Emotional Autonomy Scale [PDF]

Citation preview

Child Development, January/February 2001, Volume 72, Number 1, Pages 207–219



The Vicissitudes of Measurement: A Confirmatory Factor Analysis of the Emotional Autonomy Scale Mark F. Schmitz and Judith C. Baer



This study examined the factor structure of the Emotional Autonomy Scale (EAS) as proposed by Steinberg and Silverberg. Participants were from three independent samples of adolescents in grades 6 (n  1,842), 8 (n  1,769), and 10 (n  1,232), with each sample consisting of three ethnic groups: African American, European American, and Mexican American. None of the confirmatory factor analyses for these samples supported the factor structure proposed by Steinberg and Silverberg. From the three models tested, the EAS is best described by the four originally proposed factors, combined with two method factors, one consisting of the positively worded scale items and one consisting of the negatively worded scale items. Results show that the EAS exhibits poor construct validity and behaves quite differently for the different grade and ethnic groups. The strong impact of method variance on the factor structure is discussed. Although various alternative solutions to the psychometric problems in the EAS are proposed, the most credible solution may be to reexamine the conceptual foundations of emotional autonomy and develop better measures of those concepts for adolescents.



INTRODUCTION The Emotional Autonomy Scale (EAS) constructed by Steinberg and Silverberg (1986) has been the impetus for a number of cross-sectional studies about the development of autonomy during adolescence. At the time of its inception, the EAS addressed an important void, because there were few empirical studies that measured the process-oriented and developmental aspects of attachment and autonomy (Hill & Holmbeck, 1986). Autonomy had been conceptualized as a multidimensional construct involving affective, behavioral, and cognitive domains; however, research prior to the EAS was often conceptually vague and consisted of a variety of sparsely measured operational definitions. The result was empirical investigations driven by disparate theoretical foci, including psychoanalytic, social learning, and relational perspectives (Hill & Holmbeck, 1986). There was a great need to interrelate aspects of autonomy and tie theory to investigations of transformations in family relationships. Steinberg and Silverberg partially addressed these concerns in their construction of the EAS. Subsequent studies using the EAS have made important contributions by identifying some of the complex interrelations involved in the dynamic emotional bonds between adolescents and their parents (Fuhrman & Holmbeck, 1995; Lamborn & Steinberg, 1993; Ryan & Lynch, 1989; Steinberg & Silverberg, 1986). Unfortunately, the psychometric properties of the EAS have never been presented beyond the original findings reported by Steinberg and Silverberg (1986), which noted that exploratory factor analysis had been used



to examine the proposed four-factor structure of the scale. No other studies have published anything about the psychometric properties of the EAS, other than coefficient . In this study, confirmatory factor analytic techniques were used to examine the factor structure of the EAS in three independent samples of adolescents. Previous use of the EAS. The EAS consists of a 20item Likert scale composed of four subscales developed from Blos’s (1979) theoretical perspective on individuation. The subscales are “perceives parents as people,” “parental deidealization,” “nondependency on parents,” and “individuation” (Steinberg & Silverberg, 1986). Although the EAS has been used in a variety of studies during the past decade, a number of important controversies about the scale have arisen. The primary debate stems from the interpretation of the EAS as either a form of autonomy or unhealthy detachment. Central to the controversy are competing theories about whether emotional distancing from the parents by the adolescent leads to healthy self reliance and individuation or poor developmental outcomes. The emotional distancing model is based on the work of Douvan and Adelson (1966) and Blos (1962, 1979), theorists who view development as a process wherein the adolescent reduces parental dependency and relinquishes omnipotent ideas about parents in favor of more mature and balanced ones (Blos, 1962, 1979; Douvan & Adelson, 1966). The opposing position articulated by Ryan and Lynch (1989) is that ado© 2001 by the Society for Research in Child Development, Inc. All rights reserved. 0009-3920/2001/7201-0013



208



Child Development



lescent development is facilitated by parental attachment and involves emotional closeness and a sense of support. Ryan and Lynch argue that the EAS measures emotional detachment and is not an indicator of self-regulation and self-reliance as originally operationalized by Steinberg and Silverberg. This detachment, according to Ryan and Lynch, actually is a maladaptive reluctance by the adolescent to rely on parents. Studies of adolescent autonomy using the EAS have yielded equivocal findings that are often interpreted according to one of these two competing positions. The work of Ryan and Lynch (1989) showed that adolescents who had higher EAS scores felt less secure and were less willing to use parental resources for support. Ryan and Lynch (p. 353) concluded that their findings were consistent with the attachment perspective and suggested that emotional autonomy as indexed by the EAS was a measure of negative family functioning. “The more ‘emotional autonomy’ teenagers or young adults express, the less connected or secure they feel within the family, the less they experience their parents as conveying love and understanding, and the less they report willingness to draw upon parental resources.” Subsequent studies by Lamborn (1990) and Lamborn and Steinberg (1993) examined the relationship between the EAS and adolescent adjustment. These authors posited that the correlations between emotional autonomy and adjustment are moderated by parental support. Lamborn (1990) found that adolescents whose parental relationships were characterized as avoidant or anxious were better adjusted when they also had high EAS scores. Lamborn’s conclusion was that emotional autonomy is adaptive for adolescents in stressful family environments. In a subsequent study by Lamborn and Steinberg (1993), however, just the opposite conclusion was made, in that emotional autonomy was found to be more functional in the context of a supportive family. A recent study of the EAS by Fuhrman and Holmbeck (1995) indicated results similar to Ryan and Lynch’s (1989) in that adolescents (10–18 years old) having high EAS scores were also less securely attached and did not use parents for emotional support. Importantly, their study showed the role of context on the adaptiveness of emotional autonomy for developmental outcomes. Fuhrman and Holmbeck, however, concluded that it is unclear whether the EAS measures earlier relational (attachment) patterns or adaptive strategies arising as a function of adolescent development. Given that we have very little understanding of the psychometric properties of the EAS, basic measurement theory tells us that we really have



very little knowledge about what the EAS actually measures (Nunnally & Bernstein, 1994). Factor analytic issues and the EAS. Conceptualization and measurement have long been issues of great concern for social scientists (Blalock, 1979), yet as the literature cited earlier indicates, they remain curiously neglected in the examination of adolescent autonomy. One of the most important stages in the development of a measure is to examine the impact of random and systematic errors on the measurement of the concept being studied. Systematic errors can be particularly devastating in their potential to confound substantive associations of the concept of interest with other variables (Nunnally & Bernstein, 1994) while not being correctable with techniques for handling the effects of random errors (Bollen & Paxton, 1998). The method of measurement represents possible systematic error, and so researchers have called for the use of multiple methods for assessing substantive concepts (Bank, Dishion, Skinner, & Patterson, 1990). The most well known approach for dealing with these issues has been the multitrait-multimethod (MTMM) research design of Campbell and Fiske (1959), an approach typically involving the use of multiple reporters of multiple substantive traits. Because the data requirements for the classic MTMM design are quite demanding, Bollen and Paxton (1998) have shown the suitability of using a similar approach that uses more relaxed conditions for assessing the biases resulting from systematic measurement error. Their approach essentially involves the inclusion of method-specific factors within the factor structure of the measure being examined. The influence of method factors is examined by comparing two models, one with method factors and one without. If there are no method factors in the measure of interest, then the fit of the simpler model should essentially be the same as the model with method factors. One potentially problematic method effect comes from the standard practice of using oppositely worded items within a scale to reduce acquiescent response bias (e.g., “I am happy” and “I am sad” in a measure of affect). A significant body of literature has begun to examine the pitfalls of this approach (Green, Goldman, & Salovey, 1993; Levin & Montag, 1989; Marsh, 1986, 1996; Russell, 1996). The main assumption when using both positively and negatively worded items in a single scale is that they all measure the same construct; yet, this assumption is rarely tested (Marsh, 1996). Factor analyses that reveal separate factors for positively and negatively worded items indicate that the validity of this assumption is questionable for that particular scale. In addition,



