Sei sulla pagina 1di 22

Child Development, March/April 1999, Volume 70, Number 2, Pages 513534

Parental Reactions to Childrens Negative Emotions: Longitudinal Relations


to Quality of Childrens Social Functioning
Nancy Eisenberg, Richard A. Fabes, Stephanie A. Shepard, Ivanna K. Guthrie,
Bridget C. Murphy, and Mark Reiser

Relations between self-reported parental reactions to childrens negative emotions (PNRs) and childrens so-
cially appropriate/problem behavior and negative emotionality were examined longitudinally. Evidence was
consistent with the conclusion that relations between childrens externalizing (but not internalizing) emotion
and parental punitive reactions to childrens negative emotions are bidirectional. Reports of PNRs generally
were correlated with low quality of social functioning. In structural models, mother-reported problem behav-
ior at ages 1012 was at least marginally predicted from mother-reported problem behavior, childrens regula-
tion, and parental punitive or distress reactions. Moreover, parental distress and punitive reactions at ages 68
predicted reports of childrens regulation at ages 810, and regulation predicted parental punitive reactions at
ages 1012. Father reports of problem behavior at ages 1012 were predicted by earlier problem behavior and
parental distress or punitive reactions; some of the relations between regulation and parental reactions were
similar to those in the models for mother-reported problem behavior. Parental perceptions of their reactions
were substantially correlated over 6 years. Some nonsupportive reactions declined in the early to mid-school
years, but all increased into late childhood/early adolescence.

INTRODUCTION situations due to prior repeated associations between


punishment and emotional expressivity. Such anxiety
Although socialization has been a central topic in de-
might be reflected in the presence of internalizing
velopmental psychology for decades, only recently
emotion. Analogously, Tomkins (1963) suggested that
have psychologists focused much attention on how
children learn to express distress without shame and to
parents socialize childrens emotion-related capaci-
respond sympathetically to others if their parents re-
ties and the relation of such socialization to childrens
spond openly with sympathy and nurturance to chil-
socioemotional competence (e.g., Denham, 1993; Dunn
drens feelings of distress and helplessness.
& Brown, 1994; Garner, Jones, & Miner, 1994; Parke,
Eisenberg, Fabes, & Murphy. (1996) further sug-
Cassidy, Burks, Carson, & Boyum, 1992). Albeit infre-
gested that children who receive nonsupportive reac-
quently examined, an important aspect of emotion
tions (e.g., punitive or minimizing responses) to their
socialization may be parents negative reactions to
negative emotion are likely to remain emotionally
childrens experience and displays of normally occur-
aroused and become dysregulated in their behavior
ring negative emotion. Examples of such parental re-
when they experience negative emotions. Parental
actions include parental experience of aversive, self-
practices that heighten or extend childrens negative
focused emotion (which can lead to the expression of
emotional arousal, both in the specific context and in
negative emotion and parental avoidance) and ac-
future situations, would be expected to undermine
tions that punish or minimize the legitimacy of the
learning about emotions and their management dur-
childs emotional experience (e.g., Eisenberg & Fabes,
ing emotional events (Hoffman, 1983). In addition,
1994; Kliewer, Fearnow, & Miller, 1996).
the quality of parents reactions to childrens negative
Parents negative reactions to childrens experience
emotion likely affects childrens emotional security
of emotions such as distress, fear, anxiety, and sadness
and feelings about social interactions in general that, in
generally are believed to be associated with negative
turn, may influence the quality of their emotional re-
socioemotional outcomes for children (Eisenberg,
sponses and behavior in social encounters. Albrecht,
Fabes, & Murphy, 1996; Gottman, Katz, & Hooven,
Burleson, and Goldsmith (1994) suggested that nur-
1996). For example, Buck (1984) hypothesized that
turant communications during socialization encoun-
sanctions for emotional expressiveness are associated
ters foster a sense of security and self-confidence, a
with physiological but not external markers of emo-
secure attachment, and positive working models of
tional responding in adults. This purportedly is be-
relationships with others (also see Davies & Cum-
cause children who receive negative reactions to their
displays of emotion gradually learn to hide their emo- 1999 by the Society for Research in Child Development, Inc.
tions but feel anxious when in emotionally evocative All rights reserved. 0009-3920/99/7002-0018
514 Child Development

