Sie sind auf Seite 1von 15

Journal of Educational Psychology 1986, Vol. 78, No.

5, 381-395

Copyright 19B6 by the American Psychological Association, Inc. OO22-0663/86/SO0.75

Effects of Single-Sex Secondary Schools on Student Achievement and Attitudes


Valerie E. Lee
School of Education, University of Michigan

Anthony S. Bryk
Department of Education, University of Chicago

A movement away from single-sex education occurred in the 1960s and 1970s, just as research was beginning to document positive effects of women's colleges. There has been very little investigation of single-sex education at the secondary level, however. In this study, we compared the effects of single-sex and coeducational secondary schooling, using a random sample of 1,807 students in 75 Catholic high schools, 45 of which were singlesex institutions, drawn from High School and Beyond. Whether concerning academic achievement, achievement gains, educational aspirations, locus of control, sex role stereotyping, or attitudes and behaviors related to academics, results indicate that single-sex schools deliver specific advantages to their students, especially female students. In the recent focus on American secondary education, the relation between school organization and students' academic performance has been looked at critically. What has been considered by some to be an anachronistic organizational feature of schools may actually facilitate adolescent academic development by providing an environment where social and academic concerns are separated. Perhaps a second look at this disappearing school type is warranted.

In the 1960s and 1970s, there was a movement away from single-sex education at both the secondary and postsecondary levels. Social and economic reasons motivated these developments. Increasingly, single-sex education was viewed as a barrier to successful adolescent cross-sex socialization. Further, as the demand for single-sex education began to decline, institutions either closed or converted to coeducation to stabilize enrollments (Astin, 1977b; Block, 1984; Hyde, 1971). This trend occurred at precisely the time that research on postsecondary institutions was beginning to document positive effects for single-sex education on students' occupational achievement patterns, self-image, and career choice (Astin, 1977a, 1977b; Furniss & Graham, 1974). These effects appeared particularly strong for young women (Graham, 1970, 1974; Oates & Williamson, 1978; Tidball & Kistiakowsky, 1976). Compared with the postsecondary level, there has been
This research was funded primarily by a grant from the Radcliffe Research Support Program through the Henry A. Murray Research Center of Radcliffe College. Additional support came from the Educational Testing Service, where the first author served as a research fellow from 1984 to 1986. The conclusions, opinions, and findings expressed herein are those of the authors and not necessarily those of either the Murray Research Center or the Educational Testing Service. We wish to acknowledge the helpful comments offered by the reviewers in response to an earlier version of this article. We also wish to thank Letitia Pena for her assistance on the early parts of this project. Correspondence concerning this article should be addressed to Valerie E. Lee, School of Education, University of Michigan, Ann Arbor, Michigan 48109.
381

very little research on the effects of single-sex schooling at the secondary level. The limited available evidence on this topic is primarily from schools outside the United States. In large part, this is due to American single-sex education's now being confined almost entirely to the private sector. The recent and often critical focus on American secondary education has, however, renewed interest in questions about school organization. At least one senior educational researcher (Goodlad, 1984) had some doubts about the wisdom of coeducational secondary schooling and encouraged a closer empirical examination of this question. In essence, Goodlad expressed the same concern voiced by Coleman (1961) some two decades ago in The Adolescent Society. In this landmark study, which detailed the social life of teenagers in the 1950s, Coleman concluded that coeducation may be inimical to both academic achievement and social adjustment. Both of these studies suggested research on the question of single-sex schooling at the secondary level, particularly as it relates to the question of academic performance. In this article, we report on an investigation of the effects of single-sex and coeducational secondary school organization; whether these effects are different for male and female students; and whether the effects are sustained once adjustments are introduced for differences in student background, curriculum-track placement, and school social context. We compared the effects of these two types of school organization on the nature of students' engagement in school life in terms of their social and academic attitudes, schoolrelated behaviors, and courses of study. We also investigated the impact of single-sex education on students' academic achievement, educational aspirations, self-concept, locus of control, and views of adult sex roles. The data for

382

VALERIE E. LEE AND ANTHONY S BRYK

this study were drawn from High School and Beyond (HS&B), a national survey data base on secondary schooling in America. The analytic sample for this study consisted of a random sample of 1,807 students in 75 Catholic secondary schools, 45 of which were single-sex institutions. Data were gathered during the students' sophomore (1980) and senior (1982) years of high school. Background Postsecondary-Level Research

portant, a teacher can support and encourage her [the women's college student] vigorously in a fashion that is rare in institutions where both faculty and students often tend to regard male students as more meritorious of academic consideration, (p. 5)

Secondary-Level

Research

In a large study of more than 200,000 students from over 300 colleges, followed longitudinally, Astin (1977b) concluded, "single-sex colleges show a pattern of effects that is almost universally positive" (p. 41). The effects, generally stronger for women than men, included greater academic involvement, student interaction with faculty, verbal aggressiveness (for women), intellectual self-esteem, and satisfaction with most aspects of college life (except for social life for men). It was also found that graduates from men's colleges were more likely to complete career plans in law, business, and university teaching. Graduates from women's colleges were more likely to complete their bachelor's degree, to aspire to advanced graduate degrees, and to attain positions of leadership. Astin speculated that these advantages were probably due to the more circumscribed heterosexual activity and to a greater sense of identification and communal feeling when both students and faculty were predominantly of the same sex. Related research has been focused on the educational histories of "women achievers," defined by appearances in Who's Who (Oates & Williamson, 1978; Tidball & Kistiakowsky, 1976), by numbers of earned doctorates (Graham, 1970; Tidball, 1980; Tidball & Kistiakowsky), or by positions of leadership in the field of higher education (Graham, 1974). These researchers found that undergraduate education at a women's college, particularly at the Seven Sisters Colleges (Barnard, Bryn Mawr, Mount Holyoke, Radcliffe, Smith, Vassar, and Wellesley), was a major explanatory factor for women's patterns of career achievement. In none of these studies, however, were results adjusted cither for the greater affluence of students in private colleges or for organizational or resource differences between colleges. Further, Block (1984) suggested that these researchers failed to consider the possible greater career and academic motivation that might also be associated with attendance at one of the Seven Sisters Colleges. What are the characteristics of single-sex education that produce these beneficial effects for women? According to Tidball and Kistiakowsky (1976), the women's college provides "a favorable climate for women students . . . that conveys to them a sense of being in an environment where there are many other women seriously involved in a variety of academic pursuits" (p. 652). The presence of a high proportion of female faculty is another frequently cited factor. Graham (1974) stated the following: The teenage academic experience can be isolated from other compelling adventures at that age. Perhaps even more im-

The research comparing single-sex and coeducational secondary schooling has largely been conducted outside the United States (Dale, 1969, 1971, 1974, in the United Kingdom; Feather, 1974, in Australia; Finn, 1980, comparing U.S. and British schools; Jones, Shallcross, & Dennis, 1972, in New Zealand; Schneider & Coutts, 1982, in Canada). Much of this research has been focused on students' attitudes about the social and psychological environments of their schools rather than on the impact of this specific organizational factor on students' academic attitudes and behaviors. In few of the studies have achievement differences between the two school types been measured. These studies have typically involved a small, nonrandom sample of schools (i.e., the typical sample size ranges from 3 through 15 schools), and the data analyses have been primarily descriptive. The researchers generally have not adjusted for differences in the background characteristics of students attending coeducational versus single-sex schools and often have not considered whether the effects might be different for boys and girls. The studies are somewhat dated and may not reflect the effects of attitude changes about sex roles within the last decade. Finally, some authors (e.g., Dale, 1969, 1971, 1974; Hyde, 1971) are such zealous advocates of coeducation that some doubt is cast on the objectivity of their research findings. In studies focused on students' perceptions of their schools' environments, coeducational schools were generally found to have a friendlier and more relaxed atmosphere, providing more opportunities for pleasure-centered social contact (Dale, 1969, 1971; Feather, 1974; Jones ct al., 1972; Schneider & Coutts, 1982). Single-sex schools, especially those for girls, were considered to emphasize control and discipline to a greater extent. There was some disagreement, however, about actual academic emphasis between the two school types. Some researchers found girls' schools to evidence a more academic orientationa more competitive atmosphere, less free time, greater task emphasis, greater interest in academics, and more time spent on homework (Jones et al.; Trickett, Castro, Trickett, & Schaffner, 1982). Other research results contradict these conclusions, finding no positive relation between single-sex status and academic orientation (Dale, 1969, 1971; Dale & Miller, 1972; Feather, 1974). It is important to view this research within the socialhistorical context of the late 1960s and 1970s, when most of it was conducted. Many educational critics saw schools as oppressive institutions and argued for school reform that would make students' lives more enjoyable. The movement toward coeducation drew in part on such arguments. Order and discipline, which now tend to be viewed as characteristics of good schools (e.g. Colemen, Hoffer, & Kilgore,

