Academic Attitudes and Achievement in Students of Urban Public Single-Sex and Mixed-Sex High Schools

Abstract

Publicly funded single-sex schooling (SSS) has proliferated in recent years and is touted as a remedy to gaps in academic attitudes and achievement, particularly for low-income students of color. Research on SSS is rife with limitations, stemming from selective admissions processes, selection effects related to socioeconomic status, a lack of ethnic diversity among students, and a neglect of boys’ schools. Addressing those concerns, the current study is a quasi-experimental investigation of the academic attitudes and achievement among 11th-grade low-income students of color enrolled in nonselective, urban neighborhood public single-sex and mixed-sex high schools. Students in SSS reported significantly more negative attitudes about English/reading compared to students in mixed-sex schools (MSS), while there were no differences in math or science attitudes. Data from standardized tests indicate that SSS was associated with poorer achievement among boys in reading and math but higher achievement among girls on math, science, reading, and writing.

Keywords

academic attitudes achievement coeducational human sex differences race single-sex

Introduction

Public single-sex schooling (SSS) has proliferated in recent years, following regulations in response to the No Child Left Behind Act to ease Title IX restrictions on publicly funded SSS (Brown, 2013). This rapid expansion— which has been accompanied by a contentious debate among politicians, advocacy groups, and educational researchers—has been touted by some (e.g., Meyer, 2008; Mitchell & Stewart, 2013; Riordan, 1990) as an approach that will improve the academic achievement of low-income students of color (Fergus, Sciurba, Martin, & Noguera, 2009; Goodkind, Schelbe, Joseph, Beers, & Pinsky, 2013; McCreary, 2011). In an effort to contribute data to this debate, the current study is a quasi-experimental examination of the reading, math, and science attitudes and achievement of low-income students of color enrolled in single-sex and mixed-sex urban public high schools.

The Rationale for Single-Sex Schooling

Advocates of SSS generally have used two lines of reasoning in their support of SSS. One is that boys and girls have meaningfully different learning styles—which are rooted in structural and physiological differences in the brains of boys and girls—that require different learning environments (e.g., Gurian, Henley, & Trueman, 2001; James, 2009; Sax, 2006). Yet, Eliot (2011) reviewed neuroscience research and concluded there are “few reliable differences between boys’ and girls’ brains relevant to learning or education” (p. 363). The other line of reasoning is that sexism in coeducational classrooms—for example, in the form of sexual harassment or distraction by peers—prevents students from realizing their academic potential (e.g., Sadker & Sadker, 1994). Similarly, advocates maintain that SSS reduces sexual distraction that arises from being near or interacting with members of the other gender. In an op-ed in the Wall Street Journal, Senators Kay Bailey Hutchison (R-TX) and Barbara Mikulski (D-MD) (2012) claimed, “As studies have confirmed—and as any parent can tell you—negative gender roles are often sharpened in coeducational environments. Boys are more likely, for instance, to buy into the notion that reading isn’t masculine when they’re surrounded by (and showing off for) girls.” Advocates contend that teaching boys and girls in a single-sex environment improves students’ self-concept of ability in school subjects that are usually considered appropriate for the other gender (Crombie & Armstrong, 1999; Hoffman, Badgett, & Parker, 2008; Kessels & Hannover, 2008; Lee & Marks, 1990; Norfleet & Richards, 2003; Stables, 1990). In other words, educating boys and girls in separate spaces may permit them to engage more freely in subjects that are otherwise loaded with gender stereotypes.

The proliferation of SSS within urban school districts in particular has been part of the larger movement to improve academic performance among low-income students and students of color, especially African American boys (Fergus et al., 2009; Goodkind et al., 2013; McCreary, 2011; Mitchell & Stewart, 2013). Its political appeal draws on the school choice movement, in that making public SSS available to low-income students of color suggests some degree of equity with middle- and upper-income White families, who have greater access to private SSS (McCreary, 2011; Meyer, 2008). Because explicitly race-based education policies are politically and legally unacceptable, sex-based policies in the form of SSS—which were previously viewed as noncompliant with Title IX but are now permitted via No Child Left Behind—have multiplied with the goal of improving the outcomes of African American boys as well as girls (McCreary, 2011). Some advocates, such as Riordan (1994), have claimed that SSS is particularly effective with students of color, as “research has demonstrated that the effects of single-gender schools are greatest among students who have been disadvantaged historically—females and racial/ethnic/religious minorities (both males and females)” (p. 48).

Yet, the evidence supporting this claim is fraught with limitations. While some (e.g., Meyer, 2008; Riordan, 1990) maintain that SSS improves the academic performance of low-income students and students of color, others argue that the evidence supporting such claims remains insufficient (Fergus et al., 2009). In the absence of a strong theoretical foundation or consistent empirical evidence, the reasoning for SSS among students of color is suspect. Some scholars have argued that fundamental to the rationale for SSS for African American and Latino boys are assumptions about the characteristics and needs of such youth (Fergus et al., 2009; Goodkind, 2013). Those assumptions are based on problematic deficit notions that African American and Latino boys lack self-esteem, parental involvement, and positive male role models and have negative attitudes about education (Fergus et al., 2009). In a similar vein, the logic that SSS reduces sexual distraction is problematic within the context of expanding SSS for students of color, insofar as it reinforces and perpetuates sexist and racist stereotypes about the hypersexuality of African American youth (Goodkind et al., 2013). In addition, this reasoning reinforces heterosexist norms and traditional constructions of gender by assuming that all students are heterosexual and disregarding queer, intersex, and transgender students (Jackson, 2010).

Academic Achievement and Attitudes in SSS

Whether single-sex education has any impact on academic achievement remains hotly contested. Comprehensive reviews of the literature can be found elsewhere (e.g., Bracey, 2006; Haag, 1998; Mael, Alonso, Gibson, Rogers, & Smith, 2005; Pahlke, Hyde, & Allison, 2014). Mael et al. (2005) reported that the evidence is split between positive findings for SSS and no differences or null findings, with only a few studies supporting mixed-sex schooling (MSS). A meta-analytic investigation of Mael et al.’s (20050 review of SSS effects on academic achievement indicated nonsignificant links to math achievement (d = .03) and significant but small links (d = .18) to verbal performance (Signorella, Hayes, & Li, 2013). Critically, these effects were moderated by preexisting differences in socioeconomic status (SES) or cognitive performance, such that students from SSS tended to come from higher-SES homes and have higher cognitive abilities. A recent meta-analysis including published and unpublished data through 2013 described a literature on SSS that is varied in terms of research design quality (Pahlke et al., 2014). Analyzing studies of U.S. samples that used either random assignment or controls for selection effects, Pahlke et al. (2014) found negligible mean effects of SSS on achievement in math (d = .09 for girls, d = .02 for boys), science (d = .04 for girls, d = .00 for boys), and reading/language arts (d = .06 for girls, d = .01 for boys). Critically, there were too few studies of students of color to adequately address the research question posed in the current study. Cross-nationally, nations with a greater proportion of single-sex math classrooms show more variability in gender differences in math achievement, while those with a greater proportion of MSS show a pattern of gender similarities in achievement and gender equity in opportunity to learn (Wiseman, 2008). In sum, the evidence indicates that SSS per se does not appear to have positive effects on academic achievement, though this conclusion is accompanied by a caveat regarding research design quality.