Schmitz and Baer



Carmines and Zeller (1979) reasoned that the distinction between method factors is substantively meaningful if they relate differently to external constructs. The greater the degree to which the method factors relate to substantively interesting variables, the greater the danger for spurious effects within the model being tested (Bollen & Paxton, 1998). Of particular concern for users of the EAS is empirical evidence indicating that negatively worded items may be differentially interpreted by younger respondents and by respondents having weaker reading skills (Marsh, 1986). These respondents may have greater difficulty understanding and responding to negatively worded items, especially if the items involve the use of double negative logic. The oppositely worded items in the EAS do not follow the doublenegative logic that Marsh found most problematic; however, Steinberg and Silverberg clearly identified specific items as positive and negative measures of the four subscales of the EAS. For example, a positively worded item from the individuation subscale is “There are some things about me that my parents don’t know.” A negatively worded item from that subscale is “My parents know everything there is to know about me.” A serious difficulty arises in attempting to differentiate the effects due to cognitive development of the adolescent from the substantive effects of change in adolescent autonomy and its correlates. How this methodology influences the factor structure of the EAS remains untested. The most glaring inadequacy of the previous EAS literature is the lack of any reported examination of the originally proposed factor structure. Several basic questions arise: Do the EAS items load on their proposed factors? Given the use of both positively and negatively worded items within the scale, what are the influences of the corresponding method factors? Finally, given that emotional autonomy is supposed to evolve over time, is this factor structure consistent for different age groups of adolescents? METHOD Sample and measure. Participants in the study were surveyed in five school districts in a large metropolitan area in the southwestern United States. Three cohorts of adolescents beginning in the sixth, seventh and eighth grades were tracked for 3 years. Survey instruments were administered during regular school hours to all students present at the time of data collection; no compensation was given for participation in the study. There was an absenteeism rate of 12% during the data collection; the response rate of students who were actually in the classroom was 95%. The



209



three samples used in this study for testing the EAS factor structure are the sixth-grade cohort at wave 1 (“Grade 6” in tables and discussion), the seventhgrade cohort at wave 2 (“Grade 8”), and the eighthgrade cohort at wave 3 (“Grade 10”). Demographic characteristics of the samples are shown in Table 1. The version of the EAS used in the present study is identical to the version used by Steinberg and Silverberg (1986), except that a 5-point Likert response set was used, rather than the original 4-point set. The rationale for this modification was that more scale points generally increase variability in response and that it made the EAS consistent with the other 5-point Likert scales in the survey instrument (P. E. Baer, personal communication, June 17, 1996). Because the original scale was an attempt to operationalize Blos’s model, which had not been empirically tested with nonclinical populations, adding the neutral response option gave respondents an opportunity to indicate that ideas within the model did not apply to their experience. The items of the EAS, along with the originally proposed factor structure, are presented in the Appendix. Analytical procedures. Confirmatory factor analyses of a series of models were conducted by using maximum likelihood estimation procedures of Amos 3.61 (Arbuckle, 1997) after listwise deletion of missing data. (Analyses using the raw data and full information maximum likelihood estimation procedures available in Amos produced results equivalent to those obtained by using listwise deletion of missing data and maximum likelihood estimation.) In these analyses, the resultant parameters consist of factor loadings, factor variances and covariances, errors for the measured variables, and correlations among errors. The factor loadings are of primary interest and are indicative of the validity of each measured variable relative to the given construct (Bollen, 1989). The error terms for the measured variables consist of both random and systematic variance not accounted for by the underlying factor structure. Correlations among the measured variable error terms are indicative of systematic covariation between measured variables that cannot be explained in terms of the proposed factor structure. A more extensive discussion of confirmatory factor analysis is beyond the scope of this paper but is widely available elsewhere (Bollen, 1989; Byrne, 1989; Hayduk, 1987; Long, 1983; McDonald, 1985; Pedhazur & Schmelkin, 1991). Three proposed models were tested. These were M1, the 4 subscales originally proposed by Steinberg and Silverberg (1986); M2, two method factors, one for the positively worded items and one for the negatively worded items; and M3, the two method factors



210



Child Development



Table 1



Demographic Characteristics of Sample



Grade 6 (n  1,842) Gender Male Female Race African American European American Mexican American Age 11 years 12 years 13 years Grade 8 (n  1,769) Gender Male Female Race African American European American Mexican American Age 13 years 14 years 15 years Grade 10 (n  1,232) Gender Male Female Race African American European American Mexican American Age 15 years 16 years 17 years



51% 49% 25% 28% 37% 47% 43% 10%



51% 49% 28% 42% 30% 44% 45% 11%



46% 54% 29% 41% 30% 41% 47% 12% African European Mexican American American American (%) (%) (%)



Mother’s education  High school degree 10 High school degree/GED 26 Some college 24 College degree or higher 40 2(6, N  4,843)  1,344, p  .001



19 34 22 25



54 26 24 11



Father’s education  High school degree 13 High school degree/GED 25 Some college 20 College degree or higher 42 2(6, N  4,843)  1,158, p  .001



18 32 22 28



54 23 10 13



15 57



13 68



28



19



Family structure Single parent 30 Biological mother and 41 biological father Other 29 2(4, N  4,843)  399, p  .001



and the four original subscales. The correlations among factors were kept fairly restrictive. The four original subscales were allowed to correlate with each other, and the positive and negative method factors were allowed to correlate with each other, but correlations between the method factors and original subscales were constrained to zero. Comparing the matrix of the observed variances and covariances of the measured variables with the matrix of the variances and covariances of the measured variables resulting from the proposed model tests the suitability of that model. To perform this comparison, several measures of overall model fit are available, with varying strengths and weaknesses. The measures used in the present study are well known and widely accepted as robust measures of fit (Hoyle, 1995): 2, GFI (Goodness of Fit Index), CFI (Comparative Fit Index), ECVI (Expected Cross-Validation Index), and RMSEA (Root Mean Square Error of Approximation). The model providing the best fit to the data is determined by comparing measures of overall fit for each of the models in the proposed series, typically by testing for significant differences in the model 2. Many other models could be tested; however, the three models examined here are sufficient for providing a basic test of the EAS factor structure. Indeed, an indication of good construct validity of the EAS would be the strength of Model M1 and a lack of improvement in overall fit in the subsequent models, M2 and M3. The series of proposed models was tested as multiple group analyses of the three ethnic groups within the three grade cohorts. (To reduce the potential impact of English-language proficiency on the results, those Mexican Americans in the sample who spoke only Spanish with both of their parents were eliminated from the analyses.) This provided an examination of the EAS factor structure for three independent samples of adolescents in grades 6, 8, and 10, and for the examination of ethnic differences in the EAS structure. The European American ethnic groups allow for the comparison of these results with samples that most closely resemble the samples originally used by Steinberg and Silverberg (1986). RESULTS Descriptions and overall fit for the three models tested are presented in Table 2. Clearly, the four-factor structure (Model M1) originally proposed by Steinberg and Silverberg (1986) provided a very poor fit to the data, as indicated by all the measures of overall fit. The model 2 remained significant for all nine models tested, whereas the other fit indicators showed ac-