EFFECTS OF SINGLE-SEX SECONDARY SCHOOLS

383 Method

1982), were generally interpreted as negative features within this research stream. Given this background, it is not surprising that researchers would examine the social and psychological environment of schools irrespective of its impact on academic pursuits. The relation between sex-stereotyped attitudes and behaviors and sex-segregated education has received some research attention. Trickett et al. (1982) found that girls in single-sex high schools in the United States showed a significantly higher level of interest in the feminist movement than did girls in coeducational schools. In a study of three Jewish high schools in New York, Winchel, Fenner, and Shaver (1974) found that fear of success (Horner, 1972) was greater among girls at a coeducational school than among those at a girls-only school. Further, the fear of success was even lower for those girls who had also had a singlesex elementary school experience. In a laboratory experiment, Lockheed (1976) found that adolescent female participation rates and leadership behavior in a game-playing situation were significantly increased by previous game experience in a single-sex condition. No significant singlesex effects for boys were reported in any of these studies. Relatively little research exists on academic achievement differences between single-sex and coeducational schools. Dale (1974) summarized several studies of British secondary schooling, using percentage success on O-level national examinations in various academic areas as a measure of school effects. He found a modest but consistent advantage for boys from coeducational schools but little difference for girls in the two types of schools. Because no statistical controls were introduced for either student or school background differences or for the percentage of students who actually attempted the examinations in each area, it is difficult to interpret the observed attainment differences. Finn (1980) reported on a cross-national study, using 1970 data from the International Association for Evaluation of Educational Achievement on attitudes and achievement in science and reading among 14-year-olds in both singlesex and coeducational schools in the United States and England. Although no differences were found between coeducational and a small sample (8) of single-sex schools in the United States, noticeable effects appeared in the British sample, which included 47 single-sex schools. Girls in English coeducational schools showed a decline in science and vocabulary, relative to male peers, but girls in single-sex schools excelled in reading and science. As Finn cautioned, however, interpreting these differences is again problematic because there are considerable other student- and schoollevel differences between coeducational and single-sex schools in England. In a recent study, Riordan (1985) compared white students in public schools with those in Catholic single-sex and coeducational schools. Using the 1972 National Longitudinal Study (NLS) data, Riordan found positive effects in achievement for single-sex schools, particularly for girls. The only statistical controls used were for students1 social class and census region. These statistical adjustments were limited of necessity in that only senior-year data are available in the 1972 NLS.

Sample and Data


Because single-sex secondary education in the United States is rarely available within the public sector, we turned to private secondary schools for our investigation. Fortunately, in a national survey on American secondary education, High School and Beyond (HS&B), sponsored by the National Center for Education Statistics in 1980, private schools were deliberately oversampled. The HS&B survey design involved a two-stage probability sample of high schools and students within them. The original sample consisted of 1,015 schools and approximately 36 sophomores and 36 seniors within each school. The sophomores in 1980 were followed up as seniors in 1982, and it was from this panel file that the data for this study were drawn. We limited the investigation to Catholic schools for three reasons. First, the number of non-Catholic private schools (27) in the HS&B sample was small. Second, previous analyses of these data indicated that this group of schools is quite heterogeneous and that they are different in important ways from the Catholic sector (Coleman et al., 1982). As a result, it was not sensible to pool the Catholic and other private secondary schools. Similarly, because of the heterogeneity among the non-Catholic private schools and the small size, the data were insufficient to sustain separate analysis (Bryk, 1981; Goldberger & Cain, 1982; Murnane, 1981). Third, the subsample of Catholic schools provided a nearly ideal natural experiment for investigating the effects of single-sex education. The number of students and schools of each type was large (see Table 1), relative to the previous research on this topic. Further, although single-sex and coeducational Catholic schools differ in a number of ways besides their sex composition (see description following), the degree of confounding was modest, and the HS&B data set was sufficiently extensive for statistical controls to be introduced during the course of the analysis to adjust for differences in student background, curriculum track, and the social context of the schools. Finally, because the survey design for HS&B consisted of random sampling of both schools and students within the Catholic sector, the results of our investigation can be generalized to the majority population of American Catholic schools. In developing our analysis plan, we decided to conduct separate analyses by gender because the existing research suggests that the effects of single-sex education may be different for male and female students. Tables 2 and 3 present a detailed statistical profile for each sex-by-school-type group on the background and outcome variables of interest in this study. In developing the analytic sample, we applied the following data filters: Only students enrolled in the same school in 1980 and 1982 were included. Students in the HS&B baseline Catholic sample who transferred, graduated early, or dropped out were not included. This resulted in the loss of 305 students, or 11% of the total. In addition, we deleted eight schools (230 students) from the Catholic sample, which had a disproportionately large enrollment of vocational students (over 25%). In another recent study of Catholic secondary schools (Bryk, Holland, Lee, & Carriedo, 1984), it was found in field research that a small group of girls' schools specialized in stenographic and clerical training and was, therefore, atypical of the Catholic sector as a whole, which emphasizes a traditional academic program. Of the eight Catholic schools on the HS&B file with more than 25% vocational enrollment, five were girls' schools, and three were coeducational schools. No boys' schools fell into this category. Thus, by eliminating these schools, we reduced the sample to a relatively homogeneous set of schools with similar instructional purposes across the four

384 Table 1 Definition of the Analytic Sample


Type of school Student sample Coeducational Girls Boys Single-sex Girls Boys Total School sample Boys only Coeducational Girls only Total

VALERIE E. LEE AND ANTHONY S. BRYK Thus, a picture showing somewhat different types of students in single-sex and coeducational schools emerged. Although boys1 schools had the highest minority enrollment, their students were also more advantaged than their coeducational counterparts, as measured by ratings of social class. Students at girls' schools, on the other hand, were less advantaged and, generally, were more similar to girls in coeducational schools. Blacks in Catholic secondary schools were most likely to be found in boys' schools, whereas Hispanics were more likely to be found in girls' schools. In general, coeducational schools fell between the two types of single-sex schools on most dimensions that involved background measures. Curriculum-track placements were quite similar for girls across the two school types, whereas boys in boys' schools were somewhat more likely to be in the academic (i.e., college preparatory) track than were their coeducational counterparts. Almost no students in boys' schools were in the vocational track. Those students in high schools with high vocational enrollments that were deleted from the original HS&B sample of Catholic high schools (202 girls and 28 boys) were similar to the remaining analytic sample in regard to race and ethnic composition, religious orientation, grade-repeating history, and family financial sacrifice. They differed substantially, however, from the remaining sample on two dimensionssocial class (averaging - .361) and percentage of students with college plans in the eighth grade (averaging 49.9%). That this group of students, 88% female and 68% from single-sex schools, was both less advantaged and less educationally ambitious at entry into high school argued for its atypicality from the majority of Catholic school students and justified its exclusion from this study. To examine more fully the nature of the relation between background variables and the type of school a student attends, we undertook two types of multivariate analyses. The first approach involved a discriminant analysis relating personal, family, and academic background characteristics to the four sex-by-schooltype groups. Two statistically significant functions emerged. The first distinguished the two school types, with having higher social class, being black, coming from a two-parent family, having a public elementary school background, and being less religious predicting attendance at a single-sex school. The second function discriminated between the sexes, with greater grade repeating, less financial sacrifice, and higher social class more likely characteristics of boys than girls. That two functions were required to distinguish among the four groups provided some additional empirical support for the analytic decision to treat the effects of single-sex schools separately by sex. Second, we performed a regression analysis for each gender group to predict attendance at a single-sex school as a function of personal, family, and academic background variables. These "selection equations" were particularly helpful in identifying the pattern and magnitude of the confounding of background characteristics with attendance at a single-sex school. We identified and adjusted for these confounding relations before any attempt was made to interpret outcome differences among the four groups. Failure to consider such confounding variables is a major weakness of much of the past research on effects of single-sex schooling. In general, the degree of confounding in this natural experiment was slight, although it was somewhat stronger for boys. The selection equation for boys' schools explained 6.9% of the variance in the dichotomous-choice variable versus 3.1% for the girls' schools. College expectations at the eighth grade and public elementary school experience were statistically associated with attending a single-sex school for boys but not for girls. Greater family financial sacrifice and a less religious orientation were significantly linked to attending a single-sex school for both sexes.

