believing in the benefits of the American Dream for personal fulfillment. These results were interpreted as supporting Dienstbier's theory of moral development.
Journal orPereonality and Social Psychology 1985, Vol. 49, No. 4,'1114-1121
Copyright 1985 by the American Psychological Association, Inc. 0022-3514/85/100.75
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.
Social Contract and Social Integration in Adolescent Development W. Scott Hilles
Lynn R. Kahle
Duke University
University of Oregon
Eighty-nine subjects from two high schools were tested during the spring of their sophomore and senior years, when their mean ages were 16 years, 1 month, and 18 years, 1 month, respectively. Composites measured social contract with: (a) independence, (b) implicit social contract, societal norms and expectations, and (c) explicit social contracts, rules. Composites and single items measured social integration with: (d) role commitment, (e) social-American Dream, accepting the belief in the American Dream that hard work would lead to social success, ( f ) self-American Dream, belief that hard work will produce personal satisfaction and success, (g) raw deal, perceptions of being treated unfairly, (h) self-blame, and (i) feelings of hopelessness. The results of the cross-lagged panel correlations generally support the hypothesis that students respond to implicit social contracts through role commitment, which is further expressed by a belief in the American Dream for social fulfillment, while responding to the perception of explicit social contracts by not believing in the benefits of the American Dream for personal fulfillment. These results were interpreted as supporting Dienstbier's theory of moral development.
Dienstbier, D. Hillman, Lehnhoff, J. Hillman, and Valkenaar (1975) have articulated a theory of the development of individual differences in moral behavior, which dialectically integrates cognitive and social learning theories about moral thinking and emotion through an examination of mechanisms of social control. Dienstbier et al. identified two theoretically crucial mechanisms of social control: (a) shame, which results when socializing agents draw attention to themselves as the source of negative emotional arousal (usually through severe punishment) after some infraction, and (b) guilt, which results when socializing agents draw attention to the infraction of the perpetrator as the source of negative emotional arousal after some infraction. Externally oriented shame is considered less effective than internally oriented guilt for inducing moral thought and for controlling subsequent moral behavior in situations in which detection is unlikely. The prediction of moral behavior then becomes based on the extent of internalization of society's norms, with greater inter-
The authors appreciate comments from Dave Klingel and Debra Eisert. Requests for reprints should be sent to Lynn R. Kahle, 388 Gilbert Hall, University of Oregon, Eugene, Oregon 97403-1208.
nalization implying greater cooperation and integration with the goals of society. If one perceives one's relationship with society as based on an internally oriented social contract (cf. Douvan, 1974), behavior ought to be more integrated with society than if an externally oriented, explicit social contract is perceived as ruling behavior. Dienstbier et al. have shown how these simple assumptions can integrate much of what we have learned from moral development and socialization research. Dienstbier et al. (1975) tested their theory by inducing 8-year-old twins, run separately, to promise to monitor a slot car, which was slowly circling a track, in order to stop the car if it approached a dangerous speed. When the twin became distracted, the slot car sped up, flew off the track, and crashed. The experimenter returned immediately to discover the twin's transgression. The experimenter told one twin in each pair, "I bet you feel bad now that the car fell off," focusing attention on the infraction (guilt); he told the other twin, "I bet you feel bad that I found out the car fell off," focusing attention on the detection of the infraction by the socializing agent (shame). When subsequently placed in a situation in which "undetectable" moral behavior was possible, the guilty twin was twice as likely to exhibit moral behavior in any given time unit
1114
1115
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.
