Content uploaded by Craig Alan Mason
Author content
All content in this area was uploaded by Craig Alan Mason
Content may be subject to copyright.
Popularity, Friendship Quantity, and Friendship Quality: Interactive
Influences on Children’s Loneliness and Depression
Douglas W. Nangle, Cynthia A. Erdley, Julie E. Newman, Craig A. Mason, and
Erika M. Carpenter
Department of Psychology, University of Maine
A mediational model positing that the effects of popularity on children’s loneliness
and depression are passed through indexes of friendship experiences was tested using
structural equation modeling. Children (193 3rd through 6th graders) completed a
battery of sociometric and self-report questionnaires from which measures of popu
-
larity, multiple friendship dimensions (i.e., quantity and quality of best and good
friendships), and loneliness and depression were derived. Confirmation of a slightly
modified model supported the mediational hypothesis. Although popularity exerted
no direct impact on the adjustment indexes, it strongly influenced friendship, which,
in turn, affected depression through its strong association with loneliness. It appears
that popularity is important for setting the stage for relationship development, but
that it is dyadic friendship experiences that most directly influence feelings of loneli
-
ness and depression.
The bulk of empirical support for the association
between children’s peer relationships and their psy-
chological adjustment comes from investigations of
popularity, a unilateral construct referring to the de-
gree to which a child is liked by the peer group (see
Newcomb, Bukowski, & Pattee, 1993, for review). Re-
searchers have made another distinction in children’s
peer experiences by examining friendship, a bilateral
construct referring to a child’s participation in a close,
mutual, dyadic relationship. This distinction is impor
-
tant, as popularity and friendship are constructs with
overlapping yet unique influences on development and
adjustment (Buhrmester & Furman, 1986; Furman &
Robbins, 1985; Sullivan, 1953). Sullivan proposed that
the relative importance of popularity and friendship
changes across development as different social needs
emerge. The need for group acceptance is paramount
in middle childhood and gives way to an increasing re
-
liance on intimacy as children enter preadolescence.
Despite their changing roles and importance,
Sullivan (1953) believed that both popularity and
friendship continue to impact adjustment throughout
childhood. Consistent with this notion, even studies of
young children demonstrate the unique contribution of
friendship to adjustment (e.g., Ladd, Kochenderfer, &
Coleman, 1997; Parker & Asher, 1993). For example,
using a middle childhood sample, Parker and Asher
found that having a friend, friendship quality, and pop-
ularity made unique contributions to the prediction of
loneliness. Interestingly, not all high-accepted children
had friends, and many low-accepted children did have
friends. Furthermore, regardless of their level of group
acceptance, children without friends were lonelier than
children who were involved in at least one mutual
friendship. Nevertheless, in a demonstration of the
overlap in the peer experience dimensions, better ac
-
cepted children were almost twice as likely to have a
friend as their less accepted peers. In addition, the
better accepted children tended to perceive their
friendships as being higher in quality. Examining the
close association between popularity and friendship,
Bukowski, Pizzamiglio, Newcomb, and Hoza (1996)
found support for the idea that being liked by the group
is antecedent to friendship development. That is, a
larger social network affords a child more opportuni
-
ties for friendship formation.
Further discriminations in the friendship construct
have been proposed in the hopes of more adequately
capturing the fullness of children’s peer experiences
(Bukowski & Hoza, 1989; Hartup, 1996). As part of
their hierarchical friendship model, Bukowski and
Hoza recommended a multilevel assessment com
-
posed of determining whether a child has a friend, how
many mutual friendships that child has, and what the
quality of those friendships is. The importance of hav
-
ing a friend to a child’s development and adjustment is
well established (e.g., Parker & Asher, 1993; Sander
-
son & Siegal, 1995). The value of the quantity dimen
-
Journal of Clinical Child and Adolescent Psychology
2003, Vol. 32, No. 4, 546–555
Copyright © 2003 by
Lawrence Erlbaum Associates, Inc.
546
Thanks to Bill Bukowski for his valuable input in developing the
conceptual basis for this project. We would also like to acknowledge
the assistance of the principals and teachers at the Hermon Elemen
-
tary, Caravel Middle, Levant Elementary, and Holbrook Schools.
Without their efforts, we could not have completed the study.
Requests for reprints should be sent to Douglas W. Nangle, De
-
partment of Psychology, University of Maine, Orono,
ME 04469–5742. E-mail doug.nangle@umit.maine.edu
sion (i.e., the number of friendships in which a child is
involved), however, is not as clear. In one study, Ladd
et al. (1997) determined that although friendship quan
-
tity did not add to the prediction of children’s loneli
-
ness as they made the transition to kindergarten, it did
add to the prediction of other adjustment indexes, such
as academic readiness and school involvement. In
comparison, evaluations of the quality dimension, a
child’s impression of the degree to which a given rela
-
tionship meets specified needs ranging from the provi
-
sion of opportunities for play and companionship to al
-
lowing for intimate disclosure and exchange (e.g.,
Berndt & Perry, 1986; Parker & Asher, 1993), have es
-
tablished its worth as a distinct component of friend
-
ship with implications for adjustment (e.g., Hoza,
Bukowski, & Beery, 2000; Oldenburg & Kerns, 1997;
Parker & Asher, 1993).
