If a choice to participate or not, based on process and style, affects the outcomes of students do you think this would show up on outcome-oriented studies? Or are you contending that this effect would only be visible in individual outcomes, and cancel out across groups of any size?

You may also wish to read the second link from the above comment. It could contain the evidence you seek! http://steinhardtapps.es.its.nyu.edu/create/courses/2174/rea...