No. students3 or schools

452 382 474 499

1,807

21 30 24 75

Note. The sample sizes reported in this table are unweighted. Because certain schools in the High School and Beyond (HS&B) original sample, particularly in the Catholic sector, were oversampled, the weighted and unweighted sample sizes for any breakdown such as the above will vary. We weighted all analyses reported in subsequent tables by using the PNLTSTWT from the HS&B public-use files. a For this first sample, we excluded all students in schools with more than 25% vocational-track enrollment for reasons explained in the text. We excluded eight schools (three coeducational and five girls' schools) and 230 students (155 from girls' schools, 47 girls and 28 boys from coeducational schools). We did not exclude any boys' schools.

sex-by-school-type categories. As the data in Table 2 document, the four groups were well matched on academic-track distribution, which we interpreted as a reflection of the homogeneity of students' (and schools1) educational purpose. Background differences. Examination of the personal and family background variables in Table 2 indicates that although the four sex-by-school-type groups were well matched on academictrack distribution, there were some background differences worth noting. In general, boys were more likely to have repeated a grade in elementary school, whereas girls in coeducational schools were particularly low on this measure. Plans for college attendance at entry into high school were quite similar across school types, with boys in coeducational schools somewhat lower than other groups. Boys in Catholic high schools, especially those in boys' schools, came from more advantaged homes, as indicated by social class. Financial sacrifice, represented by tuition as a proportion of discretionary family income, was slightly higher in single-sex schools (9.2%) than in coeducational schools (7.8%). This indicated that the families of students at single-sex schools made a greater financial sacrifice in sending their children to such schools. Students who transferred to the Catholic sector from a totally public elementary school background were more likely to choose single-sex schools, and those from Catholic elementary schools were more likely to be found in coeducational schools. Paralleling the data on elementary school experience, students in single-sex schools considered themselves to be less religious than their counterparts at coeducational schools. All three of these trends were stronger for boys than girls. The enrollment of minority students was somewhat higher in single-sex schools, especially in boys' schools. Girls in coeducational schools were somewhat more likely to come from single-parent homes.

EFFECTS OF SINGLE-SEX SECONDARY SCHOOLS

385

Table 2 Characteristics of Student Sample


Girls in coed schools 78.2% 13.7% 8.1% .094 4.6% 6.4% 20.1% 8.1% .407 62.3% 16.2% 4.8% 76.6% .082 Girls in girls' schools 76.7% 17.4% 5.8% .116 2.7% 10.1% 16.1% 9.9% .129 55.6% 23.1% 6.7% 76.4% .094 Boys in coed schools 72.0% 21.6% 6.4% .192 3.3% 4.0% 14.3% 8.5% .352 64.5% 16.6% 9.3% 68.0% .073 Boys in boys' schools 80.5% 16.2% 3.3% .234 7.1% 8.5% 12.4% 7.8% .073 54.2% 27.6% 8.2% 77.4% .089

Variable Academic tracks % academic track % general track % vocational track Personal and family background Social classa>b % black % Hispanic % single-parent family Religious characteristics % non-Catholic Religiousnessa>c Elementary school experience All Catholic All public Academic background and orientation % repeated elementary grade % college plans at Grade 8 Financial sacrificed
a

These variables were standardized (M - 0, SD = 1.0) on the basis of the entire Catholic-school sample, without consideration for weighting. Because a portion of that sample was excluded from this investigation and because the above means were computed with the weighting factor applied, the statistics reported above deviate somewhat from those parameter values. ^The social class variable is the sum of standardized values for (a) parental income, (b) parental education (higher of mother's or father's), (e) parental occupation (higher of mother's or father's), and (d) number of educationally related household possessions. This is a composite factor that included student self-reports of frequency of attendance at religious services and degree of religious conviction. ^Financial sacrifice was operationalized as the school's tuition as a percentage of disposable family income (adjusted for family size). The specific income adjustments were derived from subsistence-income figures supplied by the Department of Labor. The base for a single person was $3,400, and there was an increment of $1,100 per additional family member. Catholic high school tuitions averaged $851 in 1980. Outcome differences. The breadth of outcome measures contained in HS&B was one of the major strengths of this data base. It permitted us to investigate the effects of single-sex schooling on academic and social attitudes; school-related behaviors; course enrollment; academic achievement in reading, writing, mathematics, and science; future educational aspirations; locus of control; self-concept; and sex role attitudes. In addition, the longitudinal nature of the HS&B data permitted, at least for selected measures, an assessment of change from sophomore to senior year as well as an assessment of status at each time point. The mean outcome differences across school types presented in Table 3 suggest a wide range of potential effects of single-sex schooling. Girls in girls' schools expressed a more positive attitude toward academics, whether characterized by their friends or their interests in specific subjects, such as English and mathematics. This was true in comparison with both boys in general and their coeducational counterparts in particular. They engaged in more behavior generally considered to be supportive of schoolwork; that is, girls in single-sex schools spent more time on homework, watched less television, and were less likely to be absent or experience discipline problems. Boys in the two types of schools showed certain differences in attitudes and behaviors that were not seen for girls. For example, boys in coeducational schools evidenced less positive attitudes toward academic peers, athletes, and socially active peers than did students at boys' schools. Boys in boys' schools took considerably more mathematics and science courses and fewer vocational and social science courses than did either their counterparts in coeducational schools or girls in either type of school. Boys in single-sex schools showed somewhat higher achievement levels than did their counterparts in coeducational schools at both sophomore and senior year. Although achievement for girls at sophomore year in coeducational and single-sex schools looked similar, by senior year the achievement in single-sex schools was higher in reading, writing, and science. Both single-sex groups were less likely to hold stereotypic views of women's sex roles, and these differences were especially large for girls. Girls in both types of schools gained less in mathematics and more in reading than did boys. Regardless of school type, boys scored higher than girls in math and science and lower in writing.

Analytic Approach
We chose to report both separate cross-sectional results at sophomore and senior years and sophomore-to-senior change for several reasons. First, some of the outcome variables of interest were measured at either sophomore or senior year only, so change analyses were not possible for these. Second, for certain outcomes, such as attitudes toward academics or amount of homework, it was reasonable to hypothesize that these attitudes and behaviors are formed early in the secondary school experience and remain relatively constant throughout the secondary years. Under such a model of school effects, we would expect to find differences at sophomore year that persist through senior year but do not necessarily increase. To use only change from sophomore to senior as a measure of school effects would yield an underestimate of the schools' true effects. Third, although drawing inferences about school effects from cross-sectional data can be problematic, difficulties are also encountered with change analyses. In particular, sophomore year was purposely used as a base for HS&B because

386
Table 3

VALERIE E. LEE AND ANTHONY S. BRYK

Outcome Differences of Student Sample


Girls in coed schools .341 .364 .100 .291 .147 Girls in girls' schools .438 .482 .433 .459 .206 Boys in coed schools .381 .269 .130 .367 .149 Boys in boys' schools .415 .325 .003 .240 .319

Variable School-related attitudes (sophomore year only) Interest in mathb b Interest in English Association with academically oriented friends3 Attitudes toward socially active peers3 Attitude toward student athletes3 School-related behaviors (sophomore year only) Incidence of disciplinary problems3 Uncxcuscd absences (number in 3-month period) Homework (hours/week) Television watched (hours per weekday) Course enrollment (measured only at senior year) Mathematics (years) Physical science (years) Vocational (no. courses) Social studies and history (years) Reading achievement Sophomore year Senior year Reading gainc Mathematics achievement Sophomore year Senior year Mathematics gain Science achievement Sophomore year Senior year Science gain Writing achievement Sophomore year Senior year Writing gain Educational aspirations'1 Sophomore year Senior year Aspirations gain Locus of controF Sophomore year Senior year Locus gain Self-concept6 Sophomore year Senior year Concept gain Sex role stereotypingf Sophomore year Senior year Stereotyping gain