Our rationale for assessing academic attitudes in the context of SSS stems from several theoretical perspectives positing that academic attitudes play an important role in the decision to pursue related education and careers. For example, self-efficacy theory (Bandura, 1977) proposes that individuals choose to undertake activities or tasks in which they believe they will be successful. Social-cognitive career theory (SCCT; Lent, Brown, & Hackett, 2000) maintains that our choice about whether to pursue a particular career depends on our self-efficacy and the perceived outcomes of pursuing that career. Eccles’s (1994) expectancy-value theory argues that achievement behaviors are the product of an individual’s expectations for success and the perceived value or importance of the behavior. Moreover, each of these perspectives maintains that social factors (e.g., gender role ideologies, stereotypes, and gender segregation in careers) contribute to these attitudes. Thus, it is critical to identify the contexts in which positive academic attitudes are fostered. According to SSS advocates, single-sex classrooms and schools are potentially such contexts.

Some studies have reported that single-sex classrooms and schools are linked to less gender-stereotyped academic attitudes (e.g., Kessels & Hannover, 2008; Stables, 1990). For example, in a report of SSS and MSS students’ self-concept of academic abilities, Sullivan (2009) found that although boys had more positive math and science attitudes and girls had more positive attitudes about English, SSS was linked to smaller gender gaps in these attitudes. Using retrospective reports of female graduates of private single-sex and mixed-sex high schools, Sax, Arms, Woodruff, Riggers, and Eagan (2009) reported that female graduates of SSS showed higher levels of academic engagement, interest in engineering careers, and confidence in mathematical ability and computational skills. However, the differences that favored single-sex alumnae did not depend directly on the single-sex nature of the high school but rather on the characteristics of the schools (e.g., size, selectivity, and geographic region) and students (e.g., socioeconomic class and parental education).

Advocates of SSS point to these results as evidence of beneficial outcomes from SSS, but such findings also highlight the importance of distinguishing effects of SSS from other school, student, family, and community characteristics. Other studies indicate that SSS is not consistently associated with more positive academic attitudes. For example, using longitudinal data from an intervention of SSS math and science instruction in public MSS, Shapka and Keating (2003) found no positive effects for math anxiety, math competence, or effort expended on math, despite positive effects on math and science achievement and physics self-concept (Karpiak, Buchanan, Hosey, & Smith, 2007). Thus, at first glance, the literature appears to suggest that SSS is linked to more positive academic attitudes; yet, effects from school, student, family, and community characteristics are entrenched in those findings and point to a number of serious methodological limitations. Moreover, recent meta-analytic findings indicate that when only studies of U.S. samples using either random assignment or controls for selection effects are included, the effects of SSS on math attitudes (d = .07 for girls, insufficient data for boys) and self-concept (d = −.03 for girls, d = −.02 for boys) are negligible (Pahlke et al., 2014). In contrast, effects on school attitudes were found to be small and negative (d = −.23 for boys, insufficient data for girls), such that boys in MSS had more positive school attitudes than boys in SSS. Yet, on some attitudinal measures, findings varied considerably between these controlled studies and studies that did not account for selection effects or use random assignment, highlighting the methodological limitations of existing SSS literature.

Methodological Limitations in SSS Literature

The literature on SSS and academic achievement is rife with limitations (Halpern et al., 2011; Pahlke et al., 2014). Most critically, many positive findings regarding the effects of SSS stem, at least in part, from nonrandom assignment (Halpern et al., 2011). Most single-sex schools have selective admissions processes in which only high-achieving students are admitted (Hayes, Pahlke, & Bigler, 2011). Even if students are admitted by lottery (the nearest process that U.S. public schools have to random assignment), some schools may still implement a selectivity bias by expelling students who do not meet certain standards of academic achievement and behavior. A notable exception to the pattern of these selection effects is the case of Korea, in which children were randomly assigned to SSS and MSS. Analyses of Korean student performance on the 2003 and 2007 Trends in International Mathematics and Science Study (TIMSS) indicate no significant differences between SSS and MSS in math and science achievement (Pahlke, Hyde, & Mertz, 2013). In one study of a single-sex middle school, when selection effects and peer quality were taken into account, school gender composition was not significantly linked to student achievement (Hayes et al., 2011). In Pahlke et al.’s (2014) meta-analysis of SSS effects, selection effects were so pervasive in the literature that studies were grouped and analyzed according to design features that addressed the problem. Thus, selection effects due to student characteristics (e.g., preexisting differences in ability or aptitude) are a legitimate concern to address.

Given that SSS is most often privately funded, and thus financially unfeasible for many families, findings from studies of SSS are also frequently limited by selection effects related to socioeconomic factors (Hayes et al., 2011). Whereas upper-middle-class parents may be able to choose whether to send their child to a mixed-sex or single-sex school, poor and working-class parents, who have fewer financial resources, typically do not have this option. Indeed, the meta-analytic critique by Signorella et al. (2013) found that SSS effects are rooted in preexisting group differences in SES and cognitive performance. The selection effects related to socioeconomic factors are particularly problematic when attempting to generalize SSS findings to low-income students and students of color.

Similarly, while advocates and policymakers argue that SSS might improve the educational experience of low-income students and students of color (e.g., Riordan, 1990), the effects of gender-segregated education for youth attending urban public high schools is rarely studied (Pahlke et al., 2014), and so it is unclear if SSS can benefit all students. A two-year ethnographic study of low-income students and students of color who attended SSS in California maintained that simply separating students by gender was insufficient to improve the educational experiences and achievement of low-income students and students of color (Hubbard & Datnow, 2005). Indeed, researchers have speculated that the higher achievement and more positive attitudes among SSS students result from school characteristics unrelated to single-sex composition, such as a religious affiliation, private school status, or additional funding (Hubbard & Datnow, 2005; Jimenez & Lockheed, 1989).

Another major limitation in the SSS literature is that boys’ schools are underrepresented (Mael et al., 2005). For example, of the 76 studies on SSS reviewed by the U.S. Department of Education (Mael et al., 2005), only 3 were devoted exclusively to boys’ schools; 52 studies included both boys and girls. Despite the lack of evidence, advocates maintain that single-sex education can boost grades and test scores for both genders (e.g., Sax, 2006). For example, the National Association for Single Sex Public Education (2010; since renamed National Association for Choice in Education) argues that because single-sex classrooms can break down gender stereotypes, boys in SSS are much more likely to pursue interests in art, music, drama, and foreign languages. Given that much of the rationale for SSS among low-income students and students of color emphasizes the need to target African American and Latino boys, this is a significant limitation to the literature.

The Current Study

These methodological challenges—including concerns about selective admissions processes, selection effects related to socioeconomic status, lack of ethnic diversity, and overreliance on girls’ schools—make it difficult to draw reliable conclusions from existing studies on the effects of SSS on the academic attitudes and achievement among low-income youth and youth of color in urban neighborhood public schools. Therefore, the current study seeks to address several of these methodological issues to provide new knowledge that can inform educational policy.

The current study is a quasi-experimental comparison of the academic attitudes and achievement of students enrolled in nonselective, urban public single-sex and mixed-sex high schools. We sampled from two public SSS that serve as neighborhood (i.e., default, non-magnet, nonselective) schools for children in a specific geographic region of a large urban school district and from MSS in comparable neighborhoods within the same district. Similar to the majority of students in that school district, our sample is composed of predominantly low-income and African American students; at the time of sample recruitment, 75% of students in the district qualified for free or reduced-price lunch, and 65% of students in the district were African American. We include students from both all-male and all-female SSS. Our aim was to assess differences in attitudes toward and achievement in reading, math, and science and test the hypothesis that SSS is linked to more positive attitudes and higher achievement. In addition, we examined whether differences between school types were similar for male and female students.

