al

1994 1998

Duncan

et al

1998 Ary

et al

1999a 1999b Mott

et al



et al


7


6


et al


7



29


Addiction

98

(Suppl 1) 21ndash36



et al





et al





Addiction

98

(Suppl 1) 21ndash36

30



treated as a block




et al





8




8



31


Addiction

98

(Suppl 1) 21ndash36




et al




community


niche




Addiction

98

(Suppl 1) 21ndash36

32




et al



confluence






33


Addiction

98

(Suppl 1) 21ndash36


in this situation



CONCLUSION





Addiction

98

(Suppl 1) 21ndash36


REFERENCES








































































































25


Addiction

98

(Suppl 1) 21ndash36



5


et al


et al


et al

1993 Flay

et al al

1994 Biglan

et al


et al

1997 Chassin

et al



et al





et al

1989 Dishion

et al


et al


et al


5


et al


low



Addiction

98

(Suppl 1) 21ndash36

26



et al




et al


et al





et al



et al

1989 Steinberg

et al


et al

1999) Anda

et al



27


Addiction

98

(Suppl 1) 21ndash36





for children



et al







Addiction

98

(Suppl 1) 21ndash36

28



et al


et al


et al


6


et al

1993 Bauman

et al


et al

1999) Farkas

et al




et al

1991 Metzler

et al

1994 1998

Duncan

et al

1998 Ary

et al

1999a 1999b Mott

et al



et al


7


6


et al


7



29


Addiction

98

(Suppl 1) 21ndash36



et al





et al





Addiction

98

(Suppl 1) 21ndash36

30



treated as a block




et al





8




8



31


Addiction

98

(Suppl 1) 21ndash36




et al




community


niche




Addiction

98

(Suppl 1) 21ndash36

32




et al



confluence






33


Addiction

98

(Suppl 1) 21ndash36


in this situation



CONCLUSION





Addiction

98

(Suppl 1) 21ndash36


REFERENCES








































































































Addiction

98

(Suppl 1) 21ndash36

26



et al




et al


et al





et al



et al

1989 Steinberg

et al


et al

1999) Anda

et al



27


Addiction

98

(Suppl 1) 21ndash36





for children



et al







Addiction

98

(Suppl 1) 21ndash36

28



et al


et al


et al


6


et al

1993 Bauman

et al


et al

1999) Farkas

et al




et al

1991 Metzler

et al

1994 1998

Duncan

et al

1998 Ary

et al

1999a 1999b Mott

et al



et al


7


6


et al


7



29


Addiction

98

(Suppl 1) 21ndash36



et al





et al





Addiction

98

(Suppl 1) 21ndash36

30



treated as a block




et al





8




8



31


Addiction

98

(Suppl 1) 21ndash36




et al




community


niche




Addiction

98

(Suppl 1) 21ndash36

32




et al



confluence






33


Addiction

98

(Suppl 1) 21ndash36


in this situation



CONCLUSION





Addiction

98

(Suppl 1) 21ndash36


REFERENCES








































































































27


Addiction

98

(Suppl 1) 21ndash36





for children



et al







Addiction

98

(Suppl 1) 21ndash36

28



et al


et al


et al


6


et al

1993 Bauman

et al


et al

1999) Farkas

et al




et al

1991 Metzler

et al

1994 1998

Duncan

et al

1998 Ary

et al

1999a 1999b Mott

et al



et al


7


6


et al


7



29


Addiction

98

(Suppl 1) 21ndash36



et al





et al





Addiction

98

(Suppl 1) 21ndash36

30



treated as a block




et al





8




8



31


Addiction

98

(Suppl 1) 21ndash36




et al




community


niche




Addiction

98

(Suppl 1) 21ndash36

32




et al



confluence






33


Addiction

98

(Suppl 1) 21ndash36


in this situation



CONCLUSION





Addiction

98

(Suppl 1) 21ndash36


REFERENCES








































































































Addiction

98

(Suppl 1) 21ndash36

28



et al


et al


et al


6


et al

1993 Bauman

et al


et al

1999) Farkas

et al




et al

1991 Metzler

et al

1994 1998

Duncan

et al

1998 Ary

et al

1999a 1999b Mott

et al



et al


7


6


et al


7



29


Addiction

98

(Suppl 1) 21ndash36



et al





et al





Addiction

98

(Suppl 1) 21ndash36

30



treated as a block




et al





8




8



31


Addiction

98

(Suppl 1) 21ndash36




et al




community


niche




Addiction

98

(Suppl 1) 21ndash36

32




et al



confluence






33


Addiction

98

(Suppl 1) 21ndash36


in this situation



CONCLUSION





Addiction

98

(Suppl 1) 21ndash36


REFERENCES








































































































29


Addiction

98

(Suppl 1) 21ndash36



et al





et al





Addiction

98

(Suppl 1) 21ndash36

30



treated as a block




et al





8




8



31


Addiction

98

(Suppl 1) 21ndash36




et al




community


niche




Addiction

98

(Suppl 1) 21ndash36

32




et al



confluence






33


Addiction

98

(Suppl 1) 21ndash36


in this situation



CONCLUSION





Addiction

98

(Suppl 1) 21ndash36


REFERENCES








































































































Addiction

98

(Suppl 1) 21ndash36

30



treated as a block




et al





8




8



31


Addiction

98

(Suppl 1) 21ndash36




et al




community


niche




Addiction

98

(Suppl 1) 21ndash36

32




et al



confluence






33


Addiction

98

(Suppl 1) 21ndash36


in this situation



CONCLUSION





Addiction

98

(Suppl 1) 21ndash36


REFERENCES








































































































31


Addiction

98

(Suppl 1) 21ndash36




et al




community


niche




Addiction

98

(Suppl 1) 21ndash36

32




et al



confluence






33


Addiction

98

(Suppl 1) 21ndash36


in this situation



CONCLUSION





Addiction

98

(Suppl 1) 21ndash36


REFERENCES








































































































Addiction

98

(Suppl 1) 21ndash36

32




et al



confluence






33


Addiction

98

(Suppl 1) 21ndash36


in this situation



CONCLUSION





Addiction

98

(Suppl 1) 21ndash36


REFERENCES








































































































33


Addiction

98

(Suppl 1) 21ndash36


in this situation



CONCLUSION





Addiction

98

(Suppl 1) 21ndash36


REFERENCES








































































































Addiction

98

(Suppl 1) 21ndash36


REFERENCES







































































































Theory, measurement, and methods in the study of family influences on adolescent smoking

Documents

Transcript of Theory, measurement, and methods in the study of family influences on adolescent smoking