Schmitz and Baer Table 2 Overall Fit Statistics for Emotional Autonomy Scale (EAS) Confirmatory Factor Analyses



Grade 6 Model 1 Model 2 Model 3 Grade 8 Model 1 Model 2 Model 3 Grade 10 Model 1 Model 2 Model 3



2(df)



GFI



CFI



3026.8 (492) 2260.5 (507) 1069.4 (429)



.815



.675



.871



.775



.938



.918



.819



.752



.835



.757



.931



.924



.819



.742



.815



.728



.918



.913



2835.5 (492) 2804.8 (507) 1150.4 (429) 2191.8 (492) 2293.3 (507) 1000.7 (429)



ECVI (90% CI)



RMSEA (90% CI)



2.039 (1.934, 2.148) 1.547 (1.459, 1.641) 0.907 (0.850, 0.969)



.056 (.054, .058) .046 (.044, .048) .030 (.028, .033)



1.912 (1.812, 2.018) 1.875 (1.775, 1.980) 0.954 (0.895, 1.018)



.054 (.052, .056) .053 (.051, .055) .032 (.030, .034)



2.176 (2.052, 2.308) 2.239 (2.111, 2.374) 1.235 (1.157, 1.320)



.055 (.053, .058) .056 (.053, .058) .034 (.031, .037)



Note: Model 1: Four EAS Subscales—Parent Deidealization, Nondependence, Perceive Parents as People, Individuation. Model 2: Two Method Factors—Same Direction as EAS Subscale, Opposite Direction from EAS Subscale. Model 3: Combination of Model 1 and Model 2—Four EAS Subscales and Two Method Factors.



ceptable fit only for Model M3 in each of the three samples (using as values for acceptable fit GFI  .90, CFI  .90, and RMSEA  .05). Because models M1, M2, and M3 were not hierarchically nested, the 2 difference test was not appropriate for testing differences in fit between models. The 90% Confidence Intervals, however, provided with the ECVI and RMSEA, indicate the degree of overlap between models. If the 90% CIs do not overlap (e.g., between M1 and M2 in the sixth-grade sample), then the two comparison models fit the data significantly differently, with the model having the smaller point estimate providing the better fit to the data. This suggests that Model M2 fit the data significantly better than Model M1 in the sixth-grade sample, that Models M1 and M2 fit the data equally well for the eighth-grade and tenth-grade samples, and that Model M3 provided the best fit for all three samples. Standardized and unstandardized factor loadings obtained for all of the models are shown in Table 3 for the sixth-grade sample, Table 4 for the eighth-grade sample, and Table 5 for the tenth-grade sample. Examining the standardized factor loadings for Model M1 can help determine where the lack of fit in



211



that model occurs. In Table 3, results for Model M1 for the sixth-grade sample showed standardized loadings for the Parent Deidealization construct that were good (  .40) for all three ethnic groups. This result was consistent with the other two samples, as can be seen for the eighth-grade sample in Table 4 and the tenth-grade sample in Table 5. For the Nondependency construct, EAS5 was weak in all three ethnic groups in the sixth-grade sample and was weak in the African American and Mexican American groups for the eighth- and tenth-grade samples. All other loadings for the Nondependency construct were good (  .40), with the exception of EAS6 in the European American eighth-grade sample. EAS7 and EAS9 showed very strong loadings (  .60) on the Individuation construct for all three ethnic groups in all three samples. The other items indicated weaker but generally still adequate loadings on that construct. The Perceives Parents as People construct was quite problematic for the sixth- and eighth-grade samples, with generally very weak loadings for all of the ethnic groups. This construct did exhibit somewhat better loadings for the tenth-grade sample. The standardized factor loadings for Model M2 were consistently higher than the factor loadings in Model M1, in all ethnic groups and for all three samples. In the sixth-grade sample, the EAS3 item did not load very well for any of the ethnic groups (Table 3), and the EAS16 item did not load well for the African American group. All other standardized loadings for Model M2 in the sixth-grade sample were good (  .40). In the eighth-grade sample, the EAS3 item was also weak in the European American group, as was the EAS5 item for the African American group. In the tenth-grade sample, only one loading was less than .40: the EAS10 item for the African American group. Clearly, the method factors exhibited greater stability across ages and ethnicity than did the originally proposed four-factor structure. These hints obtained from the first two models help when examining the results of Model M3, which is the most appropriate model for interpreting the resulting factor structures, having shown the best overall fit (Table 2) of the three basic models tested. In the sixth-grade sample (Table 3) the Parent Deidealization construct held only for the African American group, whereas the Nondependency construct held for the Mexican American group. The Perceives Parents as People and Individuation constructs were weak for all three ethnic groups in this sample. The European American sixth graders clearly showed the strongest and most consistent standardized factor loadings for the two method factors but presented considerably weaker loadings for the four EAS subscales.



212



Child Development



Table 3 Grade 6, Models 1, 2, and 3 Unstandardized (Standardized) Factor Loadings Model 2 Model 1



Positive



A-A



E-A



M-A



PDI



PDI



PDI



A-A



E-A



Model 3 Negative



M-A



A-A



E-A



M-A



Positive A-A



E-A



M-A



A-A



E-A



Negative M-A



A-A



E-A



M-A



PDI



PDI



PDI



EAS1



.99 .87 .99 (.61*) (.60*) (.65*)



.65 .52 .65 (.60*) (.57*) (.64*)



.90 (.48*)



.55 (.39*)



1.09 (.54*)



1.42 .48 .50 (.45*) (.43*) (.37*)



EAS4



1.33 1.30 1.03 (.60*) (.61*) (.54*)



.85 .80 .68 (.59*) (.60*) (.53*)



1.24 (.50*)



.94 (.45*)



1.15 (.46*)



1.63 .71 .48 (.39*) (.43*) (.29*)



EAS11 1.00 1.00 1.00 (.49) (.54) (.58)



.63 .60 .69 (.46*) (.56*) (.59*)



1.00 (.43)



1.00 (.55)



1.00 (.44)



.76 .41 .61 (.19*) (.29*) (.39*)



EAS15 1.36 1.52 1.47 (.53*) (.62*) (.66*)



.92 .93 .99 (.54*) (.60*) (.67*)



1.76 (.60*)



.65 (.27*)



1.33 (.45*)