.370 1.47 5.76 3.23 3.14 0.75 1.56 3.21 10.47 11.88 1.41 20.89 22.74 1.80 11.85 12.52 0.68 12.98 14.25 1.27 7.03 6.98 -0.02 .230 .215 -.015 -.031 .029 .067 6.80 6.29 -0.51

.351 1.27 6.93 2.91 3.30 0.71 1.72 4.08 10.89 12.66 1.72 20.58 22.66 1.99 11.75 13.04 1.26 13.00 14.43 1.36 7.40 7.46 0.02 .312 .317 .007 .095 .103 -.020 6.52 5.85 -0.73

.058 1.50 4.89 3.38 3.15 0.84 1.65 3.90 10.59 11.85 1.26 21.53 24.24 2.70 12.84 13.59 0.77 11.13 12.75 1.63 7.48 7.68 0.22 .322 .378 .056 .106 .035 -.074 7.65 7.41 -0.27

.229 1.05 5.99 3.59 3.78 1.14 1.25 3.03 11.64 12.78 1.15 23.30 25.79 2.49 12.97 13.79 0.83 12.00 13.19 1.22 8.26 8.42 0.11 .198 .199 -.047 .173 .163 -.013 7.63 7.21 -0.44

These variables were standardized (M = 0, SD = 1.0) on the basis of the entire Catholic-school sample, without consideration for weighting. Because a portion of that sample was excluded from this investigation and because the above means were computed with the weighting factor applied, the statistics reported above deviate somewhat from those parameter values. ''These are dummy-coded variables, so means are percentages of group answering yes. cGains were computed as the difference between scores at senior and sophomore years. Means for gains do not always equal the exact difference, because of possible missing data at either year.

EFFECTS OF SINGLE-SEX SECONDARY SCHOOLS many public high schools enroll students in Grades 10 through 12 only. Catholic secondary schools, however, typically begin at Grade 9. Thus, the status at the end of sophomore year includes 2 years of a potential effect of a single-sex school. Consequently, the sophomore results are not pure pretreatment measures. Controlling on the sophomore result in fact partialed out a portion of the effect of a single-sex school that was expected to result between sophomore and senior years. (For a more detailed discussion of this problem, see Anderson et al., 1980, chapter 12.) In sum, there are reasons to believe that the change analyses would tend to underestimate the true effects of single-sex schools. On the other hand, it is widely assumed that the statistical adjustments used with cross-sectional analyses are often inadequate and underadjust for differences in student characteristics across schools. This argument implies that the cross-sectional results might tend in this case to overestimate the effects of single-sex schools. Thus, the combination of cross-sectional and longitudinal analyses provides a degree of balance to these competing biases. Before we could conclude, however, that the observed outcome differences reported in Table 3 were attributable to single-sex education, we had to consider two major alternative explanations. First, as already mentioned, although this was a well-matched natural experiment, there were some differences in student background among the four groups that could potentially explain part or all of the observed outcome differences in Table 3. Thus, the first analytic problem involved estimating effects of single-sex schools that were adjusted for these potentially confounding background differences. In addition to the adjustments on student-level background variables, we introduced statistical controls for school social context and curriculum track. Previous researchers who have used HS&B have shown that school social context is an important predictor of academic achievement above and beyond the effects of studentlevel social demographic characteristics. In research on privateschool effects, the inclusion of social context variables has resulted in substantially different effect estimates (Hoffer, Greeley, & Coleman, 1985; Willms, 1985). Similarly, curriculum track has also been found to be strongly associated with course enrollment, achievement, and educational aspirations (e.g., see Alexander & Pallas, 1983; Coleman etal., 1982; Hoffer et al., 1985). Although the four groups were fairly well balanced on academic-track distribution (see Table 2), we included it because of the strength of its relation to the academic outcomes of interest in this study. Specifically, we regressed each outcome variable against a model consisting of a dummy variable for single-sex versus coeducational schools, personal and background characteristics (socioeconomic status, black and Hispanic racial/ethnic status, single-parent family, non-Catholic, religiousness, all-Catholic or all-public elementary school experience, whether student repeated an elementary grade, whether student had plans at eighth grade to attend college, and financial sacrifice), two dummy variables for academic- and general-track membership, and three measures of school social context (average social class, percentage black students, and percentages Hispanic students). For the change analyses, we

387

introduced the sophomore status as an additional covariate to the background, track, and school-composition measures. Assuming that single-sex effects persisted after statistical adjustments, we still had to consider whether school factors other than single-sex-school organization could account for the observed differences. Unfortunately, the school sample size precluded any further regression modeling at this level of analysis. The adjustment model just described already had four school-level variables in it. As an alternative, we undertook a descriptive analysis that allowed us to at least explore other possible dimensions along which single-sex and coeducational schools vary and to consider whether these variations offer plausible alternative explanations for the observed results.

Results Estimating Single-Sex-School Effects After Adjustment for Personal and Family Background, Curriculum Track, and School Social Composition
Tables 4 through 6 present estimates of the effects of single-sex schools on a range of outcome measures separately for boys and girls. The effect estimate presented for each outcome measure includes adjustments for family and personal background, curriculum track, and school social composition. AIL estimates are reported in an effect-size metric (Light & Pillemer, 1984) to facilitate comparison across outcome domains. Unstandardized regression coefficients and their standard errors are also reported. In addition, significance testing included a design-factor correction for the two two-stage sampling plan used in HS&B. We chose to use one-tailed tests of the statistical significance of these effects because the hypotheses we were testing were directional in nature. That is, we hypothesized that students in single-sex schools would significantly outperform their counterparts in coeducational schools on the wide array of outcomes considered. Single-sex-school effects on attitudes, behaviors, and course enrollment. Girls' schools evidenced consistent and positive effects on student attitudes toward academics (see Table 4). These students were more likely to associate with academically oriented peers and to express specific interests in both mathematics and English. The effects of boys' schools on academic attitudes were also generally positive but weaker than in the girls' schools and not statistically significant. Students in boys' schools were more positive toward socially active peers and athletes. With regard to school-related behaviors, students in single-sex schools did somewhat more homework, and this was especially true for girls. Students in boys' schools were much more likely to enroll in a larger number of mathe-

A 9-point scale of students' self-assessments of the educational level that they think they will attain, ranging from less than high school (1) to PhD or MD (9). Graduating from college was rated 7. These variables were' standardized (M = 0, SD = 1.0) on the entire High School and Beyond sample, the vast majority of which was in public school. Ratings for Catholic-school students on these variables were generally higher than those for public-school students. This is a composite factor based on the degree of students' agreement with the following statements: (a) Working mothers of preschool children can still be good mothers (reverse coded); (b) men should work while women care for the home and family; and (c) women are happiest making a home and caring for children. Each item was rated on a 4-point scale. The reported scores are the raw sums across the three items (mininum = 3, maximum = 12), with higher values indicating more stereotyping attitudes.

