"Active learning" has been used to describe classrooms that have varied widely with respect to instructional topics, age of learners, and the procedures used to operationalize the general notion of the term. In most cases, the specific variant of active learning under investigation has been more effective than the particular control used for comparison. The goal of the current study was to unambiguously describe, implement, and assess 4 different active learning implementations that varied based on the instructional technique employed by the teacher. The specific topic taught was the procedure for constructing experiments in which a single causal factor is clearly identified and there are no confounds. The procedure is commonly known in the literature on early scientific thinking as the "control of variables strategy" (CVS). The sample consisted of 145 3rd- and 4th-grade students from 3 schools. Students in each grade at each school were randomly assigned to 1 of 4 active learning conditions. Learning of CVS was measured through a hands-on, active learning activity and a written pre- and posttest. Results indicated that compared to minimal guidance/minimal guidance/activity, modeling/direct guidance/activity resulted in significantly higher levels of CVS knowledge on the hands-on activity. When examining student learning from pre- to posttest, students in all conditions had significant learning gains. However, the largest effect sizes were for modeling/direct guidance/activity followed by modeling/modeling/activity, and the weakest effect size was for minimal guidance/minimal guidance/activity. Thus, more direct/explicit forms of active learning promoted higher learning of CVS than more inquiry-based forms.