EAS18 1.02 .96 .78 (.48*) (.50*) (.45*)



.69 .61 .53 (.49*) (.51*) (.46*)



1.12 (.46*)



.37 (.20*)



.66 (.29*)



ND



ND



ND



.74 (.48*)



.95 (.49*)



.92 (.62*)



.59 (.35*)



.60 (.26*)



.57 .70 1.41 1.60 (.32*) (.31*) (.44*) (.43*)



.65 .56 .72 (.48*) (.46*) (.60*)



.50 (.35*)



.69 (.38*)



.64 (.46*)



1.74 .46 .61 (.44*) (.31*) (.38*)



.86 .82 .89 (.52*) (.56*) (.63*)



1.00 (.58)



1.00 (.45)



1.00 (.61)



.33 (.07)



.70 .59 (.39*) (.31*)



PP



PP



PP



.68 (.14)



1.32 (.54*)



1.50 1.00 1.00 1.00 (.60*) (.45) (.33) (.31)



.16 (.04)



1.11 1.18 (.66*) (.66*)



.25 (.05)



1.21 1.28 (.69*) (.71*)



ND



ND



ND



EAS2



.84 .93 .87 (.51*) (.63*) (.66*)



EAS5



.50 .68 .50 (.28*) (.39*) (.32*)



EAS6



.71 .65 .77 (.46*) (.46*) (.61*)



.74 .74 .78 (.50*) (.58*) (.61*) 1.27 1.36 1.37 (.41*) (.46*) (.42*)



EAS13 1.00 1.00 1.00 (.55) (.59) (.68) PP



PP



PP



.09 (.02)



.71 .56 .60 (.15 ) (.48*) (.36*)



.99 .62 .47 (.23*) (.39*) (.28*)



EAS3



.42 .46 (.26*) (.19*)



.68 1.00 1.00 1.00 (.50*) (.33) (.35) (.34)



EAS8



1.09 1.32 (.63*) (.50*)



.86 1.97 1.72 1.92 (.55*) (.61*) (.55*) (.57*)



1.24 (.24*)



.36 (.14*)



.32 1.38 1.94 2.18 (.11*) (.59*) (.58*) (.58*)



1.00 1.72 1.67 1.76 (.65) (.52*) (.54*) (.54*)



1.00 (.19)



1.00 (.38)



1.00 (.35)



EAS10 1.00 1.00 (.56) (.38) EAS12



.52 1.43 –.02 (.33*) (.64*) (–.02)



EAS16



.38 .65 (.24*) (.29*)



.60 .87 1.13 1.24 (.44*) (.29*) (.42*) (.43*)



EAS20



.55 1.68 (.34*) (.71*)



.01 (.01)



EAS7



1.00 1.00 (.62) (.70)



1.00 1.78 1.76 1.99 (.68) (.53*) (.54*) (.58*)



EAS9



.82 1.06 (.53*) (.77*)



.81 (.54*)



EAS14



.64 .53 (.42*) (.38*)



EAS17 EAS19



.80 .76 .77 (.54*) (.56*) (.57*)



3.51 .07 (.75*) (.03) .37 (.08)



.74 .88 .75 (.48*) (.62*) (.55*)



.40 (.18*)



1.23 1.77 1.87 (.51*) (.53*) (.50*)



.63 (.25*) .31 .68 1.16 1.36 (.12*) (.32*) (.40*) (.41*)



3.32 .39 .36 (.69*) (.16*) (.14*) I



I



I



1.00 (.51)



1.00 (.50)



1.00 (.47)



1.55 (.82*)



1.66 (.86*)



1.86 (.84*)



.66 1.75 1.64 1.79 (.49*) (.54*) (.53*) (.57*)



.22 (.12*)



.15 (.08*)



.06 (.03)



.59 .53 (.42*) (.42*)



.70 1.23 1.54 1.56 (.55*) (.42*) (.55*) (.53*)



.42 .09 (.25*) (.05)



.24 .65 1.78 1.69 (.13*) (.31*) (.59*) (.51*)



.65 .59 (.44*) (.45*)



.64 1.52 1.66 1.73 (.49*) (.49*) (.56*) (.57*)



.45 (.25*)



.10 (.05)



I



I



I



1.00 1.00 1.00 (.60) (.64) (.63)



.28 (.15*)



1.05 .99 (.55*) (.50*)



.94 1.71 1.97 (.39*) (.49*) (.52*) 1.00 (.21)



1.00 (.52)



1.00 (.47)



1.32 1.69 1.98 (.56*) (.50*) (.56*)



.92 1.67 1.91 (.40*) (.52*) (.56*)



Note: n (A-A)  392; n (E-A)  632; n (M-A)  599. A-A  African American; E-A  European American; M-A  Mexican American. PDI  Parental Deidealization; ND  Nondependency on Parents; PP  Perceives Parents as People; I  Individuation. * p  .05; p  .10.



Schmitz and Baer



213



Table 4 Grade 8, Models 1, 2, and 3 Unstandardized (Standardized) Factor Loadings Model 2 Model 1



Positive



A-A



E-A



M-A



PDI



PDI



PDI



A-A



E-A



Model 3 Negative



M-A



A-A



E-A



Positive M-A



A-A



E-A



PDI



M-A



A-A



E-A



Negative M-A



A-A



E-A



M-A



PDI



PDI



EAS1



.81 .96 .90 (.65*) (.69*) (.61*)



.67 .69 .72 .72 (.63*) (.67*) (.60*) (.45*)



.83 (.37*)



.73 (.49*)



.64 (.47*)



.71 .71 (.59*) (.35*)



EAS4



1.06 1.08 1.21 (.61*) (.59*) (.63*)



.86 .77 .97 1.36 (.58*) (.58*) (.63*) (.60*)



1.09 (.37*)



.99 (.52*)



.59 (.31*)



.78 .99 (.50*) (.38*)



.84 .72 .76 1.00 (.63*) (.58*) (.56*) (.49)



1.00 (.37)



1.00 (.60)



.74 (.43*)



.74 .48 (.51*) (.21*)



EAS11 1.00 (.65)



1.00 (.60)



1.00 (.60)



EAS15 1.24 1.47 1.24 (.66*) (.73*) (.59*)



1.06 1.07 1.00 .87 .55 (.66*) (.72*) (.59*) (.35*) (.17*)



.79 (.38*)



1.15 1.59 1.35 (.55*) (.91*) (.47*)



EAS18



.67 .52 .65 .83 (.53*) (.47*) (.50*) (.43*)



.30 (.12*)



.58 (.36*)



.60 (.37)



.59 .82 (.45*) (.37*)



ND



ND



.97 .89 .96 1.02 (.72*) (.70*) (.67*) (.61*)



1.09 (.50*)



.89 (.64*)



.85 (.49*)



.88 .86 (.58*) (.35*)



.57 (.28*)



.71 (.28*)



.47 (.28*)



.75 .45 .74 .68 (.56*) (.36*) (.54*) (.41*)



.41 (.19*)



.61 (.46*)



.65 (.38*)



.46 .69 (.32*) (.30*)



1.02 .98 .99 1.00 (.67*) (.67*) (.61*) (.53)



