Child Development, July/August 2006, Volume 77, Number 4, Pages 822 – 846
Peer Rejection, Aggressive or Withdrawn Behavior, and Psychological
Maladjustment from Ages 5 to 12: An Examination of Four Predictive Models
Gary W. Ladd
Arizona State University
Findings yielded a comprehensive portrait of the predictive relations among children’s aggressive or withdrawn
behaviors, peer rejection, and psychological maladjustment across the 5 – 12 age period. Examination of peer rejection in different variable contexts and across repeated intervals throughout childhood revealed differences in the
timing, strength, and consistency of this risk factor as a distinct (additive) predictor of externalizing versus internalizing problems. In conjunction with aggressive behavior, peer rejection proved to be a stronger additive predictor
of externalizing problems during early rather than later childhood. Relative to withdrawn behavior, rejection’s efficacy as a distinct predictor of internalizing problems was significant early in childhood and increased progressively thereafter. These additive path models fit the data better than did disorder-driven or transactional models.
The assumption that children’s early behavioral
propensities or their peer experiences engendered
psychological maladjustment gained prominence
during the 1950s and 1960s (cf. Freud & Dann, 1951;
Morris, Soroker, & Burruss, 1954; Robins, 1966; Roff,
1961, 1963), and fostered separate empirical traditions that were devoted to explicating each of these
constellations of variables as risk factors for earlyand later-emerging dysfunction (for reviews, see
Kupersmidt, Coie, & Dodge, 1990; Ladd, 2005;
McDougall, Hymel, Vaillancourt, & Mercer, 2001;
Parker & Asher, 1987). From the outset, researchers
utilized longitudinal studiesFthe most valuable of
which were based on follow-up or follow-through
(prospective) designs (see Parker & Asher, 1987)Fto
discern which facets of children’s behavior, or their
peer relationships, were most predictive of lateremerging dysfunctions (e.g., see Roff, Sells, &
Golden, 1972; Roff & Wirt, 1984).
Children’s Behavior Versus Peer Relations: Main Effect
Predictive Models
Within one empirical tradition, investigators
worked from the hypothesis that children’s behavPortions of this study were conducted as part of the Pathways
Project, a larger longitudinal investigation of children’s social/
psychological/scholastic adjustment in school contexts that is
supported by the National Institutes of Health (1 RO1MH-49223,
2-RO1MH-49223, R01HD-045906 to Gary W. Ladd). Special appreciation is expressed to all the children and parents who made
this study possible, and to members of the Pathways Project for
assistance with data collection.
Correspondence concerning this article should be addressed to
Gary Ladd, P.O. Box 852502, Arizona State University, Tempe, AZ
85287-2502. Electronic mail may be sent to [email protected].
ioral propensities were the principal determinants of
psychopathology, and a sizable proportion of this
research was designed to elucidate the predictive
contributions of aggressive and withdrawn behavior
to later dysfunction. An influential premise was that
children’s propensity to engage in confrontive forms
of aggression became an enduring behavioral orientation that eventually brought about dysfunction
(see Caspi, Elder, & Bem, 1987; Moss & Susman,
1980; Olweus, 1979). Early and recent longitudinal
findings corroborated this premise by showing that
aggressive behavior, even at early ages (e.g., see
Tremblay, Pihl, Vitaro, & Dobkin, 1994), was fairly
stable for boys and girls (see Cairns, Cairns, Neckerman, Ferguson, & Gariepy, 1989; Olweus, 1979)
and moderately predictive of maladjustment, particularly externalizing problems (e.g., Cairns &
Cairns, 1994; Caspi et al., 1987). Another influential
premise was that children’s propensity to engage in
withdrawn behavior led to maladjustment. Although early findings suggested that withdrawn
behavior was not particularly stable or predictive of
dysfunction (see Michael, Morris, & Soroker, 1957;
Morris et al., 1954; Wanlass & Prinz, 1982), later
longitudinal evidence revealed that certain subtypes
of this behavior (e.g., anxious – depressed, anxious –
solitary styles) were moderately stable (Gazelle &
Ladd, 2003; Harrist, Zaia, Bates, Dodge, & Pettit,
1997) and modestly predictive of internalizing
problems (e.g., Caspi, Elder, & Bem, 1988; Gazelle &
Ladd, 2003).
Investigators within another research tradition
were guided by the hypothesis that children’s
r 2006 by the Society for Research in Child Development, Inc.
All rights reserved. 0009-3920/2006/7704-0002
Predicting Maladjustment
exposure to adverse peer relations (e.g., peer group
rejection, low peer acceptance) was a fundamental
determinant of later maladjustment. Initial efforts to
pursue this research agenda were influenced by
ideological perspectives in which children’s development was seen as malleable and largely influenced
by their social relationships, including those with
age-mates (e.g., see Hartup, 1970; Sullivan, 1953).
This premise was strengthened by experimental evidence indicating that age-mates influenced children’s (and some primate’s) socialization, and that
aberrant peer socialization predicted later-emerging
psychopathology (e.g., Freud & Dann, 1951; Furman,
Rahe, & Hartup, 1979; Harlow, Harlow, Dodsworth,
& Arling, 1966; Suomi & Harlow, 1972). Past and
recent efforts to elucidate the contributions of children’s peer relationships to the development of
psychopathology have consistently implicated poor
peer group relations, and peer group rejection in
particular, as a precursor of externalizing and internalizing problems during the grade school years and
thereafter (e.g., DeRosier, Kupersmidt, & Patterson,
1994; Hymel, Rubin, Rowden, & LeMare, 1990;
Kupersmidt & Coie, 1990; Ladd & Troop-Gordon,
2003). Moreover, it was discovered that the predictive link between children’s peer acceptance/rejection and subsequent psychological adjustment was
robust even when it was evaluated in conjunction
with other types of peer relationships (e.g., friendship, friendlessness, peer victimization; see Bagwell,
Newcomb, & Bukowski, 1998; Ladd, Herald, & Andrews, 2006; Ladd & Troop-Gordon, 2003).
One consequence of this bifurcation in investigative paradigms was that two rather distinct empirical
literatures developed around these premises. Especially during the first generation of longitudinal
studies, researchersFdepending on their theoretical
orientationsFtended to favor one of the aforementioned premises over the other and, thus, investigated the predictors of psychological maladjustment
from a ‘‘main effects’’ perspective. As might be surmised, the evidence that accrued within these two
bodies of evidence tended to implicate either aggression or peer group rejection as a principal risk
factor for later dysfunction.
Children’s Behavior and Peer Relations: Child and
Environment Predictive Models
As the limitations of main effects perspectives
became apparent, investigators began to devise more
complex prediction models in which multiple risk
factors were recognized as potential determinants of
dysfunction (see Coie et al., 1993). These models
823
were based on the premise that specific child behaviors (e.g., aggression, withdrawal) and adverse
peer experiences (e.g., peer rejection) co-determine
children’s future psychological adjustment (see
Ladd, 1989, 2003; McDougall et al., 2001; Parker,
Rubin, Price, & DeRosier, 1995). This paradigm shift
was first evident in a critical review published by
Parker and Asher (1987) in which the authors encouraged researchers to evaluate competing explanationsFframed as ‘‘causal’’ versus ‘‘incidental’’
modelsFof how children’s behavior and peer relationships contributed to their future adjustment. The
incidental model was essentially a restatement of the
hypothesis that children’s preexisting characteristics,
including their behavioral dispositions, were the
principal causes of later maladaptive outcomes. It
was assumed that (a) premorbid forms of a lateremerging disorder, children’s behavioral propensities, or both were the principal precursors of later
maladjustment and (b) these same child attributes
caused children to develop poor peer relationshipsFthat is, children’s peer relational difficulties
were ‘‘incidental’’ consequences (i.e., markers) of
their behavioral dispositions, and played no causal
role in the development of later dysfunction. In
contrast, the main assumption of the ‘‘causal’’ model
was that, along with children’s behavioral dispositions, participation in adverse peer relationships (i.e.,
exposure to peer group rejection) was influential in
shaping children’s later adjustment.
In part, this proposal was a forerunner of contemporary ‘‘child and environment’’ frameworks.
Although several variants of child and environment
models have been proposed to elucidate the distinct
versus conjoint predictive contributions of children’s
behavior and peer relationships to later psychological maladjustment (see Ladd, 2003), most of the evidence amassed thus far conforms to an additive
model. Additive models imply that, separate (partially independent) from the contributions of children’s behavior, the experiences they have in peer
relationships increase or decrease the probability of
maladjustment.
Additive Child by Environment Prediction Models:
Principal Premises and Extant Evidence
The hypothesis that aggression and peer group
rejection additively predict later maladjustment has
been based on assumptions similar to those depicted
in the following exemplary model: Children who
frequently engage in confrontive aggression among
peersFwhether because of inheritance, learning, or
some transaction between the two (see Caspi et al.,
824
Ladd
1987)Fdevelop a style of responding to interpersonal contingencies that is eventually generalized to
a larger range of environmental contingencies (e.g.,
constraints that require delay of gratification; prohibitions or rules for behavior, etc.). Over time, this
progression causes children to develop broader and
more serious forms of maladjustment, such as externalizing problems. Further, because children who
possess aggressive behavioral orientations often alienate peers, they are exposed to another (relational)
risk factorFpeer group rejection. Once in a rejecting
peer milieu, children are exposed to debilitating relational processes (e.g., maltreatment by peers, exclusion from peer activities, isolation from peer
affirmation or support, etc.; see Asher, Rose, &
Gabriel, 2001; Buhs & Ladd, 2001; Coie, 1990, 2004)
that take an additional toll on their future adjustment
(i.e., these stressors contribute to maladjustment in
ways that are distinct from aggressive behavior).
Consistent with this model, there is evidence indicating that aggression and peer rejection during
childhood make additive contributions to the prediction of future externalizing problems. For example, findings from recent longitudinal studies have
shown that aggression and peer rejection measured
in grade school made distinct contributions to the
prediction of externalizing problems later in grade
school and during preadolescence and adolescence
(Coie, Lochman, Terry, & Hyman, 1992; Cowan &
Cowan, 2004; Hymel et al., 1990).
The hypothesis that withdrawn behavioral styles
and peer group rejection additively predict later
maladjustment has been less well investigated. One
of the principal assumptions guiding this line of investigation can be portrayed as follows: Children
who manifest a withdrawn behavioral styleF
whether rooted in temperament, socialization, or the
interaction of the two (see Rubin & Asendorpf,
1993)Fare prone to respond to social contingencies
by seeking solitude, avoiding contact, or interacting
infrequently with peers. Because withdrawn behavior may be interpreted by peers as disinterest or a
lack of social responsiveness, it may foster adverse
interpersonal reactions such as disliking and rejection. In turn, exposure to peer rejection increases the
probability that withdrawn children will develop
internalizing problems because the stressors associated with this social position cause children to feel
anxious or distressed and, thus, instigate and maintain internalizing problems (Gazelle & Ladd, 2003;
Rubin, LeMare, & Lollis, 1990).
Although sparse, corroborating evidence for this
model has begun to accrue. Findings from a 1-year
longitudinal study conducted by Renshaw and
Brown (1993) revealed that both withdrawn behavior
and low peer group acceptance made distinct predictive contributions to children’s loneliness over a
1-year interval. Similar linkages were reported by
Boivin and Hymel (1997) for data gathered concurrently rather than longitudinally. In a 5-year prospective longitudinal study, Gazelle and Ladd (2003)
examined how withdrawn behavior and peer group
exclusion (one manifestation of peer group rejection)
were linked with children’s trajectories toward depression. They found that children who were prone
toward anxious solitary behavior during the early
school years were more likely to manifest and
maintain depressive symptoms if they also had been
subjected to higher levels of peer exclusion.
In sum, there is a modicum of evidence to suggest
that specific behavioral orientations (i.e., aggressive,
withdrawn) and peer group rejection during childhood constitute related, but not entirely redundant,
risk factors for later psychological dysfunction. At
this juncture, the overarching pattern of findings
suggests that, when aggression and rejection are
evaluated conjointly as predictors of later externalizing problems, aggression typically emerges as the
stronger of the two risk factors or tends to be the only
predictor to account for significant nonshared variance in externalizing problems (for reviews, see Coie,
2004; Ladd, 2003; McDougall et al., 2001). When
these same two risk factors are evaluated as predictors of internalizing problems, the additive contribution of peer group rejection (relative to aggression)
tends to be larger (McDougall et al., 2001). In contrast, the hypothesis that withdrawn behavioral
styles and peer group rejection contribute additively
to later maladjustment has been less well investigated. At present, there is preliminary evidence to
suggest that children disposed toward withdrawn
behavioral patterns are at risk for later internalizing
problems (e.g., Boivin, Hymel, & Bukowski, 1995;
Caspi et al., 1988; Ladd & Troop-Gordon, 2003) and
that exposure to peer group rejection additively enhances the prediction of these problems (see Ladd,
2003; McDougall et al., 2001). Next to nothing is
known about whether children who are disposed
toward withdrawn behavior are at risk for externalizing problems and whether peer group rejection
exacerbates this risk (see Ladd, 2003).
