In this meta-analysis, the authors attempted to develop a parsimonious mode
l of factors that account for the significant variability in the findings o
n the effects of within-class grouping on student achievement. Two weighted
least squares regression models were tested using 103 independent findings
from 51 studies at elementary through postsecondary grades. Results indica
te that the most important study features that accounted for 48% of the tot
al variance include outcome measure source, teacher training equivalence, g
rouping basis, type of small-group instruction method, grade level, and rel
ative ability of students. Goodness-of-fit statistics indicate that the mod
el fits the data and that the remaining variance may be explained by sampli
ng errors.