Copyright 1991 by the American Psychological Association, Inc. 0022-3514/91/S3.00
Journal of Personality and Social Psychology 1991, Vol. 61, No. 4, 614-628
Influence of Sensation Seeking on General Deviance and Specific Problem Behaviors From Adolescence to \bung Adulthood Michael D. Newcomb and Linda McGee
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Division of Counseling and Educational Psychology University of Southern California Indicators of a Sensation Seeking latent construct were assessed during adolescence and used to predict changes in repeatedly gathered measures of a latent construct of General Deviance. A community sample of 595 male and female Ss was assessed 3 times over a 5-year period from late adolescence to young adulthood. Most Ss reported use of licit drugs, about one half had tried illicit drugs, and a substantial minority had engaged in other delinquent or criminal activities. The General Deviance construct was stable over time, although specific cross-effects were found. Sensation Seeking was moderately correlated with General Deviance at all 3 levels but did not predict directly the General Deviance construct over time; the effects of Sensation Seeking on later deviant behavior and attitudes were specific rather than general.
General and Specific Deviance
Crime, delinquency, and deviance among teenagers may take several specific forms, such as drug and alcohol use or abuse, vandalism, disciplinary problems in school, sexual precocity, or aggression. These may also reflect certain nontraditional attitudes, such as low religiosity, rebelliousness, and disregard for the law. These types of behavior and attitudes have often been observed among teenagers and seem to reflect a syndrome and latent construct of problem behavior or General Deviance (Donovan & Jessor, 1985; Osgood, Johnston, O'Malley, & Bachman, 1988). However, the developmental stability and integrity of this construct are not well established. The extent of any reciprocal causal effects among the specific deviant attitudes and behavior over time has rarely been studied. Little attention has been devoted to understanding what factors affect this general construct of deviance over time distinct from the constituent variables. We examined these issues empirically using multiwave panel data. First, we examined the stability and integrity of a latent construct of General Deviance and reciprocal influences among its constituent components in three waves of data over a 5-year period from late adolescence to young adulthood. Second, some suggest that personality or individual differences may be responsible for the initiation and maintenance of these deviant attitudes and behaviors (Huba & Bentler, 1983). Sensation seeking has often been suggested as one such possible personal characteristic, and we examined its effects on various types of problem behavior from late adolescence to young adulthood.
The transition from adolescence to young adulthood is a particularly critical stage of development for successful acquisition and performance of adult roles (e.g., Havighurst, 1972; Newcomb, 1987; Newcomb & Bentler, 1988a). This period often involves experimentation with new attitudes and behaviors, several of which may conflict with traditional society and be seen as problems. Problem behavior has been defined by Jessor and Jessor (1977) as "behavior that is socially defined as a problem, a source of concern, or as undesirable. . . and its occurrence usually elicits some kind of social control response" (p. 33). According to problem behavior theory, a variety of behaviors considered problematic during adolescence are associated with one another (Donovan & Jessor, 1985; Jessor & Jessor, 1977). If these various deviant behaviors are generated by one underlying motive or propensity, they can be construed as a general syndrome of deviance (Donovan & Jessor, 1985). Empirical data confirm a substantial, positive covariation among diverse adolescent problem behaviors such as illicit drug use, general delinquent behavior (e.g., cheating, lying, petty theft, and vandalism), problem drinking, precocious sexual activity, and other rebellious behaviors. Even frequent use of licit drugs (cigarettes and alcohol) constitutes part of this general deviance syndrome, which involves a wide range of norm-violating behaviors (Bachman, O'Malley, & Johnston, 1978; Donovan & Jessor, 1985; Elliott & Ageton, 1976; Elliott & Huizinga, 1984; Hindelang & Weis, 1972; Jessor & Jessor, 1977; Kandel, 1980; Robbins & Murphy, 1967). These repeated observations suggest that a common set of causes may underlie these specific behaviors (Barnes & Welte, 1986; Elliott, Huizinga, & Ageton, 1985; Huba & Bentler, 1983). Donovan and Jessor (1985) demonstrated that several measures of deviant behavior were generated by an underlying tendency toward or latent factor of unconventionality for both adolescents and college students. This general syndrome of problem behavior among teenagers seems to be a meaningful, reliable,
This research was supported by Grant DA01070 from the National Institute on Drug Abuse. We warmly acknowledge the production assistance of Tove Davison and Julie Speckart. Correspondence concerning this article should be addressed to Michael D. Newcomb, Division of Counseling and Educational Psychology, University of Southern California, Los Angeles, California 90089-0031.
614
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
GENERAL DEVIANCE AND SPECIFIC PROBLEM BEHAVIORS
and valid construct during adolescence, but may be less relevant for other stages in life (e.g., Donovan & lessor, 1985). It may diversify and fragment during the transition from older adolescence to young adulthood. Osgood et al. (1988) examined the longitudinal structural relationships among several repeated measures of deviant behavior (criminal behavior, drug use, and problem driving). A common factor or underlying (latent) construct of these deviant behaviors remained stable over a 4-year period from age 18 to 22. This construct of General Deviance was stable over time, and different types of specific deviance (i.e., problem drinking) were also stable over time, but there were few reciprocal effects between specific behaviors over time. That is, one type of deviance had little unique effect on other types of deviance over time. Osgood et al. (1988) concluded that "theories that treat different deviant behaviors as alternative manifestations of a single general tendency can account for some, but not for all, of the meaningful variance in these behaviors" (p. 81). Newcomb and Bentler (1988a) found decreasing cohesiveness among deviant attitudes and behavior as their adolescent population grew into young adulthood, although the specific types of deviance were moderately stable. This suggests that the syndrome of problem behaviors present in an adolescent population may not remain a unified construct in young adulthood or later life. This construct seemed to differentiate into at least two sets of deviant behavior over time, one set of behaviors including problems related to drug use and another set including other forms of deviance. This also supports the conclusion of Osgood et al. (1988) that deviant behaviors become more "specialized" as the adolescent moves into young adulthood.