Clearly, much remains to be learned about how
peer group rejection, in conjunction with children’s
aggressive or withdrawn behavioral styles, contributes to the prediction of childhood externalizing
and internalizing problems. Accordingly, this investigation was undertaken to address two principal
aims.
Predicting Maladjustment
Aim 1: Conduct a More Comprehensive Analysis of Peer
Rejection as an Additive Risk Factor
At present, knowledge remains limited with respect to when, to what extent, and how consistently peer
group rejection adds to (beyond aggressive or withdrawn behavioral styles) the prediction of externalizing or internalizing problems during the grade
school years (i.e., from age 5 – 12). In large part, this
void stems from an overreliance on nonpanel, singleindicator, follow-up longitudinal designs. That is, in
studies of additive risk factors, investigators have
tended to assess single-indicator predictors at one
time point (e.g., childhood) and use this information
to forecast single-indicator criteria at one later time
point (e.g., preadolescence or adolescence). Designs
such as these do not enable investigators to assess
patterns of continuity or change in predictor – criteria
relations across successive intervals in children’s
development (e.g., age, grade levels), evaluate or
control for the stability of predictors or criteria when
examining predictive associations (e.g., controlling
for current predictor – criteria associations when examining antecedent predictor-criteria linkages),
draw clear distinctions between constructs, or maximize the psychometric properties of their measures
(e.g., reliability, construct invariance over time, etc.).
Certain design problems must be avoided as well.
There is a need to both distinguish among (rather
than confound) the predictive contributions of behavioral and relational risk factors, and ensure that
risk factors are not imbued with information about
the to-be-predicted maladjustment criteria. To illustrate, when investigators attempt to estimate adjustment consequences for certain ‘‘subtypes’’ of
children, such as those who manifest multiple behavioral and relational risks (e.g., aggressive – rejected children) or those who exhibit a behavioral
risk and symptoms of maladjustment (e.g., anxious –
withdrawn, depressed – withdrawn children), they
essentially confound (likely overestimate) the predictive contributions of specific risk factors.
To address these limitations, a full-panel, followthrough longitudinal design was implemented in
this investigation. Multiple indicators of each risk
factor and maladjustment criterion were obtained
yearly across a 7-year period, and used to formulate
latent variables for each construct at each time of
assessment. This measurement model was utilized
so that all of the risk and maladjustment constructs
were operationalized as separate latent variables
(rather than confounding them, as occurs in some
risk-subtype designs) and the hypothesized predictor – criteria linkages could be examined within
825
latent stability models across six (lagged) 1-year
intervals. One set of specific aims was to determine
(1) whether, and when (i.e., how early) during the
grade school years, exposure to peer group rejection
forecasts later externalizing or internalizing problems, beyond that which can be predicted by children’s aggressive or withdrawn behavioral
propensities, and (2) whether the additive contribution of peer group rejection to the prediction of externalizing or internalizing problems (relative to
behavioral styles) remains constant versus grows
stronger or weaker across the grade school years
(determine whether there is continuity vs. discontinuity in its additive, predictive contributions). Illustrations of the type of additive, latent stability
models that was used to address these objectives are
depicted in Figure 1. The scientific merit of the
findings was expected to be substantial because
neither of these aims has previously been examined
with this age group, and the results would reflect on
past and contemporary assumptions about (a) when,
during childhood, peer group rejection (and its attendant stressors) creates adjustment consequences
that are distinct from those attributable to children’s
behavioral styles and (b) the strength and continuity
with which this form of peer adversity predisposes
children to externalizing versus internalizing forms
of dysfunction across early, middle, and later childhood.
It has been proposed that attaining acceptance in
peer groups is a critical social task for children as
they enter grade school and progress through the
primary grades, and that failure to do so (i.e., peer
group rejection) creates sustained risk for psychological difficulties. It is during the early grade school
years that children are challenged to establish
themselves (become accepted) as members of large,
age-segregated peer groups (Ladd, 1989, 1996). Furthermore, in addition to being a key psychosocial
task, acceptance into peer society can be seen as a
powerful social need that emerges during the 6 to 9year age period (Buhrmester & Furman, 1986). On
the basis of these rationales, it was hypothesized that
peer group rejection would additively contribute to
the prediction of maladjustment as early as kindergarten (children’s 1st year in grade school), and that
continuity would be apparent in this additive linkage
across the primary school years (childhood; ages 5 –
9) and, to a lesser extent, across the later grade school
years (preadolescence; ages 10 – 12). Moreover, consonant with extant empirical evidence, it was anticipated that the strength and continuity of these
additive contributions would be greater for internalizing than for externalizing problems.
826
Ladd
Kindergarten
Grade 1
Grade 2
Grade 3
Grade 4
Grade 5
Grade 6
AG
AG
AG
AG
AG
AG
AG
REJ
REJ
REJ
REJ
REJ
REJ
REJ
EXT
EXT
EXT
EXT
EXT
EXT
EXT
Grade 1
Grade 2
Grade 3
Grade 4
Grade 5
Kindergarten
Grade 6
WD
WD
WD
WD
WD
WD
WD
REJ
REJ
REJ
REJ
REJ
REJ
REJ
INT
INT
INT
INT
INT
INT
INT
Figure 1. The upper and lower panels of this figure illustrate the types of additive prediction models that were evaluated in this investigation. The model shown in the upper panel depicts aggressive behavior and peer rejection as additive predictors of externalizing
problems (AG1REJ model) across each of six targeted time lags. The model shown in the lower panel depicts withdrawn behavior and
peer rejection as additive predictors of internalizing problems (WD1REJ model) across each of six targeted time lags. AG 5 aggressive
behavior; WD 5 withdrawn behavior; REJ 5 peer rejection.
Aim 2: Evaluate Alternatives to Additive Prediction
Models
Knowledge may also be incomplete because it has
not been a priority to validate the prevailing predictive paradigm (e.g., ‘‘additive risk models’’)
against alternative models that stem from different
causal premises. A predominant assumption within
this area of inquiry has been that children’s behavior
and/or peer group rejection are determinants of
psychological maladjustment. This theoretical hegemony has dissuaded investigators from considering, or empirically testing, models that imply
alternative directions of prediction. Two such models
were identified and targeted for investigation. First,
it is conceivable that premorbid forms of a lateremerging disorder may predict children’s behavioral
propensities, peer group rejection, or both (see
Parker & Asher, 1987). Second, it is possible that the
causal relations that exist among these variables may
not be temporally invariant, but rather shift as a
function of age-related changes in organismic factors
and environmental circumstances (see Sameroff,
1987; Sameroff & Chandler, 1975). Were this the case,
permutations in the predictive relations among
children’s behavioral dispositions, classroom peer
relations, and psychological adjustment would be
expected to emerge over time, and these patterns
would be better described by transactional rather
than unidirectional models (see Parker et al., 1995).
Accordingly, a second set of specific aims was to
determine whether there was empirical support for
these alternate models, which for ease of reference
were termed ‘‘disorder-driven’’ and ‘‘transactional’’
prediction models, respectively.
The central premise of the disorder-driven model
is that early-emerging dysfunctionsFin this case,
incipient externalizing or internalizing problemsFcause children to act in aggressive or withdrawn ways and to alienate their classmates. For
example, in classroom contexts, it is likely that children who manifest externalizing problems (a general
pattern of emotional/behavioral undercontrol; e.g.,
angry, explosive, impulsive, distractible, overactive
behaviors, etc.) will devise forceful tactics or react
aggressively when confronted with peer problems or
Predicting Maladjustment
interpersonal constraints. Moreover, children who
under-regulate their behavior or emotions may, due
to their impulsiveness and insensitivity toward
others, irritate or alienate peers and, thus, become
disliked and rejected by peers (see Figure 2, upper
panel). Likewise, children’s early internalizing
problems may cause them to act in an anxious –
withdrawn manner among peers (emit anxious cues,
avoid peers), and their lack of social responsiveness
(e.g., reservedness, diffidence, melancholy) may engender peer dislike or rejection (see Figure 2, lower
panel). However, because externalizing problems
and peer aggression may stem from common underpinnings and therefore partially overlap as constructs (i.e., externalizing disorder encompasses
some forms of aggression, particularly confrontive,
angry, explosive, reactive subtypes), empirical support for disorder-driven models was expected to be
stronger for children disposed toward early externalizing rather than internalizing problems.
For purposes of this investigation, the concept of a
transactional model was based on the principle that,
temporally, the relations that transpire between risk
Kindergarten
Grade 1
factors and dysfunction are dynamic rather than
static and unidirectionalFthat is, subject to ‘‘turnabouts’’ or reversals in causal priority or directions of
effect (Rutter, 1996). The two transactional predictive
models evaluated in this investigation were based on
the premise that risk factors and maladjustment influence each other cyclically over time (Ladd, 1989).
The first of these models constitutes a synthesis (sequential paring) of the ‘‘additive risk’’ and ‘‘disorder-driven’’ models, respectively (see Figure 3,
upper panel). To illustrate, children who initially
succeed in addressing classroom social challenges
with aggression and are repeatedly subjected to
peers’ rejecting behaviors may generalize their use of
force toward broader social contingencies and become desensitized to various forms of social censure
(i.e., develop pervasive, more serious forms of externalizing problems). In turn, movement into this
form of dysfunction (i.e., acting/reacting impulsively, pervasively violating rules and norms, minimizing or disregarding broader social consequences)
may become the proximal cause for later aggression
(an increasing reliance on the use of force) and peer
Grade 2
Grade 3
Grade 4
Grade 5
Grade 6
AG
AG
AG
AG
AG
AG
AG
REJ
REJ
REJ
REJ
REJ
REJ
REJ
EXT
EXT
EXT
EXT
EXT
EXT
EXT
Grade 1
Grade 2
Grade 3
Grade 4
Grade 5
Kindergarten
827
Grade 6
WD
WD
WD
WD
WD
WD
WD
REJ
REJ
REJ
REJ
REJ
REJ
REJ
INT
INT
INT
INT
INT
INT
INT
Figure 2. The upper and lower panels of this figure provide illustrations of the disorder-driven models that were evaluated in this investigation. The model shown in the upper panel depicts externalizing problems as a predictor of aggression and rejection across each of
six targeted time lags. The model shown in the lower panel depicts internalizing problems as a predictor of withdrawn behavior and peer
rejection across each of six targeted time lags.
828
Ladd
Kindergarten
Grade 1
Grade 2
Grade 3
Grade 4
Grade 5
Grade 6
AG
AG
AG
AG
AG
AG
AG
REJ
REJ
REJ
REJ
REJ
REJ
REJ
EXT
EXT
EXT
EXT
EXT
EXT
EXT
Grade 1
Grade 2
Grade 3
Grade 4
Grade 5
Kindergarten
Grade 6
WD
WD
WD
WD
WD
WD
WD
REJ
REJ
REJ
REJ
REJ
REJ
REJ
INT
INT
INT
INT
INT
INT
INT
Figure 3. The upper and lower panels of this figure provide illustrations of the transactional prediction models that were evaluated in this
investigation. The risk ! maladjustment ! risk model is depicted in the upper panel and the maladjustment ! risk ! maladjustment
model is shown in the lower panel.
group rejection (acting in ways that generate negative peer sentiments such as dislike, distrust, wariness, fearfulness, etc.; see Coie, 2004). Corroboration
for this ‘‘risk ! maladjustment ! risk’’ model
would be obtained if observed predictive linkages
conformed to a pattern in which children’s behavioral orientations and peer group rejection during
early grade school anteceded externalizing or internalizing problems during middle childhood and,
thereafter, the dysfunction forecasted children’s behavioral propensities and rejection by peers.