Evaluating the proposed multidimensionality of
friendship would be facilitated by the use of more re
-
cently developed statistical procedures. To date, most
studies have employed covariance techniques, such as
regression, to determine whether friendship has any ef
-
fect on single adjustment indicators after the effects at
-
tributable to popularity, the more established predictor,
are partialed out (Bukowski & Hoza, 1989). Because
friendship shares variance with popularity, however, its
overall effects on adjustment are underestimated by
considering only nonshared variance (Bukowski &
Hoza, 1989). Alternatively, structural equation model-
ing allows for the examination of simultaneous effects
of multiple predictors on one or more outcome mea-
sures while taking the interrelations among the predic-
tors into consideration (see Biddle & Marlin, 1987, for
a review). Another advantage is that the “fit” of the
overall model is assessed. For example, in a study with
early adolescents, Bukowski, Hoza, and Boivin (1993)
found support for a complex model positing that popu
-
larity exerts an indirect influence on loneliness through
its effects on mutual friendship and feelings of
belongingness.
This study examined the simultaneous influences of
popularity and multiple friendship dimensions of
friendship on children’s loneliness and depression us
-
ing structural equation modeling. Measures of popu
-
larity, friendship quantity, friendship quality, and lone
-
liness and depression were derived from sociometric
procedures and psychological adjustment question
-
naires. The longstanding question of whether friend
-
ship is a categorical or continuous construct (Hartup,
1996; Price & Ladd, 1986) was addressed by the use of
two different friendship levels: (a) “good” friendships,
defined as whether either of two peers nominated the
other and the two gave one another peer ratings of at
least 4 and (b) “best” friendships, defined by reciprocal
best friend nominations (see Erdley, Nangle, & Gold,
1998, for a comparison of different friendship defini
-
tions). Another unique aspect of this study was the
evaluation of the contributions of having more than one
friendship and of having friendships of varying quality
to adjustment.
The use of middle childhood participants and selec
-
tion of loneliness and depression as adjustment indica
-
tors was based on several considerations. As outlined
earlier, during this developmental stage, children are
just beginning to experience a shift in the priority of
their social needs from acceptance to intimacy. As
such, this is an opportune time to evaluate a model pos
-
iting a mediational role for friendship. Furthermore, in
addition to the solid body of studies linking popularity
and the various friendship dimensions with this age
group, there is a well-developed literature connecting
peer variables to children’s loneliness and depression.
Numerous studies have shown that children who are
rejected by their peer group experience higher levels of
loneliness and social dissatisfaction than their better
accepted peers (see Asher, Parkhurst, Hymel, & Wil
-
liams, 1990, for a review). Even so, there is consider
-
able variability in rejected children’s reports of loneli
-
ness. In addition to the actual degree of rejection
experienced by children and its chronicity, Asher et al.
suggested that this variability may be due to the fact
that some rejected children have at least one friend who
provides an important source of emotional support, an
assertion validated in a number of more recent studies
(e.g., Hoza et al., 2000; Parker & Asher, 1993; Sander-
son & Siegal, 1995). Depression is also associated with
problematic peer relationships (Boivin, Hymel, &
Bukowski, 1995; Boivin, Poulin, & Vitaro, 1994;
Burks, Dodge, & Price, 1995). As with loneliness,
friendship experiences can serve as a buffer between
peer group rejection and depression (Bagwell, New
-
comb, & Bukowski, 1998; Oldenburg & Kerns, 1997).
In summary, we tested a mediational model in which
the effect of popularity on the adjustment indexes is
passed through the friendship variables (see Figure 1).
Popularity was hypothesizedto form direct associations
with the friendship variables, but not with loneliness or
depression (Bukowski et al., 1993; Bukowski et al.,
1996). Reflective of the quantity–quality distinction,
good and best friend quantity and quality were expected
to form two distinct latent variables representing the
friendship construct. Direct links between the latent
friendship variables and loneliness were also posited.
Finally, our hypothesis that loneliness would mediate
the relation between the peer variables and depression
was based on the findings of Boivin et al. (1995). In
model testing with longitudinal data, these authors
found that loneliness mediated the relation between
negative peer experiences, such as withdrawal and re
-
jection, and depressed mood. Such experiences appear
to lead to depressed mood when they cause children to
feel badly about their social situations. No a priori sex
hypotheses were generated. Despite speculation that in
-
ternalizing problems may be more closely related to
547
FRIENDSHIP AND ADJUSTMENT
group acceptance for boys and more closely related to
friendship for girls (Burks et al., 1995), some studies
have found sex differences in how peer variables relate
to internalizing difficulties (e.g., Oldenburg & Kerns,
1997) and others havenot(e.g.,Parker&Asher,1993).
Method
Participants
The participants included 193 third- through
sixth-grade students (103 boys, 90 girls; 99% Euro
-
pean American) who were recruited from elementary
schools located in low- to middle-income communities
in northern New England. Data were collected from all
students who had permission from their parent or
guardian to participate in this project and who gave
their own assent (76% participation rate).
Procedure
The data were collected across two testing sessions,
spaced approximately 2 weeks apart. All experimental
sessions lasted approximately 45 min and were con
-
ducted in the children’s classrooms using group ad
-
ministration of the measures. In the first session, chil-
dren were asked to indicate how much they liked to
play with each of their classmates (peer acceptance rat
-
ings). They were also instructed to circle the names of
their three best friends (limited positive nominations).
Then the children completed measures that assessed
feelings of loneliness and depression.