Method

Participants

The sample is composed of n = 46 (23 girls, 23 boys) 11th-grade students of single-sex schools and n = 42 (21 girls, 21 boys) students of mixed-sex high schools. Regarding race/ethnicity, 93.1% reported being Black or African American, 5.7% reported being multiracial/multiethnic, and 1.1% did not indicate their race/ethnicity. Students were, on average, 17.12 years old (SD = .59). Descriptive statistics for all demographic variables, by school type, appear in Table 1.

Table 1

Demographic Variables in Students From Single-Sex and Mixed-Sex Schools

	Single-Sex Schools	Mixed-Sex Schools
Number of books in home
0–10	5 (10.8)	9 (22.0)
10–50	17 (37.0)	13 (31.7)
50–100	13 (28.3)	9 (22.0)
Over 100	11 (23.9)	10 (24.4)
Highest level of education completed by mother
Less than high school	8 (18.2)	11 (26.8)
High school graduate or GED	19 (43.2)	20 (48.8)
Trade school or some college	7 (15.9)	7 (17.1)
College graduate	10 (22.7)	2 (4.9)
Graduate school	0 (0)	1 (2.4)
Reason for attending particular high school
It’s in my neighborhood	33 (71.7)	29 (70.7)
My parents wanted me to go there	7 (15.2)	3 (7.3)
I wanted to go there	5 (10.9)	8 (19.5)
Other	1 (2.2)	1 (2.4)

Note. Values are frequencies, percentages appear in parentheses. No significant differences between single-sex school and mixed-sex school students were found in any of these demographic variables.

Materials and Procedures

We recruited 11th-grade students from two single-sex high schools and eight mixed-sex high schools, all within a school district from a large city in the northeastern United States. None of the schools could be characterized as a magnet school with selective admissions; all schools were default schools for youth living within specific geographical zones within the city (i.e., “neighborhood schools”). Three mixed-sex schools were selected on the basis of comparability to the two single-sex schools. This included comparability of socioeconomic status (i.e., ≥90% of students eligible for free/reduced price meals), geography (i.e., in proximate neighborhoods of the city), ethnicity (i.e., ≥97% described as African American by school district), school enrollments, and academic needs (i.e., ≥81% without individualized education plans). Seven students from five additional mixed-sex schools were included in our sample because they transferred out of our targeted schools during the academic year but remained on the official roster; those five additional schools were comparable on the aforementioned criteria to the other three mixed-sex schools and the two single-sex schools. The two single-sex schools were in the same neighborhood and had the same feeder elementary schools.

Recruitment packets, which were addressed to parents/guardians of 11th-grade students and included parental consent forms and invitation letters, were distributed by school officials to students in homeroom classes or were mailed directly to the students’ homes, depending on the preference of the school principals. We estimate that n = 150 families received the recruitment packets; of those, n = 96 (64%) consented to participate. Our estimated response rate is based on school enrollment at the two single-sex schools (each had approximately n = 45 11th graders at the start of the school year) as well as our distribution of 100 recruitment packets to 11th graders officially enrolled in the three mixed-sex schools selected (as noted previously, 7 students had transferred to five other schools within the district by the time of data collection). Twenty-eight recruitment letters were returned by the post office as being undeliverable or having no forwarding address.

Parents/guardians provided informed consent and permission for their child’s participation and the release of standardized test scores and grades; these forms were mailed by parents/guardians to the study office. Upon receipt of parental consent and permission, surveys and student assent forms were mailed to students’ homes, and students returned completed materials by mail. Of those who consented, 100% completed their surveys. Students received $50 for completing their surveys, which included demographic items and measures of academic attitudes, in addition to several other outcomes not reported here. Data were collected during the winter and spring of the students’ 11th-grade year.

Measures

Demographic variables

At the end of the survey packet, participants were asked to report date of birth, gender, ethnicity, name of high school attended, long-term educational goals, as well as the number of books in the home and mother’s education. In addition, because transfers to other neighborhood schools are sometimes permitted by the district, we asked students for the reason they were attending their specific high school. Response options included “it’s in my neighborhood,” “my parents wanted me to go there,” “I wanted to go there,” and “other” with an open response. In subsequent analyses, we collapsed the latter three response levels into one level, which is conceptualized as attending the specific school by choice.

English/reading attitudes

Three scales developed by Eccles and colleagues (e.g., Fredricks & Eccles, 2002) comprise the measures of English and reading attitudes. Scales included reading self-concept (α = .85; e.g., “How good at reading/English are you?”), task value in reading/English (α = .77; e.g., “How interesting is reading/English to you?”), and expectations of success in reading/English (α = .87; e.g., “How successful do you think you’d be in a career that required reading/English ability?”). Participants responded to all items on a scale from 1 (e.g., not very interesting) to 4 (e.g., very interesting). Scale scores are the average of two to three items each; higher scores indicate more positive attitudes.

Math attitudes

Similarly, scales developed by Eccles and colleagues (e.g., Fredricks & Eccles, 2002) comprise three of our seven measures of math attitudes. Scales included math self-concept (α = .80; e.g., “How good at math are you?”), task value in math (α = .51; e.g., “How interesting is math to you?”), and expectations of success in math (α = .63; e.g., “How successful do you think you’d be in a career that required math ability?”). Participants responded to these items on a scale from 1 (e.g., not very interesting) to 4 (e.g., very interesting). Scale scores are the average of two to three items each; higher scores indicate more positive attitudes.

In addition, items from PISA (Programme for International Student Assessment) 2003 student surveys comprised four scales of attitudes about math (OECD, 2004a, 2004b). Scales were based on the mean of three to five items each, which were rated from 1 (strongly disagree) to 4 (strongly agree). Scales include interest and enjoyment in math (α = .84; e.g., “I do math because I enjoy it”), instrumental motivation to learn math (α = .75; e.g., “I will learn many things in math that will help me get a job”), math anxiety (α = .78; e.g., “I get very nervous doing math problems”), and self-concept of math ability (α = .78; e.g., “I learn math very quickly”).

Science attitudes

Likewise, scales developed by Eccles and colleagues (e.g., Fredricks & Eccles, 2002) comprised three of our eight measures of science attitudes. Scales included science self-concept (α = .85; e.g., “How good at science are you?”), task value in science (α = .79; e.g., “How interesting is science to you?”), and expectations of success in science (α = .83; e.g., “How successful do you think you’d be in a career that required science ability?”). Participants responded to these items on a scale from 1 (e.g., not very interesting) to 4 (e.g., very interesting). Scale scores are the average of two to three items each; higher scores indicate more positive values.

Items from PISA 2006 student surveys comprised five scales of attitudes about science (OECD, 2007). Scales include general value of science (α = .77; e.g., “Science is valuable to society”), personal value of science (α = .65; e.g., “Science is very relevant to me”), interest and enjoyment in science (α = .92; e.g., “I generally have fun when I am learning science”), instrumental motivation in science (α = .89; e.g., “I study science because I know it is useful for me”), and self-concept of science ability (α = .90; e.g., “I can easily understand new ideas in science”). Scales were based on the mean of three to five items each, which were rated from 1 (strongly disagree) to 4 (strongly agree).

Academic achievement

The math, science, reading, and writing tests of the Pennsylvania System of School Assessment (PSSA) were used as indicators of academic achievement. The PSSA is an annual statewide standards-based, criterion-referenced assessment used to measure student attainment of academic standards and also is among the components used in the determination of a school’s Adequate Yearly Progress (AYP), as mandated by the No Child Left Behind Act of 2001. It was administered to our sample in the spring of the 11th-grade year; it is not administered at any other time in high school. The math (α = .94), science (α = .92), reading (α = .90), and writing (α = .82) tests of the PSSA used in the current study demonstrated good to excellent internal consistency statistics (Pennsylvania Department of Education, 2012). Additional technical details about the PSSA, including interrater agreement and scaling procedures, are available from the Pennsylvania Department of Education (2015). Students’ scaled scores and proficiency level (below basic, basic, proficient, or advanced) on the four components of the PSSA (math, science, reading, and writing) were obtained from the school district; 70.5% of the students in our sample completed at least one of the PSSA tests.

