This study explores ways to make action research “scientific
research.” Specifically, instead of the one-group posttest-only and
one-group pretest-posttest design that most previous action research
studies have adopted, I suggest a posttest-only design with
nonequivalent groups. To solve the problem of comparing nonequivalent
groups, I suggest an analysis of covariance (ANCOVA) model that can
control for learners’ general academic abilities.