Вы находитесь на странице: 1из 5

_____________________________________________________________________________________________________________

Fam Proc 36:369-373, 1997

The Family Environment Scale: Reliability and Normative Data for an Adolescent Sample
CANDICE P. BOYD, M.Psych.a ELEONORA GULLONE, Ph.D.b GILLIAN L. NEEDLEMAN, M.Psych.c TRUDY BURT, M.Psych.c
aLecturer in Psychology and Special Education, Faculty of Education, Monash University. Send correspondence to first author: Faculty of Education, Monash University, Clayton 3168, Australia; e-mail: Candice.Boyd@Education.monash.edu.au. bSenior Lecturer in Psychology, Department of Psychology, Monash University, Australia. cPracticing psychologists, Melbourne, Australia.

The Family Environment Scale has been used extensively in family research since first being published. However, despite its appeal both conceptually and empirically, doubts have been raised over the scale's reliability. This article presents normative and reliability data for the Family Environment Scale from a large, combined sample of adolescents. Means and standard deviations were generally found to be in line with those reported in the scale's manual; however, estimates of internal consistency for most subscales could be considered inadequate for research purposes. The Family Environment Scale (FES; R.H. Moos & B.S. Moos, 1986) is one of ten Social Climate scales published by its primary author. The scale comprises 90 true-false items and was designed to assess three dimensions of family environmenta relationship dimension, a personal growth dimension, and a system maintenance dimensionencompassing ten subscales in all. The scale has featured widely in family research since its publication over 10 years ago. While the scale has been traditionally used to assess family environment from different perspectives within the family (see Cole & McPherson, 1993), the scale has also been used to assess family environment from just one family member's perspective (see Nelson, 1984). Despite its wide use, the reliability and validity of the Family Environment Scale have been the subject of some criticism. This was notable in a series of articles by Roosa and Beals (1990a,b). In the first of these articles, Roosa and Beals presented data that challenged previously reported reliability estimates for the scale. Using data from a sample of 311 distressed and 74 control families, they found that estimates of internal consistency (Cronbach's alpha) for five subscales of the FES were not only lower than those reported in the scale's manual (Moos & Moos, 1986) but were well below what is considered to be an acceptable level for research purposes (that is, 0.70 and above; see Tabachnick & Fidell, 1989, p. 636). In questioning the reliability of the FES with respect to small, clinical populations, the authors suggested that researchers be cautious in using the scale until more is known about its reliability for specific samples. In response to Roosa and Beals, Moos (1990) defended the FES by highlighting the scale's conceptual and empirical development. In particular, he emphasized how the content and face validity of the subscales were based on conceptually derived definitions of central constructs. In addition to summarizing strengths in the scale's development, Moos (1990) suggested that Roosa and Beals may have obtained low estimates of internal consistency because of the restricted nature of their samples. In order to demonstrate this point, Moos presented alpha coefficients for five subscales of the FES from four ongoing projects with which he had involvement. These coefficients ranged from 0.58 for Control subscale to 0.79 for Cohesion. He also stated that a considerable body of evidence existed showing that the FES was reasonably internally consistent, stable, and valid, but, in making this point, he did not refer the reader to specific studies. In their second article, while accepting some of Moos' arguments, Roosa and Beals (1990b) still expressed concern over what they believed to be the scale's inadequate reliability. In addition, Roosa and Beals (1990b) noted that they had chosen the FES for use with their sample based on the reliability estimates reported in the scale's manual as well as its wide use in family-based research. In so doing, they highlighted the need for FES reliability data to be published in order to accumulate evidence regarding the performance of the scale. The issues raised by Roosa and Beals (1990a,b) provided the rationale for this article. We chose to present reliability and normative data for the FES for two reasons: first, to provide researchers with normative data from a substantially sized sample of adolescents; second, in line with Roosa and Beals' request, to provide researchers with reliability estimates for the scale from a sizable normative (and hence diverse) sample from which the performance of the scale can be judged.

_____________________________________________________________________________________________________________

METHOD
As part of two separate studies (Boyd & Gullone, 1997; Needleman, 1995), the FES was administered to 1,289 adolescents615 female and 674 maleranging in age from 11 to 18 years. In order to maximize generalizability, participants were selected from a total of 16 schools in the metropolitan and country areas of Victoria, Australia. Letters were sent home with students, and parental permission was obtained. Study 1 (Boyd & Gullone, 1997; 783 adolescents) recorded a 88% participation rate. Students were administered the FES on a group basis during school hours. Participation by the students themselves was anonymous and voluntary. There were no specific inclusion or exclusion criteria. Students from non-English speaking backgrounds or with low literacy levels were assisted by teachers in reading the items; however, the proportion of participants in this group was small. Demographic characteristics of the two samples were assessed in terms of sex distribution, ethnicity, couple status, and social status. In Study 1, 52.1% of participants were female, compared to 40.9% in Study 2. In Study 2, 72.5% of participants were Caucasian, compared to 74.2% in Study 1. In Study 1, 82.8% of participants came from couple homes, compared with 66.8% in Study 2. Different measures of parents' social status were employed in the two studies: Study 1 used the Warner, Meeker, and Eeel's Revised Scale for Rating Occupation (1949; cited in Miller, 1991) and Study 2 used Hollingshead's Two-Factor Index of Social Position (1957, cited in Miller, 1991). Nevertheless, for both samples the distribution of occupations was generally in line with census data published in the same year (Australian Bureau of Statistics, 1994), with the exception of unemployment level, which was slightly overrepresented in our samples compared to the state average.