Data Analyses

First, we assessed the potential for selection effects. We compared students from SSS and MSS on socioeconomic variables (i.e., number of books in the home and maternal level of education) to determine if students were comparable in socioeconomic status. We also compared SSS and MSS students on their reasons for attending their specific high school; a chi-square analysis compared students who attended SSS and MSS as a neighborhood school or by choice. The possibility of selection effects was further probed by comparing students by reason for attending their specific high school within each school type on socioeconomic variables and academic attitudes and achievement. We ran a series of 2 (school type: mixed-sex vs. single-sex) × 2 (reason for attending specific school: neighborhood vs. choice) chi-square analyses and between-subjects analyses of variance (ANOVAs) using socioeconomic variables and academic attitudes and achievement as outcome variables.

Next, a series of 2 (school type: mixed-sex vs. single-sex) × 2 (student gender) between-subjects ANOVAs assessed group differences on English/reading, math, and science attitudes and math, science, reading, and writing PSSA scaled scores. Chi-square analyses were conducted with the math, science, reading, and writing PSSA proficiency levels to assess the proportion of students who achieved proficiency in a subject within each school type. We tested the hypothesis that SSS would be associated with more positive academic attitudes and higher achievement. Further, we examined the Gender × School Type interaction to assess if effects of school type were similar for male and female students.

This data analytic strategy was chosen to maximize statistical power in light of our relatively small sample for a 2 × 2 between-subjects design. That is, the sample size is adequate to detect large differences at α = .05. Thus, we also examined effect sizes (Cohen’s d) to inform our conclusions. Consistent with Cohen’s (1988) guidelines, we characterize effect sizes of 0.20, 0.50, and 0.80 as small, medium, and large, respectively. Hyde (2005) has characterized effect sizes of 0.10 or less as near zero or trivial and of 1.00 as very large; we follow these interpretations as well.

Results

Preliminary Analyses on SSS and MSS Selection Effects

SSS and MSS students did not differ significantly in their reasons for attending their specific high school, χ²(1) = 0.01, n = 87, p = .92. Regarding group differences in socioeconomic variables, students from single-sex and mixed-sex high schools did not differ significantly on number of books in the home, χ²(3) = 2.17, n = 87, p = .54, or mother’s highest level of education, χ²(4) = 6.74, n = 85, p = .15. In addition, students who attended a specific school by choice did not differ from students attending their neighborhood school on number of books in the home, χ²(3) = 4.13, n = 87, p = .25, or mother’s highest level of education, χ²(4) = 1.01, n = 85, p = .91. And, within school type, reasons for attending a specific school were not significantly associated with number of books in the home, MSS: χ²(3) = 7.01, n = 41, p = .07; SSS: χ²(3) = 1.18, n = 46, p = .76, or mother’s highest level of education, MSS: χ²(4) = 2.59, n = 41, p = .63; SSS: χ²(4) = 0.70, n = 44, p = .87.

The interaction effects in the 2 (school type) × 2 (reason for attending specific school: neighborhood vs. choice) ANOVAs on academic attitudes and achievement probe for selection effects by indicating whether students who attended SSS or MSS because it was their neighborhood school differ from their classmates who attended that school by choice. In three of these ANOVAs, the School Type × Reason interaction was significantly linked to study variables, though the nature of the interaction varied. Students who attended MSS by choice reported significantly greater value of reading compared to other students, F(1, 83) = 6.07, p = .02, η²_p = .07. In contrast, students who attended SSS because it was in their neighborhood reported significantly greater general value of science, F(1, 83) = 7.66, p = .007, η²_p = .08, and personal value of science, F(1, 83) = 12.12, p = .001, η²_p = .13, compared to students who attended SSS by choice. In the remaining analyses—namely, those in which the interaction was not significant—effect sizes were very small (η²_p < .02). In sum, there is limited evidence of clear, systematic differences based on MSS and SSS students’ reasons for attending their specific high school that might generate selection effects in our analyses.

Academic Attitudes

The results of the ANOVAs testing how school type and gender are linked to academic attitudes, including descriptive statistics, appear in Table 2. Across the measures of math, science, and reading attitudes, school type was significantly linked to only two outcomes. Students from MSS reported significantly more positive reading self-concept (d = −.45) and reading expectations of success (d = −.51), contrary to the hypothesis that SSS would be associated with more positive academic attitudes. Of the other outcomes, one showed a small advantage for MSS students (reading task value, d = −.36), and nine showed negligible or very small differences (i.e., d < .11; math self-concept [Eccles], math task value, math expectations, math self-concept [PISA], interest and enjoyment in math, general value of science, personal value of science, enjoyment of science, science expectations). Six outcomes showed small but nonsignificant advantages for SSS (ds = .21-.36; instrumental motivation in math, math anxiety, science self-concept [Eccles], science task value, science self-concept [PISA], and instrumental motivation in science). Yet, none of the tests of the interaction between school type and gender was significant, which indicates that differences by school type are not moderated by gender. In sum, results indicate that SSS is linked to significant negative, small to medium effects in reading attitudes but nonsignificant positive, negligible to small effects on science and math attitudes, regardless of student gender.

Table 2

Group Means and Standard Deviations in Math, Science, and Reading Attitudes, by Gender and School Type, Effect Sizes, and F Statistics for Main Effects and Interaction of Gender and School Type

	Girls SSS M, SD	MSS M, SD	Boys SSS M, SD	MSS M, SD	SSS-MSS Girls d	SSS-MSS Boys d	School Type F	Gender F	School Type × Gender F
Math self-concept (Eccles)	2.78	2.76	2.96	3.07	0.03	−0.19	0.11	3.06	0.23
Math self-concept (Eccles)	0.76	0.62	0.51	0.64
Math task value	3.33	3.11	3.41	3.55	0.39	−0.36	0.13	5.91*	3.03
Math task value	0.56	0.57	0.39	0.41
Math expectations	2.83	2.76	2.91	3.13	0.09	−0.36	0.27	2.54	0.96
Math expectations	0.81	0.62	0.47	0.69
Interest and enjoyment in math	2.58	2.55	2.68	2.74	0.04	−0.07	0.01	0.94	0.07
Interest and enjoyment in math	0.67	0.71	0.67	0.83
Instrumental motivation in math	3.35	3.23	3.29	3.19	0.23	0.19	0.93	0.17	0.01
Instrumental motivation in math	0.53	0.55	0.35	0.70
Math anxiety	2.17	2.40	2.04	2.10	−0.39	−0.09	1.17	2.60	0.42
Math anxiety	0.62	0.55	0.62	0.66
Math self-concept (PISA)	2.79	2.62	2.76	2.95	0.21	−0.31	0.00	0.98	1.38
Math self-concept (PISA)	0.85	0.71	0.66	0.52
Science self-concept (Eccles)	2.91	2.63	3.13	2.92	0.37	0.34	2.79	2.87	0.05
	0.75	0.74	0.49	0.74
Science task value	2.81	2.92	3.25	2.78	−0.13	0.65	1.14	0.81	2.98
Science task value	0.89	0.82	0.60	0.75
Science expectations	2.83	2.67	3.00	3.00	0.18	0.00	0.19	1.95	0.19
Science expectations	0.94	0.87	0.72	0.84
General value of science	3.21	3.24	3.33	3.34	−0.05	−0.03	0.11	3.06	0.23
General value of science	0.48	0.56	0.38	0.43
Personal value of science	2.64	2.90	2.94	2.78	−0.43	0.33	0.19	0.57	3.23
Personal value of science	0.71	0.44	0.48	0.47
Enjoyment of science	2.77	2.88	3.02	2.85	−0.14	0.23	0.03	0.46	0.72
	0.92	0.65	0.70	0.75
Instrumental motivation in science	2.86	2.95	3.07	2.70	−0.13	0.62	1.00	0.02	2.73
	0.78	0.67	0.52	0.64
Science self-concept (PISA)	2.76	2.70	2.97	2.55	0.08	0.52	2.16	0.03	1.23
	0.76	0.69	0.68	0.88
Reading self-concept	3.12	3.43	3.17	3.42	−0.45	−0.45	4.47*	0.03	0.07
	0.78	0.55	0.61	0.43
Reading task value	3.33	3.63	3.59	3.70	−0.43	−0.26	2.84	1.83	0.56
	0.71	0.61	0.40	0.46
Reading expectations	3.09	3.64	3.41	3.60	−0.73	−0.27	6.05*	0.88	1.49
	0.79	0.62	0.79	0.55