The second of these transactional models depicts a
cycle in which the direction of prediction for the
disorder-driven model precedes that specified for
the additive risk model (i.e., a ‘‘maladjustment ! risk ! maladjustment’’ model; see Figure 3,
lower panel). As an example, it may be the case that
children’s initial internalizing tendencies (e.g.,
emitting anxious social signals, being emotionally
unresponsive to peers) cause them to act in a withdrawn manner and be rejected by peers, but it is the
development of withdrawn behavior and the experience of being rejected (exposure to peer maltreatment, and the accompanying emotional scars or
debilitating cognitions children acquire from it; see
Pomerantz & Rudolph, 2003) that subsequently
drives the development of later internalizing problems (i.e., predictive relations are congruent with a
diathesis-stress interpretation; e.g., isolation from
peers and rejecting peer reactions alienate children
and confirm their anxieties about social rejection,
and thus perpetuate or increase their wariness,
fearfulness, propensity toward depression, etc.).
Support for this maladjustment ! risk ! maladjustment model would be obtained if evidence fit a
pattern in which early externalizing or internalizing
problems forecasted peer group rejection and/or
children’s behavioral orientations in middle childhood but, thereafter, one or both of these risk factors
became the principal antecedents of maladjustment
in later childhood.
Method
Participants
Participants were 399 children (206 boys; 193 girls)
who were recruited as they entered kindergarten
Predicting Maladjustment
(average age 5 5.62 years) and followed prospectively until they completed sixth grade (average
age 5 11.71 years). Children and their parents were
recruited at kindergarten preregistration meetings
that were held in school systems within multiple
urban, suburban, and rural U.S. locations. Written
informed parental consent was obtained before
children’s participation, and 95% of the recruited
families agreed to participate. The sample contained
children from different ethnic groups (77.4% European American, 17.3% African American, 5.3% Latina, mixed race, or other) and family socioeconomic
backgrounds (36.8% lower to middle income, range 5
$0 – 20,000; 30.6% middle income, range 5 $21,000 –
40,000; and 32.6% upper-middle to high income;
range 5 $41,000 and higher). At the time of the sixthgrade assessment, 96.49% of the children were still
active participants in the longitudinal study
(n 5 385). Data were also collected from children’s
teachers and classmates.
Over the course of this investigation, children
became increasingly dispersed across classrooms
and school districts. When children moved or changed schools, permission was sought from administrators, teachers, and parents to extend the project
into their schools. After these consents were obtained, the parents of project children’s classmates
were contacted and asked to provide consent for
their child’s participation. Only those classmates
who had written informed parental consent (and
child assent) participated in the study. The number
of teachers who participated in this study per assessment period ranged from 32 to 282, and the
number of classmates who contributed data per assessment ranged from 964 to 3,030.
Measures
Multiple indicators of the targeted constructs were
obtained. To increase the probability that these indicators could be used to form latent variables that
were invariant across measurement occasions, identical measures were administered at each time of
assessment. Indicators were measured with instruments that are known to yield reliable and valid data.
Behavioral Orientations
Confrontive aggression. Indicators of this construct
included two peer nomination measures (one for
verbal and another for physical aggression) and one
teacher report measure of children’s aggressive behavior toward peers. The reliability and validity of
all three measures have been substantiated (see Ladd
829
& Profilet, 1996). Peer nominations of verbal aggression (PN-VAG) and physical aggression (PNPAG) were obtained by presenting children with a
roster containing classmates’ names (those with
permission) and asking them to circle the names of
up to three classmates who could be described by
each of the following criteria, respectively (order of
administration counterbalanced): (1) ‘‘Someone who
teases, calls names, or makes fun of other kids’’ (i.e.,
verbal aggression toward peers) and (2) ‘‘Someone
who hits, kicks, or pushes other kids’’ (i.e., physical
aggression toward peers). For children in kindergarten to grade 2, a felt board containing classmates’
pictures was used instead of a roster and, during
individual interviews, respondents were asked to
designate (by pointing) up to three nominees. Before
administering these measures, children were presented with class rosters/picture boards, tested until
they could identify all classmates, and trained to
make nominations contingent on a series of practice
criteria. The third measure of aggression (CBS-AG)
was obtained by asking teachers to rate, on a 3-point
scale (1 5 doesn’t apply; 2 5 applies sometimes; 3 5 certainly applies), each of the seven-items included in the
Aggressive with Peers subscale of the Child Behavior
Scale (CBS; Ladd & Profilet, 1996). Items on this
subscale describe children’s use of verbal and
physical aggression against peers, and as across
testing occasions ranged from .90 to .93.
Withdrawn, asocial behavior. Two teacher-rating
measures served as indicators of this construct, including the Withdrawn Behavior subscale of the
Achenbach Teacher Report Form (TRF-WD; Achenbach, 1991) and the Asocial Behavior Subscale of the
CBS (CBS-AB; Ladd & Profilet, 1996). Items on the
TRF-WD and CBS-AB indexed children’s tendency
to engage in shy or solitary behaviors, avoid or
withdraw from peer activities, and abstain from social overtures (e.g., talking to peers). One item from
the TRF-WD subscale (i.e., ‘‘sad’’) was eliminated
due to its semantic similarity with indicators selected
to tap internalizing problems (anxiety/depression,
see below). The remaining eight TRF-WD items (e.g.,
withdrawn, shy, rather be alone) were rated on a 3point scale (e.g., 0 5 not true; 1 5 sometimes true;
2 5 often true), as were the six items included in the
CBS-AB subscale (e.g., solitary child; keeps peers at a
distance, likes to play alone; 1 5 doesn’t apply;
2 5 applies sometimes; 3 5 certainly applies). Internal
consistency was acceptable across measurement occasions for both measures (range 5 .90 – .93 and .87 –
.92, respectively).
Although it was of interest to obtain indicators of
this construct from the same informants used to tap
830
Ladd
aggressive behavior (i.e., peers and teachers), this
was infeasible because peers appear to lack the
necessary schema to distinguish and report this form
of behavior reliably until middle childhood (approximately 8 years of age; see Ladd & Mars, 1986;
Younger & Boyko, 1987; Younger, Gentile, & Burgess,
1993). However, for prediction purposes, problems
associated with shared-informant bias were reduced
because different teachers rated children on these
measures at each time of assessment. Further, as a
check against the validity of these indicators, scores
for the TRF-WD and CBS-AB were correlated with
scores from a peer nomination measure of children’s
withdrawn, asocial behavior (PN-ASOC) that were
obtained during grades 3 – 6. The procedures used to
obtain this index were the same as those described
for the PN-VAG and PN-PAG measures, except that
peers responded to the criterion ‘‘Someone who is
alone a lot.’’ Moderate convergence was found between the peer- and the teacher-report measures.
Correlations (all po.01) ranged from .37 to .53 between the PN-ASOC and TRF-WD, and from .30 to
.48 between the PN-ASOC and CBS-AB.
Peer Group Rejection
The three measures used as indicators of this
construct were obtained from two types of informants: peers and teachers. Peer reports of children’s
group acceptance/rejection were obtained with roster-and-rating and peer nomination sociometric
measures (see Asher & Dodge, 1986; Asher, Singleton, Tinsley, & Hymel, 1979). The roster-and-rating
measure (RR-SMS) was administered in each classroom by asking children (those with permission) to
rate each of their classmates on the following criterion: ‘‘How much do you like to play with this person in school?’’ At all grades, ratings were made on a
3-point scale, with 1 5 not much, 2 5 sometimes, and
3 5 a lot (for details, see Asher et al., 1979). For children in kindergarten to grade 2, a felt board containing classmates’ pictures was used instead of a
class roster, and respondents rated classmates during
individual interviews by sorting their pictures into
boxes marked (in text and graphics) with the same 3
scale points. A second measure of peer group rejection was acquired via a peer nomination procedure
(PN-NEG) in which children nominated up to three
classmates (i.e., ‘‘Kids who you don’t like to hang out
with [play with] at school’’) by circling names on a
roster (pointing to individual photos). The ratings
and the nominations children received from classmates were separately averaged and standardized
within classrooms. It has been established that the
RR-SMS and PN-NEG measures yield reliable and
valid data with both younger and older children (see
Asher et al., 1979; Cillessen & Bukowski, 2000; Ladd
et al., 2006). The third measure of peer group rejection was obtained by asking teachers to rate participants on the seven-item Excluded by Peers subscale
of the CBS (CBS-EP). Peer exclusion constitutes one
manifestation of peer group rejection (i.e., peers’ attempts to exclude children from participation in
classroom social activities; see Buhs & Ladd, 2001)
that can be reliably observed by teachers in classroom settings and that has been shown to correlate
substantially with peer report indices of peer group
rejection (see Buhs & Ladd, 2001; Ladd & Profilet,
1996).
Psychological Maladjustment
Externalizing problems. Indicators of this broadband construct consisted of three forms of child
misconductFclassroom disruptiveness, hyperactive – distractible behavior, and ‘‘delinquent’’ behaviorFall of which were measured at each time of
assessment. Collectively, these measures were seen
as indicators of children’s propensity to respond to a
broad range of classroom contingencies in an emotionally and behaviorally impulsive or undercontrolled manner. To acquire a measure of the first
indicator, classroom disruptiveness (CD), teachers
rated children on five-items taken from the Teacher
Rating Scale of School Adjustment (TRSSA; see
Ladd, Buhs, & Seid, 2000; e.g., ‘‘Has discipline
problems’’; ‘‘Is easy for teacher to manage (reverse
scored)’’; ‘‘Needs constant supervision’’), each of
which was accompanied by a 3-point scale
(1 5 doesn’t apply; 2 5 applies sometimes; 3 5 certainly
applies). To tap the second indicator, teachers rated
each of four items on the CBS Hyperactive – Distractible subscale (CBS-HD; e.g., ‘‘Restless, doesn’t
keep still’’; ‘‘Poor concentration’’) on a 3-point scale
(1 5 doesn’t apply; 2 5 applies sometimes; 3 5 certainly
applies). The nine-item TRF Delinquent Behavior
subscale (TRF-DB) was used to tap the third indicator. Teachers rated these items (‘‘Lying or cheating’’;
‘‘Steals’’) on a 3-point scale (e.g., 1 5 not true;
2 5 sometimes true; 3 5 very or often true). Scores for
each measure were created by averaging teachers’
ratings across the items within each subscale. as
calculated for these measures across assessment occasions fell within the following ranges: .79 – .84
(CD), .84 – .91(CBS-HD), and .68 – .78 (TRF-DB).
Internalizing problems. Indicators for this broadband construct included measures of anxiety and
depression (i.e., emotional/behavioral manifesta-
Predicting Maladjustment
tions of anxious and sad affect). Teachers rated each
child on 17 items from the TRF Anxious/Depressed
subscale (TRF-AD; one item was deleted due to
content that overlapped with peer group rejection)
using a 3-point scale (e.g., 0 5 not true; 1 5 sometimes
true; 2 5 very or often true). as for the TRF-AD ranged
from .82 to .89 across all times of assessment.
Teachers also rated children on the 4-item (e.g.,
‘‘Appears miserable, distressed’’; Fearful or afraid’’)
CBS Anxious Fearful subscale (CBS-AF) using a 3point scale (1 5 doesn’t apply; 2 5 applies sometimes;
3 5 certainly applies). Scores for the CBS-AF are
known to be reliable (see Ladd & Profilet, 1996; Ladd
& Troop-Gordon, 2003) and yielded as that ranged
from .69 to .78 with this sample.
Procedure
Peers and teachers completed each of the behavioral orientation, peer group rejection, and psychological maladjustment measures each year, for a
period of 7 consecutive years (from kindergarten to
grade 6). Trained examiners administered these
measures in children’s and teachers’ classrooms
during the spring of each school year (i.e., approximately March through May). Details about the exact
procedures used to administer each measure are reported in the Measures section. During each assessment interval, children were trained, and teachers
received instructions about how to use the rating
scales that accompanied the items contained in each
of their assigned measures. Upon completion of their
assigned measures, children and their classmates
were thanked and given a small gift (e.g., pencil,
stickers), and teachers received a cash honorarium.
Results
Data Analytic Strategy and Results of Preliminary
Analyses
Preliminary analyses were conducted to examine
variable distributions, relations, and patterns of
missing data, and to determine whether these and
other properties of the data set conformed to the
assumptions underlying parametric statistics and
data augmentation proceduresFspecifically, multiple imputation (e.g., values missing at random;
variables normally distributed, etc.; Rubin, 1987;
Schafer, 1999; Schafer & Olsen, 1998). Because it was
possible to satisfy these prerequisites, missing values
were replaced with estimated scores, and additional
preparatory analysis was undertaken. Multicollinearity was evaluated by examining correlations
831
among predictors or criteria. Although substantial
covariance was evident among some predictors (e.g.,
indicators of aggressive behavior and peer rejection),
it was not of a magnitude that constituted serious
multicollinearity (e.g., destabilized matrix inversion). Measurement models were initially evaluated
to determine whether the hypothesized alignments
of construct indicators and corresponding latent
variables fit the data well, and whether there was
evidence of factorial invariance or constancy in the
relations of indicators to constructs over time.