Sensation Seeking During the past 2 decades, Sensation Seeking has emerged as an important explanatory construct for a variety of behaviors. The sensation-seeking trait has been defined by Zuckerman (1979) as "the need for varied, novel, and complex sensations and experiences and the willingness to take physical and social risks for the sake of such experiences" (p. 10). The theoretical basis for Sensation Seeking as a biosocial dimension of personality has changed over the years to incorporate new biological findings. Zuckerman originally proposed that individuals with high sensation-seeking needs feel and function best at a high level of cortical arousal and that they are motivated to engage in activities that increase their arousal level. He later rejected this optimal arousal basis for Sensation Seeking and reduced arousal to a secondary role. He now hypothesizes that Sensation Seeking is a general sensitivity to reinforcement that may reflect levels of catecholamines and their enzymes (Zuckerman, 1983,1984; Zuckerman, Buchsbaum, & Murphy, 1980). Zuckerman (1979) has documented an impressive array of correlations between the Sensation-Seeking Scale (SSS) he developed and a variety of psychological measures. For instance, individuals high on the SSS are more oriented to body sensation, extraverted, impulsive, antisocial, nonconformist, and less anxious (Zuckerman & Link, 1968). It is, therefore, reasonable to suspect that these psychological associations with the
615
SSS are reflected in correlations between the SSS and analogous overt behavior (Segal, Huba, & Singer, 1980). In fact, the SSS has been significantly related to many diverse behaviors, such as drug use, alcohol use, and criminality (Bates, White, & Labouvie, 1985; Huba, Newcomb, & Bentler, 1981; White, Labouvie, & Bates, 1985; Zuckerman, 1983). Unfortunately, the majority of this research on sensation seeking has been conducted with cross-sectional samples of college students, which severely limits causal interpretations and generalization of results. Also, studies that explore the influence of adolescent sensation-seeking needs on adult behavior are lacking.
The Present Study An appropriate and accurate conceptualization of adolescent deviant behaviors has important implications not only for understanding the causal factors leading to potential problems, but also for establishing effective intervention, prevention, and treatment programs. If certain behaviors are highly associated, the efficacy and usefulness of targeting only one behavior to the exclusion of others is questionable (Barnes & Welte, 1986). If Sensation Seeking contributes to these problem behaviors, then therapeutic environments and treatment plans could help meet these needs in less injurious and perhaps more socially approved ways. According to Zuckerman (1979), Sensation Seeking is a displaceable motive that can be satisfied by a variety of activities. This characteristic of Sensation Seeking suggests one way to change behavior through substitution of stimulating activities. Our longitudinal data provided the opportunity to explore across-time influences of adolescent sensation-seeking needs on young adult deviant behavior. Our primary method of analysis was latent-variable structural equation models (SEMs; Bentler, 1980; Bentler & Newcomb, 1986; Newcomb, 1990). This method allows examination of relationships among constructs that are unobserved (latent) and not influenced by errors of measurement. Latent variables capture the common portion of the covariation among multiple observed indicators of the construct. SEMs are particularly useful in longitudinal research (Bentler, 1980). In addition, we examined possible nonstandard effects; these are causal paths that involve observed (indicator) variables and that are not restricted to between latent constructs only (Newcomb, 1988, 1990; Newcomb & Bentler, 1988a, 1988b). On the basis of existing literature (e.g., Donovan & Jessor, 1985), we selected several measures of deviant behaviors and attitudes to capture various aspects of this problem behavior syndrome, and we used these as indicators of a latent construct of General Deviance. A latent variable was also hypothesized to represent the construct of Sensation Seeking, as reflected by the four observed SSS subscales of boredom susceptibility, disinhibition, experience seeking, and thrill and adventure seeking. We used longitudinal latent-variable structural models to examine the developmental course of the General Deviance construct and the potential impact of Sensation Seeking on this course. At the same time, we considered the predictive relationships among the specific measured (observed) variables that reflect these latent factors.
MICHAEL D. NEWCOMB AND LINDA McGEE
616 Method
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Subjects Data were from 595 participants and were gathered in Year 4, Year 5, and Year 9 follow-ups of a prospective study of adolescent growth and development (see Newcomb & Bentler, 1988a). Year 4 was the first wave in which extensive information on deviant activities was assessed and was the only occasion the SSS subscales were included. Year 4 data were collected in 1979, when subjects were high school students in Grades 10-12. Year 5 data were gathered in 1980, when subjects were in late adolescence, either recently graduated or in the last 2 years of high school. Year 9 data were collected 4 years later in 1984, when participants were in their early 20s. Table 1 presents a demographic description of the 595 subjects. Newcomb and Bentler (1988a) compared the characteristics of these participants in Year 9 with data from other national and large surveys of young adults. They noted very similar patterns of characteristics related to life activity and living arrangements. The main difference was that our sample had a greater percentage of women than men, as it had since the study began. Newcomb and Bentler (1988a) conducted an extensive series of attrition analyses to determine whether the attrition in sample size from Year 1 (1976) to Year 9 (1984) was due to any systematic influences. Those who participated in 1984 were compared with those who were not assessed in 1984 in terms of 38 measures from the 1976 data set. Using the Bonferroni procedure to adjust for simultaneous comparisons, Newcomb and Bentler found that none of the 38 variables were able to differentiate significantly the new sample from those lost, at the .05 level of significance (see Newcomb and Bentler, 1988a, for a detailed description of further attrition analyses).