Prior to the second session, every possible dyad
within each classroom was examined to determine
whether that dyad met fairly liberal criteria for friend
-
ship (i.e., either of the two peers nominated the other
and the two had an average mutual peer rating of 4 or
greater; Berndt & Perry, 1986). In the second session,
children completed a friendship quality questionnaire
for each of their previously identified friendships. All
children rated at least three friendships. If children did
not have three friendships that met our criteria, they
completed friendship quality questionnaires concern
-
ing the three children they circled as their best friends,
three children they rated highly, or some combination
of the two. Friendship quality data for such
nonreciprocated friendships were discarded from fur
-
ther analyses.
With the exception of the friendship quality mea
-
sure, all questionnaires were read out loud to ensure
that children understood, regardless of their reading
548
NANGLE ET AL.
Figure 1. Proposed model linking the peer and adjustment variables.
level. Children were taught how to use each response
scale via several practice items, and they reported their
responses in individual packets.
Measures
Sociometric assessment. Children’s level of
peer acceptance was assessed by presenting students
with a class roster (listing only those students who had
permission to participate) and asking them to rate on a
5-point Likert scale ranging from 1 (I don’t like to)to5
(Iliketoalot) how much they like to play with each of
their classmates. A child’s peer acceptance score was
the mean rating received from all participants in the
classroom who rated him or her. For the peer nomina
-
tion measure, children were asked to circle the names
of their three best friends on a separate class list.
Best-friend quantity was defined as the total number
of reciprocated best friend nominations for each child.
Good friend quantity was defined as the total number
of friendships for each child that met the following cri
-
teria: either of the two peers nominated the other, the
two children gave each other a peer rating of 4 or
greater, and the pair had not met the criterion of recip-
rocated positive nominations. We had originally in-
tended to use the good friend definition put forth by
Berndt and Perry (1986) as described previously but
chose an alternative because of the fact that requiring
only an average peer rating of 4 leaves room for neutral
ratings.
Friendship quality. Children reported their per-
ceptions of various qualitative aspects of their previ-
ously identified friendship dyads using a modified ver
-
sion of the Friendship Quality Questionnaire–Revised
(Parker & Asher, 1993). This modified questionnaire
included the three items that, according to Parker and
Asher, showed the strongest loadings on each of the
following six factors: validation (e.g., “My friend
makes me feel good about my ideas”), conflict (e.g.,
“My friend and I argue a lot”), conflict resolution (e.g.,
“My friend and I make up easily when we have a
fight”), help and guidance (e.g., “My friend helps me
so I can get done quicker”), companionship (e.g., “My
friend and I always sit together at lunch”), and inti
-
macy (e.g., “My friend and I always tell each other our
problems”).
Children responded to these items on a 5-point
scale ranging from 1 (not at all true)to5(really
true). Each child rated at least three friendships. For
each questionnaire, the name of a specific friend was
inserted into each individual item using word-pro
-
cessing software. According to Parker and Asher
(1993), this method is used to reduce the likelihood
that children will complete the questionnaire based
on an ideal friendship or mental representation of a
combination of many different friendships. Consis
-
tent with previous research using this measure, a total
friendship quality score was used in this study
(Parker & Asher, 1993). Best friend quality was the
mean of the average ratings for each of the six
Friendship Quality Questionnaire–Revised factor
scores across each child’s identified best friendships
(α = .90 for boys and .95 for girls using the modified
version). Likewise, good friend quality was the mean
of the average ratings for each of the six Friendship
Quality Questionnaire–Revised factor scores across
each child’s identified good friendships (α =.81for
boys and .87 for girls using the modified version).
Loneliness. Using the Asher and Wheeler
(1985) Loneliness and Social Dissatisfaction Ques
-
tionnaire, children were asked to rate themselves on a
5-point scale ranging from 1 (that’s not true at all
about me)to5(that’s always true about me) across
24 items, 16 of which assess feelings of loneliness
and social dissatisfaction at school (e.g., “I feel alone
at school” and “There are no other kids I can go to
when I need help at school”) and 8 of which are filler
items. This measure has been used extensively with
third- through sixth-grade children (Asher et al.,
1990). Results of factor analyses reported by Asher et
al. revealed a single factor comprising 16 primary
loneliness and social dissatisfaction items. Consistent
with previous research using this measure (e.g.,
Asher et al., 1990), reliability analyses based on data
collected for this study revealed high internal consis-
tency (coefficient α = .93).
Depression. Children’s feelings of depression
were assessed using the Children’s Depression Inven
-
tory (Kovacs, 1985), a 27-item self-report question
-
naire that assesses the presence and severity of affec
-
tive, cognitive, and behavioral symptoms of
depression. For each item, participants chose one of
three responses that described their feelings and ideas
during the past 2 weeks (e.g., “I am sad once in a
while,” “I am sad many times,” “I am sad all the time”).
Responses were scored on a 3-point scale ranging from
0(symptom is absent)to2(symptom is present most of
the time). For this study, the Institutional Review
Board asked us to remove the suicidal ideation item.
As such, children completed 26 items and could obtain
a total score that ranged from 0 to 52, with higher
scores indicating greater incidence and severity of de
-
pressive symptoms. Adequate psychometric properties
for the Children’s Depression Inventory have been re
-
ported across several studies (Carey, Gresham,
Ruggiero, Faulstich, & Enyart, 1987; Kovacs, 1985;
Smucker, Craighead, Craighead, & Green, 1986). In
-
ternal consistency for the Children’s Depression Inven
-
tory in this study was .90.