Structural equations modeling was then utilized to
examine the fit between the data and the proposed
(i.e., hypothesized and alternative) prediction models. Two criteria were used to judge model fit (see Hu
& Bentler, 1999): the root mean square error of approximation (RMSEA .08) and the standardized
root mean square residual (SRMRo.08). CFI was not
utilized because it tends to be a conservative fit indicator when measurement models are complex. All
estimated models were compared by gender using
multisample SEM analyses, and the same analytic
strategy was used to compare model fit for the entire
sample versus a reduced sample in which a small
number (n 5 4) of dual behavioral risk children (i.e.,
those with scores41.00 SD on the averaged indicator
scores for both aggressive and withdrawn behavior
across four or more assessment occasions) had been
eliminated. Because there was no instance in which
model fit differed significantly by gender or by risk
status, the full sample was utilized in all analyses.
Estimation of Missing Values Via Multiple Imputation
Attrition was minimal, with 385 (96.49%) of the
original 399 participants remaining in the sample
from kindergarten to grade 6. Constraints on data
collection (e.g., absences, moves, dropouts, etc.)
caused 3.49% of the data points within the entire
longitudinal data set to be missing. The percentage
of missing data by measure across all times of administration averaged 3.52% (range 5 0.00 – 9.00%).
Missing data were estimated via NORM (Schafer,
1999), a multiple imputation program in which
missing multivariate data are simulated m41 times.
Three imputations were generated (n 5 3) because,
for data sets containing 10% or fewer missing data
points, it has been established that n 5 3 imputations
yield sufficient estimation efficiency (i.e., 97%; Rubin, 1987; Schafer, 1999). Mplus 3.01 (Muthén &
Muthén, 1998 – 2004) was used to estimate each of
the hypothesized models from the multiple imputed
data sets, and to average parameter estimates and
obtain combined standard errors.
832
Ladd
Formation of Latent Variables and Factorial Invariance
The multiple prediction models that were stipulated for this investigation necessitated the creation
of five latent variables: confrontive aggression,
withdrawn behavior, peer group rejection, internalizing problems, and externalizing problems. Each
latent variable was constructed from its manifest
indicators (i.e., see measures listed under corresponding constructs in the Method section). For example, the latent confrontive aggression measure
was formed from three indicators: the PN-VAG, PNPAG, and CBS-AG measures. The coherence between
indicators and their respective latent variables was
evaluated as part of the SEM analyses that were
conducted for each prediction model; without exception, indicators loaded only on the latent variables that they were intended to represent (see
measurement model results in subsequent sections).
To determine whether the factor structure for each
latent variable was invariant over time, five multisample CFAs were calculated. In each of these
analyses, grade level was the grouping variable and
the measurement model contained a single latent
variable that was constructed from its corresponding
indicators. Using a marker variable strategy, the
unstandardized loading for one indicator (per factor)
was fixed at 1.0, and all remaining factor loadings
were freely estimated, as were error covariances between identical indicators across measurement occasions (Rensvold & Cheung, 2001). The results of
these analyses (ranges for fit statistics: w2/df 5 1.44 –
3.85; RMSEA 5 .03 – .08; SRMR 5 .03 – .07) supported
the assumption that each latent variable’s indicators
formed a single factor that was invariant across
measurement occasions.
Stabilities of the Latent Variables
SEM analyses were also used to estimate the stability of each latent variable across the seven assessment occasions. Each of these analyses was
conducted on a model in which the same latent
variable was estimated on each of seven measurement occasions from its corresponding indicators.
Factor loadings were freely estimated, as were error
covariances for identical indicators across measurement occasions. The results are shown by construct
in Figure 4. Of the five latent variables, aggression
and peer rejection were the most stable from kindergarten to grade 6; the stability coefficients for
these variables were substantial at lag 1 and remained high across all subsequent lags. The withdrawn behavior and externalizing latent variables
exhibited a pattern of increasing stability over time;
the stability coefficients for these variables were low
to moderate in magnitude during the early grades
(across lags 1 – 2) but became progressively stronger
thereafter (e.g., across lags 3 – 6). Only modest stability was found for internalizing problems and, over
time (lags 1 – 6), the stability coefficients for this latent variable increased only slightly.
Estimation of Additive Continuity Models
The hypothesis that each of two variables makes a
distinct (additive) predictive contribution to a criterion rests on the assumption that both variables are,
independently, predictors of that criterion. Thus, as a
precondition to estimating the proposed additive
models, six ‘‘main effects’’ models were tested via
SEM to determine whether each of the latent behavioral and peer rejection variables, when evaluated
independently, were significant predictors of the latent
externalizing or internalizing variables across one or
more of the six lagged, 1-year intervals. These analyses were conducted on a model that contained one
latent predictor variable measured on seven occasions (either aggressive behavior, withdrawn behavior, or peer rejection at each of the seven grade
levels) and one latent criterion variable measured on
seven occasions (either externalizing or internalizing
behavior at each of the 7 grade levels). Each of the
latent variables included in these analyses was estimated from its corresponding indicators. Factor
loadings for latent variables were freely estimated, as
were error covariances for matching indicators
across measurement occasions and same-informant
indicators within measurement occasions (see Cole
& Maxwell, 2003). For example, in the analysis where
aggressive behavior predicted externalizing problems, structural paths were estimated across each of
six lagged, 1-year intervalsFthat is, from the latent
aggressive behavior predictor measured at one grade
level to latent externalizing problems criterion measured at the next grade level (e.g., AG-K ! EXT-G1;
AG-G1 ! EXT-G2; AG-G2 ! EXT-G3; AG-G3 !
EXT-G4; AG-G4 ! EXT-G5; & AG-G5 ! EXT-G6F
K, kindergarten; G1, grade 1).
For externalizing problems, the model in which
aggression was the sole predictor fit the data adequately, w2(721) 5 2,362.04, po.01; RMSEA 5 .08,
SRMR 5 .07, as did the model in which peer rejection
was a lone predictor, w2(721) 5 1,922.32, po.01;
RMSEA 5 .06, SRMR 5 .07. Positive and statistically
significant path coefficients were found from aggression to externalizing problems, and these
weights increased in magnitude across the six lagged
Predicting Maladjustment
Kindergarten
Grade 1
.90
AG
.38
AG
.31
WD
.87
REJ
.67
.26
.87
.27
INT
.40
INT
.95
.83
REJ
.91
EXT
.27
INT
WD
REJ
EXT
.37
INT
.60
.91
.78
AG
WD
REJ
EXT
.93
.48
.93
Grade 6
AG
WD
REJ
EXT
.93
.56
.82
.63
Grade 5
AG
WD
REJ
EXT
.91
.56
.73
Grade 4
AG
WD
REJ
EXT
Grade 3
.95
.82
AG
WD
INT
Grade 2
833
EXT
.33
INT
INT
Figure 4. Stability coefficients are shown for each of the six latent variables across each of the six lags that were examined in this investigation. AG 5 aggressive behavior; WD 5 withdrawn behavior; REJ 5 peer rejection; EXT 5externalizing problems; INT 5 internalizing
problems.
intervals (e.g., .67, .73, .78, .79, .80, .89, respectively).
The path weights from peer rejection to externalizing
problems were also positive and significant, but
these coefficients declined in magnitude across the
latter of the six lags (i.e., .73, .70, .74, .68, .54, .55,
respectively). In contrast, the model for withdrawn
behavior did not fit the data, w2(487) 5 2,207.40,
po.01; RMSEA 5 .10, SRMR 5 .18, and none of the
path coefficients were significant.
For internalizing problems, model fit was adequate for the withdrawn model, w2(279) 5 795.27,
po.01; RMSEA 5 .07, SRMR 5 .07, and for the peer
po.01;
rejection
model,
w2(487) 5 1,176.78,
RMSEA 5 .06, SRMR 5 .07. All but the first of the six
path coefficients from withdrawn behavior to internalizing problems were significant, and these
weights increased in magnitude across the six lags
(.09, .15, .19, .31, .26, .40, respectively). Similarly, all
path coefficients from peer rejection to internalizing
problems were significant and tended to increase in
magnitude across the six lags (.14, .22, .43, .38, .32,
.42, respectively). Although the aggression model
approximated a fit to the data, w2(487) 5 1,302.18,
po.01; RMSEA 5 .07, SRMR 5 .10, none of the path
weights from aggression to internalizing problems
were significant.
These findings indicated that it was feasible to
evaluate two of the hypothesized additive models.
The first of these models (i.e., the AG1REJ additive
model) provided a test of the hypothesis that, in
addition to children’s propensity to engage in aggressive behavior, peer group rejection contributes
additively to the prediction of externalizing problems. The second model (i.e., the WD1REJ additive
model) addressed the hypothesis that beyond children’s propensity to engage in asocial – withdrawn
behavior, peer group rejection contributes additively
to internalizing problems. In each case, it was of
interest to determine (1) whether peer rejection
contributed additively to the prediction of maladjustment, (2) when during childhood such contributions were evident, and (3) how the observed
contributions differed in strength and continuity
across grade levels (i.e., the six lagged, 1-year intervals). Because results from the main effects analyses
did not implicate withdrawn behavior as a signi-
834
Ladd
ficant predictor of externalizing problems, nor
aggressive behavior as a predictor of internalizing
problems, the predictive contributions of peer rejection were not estimated in conjunction with these
latent variables.
Both the measurement and structural components
of the AG1REJ and WD1REJ models are depicted
graphically in Figures 5 and 6. In the SEM analyses
performed on these models, the latent predictor and
criterion variables were estimated from their corresponding indicators. Factor loadings for all latent
variables were freely estimated, as were error covariances for matching indicators across measurement occasions and for same-informant indicators
within measurement occasions.
AG1REJ additive model. Calculation of the correlations among the multiple indicators of the latent
variables included in this model (averaged across
imputed data sets) yielded a very large number of
coefficients (n 5 1,953). For the sake of brevity, the
ranges of coefficients obtained between predictors
(AG & REJ) across all lags, and between the predictors and the criterion (AG & EXT; REJ & EXT) across
all lags, are reported by measure in the left column of
Table 1. SEM findings (see Figure 5) showed that this
model fit the data adequately, w2(1,636) 5 4,138.90,
po.01; RMSEA 5 .06, SRMR 5 .06. Indicators loaded
significantly (all ls po.01) on their designated latent
variables. Across each of the six lags (1-year intervals), significant positive path weights were found
not only from aggression to externalizing problems
but also from peer rejection to externalizing problems (po.01). Thus, from kindergarten to grade 6,
both aggression and peer rejection at one grade level
were found to be distinct and significant predictors
of externalizing problems at the next grade level.
There was evidence of continuity in the predictive
(additive) contributions of peer group rejection to
externalizing problems in the sense that the path
coefficients for peer rejection were significant (after
adjusting for aggression) across every grade level
from kindergarten to grade 6. Discontinuity was
observed, however, in the sense that the magnitudes
of these path coefficients varied across grade levels;
of the weights obtained, those for the early primary
grades (particularly those spanning grades 1 – 2 and
2 – 3) were among the strongest.
WD1REJ additive model. In this model (see Figure
6), withdrawn behavior and peer rejection served as
predictors, and it was of interest to examine the
continuity and strength of peer rejection as an additive predictor of internalizing problems across the
investigated grade intervals. The ranges of the correlations obtained between predictors (WD & REJ)
across all lags, and between the predictors and the
criterion (WD & INT; REJ & INT) across all lags, are
reported by measure in the right column of Table 1.
SEM results indicated an adequate fit between this
model and the data, w2(924) 5 2,213.76, po.01;
RMSEA 5 .06, SRMR 5 .07. Measurement model results showed that indicators loaded on their designated latent variables (all ls po.01). Results for the
structural model revealed substantial continuity in
the predictive (additive) contributions of peer group
rejection to internalizing problems in the sense that
the path coefficients obtained for this latent variable
were significant (after adjusting for asocial – withdrawn behavior) across all grade intervals except
K ! G1. In contrast, the path weights for asocial –
withdrawn behavior were significant only for the
last three grade intervals (i.e., G3 ! G4, G4 ! G5,
and G5 ! G6), suggesting that this variable’s additive contribution to the prediction of internalizing
problems was not substantial until later in childhood. The importance of peer rejection as a distinct
predictor of internalizing problems was also illustrated by the fact that, across each of the lagged intervals, the path weights obtained for this variable
were larger in magnitude than those obtained for
asocial – withdrawn behavior.