RESULTS
Data from the two studies were analyzed using the Statistical Package for Social Sciences for Windows (SPSS, 1993). Initially, means and standard deviations were calculated for the two samples separately and a multivariate analysis of variance (MANOVA) by sample conducted to see whether the means significantly differed. The MANOVA yielded a significant multivariate effect [Wilk's (10, 1236) = 0.91; p < 0.001] and significant univariate effects for all FES subscales. However, the multivariate effect size was 0.09 (at the 0.05 level), hence the statistical significance obtained is likely to be a function of sample size rather than representing any practical difference. Because little true difference between samples was found, data from separate samples were then combined. Means and standard deviations for the combined sample, as well as for age and gender subgroups, are provided in Table 1. A MANOVA by age-group and gender was conducted in order to determine whether any significant differences were present in the data. This analysis yielded a significant multivariate effect for gender [Wilks (10, 1093) = 0.98; p < .05] and a significant univariate effect by gender for the Achievement Orientation subscale [F (1, 1102) = 7.10; p < 0.01]. A multivariate effect for age-group was also noted [Wilk's (10, 1093) = 0.94; p < 0.001] as well as significant univariate effects for the Achievement Orientation, Conflict, Control, Expressiveness, and Independence subscales. However, despite reaching statistical significance, the multivariate effects found for both age and gender yielded very small effect sizes (0.06 and 0.02, respectively). No age by gender interaction effects were found.
Table 1 Means and Standard Deviations by Age and Sex for FES Subscales (Combined Sample) Dimension/Sex Mean SD Cohesion Female Male 11-14 year olds 15-18 year olds Entire Sample Expressiveness Female Male 11-14 year olds 15-18 year olds Entire Sample Conflict Female
2

5.98 5.74 6.11 5.57 5.56 4.24 4.25 4.07 4.46 4.36 3.51

2.30 2.35 2.10 2.55 2.22 1.84 1.84 1.80 1.88 1.83 2.31

580 619 508 691 1199 589 623 516 696 1212 591

_____________________________________________________________________________________________________________

Male 11-14 year olds 15-18 year olds Entire Sample Independence Female Male 11-14 year olds 15-18 year olds Entire Sample Achievement Female Male 11-14 year olds 15-18 year olds Entire Sample Intellectual-Cultural Orientation Female Male 11-14 year olds 15-18 year olds Entire Sample Active-Recreational Orientation Female Male 11-14 year olds 15-18 year olds Entire Sample Moral-Religious Emphasis Female Male 11-14 year olds 15-18 year olds Entire Sample Organization Female Male 11-14 year olds 15-18 year olds Entire Sample Control Female Male 11-14 year olds 15-18 year olds Entire Sample

3.84 3.39 3.98 3.92 5.79 5.82 5.67 5.97 5.90 5.50 5.87 5.57 5.79 5.69 4.36 4.33 4.56 3.70 4.42 5.25 5.52 5.38 5.48 5.65 3.74 3.47 3.72 3.49 3.53 5.50 5.40 5.50 5.40 5.24 4.17 4.47 4.33 4.27 4.23

2.40 2.27 2.42 2.39 1.59 1.62 1.55 1.65 1.64 1.67 1.77 1.68 1.78 1.74 1.83 1.90 1.84 1.84 1.87 2.05 2.14 2.13 2.06 2.04 2.15 2.22 2.15 2.23 2.20 2.27 2.34 2.30 2.31 2.25 2.13 2.04 2.10 2.08 2.13

623 514 700 1214 592 615 512 697 1207 592 620 514 700 1214 586 621 511 696 1207 596 630 520 706 1226 584 612 505 691 1196 600 625 518 707 1225 591 626 517 700 1217