549
FRIENDSHIP AND ADJUSTMENT
Results
Preliminary Analyses
A correlation matrix that includes means and stan
-
dard deviations for the peer and adjustment variables is
presented in Table 1. As expected, for both boys and
girls, popularity was generally highly correlated with
each of the friendship dimensions. Correlations among
the friendship variables ranged from highly significant
for the good friend quantity–best friend quantity corre
-
lation, to nonsignificant for the good friend quan
-
tity–best friend quality correlation. Correlations be
-
tween the friendship variables and loneliness were in
the expected directions, although the correlations be
-
tween the quality variables and loneliness were not sig
-
nificant among girls. Finally, loneliness was highly
correlated with depression for both boys and girls.
Proposed Model
Based on the correlations, it was decided that the er
-
rors for the two best friend variables and the errors for
the two good friend variables be allowed to covary.
Preliminary structural equation analyses also sug-
gested that the two separate factors representing
Friendship Quality and Friendship Quantity either be
integrated into a single factor or combined under a sec-
ond-order Friendship factor. Both of these models
were tested and resulted in similar coefficients be-
tween the key constructs of Popularity, Friendship,
Loneliness, and Depression. However, an examination
of the fit indexes and the various factor loadings sug-
gested that a model based on a second-order factor
comprising the friendship quality and friendship quan
-
tity latent variables was preferred.
With these modifications, the final model was a
good fit to the data. Although the sample-size sensitive
chi-square test was significant, χ
2
(10) = 27.102, p =
.003, the Normed Fit Index and comparative fit index
suggested an excellent fit (.990 and .994, respectively).
The final model, with standardized path coefficients, is
presented in Figure 2. Popularity and the friendship
variables accounted for 33.7% of the variance in loneli
-
ness scores, and the entire model accounted for 36.8%
of the variance in depression scores.
Alternative Direct-Effects Model
Even though the proposed model proved to be a
good fit with the data, additional analyses were per
-
formed comparing it to a model in which popularity
had a direct effect on loneliness and both popularity
and friendship had direct effects on depression. In es
-
sence, although the proposed model posits an entirely
mediational model, this alternative model allows for
both mediation and direct effects among popularity,
friendship, loneliness, and depression. The inclusion
of these additional paths did not improve the fit of the
model, ∆χ
2
(3) = 4.827, p > .10, and none of the addi
-
tional paths were significant. In the interest of parsi
-
mony, the more restricted proposed model is therefore
preferred, suggesting a strong mediational pathway.
Sex Equivalence
Finally, analyses were performed examining the sex
equivalence of the model. A multigroup analysis of the
final model with all means, intercepts, path coeffi-
cients, variances, and covariances required to be equal
for both boys and girls was tested. This highly restric-
tive model forces all standardized and unstandardized
path coefficients and covariances to be identical across
sex, but nevertheless resulted in a marginally good fit
to the data, χ
2
(45) = 72.344, p = .006, Normed Fit In-
dex = .974, comparative fit index = .990. Allowing
means and intercepts for all measured variables to vary
based on sex resulted in a significant improvement in
the fit of the model, ∆χ
2
(8) = 18.089, p < .05. Allowing
the coefficients for the three key mediational paths
(popularity to friendship, friendship to loneliness, and
loneliness to depression) to vary resulted in a margin
-
ally significant improvement in fit, ∆χ
2
(3) = 6.965, p =
.07). Closer examination suggested that this was due to
550
NANGLE ET AL.
Table 1. Means, Standard Deviations, and Correlations Amongst the Peer and Adjustment Variables
MSD
Variables 1 2 3 4 5 6 7 Boys Girls Boys Girls
1. Popularity — .55*** .72*** .49*** .43*** –.54*** –.25* 2.78 2.71 .74 .55
2. Best friend quantity .55*** — .69*** .61*** .06 –.50*** –.27** 1.62 1.61 1.12 .94
3. Good friend quantity .65*** .64 — .57*** .52*** –.57*** –.31*** 1.59 1.32 1.49 1.26
4. Best friend quality .38** .47*** .36*** — .23* –.54*** –.36*** 2.97 3.27 1.30 1.27
5. Good friend quality .28 .11 .54*** .11 — –.26** –.16 2.50 2.34 1.15 1.12
6. Loneliness –.22* –.37*** –.39*** –.20 –.07 — .68*** 2.02 2.05 .61 .60
7. Depression –.12 –.15 –.19 –.19 –.17 .51*** — 7.48 6.56 7.63 6.08
Note: Correlations for boys (n = 103) are reported above the diagonal, girls (n = 90) below the diagonal. Children without best (n = 33) or good (n
= 54) friendships were scored 0 on the respective quality measure.
*p < .05. **p < .01. ***p < .001.
sex differences in the loneliness to depression pathway,
∆χ
2
(1) = 6.527, p = .01. Specifically, the pathway be-
tween loneliness and depression was greater for boys
(standardized path = .700) than for girls (standardized
path = .490). A series of analyses examining potential
sex differences in the factor loadings of the friendship
quality or the friendship quantity latent variables, or
the Friendship second-order factor, found no evidence
of sex differences—with the exception of the previ
-
ously noted difference in intercepts for the measured
variables.
