Baron and Kenny (1986) nicely highlighted the distinction
between moderator and mediator variables in psychological
research. A moderator model is operative when the strength or
direction of a relation between two variables varies as a function
of a third variable, called the moderator variable. The moderator
need not affect the outcome variable itself. Rather, it
represents a set of conditions that qualify the relation between
two or more other variables. A mediational model is operative
when one variable affects a second (outcome) variable only indirectly,
through an intermediate step or process, represented by a
third variable, called the mediator variable. The mediator variable
is simultaneously conceived as being a consequence of the
first variable and a cause of the outcome variable. In a completely
mediational model, the entire effect of the first variable
on the outcome variable is translated through the mediator
variable. Baron and Kenny (1986) further noted that moderational
and mediational models are not mutually exclusive but
can be integrated to form more complex models reflective of
more sophisticated theories.