1.00 (.40)



1.00 (.63)



.95 (.48)



1.01 .74 (.58*) (.27*)



1.33 (.65)



.75 1.77 (.45*) (.67)



1.53 (.76*)



.89 1.89 (.53*) (.71*)



1.00 (.50)



1.00 1.00 (.60) (.37)



.80 .71 .81 (.54*) (.47*) (.50*) ND



ND



ND



EAS2



.96 .93 .97 (.78*) (.74*) (.73*)



EAS5



.48 .84 .60 .94 2.19 1.16 (.33*) (.57*) (.38*) (.39*) (.62*) (.46*)



EAS6



.67 .44 .68 (.55*) (.36*) (.54*)



EAS13 1.00 (.72) PP



1.00 1.00 (.69) (.67) PP



ND



PP



PP



PP



PP



1.22 2.65 1.30 (.43*) (.58*) (.63*)



EAS3



.92 .75 1.01 1.00 1.00 1.00 (.24*) (.35*) (.51*) (.44) (.31) (.45)



.74 (.44*)



.84 (.38*)



1.01 (.48*)



1.00 1.00 1.00 (.37) (.24) (.40)



EAS8



1.78 1.50 1.54 1.29 1.80 1.55 (.43*) (.64*) (.72*) (.53*) (.52*) (.64*)



.46 (.26*)



1.04 (.44*)



.48 (.21*)



1.41 2.10 1.67 (.49*) (.47*) (.63*)



1.00 (.56)



1.00 (.45)



1.00 (.48)



1.23 1.30 .94 (.43*) (.31*) (.38*)



.77 .63 .79 .05 (.50*) (.45*) (.50*) (.03)



1.07 (.47)



.36 (.17*)



EAS10 1.00 (.25)



1.00 1.00 1.19 1.26 .95 (.45) (.51) (.50*) (.39*) (.42*)



EAS12 2.71 1.39 .56 (.69*) (.63*) (.28*) EAS16



.73 .99 .92 1.33 1.02 .47 (.12 ) (.33*) (.51*) (.40*) (.41*) (.47*)



EAS20 2.82 1.66 .76 (.73*) (.74*) (.38*) I



I



.57 (.34*) .81 .74 .70 .21 (.53*) (.53*) (.44*) (.13 ) I



I



EAS7



1.00 (.57)



1.00 1.00 1.10 2.22 1.30 (.71) (.70) (.47*) (.64*) (.55*)



EAS9



1.26 1.05 .98 (.71*) (.79*) (.71*)



EAS14



1.00 (.34)



.43 (.20*)



.63 (.31*)



.86 1.56 1.06 (.32*) (.38*) (.44*)



1.31 .04 (.57*) (.02) I



I



1.00 (.62)



1.00 (.45)



1.00 1.00 2.76 (.71) (.63) (.94*)



.95 (.62*)



2.19 (.99*)



.61 .83 .63 1.13 2.26 1.14 (.34*) (.58*) (.43*) (.49*) (.64*) (.47*)



.36 (.12*)



.53 (.33*)



.37 (.16*)



1.24 2.37 1.16 (.45*) (.52*) (.43*)



EAS17



.79 .73 .56 1.16 2.05 .99 (.49*) (.56*) (.44*) (.55*) (.63*) (.46*)



.32 .28 (.12*) (.19*)



.17 (.08*)



1.45 3.43 1.14 (.58*) (.82*) (.48*)



EAS19



.65 .57 .55 1.28 1.69 1.20 (.38*) (.41*) (.40*) (.58*) (.49*) (.53*)



.07 (.03)



.23 (.10*)



1.62 2.08 1.25 (.63*) (.47*) (.50*)



1.00 (.65)



.11 (.07 )



1.33 2.36 1.30 (.48*) (.53*) (.50*)



Note: n (A-A)  437; n (E-A)  701; n (M-A)  492. A-A  African American; E-A  European American; M-A  Mexican American. PDI  Parental Deidealization; ND  Nondependency on Parents; PP  Perceives Parents as People; I  Individuation. * p  .05; p  .10.



214



Child Development



Table 5 Grade 10, Models 1, 2, and 3 Unstandardized (Standardized) Factor Loadings Model 2 Model 1



Positive



A-A



E-A



M-A



PDI



PDI



PDI



A-A



E-A



Model 3 Negative



M-A



A-A



E-A



Positive M-A



A-A



E-A



M-A



A-A



E-A



Negative M-A



A-A



E-A



M-A



PDI



PDI



PDI



EAS1



.91 .91 .89 (.62*) (.64*) (.72*)



.73 .80 .81 (.60*) (.63*) (.69*)



.55 (.24*)



.60 (.17*)



.82 (.41*)



.75 .79 (.55*) (.61*)



.87 (.57*)



EAS4



1.09 1.02 1.15 (.60*) (.60*) (.67*)



.87 .87 1.06 (.57*) (.57*) (.66*)



1.39 (.50*)



.97 1.06 (.22*) (.38*)



.83 .89 (.49*) (.58*)



1.19 (.57*)



EAS11 1.00 1.00 1.00 (.57) (.59) (.70)



.81 .86 .89 (.56*) (.57*) (.66*)



1.00 (.37)



1.00 (.23)



.81 .90 (.49*) (.59*)



.99 (.57*)



EAS15 1.41 1.33 1.20 (.68*) (.69*) (.64*)



1.13 1.18 1.12 .40 (.65*) (.68*) (.63*) (.12)



EAS18



.68 .72 .67 (.51*) (.53*) (.55*)



.85 .85 .77 (.53*) (.56*) (.59*) ND



ND



1.08 .99 1.09 (.81*) (.72*) (.79*)



EAS5



.57 .68 .49 1.18 1.30 .90 (.37*) (.44*) (.35*) (.48*) (.56*) (.45*)



EAS6



.61 .55 .78 (.46*) (.41*) (.61*)



PP



PP



1.53 1.45 .02 (.79*) (.82*) (.01)



.34 (.09)



.86 (.41*)



ND



ND



ND



.69 (.48*)



.83 (.48*)



.93 (.61*)



.26 (.16*)



.35 (.17*)



.62 1.62 1.75 .60 (.39*) (.50*) (.52*) (.31*)



.73 .62 .89 (.49*) (.43*) (.62*)



.13 (.09)



.12 (.07)



.94 .95 1.02 (.58*) (.58*) (.64*)



1.00 (.65) PP



.99 .95 1.12 (.66*) (.63*) (.73*)



EAS13 1.00 1.00 1.00 (.68) (.59) (.70)



2.43 3.04 (.49 ) (.99*)



.83 (.34*)



ND



EAS2



1.00 (.43)



PP



.67 .76 (.46*) (.55*)



.65 (.41*)



.98 .88 (.59*) (.58*)



.98 (.50*)



.56 (.39*)



.79 .62 (.48*) (.42*)



.92 (.50*)



1.00 (.52)



1.00 (.63)



.86 .85 (.47*) (.51*)



.70 (.34*)



PP



PP



.94 .74 (.49*) (.42*)



.68 (.34*)



.92 .76 (.49*) (.44*)



.34 (.18*)