SOCIAL CONTRACT
as was the shamed twin. Thus, the Dienstbier et at. theory seems plausible in this context. Additional support for the theory has been found in a test-cheating situation with college students (Dienstbier, Kahle, Willis, & Tunnell, 1980) and in other research paradigms (Dienstbier, 1978). Several aspects of the research this theory has generated, however, have been incomplete. First, the research has not investigated the behavior of adolescents, a group whose moral behavior has been a source of social concern. For example, adolescents lead all other age groups in many categories of crime, including initiation into homicide. Second, the implied link between social control and social integration, which is a goal of the socialization process, has not been tested directly, although this link is central to the viability of the Dienstbier et al. theory. Third, studies to date have relied on short-l,erm laboratory deception studies, which have been the subject of considerable criticism in recent years (e.g., Silverman, 1977). Regardless of whether one objects to these types of studies, evidence from a different paradigm that converges with existing evidence ought to increase our confidence in the phenomenon (Cook & Campbell, 1979). Finally, a correlate of the last two points is that, although the short-term phenomenon of immediate moral behavior has been investigated, the long-term phenomenon of integration into society's goals has not. The research reported here addresses these concerns. Social Contract The present study measured social contract in several ways. Externally oriented social contract was operationalized as rules, and internally oriented social contract was operationalized as expectations. Whereas the violating of rules brings explicit consequences from an external source, the violating of expectations brings more ambiguous and selfadministered consequences. We may venture two different hypotheses about the relation between social contract and social integration. From some theoretical perspectives the clarity of relation between the instrumental response and consequence contained in externally oriented social contracts ought to facilitate social integration more than internally oriented so-
cial contracts; however, the theory of Dienstbier et al. clearly hypothesizes the opposite. Rules and expectations in a sense define society's demands or benefits in social contracts. Two of the individual's contractual demands or benefits also were measured in the present study: personal efficacy and independence. We know from previous research that various aspects of the self concept, such as personal efficacy, are important in adolescent social integration (Eisert & Kahle, 1982; Kahle, Kulka, & Klingel, 1980). Independence provides a particularly interesting construct within the present study because independence may hamper attempts to foster social integration from some perspectives due to psychological reactance (cf. Baumrind, 1975) or may facilitate social integration because independence is such a strongly held value among socially integrated Americans. Social Integration The present study measured social integration in several ways. Role commitment may be viewed as a crucial aspect of adolescence in our society (Katz & Kahn, 1966). "The individual must be able to pick up his cues and play his part" (p. 132). Without playing the part and without role commitment, the individual cannot become socially integrated into the relationships mandated by the larger culture (Douvan, 1974). Another possible manifestation of social integration in the United States is acceptance of that hackneyed aspiration, the American Dream. It stems from the belief that if one possesses some intelligence, works hard, and perhaps has some good fortune, he or she will deservedly attain personal and social fulfillment in the future. The fundamental idea is that one has control over his or her future through the ability to work diligently and the anticipation that fate will reward such work. The anticipated consequences of diligence may be manifested both in social popularity and in self-oriented consequences, such as material success. Belief in this dream often motivates acceptance of the goals necessary for social integration because the dream promises a better future. Dienstbier et al. do not offer an explicit hypothesis about possible sequential relations
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.
1116
W. SCOTT HILLES AND LYNN R. KAHLE
among these variables other than that social contracts should precede social integration. Other theorists, however, have proposed more precise sequences. For example, Hogan (1982) hypothesizes three developmental stages: (a) accommodation to socializing agents, (b) accommodation to peers, and (c) accommodation to institutions. This sequence implies that social contracts should lead to role commitment, which in turn should lead to belief in the institutional value, the American Dream. At least some evidence supports this hypothesized sequence (Kahle, 1983). Method Subjects High school students from a probability sample of two suburban Detroit high schools participated in this survey. The study was conducted by the University of Michigan's Institute for Social Research in the spring of the students' sophomore year, when the students' mean age was 16 years, 1 month, and again in the spring of their senior year, when the mean age was 18 years, I month. The students were largely from working-class families at one school and uppermiddle-class families at the other school. The total number of participants who were surveyed at both times was 89, 43 of whom were male. This small sample size precluded analyses contrasting the two schools; however, Kelly (1979) has thoroughly described differences between these schools.