Measures Variables were chosen to reflect the latent constructs of Sensation Seeking and General Deviance. Our approach to structural equation models allowed us to test the relationships among unobserved (latent) constructs and thereby capture common factors and also to incorporate the unique components of each variable as captured in the residuals or between-measured variables and latent factors (Newcomb, 1990; Newcomb & Bentler, 1988a, 1988b). SSX A short 16-item version of Zuckerman's SSS was included at Year 4 (Huba, Newcomb, & Bentler, 1981). Individuals were asked to answer, "How often do you feel the following way?" on an anchored 5-point scale (1 = never, 2 = rarely, 3 = sometimes, 4 = usually, 5 =
always). Zuckerman's original scale consists of 10 forced-choice items for each of the four subscales. Huba et al. (1981) selected 4 positive choice statements from each subscale on the basis of high factor loadings in both American and British samples and also on the basis of relevance to adolescents. Any item related to alcohol or drug use was excluded. A latent construct of Sensation Seeking was hypothesized to reflect the four observed subscales: boredom susceptibility, disinhibition, experience seeking, and thrill and adventure seeking. General deviance. At least five measures were chosen to reflect the latent construct of General Deviance at each of the 3 years: licit and illicit substance use, sexual events, attitudes toward law abidance, religiosity, and specific deviant behaviors. Each of these measures is described below. Licit and illicit substance use. Frequency of drug use measures were obtained for cigarettes, alcohol (beer, wine, and liquor), cannabis (marijuana and hashish), and hard drugs (e.g., hypnotics, cocaine, stimulants, psychedelics, inhalants, narcotics, and PCP). Frequency of drug use for the previous 6 months was rated on 7-point anchored scales, with never (1), once (2), a few times (3), once a month (4), once a week (5), once a day (6), and more than once a day (7). For the present analyses, substance use items were summed into two scales: licit drug use (cigarettes and alcohol) and illicit drug use (cannabis and hard drugs). Sexual events. Sexual behavior was assessed by a 7-item scale in which the subjects were asked to indicate in a dichotomous manner (0 = no and 1 = yes) whether they had experienced an event (Newcomb, Huba, & Bentler, 1981). For Year 4 data, lifetime occurrence was assessed, whereas for Year 5, occurrence was limited to the past year, and in Year 9 occurrence was for the previous 6 months. These events included falling in love, getting pregnant or getting someone pregnant, getting or giving venereal disease, breaking up with a boyfriend or girlfriend, going out on dates regularly, having a gay or lesbian experience, and losing virginity. Responses to these seven items were summed into one scale. Religiosity. The Religiosity scale consisted of four items, each with a 5-choice semantic differential response format with opposite adjective phrases as anchor points. We asked the subjects, "Which describes you most of the time?" (Castro, Maddahian, Newcomb, & Bentler, 1987; Newcomb & Bentler, 1988a, 1988b; Stein, Newcomb, & Bentler, 1986). These adjective phrases were: (a) am not religious-am a religious person, (b) believe in religion or the Bible-believe in science, (c) feel that prayers are answered-feel that praying is a waste, and (d) think religion is outdated-think religion is not outdated. In creating a total score, the keying was reversed on Items b and c. Law abidance. For the Law Abidance scale, we used the same se-
Table 1 Description of Sample Variable
Number of subjects
% of sample
Total Sex Male Female Ethnicity Black Hispanic White Asian Grade level at Year 4 10th llth 12th
595
100
179 416
30 70
82 80 377 56
14 13 64 9
266 208 121
45 35 20
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
GENERAL DEVIANCE AND SPECIFIC PROBLEM BEHAVIORS mantic differential format and the same instructions to subjects as the Religiosity scale discussed above. This scale consisted of four items: (a) return incorrect change-willing to keep extra change, (b) might use a false ID-afraid of getting caught, (c) might shoplift-wouldn't know how or want to, (d) goodie-goodie honest type-not quite so honest. Items 1 and 4 were reverse coded (Castro et al, 1987; Newcomb & Bentler, 1988a, 1988b; Stein et al, 1986). Specific deviant behaviors. The measures of specific deviant behaviors evolved across time to take into account the increasing age of the subjects. During \ear 4, a measure of deviance events was used that consisted of three dichotomous items (Newcomb, Huba, & Bentler, 1981): (a) got in trouble with the law, (b) stole something valuable, and (c) got in trouble at school. These items, as well as a more complete measure of criminal activities, were also measured during Year 5. The three deviance items were dropped in Year 9, and the measure of criminal activities was included again. This pattern of measuring deviant behaviors over the study allowed the measures of adolescent deviant behavior to be compared with the measures of young adult criminal activities. In Year 5 and Year 9, the criminal activities measures asked how often during the prior 6 months the respondents had committed each of 16 acts that ranged from minor fights to major acts of vandalism (Huba & Bentler, 1984). On the basis of factor analyses, these items were summed into three scales that reflected confrontational acts, stealing episodes, and property damage. Confrontational acts consisted of the number of times an individual got into serious fights, got into gang fights, caused a serious injury, and used a deadly weapon against someone. Stealing episodes included the number of times the adolescent had stolen something worth less than $50, stolen something worth more than $50, and shoplifted. Property damage included the number of times the teenager committed vandalism at school or at work, committed arson, or trespassed (Newcomb & Bentler, 1988a).
Analysis In the initial analyses, we examined the means and prevalence rates with regard to sex differences and changes over time. We then presented the cross-sectional intercorrelations for measures within Year 4, Year 5, and Year 9. The prevalence and correlational analyses are presented separately by sex. The principal method for studying the longitudinal relationships was latent-variable structural equation models (Bentler, 1980; Bentler & Newcomb, 1986; Newcomb, 1988,1990; Newcomb & Bentler, 1988a, 1988b). All structural model analyses were performed using the EQS computer program (Bentler, 1989). Initially, the adequacy of the variables chosen to reflect the latent factors was tested by using a confirmatory factor analysis. After this, we used across-time structural models to examine the plausible causal effects between sensation seeking and deviance and between various specific types of deviance.