1.00 1.00 (.54) (.59)



1.00 (.48)



EAS3



1.06 1.02 .99 1.00 1.00 1.00 (.51*) (.57*) (.56*) (.41) (.42) (.51)



.77 (.45*)



1.04 (.56*)



EAS8



1.25 1.20 1.18 1.17 1.25 1.08 (.56*) (.63*) (.65*) (.45*) (.49*) (.54*)



.65 (.35*)



.99 1.39 1.42 1.41 .82 (.50*) (.45*) (.42*) (.38*) (.43*)



1.00 (.60)



1.00 (.54)



.52 (.30*)



.90 1.81 (.49*) (.60*)



.43 (.25*)



.43 (.23*)



.83 (.48*)



1.08 2.15 (.59*) (.73*)



EAS10 1.00 1.00 1.00 (.50) (.56) (.58) EAS12



.79 1.05 .94 (.34*) (.44*) (.50*)



.69 1.01 .78 (.33*) (.58*) (.44*)



.74 .72 .73 (.44*) (.43*) (.48*)



EAS16 1.05 .68 .68 1.24 1.05 .93 (.51*) (.38*) (.37*) (.52*) (.44*) (.47*)



1.00 (.34)



.68 1.11 .92 (.22*) (.32*) (.51*)



.37 1.41 1.29 .87 (.12*) (.45*) (.37*) (.46*)



EAS20



.99 1.14 .86 (.49*) (.67*) (.50*)



EAS7



1.00 1.00 1.00 1.28 1.43 1.21 (.62) (.66) (.61) (.53*) (.56*) (.54*)



EAS9



.83 .94 .99 (.53*) (.70*) (.65*)



EAS14



.84 .96 .86 1.39 1.57 1.19 (.49*) (.62*) (.54*) (.55*) (.61*) (.55*)



EAS17



.95 .71 .86 1.38 1.33 1.22 (.62*) (.55*) (.59*) (.61*) (.62*) (.61*)



.29 .39 2.18 2.14 2.58 .70 (.15*) (.25*) (.66*) (.73*) (.83*) (.36*)



EAS19



.92 .60 .77 1.66 1.19 1.21 (.49*) (.46*) (.51*) (.68*) (.54*) (.59*)



.07 (.04)



I



I



.65 .74 .65 (.40*) (.44*) (.44*)



.90 1.00 1.00 1.00 (.30*) (.31) (.29) (.53)



I



1.00 1.00 1.00 (.62) (.61) (.62)



I



I



1.00 (.48)



1.00 (.55)



I



1.37 (.68*)



.90 1.38 (.55*) (.41*)



.45 (.21*)



.86 (.46*)



1.00 1.83 1.90 1.03 (.27) (.57*) (.52*) (.48*)



.67 1.65 1.91 1.18 (.19*) (.50*) (.51*) (.56*)



.16 1.25 2.05 1.57 .96 (.10*) (.37*) (.64*) (.50*) (.48*)



Note: n (A-A)  325; n (E-A)  467; n (M-A)  345. A-A  African American; E-A  European American; M-A  Mexican American. PDI  Parental Deidealization; ND  Nondependency on Parents; PP  Perceives Parents as People; I  Individuation. * p  .05; p  .10.



Schmitz and Baer



215



Table 6 Coefficient  of Emotional Autonomy Scale and Subscales PDI



ND



PP



I



Positive



Negative



Total



Grade 6 A-A E-A M-A



.67 .70 .73



.49 .56 .66



.50 .66 .51



.59 .69 .69



.70 .75 .75



.80 .83 .86



.76 .84 .80



Grade 8 A-A E-A M-A



.75 .75 .76



.64 .65 .68



.63 .70 .62



.65 .76 .68



.72 .78 .74



.86 .85 .86



.82 .87 .83



Grade 10 A-A E-A M-A



.74 .75 .78



.63 .63 .68



.66 .74 .69



.68 .74 .75



.76 .77 .77



.83 .83 .86



.83 .85 .86



Note: A-A  African American; E-A  European American; M-A  Mexican American. PDI  Parental Deidealization; ND  Nondependency on Parents; PP  Perceives Parents as People; I  Individuation.



In the eighth-grade sample, the Perceives Parents as People and Individuation constructs were again weak for all three ethnic groups. The Parent Deidealization and Nondependency constructs were strong in the Mexican American and African American groups, but those items loaded more strongly on the method factors for the European American group. The African American eighth-graders showed stronger loadings on the two method factors, as compared with the sixth-grade African Americans; the Mexican American group showed strong loadings on the Positive method factor but not on the Negative method factor. The Parent Deidealization, Nondependency, and Individuation constructs were quite strong for the tenth-grade Mexican Americans, whereas the items making up those constructs tended to load more strongly on the two method factors for the tenthgrade African American and European American groups. The Perceives Parents as People construct exhibited good loadings for the European American tenth graders but only marginal loadings for the other two tenth-grade ethnic groups. Essentially, for the tenth-grade sample, the Mexican American group tended to support the four-factor structure, whereas the African American and European American groups tended to support the method factors. One of the most commonly reported statistics regarding the EAS is an estimate of reliability, coefficient . Table 6 shows values of coefficient  for the total EAS and the different subscales, including the two method factors. Coefficient  for the total EAS was quite good for all ages and ethnic groups (  .75); however, the two method factors showed almost equally as strong values as for the total EAS and consistently higher values than did any of the EAS subscales.



A primary concern in these analyses was whether the positively worded and negatively worded items measure the same construct. A comparison of the mean values for the negatively worded items (after reverse coding) with the mean values for the positively worded items indicated a distinct difference in the patterns of the means over time (Figure 1). The sixth-grade adolescents showed a significantly smaller mean for the negatively worded items than for the positively worded items (p  .05). This statistically significant difference in means at the sixth grade disappeared by the eighth. The difference in the pattern for the items may result from the shift in understanding of the negatively worded items as the adolescents develop greater cognitive sophistication. Another pos-



Figure 1 Means of positive and negative item subscales from the Emotional Autonomy Scale.



216



Child Development



sibility is that the negatively worded items and the positively worded items measure substantively different constructs. DISCUSSION None of the confirmatory factor analyses presented here, using three independent samples, supported the factor structure originally proposed by Steinberg and Silverberg (1986). Indeed, the analyses showed that the EAS has very poor construct validity for any of the age and ethnic groups examined in this study, even in the European American samples, which most closely resemble the samples used by Steinberg and Silverberg (1986) in developing the EAS. The factor structure found in this study was often dominated by two method factors, one consisting of positively worded items and one consisting of negatively worded items, rather than the four-factor structure originally proposed by Steinberg and Silverberg (1986). Ethnic differences in the factor structure were present, most broadly in that the Mexican American groups tended to be somewhat different from both the African American and European American groups, particularly for the eighth-grade and tenth-grade samples. Among the four subscales, the Perceives Parents as People factor was consistently the weakest, except in the tenth-grade European American sample. The particularly poor construct validity for this subscale should not be too surprising. Other studies that have used the EAS subscales have dropped that subscale from most analyses, usually because it has been viewed as showing slow development during adolescence (Chen & Dornbusch, 1998; Lamborn & Steinberg, 1993; Steinberg & Silverberg, 1986). The results of this study indicate that this may actually have been due to the very poor construct validity of that particular subscale, rather than to any developmental issues within that aspect of autonomy. The Individuation factor also exhibited very weak factor loadings, except in the tenth-grade Mexican American sample. The African American and European American samples showed substantially stronger loadings of the Individuation items on the two method factors. All three Mexican American samples tended to differentiate between the Parent Deidealization and Nondependency constructs, whereas the African American and European American samples (particularly the eighth- and tenth-grade samples) blended these constructs into the method factors. Analyses of the mean scores for negatively worded and positively worded items showed very different trajectories through time. The negative items showed a significant increase in their mean values between