Note. For all tests, df = 1, 83. SSS = single-sex school; MSS = mixed-sex school; PISA = Programme for International Student Assessment.

p < .05.

Regarding the main effect of gender, only math task value showed a significant gender difference, which was medium (d = .49) and favored boys. Four nonsignificant gender comparisons were negligible (i.e., d < .09), and the remainder were small and favored boys (ds = .16–.37). In other words, boys tended to report somewhat more positive academic attitudes, regardless of school type.

Academic Achievement

The results of the ANOVAs testing group differences by school type and gender in academic achievement, as measured by PSSA scaled scores, appear in Table 3. The main effect of school type was significant in science and reading but not in math or writing. In addition, the main effect of gender was significant in math and reading but not in science or writing. However, these main effects of school type and gender are qualified by significant School Type × Gender interactions in math, reading, and writing. In general, SSS was linked to negative effects for boys and positive effects for girls. That is, while girls in SSS outperformed girls in MSS in math, science, reading, and writing, boys in SSS scored lower than boys in MSS on the PSSA tests of math and writing. As shown in Table 3, these effects ranged in magnitude from negligible to medium among boys and large among girls.

Table 3

Group Means and Standard Deviations in Math, Science, Reading, and Writing Tests of the PSSA, by Gender and School Type, Effect Sizes, and F Statistics for Main Effects and Interaction of Gender and School Type

	Girls SSS M, SD	MSS M, SD	Boys SSS M, SD	MSS M, SD	SSS-MSS Girls d	SSS-MSS Boys d	School Type F	Gender F	School Type × Gender F	df
Math	1,250.57	1,082.28	1,030.00	1,106.87	0.87	−0.42	1.00	4.61*	7.21**	1, 58
	188.49	167.15	116.27	228.54
Science	1,142.21	1,100.77	1,131.67	1,098.00	0.70	0.67	5.79*	0.18	0.06	1, 44
	61.28	51.24	47.26	49.53
Reading	1,331.14	1,050.00	1,044.08	1,050.38	1.27	−0.03	6.47*	7.04**	7.08**	1, 57
	141.29	194.26	143.35	300.09
Writing	1,402.54	1,115.31	1,075.67	1,205.08	1.10	−0.57	1.72	3.89	12.00***	1, 50
	129.54	270.73	181.17	250.45

Note. SSS = single-sex school; MSS = mixed-sex school; PSSA = Pennsylvania System of School Assessment.

p < .05. **p < .01. ***p < .001.

We further examined academic achievement among SSS and MSS students by conducting a series of chi-square analyses on math, science, reading, and writing proficiency levels. Table 4 shows the percentage of male and female students within SSS and MSS schools whose achievement was categorized as below or basic (i.e., not proficient) and proficient or advanced (i.e., at or above proficient) on each of the PSSA tests. On the math test, the chi-square test was significant for girls, χ²(1) = 4.23, n = 32, p = .04, but not for boys, χ²(1) = 3.33, n = 30, p = .07, indicating that SSS was linked to greater percentage of girls, but not boys, achieving proficiency in math on the PSSA. In science, no students achieved proficiency; thus, a chi-square statistic could not be computed. On the reading test, the chi-square test was significant for girls, χ²(1) = 12.22, n = 32, p < .001, but not for boys, χ²(1) = 0.74, n = 29, p = .39, indicating that SSS was linked to greater percentage of girls, but not boys, being proficient in reading on the PSSA. On the writing test, the chi-square test was significant for girls, χ²(1) = 12.46, n = 29, p < .001, but not for boys, χ²(1) = 2.49, n = 27, p = .11, indicating that SSS was linked to greater percentage of girls, but not boys, being proficient in writing on the PSSA. In sum, the effects of school type on academic achievement were moderated by gender such that SSS effects were significant among girls but not among boys.

Table 4

Quantity of Female and Male Students Within SSS and MSS at Proficiency Levels of PSSA Tests

		Girls SSS	MSS	Boys SSS	MSS
Math	Below/basic	8 (57.1)	16 (88.9)	15 (100.0)	12 (80.0)
	Proficient/advanced	6 (42.9)	2 (11.1)	0 (0.0)	3 (20.0)
Science	Below/basic	14 (100.0)	13 (100.0)	9 (100.0)	12 (100.0)
	Proficient/advanced	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)
Reading	Below/basic	4 (28.6)	16 (88.9)	12 (92.3)	13 (81.3)
	Proficient/advanced	10 (71.4)	2 (11.1)	1 (7.7)	3 (18.8)
Writing	Below/basic	2 (15.4)	13 (81.3)	13 (92.9)	9 (69.2)
	Proficient/advanced	11 (84.6)	3 (18.8)	1 (7.1)	4 (30.8)

Note. Values are frequencies; percentages appear in parentheses and are within school type. SSS = single-sex school; MSS = mixed-sex school; PSSA = Pennsylvania System of School Assessment.

Discussion

As SSS continues to be touted as a remedy for the underachievement of low-income youth of color, especially boys, the need for empirical evidence of such educational reforms within urban public schools is greater than ever. To that end, we compared 11th-grade boys and girls from nonselective urban public SSS and MSS. There was some evidence that SSS was associated with higher standardized test scores among girls, yet the findings for boys are a stark contrast in that SSS was linked to poorer achievement. That is, the results from this quasi-experimental study indicate that while SSS was linked to higher achievement on the math, reading, and writing tests for girls, SSS was linked to poorer achievement in math and writing for boys. Similarly, compared with girls in MSS, girls in SSS were significantly more likely to achieve proficiency on all of the PSSA tests; however, school gender segregation was not linked to boys’ proficiency levels on any of the PSSA tests. In sum, although the expansion of urban public SSS has been promoted by some (e.g., Meyer, 2008; Mitchell & Stewart, 2013; Riordan, 1990) as a method to improve the academic performance of African American boys in particular (as described by Fergus et al., 2009; Goodkind et al., 2013; McCreary, 2011), our findings question the veracity of this claim.

Evidence that gender segregation was linked to academic attitudes for boys or girls was inconsistent; moreover, the group differences tended to be nonsignificant and negligible or very small. In the few significant main effects of school type on attitudes, SSS was linked to more negative reading attitudes but slightly more positive science and math attitudes; these effects were not significantly different between boys and girls. A notable exception to the pattern of similarities in attitudes between SSS and MSS is that students from SSS scored about one-half standard deviation lower than students from MSS in English/reading self-concept and expectations of success in English/reading. Nonetheless, given higher academic achievement for girls in SSS, the lack of clear, consistent links to academic attitudes points to the discrepancy between academic attitudes and achievement.

