Group Work in the Science Classroom: How Gender Composition May Affect Individual Performance

Current research on how gender composition within groups influences individual outcomes is both sparse and conflicting. We examined how gender composition within groups affects learning outcomes. Students from sixth, seventh, and eighth grade classes from three US Midwestern public school districts with diverse demographic compositions (N=637, 314 boys and 323 girls) participated in this study as a part of their regular science class during a 12-week design-based physics curriculum, CoMPASS. We conducted two 5 x 2 analyses of covariance to evaluate the effect of group gender ratio and gender on students’ physics learning and science practice outcomes. Results indicate that group gender ratio does influence students’ science learning and practices as measured by posttest differences. Students in mixed-gender groups performed significantly better than students in same-gender groups. Having at least one group member of the opposite gender increased individual students’ posttest performance. Limitations and implications for practice are discussed. Engaging students in group work during inquiry-based and project-based learning activities has become an increasingly common practice in science classrooms. However, as research suggests, students may not always effectively collaborate in ways that foster learning (Barron, 2003; Rummel & Spada, 2005). Further, collaborative learning may not always result in equivalent learning gains for each individual (Teasley & Fischer, 2008; Gnesdilow, Bopardikar, Sullivan, & Puntambekar, 2010). Several factors such as group size, context, gender, prior knowledge, and individual abilities may affect the collaboration in groups (e.g., Apedoe, Ellefson, & Schunn, 2012, Hawkins & Power, 1999). In this paper we focus on understanding how the gender composition in groups affects students’ learning outcomes in science. The current research on how gender composition in groups influences individual outcomes is both sparse and conflicting. Ding, Bosker, and Harskamp (2011) discussed that while Computer Supported Collaborative Learning (CSCL) has the potential to lessen the gender gap between male and female performance and persistence in physics, the positive findings from CSCL research “are controversial where gender is concerned” (p.325). Leman (2010) pointed out that there is a scarcity of empirical research linking “interactions and collaboration to gender and learning outcomes” (p.218). Research has indicated that there are differences between how boys and girls learn, converse, and interact (Leman, 2010; Kommer, 2006; Rice & Dolgin, 2002), including when within mixed-gender groups (Hawkins & Power, 1999) and also within mixedgender dyads (Ding, Bosker, & Harskamp, 2011; Harskamp, Ding, & Suhre, 2012). Some studies have found that girls in mixed-gender groups do not perform as well as girls in same-gender groups (e.g. Light, Littleton, Bale, Joyner & Messer, 2000). Similarly, other studies have revealed that high school girls learning physics in mixed-dyads scored significantly lower on posttests than the boys working in the mixed-dyads, as well as the boys and girls who worked in same-sex dyads (Ding et al., 2011; Harskamp et al., 2012). Alternatively, one of the key findings highlighted by Bennett, Hogarth, Lubben, Campbell, and Robinson’s (2010) review of studies of small groups in science classrooms was that students in single-sex groups were more purposeful than mixedgender groups, but ultimately group gender composition did not affect understanding. In another study, girls participated more actively and persistently on collaborative learning activities when in mixed-gender groups, including generating more science and group orchestration talk during computer-based learning activities (Goldstein & Puntambekar, 2004). Given the contradictions between the findings outlined above, as well as the lack of overall evidence about how gender composition affects students’ learning in groups, we believe that understanding these relationships could lead to strategic and easy-to-implement teaching decisions for enhancing collaboration and learning. In this study we examined how gender composition in groups affects students’ learning outcomes and attempt to answer the research question: Do differences in gender composition affect middle school science students’ learning in groups? We explored this question by examining students’ science content knowledge and practices outcomes. Methods Participants and Instructional Context Two hundred sixth grade, 143 seventh grade, and 294 eighth grade students (N=637, 314 boys and 323 girls) from three US Midwestern public school districts with diverse demographic compositions participated in this study as a part of their regular science class. All students took part in the CoMPASS roller coaster unit, a 12week design-based science curriculum, to learn about forces, motion, work, and energy. They participated in a variety of physical science activities in order to design a fun, safe, and efficient roller coaster for an amusement park whose attendance is waning. Students worked in the same group of three of four throughout the 12-week unit (Group N=178, 54 sixth, 41 seventh, and 83 eighth grade groups), with group composition determined prior to this study by our collaborating teachers. Students took separate preand posttests for science content and practices (described below). Students took these tests before starting and after finishing the CoMPASS roller coaster curriculum in their classes. Data Sources and Analysis Measures We used two tests: the Physics Fiesta measured students’ content knowledge in physics and the Scientist’s IQ tested science practices. The Physics Fiesta consisted of 29 multiple-choice questions and addressed a range of physics concepts and relationships such as mass, work, force, potential and kinetic energy, velocity, acceleration, efficiency, the law of conservation of energy, and Newton’s Laws. Each correct item earned a score of one point and incorrect answers were scored as zero, with 29 points being the highest score possible. The Scientist’s IQ consisted of 13 multiple-choice and five open-ended questions that assessed students’ skills in areas such as interpreting, making inferences, setting up data in graphs and charts, hypothesis writing, variable identification in setting up experiments, using data to back up reasoning and explanations, and identifying measurement and other sources of error in investigational scenarios. Correct multiple-choice responses on the Scientist’s IQ earned one point and incorrect responses were scored as zero. The open-ended questions were graded from 0 to 2 or 3 points. A score of 2 (or 3) indicated a more sophisticated, elaborate response or explanation, while a zero indicated that the answer was incorrect, blank, or unintelligible. Answers coded as 1 or 2 points (for 3-point questions) were correct but not explained well, supported, or were partial responses. The maximum score for the Scientist’s IQ was 24 points. Interrater reliability for scoring the openended responses on the Scientist’s IQ pretest was 94.35% and 92.5% for the posttest. Gender Ratio Group Categories To examine how the gender composition of groups influenced each student’s learning outcome, we used five different Gender Ratio categories. The five categories were: 1) all boys, 2) mostly boys (i.e. 2 or 3 boys in a group of 3 or 4, respectively), 3) even split between boys and girls, 4) mostly girls (i.e. 2 or 3 girls in a group of 3 or 4, respectively), and 5) all girls. Based on the gender composition of the group that a student worked in throughout the CoMPASS curriculum, he or she was labeled as belonging in one of the five categories. For example, if a group consisted of two girls and one boy, each of the three students was labeled as belonging to a mostly girl group. Due to missing data, a total of 574 students (280 boys and 294 girls) completed both preand posttests for the Physics Fiesta and 530 students (259 boys and 271 girls) completed both preand posttests for the Scientist’s IQ. Only the scores of students who completed both preand posttests for a given measure were included in our analysis.