* noted as presenting author
Jill Chonody, PhD, Lecturer, University of South Australia, Adelaide, Australia, Carl F. Siebert, MBA, MS, Statistical Consultant and Research Instructor, Rutgers University, New Brunswick, NJ, Darcy Clay Siebert, PhD, Associate Professor, Rutgers University, New Brunswick, NJ and Scott Edward Rutledge, PhD, Assistant Professor, Temple University, Philadelphia, PA
Background and Purpose. Measurement of antigay bias began shortly after Weinberg coined the term “homophobia” in 1972, and theoretical and statistical improvements to early psychometric endeavors have produced sound indicators of this latent construct. Foremost amongst 22 reviewed instruments is the Attitudes Toward Lesbians and Gay Men scale (ATLG) developed and validated by Herek in 1988, which was the first instrument to take a bidimensional approach with separate scales measuring bias against gay men versus lesbians. This enabled investigation of attitudes based on gender and relationships with other constructs, such as sexism. To create a short form of the ATLG (ATLG-S) in 1994, Herek chose five items from each of his gendered scales that were highly correlated with the ATLG global score, and tested them on several nonstudent samples. Internal consistency remained high for both the gay male and lesbian subscales and the short forms were highly correlated with the long versions. Herek reported construct validity as significant associations with measures of traditional sex roles, traditional family ideology, authoritarianism, attendance at religious services, affiliation with conservative religious denominations, and conservative religious beliefs. However, a recent review of the literature did not yield any factor analytic studies of the ATLG-S. We sought to verify the factor structure of the ATLG-S by conducting two independent studies with separate samples. First, we performed an exploratory factor analysis (EFA) to investigate the measure suggested by Herek (1994). Then, we used confirmatory factor analysis (CFA) to test the resulting model.
Methods. First, we surveyed 394 students in 19 classes at a Southeastern university and conducted an EFA to test the hypothesized factor structure of the ATLG-S. Then, we surveyed 851 students at four geographically diverse universities. This survey included the ATLG-S, standardized scales, and demographic questions, which were used to test for evidence of factorial validity (CFA), construct validity, known groups, and predictive validity along with a reliability analysis.
Results. Study 1. Data met the assumptions of the analyses. Correlation analysis revealed that all items were correlated with at least one other item (> .30). Results of a principal axis factor analysis with an orthogonal rotation indicated a one factor solution that explained 61% of the variance. Bartlett's Test of Sphericity and the Kaiser-Meyer-Olkin measures were excellent. The measure appeared to be valid. Study 2. The CFAs of both single factor and two-factor models resulted in a highly unacceptable model fit, and minimal improvement was made after implementing the suggested modifications. Because the number of correlations required for improvement defied a logical conceptualization, our conclusion is that the ATLG-S is not valid with our sample of college students.
Conclusion and Implications. This study highlights the importance of confirming the validity of measures selected for use in research and practice. Scholars regularly choose measures that have been frequently used in previous studies, without considering the extent of confirmatory validation on samples that are similar to those they wish to study. We will discuss the implications of our findings, and their application to social work research.
![[ Visit Client Website ]](images/banner.gif)