Estimation of Disorder-Driven Models
Another series of SEM analyses were conducted to
evaluate models founded on the premise that earlyemerging maladjustment predicts children’s subsequent propensities to engage in specific behaviors
(aggressive or withdrawn) and be rejected by their
classmates. Here again, main effects models were
initially estimated to determine whether early externalizing problems or internalizing problems,
when evaluated independently, were significant
predictors of each of the latent behavioral risk and
peer rejection variables across one or more of the six
lagged, 1-year intervals. Each of these analyses was
conducted on a model that contained one latent
predictor variable measured on seven occasions (either externalizing or internalizing behavior at each of
the seven grade levels) and one latent criterion variable measured on seven occasions (either aggressive
behavior, withdrawn behavior, or peer rejection at
each of the seven grade levels). Each latent variable
was estimated from its corresponding indicators,
and factor loadings for latent variables were freely
estimated, as were error covariances for matching
indicators across measurement occasions and sameinformant indicators within measurement occasions.
To illustrate, in the analysis where externalizing
3
2
1
3
.84
.27
.68
3
.47
1
.85
1
.92
1
.91
Lag 1
.74
EXT
1
REJ
1
–.82
2
AG
1
2
3
3
.76
.47
.52
3
.85 .71
2
Grade 1
.56
1
.87
1
.87
1
.88
Lag 2
.72
EXT
2
REJ
2
–.76
2
AG
2
2
3
3
.73
1
.84
1
.82
1
.82
Lag 3
.64
.36
.64
3
.82 .75
2
Grade 2
.76
EXT
3
REJ
3
–.77
2
AG
3
2
3
3
.74
1
.66
1
.76
1
.85
Lag 4
.68
.23
.64
3
.79 .77
2
Grade 3
2
.76
EXT
4
REJ
4
–.59
2
AG
4
.65
2
3
3
.77
.72
1
.78
1
.79
1
.78
Lag 5
.18
.63
3
.78
Grade 4
2
.73
EXT
5
REJ
5
–.75
2
AG
5
.71
2
3
3
.68
1
.83
1
.90
1
.76
Lag 6
.76
.20
.67
3
.72
Grade 5
.71
EXT
6
REJ
6
–.76
2
AG
6
2
3
3
.72
.69
3
.85 .68
2
Grade 6
Figure 5. Path diagram for the AG1REJ additive model. AG 5 latent variable for aggressive behavior; REJ 5 latent variable for peer rejection; EXT 5 latent variable for externalizing
problems. Indicators for latent variables: AG (1: PN-PAG; 2: PN-VAG; 3: CBS-AG); REJ (1: PN-NEG; 2: RR-SMS; 3: CBS-EP); and EXT (1: CD; 2: TRF-DB; 3: CBS-HD). All path weights
shown in this model were statistically significant (po.01).
.75
EXT
K
REJ
K
–.70
2
AG
K
.85 .76
2
.88
.71
1
.87
1
Kindergarten
Predicting Maladjustment
835
1
INT
K
REJ
K
2
–.90
2
WD
K
.11
.58
3
.91
.73
2
.05
Lag 1
.89
.83
1
.60
1
1
INT
1
REJ
1
2
–.86
2
WD
1
.09
Lag 2
.18
.56
3
.84
.91
2
Grade 1
.82
.76
1
.85
1
1
INT
2
REJ
2
2
–.76
2
WD
2
.13
Lag 3
.37
.70
3
.88
.80
2
Grade 2
.83
.78
1
.76
1
1
INT
3
REJ
3
2
–.80
2
WD
3
Lag 4
.17
.28
.75
3
.91
.87
2
Grade 3
.77
.73
1
.64
1
1
INT
4
REJ
4
2
–.73
2
WD
4
.16
Lag 5
.23
.67
3
.98
.84
2
Grade 4
.91
.68
1
.75
1
1
INT
5
REJ
5
2
–.73
2
WD
5
Lag 6
.17
.32
.70
3
.83
.67
2
Grade 5
.84
.75
1
.84
1
1
INT
6
REJ
6
2
–.78
2
WD
6
.75
3
.86
.80
2
Grade 6
Figure 6. Path diagram for the WD1REJ additive model. WD 5 latent variable for withdrawn behavior; REJ 5 latent variable for peer rejection; INT 5 latent variable for internalizing
problems. Indicators for latent variables: WD (1: CBS-AB; 2: TRF-WD); REJ (1: PN-NEG; 2: RR-SMS; 3: CBS-EP); and INT (1: TRF-AD; 2: CBS-AF). All path weights above .16 were
statistically significant (po.01).
.69
.83
1
.67
1
Kindergarten
836
Ladd
Predicting Maladjustment
837
Table 1
Range of Correlations Obtained Between Latent Variable Indicators Across Six Assessment Lags
AG1REJ model
Aggressive behavior and peer rejection
PN-PAG & RR-SMS
PN-PAG & PN-NEG
PN-PAG & CBS-EP
PN-VAG & RR-SMS
PN-VAG & PN-NEG
PN-VAG & CBS-EP
CBS-AG & RR-SMS
CBS-AG & PN-NEG
CBS-AG & CBS-EP
WD1REJ model
Withdrawn behavior and peer rejection
TRF-WD & RR-SMS
TRF-WD & PN-NEG
TRF-WD & CBS-EP
CBS-AB & RR-SMS
CBS-AB & PN-NEG
CBS-AB & CBS-EP
.21 to .56
.21 to .60
.11 to .35
.19 to .51
.21 to .58
.11 to .40
.23 to .45
.21 to .56
.11 to .57
.12
.05
.10
.08
.07
.07
to
to
to
to
to
to
.39
.32
.53
.44
.37
.60
to
to
to
to
.58
.56
.42
.48
Aggressive behavior and externalizing problems
PN-PAG & CD
.36 to .61
PN-PAG & CBS-HD
.27 to .58
PN-PAG & TRF-DB
.38 to .60
PN-VAG & CD
.31 to .63
PN-VAG & CBS-HD
.22 to .52
PN-VAG & TRF-DB
.33 to .57
CBS-AG & CD
.35 to .61
CBS-AG & CBS-HD
.21 to .60
CBS-AG & TRF-DB
.35 to .62
Withdrawn behavior and internalizing problems
TRF-WD & TRF-AD
TRF-WD & CBS-AF
CBS-AB & TRF-AD
CBS-AB & CBS-AF
.04
.04
.01
.01
Peer rejection and externalizing problems
RR-SMS & CD
RR-SMS & CBS-HD
RR-SMS & TRF-DB
PN-NEG & CD
PN-NEG & CBS-HD
PN-NEG & TRF-DB
CBS-EP & CD
CBS-EP & CBS-HD
CBS-EP & TRF-AD
Peer rejection and internalizing problems
RR-SMS & TRF-AD
RR-SMS & CBS-AF
.01 to .26
.01 to .29
.18
.14
.14
.19
.20
.18
.13
.16
.16
to
to
to
to
to
to
to
to
to
.54
.56
.44
.55
.53
.56
.53
.50
.46
PN-NEG & TRF-AD
PN-NEG & CBS-AF
.03 to .24
.05 to .28
CBS-EX & TRF-AD
CBS-EX & CBS-AF
.03 to .52
.05 to .51
AG 5 aggressive behavior; WD 5 withdrawn behavior; REJ 5 peer rejection.
problems were evaluated as predictors of aggressive
behavior, structural paths were estimated across
each of six lagged, 1-year intervalsFthat is, from the
latent externalizing predictor measured at one grade
level to latent aggressive variable measured at the
next grade level (e.g., EXT-K ! AG-G1; EXTG1 ! AG-G2; EXT-G2 ! AG-G3; EXT-G3 ! AGG4; EXT-G4 ! AG-G5; & EXT-G5 ! AG-G6).
As the following fit statistics reveal, there was less
than adequate coherence between the data and each of
these six models: externalizing problems ! aggressive behavior, w2(721) 5 2,821.55, po.01; RMSEA 5 .09,
SRMR 5 .14; externalizing problems ! withdrawn
behavior, w2(487) 5 1,976.28, po.01; RMSEA 5 .09,
SRMR 5 .09; externalizing problems ! peer rejection, w2(721) 5 2,073.41, po.01; RMSEA 5 .07,
SRMR 5 .14; internalizing problems ! aggressive
behavior, w2(487) 5 1,679.58, po.01; RMSEA 5 .08,
SRMR 5 .17; internalizing problems ! withdrawn
behavior, w2(279) 5 1,106.77, po.01; RMSEA 5 .09,
SRMR 5 .13; internalizing problems ! peer rejection,
w2(487) 5 1,973.89, po.01; RMSEA 5 .09, SRMR 5 .21.
Because none of these models fit the data well, no
attempt was made to evaluate larger disorder-driven
models such as those in which antecedent maladjustment variables predicted both children’s behavior
(e.g., aggressive, withdrawn behavior) and peer rejection across multiple grade intervals.
Estimation of Transactional Models
Two final SEM analyses were undertaken to determine whether the predictive relations observed
between the investigated risk factors and dysfunctions followed a cyclical pattern that conformed to
one of the two proposed transactional models. Both
of these models were estimated such that the direction of effect between risk factors and maladjustment
838
Ladd
were reversed beginning at grade 3 (at lag 4, as
shown in the upper and lower panels of Figure 3).
The first of these modelsFthe risk ! maladjustment ! risk model (see Figure 3, upper panel)Fcontained three latent variables, each of which
was measured on seven occasions (at each of the
seven grade levels). This model was estimated such
that (a) in kindergarten to grade 2, the latent aggressive behavior and peer rejection variables that
were measured at each of these grade levels predicted the latent externalizing problems variable that
was measured at the next grade level (each of these
structural paths traversed a 1-year lag), and (b) in
grades 3 – 6, the direction of prediction reversed so
that the latent externalizing problems variables that
were measured at each of these grade levels predicted the latent aggressive behavior and peer rejection variables that were measured at each
subsequent grade level (each path traversed a 1-year
lag). The second of the two modelsFthe maladjustment ! risk ! maladjustment model (see Figure 3, lower panel) was estimated as follows: (a) in
kindergarten to grade 2, the latent internalizing
problems variable that was measured at each of
these grade levels predicted the latent withdrawn
behavior and peer rejection variables that were
measured at the next grade level (each of these
structural paths traversed a 1-year lag), and (b) in
grades 3 – 6, the direction of prediction reversed so
that the latent withdrawn behavior and peer rejection variables that were measured at each of these
grade levels predicted the latent internalizing problems variable that was measured at each subsequent
grade level (each path traversed a 1-year lag). In both
analyses, the latent predictor and criterion variables
were estimated from their corresponding indicators.
Factor loadings for all latent variables were freely
estimated, as were error covariances for matching
indicators across measurement occasions and for
same-informant indicators within measurement occasions. Results showed that neither the risk !
maladjustment ! risk model, w2(1,664) 5 4,626.61,
po.01; RMSEA 5 .07, SRMR 5 .11, nor the maladjustment ! risk ! maladjustment model, w2(1,672)
5 5,099.54, po.01; RMSEA 5 .07, SRMR 5 .13, fit the
data adequately. Altering the structural paths such
that the reversals occurred at earlier or later grade
levels did not improve the fit of either model.
Discussion
The results of this investigation advance extant
knowledge in several novel, specific ways. Unlike
the evidence obtained in past investigations, the data
obtained from this study yielded a more comprehensive, developmental portrait of the relations that
evolve among children’s behavioral orientations,
peer rejection, and psychological maladjustment
across the entire grade school period (ages 5 – 12). In
addition, this study’s prospective longitudinal panel
design permitted an evaluation of differing premises
about the causal forces that may underlie the relations that develop among the targeted risk and
maladjustment variables. By estimating traditional
models along with conceptually viable but previously untested alternative models, it was possible to
draw stronger inferences about the relative importance (predictive significance) of peer behavioral and
relational risks in the development of childhood
psychological maladjustment and vice versa.
