The analysis of group contingency data.

Neumann (1977) wrote a how-to paper, not an experiment. The author looked at past group-contingency studies and showed they often lump all kids’ scores into one class average.

The paper gives step-by-step ways to graph each student alone instead. It warns that averages can hide kids who are getting worse while the class mean still looks good.

The paper found that class-wide averages can lie. A steady group mean may mask one child whose behavior is sliding and another who is soaring.

When you plot each child’s data separately, you see who really needs help and who is already done.

Wilson et al. (1973) ran a group contingency in a junior-high room and reported only class means. Neumann (1977) says those same numbers should be unpacked kid-by-kid.

Christophersen et al. (1972) pushed using ANOVA for single-case data, but Lydersen et al. (1974) fired back that ANOVA breaks the independence rule. Neumann (1977) sides with the critics: skip both ANOVA and class averages; graph each child instead.

Iversen (2021) and Nasr et al. (2000) still carry the torch, warning that any blend of single-case data—meta-analysis or class average—can fool you.

If you run group contingencies, stop looking only at the class line. Print one mini-graph per student before you pick the next target skill or fade the reward. You will catch the silent outliers who need more help and the quick masters ready for tougher goals.