388

VALERIE E. LEE AND ANTHONY S. BRYK

Table 4 Estimated Effects of Attending a Single-Sex School on Student Attitudes, Behaviors, and Course Enrollment
Girls Variable School-related attitudes Interest in math Interest in English Association with academically oriented friends Attitudes toward socially active peers Attitude toward student atheletes School-related behaviors Incidence of disciplinary problems Unexcused absences Homework Television watched Course enrollment Mathematics Physical science Vocational Social studies and history Effect sizea .23* .26* .23*
.10

Boys Standard error .037 .039 .116 .162 .136 Effect sizea
.12 .14 .04

Regression coefficicntb .111 .126 .383 .215 -.330

Regression coefficientb .060 .060 .071 .585 .572 -.186 -.469 .742 .319 .609 .336 -.501 -.041

Standard error .042 .039 .138 .183 .165 .099 .195 .283 .151 .089 .061 .132 .072

.26* .30* -.15 -.19 .23*


.17

-.18

-.03 -.11 36*** -.15 .16* -.02


.11 .12

-.029 - .265 1.232 -.259 .213 -.016 .144 .107

.070 .166 .259 .134 .086 .053 .101 .065

4 0 *** 40*** -.26** -.05

Note. Effects were calculated with ordinary least squares regression and include adjustments for personal and family background (socioeconomic status, black and Hispanic race/ethnicity, single-parent-family status), religious characteristics (non-Catholic, religiousness, all-Catholic or all-public elementary school experience), academic background and orientation (repeated elementary grade, college plans in Grade 8, family financial sacrifice), school social context (average socioeconomic status, percentage black students, percentage Hispanic students), and academic curricular track. Group mean differences on these variables are described in Table 2. The effect sizes, presented as standard-deviation units, were calculated by dividing the unstandardized regression coefficient for the single-sex dummy variable by the standard deviation of the appropriate dependent variable among students of that gender in coeducational schools. The proportion of variance explained by these models was not large and varied little across genders. Course enrollment was explained better than behaviors, which in turn were better explained than attitudes. The R2 figures averaged about .10 to .15. a A correction factor of 1.5 was introduced for the design factor associated with the two-stage probability-sampling plan. The use of the 1.5 design effect is based on empirical studies of the efficacy High School and Beyond of the sampling plan (National Center for Education Statistics, 1983, pp. 31-32). These conventions apply to all significance levels reported in this article. bThese are the unstandardized regression coefficients for the single-sex dummy variables in regression models. We have also reported the nominal standard errors associated with each regression coefficient. This information allows the reader to recompute probability values both for different design-effect estimates or difference significance criteria. The nominal standard error is multiplied by the design effect and used as the denominator in a standard /-test computation. *p < .05. **p < .01. ***p < .001, all one-tailed.

matics and science courses and less likely than their coeducational-school counterparts to enroll in vocational courses. The course-enrollment pattern for girls was more similar across the two school types. Girls in single-sex schools took more mathematics courses, and although it was not statistically significant, they took slightly more vocational offerings and social sciences than did their coeducationalschool counterparts. Effects on academic achievement. Table 5 presents the estimates of single-sex-school effects on reading, mathematics, science, and writing achievement separately at sophomore and senior years as well as change scores. The adjusted single-sex-school effects on academic achievement were generally positive. There were no achievement areas in which coeducational-school students surpassed their single-sex-school counterparts at either sophomore or senior year.

The pattern of effects was different for male and female students. The estimated effects of boys' schools were largest at sophomore year and, although still positive, were somewhat diminished by senior year. The boys' schools did not display any statistically significant sophomore-tosenior gains. For girls' schools, however, the estimated effects increased in size from sophomore year to senior year, and the gains in reading and science achievement were statistically significant. The girls' school effects on academic achievement were particularly salient in the light of their students' not taking more academic courses than their coeducational-school counterparts (except in mathematics, where no gain was seen). As a result, the effects could not be attributed to course taking, which was the major explanatory factor for secondary school achievement in these data (Bryk et al., 1984). Other aspects of the students' engagement in school

EFFECTS OF SINGLE-SEX SECONDARY SCHOOLS

389

Table 5 Estimated Effects of Attending a Single-Sex School on Student Achievement at Sophomore and Senior Years and on Sophomore-to-Senior Achievement Gains
Girls Variable Reading achievement Sophomore year Senior year Reading gain Mathematics achievement Sophomore year Senior year Mathematics gain Science achievement Sophomore year Senior year Science gain Writing achievement Sophomore year Senior year Writing gain Effect size
.11

Boys Standard error .255 .259 .190 .455 .502 .321 .228 .222 .172 .212 .175 .134 Effect size .20*
.18 .05

Regression coefficient" .410 .777 .494 -.261 .047 .268 -.139 .543 .630 .030 .205 ,189

Regression coefficient3 .735 .692 .196 1.801 1.351 .014 .107 .033 .024 .830 .255 -.165

Standard error .280 .294 .227 .511 .545 .394 .243 .237 .201 .263 .249 .213

.21* .14* -.04


.01 .04

26** .18*
.00 .01 .01 .01

-.05
.17

.20**
.01 .08 .07

.24*
.08

-.05

Note. Effects were calculated with ordinary least squares regression and include adjustments for the variables listed in the note of Table 4. Gains include additional control for sophomore status on particular dependent measures. The effect sizes, presented as standard-deviation units, were calculated by dividing the unstandardized regression coefficient for the single-sex dummy variable by the standard deviation of the appropriate dependent variable among students of that gender in coeducational schools. a These are the unstandardized regression coefficients for the singlc-scx dummy variables in regressions described in the note above. *p < .05. **p < .01, both one-tailed.

life, the nature of the environmental press, and social organizational features of single-sex schools are all possible explanations for the observed effects. Effects on educational aspirations, locus of control, selfconcept, and sex role attitudes. Here, too, the estimated effects generally favored single-sex schools, with larger effects accruing for girls' schools (see Table 6). The estimated girls' school effect on educational aspirations was statistically significant at sophomore and senior years and for the sophomore-to-senior gains. Girls' schools also displayed a statistically significant positive effect in senioryear locus of control and sophomore-year self-concept. None of the estimated boys' school effects were statistically significant. The final outcome variable measured students' views of adult sex roles. This was a composite of attitudes about the compatibility of work and motherhood, the traditional role of men as achievers, and whether most women are satisfied with home and child care rather than careers. Girls' school students were considerably less likely to evidence stereotyped sex role attitudes than were comparable girls in coeducational schools. Further, the stereotyping attitudes of students in girls' schools showed a statistically significant decline from sophomore to senior years. Interestingly, although students in boys' schools were slightly more sex role stereotyping in sophomore year, this difference disappeared by senior year. These results suggest that stereotyping attitudes are not an inevitable consequence of an all-male environment.

Are these effects large or small? The reporting of research results in standardized effect sizes is becoming increasingly common because it provides a convenient metric for research synthesis (see, e.g., Light & Pillemer, 1984). It was for this reason that we reported our results in Tables 4 through 6 in this manner. A proper interpretation of the substantive significance of any estimated effect, however, requires reference to some external standard, particularly because the nature of tests of statistical significance is approximate. Depending on the standard chosen, one can potentially draw very different conclusions about the importance of the estimated effects. For example, an effect size of .20 means that the probability that the score of a randomly chosen student from a single-sex school will exceed the score of a randomly chosen counterpart from a coeducational school is approximately .54. From this point of view, the statistically significant results reported in this article might appear small. Hoffer et al. (1985), however, showed that a standardized effect size of .20 on the gain from sophomore to senior year in academic achievement is equivalent to the amount of learning of the average public-school student during 1 full year of secondary school instruction. That is, when compared to what is typically learned in a year of instruction, an effect size of .20 is quite large. It implies that attendance at a girls' school would increase science achievement by approximately 1 year's growth, or half of the amount typically learned over the last 2 years of secondary schooling.

390

VALERIE E. LEE AND ANTHONY S. BRYK

Table 6 Estimated Effects of Attending a Single-Sex School on Educational Aspirations, Locus of Control, Self-Concept, and Sex Role Stereotyping (Sophomore and Senior Years and Sophomore-to-Senior Gains) Girls Variable Educational aspirations Sophomore year Senior year Aspirations gain Locus of control Sophomore year Senior year Locus gain Self-concept Sophomore year Senior year Concept gain Sex role stereotyping Sophomore year Senior year Stereotyping gain Effect size .19* 23** .15* .16 .21* .14 .18* .10 .02 .16 -.25* -.17* Regression coefficient* .374 .408 .263 .084 .113 .073 .133 .072 .017 -.308 -.431 -.298 Standard error .123 .115 .106 .039 .041 .036 .054 .055 .050 .140 .124 .109 Effect size .13 .11 .07 .18 -.04 -.12 .08 .12 .09 .15 -.03 -.09 Boys Regression coefficient8 .266 .207 .120 .108 -.026 -.073 .049 .081 .062 .227 -.040 -.138 Standard error .207 .122 .114 .047 .049 .045 .050 .056 .053 .135 .136 .123

