A Phase II exclusion based on obvious methodological considerations (e
A primary criticism of previous single-sex literature has been the confounding of single-sex effects with the effects of religious values, financial privilege, selective admissions, or other advantages associated with the single-sex school being studied
g., nonstudy, weak study). On the basis of titles and abstracts, citations that appeared to be essays, reviews, opinion pieces, and similar items were excluded, and only qualitative and quantitative studies that were likely to be codable in Phase III were retained. During Phase II, 114 citations were culled from the 379 items and coded as appropriate for review as quantitative (88) or qualitative (26) studies. Of the 26 qualitative studies, 4 met the criteria for final inclusion and were reviewed separately.
A Phase III evaluation and coding of the remaining quantitative articles. According to the guidelines of the WWC, all studies other than randomized controlled trials, quasi-experimental designs (QED) with matching, or regression discontinuity designs would be excluded prior to Phase III. Under the WWC criteria for inclusion, virtually all single-sex studies would have been eliminated from the review process because of the lack of experimental research on this topic. Therefore, for this review, a conscious decision was made to relax these standards and include all correlational studies that employed statistical controls. By relaxing the WWC standards, the number of candidate studies to be screened in Phase III was greatly increased. A more streamlined and efficient checklist was developed requiring dichotomous responses rather than descriptive responses in order to facilitate rater decision making. To be included in the quantitative review, a study had to use appropriate measurement and statistical principles. Therefore, in particular a study had to include statistical controls to account for individual differences (e.g., socioeconomic status [SES], individual ability, and age) as well as school and class differences that might account for the differences between single-sex and coeducational schools. Even so, many studies that included at least one covariate lacked other important covariates such as ethnic or racial minority status, socioeconomic status, and grade level or age. Also, it is important to note that the inclusion of covariates cannot control for important unobservable differences between the groups, such as motivation. Because correlational studies cannot adequately address the issue of differences in unobservables (or selection bias), the studies in this review may over or understate the true effects of SS schooling.
The Quantitative Review
Two reviewers coded each study independently, using a quantitative coding guide. A quantitative study was coded for its treatment of the following broad issues: sample characteristics, psychometric properties, internal validity, effect, and bias. Each of these categories had several criteria by which they were coded. To be retained, a study did not have to meet all criteria.
Of the 88 quantitative studies, 48 were eliminated after further review using the coding guide, and 40 studies met the inclusion criteria and were retained. The reasons for the exclusion of these articles were 1) failure to operationalize the intervention properly; 2) failure to apply statistical controls during the analyses; 3) work that was actually qualitative in nature rather than quantitative; 4) work performed in a non-Westernized country and therefore not comparable; 5) work written in a foreign language and therefore not codable by the researchers; 6) failure to draw comparisons between SS and CE schools; and 7) participants not of high school, middle, or elementary school age. In all, 40 studies met the inclusion criteria and were retained in the quantitative review. The following table shows results of each study according to the seven broad questions listed above and is broken into specific criteria within each larger category. Because some studies addressed multiple criteria, the total number of findings is greater than 40. Specifically, there are 112 findings considered singleparentmeet in the 40 quantitative studies.