Figure 1. Interpretation of the interaction effect
between child's behavior problems and fathers'
self-efficacy on fathers' anxiety. The circle indicates
high self-efficacy; the square, mid self-efficacy;
and the triangle, low self-efficacy. HADS =
Hospital Anxiety and Depression Scale, DBC =
Developmental Behavior Checklist.
of three levels of self-efficacy scores (one SD below
the mean, at the mean value, and one SD
above the mean). Figure 1 shows that at low levels
of child behavior problems, there was little effect
of varying levels of self-efficacy. Conversely, at
higher levels of child behavior problems, those
fathers with lower feelings of self-efficacy reported
more anxiety.
Discussion
The data from the present study show high
levels of potential mental health problems in parents
of children with autism, especially mothers.
Also at the descriptive level, there were only weak
and nonsignificant associations between parents
of the same child, suggesting that mothers and
fathers may be affected differently by their child,
as has been indicated in previous research. Hierarchical
regression analyses helped to elucidate
one mechanism by which parents may be affected
differently by their child's behavior problems.
These analyses identified self-efficacy as a mediator
of the relationship between child behavior
problems and mothers' anxiety and depression,
but not fathers' anxiety and depression. Conversely,
self-efficacy moderated the effect of child behavior
problems on fathers' anxiety. Fathers with
high self-efficacy were less anxious than were
those with low self-efficacy when their child had
a high level of behavior problems. When their
child had low levels of behavior problems, father's
self-efficacy had no effect on their anxiety. There
was no evidence for the moderating effect of selfefficacy
in the analyses of mothers' anxiety or depression.
These findings complement those of research
studies in other fields, especially on parenting, in
that self-efficacy was identified as an important
variable in understanding the relationship between
child behavior problems and parents' mental
health outcomes. Analyses of mothers' data
supported the general prediction that self-efficacy
would function as a mediator variable in this context.
However, analyses of fathers' data indicated
that self-efficacy was not a mediating variable for
them. Rather, there was evidence that self-efficacy
acted as a moderator variable for fathers. In essence,
self-efficacy acted as a protective factor
(Rutter, 1985) for fathers, countering the risk to
anxiety associated with high levels of child behavior
problems. This difference in the mechanism
of action of self-efficacy between mothers and fathers
of children with autism requires replication.
However, the observed distinction emphasizes the
importance of studying mothers and fathers separately,
even where overall levels of mental health
and self-efficacy were relatively similar.
There are a number of methodological issues
that need to be emphasized before we discuss the
theoretical and practical significance of the present
findings. First, the samples of mothers and
fathers were small. This needs to be balanced
against the fact that the effects observed in the
analyses were found to be statistically significant.
With a larger sample, the effect of child behavior
problems on fathers' mental health probably was
substantial enough to emerge as statistically significant.
Even within the small sample here, there
were strong relationships between the child's behavior
problems and mothers' anxiety and depression
(see Tables 2 and 3). However, our focus
in the present study was on process, and the results
of the analyses in Tables 4 and 5 show no
change in the prediction of mental health by child
behavior problems when self-efficacy is added at
Step 3. Thus, there is not even a hint that selfefficacy
might act as a mediator variable for fathers.
Also worthy of note is that it is very difficult
to identify interaction effects in regression in applied
studies (McClelland & Judd, 1993). However,
the interaction term for fathers' anxiety was
found to be significant even in this small sample.
The second methodological issue is that the
present sample was unlikely to have been representative
on a number of levels. First, the response