At the broadest level of analysis, the combined
model evaluation results were consistent with a child
and environment perspective, and implicated both
child attributes (i.e., aggressive, withdrawn behavioral orientations) and features of the child’s peer
milieu (i.e., peer rejection) as risk factors for psychological maladjustment (Ladd, 1989, 2003, 2005;
McDougall et al., 2001; Parker & Asher, 1987; Parker
et al., 1995). More importantly, the results were
congruent with an additive child and environment
model (see Ladd, 2003). An additive model implies
that maladjustment has multiple triggering mechanisms and accelerants (exacerbating factors) that,
when concurrently operative, act as separate forces
(complements) on the inception and development of
dysfunction. The fact that the data conformed to an
additive model is significant because, theoretically,
these findings strengthen the view that peer rejection
contributes to children’s maladjustment, and does so
in a way that is at least partially independent of their
behavioral orientations. That is, the findings are
consistent with the inference that, in addition to
agentic forces (the child’s performance of aggressive
or withdrawn behaviors among peers), relational
influences (peers’ rejection of the child) shape early
psychological maladjustment and its developmental
course. On the basis of this, it can be argued that
maladjustment becomes more likely when either of
these risk factors is present, and its probability of
occurrence or severity is exacerbated when both are
operative. This is in contrast to several competing
theoretical perspectives (for a review, see Ladd,
2003), including those in which it has been argued
that (a) child (e.g., behavioral) rather than relational
forces, or vice versa, constitute the determinants of
maladjustment (i.e., main effects models), (b) peer
rejection is a consequence of deviant behavior rather
Predicting Maladjustment
than a cause of maladjustment (e.g., incidental, behavior continuity models), (c) children’s behavioral
styles and peers’ rejecting reactions derive from
emergent dysfunction rather than vice versa (i.e.,
disorder-driven models), and (d) child behavior,
peer rejection, and maladjustment influence each
other and evolve cyclically over time (i.e., transactional models).
It must be recognized, however, that even though
additive models better represented the covariance
structures within this longitudinal data set than did
the investigated alternative models, such findings do
not render implausible other hypothesized but untested effect patterns. Investigators have, for example, proposed models other than those tested here to
explain how behavioral and relational risks might
combine to influence psychological maladjustment.
Examples would include models in which one risk
factor’s effects on children’s maladjustment is hypothesized to be (a) contingent upon another variable (i.e., moderator models) or (b) transmitted
through one or more variables (i.e., mediator models;
see Ladd, 2003, for further explication of these
models). Models such as these were not evaluated in
this study, however, because they could not be operationalized or analyzed in a way that made them
comparable, from a sample size (power) or design
perspective, to the investigated main effects, additive, disorder-driven, and transactional models. For
example, evaluation of a mediated model, using this
study’s panel design, would have required estimating latent variable linkages across multiple rather
than single time lags.
Similarly, interpretive certainty is curtailed by
sampling constraints. It remains to be seen whether
the reported data patterns, which were obtained
with a normative community sample (one that was
fairly diverse with respect to gender, SES, ethnicity,
and geographic locales), will replicate in other samples, including ones that incorporate other cultural
groups, peer environments, and child populations.
For example, stronger empirical support for disorder-driven models might be found in clinical samples, such as those composed entirely of children
who manifest diagnosable early-onset psychopathology. Conceivable, too, are differences in the relative importance of additive risk factors across
samples. It is possible, for example, that the strength
of a particular child attribute as a predictor of maladjustment may vary as a function of the ‘‘normativeness’’ of that characteristic within the child’s peer
environment. The additive contribution of aggression, for example, relative to peer rejection, might be
diminished within peer environments where ag-
839
gression is widespread (e.g., exhibited by most
group members) or so frequent as to desensitize
children to its occurrence and effects (see Wright,
Giammarino, & Parad, 1986). For these reasons, it
will be necessary to undertake additional studies to
achieve a fuller, more comprehensive understanding
of how peer behavioral and relational risks antecede
psychological maladjustment.
In the meantime, however, the present findings
enhance the stature of additive modelsFthat is, they
lend support to the premise that, in addition to
specific behavioral orientations, peer rejection increases the probability that children will become
maladjusted. This inference, however, calls attention
to the question of why peer rejection forecasts or, if
construed as a determinant, ‘‘contributes’’ to dysfunction in ways that are partially distinct from those
attributable to children’s behaviors. Pertinent to this
issue is the contention that, even though children’s
behaviors contribute to their rejection (see Coie &
Kupersmidt, 1983; Dodge, 1983), the subsequent establishment of a rejecting peer milieu exposes children to other (i.e., nonshared) processes that foster
the development of psychological dysfunction. In
other words, behaviors such as aggression may cause
peer rejection, but once children come to occupy this
social position (i.e., acquire rejected status), their
experiences in the peer group may change in ways
that affect their adjustment. As Coie (2004) has argued, children’s behavioral characteristics may have
a pathogenic influence on their development, but
peer rejection has its own debilitating consequences.
Therefore, why is peer rejection debilitating to
children’s psychological health in a way that is distinct from their behavioral propensities? Insight into
this question can be found, in part, within emergent
metatheories where it has been proposed that peer
rejection is not just a social ‘‘address’’ (e.g., a child’s
position or status in a peer group) but also a social
experienceFone that exposes children to adverse or
painful interpersonal processes (see Bukowski &
Hoza, 1989; Coie, 1990, 2004; Sandstrom & Zakriski,
2004). For example, it has been hypothesized that the
acquisition of rejected status (an awareness among
peers that an individual is disliked) signals to
members of the peer group that disliked individuals
are suitable targets for maltreatment (see Buhs &
Ladd, 2001; Coie, 1990). Coie (1990, 2004), for example, contends that once children are rejected,
peers treat them more negatively, and that this abuse
and the isolation and stigma it often produces are
distinct causes of maladjustment. In this way, peer
rejection can be seen as having effects on children
that are at least partially distinct from those attrib-
840
Ladd
utable to aggression or other behaviors that may
antecede rejection. Support for this contention has
begun to accrue; evidence indicates that peers perpetrate many forms of abuse against rejected children (e.g., ignoring, exclusion, threats, mocking,
verbal and physical abuse; see Asher et al., 2001) and
that peer maltreatment predicts psychological and
school maladjustment even after controlling for individual differences in children’s aggressive and
withdrawn behaviors (see Buhs & Ladd, 2001; Gazelle & Ladd, 2003; Ladd, Birch, & Buhs, 1999; Ladd
& Troop-Gordon, 2003).
In contrast, little evidentiary support was found
for alternative or competing prediction models based
on the view that premorbid forms of psychological
dysfunction foster the development of specific behavioral propensities or peer group rejection (i.e.,
disorder-driven models; see Parker & Asher, 1987),
or that the investigated forms of risk and dysfunction influence each other cyclically over time (i.e.,
transactional models; see Parker et al., 1995). Collectively, the lack of unanimity between these models
and the data, coupled with the stronger fit found for
additive models, implies that developments within
the peer milieu (children’s behavior and rejection
among age-mates) are more influential in shaping
the course of dysfunction than vice versa. Moreover,
this implied direction of effect remained evident
over a 7-year period, suggesting that both peer behaviors and relations continue to play an important
role in the emergence and maintenance of psychological dysfunction throughout childhood.
In part, the basis for the lesser efficacy of the disorder-driven and transactional models was reflected
in the pattern of stabilities and main effect linkages
that were observed among the investigated latent
variables. Consider, for example, the relative stability
of the three latent variables that were included in the
AG1REJ and WD1REJ additive models. Within each
model, the two latent peer predictors (i.e., aggressive
behavior and peer rejection, withdrawn behavior
and peer rejection, respectively) exhibited greater
stability across the six lagged, 1-year intervals than
did the latent maladjustment criterion (i.e., externalizing or internalizing problems, respectively).
This relative difference in stabilities was particularly
apparent during the early grade school years, suggesting that individual differences in maladjustment
were subject to greater temporal variability (change
over time) than were differences in children’s behavioral orientations or peer rejection. This apparent
instability may reflect the fact that internalizing and
externalizing problems, when measured at early ages
and in normative samples such as this one, constitute
incipient rather than mature forms of dysfunction. If
so, it is likely that the symptoms, or outward expressions of these dysfunctions, were few in number,
sporadically expressed, and only moderately discrepant from age norms. For this reason, these dysfunctionsFas they were initially manifestedFmay
not have constituted a force strong or consistent enough to induce moderate to severe behavioral deviance (e.g., aggressive or withdrawn behavior) or peer
censure (e.g., peer group rejection). Additionally, results from the main effects models revealed that,
across each of the six lags, the predictive links between both of the peer variables and the maladjustment criteria remained substantial (statistically
significant) or became stronger over time. Collectively, these findings are consistent with the inference
that children’s behavioral orientations and peer rejection anteceded emergent forms of dysfunction or,
conversely, that manifestations of maladjustment
changed over time so as to become increasingly
predictable from children’s behavior and social rejection within the peer milieu.
Second, the fact that peer rejection increased the
efficacy with which both externalizing and internalizing problems could be predicted (beyond that attributable to differing behavioral orientations)
suggests that this aspect of children’s peer experience is a risk factor for both externalizing and internalizing problemsFthat is, multiple forms of
psychopathology. The findings also suggest, however, that it would be imprudent to conclude that the
strength and pattern of these predictive contributions were identical or invariant over time. Rather,
examination of peer rejection in different variable
contexts (e.g., in conjunction with aggression as a
predictor of externalizing problems, and in combination with withdrawn behavior as a predictor of
internalizing problems), and across repeated time
intervals throughout childhood (yearly, from kindergarten to grade 6), revealed differences in the
timing, strength, and consistency of peer rejection as
a distinct predictor of externalizing versus internalizing problems.
Results from the AG1REJ additive model corroborated extant evidence by indicating that peer
rejection, as well as children’s aggressive behavior,
presages the development of externalizing problems.
More importantly, these findings expanded extant
knowledge by showing that the distinct predictive
link between peer rejection and externalizing problems (1) was significant and sustained across a substantial period of child development, from childhood
through preadolescence, and (2) tended to be
stronger during the early rather than the later grade
Predicting Maladjustment
school years. Thus, along with aggressive behavior,
peer rejection was found to be an enduring risk factor throughout childhood, but one that proved to be
a stronger prognosticator of impending externalizing
problems during the early rather than later school
years. These age-related differences in prediction
were consistent with the hypothesis that peer rejection during the primary grades (ages 5 – 9) causes
children to resolve insipient social needs and tasks in
ways that orient them toward antisocial rather than
normative social and behavioral development. On
the basis of Sullivanian premises, Buhrmester and
Furman (1986) characterized this particular age period (i.e., the ‘‘juvenile era’’) as one in which children
experience a growing need for peer acceptance (inclusion) and a heightened fear of peer rejection (exclusion). As these needs become prominent, peer
rejection experiences (e.g., ignoring, exclusion, maltreatment) may become an impetus for negative
emotions (e.g., anger at exclusion), forceful behaviors
(e.g., attacks on peers, revenge seeking), rule violations (e.g., rejecting norms established or followed by
nonrejected peers), and other maladaptive reactions
that lead to the development of externalizing problems (see Coie, 2004; Kupersmidt & DeRosier, 2004).
Similarly, it has been proposed that peer rejection
during the early grades institutes thought and behavior patterns (e.g., attributional biases, negative
peer beliefs, coercive cycles) that persist over time
and increase children’s propensity to engage in
misconduct and delinquency (see Coie, 2004; Ladd &
Troop-Gordon, 2003; Sandstrom & Zakriski, 2004).
Within this same variable context, the efficacy of
aggressive behavior as a distinct predictor of externalizing problems not only remained significant but
also increased in magnitude from childhood through
preadolescence. Although not previously documented longitudinally across this entire age period,
perhaps it should not be surprising that aggressive
behavior became a stronger predictor of externalizing problems over time. The current findings corroborated the view that children’s use of force
against age-mates is an enduring behavioral propensity for both boys and girls (Cairns et al., 1989;
Ladd, 2003; Olweus, 1979), and one that is likely to
be perpetuated by its social consequences (Caspi et
al., 1987) and other mediating factors (e.g., aggressive children’s tendency to infer that forceful behaviors are effective; see Boldizar, Perry, & Perry,
1989; Crick & Ladd, 1990; Egan, Monson, & Perry,
1998; Slaby & Guerra, 1988). Eventually, the use of
confrontive aggression in the peer milieu may generalize to a larger range of environmental contingencies and cause children to develop broader and
841
more serious forms of externalizing problems.
Moreover, as children approach preadolescence, they
are less likely to be restricted by the contextual
constraints of self-contained classrooms and more
likely to engage in social niche seeking. It has been
substantiated that as children mature, those with
aggressive tendencies tend to seek the company of
other aggressive peers who, in turn, may socialize or
encourage (e.g., condone) more frequent and serious
forms of externalizing behavior (Brown, Dolcini, &
Leventhal, 1997; Cairns & Cairns, 1994; Cairns,
Cairns, Neckerman, Gest, & Gariepy, 1988). Moreover, in this context, aggressive children may find that
it is necessary to engage in more frequent and extreme conduct problems to establish and maintain
their social positions among aggressive peers (Coie,
2004).