Results Means, Prevalence Rates, and Sex
Differences
The initial analyses examined separately the univariate distributions of all of the variables for men and women. Prevalence rates represent any reported licit or illicit drug use, engagement in any of the various delinquent or criminal activities, or occurrence of any sexual event (see Measures section). No prevalence rates are reported for law abidance, religiosity, or any of the SSS subscales because there was no meaningful way to dichotomize them. The prevalence rates for men and women, related chi-square significance tests, means for men and
617
women, and associated point-biserial correlations to test for mean sex differences are presented in Table 2. In Year 4, 87.2% of the men and significantly fewer of the women (80.0%) reported use of licit drugs (alcohol or cigarettes), and about one half of both the men and the women reported illicit drug use. Over 80% of both sexes reported at least one sexual event. Significantly more men than women engaged in deviance events. Means were significantly higher for women than for men on law abidance and religiosity and lower on deviance events. No sex differences were found for prevalence or frequency of sexual events. On the SSS, the women had significantly higher scores on the Experience Seeking subscale than the men, whereas the men had significantly higher scores on the Thrill and Adventure Seeking and Disinhibition scales than the women. No sex difference was found on the Boredom Susceptibility scale. In Year 5, women continued to have higher mean scores for law abidance and religiosity, whereas men continued to be higher in deviance events. At this point the prevalence of sexual events had become significantly higher for the women. Licit drug use remained almost the same as in \fear 4. Prevalence of illicit drug use did not change from Year 4, although mean levels of illicit drug use almost doubled for both men and women over the 1-year period. This variation may reflect a plateau of illicit drug use initiation, in conjunction with greater involvement among those who began their use earlier. On all of the criminal activity subscales (first measured in "Year 5), both prevalence rates and means were significantly higher for men than for women. In Year 9, the highest prevalence rate was for licit drug use for both male subjects (92.2%) and female subjects (90.1%). The mean level of licit drug use became significantly higher for male subjects than for female subjects. On all of the criminal activity scales, prevalence rates and means continued to be significantly higher for male subjects.
Cross-Sectional Bivariate Analyses The correlations among all "Year 4, \fear 5, and \ear 9 variables were examined separately by sex. "Year 4 correlations are presented in Table 3, Year 5 correlations are in Table 4, and Year 9 correlations are in Table 5. In the tables, the correlations for the male subjects are above the diagonal and those for female subjects are below the diagonal. Year 4. As shown in Table 3, during "Yfear 4, one or more of the four SSS subscales correlated significantly and positively with licit drug use, illicit drug use, sexual events, and deviance events, and one or more of the scales was significantly negatively correlated with religious commitment and law abidance. Licit drug use (alcohol and cigarettes) correlated with all of the SSS subscales for male and female subjects, with the exception that experience seeking did not correlate with licit drug use for male subjects. Illicit drug use correlated with all SSS subscales for female subjects, while for male subjects, only one of the SSS subscales (disinhibition) correlated with illicit drug use. Sexual events correlated with disinhibition and boredom susceptibility for female subjects, and all SSS subscales, except experience seeking, for male subjects. Finally, deviance events correlated
618
MICHAEL D. NEWCOMB AND LINDA McGEE Table 2 Prevalence and Means for All Variables
M
Men
Women
Sex difference X2'
87.2 52.5
80.0 51.9
3.84* 0.00
— —
— —
Prevalence (%)
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Variable Year 4 Licit drug use Illicit drug use Law abidance0 Religiosity0 Sexual events Deviance events Experience seeking0 Thrill/adventure0 Disinhibitionc Boredom susceptibility0 YearS Licit drug use Illicit drug use Law abidance Religiosity Sexual events Deviance events Confrontational acts Stealing episodes Property damage Year 9 Licit drug use Illicit drug use Law abidance Religiosity Sexual events Confrontational acts Stealing episodes Property damage
'df=l.
81.6 39.1
85.3 25.2
— — — —
— — — —
87.2 55.9
82.7 51.0
— —
— — 80.5
70.4 31.3 24.6 45.3 29.1 92.2 48.6
90.1 50.7
— —
52.2
48.6 17.3 33.0 15.1
Women (n = 416)
9.16 10.21 12.18 14.54 1.95 0.63 13.21 12.16 12.18 11.57
9.43 10.38 13.39 15.81 2.15 0.35 13.80 10.79 10.76 11.65
9.98 20.58 12.06 14.97 1.54 0.40 0.49 1.75 0.65
9.71 20.28 13.47 15.96 1.73 0.15 0.31 0.87 0.29
-.02 -.02 .16*** .12**
12.21 20.33 13.45 14.91 0.70 0.38 0.83 0.33
11.23 19.82 14.43 15.97 0.91 0.14 0.35 0.08
-.09* -.05 .13** .12** .09* -.14*** -.16*** -.15***
— — 1.07
10.93***
— — — — 1.53 1.01
— —
6.80** 30.42*** 6.12** 13.87*** 22.63***
12.0 15.6 29.1 12.5
0.40 0.15
— —
— — 9.6
0.50 6.35** 28.36*** 10.18***
13.7
6.5
Sex difference point-biserial correlationb
Men (H = 179)
.03 .02 .14*** .15***
.07 -.17*** .11** -.17*** -.19***
.03
.07 -.22*** -.08* -.17*** -.14***
b
A positive point-biserial correlation indicates that the female subjects had the higher values. Prevalence rates are not appropriate for these variables. *p
Journal of Personality and Social Psychology 1991, Vol. 61, No. 4, 614-628
Influence of Sensation Seeking on General Deviance and Specific Problem Behaviors From Adolescence to \bung Adulthood Michael D. Newcomb and Linda McGee
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Division of Counseling and Educational Psychology University of Southern California Indicators of a Sensation Seeking latent construct were assessed during adolescence and used to predict changes in repeatedly gathered measures of a latent construct of General Deviance. A community sample of 595 male and female Ss was assessed 3 times over a 5-year period from late adolescence to young adulthood. Most Ss reported use of licit drugs, about one half had tried illicit drugs, and a substantial minority had engaged in other delinquent or criminal activities. The General Deviance construct was stable over time, although specific cross-effects were found. Sensation Seeking was moderately correlated with General Deviance at all 3 levels but did not predict directly the General Deviance construct over time; the effects of Sensation Seeking on later deviant behavior and attitudes were specific rather than general.