_____________________________________________________________________________________________________________

A measure of internal consistency (Cronbach's alpha) was then calculated for separate samples and for the two samples combined. Data for the combined sample are provided in Table 2. (Reliability data for separate samples may be obtained through correspondence with the authors.)
Table 2 Estimates of Internal Consistency by Age and Sex for FES Subscales (Combined Sample) Subscale Female Male 11-14 Year 15-18 Year Olds Olds Cohesion Expressiveness Conflict Independence Achievement Intellectual-Cultural Orientation Active-Recreational Orientation Moral-Religious Emphasis Organization Control 0.69 0.41 0.72 0.33 0.43 0.41 0.64 0.68 0.61 0.61 0.64 0.36 0.72 0.30 0.44 0.53 0.60 0.72 0.60 0.57 0.63 0.36 0.71 0.26 0.41 0.37 0.65 0.68 0.60 0.57 0.68 0.39 0.71 0.34 0.45 0.54 0.59 0.73 0.61 0.61

Entire Sample 0.67 0.39 0.72 0.31 0.44 0.47 0.62 0.71 0.60 0.59

As can be seen from Table 2, the alpha coefficients obtained for the combined sample were low to moderate. Coefficients ranged from 0.31 for the Independence subscale to 0.72 for the Conflict subscale. Lower than usual estimates might be expected due to the low number of items per subscale (nine items each); however, reliability estimates of below 0.70 are usually considered as inadequate for research purposes (Tabachnick & Fidell, 1989).

DISCUSSION
The means and standard deviations presented in this article are generally in line with previously reported data for the Family Environment Scale (see Moos & Moos, 1986). Despite limitations in sampling methods with respect to social and family status, these data provide descriptive statistics for a large, normative sample of adolescents. Researchers wanting to focus specifically on adolescent samples may find that these data provide a more valid comparison than the family data published in the FES manual. In contrast, the estimates of internal consistency reported in this article are not consistent with those previously reported by the scale's authors (Moos, 1990; Moos & Moos, 1986). This finding lends support to the results from Roosa and Beals' (1990) study and raises questions about the scale's reliability. Contrary to Moos' (1990) argument, despite the size and diversity of our combined sample, reliability estimates were still low to moderate for most subscales with the exception of Cohesion, Conflict, and Moral-Religious Emphasis. There are, however, alternative explanations for why estimates of internal consistency were low in our sample. Factors such as age, socioeconomic characteristics of the sample, and cultural differences between Australia and the U.S. might all have contributed to lower levels of internal consistency in our data. However, it is unlikely that either socioeconomic or cultural differences were significant factors since Australia and the U.S. are quite similar in terms of social, political, and economic climate. Age is the most likely factor influencing the internal consistency of our data. There may well be differences in how individual items are interpreted by young people compared with other family members. However, this assumption cannot be tested using our data because only the adolescent's perception was obtained. Future research could examine this question by comparing measures of internal consistency using complete family data to see whether the reliability of the FES varies as a function of respondent's age. To conclude, researchers are likely to continue using the FES because of its obvious conceptual appeal. Because of its wide use in past research, it also provides an excellent basis for comparison between studies. However, family researchers must balance the scale's strengths against some doubts over its reliability, particularly with respect to adolescent samples. Considering that it has now been more than 10 years since the scale was first published, it may be timely for this popular scale to be revised.

REFERENCES

_____________________________________________________________________________________________________________

1. 2. 3. 4. 5. 6. 7. 8.

9. 10. 11. 12.

Australian Bureau of Statistics, (1994) Victorian yearbook. Melbourne: Victorian Office of the Australian Bureau of Statistics. Boyd, C. P. and Gullone, E., (1997) Perceived family environment and its relationship to anxiety and depression in adolescence. Unpublished manuscript available from authors. Cole, D. A. and McPherson, A. E., (1993) Relation of family subsystems to adolescent depression: Implementing a new family assessment strategy, Journal of Family Psychology, 7, 119-133. Miller, D. C., (1991) Handbook of research design and social measurement. Newbury Park CA: Sage Publications. Moos, R. H., (1990) Conceptual and empirical approaches to developing family-based assessment procedures: Resolving the case of the Family Environment Scale, Family Process, 29, 199-208. Moos, R. H. and Moos, B. S., (1986) The Family Environment Scale: The manual. Palo Alto CA: Consulting Psychologists Press. Needleman, G., (1995) Self-esteem and quality of life: Relations with family environment in Australian adolescents. Master's dissertation, Monash University, Australia. Nelson, G., (1984) The relationship between dimensions of classroom and family environments and the self-concept, satisfaction and achievement of grade 7 and 8 students, Journal of Community Psychology, 12, 276-287. Roosa, M. W. and Beals, J., (1990a) Measurement issues in family assessment: The case of the Family Environment Scale, Family Process, 29, 191-198. Roosa, M. W. and Beals, J., (1990b) A final comment on the case of the Family Environment Scale, Family Process, 29, 209-211. SPSS, (1993) SPSS package for Windows. Chicago: SPSS, Inc. Tabachnick, B. G. and Fidell, L. S., (1989) Using multivariate statistics (2nd ed.). New York: Harper Collins.

Manuscript received December 30, 1996; Revisions submitted December 30, 1996; Accepted July 30, 1997.

Вам также может понравиться