Results from the WD1REJ additive model implied that the efficacy with which peer rejection and
children’s withdrawn behavior predicted internalizing problems increased as children matured, or were
stronger during the later rather than the early grade
school years. Compared with results for the AG1REJ
model, these findings suggest that internalizing
problems are less stable than externalizing problems
throughout grade school, and that peer rejection and
withdrawn behavior are less powerful predictors
(risk factors) for this type of dysfunction, especially
during the early elementary school years. Of the two
risk factors, peer rejection was the stronger; unlike
withdrawn behavior, it emerged as a significant
predictor of internalizing problems early in grade
school (as indicated by main effects results), and its
distinct predictive contributions increased progressively over time (implied by additive model findings). This was in contrast to rejection’s distinct
predictive contributions (when evaluated with aggressive behavior) to externalizing problems, which,
as previously noted, declined as children matured.
These contrasting data patterns were consistent with
McDougall et al.’s (2001) contention that peer rejection experiences are particularly influential in the
development of internalizing problems. Moreover,
these data extend that inference by suggesting that
rejection’s role becomes progressively more important as an antecedent of internalizing problems as
children mature. This may be the case because repeated exposure to rejection experiences (e.g., peer
maltreatment) may intensify children’s feelings of
distress and strengthen processes that instigate internalizing problems (e.g., uncertainty, mistrust, debilitating self-attributions, etc.; see Chorpita &
Barlow, 1998; Pomerantz & Rudolph, 2003; Rubin et
al., 1990; Troop-Gordon & Ladd, 2005). However, a
842
Ladd
comparison of the absolute magnitude of rejection’s
distinct predictive contributions (additive path
weights) with externalizing versus internalizing problems over time does not appear to justifyFat least for
the age period investigated hereFthe conclusion that
peer rejection is a more powerful risk factor for internalizing than for externalizing problems.
Additionally, the findings expand what is known
about withdrawn behavior as a risk factor for psychopathology. The data were consistent with the
conclusion that the risk posed by withdrawn behavior for internalizing problems was low during
the early grade school years, but higher during the
middle grade school years when, perhaps not surprisingly, peers begin to judge withdrawn behavior
as atypical or deviant (Younger & Boyko, 1987;
Younger et al., 1993). It may be the case that, as
children grow older, those prone to engage in withdrawn behavior become increasingly marginalized
from the mainstream of social interaction and, accordingly, begin to experience stronger feelings of
alienation, anxiety, loneliness, and other internalizing symptoms. Furthermore, by middle childhood
and thereafter, withdrawn behavior may become a
less mature way of coping with the social demands
of peer environments, and norms may shift such that
children who engage in such behaviors are at greater
risk for peer censure. Were this the case, both withdrawn behavior and peer rejection would become
more powerful risk factors, or predictors of subsequent internalizing problems. Indeed, the findings
reported here were congruent with this inference.
Comparatively, however, it would appear that
withdrawn behavior does not carry the same level of
risk for internalizing problems during the grade
school years as aggression does for externalizing
problems. In part, this appears to be attributable to
the fact that withdrawn behavior exhibited lesser
stability over time as compared with aggressive behavior, and the fact that internalizing problems
manifested lesser stability than did externalizing
problems. This result was consistent with past evidence indicating that, at early ages, withdrawn behavior is not a particularly powerful risk factor for
maladjustment (Ladd & Burgess, 1999; Rubin, 1982,
1993). Still, it may be judicious to reexamine this
conclusion in light of additional evidence. Whether
or not these findings accurately gauge the risk associated with early withdrawn behavior may depend, in part, on investigators’ definitions of this
construct and their ability to distinguish it conceptually and empirically from its psychological consequences (particularly, problems of an internalizing
nature). A key aim of this investigation was to define
and measure withdrawn behavior in a way that
would distinguish its attributes (i.e., avoidance of,
and infrequent interaction with peers) from those
indicative of its hypothesized or to-be-predicted
consequences (internalizing problems). Whether this
aim was fully realized using existing measures of
withdrawn behavior (e.g., TRF-WD, CBS-AB subscales) warrants further theoretical and empirical
scrutiny and, perhaps, cannot be determined until
greater precision is achieved in defining and measuring this construct and distinguishing it from its
reputed consequences. Interpretation is also complicated by the conceptual question of whether
withdrawn behaviorFonce distinguished from its
psychological consequencesFis multifaceted or
takes more than one form (see Rubin & Coplan,
2004). If the measures used to construct the latent
variable used in this study tapped more than one
form of withdrawn behavior, then the reported
findings may not accurately reflect the risk associated with constituent subtypes.
In sum, the results of this investigation reflect on a
long-standing controversy about the efficacy of two
childhood risk factorsFchildren’s behavioral
orientations (i.e., aggressive, withdrawn) and peer
group rejectionFas predictors of incipient and lateremerging forms of psychological maladjustment.
Unlike prior studies, the results provide a more nuanced, developmental perspective on when, to what
extent, and how consistently children’s behavioral
styles and peer rejection forecast the development of
externalizing and internalizing problems during the
5 – 12 age period. Peer group rejection, when evaluated in conjunction with children’s behavior, made
distinct contributions to the prediction of both externalizing and internalizing problems, and did so
consistently throughout the 5 – 12 age period. As anticipated, aggressive behavior was a stronger predictor of externalizing problems than was peer
rejection over the entire course of this age period, but
no support was found for the conclusion that only
aggression accounts for significant nonshared variance in externalizing problems. The predictive contributions of withdrawn behavior to internalizing
problems was eclipsed by peer rejection earlier in
childhood, but became progressively stronger (although not stronger than rejection) as children approached preadolescence. No support was found for
the premise that withdrawn behavior increases
children’s risk for externalizing problems, nor the
supposition that aggressive behavior increases children’s risk for internalizing problems.
Thus, these data add to a growing corpus of evidence that contests the validity of ‘‘incidental mod-
Predicting Maladjustment
els’’ (Parker & Asher, 1987; Rudolph & Asher, 2000) or
the proposition that peer rejection is simply a marker
(derivative) of the risks posed by children’s preexisting behavioral tendencies, such as aggressive and
withdrawn behavior. Rather, the overall pattern of
findings bolstered the assumption that both children’s
behavior and their exposure to peer group rejection
are determinants of psychological maladjustment.
References
Achenbach, T. M. (1991). Manual for the teacher’s report form
and 1991 profile. Burlington: Department of Psychiatry,
University of Vermont.
Asher, S. R., & Dodge, K. A. (1986). Identifying children
who are rejected by their peers. Developmental Psychology, 22, 442 – 449.
Asher, S. R., Rose, A. J., & Gabriel, S. W. (2001). Peer rejection in everyday life. In M. R. Leary (Ed.), Interpersonal
rejection (pp. 105 – 142). Oxford, UK: Oxford University
Press.
Asher, S. R., Singleton, L. C., Tinsley, B. R., & Hymel, S.
(1979). A reliable sociometric measure for preschool
children. Developmental Psychology, 15, 443 – 444.
Bagwell, C. L., Newcomb, A. F., & Bukowski, W. M. (1998).
Early adolescent friendship and peer rejection as predictors of adult adjustment. Child Development, 69,
140 – 153.
Boivin, M., & Hymel, S. (1997). Peer experiences and social
self-perceptions: A sequential model. Developmental
Psychology, 33, 135 – 145.
Boivin, M., Hymel, S., & Bukowski, W. M. (1995). The
roles of social withdrawal, peer rejection, and victimization by peers in predicting loneliness and depressed
mood in childhood. Development and Psychopathology, 7,
765 – 785.
Boldizar, J. P., Perry, D. G., & Perry, L. C. (1989). Outcome
values and aggression. Child Development, 60, 571 – 579.
Brown, B. B., Dolcini, M. N., & Leventhal, A. (1997).
Transformations in peer relationships at adolescence:
Implications for health-related behavior. In J. Schulenberg, J. L. Maggs, & K. Hurrelmann (Eds.), Health risks and
developmental transitions during adolescence (pp. 161 – 189).
New York: Cambridge University Press.
Buhrmester, D., & Furman, W. (1986). The changing functions of friends in childhood: A neo-Sullivanian perspective. In V. J. Derlega & B. A. Winstead (Eds.),
Friendship and social interaction (pp. 41 – 61). New York:
Springer-Verlag.
Buhs, E. S., & Ladd, G. W. (2001). Peer rejection as antecedent of young children’s school adjustment: An examination of mediating processes. Developmental
Psychology, 37, 550 – 560.
Bukowski, W. M., & Hoza, B. (1989). Popularity and
friendship: Issues in theory, measurement, and outcome.
In T. J. Berndt & G. W. Ladd (Eds.), Peer relationships in
child development (pp. 15 – 45). New York: Wiley.
843
Cairns, R. B., & Cairns, B. D. (1994). Lifelines and risks:
Pathways of youth in our time. New York: Cambridge
University Press.
Cairns, R. B., Cairns, B. D., Neckerman, H. J., Ferguson, L.
L., & Gariepy, J. L. (1989). Growth and aggression I:
Childhood to early adolescence. Developmental Psychology, 25, 320 – 330.
Cairns, R., Cairns, B., Neckerman, H. J., Gest, S., & Gariepy,
J. (1988). Peer networks and aggressive behavior: Peer
support or peer rejection? Developmental Psychology, 24,
815 – 823.
Caspi, A., Elder, G. H., & Bem, D. J. (1987). Moving against
the world: Life-course patterns of explosive children.
Developmental Psychology, 23, 308 – 313.
Caspi, A., Elder, G. H., & Bem, D. J. (1988). Moving away
from the world: Life-course patterns of shy children.
Developmental Psychology, 24, 824 – 831.
Chorpita, B. F., & Barlow, D. H. (1998). The development of
anxiety: The role of control in the early environment.
Psychological Bulletin, 124, 3 – 21.
Cillessen, A. H. N., & Bukowski, W. M. (Eds.) (2000). Recent
advances in the measurement of acceptance and rejection in
the peer system (New Directions for Child Development, No.
88). San Francisco: Jossey-Bass.
Coie, J. D. (1990). Toward a theory of peer rejection. In
S. R. Asher & J. D. Coie (Eds.), Peer rejection in childhood (pp. 365 – 401). New York: Cambridge University
Press.
Coie, J. D. (2004). The impact of negative social experiences on the development of antisocial behavior. In
J. B. Kupersmidt & K. A. Dodge (Eds.), Children’s
peer relations: From development to intervention (pp.
243 – 267). Washington, DC: American Psychological
Association.
Coie, J. D., & Kupersmidt, J. B. (1983). A behavioral analysis of emerging social status in boys’ groups. Child
Development, 54, 1400 – 1416.
Coie, J. D., Lochman, J. E., Terry, R., & Hyman, C. (1992).
Predicting early adolescent disorder from childhood
aggression and peer rejection. Journal of Consulting and
Clinical Psychology, 60, 783 – 792.
Coie, J. D., Watt, N. F., West, S. G., Hawkins, D., Asarnow,
J. R., Markman, H. J., et al. (1993). The science of prevention: A conceptual framework and some directions
for a national research program. American Psychologist,
48, 1013 – 1022.
Cole, D. A., & Maxwell, S. E. (2003). Testing mediational
models with longitudinal data: Questions and tips in the
use of structural equation modeling. Journal of Abnormal
Psychology, 112, 558 – 577.
Cowan, P. A., & Cowan, C. P. (2004). From family relationships to peer rejection to antisocial behavior in
middle childhood. In J. B. Kupersmidt & K. A. Dodge
(Eds.), Children’s peer relations: From development to intervention (pp. 159 – 177). Washington, DC: American Psychological Association.
Crick, N. R., & Ladd, G. W. (1990). Children’s perceptions
of the consequences of aggressive strategies: Do the ends
844
Ladd
justify being mean? Developmental Psychology, 26,
612 – 620.
DeRosier, M. E., Kupersmidt, J. B., & Patterson, C. J. (1994).
Children’s academic and behavioral adjustment as a
function of the chronicity and proximity of peer rejection. Child Development, 65, 1799 – 1813.
Dodge, K. A. (1983). Behavioral antecedents of peer social
status. Child Development, 54, 1386 – 1399.
Egan, S. E., Monson, T. C., & Perry, D. G. (1998). Social
cognitive influences on change in aggression over time.