Thus, our findings with academic attitudes contrast with some previous research on gender segregated schooling (e.g., Kessels & Hannover, 2008; Sax et al., 2009; Stables, 1990; Sullivan, 2009) but are consistent with others (e.g., Karpiak et al., 2007; Shapka & Keating, 2003). In light of the centrality of academic attitudes to educational goals and choices (e.g., Bandura, 1977; Eccles, 1994; Lent et al., 2000), our data suggest that SSS is unlikely to foster student interest or engagement any more effectively than MSS.

Implications for Research and Policy

Our findings provide no evidence that SSS frees students to have more positive or less gender stereotyped academic attitudes, in contrast to the claims made by SSS advocates. Developmental intergroup theory (Bigler & Liben, 2007) posits that when an environment increases the saliency of a social group membership such as gender, children will engage in more categorization related to that group and increase stereotypes and prejudices. Experimental research shows that making gender salient in the classroom—for example, by segregating boys and girls—increases gender stereotyping among preschoolers (Hilliard & Liben, 2010). In single-sex high schools, students’ use of heterosexism and homophobia to reinforce traditional gender roles is pervasive (Woody, 2003), which suggests that SSS do not provide the liberating environment necessary for girls and boys to pursue gender-atypical interests. Similarly, opponents have maintained that SSS is likely to exacerbate—not reduce—sexism and gender stereotyping (Halpern et al., 2011). While our findings do not indicate greater gender stereotyped academic attitudes in SSS, neither do they suggest that SSS provides a liberating atmosphere free of gender roles or expectations. Future research on SSS should examine the extent to which students endorse gender stereotypes and self-segregate by gender outside of school to explore other effects of gender-segregated schooling on adolescent development.

Several of the findings reported here—namely, that boys and girls both tend to fare better when in the company of girls—are consistent with research demonstrating gender-moderated effects of gender-segregated play on behavioral outcomes. For example, Fabes and colleagues found that the proportion of time that boys spent playing with other boys predicted an increase in their subsequent behavior problems, while the proportion of time that girls spent playing with other girls predicted a decrease in their behavior problems (Fabes, Shepard, Guthrie, & Martin, 1997). In other words, both boys’ and girls’ behavioral outcomes were better when they spent more time with girls.

In a similar vein, boys and girls in our study achieved the best academic outcomes when they had female classmates. This suggests that SSS might be linked to better academic achievement for girls. However, if gender segregation in schooling is causally linked to academic outcomes—a claim we cannot test with a quasi-experiment—benefits of SSS for girls’ academic achievement could come at the expense of boys’ academic achievement. If this pattern of findings is replicated in other communities and populations—a necessary condition to justify changes in educational policy—policymakers are faced with a dilemma: to provide MSS to maximize boys’ achievements at the expense of girls’ or to provide SSS to maximize girls’ achievements at the expense of boys’. This is a particularly salient dilemma in light of efforts to improve the representation of women in STEM fields (e.g., Halpern et al., 2011). That is, despite gender similarities in STEM abilities (Hyde, Lindberg, Linn, Ellis, & Williams, 2008; OECD, 2009), girls’ more negative attitudes about some STEM disciplines remain an obstacle to fuller participation (Ceci & Williams, 2007).

Related to this point, Baker (2002) observed and analyzed the single-sex math and science classrooms in a middle school serving predominantly Latino/a and African American youth. She reported that all-girl classrooms fostered students’ feelings of empowerment, peer support, and positive self-concept. However, she also reported that all-boy classrooms were less supportive learning environments, characterized by hostile interactions and discipline problems, despite greater engagement in technology-based activities. Notably, Baker found that teachers did not use different teaching styles or techniques, nor did they alter the course content or activities, depending on the gender of the students. Baker concluded that the pedagogy and curriculum used were generally a better fit for girls, who did well with low structure and group work requiring cooperation. In contrast, boys needed classrooms with tasks that reflected their interests better, with more structure and technology use. Thus, the all-girl classrooms were perceived as effective because the curriculum and pedagogy matched the needs of the girls and because boys were not there to disrupt the class or discourage the girls.

One implication of Baker’s (2002) findings, then, is that boys and girls should be taught in segregated classrooms, but with different pedagogy and curricula that map onto apparent psychological gender differences—in other words, classrooms should be separate and unequal. Such a conclusion is ethically problematic and unacceptable as educational policy and may ultimately do more to perpetuate academic, social, and emotional gender differences than mitigate them. Still another implication of the findings of Baker and the current study is that the effects of gender-segregated classrooms and schools should not be conflated with the effects of gender-differentiated pedagogy and curriculum.

A limitation of our study is that we did not observe classrooms to determine whether single-sex and mixed-sex classrooms showed systematic pedagogical differences or similarities, nor did we obtain such data from teachers or administrators. The question of gender-differentiated pedagogy in single-sex classrooms and schools is an important direction for future research. Nonetheless, pedagogy that is engaging, dynamic, and supportive, as well as sensitive to the developmental needs of students, and curricula that tap into students’ interests and goals need not be linked to gender, nor should pedagogy or curricula negate the goal of gender equity in education.

Strengths, Limitations, and Future Directions

Concerns about limitations in previous research on SSS, including selective admissions processes, selection effects related to socioeconomic status, and lack of ethnic diversity, were addressed in our quasi-experimental study. We were able to address some selection effects due to student characteristics and socioeconomic factors by sampling from single- and mixed-sex neighborhood high schools with comparable SES and nonselective admissions. The students enrolled in those schools and living in those neighborhoods were largely African American and from low SES homes; there is a paucity of empirical SSS research in these populations (e.g., Pahlke et al., 2014). Moreover, our analyses of selection effects suggest that when students attended SSS or MSS for reasons other than the school being in their neighborhood, they were not significantly different from their classmates on study variables. In these respects, our study design offers significant advantages in that the findings may be generalizable to similar schools and communities.

Nonetheless, although selective admissions processes were not used to determine school enrollments, students were not randomly assigned to the schools; thus, we cannot make causal inferences, and selection effects may remain. In such research designs, pretest data—for example, in the form of standardized tests from grades prior to enrollment at the single-sex school— can be analyzed to address concerns about preexisting group differences. Future quasi-experimental studies on SSS should include such data. In addition, although students’ reported reasons for attending MSS or SSS did not suggest that families were disproportionately enrolling in schools outside of their neighborhood schools’ geographical boundary, we cannot know for certain that families did not choose their school in some other way (e.g., by moving to a specific neighborhood). In any quasi-experiment, selection effects are present to some extent; future research on SSS must attend to this limitation as much as possible. For example, a deeper examination of parents’ and students’ reasons for choosing SSS should be included. Students who opted out of SSS and students who wanted to attend SSS but were denied admission also should be sampled. In addition, given that we sampled from only two SSS, it is possible that the group differences found are specific to those particular schools, teachers, or students. Thus, future studies should sample from more than two single-sex schools, which would allow for the assessment of variability in the implementation of SSS across sites.

