College teachers in the sciences and professional studies have endorsed collaborative learning group strategies for teaching undergraduate statistics courses, but few researchers provide empirical evidence that students' quantitative skills actually increase as a result of the collaborative experience. Assessment of the efficacy of collaborative learning group techniques is frequently subjectively based, and often relies on casual comments from students or faculty. Despite this shortcoming, instructors searching for new and effective ways of teaching quantitative courses continue to experiment with collaborative pedagogy. Consequently, an exploration of the effectiveness of collaborative teaching modalities on students' statistical knowledge seems warranted. This study examined the relationship between performance on collaborative learning group assignments and students' examination scores in statistics. Data were derived from 270 students enrolled in nine sections of a social statistics course, taught between 1996 and 2004. The results both challenge and support the efficacy of collaborative learning groups, and suggest that faculty modify such techniques when evidence of student learning cannot be empirically linked to the collaborative experience. (Contains 3 tables and 3 notes.)