Developmental Psychology, 34, 996 – 1006.
Freud, A., & Dann, S. (1951). An experiment in group
upbringing. In R. Eisler, A. Freud, H. Hartmann, &
E. Kris (Eds.), Psychoanalytic study of the child (Vol. 6,
pp. 127 – 168). New York: International University
Press.
Furman, W., Rahe, D., & Hartup, W. (1979). The rehabilitation of socially withdrawn preschool children through
mixed-age and same-age socialization. Child Development, 50, 915 – 922.
Gazelle, H., & Ladd, G. W. (2003). Anxious solitude and
peer exclusion: A diathesis-stress model of internalizing trajectories in childhood. Child Development, 74,
257 – 278.
Harlow, H. F., Harlow, M. K., Dodsworth, R. O., & Arling,
G. L. (1966). Maternal behavior of rhesus monkeys deprived of mothering and peer associations in infancy.
Proceedings of the American Philosophical Society, 110,
58 – 66.
Harrist, A. W., Zaia, A. F., Bates, J. E., Dodge, K. A., &
Pettit, G. S. (1997). Subtypes of social withdrawal in
early childhood: Sociometric status and social-cognitive
differences across four years. Child Development, 68,
278 – 294.
Hartup, W. (1970). Peer interaction and social organization.
In P. H. Mussen (Ed.), Carmichael’s manual of child psychology (Vol. 2, 3rd ed., pp. 360 – 456). New York: Wiley.
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indices
in covariance structure analysis: Conventional criteria
versus new alternatives. Structural Equation Modeling, 6,
1 – 55.
Hymel, S., Rubin, K. H., Rowden, L., & LeMare, L. (1990).
Children’s peer relationships: Longitudinal prediction
of internalizing and externalizing problems from middle
to late childhood. Child Development, 61, 2004 – 2021.
Kupersmidt, J. B., & Coie, J. D. (1990). Preadolescent peer
status, aggression, and school adjustment as predictors
of externalizing problems in adolescence. Child Development, 61, 1350 – 1362.
Kupersmidt, J. B., Coie, J. D., & Dodge, K. A. (1990). The
role of poor peer relationships in the development of
disorder. In S. R. Asher & J. D. Coie (Eds.), Peer rejection
in childhood (pp. 274 – 305). New York: Cambridge University Press.
Kupersmidt, J. B., & DeRosier, M. E. (2004). How peer
problems lead to negative outcomes: An integrative
mediational model. In J. B. Kupersmidt & K. A. Dodge
(Eds.), Children’s peer relations: From development to inter-
vention (pp. 119 – 138). Washington, D.C.: American
Psychological Association.
Ladd, G. W. (1989). Children’s social competence and social supports: Precursors of early school adjustment? In
B. Schneider, G. Attili, J. Nadel, & R. Weissberg (Eds.),
Social competence in developmental perspective (pp.
277 – 292). Amsterdam: Kluwer Academic Publishers.
Ladd, G. W. (1996). Shifting ecologies during the 5 – 7 year
period: Predicting children’s adjustment to grade
school. In A. Sameroff & M. Haith (Eds.) The five to seven
year shift (pp. 363 – 386). Chicago: University of Chicago
Press.
Ladd, G. W. (2003). Probing the adaptive significance of
children’s behavior and relationships in the school
context: A child by environment perspective. In R. Kail
(Ed.), Advances in child behavior and development (31, pp.
43 – 104). New York: Wiley.
Ladd, G. W. (2005). Children’s peer relationships and social
competence: A century of progress. New Haven, CT: Yale
University Press.
Ladd, G. W., Birch, S. H., & Buhs, E. (1999). Children’s
social and scholastic lives in kindergarten: Related
Spheres of Influence? Child Development, 70, 1373 –
1400.
Ladd, G. W., Buhs, E., & Seid, M. (2000). Children’s initial
sentiments about kindergarten: Is school liking an antecedent of early classroom participation and achievement? Merrill-Palmer Quarterly, 46, 255 – 279.
Ladd, G. W., & Burgess, K. B. (1999). Charting the relationship trajectories of aggressive, withdrawn, and aggressive/withdrawn children during early grade school.
Child Development, 70, 910 – 929.
Ladd, G. W., Herald, S., & Andrews, K. (2006). Young
children’s peer relations and social competence. In
B. Spodek & O. Saracho (Eds.), Handbook of research on
the education of young children (Vol. 2, pp. 23 – 54). New
York: Macmillan.
Ladd, G. W., & Mars, K. T. (1986). Reliability and validity of
preschoolers’ perceptions of peer behavior. Journal of
Clinical Child Psychology, 15, 16 – 25.
Ladd, G. W., & Profilet, S. M. (1996). The Child Behavior
Scale: A teacher-report measure of young children’s
aggressive, withdrawn, and prosocial behaviors. Developmental Psychology, 32, 1008 – 1024.
Ladd, G. W., & Troop-Gordon, W. (2003). The role of
chronic peer adversity in the development of children’s
psychological adjustment problems. Child Development,
74, 1325 – 1348.
McDougall, P., Hymel, S., Vaillancourt, T., & Mercer, L.
(2001). The consequences of childhood peer rejection. In
M. R. Leary (Ed.), Interpersonal rejection (pp. 213 – 247).
Oxford, UK: Oxford University Press.
Michael, C. M., Morris, D. P., & Soroker, E. (1957). Followup studies of shy, withdrawn children II. Relative
incidence of schizophrenia. American Journal of Orthopsychiatry, 27, 331 – 337.
Morris, D. P., Soroker, E., & Burruss, G. (1954). Follow-up
studies of shy, withdrawn children I: Evaluation of later
Predicting Maladjustment
adjustment. American Journal of Orthopsychiatry, 24,
743 – 754.
Moss, H. A., & Susman, E. J. (1980). Longitudinal study
of personality development. In O. G. Brim Jr. & J.
Kagan (Eds.), Constancy and change in human development
(pp. 530 – 595). Cambridge, MA: Harvard University
Press.
Olweus, D. (1979). Stability of aggressive reaction patterns in males: A review. Psychological Bulletin, 86, 852 –
875.
Parker, J. G., & Asher, S. R. (1987). Peer relations and later
personal adjustment: Are low-accepted children ‘‘at
risk’’? Psychological Bulletin, 102, 357 – 389.
Parker, J. G., Rubin, K. H., Price, J. M., & DeRosier, M.
(1995). Peer relationships, child development, and adjustment: A developmental psychopathology perspective. In D. Cicchetti & D. Cohen (Eds.), Developmental
psychopathology: Vol. 2. Risk, disorder and adaptation (pp.
96 – 161). New York: Wiley.
Pomerantz, E. M., & Rudolph, K. D. (2003). What ensues
from emotional distress? Implications for competence
estimation. Child Development, 74, 329 – 345.
Renshaw, P., & Brown, P. J. (1993). Loneliness in middle
childhood: Concurrent and longitudinal predictors.
Child Development, 64, 1271 – 1284.
Rensvold, R. B., & Cheung, G. W. (2001). Testing for metric
invariance using structural equations models: Solving
the standardization problem. In C. A. Schriesheim &
L. L. Neider (Eds.), Research in management: Vol 1.
Equivalence in measurement (pp. 21 – 51). Greenwich, CT:
Information Age.
Robins, L. N. (1966). Deviant children grow up. Baltimore:
Williams & Wilkins Press.
Roff, J. D., & Wirt, D. (1984). Childhood aggression and
social adjustment as antecedents of delinquency. Journal
of Abnormal Child Psychology, 12, 111 – 126.
Roff, M. (1961). Childhood social interaction and young
adult bad conduct. Journal of Abnormal Social Psychology,
63, 333 – 337.
Roff, M. (1963). Childhood social interactions and young
adult psychosis. Journal of Clinical Psychology, 19,
152 – 157.
Roff, M., Sells, S. B., & Golden, M. M. (1972). Social adjustment and personality development in children. Minneapolis: University of Minnesota Press.
Rubin, D. B. (1987). Multiple imputation for nonresponse in
surveys. New York: Wiley.
Rubin, K. H. (1982). Nonsocial play in preschoolers: Necessary evil? Child Development, 53, 651 – 657.
Rubin, K. H. (1993). The Waterloo Longitudinal Project:
Correlates and consequences of social withdrawal from
childhood to adolescence. In K. H. Rubin & J. B. Asendorpf (Eds.), Social withdrawal, inhibition, and shyness in
childhood (pp. 291 – 314). Hillsdale, NJ: Lawrence Erlbaum.
Rubin, K. H., & Asendorpf, J. (1993). Social withdrawal, inhibition, and shyness in childhood. Hillsdale, NJ: Lawrence
Erlbaum.
845
Rubin, K. H., & Coplan, R. (2004). Paying attention to and
not neglecting social withdrawal and social isolation.
Merrill-Palmer Quarterly, 50, 506 – 534.
Rubin, K. H., LeMare, L., & Lollis, S. (1990). Social withdrawal in childhood: Developmental pathways to peer
rejection. In S. R. Asher & J. D. Coie (Eds.), Peer rejection
in childhood (pp. 217 – 249). New York: Cambridge University Press.
Rudolph, K. D., & Asher, S. R. (2000). Adaptation and
maladaptation in the peer system: Developmental
processes and outcomes. In A. J. Sameroff, M. Lewis, &
S. Z. Miller (Eds.), Handbook of developmental psychopathology (pp. 157 – 175). New York: Kluwer Academic/
Plenum.
Rutter, M. (1996). Transitions and turning points in developmental psychopathology: As applied to the age span
between childhood and mid-adulthood. International
Journal of Behavioral Development, 19, 603 – 626.
Sameroff, A. J. (1987). The social context of development.
In N. Eisenberg (Ed.), Contemporary topics in developmental psychology (pp. 273 – 291). New York: Wiley.
Sameroff, A. J., & Chandler, M. J. (1975). Reproductive
risk and the continuum of caretaking causality. In F. D.
Horowitz (Ed.). Review of child development research
(Vol. 4, pp. 187 – 244). Chicago: University of Chicago
Press.
Sandstrom, M. J., & Zakriski, A. L. (2004). Understanding
the experience of peer rejection. In J. B. Kupersmidt &
K. A. Dodge (Eds.), Children’s peer relations: From development to intervention (pp. 101 – 118). Washington, DC:
American Psychological Association.
Schafer, J. L. (1999). NORM: Multiple imputation of incomplete multivariate data under a normal model, version 2. Software for Windows 95/98/NT, available from
http://www.stat.psu.edu/ jls/misoftwa.html.
Schafer, J. L., & Olsen, M. K. (1998). Multiple imputation for multivariate missing-data problems: A data
analyst’s perspective. Multivariate Behavioral Research, 33,
545 – 571.
Slaby, R. G., & Guerra, N. G. (1988). Cognitive mediators of
aggression in adolescent offenders: I. Assessment. Developmental Psychology, 24, 580 – 588.
Sullivan, H. S. (1953). The interpersonal theory of psychiatry.
New York: W.W. Norton & Company.
Suomi, S. J., & Harlow, H. F. (1972). Social rehabilitation of
isolate reared monkeys. Developmental Psychology, 6,
487 – 496.
Tremblay, R. E., Pihl, R. O., Vitaro, F., & Dobkin, P. L. (1994).
Predicting early onset of male antisocial behavior from
preschool behavior. Archives of General Psychiatry, 51,
732 – 739.
Troop-Gordon, W., & Ladd, G. W. (2005). Trajectories of
peer victimization and perceptions of the self and
schoolmates: Precursors to internalizing and externalizing problems. Child Development. 76, 1072 – 1091.
Wanlass, R. L., & Prinz, R. J. (1982). Methodological issues
in conceptualizing and treating childhood social isolation. Psychological Bulletin, 92, 39 – 55.
846
Ladd
Wright, J. C., Giammarino, M., & Parad, H. W. (1986). Social status in small groups: Individual-group similarity
and the social ‘‘misfit’’. Journal of Personality and Social
Psychology, 50, 523 – 536.
Younger, A. J., & Boyko, K. A. (1987). Aggression and
withdrawal as social schemas underlying children’s
peer perceptions. Child Development, 58, 1094 – 1100.
Younger, A. J., Gentile, C., & Burgess, K. B. (1993). Children’s perceptions of social withdrawal: changes across
age. In K. Rubin & J. Asendorpf (Eds.), Social withdrawal,
inhibition, and shyness in childhood (pp. 215 – 235). Hillsdale, NJ: Lawrence Erlbaum.