the sixth, seventh, and eighth grades and then a leveling out of the trajectory from the eighth grade until the tenth grade. The positive items showed a relatively stable mean value for all age groups in this study. These differences may be substantive in nature, or they may result from the increase in cognitive sophistication in older adolescence and thus represent the effects of method variance (Marsh, 1996). The present study, however, cannot distinguish between these alternatives. Measures of reading and language skill would be particularly helpful in discerning between the two possibilities to more systematically assess the cognitive sophistication of the respondents. Unfortunately these types of measures were not available in the present study. The influence of method variance extends even to the most commonly reported psychometric property of the EAS: coefficient . Values of coefficient  for the EAS total score were quite strong, a result consistent with most previous research, which typically reports values of .70 and greater for the total EAS (Fuhrman & Holmbeck, 1995; Lamborn & Steinberg, 1993; Ryan & Lynch, 1989; Steinberg & Silverberg, 1986). Recent research, however, has shown that method variance within the scale items inflates values for coefficient  in that the computation of coefficient  does not discriminate between the true covariances between the scale items and covariances due to systematic error (Miller, 1995; Tepper & Tepper, 1993). Indeed, the confirmatory factor analyses in the present study show how even very strong values for coefficient  can be psychometrically misleading. One solution to the influence of method variance is to use only the positively worded items (Marsh, 1986). Unfortunately, the EAS does not have an even split of wording style within the subscales so that eliminating the negatively worded items also eliminates all of the items for the Parental Deidealization (PDI) construct and all but one of the items for the Nondependency on Parents (ND) construct. This lack of balance between the oppositely worded items within the EAS subscales means that the alternative hypotheses regarding method influence or substantive influence in the method factors cannot be distinguished. The connection within the negative method factor of the PDI items and the ND items may be indicative of a common construct. This is an appealing hypothesis in that there likely is a strong connection between adolescents’ respect for their parents’ opinions and the likelihood of going to the parents for help with a problem. A fruitful line of research would be to develop items that are similar to the PDI and ND items but that are balanced in the use of positive and negative wording style. This would help to distinguish the



Schmitz and Baer



substantive variance from the method variance in the items. Significant ethnic differences in the EAS factor structure were present in all three cohorts. These differences could be due to cultural influences on adolescents’ perceptions of emotional autonomy. For example, in trying to control for English-language proficiency, we eliminated Mexican Americans who spoke only Spanish with both of their parents from the analyses. This also controls for some of the potential effects of acculturation on EAS factor structure; however, there were also significant demographic differences that could impact the presence and perception of emotional autonomy, particularly the ethnic differences in family structure and education (see Table 1). For example, the large percentage of singleparent African American families in the samples could produce the differences in EAS factor structure found for the African American adolescents. Unfortunately, the cell sizes in this study’s dataset were too small to test for group differences in the EAS factor structure on these variables. Future research should more carefully examine these cultural and social structural aspects of family relations and their impact on measures of emotional autonomy in adolescents. Study limitations. The method of obtaining the samples was a primary limitation of this study. The survey instrument was administered in the classrooms of the selected schools. If a student was not in the school on the day the survey was administered, then that student was not a participant in this study, for that particular wave. Although there was only 12% absenteeism, this aspect of missingness may have influenced our results, as indicated by the gender differences between the tenth-grade sample and the sixth- and eighth-grade samples (Table 1). The effects of missingness within the EAS, however, were tested and had no significant impact on the analyses. A second critical limitation in the present study was the use of a 5-point Likert response set, rather than the 4-point set typically used in the EAS. This, of course, impacted the variances and covariances among the EAS items and thus may produce a factor structure quite different from that found in the exploratory analysis of Steinberg and Silverberg (1986). The present analyses should thus be replicated by using different samples; confirmatory factor analyses using the 4-point set should also be done, before any further work with the EAS is performed. A third limitation was the potential impact of English-language proficiency. The English-language proficiency along with the cognitive sophistication of the respondents likely influences the amount of ran-



217



dom error present in the data and thus impacts the strength of the factor loadings in the analyses presented here. To minimize this error source, those Mexican Americans who spoke only Spanish with both of their parents were eliminated from the analyses. The similarity of coefficient  for the three ethnic groups (see Table 6) indicates that English-language proficiency likely did not impact the results very much. But the lack of adequate measures of English-language proficiency and cognitive sophistication remains a key limitation of this study. Conclusion. A crucial process in human development is the separation and individuation that occurs during adolescence. A critical aspect of this process may be the rise of autonomy in the adolescents’ relationships with parents and peers. As such, a significant body of research has been undertaken to investigate autonomy in adolescents. Much of that research has been based on the Emotional Autonomy Scale (EAS) proposed by Steinberg and Silverberg (1986). The confirmatory factor analyses presented here, which are the only published analyses of the construct validity of the EAS, indicate some problematic aspects of this scale. Indeed, the results presented in the present study indicate that the debate that has arisen about whether the EAS measures something normative or pathological (Fuhrman & Holmbeck, 1995; Lamborn & Steinberg, 1993; Ryan & Lynch, 1989; Steinberg & Silverberg, 1986) may have been fundamentally misguided by the unstable psychometric properties of the EAS. These problems in the EAS may be particularly damaging for studies that have used the EAS subscales (e.g., Chen & Dornbusch, 1998; Frank, Pirsch, & Wright, 1990). Recent recommendations proposing the examination of predictors and outcomes of the EAS subscales (Silverberg & Gondoli, 1996) are simply ill advised until significant improvement occurs in the measurement of the associated constructs. Emotional autonomy is very likely multidimensional in nature, and this should be addressed in any new or modified instrumentation. In the examination of emotional autonomy in adolescence, one helpful approach may be to use portions of the present EAS for the development of a new measure of this construct or a construct similar in nature (cf. Chen & Dornbusch, 1998). An alternative approach may be to start at the beginning by reconceptualizing what we are trying to measure. Because much of the debate has focused on what emotional autonomy represents within adolescent developmental processes, this may be an indication of the need for more work at the conceptualization level in the research process.



