Cooperative Learning Demonstrates Significant Gains in Learning

Marzano Research Laboratory

Cooperative learning provides students with opportunities to work together in small groups.

The following table presents a summary of findings from two meta-analyses of the experimental/control action research studies in Marzano Research Laboratory’s Meta-Analysis Database which utilized this strategy (for a listing of the action research studies, click here. One meta-analysis was conducted using the reported effect sizes from the action research studies in our database. The second meta-analysis (findings reported in parentheses) was conducted using effect sizes that were corrected for attenuation due to a lack of reliability often associated with teacher-designed assessments of student academic achievement (for a discussion of attenuation and meta-analysis and the method used to correct for attenuation click here. Both meta-analyses employed a random-effects model of error (for a discussion of models of error in meta-analysis click here.

Consulting a table of the normal curve, the overall percentile gain associated with the corrected weighted average effect size of 0.84 is 0.2996. This suggests that on the average, the use of cooperative learning by teachers in the action research studies was associated with a gain in student academic achievement of 30 percentile points over what was expected when teachers did not use cooperative learning. In order to illustrate this gain, consider a hypothetical student who is ranked in the middle of a control group with 100 students. Under an assumption that everything else is equal, if this student were the only one to receive instruction with the strategy, his or her ranking would improve from 50th to 20th. In other words, the student would be expected to surpass 30% of the students that did not receive instruction with the strategy.

The effect sizes reported in the table are weighted averages of all the effect sizes from the action research studies and should be considered estimates of the true effect size of the experimental condition (i.e., use of cooperative learning). For example, consider the corrected weighted average effect size reported in parentheses, 0.84 ± 1.27. This mathematical expression represents the 95% confidence interval and includes the range of effect sizes (-0.43 to 2.11) in which one can be 95% certain the true effect size falls. When the confidence interval does not include 0.00, the weighted average effect size can be considered statistically significant. In other words, an effect size of 0.00 would not be considered a reasonable assumption.

It is worth noting that these findings are consistent with the earlier meta-analytic studies on cooperative learning.

"Meta-Analytic Synthesis of Studies Conducted at Marzano Research L...