Note. Effects were calculated with ordinary least squares regression and include adjustments for those variables listed in the note of Table 4. Gains include additional control for sophomore status on particular dependent measures. The effect sizes, presented as standard-deviation units, were calculated by dividing the unstandardized regression coefficient for the single-sex dummy variable by the standard deviation of the appropriate dependent variable among students of that gender in coeducational schools. a These are the unstandardized regression coefficients for the single-sex dummy variables in regressions described in the note above. *p < .05. **p < .01, both one-tailed. The reasons for these apparently conflicting inferences were explained in detail by Jencks (1985). In brief, a great deal of learning has occurred by sophomore year in high school. As a result, student performance is extraordinarily variable, and comparing an estimated school effect against this standard produces a small standardized estimate. When the estimated effect is compared against what students typically learn in a year of instruction, however, a very different picture emerges. Another way to approach concerns about the substantive significance of the estimated effects is focused on the fact that single-sex-school organization is a school-level variable. This suggests that the estimated effects associated with single-sex organization, relative to the amount of variation naturally occurring between schools (rather than between students), be considered. If the natural variation between schools is large, then the estimated single-sex-school effects may be of limited importance, relative to other school characteristics that might be studied. Unfortunately, the typical procedures for estimating variance components in nested models produce biased estimates when the data are unbalanced, as in the HS&B sample used in this investigation. As an alternative, we used the BMDP3V program (Jennrich & Sampson, 1983), which allows estimation of the between-school variance component by a procedure known as restricted maximum likelihood estimation. The latter provides a consistent estimate of the variance between schools under the assumption that the outcome variables are normally distributed. Our analytic model treated schools as a random factor and adjusted for the same set of student-level covariates (specified as fixed factors) that were used in estimating the effects in Tables 4 through 6. Because the computing routine is iterative and, as a result, the computational costs are more substantial, we limited our investigation to only those outcomes for which statistically significant results appeared in Tables 4 through 6. We estimated the between-school variance for each of these outcomes. We then computed new effect-size estimates for each dependent measure so that the standard deviation between schools rather than between students was used as the denominator in the effect-size calculation. The results are presented in Table 7 along with the corresponding student-level effect sizes previously reported. It is clear that the estimated single-sex-school effects represented a very large portion of the total variability occurring between Catholic secondary schools. Although the effect sizes based on the between-student variation averaged only about .25 standard-deviation units, over half of the effect sizes based on the between-school variation were in excess of 1.0. When viewed against the total variation naturally occurring among schools, the estimated single-sex effects appear quite substantial. Alternative Factors That Might Account for the Observed Effects When attempting to draw causal inferences from nonexperimental research, plausible alternative explanations might

EFFECTS OF SINGLE-SEX SECONDARY SCHOOLS Table 7

391

Comparison of Estimated Effect Sizes in Terms of Student- Versus School-Level Variability


Effect size based Effect size based on variability on variability between studentsa between schools1" Girls Interest in mathematics Interest in English Association with academically oriented friends Homework, sophomore year Mathematics courses Reading achievement, senior year Reading achievement gain Science achievement gain Educational aspirations, sophomore year Educational aspirations, senior year Educational aspirations gain Locus of control, senior year Sex role stereotyping, senior year Sex role stereotyping gain Boys Attitudes toward socially active peers Attitude toward student athletes Homework, sophomore year Mathematics courses Science courses Vocational courses Reading achievement, sophomore year Math achievement, sophomore year Writing achievement, sophomore year Math achievement, senior year
a

Variable

.23* .26* .23* .36*** .16* .21* .14* .20** .19** .23* .15* .21* - .25* - .17*

1.35 1.95
.96 .67 .84 .81

1.03 2.14

1.44 1.14 1.16 -1.13

tially different from the simple mean differences on these outcomes. It is also often argued that such statistical adjustments are by their nature insufficient. This argument seems implausible in this instance, however, because the effects of the adjustments often increased the size of the estimated single-sex-school effect rather than diminished it. Further, effects persisted, at least for girls' schools, even in the relatively conservative change analysis. On balance, it is always possible that some other measured or unmeasured confounding factor could account for the observed effects. We made a conscious choice to include only those personal and background factors that are relatively static over time and, as a result, are more likely to reflect a student's status at high school entry (social class, family status, financial sacrifice, race/ethnicity, and religion) or those variables that describe a student's academicbackground status at entry (elementary school experience, grade-repeating history, college plans in Grade 8, and track placement). Other researchers might consider sophomore-year measures of academic orientation, amount of homework done, and incidence of disciplinary problems as additional background differences that should have been adjusted away. The case here is arguable. We tend to agree with Coleman et al. (1982) that to treat these data as covariates (and thereby to assume that they are a priori background factors uninfluenced by exposure to school type) seems highly questionable when the data are collected almost half way through the secondary school experience. It seems more appropriate to us to treat these variables as intermediate outcomes that are influenced by student background and school characteristics. To be sure, these variables may in turn influence such ultimate outcomes as growth in academic achievement and educational aspirations. Should such a finding emerge from subsequent research, it would not, however, negate the positive effects associated with single-sex schools. Rather, it would provide a basis for developing a better explanation for the actual source of these effects. Of course, the strength of this study (as well as any study using HS&B data) would be further reinforced if controls for students' achievement, ability, and attitudes prior to high school entry were available. Perhaps the most difficult selection hypothesis to counter is that families specifically choose single-sex schooling because of presumed benefits for their children. In particular, perhaps families with higher academic orientations and educational aspirations are more likely to choose this schooling option. The data reported in Table 2 on students' eighthgrade educational plans provide some evidence on this issue. A higher proportion of the students in boys' schools indicated that they had college plans at eighth grade than did their coeducational school counterparts. Thus, the data for boys' schools indicate that some differential selection was occurring. Although we introduced explicit statistical controls for eighth grade plans, we always remain open to the possibility that we underadjusted for the true structural effects.

.26**
30* 23*

1.82

.51

40***
26* 20*

1.31 1.15 -1.39


.84

26**
24* 18*

1.19
.93 .75

These effect sizes are identical to those reported in Tables 4, 5, and 6. bThese effect sizes were calculated by dividing the unstandardized regression coefficient for the single-sex dummy variable by an estimate of the standard deviation between schools on the corresponding dependent variable. The latter were estimated by using the BMDP3PV program (Jennrich & Sampson, 1983) with restricted maximum likelihood estimation. For some dependent measures, BMDP3PV produced an estimate of zero betweenschools variance. These are marked by in the table. */? < .05. **p < .01. ***/? < .001, all one-tailed.

be offered for any set of results. The most frequently encountered objection is a selection hypothesis; that is, the individuals within the groups vary in important ways besides the exposure to a different program. In response to this concern, we introduced statistical controls for individual background and track placement and for school-context differencesthe sorts of factors usually cited in selection hypothesesin the analyses described in the last section. The pattern of results after adjustments was not substan-

392

VALERIE E. LEE AND ANTHONY S. BRYK 8 indicate that girls' schools were smaller and had fewer students per teacher than did coeducational schools. These features suggest a potentially more intimate environment that might contribute to the girls' school effect. To the extent, however, that there were boys1 school effects, we would have to look elsewhere for an explanation in that the boys' schools had the largest size of the three groups and the greatest student-to-teacher ratio. Another possible explanation draws on the fact that single-sex schools of both gender compositions offer a less diverse array of courses. As a result, students are more likely to take a similar course of study, and this homogeneity of experiences could conceivably produce beneficial effects. The mechanism that produces such effects, however, would have to be rather subtle. It is not simply a matter of greater enrollment in academic courses; HS&B data indicated that students in girls' schools did not take more academic courses than did their coeducational school counterparts. It is possible that the general content of the courses in single-sex schools is somewhat more academic in nature. Alternatively, the curriculum homogeneity may contribute to a stronger academic press in the environment whose influences stand apart from course-taking effects. In regard to staff variables, the boys' schools had more resources, as measured by percentage of faculty with advanced degrees, low teacher turnover, and high teacherstability rates. The girls' schools, on the other hand, were the least favorably positioned on these variables. A similar pattern emerged when we focused attention on fiscal resources. Boys' schools had the highest per-pupil expenditure, but the girls' schools had the lowest. Unless a schooleffects model with significant interaction terms involving school by sex type is entertained, this area does not appear to be fertile ground for locating alternative explanations. One of the major commonalities in boys' and girls' schools is that the gender composition of the faculty reflects the nature of the student body. This parallelism provides students with a large array of adult role models of the same sex, which might foster more serious student attitudes and behaviors toward schooling. If this were a major source of the observed effects, however, it would imply that these effects are intrinsic to the single-sex form of school organization and might not be easily transferred. The data in Table 8 also indicate that single-sex schools were much more likely to be sponsored by religious orders. For girls' schools, it was impossible to separate a singlesex-organization effect from a religious-order-governance effect because these two variables were almost totally confounded. The data on boys' schools, however, were sufficient to sustain further analysis. In particular, we recomputed the effects of boys' schools, reported in Tables 4 through 6, under an analytic model that added another variable to distinguish religious-order boys' schools from those boys' schools governed by other organizations, such as parishes and dioceses. The boys' school effects remained virtually unchanged with this added control factor. Although we cannot totally discount the possibility of a governance effect, we have no evidence that favors this alternative explanation.