Survey Schedule The questionnaire measured students' responses to a variety of questions concerning attitudes about school and the future, as well as other informational questions (cf. Kelly, 1979, for a description of an earlier phase of this research). Subjects responded to questions reported here by circling a number from 1 to 7 for all composite questions and from I to 5 for the personal efficacy questions on hopelessness, unfairness, and self-blame. "Very little" corresponded with 1 and "Very great" or "Always" corresponded with a 7 for composite questions and "Always" with 1 and "Never" with 5 for the other three individual questions. Composites were developed from items relevant to the topic of this research using a principal-components factor analysis with varimax rotation. Composites were constructed only when eigenvalues exceeded 1.5. (In some circles eigenvalues are viewed as approximations of how many variables are explained by the factor. Thus, a minimal composite ought to have at least two variables within rounding. The eigenvalue of 1.5 is admittedly arbitrary.) Variables were excluded from the composites if the factor loadings were less than .50. All of the anticipated composites emerged from the factor analysis except the personal efficacy composite. Those items are therefore treated individually. Coefficient alphas were computed on the composites (Cronbach, 1951). An item from each set of questions is reported in Table 1. The item with the largest cor-
rected item-total correlation for each composite was selected to be included in Table 1. Means and standard deviations were also computed. All of these descriptive statistics are reported in Table 2. Names for the composites were determined as follows: Independence measures the degree of control that students would like their teachers to give them as evidenced through teachers' trusting, listening to, and treating students as young adults. Implicit social contract measures the perceptions of the magnitude of societal expectations imposed on students by their school and parents through social norms and interpersonal relationships in school. Explicit social contract measures perceptions of the structure and rules on the students by the teachers and principals. Unlike implicit social contract, this composite does not deal with expectations on students. Role commitment measures the belief that commitment to a role in school; such as becoming an athlete, musician, or newspaper editor; based on involvement in extracurricular activities leads to social success and personal satisfaction. Social-American Dream measures the belief that if one is intelligent, hardworking, and has good fortune then he or she will be socially fulfilled through popularity, admiration and respect, and successful interpersonal relationships, such as in marriage. Self-American Dream measures the belief that intelligence and hard work produce personal fulfillment through personal satisfaction and success in the future, specifically through the achievement of good grades and more generally through achievement in future work. Groups of Pearson product-moment correlations were computed using list-wise deletion, in which all correlations are computed only for subjects who answered all questions in a set of composites. It is important to use list-wise deletion because pair-wise deletion, which only deletes subjects who did not answer certain questions of variables that are being correlated at a certain time, would affect the correlations by giving a sample size that is inconsistent from one coefficient to another, with different subjects answering different questions. On the other hand, a total list-wise strategy using one list for the entire study would exaggerate the effect of missing data. Therefore, each set of correlations was separately computed in a list-wise fashion. (The sample sizes are in Table 3.) Cross-lagged panel correlations (CLPC) were then tested using a Pearson-Filon test for significance of differences between correlated correlations (Kenny, 1975). This methodology could be applied to the six possible first-order correlations between two panel variables, say A and B. Cross-lagged refers to the diagonal correlations between variable A at Time 1 and variable B at Time 2 and between variable B at Time I and variable A at Time 2. These are the most necessary correlations used in testing a CLPC hypothesis, but just these two are not sufficient. The synchronous correlations are the correlations between the two variables at the same time. This relation is important in determining stationarity. The autocorrelations are the correlations of each variable with itself at Time 1 and Time 2. These correlations are also important in CLPC analysis. A CLPC hypothesis predicts the influence of one variable over another variable. The null hypothesis for a CLPC study would be that the two variables are not related nonspu-
1117
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.
SOCIAL CONTRACT
riously, although they still may be correlated. Kenny (1975) has provided a mathematical model to show that with synchronicity, stationarity, and spuriousness, the cross-lagged correlations do not differ. Thus, if stationarity and synchronicity can be established, as in the present study, and if the cross-lagged correlations differ significantly, one can reject the null hypothesis of lack of nonspuriousness (Kahle &Berman, 1979). The test of significance is the Pearson-Filon test Because this test has very low power, it is often advisable to set liberal alpha levels in studies with sample sizes as small as the present study (Kenny, 1975). Here the alpha level was placed a t . 1.
Results Autocorrelations and Synchronous Correlations The pairs of synchronous correlations in Table 3 are not in any case significantly different. This finding implies that the presence of unstationarity, which might suggest a shift in the structural equations between the two time frames, is rendered less plausible. Results
from the program PANAL imply that the assumption of perfect stationarity (Kenny, 1975) was viable here, and it was observed in all of the following analyses. The results in Table 3 also indicate that the autocorrelations for all eight relations do not involve the problem of stability. The difference between the two autocorrelations in the relation in all cases is not significant at the. 1 level, using a Pearson-Filon test for significance. Although this test is not necessary to test for spuriousness (Kenny, 1975), the philosophically more rigorous assumptions of causation (Cook & Campbell, 1979) require equal autocorrelations to avoid certain alternative interpretations about the direction of causal influence (Kenny & Campbell, 1984). A significant difference would suggest instability, meaning that the variable with the larger autocorrelation may spuriously appear to be an effect (Rogosa, 1980). The issue Rogosa raised, therefore, does not provide a viable alternative explanation.