mings, 1994). Similarly, Bryant (1987) argued that dis- not affect childrens regulationa finding that sug-
tressing experiences increase childrens self-concern gests that how parents deal with children around
and can undermine the development of satisfying so- emotional issues is not simply a reflection of their
cial relationships; therefore, parental buffering of chil- more general style of interaction in teaching or social-
dren from stress (e.g., through acknowledging and le- ization contexts.
gitimizing childrens feelings of distress) may enhance In a recent study of parents reported reactions to
childrens empathic and prosocial development. One third through sixth graders negative emotions (Eisen-
would expect negative or punitive interactions to have berg, Fabes, & Murphy, 1996), maternal minimizing
the reverse effect. of childrens negative emotions or punitive reactions
Gottman, Katz, and Hooven (1997) have specifi- were related to low levels of childrens socially appro-
cally implicated childrens arousal and self-regulatory priate behavior, constructive coping, and teacher-
skills in the process of emotion socialization and in reported popularity, and high levels of mother-reported
individual differences in childrens emotional and avoidant coping. However, maternal reports of dis-
social competence. They argued that parental philoso- tressed or upset reactions were not correlated with
phies about the expression of emotion affect parents childrens coping or social competence. Moreover,
inhibition of negative affect toward the child and fa- there were few significant findings for fathers.
cilitate positive parenting, as well as directly affect In a study with 4- to 6-year-olds (the same sample
childrens regulatory physiology and ability to regu- as in the present study), parental negative reactions
late their own emotions. In their view, parents sup- were correlated with childrens real-life reactions when
portive coaching of children in regard to their emo- angered or frustrated (Eisenberg & Fabes, 1994). Pu-
tions is nested within a web of warm parenting (p. nitive and minimizing reactions (combined) were as-
90) and does not especially promote the childs ex- sociated with low venting of negative emotion and
pressiveness. Rather, parental coaching helps children high levels of escape behavior when angered; paren-
to develop the ability to inhibit negative affect, to self- tal distressed/upset reactions to their children nega-
sooth, and to focus attention (including attention in tive emotions also were associated with low levels of
social contexts). These skills are expected to enhance venting. However, parental reactions were not exam-
socially skilled rather than unregulated problem be- ined in relation to quality of childrens social behavior
havior. In contrast, children whose parents dismiss or problem behavior.
negative emotions and do not talk about them with A primary goal of the present longitudinal study
children in a supportive way are expected to be rela- was to examine the relation of parental reports of neg-
tively low in the ability to manage both their emo- ative reactions to childrens negative emotions with
tions and their attention in social situations. This is es- childrens socially appropriate and problem behavior,
pecially true if parents belittle or mock the child. both concurrently and across time. Primary caregiv-
Empirical evidence for the aforementioned asser- ing parents (usually mothers) reports of their reac-
tions is mixed. In general, punitive, nonsupportive tions to childrens negative emotions such as anxiety
parenting (or parental report of such parenting) has and sadness in everyday contexts were obtained at
been associated with negative outcomes and problem several points in time, as were teachers, mothers,
behavior for children (Gottman et al., 1996; Maccoby and fathers reports of quality of social functioning
& Martin, 1983; Parke & Buriel, 1998; Patterson, (socially appropriate and/or externalizing problem
1982). Of direct relevance, Carson and Parke (1996) behavior). Thus, unlike in previous work, we could
found that children whose fathers tended to respond examine the direction of relations between parental
with negative emotion to their childrens displays of reactions and the childs behavior and characteristics
negative emotion were less socially skilled (i.e., (i.e., negative emotionality or regulation) over time.
shared less, avoided others more, and were more ag- Based on the aforementioned reasoning, parental
gressive) than were their preschool peers. Moreover, nonsupportive reactions, especially those that were
Gottman et al. (1996, 1997) found that parents who punitive, were expected to predict externalizing prob-
were relatively aware and supportive of the expres- lem behavior and inappropriate social behavior. It
sion of emotion (rather than the reverse) had children was further hypothesized that at least part of the rela-
who were physiologically well regulated and there- tion of PNRs (parental negative reactions to chil-
fore could manage their emotional arousal in social drens negative emotions) to quality of social func-
contexts. In addition, well-regulated children were tioning would be indirect through their effect on
less aggressive with peers. Also of interest, Gottman childrens regulation. If Buck (1984) is correct in his
et al. (1996) found that general parenting, as reflected theorizing, children who receive negative reactions to
in parental derogation or scaffolding/praising, did their expression of normal negative emotion are likely
Eisenberg et al. 515