General and Specific Deviance
Crime, delinquency, and deviance among teenagers may take several specific forms, such as drug and alcohol use or abuse, vandalism, disciplinary problems in school, sexual precocity, or aggression. These may also reflect certain nontraditional attitudes, such as low religiosity, rebelliousness, and disregard for the law. These types of behavior and attitudes have often been observed among teenagers and seem to reflect a syndrome and latent construct of problem behavior or General Deviance (Donovan & Jessor, 1985; Osgood, Johnston, O'Malley, & Bachman, 1988). However, the developmental stability and integrity of this construct are not well established. The extent of any reciprocal causal effects among the specific deviant attitudes and behavior over time has rarely been studied. Little attention has been devoted to understanding what factors affect this general construct of deviance over time distinct from the constituent variables. We examined these issues empirically using multiwave panel data. First, we examined the stability and integrity of a latent construct of General Deviance and reciprocal influences among its constituent components in three waves of data over a 5-year period from late adolescence to young adulthood. Second, some suggest that personality or individual differences may be responsible for the initiation and maintenance of these deviant attitudes and behaviors (Huba & Bentler, 1983). Sensation seeking has often been suggested as one such possible personal characteristic, and we examined its effects on various types of problem behavior from late adolescence to young adulthood.
The transition from adolescence to young adulthood is a particularly critical stage of development for successful acquisition and performance of adult roles (e.g., Havighurst, 1972; Newcomb, 1987; Newcomb & Bentler, 1988a). This period often involves experimentation with new attitudes and behaviors, several of which may conflict with traditional society and be seen as problems. Problem behavior has been defined by Jessor and Jessor (1977) as "behavior that is socially defined as a problem, a source of concern, or as undesirable. . . and its occurrence usually elicits some kind of social control response" (p. 33). According to problem behavior theory, a variety of behaviors considered problematic during adolescence are associated with one another (Donovan & Jessor, 1985; Jessor & Jessor, 1977). If these various deviant behaviors are generated by one underlying motive or propensity, they can be construed as a general syndrome of deviance (Donovan & Jessor, 1985). Empirical data confirm a substantial, positive covariation among diverse adolescent problem behaviors such as illicit drug use, general delinquent behavior (e.g., cheating, lying, petty theft, and vandalism), problem drinking, precocious sexual activity, and other rebellious behaviors. Even frequent use of licit drugs (cigarettes and alcohol) constitutes part of this general deviance syndrome, which involves a wide range of norm-violating behaviors (Bachman, O'Malley, & Johnston, 1978; Donovan & Jessor, 1985; Elliott & Ageton, 1976; Elliott & Huizinga, 1984; Hindelang & Weis, 1972; Jessor & Jessor, 1977; Kandel, 1980; Robbins & Murphy, 1967). These repeated observations suggest that a common set of causes may underlie these specific behaviors (Barnes & Welte, 1986; Elliott, Huizinga, & Ageton, 1985; Huba & Bentler, 1983). Donovan and Jessor (1985) demonstrated that several measures of deviant behavior were generated by an underlying tendency toward or latent factor of unconventionality for both adolescents and college students. This general syndrome of problem behavior among teenagers seems to be a meaningful, reliable,
This research was supported by Grant DA01070 from the National Institute on Drug Abuse. We warmly acknowledge the production assistance of Tove Davison and Julie Speckart. Correspondence concerning this article should be addressed to Michael D. Newcomb, Division of Counseling and Educational Psychology, University of Southern California, Los Angeles, California 90089-0031.
614
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
GENERAL DEVIANCE AND SPECIFIC PROBLEM BEHAVIORS
and valid construct during adolescence, but may be less relevant for other stages in life (e.g., Donovan & lessor, 1985). It may diversify and fragment during the transition from older adolescence to young adulthood. Osgood et al. (1988) examined the longitudinal structural relationships among several repeated measures of deviant behavior (criminal behavior, drug use, and problem driving). A common factor or underlying (latent) construct of these deviant behaviors remained stable over a 4-year period from age 18 to 22. This construct of General Deviance was stable over time, and different types of specific deviance (i.e., problem drinking) were also stable over time, but there were few reciprocal effects between specific behaviors over time. That is, one type of deviance had little unique effect on other types of deviance over time. Osgood et al. (1988) concluded that "theories that treat different deviant behaviors as alternative manifestations of a single general tendency can account for some, but not for all, of the meaningful variance in these behaviors" (p. 81). Newcomb and Bentler (1988a) found decreasing cohesiveness among deviant attitudes and behavior as their adolescent population grew into young adulthood, although the specific types of deviance were moderately stable. This suggests that the syndrome of problem behaviors present in an adolescent population may not remain a unified construct in young adulthood or later life. This construct seemed to differentiate into at least two sets of deviant behavior over time, one set of behaviors including problems related to drug use and another set including other forms of deviance. This also supports the conclusion of Osgood et al. (1988) that deviant behaviors become more "specialized" as the adolescent moves into young adulthood.
