the first place. Furthermore, the failure to establish such beliefs in
childhood can result in a variety of problematic outcomes, including
the emergence of depressive symptoms (Cole, 1990; Cole,
Martin, & Powers, 1997; Jordan & Cole, 1996). According to this
conceptualization, the belief in one's self-competence may function
as a mediator of the relation between life events and depressive
symptoms, especially in young people (Baron & Kenny,
1986).
Mediational and moderational processes are not necessarily
mutually exclusive. On the one hand, evidence already exists that
self-competence meets three of the four criteria for being a mediator
(Kenny, Kashy, & Bolger, 1998). First, negative events relate
to depression in young people (Cole & Turner, 1993; Compas et
al., 1989; Compas, Slavin, Wagner, & Vannatta, 1986; Turner &
Cole, 1994). Second, negative events relate to self-perceived competence
(Cole, Maxwell, & Martin, 1997). Third, low levels of
perceived competence relate to depressive symptoms (Cole, Martin,
& Powers, 1997; Cole, Martin, Powers, & Truglio, 1996;
Turner & Cole, 1994). The final criterion, however, has yet to be
tested well. No one has demonstrated that the effect of negative
events on depressive symptoms diminishes after controlling for
self-perceived competence. Research has shown, however, that
closely related variables (i.e., attributional style and negative cognitive
errors) do account for part of the relation between negative
events and depressive symptoms (Cole & Turner, 1993).
On the other hand, as self-competence develops in the child or
adolescent, it may begin to serve as a moderator of the relation
between negative events and depression. The interaction between
negative events and self-competence has not been formally tested;
however, a number of similar tests suggest that such an interaction
might well exist, especially in later childhood or adolescence. In
longitudinal studies with college students, Metalsky and Joiner
(1992; Metalsky et al., 1993) reported interactions between negative
events and explanatory style, personal hopelessness, and selfesteem
in the prediction of self-reported depression in late adolescence
(i.e., college students). In a 5-year longitudinal study of
children and adolescents, Nolen-Hoeksema, Girgus, and Seligman
(1992) noted that the interaction between explanatory style and
negative events did not predict depressive symptoms in younger
children but that the interaction was significant at older ages.
Similarly, Robinson, Garber, and Hilsman (1995) described significant
interactions between stressful events and measures of
explanatory style and self-esteem during the transition from elementary
to middle school. Evidence of such interactions has been
weak or inconsistent at younger age levels (Cole & Turner, 1993;
Dixon & Ahrens, 1992; Hammen, 1988; Hammen, Adrian, &
Hiroto, 1988; Nolen-Hoeksema, Girgus, & Seligman, 1986; cf.
Conley, Haines, Hilt, & Metalsky, 1999; Hilsman & Garber,
1995).
The purpose of the present study was to examine the role of
self-competence beliefs as mediators and moderators of the relation
between life events and depressive symptoms in adolescence.
Mindful of the fact that mediational and moderational models
describe causal relations, we implemented a longitudinal design
and statistically controlled for prior levels of our dependent variables.
Aware of the multidimensional nature of self-concept
(Wylie, 1974), we assessed five different domains of selfperceived
competence. Cognizant of the problems inherent in
assessing life events in children, we focused on age-appropriate
life events and used a nomothetic system for evaluating the valence
of such events (Cole & Turner, 1993; Compas, 1987; Johnson,
1986).
Method
Participants
The participants in this study were part of a larger longitudinal study of
child and adolescent depression (Cole et al., 1996; Cole, Martin, & Powers,
1997; Cole, Peeke, Dolezal, Murray, & Canzoniero, 1999). When the
participants reached ninth grade, we began to administer a life events
inventory, making the present study possible. (Earlier administration of the
events inventory was not possible because of time constraints and school
administrative concerns.) The participants in this study were tested twice,
once in the Fall and once in the Spring of the 1996-1997 school year. The
participants were 468 ninth graders from two public high schools in a
midsize Midwestern city. This sample derived from a larger pool of 512
students. Loss of participants was due to parental nonconsent (n = 32) and
to student attrition (n ~ 12). Nonparticipants did not differ from the
participants with regard to race, gender, age, family size, family income, or
(when available) scores on the instruments administered at the onset of the
study (ps > .10). The participants' ages ranged from 13 to 17 years
(M = 14.5, SD = 0.6). The sample was 55% male