218



Child Development



ACKNOWLEDGMENTS This research is supported by a grant from the National Institute of Alcoholism and Alcohol Abuse (R01 AA08864) awarded to Paul E. Baer and James Bray, Baylor College of Medicine. ADDRESSES AND AFFILIATIONS Corresponding author: Mark F. Schmitz, School of Social Work, Rutgers, The State University of New Jersey, 536 George Street, New Brunswick, NJ 089011167; e-mail: [email protected]. Judith C. Baer is also at Rutgers, The State University of New Jersey. APPENDIX



EAS16: My parents probably talk about different things when I am around from what they talk about when I’m not. ( PP) EAS17: There are things that I will do differently from my mother and father when I become a parent. ( I) EAS18: My parents hardly ever make mistakes. ( ND) EAS19: I wish my parents would understand who I really am. ( I) EAS20: My parents act pretty much the same way when they are with their friends as they do when they are at home with me. ( PP) Note: PDI  Parental Deidealization; ND  Nondependency on Parents; PP  Perceives Parents as People; I  Individuation.



Emotional Autonomy Scale EAS1: EAS2: EAS3: EAS4: EAS5:



EAS6:



EAS7: EAS8:



EAS9: EAS10: EAS11: EAS12:



EAS13:



EAS14: EAS15:



My parents and I agree on everything. ( PDI) I go to my parents for help before trying to solve a problem myself. ( ND) I have often wondered how my parents act when I’m not around. ( PP) Even when my parents and I disagree, my parents are always right. ( PDI) It’s better for kids to go to their best friend than to their parents for advice on some things. ( ND) When I’ve done something wrong, I depend on my parents to straighten things out for me. ( ND) There are some things about me that my parents don’t know. ( I) My parents act differently when they are with their own parents from the way they do at home. ( PP) My parents know everything there is to know about me. ( I) I might be surprised to see how my parents act at a party. ( PP) I try to have the same opinions as my parents. ( PDI) When they are at work, my parents act pretty much the same way they do when they are at home. ( PP) If I were having a problem with one of my friends, I would discuss it with my mother or father before deciding what to do about it. ( ND) My parents would be surprised to know what I’m like when I’m not with them. ( I) When I become a parent, I’m going to treat my children exactly the same way that my parents have treated me. ( PDI)



REFERENCES Arbuckle, J. L. (1997). Amos users’ guide version 3.6. Chicago: SPSS, Inc. Bank, L., Dishion, T., Skinner, M., & Patterson, G. R. (1990). Method variance in structural equation modeling: Living with “glop.” In G. R. Patterson (Ed.), Depression and aggression in family interaction (pp. 247–279). Hillsdale, NJ: Erlbaum. Blalock, H. M. (1979). The presidential address: Measurement and conceptualization problems: The major obstacle to integrating theory and research. American Sociological Review, 44, 370–390. Blos, P. (1962). On adolescence: A psychoanalytic interpretation. Glencoe, IL: Free Press. Blos, P. (1979). The adolescent passage. New York: International Universities Press. Bollen, K. A. (1989). Structural equations with latent variables. New York: Wiley. Bollen, K. A., & Paxton, P. (1998). Detection and determinants of bias in subjective measures. American Sociological Review, 63, 465–478. Byrne, B. M. (1989). A primer of LISREL: Basic applications and programming for confirmatory factor analytic models. New York: Springer-Verlag. Campbell, D. T., & Fiske, D. W. (1959). Convergent and discriminant validation by the multitrait-multimethod matrix. Psychological Bulletin, 56, 81–105. Carmines, E. G., & Zeller, R. A. (1979). Reliability and validity assessment. Thousand Oaks, CA: Sage. Chen, Z. Y., & Dornbusch, S. M. (1998). Relating aspects of adolescent emotional autonomy to academic achievement and deviant behavior. Journal of Adolescent Research, 13, 293–319. Douvan, E., & Adelson, J. (1966). The adolescent experience. New York: Wiley. Frank, S. J., Pirsch, L. A., & Wright, V. C. (1990). Late adolescents’ perceptions of their relationships with their parents: Relationships among deidealization, autonomy, relatedness, and insecurity and implications for adoles-



Schmitz and Baer



cent adjustment and ego identity status. Journal of Youth and Adolescence, 19, 571–588. Fuhrman, T., & Holmbeck, G. N. (1995). A contextualmoderator analysis of emotional autonomy and adjustment in adolescence. Child Development, 66, 793–811. Green, D. P., Goldman, S. L., & Salovey, P. (1993). Measurement error masks bipolarity in affect ratings. Journal of Personality and Social Psychology, 64, 1029–1041. Hayduk, L. A. (1987). Structural equation models with LISREL: Essentials and advances. Baltimore: Johns Hopkins University Press. Hill, J. P., & Holmbeck, G. (1986). Attachment and autonomy during adolescence. In G. Whitehurst (Ed.), Annals of child development (Vol. 3). Greenwich, CT: JAI. Hoyle, R. H. (1995). Structural equation modeling: Concepts, issues, and applications. Thousand Oaks, CA: Sage. Lamborn, S. D. (1990, March). Psychosocial correlates of emotional autonomy among securely, avoidantly, and anxiously attached adolescents. Paper presented at the meeting of the Society for Research on Adolescence, Atlanta. Lamborn, S. D., & Steinberg, L. (1993). Emotional autonomy redux: Revisiting Ryan and Lynch. Child Development, 64, 483–499. Levin, J., & Montag, I. (1989). The bipolarity of the Comfrey Personality Scales: A confirmatory factor analysis. Personality and Individual Differences, 10, 1115–1120. Long, J. S. (1983). Confirmatory factor analysis. Thousand Oaks, CA: Sage. Marsh, H. W. (1986). Negative item bias in ratings scales for preadolescent children: A cognitive-developmental phenomenon. Developmental Psychology, 22, 37–49. Marsh, H. W. (1996). Positive and negative global self-



219



esteem: A substantively meaningful distinction or artifactors? Journal of Personality and Social Psychology, 70, 810–819. McDonald, R. P. (1985). Factor analysis and related methods. Hillsdale, NJ: Erlbaum. Miller, M. B. (1995). Coefficient alpha: A basic introduction from the perspectives of classical test theory and structural equation modeling. Structural Equation Modeling, 2, 255–273. Nunnally, J. C., & Bernstein, I. H. (1994). Psychometric theory. New York: McGraw Hill. Pedhazur, E. J., & Schmelkin, L. P. (1991). Measurement, design, and analysis: An integrated approach. Hillsdale, NJ: Erlbaum. Russell, D. W. (1996). UCLA loneliness scale (version 3): Reliability, validity, and factor structure. Journal of Personality Assessment, 66, 20–40. Ryan, R. M., & Lynch, J. H. (1989). Emotional autonomy versus detachment: Revisiting the vicissitudes of adolescence and early adulthood. Child Development, 60, 340– 356. Silverberg, S. B., & Gondoli, D. M. (1996). Autonomy in adolescence: A contextualized perspective. In G. R. Adams, R. Montemayor, & T. P. Gullotta (Eds.), Psychosocial development during adolescence: Progress in developmental contextualism (pp. 12–61). Thousand Oaks, CA: Sage. Steinberg, L., & Silverberg, S. B. (1986). The vicissitudes of autonomy in early adolescence. Child Development, 57, 841–851. Tepper, B. J., & Tepper, K. (1993). The effects of method variance within measures. The Journal of Psychology, 127, 293– 302.



This document is a scanned copy of a printed document. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material.