Discussion
Somewhat surprisingly, in light of the suppositions
of Sullivan (1953) and others, we found support for a
fully mediational model using a middle childhood
sample. Differing from past studies with this age
group, we incorporated a multidimensional assessment
of the friendship construct and a statistical approach al
-
lowing for an analysis of simultaneous influences.
Consistent with our proposed model, popularity
strongly influenced friendship, which emerged as a
single factor combining the two posited latent quantity
and quality variables. Also consistent with our hypoth
-
eses, the impact of the peer variables on depression
was passed through loneliness.
These findings do not detract from the importance
of popularity. Rather, they shed light on the possible
pathways through which popularity might influence
friendship and adjustment. In teasing apart the popu
-
larity and friendship constructs, Bukowski et al. (1996)
found support for the contention that popularity is an
affordance of friendship. Temporally antecedent to
friendship, popularity increases the likelihood that
children will form dyadic relationships. Better ac
-
cepted children would have increased opportunities to
form friendships because the “pool” of peers who like
them is larger. In addition to this increased exposure,
better accepted children would also be more likely to
have the social skills needed to capitalize on these op
-
portunities and form more lasting, higher quality rela
-
tionships (Asher, Parker, & Walker, 1996; Parker &
Asher, 1993). Another factor increasing the chances
that relationships would be formed is the likelihood
that the peers who like these better accepted children
are themselves more accepted and socially skilled
(Nangle, Erdley, & Gold, 1996).
In turn, it is the increased size and relative quality of
the friendship networks of more popular children that
appears to buffer them from feelings of loneliness and
551
FRIENDSHIP AND ADJUSTMENT
Figure 2. Final model linking the peer and adjustment variables. †, parameter fixed, all other paths significant at p < .001.
social dissatisfaction. Thus, even before adolescence,
the mutuality that is unique to friendships appears to be
critical. Perhaps the provisions primarily associated
with close friendships, which include affection, inti
-
macy, and sense of reliable alliance (Furman & Rob
-
bins, 1985), are more important in determining loneli
-
ness than the sense of inclusion that results from group
acceptance. In light of the theories of Sullivan and oth
-
ers suggesting that with age friendship becomes in
-
creasingly important, more research addressing devel
-
opmental differences in the relation between
friendship and adjustment is needed.
The centrality of the quantity and quality dimen
-
sions in determining adjustment strongly supports the
need for a hierarchical conceptualization of the friend
-
ship construct (Bukowski & Hoza, 1989; Hartup,
1996). Regarding the quantity dimension, the number
of both types of friends was important in predicting ad
-
justment. Considering that our findings in the context
of past research is made difficult by the fact that most
studies have investigated the role of quantity using ad
-
olescent samples. Echoing our earlier suggestion,
more research addressing possible developmental dif
-
ferences is needed. Confirmation of the significance of
the quality dimension in determining adjustment is a
robust finding that has held across a number of studies
with children of varying ages (e.g., Bukowski et al.,
1993; Hoza et al., 2000; Oldenburg & Kerns, 1997;
Parker & Asher, 1993). Our findings add to existing re-
search by showing that the beneficial effects of friend-
ship quality extend beyond that provided by one best
friendship. Children’s ability to maintain higher qual-
ity mutual relationships across multiple dyads is im-
portant. Taken together, our quantity–quality results
suggest that the positive effects of friendship are cumu
-
lative. This speaks to the longstanding question of
whether friendship is a categorical or continuous con
-
struct (Hartup, 1996; Price & Ladd, 1986). Taking a
categorical view, Bukowski and Hoza (1989) criticized
more liberal friendship definitions used to identify
“good” friendships (e.g., Berndt & Perry, 1986) on the
basis that they do not capture the more intense recipro
-
cal liking they consider to be the essential feature of
friendship. Nevertheless, in concert with the stipula
-
tions of Berndt (1981), the direct links formed between
the good friend dimensions and the latent quantity and
quality variables in the confirmed model indicate that
these somewhat less intense relationships play an im
-
portant role in determining adjustment. As such,
friendship appears to be more of a continuous con
-
struct comprised of varying levels of relationships. At a
more conceptual level, the notion of continuity fits
with the proposals of Sullivan (1953) and others (e.g.,
Furman & Robbins, 1985) that different functions are
served by different types of relationships. If mere ac
-
quaintances are thought to fulfill important functions
that sometimes overlap with those served by close
friendships, then it seems reasonable to assume that
good friendships make unique contributions to chil
-
dren’s adjustment. For instance, although children may
experience less intimacy in their good friendships,
these relationships are likely to provide valuable
sources of instrumental aid, nurturance, worth en
-
hancement, and companionship. Unfortunately, de
-
spite the debate in the literature, we are not aware of
any other studies addressing the possibility that there
may be qualitative differences in the relationships
identified using the different definitional criteria. No
doubt, progress in this area has been hampered by the
continued use of as many as five different definitions as
interchangeably representing the friendship construct
(see Erdley et al., 1998, for a review).