Table 1 Representative Item From Each Set of Questions Set
Sample item
Response range
Independence (5 items)
To what extent would you like for your teachers to have confidence in you and trust you?
Very little (l)-very great (7)
Implicit social contract (16 items)
To what extent does your school expect you to be involved in other extracurricular activities?
Very little (l)-very great (7)
Explicit social contract (7 items)
How true is this at your school? There are too many rules.
Not true (l)-very true (7)
Role commitment (5 items)
Involvement in school activities [leads to] popularity.
Never (l)-al ways (7)
Social-American Dream (7 items)
Getting good grades [leads to] being admired and respected.
Never ( 1 )-always (7)
Self-American Dream (8 items)
Working hard at school work [leads to] success in future work.
Never(l)-always(7)
Raw deal (1 item)
How often is each true for you? I feel I get a raw deal out of life.
Always ( I )-never (5)
Self-blame (1 item)
How often is each true for you? I blame myself when things go wrong.
Always (l)-never (5)
Hopeless (1 item)
How often is each true for you? Things seem hopeless.
Always (l)-never (5)
1118
W. SCOTT HILLES AND LYNN R. KAHLE
Table 2 Descriptive Statistics
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.
M
SD
a level Time 1
Time 2
.488
.79
.85
1.115
.83
.82
.445 .310
.437 .437
.87 .79
.90 .81
3.761
.239
.321
.86
.90
5.387 2.321 2.971 2.462
.063 .940 .979 .978
.224 .811 .950 .978
.79
.87
— — —
— —
Variable
Time 1
Time 2
Time 1
Independence Implicit social contract Explicit social contract Role commitment Social-American Dream Self-American Dream Raw deal Self-blame Hopeless
5.459
5.418
.472
4.296
4.282
.920
3.857 4.944
3.645 4.890
3.783 5.240 2.437 3.000 2.351
Cross-Lagged Correlations According to Kenny (1975), when perfect stationarity and synchronicity are present, the cross-lagged correlations do not differ if the relation between the two variables is spurious. Synchronicity exists for the synchronous correlations here because both variables were measured at the same time in the students' sophomore and senior years. And, as we have seen, perfect stationarity is a viable assumption here; hence, differences in cross-lagged correlations imply nonspuriousness. The first set of correlations (implicit social contract and role commitment) measures the Table 3 Correlations Among Variables and Differences
Time 2
—
relation between (a) expectations and norms and (b) belief in the benefits of involvement. The cross-lagged correlations are significantly different, indicating that a social contract based on expectations of students nonspuriously increases socially integrated motivations and attempts by the students to meet those expectations. The second set of correlations (independence and role commitment) relates (a) the student's desire to have more independence over the consequences of his or her interaction with teachers, such as wanting the teacher to treat him or her in a less authoritarian manner, with (b) the belief that involvement in school
Between Cross-lagged Correlations
Cross-lagged correlations
Autocorrelations
Synchronous correlations
Variable A/B
n
A1B2
B1A2
Z"
A1A2
BIB2
A1B1
A2B2
Implicit social contract/role commitment Independence/role commitment Raw deal/self-blame Raw deal/hopeless Hopeless/self-blame Hopeless/role commitment Role commitment/social-American Dream Explicit social contract/self-American Dream
61
59
.367 .379 -.327 .313 -.351 -.393
.130 .012 -.039 .005 .083 -.001
1.656» 2.492" 1.908* 1.949** 1.789* 2.393**
.295 .241 .288 .264 .192 .150
.240 .272 .350 .150 .350 .381
.560 .465 -.171 .413 -.370 -.297
.509 .509 -.066 .466 -.210 -.187
71
.240
.011
1.709*
.399
.283
.426
.055
2.557"
.540
.310
61
71 59 71
59
-.341
* Pearson-Filon test, following Kenny (1975), for difference between the cross-lagged correlations.
* p