Adequate measurement of academic achievement and attitudes are crucial in determining the effects of SSS. We used standardized measures of academic achievement by obtaining students’ scores on state assessments of math, science, reading, and writing. Our measures of academic attitudes have been widely used with diverse samples and provide a multidimensional assessment of students’ attitudes about math, science, and English/reading. Still, several of the math and science attitudes scales had unfavorable internal consistencies; that is, for the scales of task value in math, expectations of success in math, and personal value of science, α ≤ .65. Therefore, the findings with these scales should be interpreted cautiously. Moreover, several of the attitudinal scales used were based on two or three items; this limitation is addressed in part by the inclusion of additional attitudinal scales. For example, in an effort to bolster the construct validity within our study, we employed two measures of self-concept of math ability (Eccles and PISA). These measures differ in subtle ways; the Eccles scale has three items, including one item that compares the participant to his or her classmates on math ability, whereas the PISA scale has five items, none of which ask the participant to compare himself or herself to classmates. Academic attitudes are complex and context specific; a student may find biology interesting but be bored with physics, while another student may be fascinated by geometry because she has a dynamic and enthusiastic teacher. This complexity in attitudes may contribute to both the inadequate internal consistency of some scales as well as a general need to measure attitudes with multiple scales tapping into multiple dimensions, specific to the academic subject.

An additional limitation in our research design regards the duration of time students spent in SSS and the developmental stage at which they attend SSS. Both of the single-sex high schools covered Grades 9 through 12, which suggests that students in our study had been enrolled in gender-segregated schools for up to three years. It is possible that any effects of SSS are moderated by dosage, such that students who have spent more time in SSS will show larger losses or gains than those who have spent relatively less time in SSS. In a similar vein, the effects of single-sex schools may differ from those of single-sex classrooms within mixed-sex schools. Meta-analysis of such effects have been mixed but suggest that, if anything, SSS is more potent when delivered at the classroom level than at the school level (Pahlke et al., 2014).

SSS effects may also be moderated by developmental timing. In other words, students may be more or less responsive to a gender-segregated environment at different points in their development. Although children tend to self-segregate by gender during the elementary school years, they tend to reconnect in adolescence, particularly as heterosexual romantic relationships become more interesting to many of them (Maccoby, 1990). Given this developmental pattern of preferences and social interactions, SSS likely has different effects—in terms of magnitude, direction, and nature— depending on when in development SSS is implemented. Meta-analytic results indicate that the effects of SSS on math performance are significantly larger in middle school compared to elementary or high school and that the effects also differ in direction, such that SSS confers an advantage for middle school girls but a disadvantage for middle school boys (Pahlke et al., 2014). We cannot examine these research questions with our data, but they are important elements to address in designing future research on SSS.

Conclusions

As public SSS has proliferated in the United States, so has research on its efficacy. While the research has yet to demonstrate clear and consistent benefits, neither has it demonstrated clear and consistent harm. Our study indicates that some students may be helped by gender-segregated programs, some may show detriments, and some may not be affected at all. In the schools sampled here, girls in SSS outperformed girls in MSS, but boys in MSS fared better than boys in SSS. Within the context of urban schools serving low-income students of color, it is important to acknowledge that myriad inequities contribute to poor student outcomes. Lack of access to books and libraries; poor infrastructure and public safety concerns; inadequate nutrition, health care, and housing; as well as persistent and systemic racial/ethnic discrimination and stereotyping are powerful obstacles to student achievement and are unlikely to be remedied by gender segregation in the schools. In the absence of a sound theoretical rationale supporting SSS and the presence of theoretical rationale supporting MSS, gender-segregated schools and classrooms need to be transparent and open to research that evaluates their efficacy. Moreover, researchers without a vested interest in either form of schooling ought to be leading the evaluations of such schools to avoid conflicts of interest. Certainly, the need for replication of the current study across multiple communities is critical and speaks to the risky practice of implementing, without sufficient empirical justification, public SSS in urban communities serving marginalized populations.

Confronting a common fate among contemporary urban public schools, both of the SSS and two of the MSS schools in our study closed in the year following data collection for our project. School districts in Detroit, Chicago, Kansas City, Milwaukee, Pittsburgh, Philadelphia, and the District of Columbia have been dealing with fiscal constraints and declining enrollments by closing schools within the most impoverished neighborhoods. In this way, the schools we sampled are much like many urban public schools serving low-SES families and students of color across the United States, and our findings are informative to policymakers and researchers alike. In some cases, gender-segregated programs may be a last-ditch effort to “save” an underperforming school, or they may be initiated by charter schools replacing traditional public schools. If reforms or innovations such as SSS are insufficiently evaluated, they have the potential to further marginalize and shortchange these youth. Students of color from low-SES homes are already disadvantaged in urban public schools; evidence reported here indicates that enrolling these youth in SSS is not likely to correct that pattern. Moreover, curriculum and pedagogy that are appropriately engaging, challenging, and supportive to boys and girls alike will do more to promote equality and equity than will gender segregated classrooms or schools.

Footnotes

Notes

Nicole M. Else-Quest is associate professor and associate chair of the Department of Psychology and affiliate associate professor in the Department of Gender and Women’s Studies at the University of Maryland, Baltimore County, 1000 Hilltop Circle, Baltimore, MD 21250; e-mail: nmeq@umbc.edu . Her research uses an intersectional approach to examine gender development across the life span, with emphasis in the content areas of emotion, STEM attitudes, and financial well-being.

Oana Peterca earned her master of science in psychology at Villanova University and currently works at Weill Cornell Medical College. Her research interests are in gender issues, education, and underrepresented groups.

References

Baker

(2002). Good intentions: An experiment in middle school single-sex science and mathematics classrooms with high minority enrollment. Journal of Women and Minorities in Science and Engineering, 8, 1–23.

Bandura

(1977). Self-efficacy: Toward a unifying theory of behavioral change. Psychological Review, 84, 191–215.

Bigler

R. S.

Liben

L. S.

(2007). Developmental intergroup theory: Explaining and reducing children’s social stereotypes and prejudice. Current Directions in Psychological Science, 16, 162–171.

Bracey

G. W.

(2006). Separate but superior? A review of issues and data bearing on single-sex education. Tempe, AZ: Arizona State University Educational Policy Research Unit (EPRU).

Brown

C. S.

(2013). Legal issues surrounding single-sex schools in the U.S.: Trends, court cases, and conflicting laws. Sex Roles, 69, 356–362.

Ceci

S. J.

Williams

W. M.

(2007). Why aren’t more women in science? Top researchers debate the evidence. Washington, DC: American Psychological Association.

Cohen

(1988). Statistical power analysis for the behavioral sciences. Hillsdale, NJ: Erlbaum.

Crombie

Armstrong

P. I.

(1999). Effects of classroom gender composition on adolescents’ computer-related attitudes and future intentions. Journal of Educational Computing Research, 20, 217–327.

Eccles

J. S.

(1994). Understanding women’s educational and occupational choices: Applying the Eccles et al. model of achievement-related choices. Psychology of Women Quarterly, 18, 585–610.

10.

Eliot

(2011). Single-sex education and the brain. Sex Roles, 69, 363–381.

11.

Fabes

R. A.

Shepard

S. A.

Guthrie

I. K.

Martin

C. L.

(1997). Roles of temperamental arousal and gender-segregated play in young children’s social adjustment. Developmental Psychology, 33, 693–702.

12.

Fergus

Sciurba

Martin

Noguera

(2009). Single-sex schools for Black and Latino boys: An intervention in search of theory. In Motion Magazine. Retrieved from http://inmotionmagazine.com/er/fsmn_09.html

13.

Fredricks

J. A.

Eccles

J. S.

(2002). Children’s competence and value beliefs from childhood through adolescence: Growth trajectories in two male-sex-typed domains. Developmental Psychology, 38, 519–533.

14.

Goodkind

(2013). Single-sex public education for low-income youth of color: A critical theoretical review. Sex Roles, 69, 393–402.

15.

Goodkind

Schelbe

Joseph

A. A.

Beers

D. E.

Pinsky

S. L.

(2013). Providing new opportunities or reinforcing old stereotypes? Perceptions and experiences of single-sex public education. Children and Youth Services Review, 35, 1174–1181.

