A commentary is presented on the colloquium << Alternatives to Classic
al Statistical Procedures>> organized by Zumbo. Each presentation of t
he colloquium is summarized and briefly critically reviewed. Together,
the presentations yield an array of diverse methodological solutions
to the fact that psychological data do not conform to the ideal of the
normal distribution nor to the assumptions of parametric analysis (e.
g., independent observations). These considerations suggest that psych
ologists should question how their variables of interest are measured.
The present commentary adds to this suggestion the following one: psy
chologists should also critically review the effect size of the phenom
ena they study. The consideration of both measurement problems and eff
ect size should bring psychologists towards a coherent use of inferent
ial statistics.