Sensation Seeking During the past 2 decades, Sensation Seeking has emerged as an important explanatory construct for a variety of behaviors. The sensation-seeking trait has been defined by Zuckerman (1979) as "the need for varied, novel, and complex sensations and experiences and the willingness to take physical and social risks for the sake of such experiences" (p. 10). The theoretical basis for Sensation Seeking as a biosocial dimension of personality has changed over the years to incorporate new biological findings. Zuckerman originally proposed that individuals with high sensation-seeking needs feel and function best at a high level of cortical arousal and that they are motivated to engage in activities that increase their arousal level. He later rejected this optimal arousal basis for Sensation Seeking and reduced arousal to a secondary role. He now hypothesizes that Sensation Seeking is a general sensitivity to reinforcement that may reflect levels of catecholamines and their enzymes (Zuckerman, 1983,1984; Zuckerman, Buchsbaum, & Murphy, 1980). Zuckerman (1979) has documented an impressive array of correlations between the Sensation-Seeking Scale (SSS) he developed and a variety of psychological measures. For instance, individuals high on the SSS are more oriented to body sensation, extraverted, impulsive, antisocial, nonconformist, and less anxious (Zuckerman & Link, 1968). It is, therefore, reasonable to suspect that these psychological associations with the
615
SSS are reflected in correlations between the SSS and analogous overt behavior (Segal, Huba, & Singer, 1980). In fact, the SSS has been significantly related to many diverse behaviors, such as drug use, alcohol use, and criminality (Bates, White, & Labouvie, 1985; Huba, Newcomb, & Bentler, 1981; White, Labouvie, & Bates, 1985; Zuckerman, 1983). Unfortunately, the majority of this research on sensation seeking has been conducted with cross-sectional samples of college students, which severely limits causal interpretations and generalization of results. Also, studies that explore the influence of adolescent sensation-seeking needs on adult behavior are lacking.
The Present Study An appropriate and accurate conceptualization of adolescent deviant behaviors has important implications not only for understanding the causal factors leading to potential problems, but also for establishing effective intervention, prevention, and treatment programs. If certain behaviors are highly associated, the efficacy and usefulness of targeting only one behavior to the exclusion of others is questionable (Barnes & Welte, 1986). If Sensation Seeking contributes to these problem behaviors, then therapeutic environments and treatment plans could help meet these needs in less injurious and perhaps more socially approved ways. According to Zuckerman (1979), Sensation Seeking is a displaceable motive that can be satisfied by a variety of activities. This characteristic of Sensation Seeking suggests one way to change behavior through substitution of stimulating activities. Our longitudinal data provided the opportunity to explore across-time influences of adolescent sensation-seeking needs on young adult deviant behavior. Our primary method of analysis was latent-variable structural equation models (SEMs; Bentler, 1980; Bentler & Newcomb, 1986; Newcomb, 1990). This method allows examination of relationships among constructs that are unobserved (latent) and not influenced by errors of measurement. Latent variables capture the common portion of the covariation among multiple observed indicators of the construct. SEMs are particularly useful in longitudinal research (Bentler, 1980). In addition, we examined possible nonstandard effects; these are causal paths that involve observed (indicator) variables and that are not restricted to between latent constructs only (Newcomb, 1988, 1990; Newcomb & Bentler, 1988a, 1988b). On the basis of existing literature (e.g., Donovan & Jessor, 1985), we selected several measures of deviant behaviors and attitudes to capture various aspects of this problem behavior syndrome, and we used these as indicators of a latent construct of General Deviance. A latent variable was also hypothesized to represent the construct of Sensation Seeking, as reflected by the four observed SSS subscales of boredom susceptibility, disinhibition, experience seeking, and thrill and adventure seeking. We used longitudinal latent-variable structural models to examine the developmental course of the General Deviance construct and the potential impact of Sensation Seeking on this course. At the same time, we considered the predictive relationships among the specific measured (observed) variables that reflect these latent factors.
MICHAEL D. NEWCOMB AND LINDA McGEE
616 Method
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Subjects Data were from 595 participants and were gathered in Year 4, Year 5, and Year 9 follow-ups of a prospective study of adolescent growth and development (see Newcomb & Bentler, 1988a). Year 4 was the first wave in which extensive information on deviant activities was assessed and was the only occasion the SSS subscales were included. Year 4 data were collected in 1979, when subjects were high school students in Grades 10-12. Year 5 data were gathered in 1980, when subjects were in late adolescence, either recently graduated or in the last 2 years of high school. Year 9 data were collected 4 years later in 1984, when participants were in their early 20s. Table 1 presents a demographic description of the 595 subjects. Newcomb and Bentler (1988a) compared the characteristics of these participants in Year 9 with data from other national and large surveys of young adults. They noted very similar patterns of characteristics related to life activity and living arrangements. The main difference was that our sample had a greater percentage of women than men, as it had since the study began. Newcomb and Bentler (1988a) conducted an extensive series of attrition analyses to determine whether the attrition in sample size from Year 1 (1976) to Year 9 (1984) was due to any systematic influences. Those who participated in 1984 were compared with those who were not assessed in 1984 in terms of 38 measures from the 1976 data set. Using the Bonferroni procedure to adjust for simultaneous comparisons, Newcomb and Bentler found that none of the 38 variables were able to differentiate significantly the new sample from those lost, at the .05 level of significance (see Newcomb and Bentler, 1988a, for a detailed description of further attrition analyses).