The picture for girls, however, was quite different. There were virtually no differences in the educational aspirations at 8th grade between girls in the two types of schools. Yet we found statistically significant differences favoring girls' schools on educational aspirations at 10th and 12th grades and on the sophomore-to-senior gain in aspirations. Further, there are also structural factors that impose some limits on the possible magnitude of such a selection effect. Not all students and families who select Catholic secondary schools have a choice. In many communities, there is only a single Catholic high school, or the available schools are all of only one typeeither single-sex or coeducational. Unless single-sex schools are more likely to be located in communities with more highly motivated students, differential choice would not provide an alternative explanation, at least in these cases. In sum, although it is always possible that students and parents choose single-sex schools for other reasons that can also explain the observed results, the preponderance of the evidence that we assembled tends to support a school effect rather than a selection-hypothesis explanation. The results presented in Tables 4 through 6 indicate significant differences across a range of outcome measures between singlesex and coeducational schools. The estimated effects were stronger and more pervasive for girls than for boys. Further, as already noted, the selection hypothesis is a less tenable alternative for explaining the girls' schools effects. In our view, the observational evidence that we assembled provides strong support for concluding that there are positive effects associated with attendance at girls' schools. The picture is more ambiguous with regard to the effects of boys' schools. In a cautious interpretation of the latter there would be no indication of any negative effects on the outcomes that we considered. In fact, there may be some positive effects associated with attendance at boys' schools, although it is not possible to definitively resolve the issue on the basis of the existing data.

Possible School Factors Contributing to the SingleSex Effects


The key remaining question is "What are the critical attributes of single-sex Catholic schools that contribute to the observed effects?" Are the factors intrinsic in this form of school organization associated with a greater academic press, as hypothesized by Coleman (1961)? Or is it some other set of considerations, such as school resources and academic policies, that just happen to be associated with this particular set of Catholic schools? Although further research is needed to fully answer this question, a descriptive comparison of single-sex and coeducational Catholic schools can at least provide some clues. In the analyses reported here, we took into account the social composition of the school, so this is unlikely to be a critical factor in the observed single-sex-school effects. One obvious possibility is some set of differences between coeducational and single-sex schools in various aspects of school structure and resources. The data presented in Table

EFFECTS OF SINGLE-SEX SECONDARY SCHOOLS Table 8

393

Comparison of Single-Sex and Coeducational Schools (Weighted Means)


Variable School-descriptive characteristics School size Student/faculty ratio Diversity of curricular offerings'* *b Staff resources % female faculty % faculty with advanced degrees % annual faculty turnover % faculty at school 10 years or more School fiscal resources First salary step, BA Annual tuition Per-pupil expenditure School quality, average of student reports Perceived quality of teaching"-11 General school rating ac School governance* % religious order
a

Boys1 schools
814

Coeducational schools
616 19.7

Girls1 schools 491 16.0 -.297 87.5% 39.6% 15.1% 17.2% $9,007 1,057 1,325 .532 .143 86.8%-

20.1 -.279 15.4% 60.3% 8.7% 31.6% $8,927


990

.101 57.2%' 43.3% 13.4% 17.2%


$9,329

1,517 .731 .288 59.2%

750 1,321 -.135 -.116 8.2%

These variables were standardized (M - 0, SD = 1.0) on the basis of the entire Catholic-school sample. Because a portion of that sample was excluded from this investigation, the statistics reported above deviate slightly from those parameter values. bThis factor was constructed by (a) creating two factors from the courses students had taken in high school, centered on the English/social studies and math/science areas; (b) aggregating the two factors to the school level; (c) creating a measure of school curriculum diversity by subtracting the school's score from the student's score for each of the two factors, squaring the differences, and summing these squares over the two factors, and (d) aggregating this sum of squared difference scores to schooMevel means to create the measure of schoolcurriculum diversity. cPer-pupil expenditure was estimated by dividing the tuition rate by the percentage of a school's funds reported by the principal as coming from tuition. dThis variable is an aggregate of the following two student-level factors: (a) students' perceptions of the quality of instructions and (b) students' reports about the proportion of their instructors having certain desirable characteristics (e.g., are patient and present materials clearly). Each factor was based on several items. This is an aggregate of a student-level factor created from students' ratings of various aspects of their schools (e.g., condition of buildings, library facilities, fairness and effectiveness of discipline, and faculty interest in students). fThe wording of this question in the base-year questionnaire for principals allowed no differentiation between Catholic-school governance types except for religious order or nonreligious order. Therefore, the variable is dichotomous.

Some support for the hypothesis that it is not intrinsic features of single-sex schools that produce the observed effects can be found in student reports about the quality of teaching and students' overall ratings of the school. Both measures were based on factor score composites of several rating scale items each. Students in both boys' and girls1 schools rated both their schools and the quality of teaching much more positively than did their counterparts in coeducational schools. How to interpret these data, however, remains an open question. One explanation is that this set of schools, labeled single-sex, just happens to have unusually effective staff. But even here the interpretation can easily become muddied. It is quite plausible that singlesex-school organization could produce not only a positive environment for learning but also a positive environment for teaching. In fact, it seems likely that these two factors combine in a mutually reinforcing relation. In that case, the positive reports about teaching and school quality may be just another set of outcomes of single-sex-school organi-

zation rather than a competing explanation for the apparent effects of that organization. Finally, there is a wide array of school policies, for example, on homework, absenteeism, and discipline, that could have also contributed to the observed effects. In our view, these school policies along with structural features, such as size, and the intrinsic characteristics of single-sex schools, such as their climate and faculty composition, merit closer scrutiny in future research. Discussion At a minimum, the student reports on teaching quality and rating of their schools, in conjunction with the other results presented in this article, indicate that something significant is occurring in Catholic single-sex schools. A cursory examination of our findings indicates few negative effects of single-sex schools on the diverse array of dependent variables that we considered. In fact, all of the statis-

394

VALERIE E. LEE AND ANTHONY S. BRYK

tically significant single-sex-school effects were positive. Whether considering academic achievement in specific areas at sophomore or senior year, gains in academic achievement over the 2 years, future educational plans, affective measures of locus of control or self-image, sex role stereotyping, or attitudes and behaviors related to academics, we found that single-sex schools appear to deliver specific advantages to their students. The results are particularly strong for girls' schools, where students were generally more interested in academics and showed significantly greater gains in reading, science, and educational ambition over the course of their high school years. These girls' schools also showed some positive effects on students' locus of control, and these girls were less likely to see themselves in sex-stereotyped adult roles. As has been suggested about postsecondary single-sex education, single-sex secondary schooling may in fact serve to sensitize young women to their occupational and societal potentials in an atmosphere free of some of the social pressures that female adolescents experience in the presence of the opposite sex. Adolescence is a critical period for the formation of attitudes about oneself. It may be that some separation of students' academic and social environments removes the distractions that can interfere with the academic development of some students. Although the aims of schooling are varied, academic pursuits rightly belong at the top of the list. If the positive benefits of social contact cited by Dale (1969, 1971, 1974) and others as an advantage of coeducational schooling act to undermine the development of some students, particularly female students, then considerably more thought on the issue of coeducational versus single-sex schooling seems warranted. Further, our research raises questions about the appropriateness of "students' satisfaction with the school environment" as a primary standpoint for evaluating schools. The latter has been a focus in previous research on singlesex schooling and a central element in the argument for adoption of coeducational schools (i.e., students prefer to attend coeducational schools). Although student satisfaction is an important consideration, when uncoupled from concerns about academic attitudes, school behavior, and achievement, it offers only a partial and somewhat distorted view of the school. In particular, if improving the social environmental press is accompanied by a general decline in academic behavior and performance, then a failure to recognize these unintended negative consequences would be very troublesome. Further, our results raise questions about the presumed negative aspects of single-sex schooling. HS&B data indicated that the stronger academic attitudes, academically related behaviors, and higher achievement in Catholic single-sex schools are in fact accompanied by very positive student reports about their schools. At a minimum, the results of this study ought to encourage a more careful scrutiny of life within single-sex secondary schools. This inquiry should explore the actual practices within these institutions in an attempt to elucidate the full range of critical factors contributing to the observed effects. The study should begin at the point where students normally enter these schools (ninth grade) rather than half-