Loneliness proved to be the gateway through which
the peer variables, particularly friendship, impacted
depressed mood. Fitting well with past studies testing
multivariate models, peer difficulties appeared to in
-
crease the risk of depression when they became inter
-
nalized and adversely affected children’s perceptions
of their social environment (Boivin et al., 1995). More
-
over, these perceptions were shaped more by children’s
friendship experiences than their overall group accep-
tance (Bukowski et al., 1993). Recent research has be-
gun to explore the possibility that there are subtypes of
loneliness derived from different social relationships
(Hoza et al., 2000). For instance, Hoza et al. deter-
mined that unsatisfactory peer group experiences were
more related to peer-network loneliness (i.e., loneli-
ness associated with isolation from the peer group),
whereas the lack of a close friend was more related to
dyadic loneliness (i.e., loneliness associated with the
absence of a close friendship). In their review, these au
-
thors criticized the one-dimensional structure of the
loneliness measure (Asher & Wheeler, 1985) used in
this study because of its hypothesized lack of sensitiv
-
ity to such subtypes. At the heart of this critique was
the concern that the Asher and Wheeler measure inade
-
quately assesses the loneliness associated with close
friendships. The direct paths between friendship and
loneliness, along with the absence of any paths directly
linking popularity and loneliness, found in this study
appear to suggest otherwise. Retesting our model using
the Peer Network and Dyadic Loneliness Scale devel
-
oped by Hoza et al. (2000) might result in strengthened
connections between the friendship variables and lone
-
liness (i.e., dyadic loneliness), as well as an emerging
link between popularity and loneliness (i.e., peer-net
-
work loneliness). Another interesting possibility is that
the distinction might better explain the stronger con
-
nection between loneliness and depression found for
the boys in our sample. Hoza et al. found that boys ex
-
perienced more dyadic loneliness than did girls.
The aforementioned loneliness finding notwith
-
standing, we found no sex differences in our overall
model. Our reluctance to form any a priori sex hypoth
-
552
NANGLE ET AL.
eses was based on the mixed findings of previous stud
-
ies examining potential sex differences in peer rela
-
tionship experiences and adjustment (Burks et al.,
1995; Oldenburg & Kerns, 1997; Parker & Asher,
1993). The only one of these studies using participants
similar in age to those in our investigation and assess
-
ing friendship also found no sex effects in analyses ex
-
ploring the connection between peer variables and ad
-
justment (Parker & Asher, 1993). There has been some
speculation that friendship experiences may be more
important for the adjustment of girls than boys (Burks
et al., 1995). The peer groups of boys tend to be more
global and group based, whereas those of girls tend to
include more intimate dyadic or small-group friend
-
ships (Eder & Hallinan, 1978). Hence, girls may be
more apt to rely on alternative types of relationships,
such as close friendships, to provide them with added
protection from the negative consequences of peer re
-
jection. Burks et al. used this reasoning to explain their
finding that the extent of peer rejection experienced
over a 2-year period predicted internalizing difficulties
at 3- and 6-year follow-up assessments only for boys.
Our results, along with those of Parker and Asher
(1993), suggest that such sex differences in the ways
that peer variables impact internalizing difficulties are
not evidenced in younger children. Perhaps with time
these sex effects would emerge. Given the cross-sec-
tional longitudinal design of the Burks et al. (1995)
study, the children were older than the participants in
our study at the time of the 3-year follow-up assess-
ment and in early to middle adolescence by the 6-year
follow-up. Again, the need for more research devoted
to the investigation of potential developmental differ-
ences is underscored.
Before discussing the applied implications of our re
-
sults, we must acknowledge some of the study’s limita
-
tions. Though the fit of our model was excellent, a larger
sample size would have enhanced our ability to detect
more subtle relations among the variables. For example,
even though stratified analyses suggested that there
were no meaningful differences between boys and girls
in the overall model, a replication with a larger sample
may find more subtle or specific sex-based discrepan
-
cies in the relations among these variables.
There were also limitations posed by our popularity
and friendship measurements. Regarding the use of
peer ratings, acceptance and rejection are distinct di
-
mensions of popularity (Bukowski & Hoza, 1989).
Thus, the degree to which our findings regarding popu
-
larity as measured by peer acceptance ratings are appli
-
cable to the larger body of studies focusing on rejection
and adjustment (e.g., Asher et al., 1990; Boivin et al.,
1994; Burks et al., 1995) can be questioned. However,
unlike positive and negative nominations that assess
acceptance and rejection respectively, rating scale
sociometrics are viewed by some as composite indexes
that reflect both the acceptance and rejection dimen
-
sions of the popularity construct (Bukowski & Hoza,
1989).
As a closely related measurement consideration, the
addition of negative nominations, particularly given
our interest in internalizing syndromes, would have en
-
abled us to examine different sociometric classifica
-
tions. Our good friend definition incorporated peer rat
-
ings. Some researchers have suggested that friendship
criteria be based on nomination data to avoid con
-
founds with popularity measures (Bukowski et al.,
1996; Parker & Asher, 1993). Nonetheless, given the
shared basis in liking, the ability to avoid such con
-
founds seems limited and, perhaps, conceptually un
-
warranted. Moreover, the degree to which nominations
and high ratings differ functionally is questionable
(Bukowski et al., 1996; Erdley et al., 1998). At any
rate, the fact that our good friend operationalization
may have affected the formation of a direct path be
-
tween popularity and loneliness is worth considering.
Perhaps using unlimited friendship nominations and
allowing children to circle their “very best” friends
would have been a better alternative (A. M. La Greca,
personal communication, December 5, 2002). Finally,
an important consideration is the direction of effects.