Measures Variables were chosen to reflect the latent constructs of Sensation Seeking and General Deviance. Our approach to structural equation models allowed us to test the relationships among unobserved (latent) constructs and thereby capture common factors and also to incorporate the unique components of each variable as captured in the residuals or between-measured variables and latent factors (Newcomb, 1990; Newcomb & Bentler, 1988a, 1988b). SSX A short 16-item version of Zuckerman's SSS was included at Year 4 (Huba, Newcomb, & Bentler, 1981). Individuals were asked to answer, "How often do you feel the following way?" on an anchored 5-point scale (1 = never, 2 = rarely, 3 = sometimes, 4 = usually, 5 =
always). Zuckerman's original scale consists of 10 forced-choice items for each of the four subscales. Huba et al. (1981) selected 4 positive choice statements from each subscale on the basis of high factor loadings in both American and British samples and also on the basis of relevance to adolescents. Any item related to alcohol or drug use was excluded. A latent construct of Sensation Seeking was hypothesized to reflect the four observed subscales: boredom susceptibility, disinhibition, experience seeking, and thrill and adventure seeking. General deviance. At least five measures were chosen to reflect the latent construct of General Deviance at each of the 3 years: licit and illicit substance use, sexual events, attitudes toward law abidance, religiosity, and specific deviant behaviors. Each of these measures is described below. Licit and illicit substance use. Frequency of drug use measures were obtained for cigarettes, alcohol (beer, wine, and liquor), cannabis (marijuana and hashish), and hard drugs (e.g., hypnotics, cocaine, stimulants, psychedelics, inhalants, narcotics, and PCP). Frequency of drug use for the previous 6 months was rated on 7-point anchored scales, with never (1), once (2), a few times (3), once a month (4), once a week (5), once a day (6), and more than once a day (7). For the present analyses, substance use items were summed into two scales: licit drug use (cigarettes and alcohol) and illicit drug use (cannabis and hard drugs). Sexual events. Sexual behavior was assessed by a 7-item scale in which the subjects were asked to indicate in a dichotomous manner (0 = no and 1 = yes) whether they had experienced an event (Newcomb, Huba, & Bentler, 1981). For Year 4 data, lifetime occurrence was assessed, whereas for Year 5, occurrence was limited to the past year, and in Year 9 occurrence was for the previous 6 months. These events included falling in love, getting pregnant or getting someone pregnant, getting or giving venereal disease, breaking up with a boyfriend or girlfriend, going out on dates regularly, having a gay or lesbian experience, and losing virginity. Responses to these seven items were summed into one scale. Religiosity. The Religiosity scale consisted of four items, each with a 5-choice semantic differential response format with opposite adjective phrases as anchor points. We asked the subjects, "Which describes you most of the time?" (Castro, Maddahian, Newcomb, & Bentler, 1987; Newcomb & Bentler, 1988a, 1988b; Stein, Newcomb, & Bentler, 1986). These adjective phrases were: (a) am not religious-am a religious person, (b) believe in religion or the Bible-believe in science, (c) feel that prayers are answered-feel that praying is a waste, and (d) think religion is outdated-think religion is not outdated. In creating a total score, the keying was reversed on Items b and c. Law abidance. For the Law Abidance scale, we used the same se-
Table 1 Description of Sample Variable
Number of subjects
% of sample
Total Sex Male Female Ethnicity Black Hispanic White Asian Grade level at Year 4 10th llth 12th
595
100
179 416
30 70
82 80 377 56
14 13 64 9
266 208 121
45 35 20
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
GENERAL DEVIANCE AND SPECIFIC PROBLEM BEHAVIORS mantic differential format and the same instructions to subjects as the Religiosity scale discussed above. This scale consisted of four items: (a) return incorrect change-willing to keep extra change, (b) might use a false ID-afraid of getting caught, (c) might shoplift-wouldn't know how or want to, (d) goodie-goodie honest type-not quite so honest. Items 1 and 4 were reverse coded (Castro et al, 1987; Newcomb & Bentler, 1988a, 1988b; Stein et al, 1986). Specific deviant behaviors. The measures of specific deviant behaviors evolved across time to take into account the increasing age of the subjects. During \ear 4, a measure of deviance events was used that consisted of three dichotomous items (Newcomb, Huba, & Bentler, 1981): (a) got in trouble with the law, (b) stole something valuable, and (c) got in trouble at school. These items, as well as a more complete measure of criminal activities, were also measured during Year 5. The three deviance items were dropped in Year 9, and the measure of criminal activities was included again. This pattern of measuring deviant behaviors over the study allowed the measures of adolescent deviant behavior to be compared with the measures of young adult criminal activities. In Year 5 and Year 9, the criminal activities measures asked how often during the prior 6 months the respondents had committed each of 16 acts that ranged from minor fights to major acts of vandalism (Huba & Bentler, 1984). On the basis of factor analyses, these items were summed into three scales that reflected confrontational acts, stealing episodes, and property damage. Confrontational acts consisted of the number of times an individual got into serious fights, got into gang fights, caused a serious injury, and used a deadly weapon against someone. Stealing episodes included the number of times the adolescent had stolen something worth less than $50, stolen something worth more than $50, and shoplifted. Property damage included the number of times the teenager committed vandalism at school or at work, committed arson, or trespassed (Newcomb & Bentler, 1988a).
Analysis In the initial analyses, we examined the means and prevalence rates with regard to sex differences and changes over time. We then presented the cross-sectional intercorrelations for measures within Year 4, Year 5, and Year 9. The prevalence and correlational analyses are presented separately by sex. The principal method for studying the longitudinal relationships was latent-variable structural equation models (Bentler, 1980; Bentler & Newcomb, 1986; Newcomb, 1988,1990; Newcomb & Bentler, 1988a, 1988b). All structural model analyses were performed using the EQS computer program (Bentler, 1989). Initially, the adequacy of the variables chosen to reflect the latent factors was tested by using a confirmatory factor analysis. After this, we used across-time structural models to examine the plausible causal effects between sensation seeking and deviance and between various specific types of deviance.