to become overly emotionally aroused and, hence, of negative emotions. Parents who typically react
less regulated, in situations involving negative emo- negatively to childrens expression of normal dis-
tion. As discussed previously, Gottman et al. (1996) plays of negative emotion (especially those that do
argued that parental coaching in regard to emotion not harm others) may elicit anger and frustration
and support of emotion directly affects childrens reg- from children. Of course, children who are prone to
ulatory physiology and ability to regulate their own intense and frequent negative emotions also may
emotions. Although they found evidence consistent elicit relatively high levels of parental distress, upset,
with this view, in their structural equation models and punitive and minimizing reactions in response to
physiological regulation also predicted parental coach- these emotions.
ing when the path was reversed. Because Gottman et Indeed, it is likely that the relation between chil-
al.s measures of regulation and parental coaching drens dispositional negative emotionality and PNRs
were taken at the same point in time, they could not is bidirectional. Given that negative emotionality ap-
examine whether earlier parental variables predicted pears to have a strong temperamental basis (Emde et
later child regulation, if early child regulation pre- al., 1992; Plomin & Stocker, 1989), it is logical to expect
dicted later parental coaching, or whether there ap- childrens negative emotionality at a young age to have
peared to be bidirectional causation (i.e., both paths some effect on parents negative reactions to chil-
held across time). drens emotions. As evidence of this, in one study
Based on prior theory and findings such as those mothers engaged in more comforting of their toddlers
obtained by Gottman et al. (1996), parental reactions if the children avoided a fear-eliciting stimulus, a
to childrens negative emotions were hypothesized to finding consistent with the view that mothers altered
influence childrens regulation, which would be ex- their behavior as a function of the childs fearfulness
pected to affect the quality of childrens social behav- (Diener, Mangelsdorf, Fosnot, & Kienstra, 1997; see
ior at home and school (i.e., childrens regulation was also Fabes et al., 1994). Moreover, Cook, Kenny, and
expected to mediate the effects of PNRs on social Goldstein (1991) found that adolescents expression
functioning). Of course, it also is plausible that indi- of negative affect toward their parents seemed to
vidual differences in childrens regulation influence elicit negative parental affect directed toward them.
the degree to which parents react negatively to chil- However, as children develop, it also is likely that
drens negative emotions, which might then affect their negative emotionality is influenced to some de-
childrens social behavior (Bell, 1977; see Bugental & gree by interactions with parents. Belsky et al. (1991)
Goodnow, 1998). Thus, bidirectional relations be- found that infants who changed from high to low
tween PNRs and childrens regulation were predicted negative emotionality from 3 to 9 months of age had
and we hypothesized that both childrens regulatory greater than expected frequencies of complementary,
capacities and PNRs would have unique effects on responsive, and sensitive interactions with mothers
the quality of childrens social behavior. Relations and fewer than expected disharmonious, unrespon-
from PNRs to regulation, and vice versa, were pitted sive, and poorly coordinated exchanges. The reverse
against one another, as was the possibility that child was true of mother-infant dyads in which the infant
regulation and PNRs were correlated at each time pe- was highly distressed at ages 3 and 9 months (also see
riod, by including all of these possibilities in struc- Matheny, 1986). Thus, it was predicted that the rela-
tural equation models with longitudinal data. Thus, tion between PNRs and school-aged childrens nega-
this is one of the few studies in which bidirectional tive emotionality would be bidirectional over time.
parent-child effects were examined over time. Moreover, because childrens externalizing nega-
A second goal was to examine the relation of chil- tive emotions (e.g., anger, hostility) are likely to be
drens dispositional negative emotionality to PNRs. more aversive and worrisome to parents than are
These two constructs have been found to be posi- childrens internalizing negative emotions, relations
tively correlated when measured concurrently (Eisen- of parent-reported PNRs with childrens negative
berg & Fabes, 1994; Eisenberg, Fabes, & Murphy, emotionality were expected to be more consistent for
1996; cf. Gottman et al., 1996). Some psychologists childrens externalizing than internalizing negative
have noted that despite the stability obtained for emotions.
measures of temperament (including negative emo- The third goal of the present study was to examine
tionality), it is important to identify environmental interindividual consistency over time in parents re-
factors that influence change in childrens expression ports of various reactions to childrens negative emo-
of temperament (Belsky, Fish, & Isabella, 1991; Gott- tions, as well as change in mean levels of parents re-
man et al., 1997; Thomas, 1984). One such factor is ports of their reactions. Parents reported reactions
likely to be parental reactions to childrens expression were expected to be relatively consistent over time
516 Child Development