The flow of effects is likely to be more bidirectional
than is suggested in our model. Depression may inten-
sify feelings of loneliness and adversely impact accep-
tance and friendships. We did perform additional anal-
yses examining multiple directions, and none
suggested a backward pathway between depression
and friendship. However, support was found for a pos-
sible pathway from loneliness to friendship.
Clinically, our results suggest that intervention ef-
forts might be better directed at developing and im
-
proving dyadic friendships than enhancing children’s
overall peer acceptance. In addition to the limited abil
-
ity of most social skills training interventions to alter
peer group status (Asher et al., 1996), our results indi
-
cate that, even when successful, these treatments are
only likely to have an indirect effect on children’s so
-
cial adjustment through their impact on friendship.
When children become more popular, they are more
likely to form more friendships. More friendships of
relatively higher quality provide these now more popu
-
lar children with greater protection from loneliness,
which, in turn, lessens their overall risk of depression.
Attention to improving friendship quality is particu
-
larly important in light of findings showing that friend
-
ships that are higher in conflict might place children at
increased risk for negative outcomes (Berndt, 1996).
However, despite the demonstrated ability of even one
close friendship to compensate for the negative effects
of peer rejection (e.g., Parker & Asher, 1993), there
have been very few studies evaluating interventions
that seek to enhance children’s friendship development
(see Frankel, Myatt, & Cantwell, 1995; Mrug, Hoza, &
Gerdes, 2001, as example interventions).
553
FRIENDSHIP AND ADJUSTMENT
In closing, a cautionary note is in order. The relation
between friendship and adjustment is not always posi
-
tive. For example, friendships among certain children
may actually have deleterious effects. In systematic re
-
search, Dishion Andrews, and Crosby (1995) have
shown that associations between deviant adolescents
predict escalations in substance abuse, delinquency,
and violent behavior. Other researchers examining the
impact of friendship on externalizing difficulties have
found similar effects (e.g., Berndt & Keefe, 1995;
Hoza, Molina, Bukowski, & Sippola, 1995). An inter
-
esting possibility is that the same friendships that pro
-
tect some children from internalizing problems also
encourage escalations in antisocial behavior. Ulti
-
mately, determining the full impact of friendship on
adjustment will be a very complex undertaking incor
-
porating measures of who is included in children’s
friendship networks (Hartup, 1996), friendship quan
-
tity and quality, and multiple adjustment indexes (e.g.,
internalizing, externalizing). All in all, it is likely that
such research will discover multiple and variant path
-
ways linking children’s friendship experiences to ad
-
justment.
References
Asher, S. R., Parker, J. G., & Walker, D. L. (1996). Distinguishing
friendship from acceptance: Implications for intervention and
assessment. In W. M. Bukowski, A. F. Newcomb, & W. W.
Hartup (Eds.), The company they keep: Friendship during
childhood and adolescence (pp. 366–405). New York: Cam-
bridge University Press.
Asher, S. R., Parkhurst, J. T., Hymel, S., & Williams, G. A. (1990).
Peer rejection and loneliness in childhood. In S. R. Asher & J.
D. Coie (Eds.), Peer rejection in childhood (pp. 253–273). New
York: Cambridge University Press.
Asher, S. R., & Wheeler, V. A. (1985). Children’s loneliness: A com
-
parison of rejected and neglected peer status. Journal of Con
-
sulting and Clinical Psychology, 53, 500–505.
Bagwell, C. L., Newcomb, A. F., & Bukowski, W. M. (1998).
Preadolescent friendship and peer rejection as predictors of
adult adjustment. Child Development, 69, 140–153.
Berndt, T. J. (1981). Age changes and changes over time in prosocial
intention and behavior between friends. Developmental Psy
-
chology, 17, 408–416.
Berndt, T. J. (1996). Exploring the effects of friendship quality on so
-
cial development. In W. M. Bukowski, A. F. Newcomb, & W.
W. Hartup (Eds.), The company they keep: Friendship in child
-
hood and adolescence (pp. 346–365). New York: Cambridge
University Press.
Berndt, T. J., & Keefe, K. (1995). Friends’ influence on adolescents’
adjustment to school. Child Development, 66, 1312–1329.
Berndt, T. J., & Perry, T. B. (1986). Children’s perceptions of friend
-
ships as supportive relationships. Developmental Psychology,
22, 640–648.
Biddle, B. J., & Marlin, M. J. (1987). Causality, confirmation, credu
-
lity, and structural equation modeling. Child Development, 58,
4–17.
Boivin, M., Hymel, S., & Bukowski, W. M. (1995). The roles of so
-
cial withdrawal, peer rejection, and victimization by peers in
predicting loneliness and depression. Development and
Psychopathology, 7, 765–785.
Boivin, M., Poulin, F., & Vitaro, F. (1994). Depressed mood and peer
rejection in childhood. Development and Psychopathology, 6,
483–498.
Buhrmester, D., & Furman, W. (1986). The changing functions of
friends in childhood: A neo-Sullivanian perspective. In V. G.
Derlega & B. A. Winstead (Eds.), Friendship and social inter
-
action (pp. 41–62). New York: Springer-Verlag.
Bukowski, W. M., & Hoza, B. (1989). Popularity and friendship: Is
-
sues in theory, measurement, and outcome. In T. J. Berndt & G.
W. Ladd (Eds.), Peer relationships in child development (pp.
15– 45). New York: Wiley.
