Qin, Johnson, and Johnson compared the impacts of cooperative and competitive efforts on problem solving. They integrated the findings from 48 studies by means of a meta-analysis. Meta-analysis is a statistical method that combines the results of different studies of a process to draw conclusions about the overall results of implementing the process. Its purpose is to determine the size of the effect of independent variables on a dependent variable. In this case, the dependent variable is effectiveness in individualized problem solving. Some independent variables of interest to chemists are cooperation versus competition and the type of problemfor example, linguistic versus nonlinguistic or well-defined versus ill-defined. Linguistic problems are primarily presented and solved in written or oral language; nonlinguistic problems are primarily presented and solved in pictures, mathematics, graphs, symbols, etc. A well-defined problem has a clearly specified goal and representation. An ill-defined problem contains some uncertainty about its procedures and goals; a research project would be an example in chemistry.

The authors restricted their analysis to studies of problem solving and excluded those that simply involved training to do exercises. They took the following components as the essential elements of problem solving: the solution process requires solvers to interpret the problem, not simply to recognize it as an example of a type; to plan potential sequences of actions or procedures to solve the problem; and to execute the plan and check the results. The authors were equally careful to limit their analysis to studies that actually compared cooperation (as opposed to group work) with competition. They defined cooperation as the presence of joint goals, mutual rewards, shared resources, and complementary roles among members of a group. Competition was defined as the presence of a goal or reward that only one or a few group members could achieve by outperforming the others (grading on a curve is one example). Finally, the research studies used were ranked by their methodological adequacy based on a random assignment of participants to treatment (cooperative learning sections) and control sections (competitive sections), clarity of definition of the nature of the control section (an explanation of what the control section actually did, rather than simply "traditional instruction"), control for experimenter effects (experimenters or teachers rotated through both sections), use of the same content and curriculum for both treatment and control sections, and verification that the pedagogy proposed for the treatment and control sections was actually used.

The analysis shows that cooperation resulted in higher-quality individual problem solving than did competition. The average individual at the 50th percentile from the cooperative groups solved problems better than 72.5% of the population from the competitive groups. The effect of cooperation was greater on all four types of problems: linguistic, nonlinguistic, well-defined, and ill-defined. Interestingly, cooperation appeared to be more effective on problems that might be of interest to chemistssolving nonlinguistic problems benefited more than solving linguistic problems.