Results Means, Prevalence Rates, and Sex
Differences
The initial analyses examined separately the univariate distributions of all of the variables for men and women. Prevalence rates represent any reported licit or illicit drug use, engagement in any of the various delinquent or criminal activities, or occurrence of any sexual event (see Measures section). No prevalence rates are reported for law abidance, religiosity, or any of the SSS subscales because there was no meaningful way to dichotomize them. The prevalence rates for men and women, related chi-square significance tests, means for men and
617
women, and associated point-biserial correlations to test for mean sex differences are presented in Table 2. In Year 4, 87.2% of the men and significantly fewer of the women (80.0%) reported use of licit drugs (alcohol or cigarettes), and about one half of both the men and the women reported illicit drug use. Over 80% of both sexes reported at least one sexual event. Significantly more men than women engaged in deviance events. Means were significantly higher for women than for men on law abidance and religiosity and lower on deviance events. No sex differences were found for prevalence or frequency of sexual events. On the SSS, the women had significantly higher scores on the Experience Seeking subscale than the men, whereas the men had significantly higher scores on the Thrill and Adventure Seeking and Disinhibition scales than the women. No sex difference was found on the Boredom Susceptibility scale. In Year 5, women continued to have higher mean scores for law abidance and religiosity, whereas men continued to be higher in deviance events. At this point the prevalence of sexual events had become significantly higher for the women. Licit drug use remained almost the same as in \fear 4. Prevalence of illicit drug use did not change from Year 4, although mean levels of illicit drug use almost doubled for both men and women over the 1-year period. This variation may reflect a plateau of illicit drug use initiation, in conjunction with greater involvement among those who began their use earlier. On all of the criminal activity subscales (first measured in "Year 5), both prevalence rates and means were significantly higher for men than for women. In Year 9, the highest prevalence rate was for licit drug use for both male subjects (92.2%) and female subjects (90.1%). The mean level of licit drug use became significantly higher for male subjects than for female subjects. On all of the criminal activity scales, prevalence rates and means continued to be significantly higher for male subjects.
Cross-Sectional Bivariate Analyses The correlations among all "Year 4, \fear 5, and \ear 9 variables were examined separately by sex. "Year 4 correlations are presented in Table 3, Year 5 correlations are in Table 4, and Year 9 correlations are in Table 5. In the tables, the correlations for the male subjects are above the diagonal and those for female subjects are below the diagonal. Year 4. As shown in Table 3, during "Yfear 4, one or more of the four SSS subscales correlated significantly and positively with licit drug use, illicit drug use, sexual events, and deviance events, and one or more of the scales was significantly negatively correlated with religious commitment and law abidance. Licit drug use (alcohol and cigarettes) correlated with all of the SSS subscales for male and female subjects, with the exception that experience seeking did not correlate with licit drug use for male subjects. Illicit drug use correlated with all SSS subscales for female subjects, while for male subjects, only one of the SSS subscales (disinhibition) correlated with illicit drug use. Sexual events correlated with disinhibition and boredom susceptibility for female subjects, and all SSS subscales, except experience seeking, for male subjects. Finally, deviance events correlated
618
MICHAEL D. NEWCOMB AND LINDA McGEE Table 2 Prevalence and Means for All Variables
M
Men
Women
Sex difference X2'
87.2 52.5
80.0 51.9
3.84* 0.00
— —
— —
Prevalence (%)
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Variable Year 4 Licit drug use Illicit drug use Law abidance0 Religiosity0 Sexual events Deviance events Experience seeking0 Thrill/adventure0 Disinhibitionc Boredom susceptibility0 YearS Licit drug use Illicit drug use Law abidance Religiosity Sexual events Deviance events Confrontational acts Stealing episodes Property damage Year 9 Licit drug use Illicit drug use Law abidance Religiosity Sexual events Confrontational acts Stealing episodes Property damage
'df=l.
81.6 39.1
85.3 25.2
— — — —
— — — —
87.2 55.9
82.7 51.0
— —
— — 80.5
70.4 31.3 24.6 45.3 29.1 92.2 48.6
90.1 50.7
— —
52.2
48.6 17.3 33.0 15.1
Women (n = 416)
9.16 10.21 12.18 14.54 1.95 0.63 13.21 12.16 12.18 11.57
9.43 10.38 13.39 15.81 2.15 0.35 13.80 10.79 10.76 11.65
9.98 20.58 12.06 14.97 1.54 0.40 0.49 1.75 0.65
9.71 20.28 13.47 15.96 1.73 0.15 0.31 0.87 0.29
-.02 -.02 .16*** .12**
12.21 20.33 13.45 14.91 0.70 0.38 0.83 0.33
11.23 19.82 14.43 15.97 0.91 0.14 0.35 0.08
-.09* -.05 .13** .12** .09* -.14*** -.16*** -.15***
— — 1.07
10.93***
— — — — 1.53 1.01
— —
6.80** 30.42*** 6.12** 13.87*** 22.63***
12.0 15.6 29.1 12.5
0.40 0.15
— —
— — 9.6
0.50 6.35** 28.36*** 10.18***
13.7
6.5
Sex difference point-biserial correlationb
Men (H = 179)
.03 .02 .14*** .15***
.07 -.17*** .11** -.17*** -.19***
.03
.07 -.22*** -.08* -.17*** -.14***
b
A positive point-biserial correlation indicates that the female subjects had the higher values. Prevalence rates are not appropriate for these variables. *p