Bukowski, W. M., Hoza, B., & Boivin, M. (1993). Popularity,
friendship, and emotional adjustment during adolescence. In
B. Laursen (Vol. Ed.) & W. Damon (Series Ed.), New direc
-
tions for child development: No. 60. Close friendships in ado
-
lescence (pp. 23–37). San Francisco: Jossey-Bass.
Bukowksi, W. M., Pizzamiglio, M. T., Newcomb, A. F., & Hoza, B.
(1996). Popularity as an affordance of friendship: The link be
-
tween group and dyadic experience. Social Development, 5,
189–202.
Burks, V. S., Dodge, K. A., & Price, J. M. (1995). Models of internal
-
izing outcomes of early rejection. Development and
Psychopathology, 7, 683–695.
Carey, M. P., Gresham, F. M., Ruggiero, L., Faulstich, M. E., &
Enyart, P. (1987). Children’s Depression Inventory: Construct
and discriminant validity across clinical and nonreferred (con
-
trol) populations. Journal of Consulting and Clinical Psychol
-
ogy, 55, 755–761.
Dishion, T. J., Andrews, D. W., & Crosby, L. (1995). Antisocial boys
and their friends in early adolescence: Relationship characteris-
tics, quality, and interactional process. Child Development, 63,
538–548.
Eder, D., & Hallinan, M. T. (1978). Sex differences in children’s
friendship. American Sociological Review, 43, 237–250.
Erdley, C. A., Nangle, D. W., & Gold, J. A. (1998). Operationalizing
the construct of friendship among children: A psychometric
comparison of sociometric-based definitional methodologies.
Social Development, 7, 62–71.
Frankel, F., Myatt, R., & Cantwell, D. P. (1995). Training outpatient
boys to conform with the social ecology of popular peers: Ef
-
fects on parent and teacher ratings. Journal of Clinical Child
Psychology, 24, 300–310.
Furman, W., & Robbins, P. (1985). What’s the point? Issues in the
selection of treatment objectives. In B. H. Schneider, K. H.
Rubin, & J. E. Ledingham (Eds.), Children’s peer relations:
Issues in assessment and intervention (pp. 41–54). New York:
Springer-Verlag.
Hartup, W. W. (1996). The company they keep: Friendships and
their developmental significance. Child Development, 67,
1–13.
Hoza, B., Bukowski, W. M., & Beery, S. (2000). Assessing peer net
-
work and dyadic loneliness. Journal of Clinical Child Psychol
-
ogy, 29, 119–128.
Hoza, B., Molina, B. S. G., Bukowski, W. M., & Sippola, L. K.
(1995). Peer variables as predictors of later childhood adjust
-
ment. Development and Psychopathology, 7, 787–802.
Kovacs, M. (1985). The Children’s Depression Inventory.
Psychopharmacology Bulletin, 21, 995–998.
Ladd, G. W., Kochenderfer, B. J., & Coleman, C. C. (1997). Class
-
room peer acceptance, friendship, and victimization: Distinct
relational systems that contribute uniquely to children’s school
adjustment? Child Development, 68, 1181–1197.
Mrug, S., Hoza, B., & Gerdes, A. C. (2001). Children with atten
-
tion-deficit/hyperactivity disorder: Peer relationships and
peer-oriented interventions. In D. W. Nangle & C. A. Erdley
(Eds.), New directions for child and adolescent development:
The role of friendship in psychological adjustment (pp. 51–78).
San Francisco: Jossey-Bass.
554
NANGLE ET AL.
Nangle, D. W., Erdley, C. A., & Gold, J. A. (1996). A reflection on
the popularity construct: The importance of who likes and dis
-
likes a child. Behavior Therapy, 27, 337–352.
Newcomb, A. F., Bukowski, W. M., & Pattee, L. (1993). Children’s
peer relations: A meta-analytic review of popular, rejected, ne
-
glected, controversial, and average sociometric status. Psycho
-
logical Bulletin, 113, 99–128.
Oldenburg, C. M., & Kerns, K. A. (1997). Associations between peer
relationships and depressive symptoms: Testing moderator ef
-
fects of gender and age. Journal of Early Adolescence, 17,
319–337.
Parker, J. G., & Asher, S. R. (1993). Friendship and friendship qual
-
ity in middle childhood: Links with peer group acceptance and
feelings of loneliness and social dissatisfaction. Developmental
Psychology, 29, 611–621.
Price, J. M., & Ladd, G. W. (1986). Assessment of children’s friend
-
ships: Implications for social competence and social adjust
-
ment. In R. J. Prinz (Ed.), Advances in behavioral assessment of
children and families (Vol. 2, pp. 121–149). Greenwich, CT:
JAI.
Sanderson, J. A., & Siegal, M. (1995). Loneliness and stable friend
-
ship in rejected and nonrejected preschoolers. Journal of Ap
-
plied Developmental Psychology, 16, 555–567.
Smucker, M. R., Craighead, W. E., Craighead, L. W., & Green, B. J.
(1986). Normative and reliability data for the children’s depres
-
sion inventory. Journal of Abnormal Child Psychology, 14,
25–39.
Sullivan, H. S. (1953). The interpersonal theory of psychiatry. New
York: Norton.
Received October 15, 2002
Accepted June 13, 2003
555
FRIENDSHIP AND ADJUSTMENT