16.

Gurian

Henley

Trueman

(2001). Boys and girls learn differently! A guide for teachers and parents. New York, NY: Jossey-Bass.

17.

Haag

(1998). Single-sex education in grades K–12: What does the research tell us? In Morse

(Ed.), Separated by sex: A critical look at single-sex education for girls (pp. 13–38). Washington, DC: AAUW Educational Foundation.

18.

Halpern

D. F.

Eliot

Bigler

R. S.

Fabes

R. A.

Hanish

L. D.

Hyde

J. S.

. . . Martin

C. L.

(2011). The pseudoscience of single-sex schooling. Science, 333, 1706–1707.

19.

Hayes

A. R.

Pahlke

E. E.

Bigler

R. S.

(2011). The efficacy of single-sex education: Testing for selection and peer quality effects. Sex Roles, 65, 693–703.

20.

Hilliard

L. J.

Liben

L. S.

(2010). Differing levels of gender salience in preschool classrooms: Effects on children’s gender attitudes and intergroup bias. Child Development, 81, 1787–1798.

21.

Hoffman

B. H.

Badgett

B. A.

Parker

R. P.

(2008). The effect of single-sex instruction in a large, urban, at-risk high school. The Journal of Educational Research, 102, 15–35.

22.

Hubbard

Datnow

(2005). Do single-sex schools improve the education of low-income and minority students? An investigation of California’s public single-gender academies. Anthropology and Education Quarterly, 36, 115–131.

23.

Hutchison

K. B.

Mikulski

(2012, October 16). A right to choose single-sex education. Wall Street Journal. Retrieved from http://www.wsj.com/articles/SB10000872396390443768804578038191947302764

24.

Hyde

J. S.

(2005). The gender similarities hypothesis. American Psychologist, 60, 581–592.

25.

Hyde

J. S.

Lindberg

S. M.

Linn

M. C.

Ellis

Williams

(2008). Gender similarities characterize math performance. Science, 321, 494–495.

26.

Jackson

(2010). “Dangerous presumptions”: How single-sex schooling reifies false notions of sex, gender, and sexuality. Gender and Education, 22, 227–238.

27.

James

A. N.

(2009). Teaching the female brain: How girls learn math and science. Thousand Oaks, CA: Corwin Press.

28.

Jimenez

Lockheed

M. E.

(1989). Enhancing girls’ learning trough single-sex education: Evidence and a policy conundrum. Education Evaluation and Policy Analysis, 11, 117–142.

29.

Karpiak

C. P.

Buchanan

J. P.

Hosey

Smith

(2007). University students from single-sex and coeducational high schools: Differences in majors and attitudes at a Catholic university. Psychology of Women Quarterly, 31, 282–289.

30.

Kessels

Hannover

(2008). When being a girl matters less: Accessibility of gender-related self-knowledge in single-sex and coeducational classes and its impact on students’ physics-related self-concept of ability. British Journal of Educational Psychology, 78, 273–289.

31.

Lee

V. E.

Marks

H. M.

(1990). Sustained effects of the single-sex secondary-school experience on attitudes, behaviors, and values in college. Journal of Educational Psychology, 82, 578–592.

32.

Lent

R. W.

Brown

S. D.

Hackett

(2000). Contextual supports and barriers to career choice: A social cognitive analysis. Journal of Counseling Psychology, 47, 36–49.

33.

Maccoby

E. E.

(1990). Gender and relationships: A developmental account. American Psychologist, 45, 513–520.

34.

Mael

Alonso

Gibson

Rogers

Smith

(2005). Single-sex versus coeducational schooling: A systematic review. Washington, DC: US Department of Education.

35.

McCreary

A. J.

(2011). Public single-sex K–12 education: The renewal of sex-based policy by post-race politics, 1986–2006. Journal of Law and Education, 40, 461–497.

36.

Meyer

(2008). Learning separately: The case for single-sex schools. Education Next, 8, 10–21.

37.

Mitchell

A. B.

Stewart

J. B.

(2013). The efficacy of all-male academies: Insights from critical race theory (CRT). Sex Roles, 69, 382–392.

38.

National Association for Single Sex Education. (2010). In the news. Retrieved from http://www.singlesexschools.org/links-articles.htm.

39.

Norfleet

J. A.

Richards

H. C.

(2003). Escaping stereotyping: Educational attitudes of male alumni of single-sex and coed schools. Psychology of Men and Masculinity, 4, 136–148.

40.

OECD. (2004a). Learning for tomorrow’s world: First results from PISA 2003. Paris: Author.

41.

OECD. (2004b). PISA 2003 Technical Report. Paris: Author.

42.

OECD. (2007). PISA 2006: Science competencies for tomorrow’s world. Volume 1: Analysis. Paris: Author.

43.

OECD. (2009). Equally prepared for life? How 15-year-old boys and girls perform in school. Paris: Author.

44.

Pahlke

Hyde

J. S.

Allison

C. M.

(2014). The effects of single-sex compared with coeducational schooling on students’ performance and attitudes: A meta-analysis. Psychological Bulletin, 140, 1042–1072.

45.

Pahlke

Hyde

J. S.

Mertz

J. E.

(2013). The effects of single-sex compared with coeducational schooling on mathematics and science achievement: Data from Korea. Journal of Educational Psychology, 105, 444–452.

46.

Pennsylvania Department of Education. (2012). Technical report for the 2011 Pennsylvania System of School Assessment. Retrieved from http://www.portal.state.pa.us/portal/server.pt/community/technical_analysis/7447

47.

Pennsylvania Department of Education. (2015). Home page. Retrieved from (http://www.portal.state.pa.us/portal/server.pt/community/technical_analysis/7447)

48.

Riordan

(1990). Girls and boys in school: Together or separate? New York, NY: Teachers College Press.

49.

Riordan

(1994, February 23). Reconsidering single-gender schools: The VMI case and beyond. Education Week, 13, 48. Retrieved from http://www.edweek.org/ew/articles/1994/02/23/22rior.h13.html

50.

Sadker

(1994). Failing at fairness: How our schools cheat girls. New York, NY: Simon and Schuster.

51.

Sax

(2006). Six degrees of separation: What teachers need to know about the emerging science of sex differences. Educational Horizons, 84, 190–212.

52.

Sax

Arms

Woodruff

Riggers

Eagan

(2009). Women graduates of single-sex and co-educational high schools: Differences in their characteristics and the transition to college. Los Angeles, CA: Sudikoff Family Institute for Education and New Media.

53.

Shapka

J. D.

Keating

D. P.

(2003). Effects of a girls-only curriculum during adolescence: Performance, persistence, and engagement in mathematics and science. American Educational Research Journal, 40, 929–960.

54.

Signorella

M. L.

Hayes

A. R.

(2013). A meta-analytic critique of Mael et al.’s (2005) review of single-sex schooling. Sex Roles, 69, 423–441.

55.

Stables

(1990). Differences between pupils from mixed and single-sex schools in their enjoyment of school subjects and in their attitudes to science and to school. Educational Review, 42, 221–230.

56.

Sullivan

(2009). Academic self-concept, gender and single-sex schooling. British Educational Research Journal, 35, 259–288.

57.

Wiseman

A. W.

(2008). A culture of (in) equality?: A cross-national study of gender parity and gender segregation in national school systems. Research in Comparative and International Education, 3, 179–201.

58.

Woody

E. L.

(2003). Homophobia and heterosexism in public school reform: Constructions of gender and sexuality in California’s single gender academies. Equity & Excellence in Education, 36, 148–160.