way through the high school years, as was the case in HS&B. A combination of field research and quantitative observational methods is likely to be required. In an era in which single-sex secondary schools are often looked at as anachronistic and in which these same schools are often merged with opposite-sex schools to create coeducational institutions thought to be more viable economically and socially, it is striking that there has been so little empirical investigation of this form of school organization. At least the results of this study ought to encourage educators to ask, "Is secondary education in America enriched or improverished by the gradual disappearance of the single-sex school?" To be sure, the relevant policy consideration is not whether all secondary schools should be single sex. Rather, if subsequent research supports the positive findings of our investigation and also concludes that the critical factors for this success are intrinsic to the singlesex organizational form, then the practical issue is to find ways to preserve existing single-sex schools and to encourage their development in contexts where the option does not currently exist. References
Alexander, K. L., & Pallas, A. M. (1983). Private schools and public policy: New evidence on cognitive achievement in public and private schools. Sociology of Education, 170-182. Anderson, S., Auquier, A., Hauch, W. W., Oakes, D., Vandaele, W., & Weisberg, H. I. (1980). Statistical methods for comparative studies: Techniques for bias reduction. New York: Wiley. Astin, A. W. (1977a). Four critical years; Effects of college on beliefs, attitudes, and knowledge. San Francisco: Jossey-Bass. Astin, A. W. (1977b). On the failure of educational policy. Change, 9, 40-45. Block, J. H. (1984). Gender differences and implications for educational policy. In Sex role identity and ego development, (pp. 207-252). San Francisco: Jossey-Bass. Bryk, A. S. (1981). Disciplined inquiry or policy argument? Harvard Educational Review, 51, 497-509, Bryk, A. S., Holland, P. B., Lee, V. E , & Carriedo, R. A. (1984). Effective Catholic schools: An exploration. Washington, DC: National Catholic Education Association. Coleman, J. S. (1961). The adolescent society. New York: Free Press of Glencoe. Coleman, J. S., Hoffer, T., & Kilgore, S. (1982). High school achievement: Public, Catholic, and private schools compared. New York: Basic Books. Dale, R. R. (1969). Mixed or single-sex school? Volume I: A research study about pupil-teacher relationships. London: Routledge & Kegan Paul. Dale, R. R. (1971). Mixed or single-sex school? Volume II: Some social aspects. London: Routledge & Kegan Paul. Dale, R. R. (1974). Mixed or single-sex school? Volume III: Attainment, attitudes and overview. London: Routledge & Kegan Paul. Dale, R. R., & Miller, P. M. (1972). The academic progress of university students from co-educational and single-sex schools. British Journal of Educational Psychology, 42, 317-319. Feather, N. T. (1974). Coeducation, values, and satisfaction with school. Journal of Educational Psychology. 66, 915.

EFFECTS OF SINGLE-SEX SECONDARY SCHOOLS Finn, J. D. (1980). Sex differences in educational outcomes: A cross-national study. Sex Roles, 6, 9-25. Furniss, W. T., & Graham, P. A. (Eds.)- (1974). Women in higher education. Washington, DC: American Council on Education. Goldberger, A. S., & Cain, G. G. (1982). The causal analysis of cognitive outcomes in the Colcman, Hoffer and Kilgore report. Sociology of Education, 55, 103-122. Goodlad, J. (1984). A place called school. New York: McGrawHill. Graham, P. A. (1970). Women in academe. Science, 69, 12841290. Graham, P. A. (1974). Women in higher education: A biographical inquiry. New York: Columbia University, Barnard College. (ERIC Document Reproduction Service No. ED 095 742). Hoffer, T., Greeley, A., & Coleman, J. (1985). Achievement growth in public schools. Sociology of Education, 58, 1491. Horner, M. S. (1972). Toward an understanding of achievementrelated conflicts in women. Journal of Social Issues, 28, 157175. Hyde, S. (1971). The case of coeducation. Independent School Bulletin, 31, 20-24. Jencks, C. (1985). How much do high school students learn? Sociology of Education, 58, 128135. Jennrich, R., & Sampson, P. (1983). P3V: General mixed model analysis of variance. In W. J. Dixon (Ed.), BMDP statistical software (pp. 413-426). Berkeley: University of California Press. Jones, J. C , Shallcross, J., & Dennis, C. L. (1972). Coeducation and adolescent values. Journal of Educational Psychology, 63, 334-341. Light, R. J., & Pillemer, D. B. (1984). Summing up. Cambridge, MA: Harvard University Press. Lockheed, M. E. (1976). The modification of female leadership behavior in the presence of males. Princeton, NJ: Educational Testing Service.

395

Murnane, R. J. (1981). Evidence, analysis, and unanswered questions. Harvard Educational Review, 51, 483-^-89. National Center for Education Statistics (1983). High School and Beyond 1980 sophomore cohort: First follow-up (1982). Data file user's manual. Washington, DC: Author. Oates, M. J., & Williamson, S. (1978). Women's colleges and women achievers. Journal of Women in Culture and Society, 3, 795-806. Riordan, C. (1985). Public and Catholic schooling: The effects of gender context policy. American Journal of Education, 5, 518 540. Schneider, F. W., & Coutts, L. M. (1982). The high school environment: A comparison of coeducational and single-sex schools. Journal of Educational Psychology, 74, 898-906. Tidball, M. E. (1980). Women's colleges and women achievers revisited. Signs: Journal of Women in Culture and Society, 5, 504-515. Tidball, M.D., & Kistiakowsky, V. (1976). Baccalaureate origins of American scientists and scholars. Science, 193, 646-652. Trickett, E. J., Castro, J. J., Trickett, P. K., & Shaffner, P. (1982). The independent school experience: Aspects of the normative environments of single-sex and coed secondary schools. Journal of Educational Psychology, 74, 374381. Willms, J. D. (1985). Catholic-school effects on academic achievement: New evidence from the High School and Beyond follow-up study. Sociology of Education, 55(2), 98-114. Winchel, R., Fenner, D., & Shaver, P. (1974). Impact of coeducation on "fear of success'1 imagery expressed by male and female high school students. Journal of Educational Psychology, 66, 726-730.

Received January 30, 19B5 Revision received April 29, 1986

Integrating Personality and Social Psychology: Call for Papers


The editors of the Journal of Personality and Social Psychology intend to publish a special issue devoted to papers demonstrating that social behavior is best understood by integrating the diverse concerns of the three sections of JPSP. Papers are invited that deal with particular substantive issues crossing the boundaries of individual differences, social cognition, and interpersonal relations. Papers should represent practical demonstrations that the diverse concerns of this journal belong together in a full understanding of social behavior. We seek previously unpublished contributions, primarily empirical studies, but we are also amenable to syntheses of long-term research programs and to innovative theoretical slatements. Contributions intended for the special issue should be sent to the guest editor: John F. Kihlstrom, PhD W. J. Brogden Psychology Building University of Wisconsin 1202 West Johnson Street Madison, Wisconsin 53706 Interest authors should send an abstract of their article to the guest editor by September 1, 1986 and plan to submit a completed manuscript by December 1, 1986, at which time the paper will become subject to the usual peer review process. Journal of Personality and Social Psychology editors: Attitudes and Social Interpersonal Relations Cognition and Group Processes Steven J. Sherman Harry T. Reis Charles M. Judd Norbert L. Kerr Personality Processes and Individual Differences Irwin G. Sarason Edward F. Diener Warren H. Jones

Das könnte Ihnen auch gefallen