due to consistencies in their own philosophies about tionality sometimes predicted negative outcomes for
emotions, in their emotion-related temperamental or boys but not girls (e.g., Eisenberg, Fabes, Murphy,
personality characteristics, and in childrens temper- Karbon, et al., 1996). Thus, sex differences were exam-
amentally based emotionality, which could serve to ined in some analyses.
elicit consistent parental responding (Murphy, Eisen-
berg, Fabes, Shepard, & Guthrie, 1998; Plomin &
Stocker, 1989). Parents beliefs about their own reac-
METHOD
tions likely are more consistent than their actual be-
havior; nonetheless, consistency in parental beliefs Participants. Participants were involved in a longi-
about their own child rearing reactions probably is re- tudinal study that included five assessments, two of
lated to consistency in actual behavior. which were close in time and combined (see Eisen-
We were not sure what to hypothesize in regard to berg, Fabes, Nyman, Bernzweig, & Pinuelas, 1994;
changes in mean level of parental negative reactions Eisenberg et al., 1995, 1997; Eisenberg & Fabes, 1994).
with age of the child. On the one hand, it is possible The initial sample of 94 children was studied for two
that parents negative reactions become less negative consecutive semesters (henceforth called T1 and T2)
or less frequent in occurrence as children become when in preschool or kindergarten (the kindergarten
more regulated, especially when moving from the was at a preschool; see Eisenberg, Fabes, Nyman, et
preschool into the elementary school years. Parents in al. 1994). The first follow-up was 2 years after the be-
this sample reported that their children displayed ginning of T1 (henceforth called T3); at this time the
somewhat less negative emotionality with age in the sample included 82 kindergarten to second grade
elementary school years (Murphy et al., in press); children. The families were followed up again 2 and 4
thus, it was reasonable to hypothesize that parental years later (at T4 and T5). At T4, participants were 77
nonsupportive/negative reactions decline in fre- children (36 girls, 41 boys). Sixty-five came to the lab-
quency as children age. On the other hand, parents oratory; data were obtained by mail from teachers
may feel that with age it is increasingly important for and parents for 12 more children (5 girls, 7 boys). At
children to regulate their negative emotions and, con- T5, all data were obtained by mail from either parents
sequently, parents may become less tolerant of nega- (61) or teachers (59) on 64 children (33 girls, 31 boys).
tive emotional reactions as children develop. In either Only children with parental reports of socializa-
case, early adolescence may be a time in which paren- tion of negative emotion could be included in the
tal reactions become somewhat more negative due to analyses in the present study. Numbers of children
increases in both negative conflictual exchanges be- with these data were 79 at T1/T2 and T3 (37 girls, 42
tween parents and children (Collins, 1990) and chil- boys), 71 at T4 (34 girls, 37 boys), and 58 at T5 (31
drens negative emotionality (Larson & Richards, girls, 27 boys). Parental reports of their reactions to
1991; Murphy et al., 1998). Thus, a quadratic pattern childrens negative emotions were available for at
of change was predicted such that PNRs were ex- least one time period for 90 children (44 girls, 46
pected to decrease with age in middle childhood boys). Mean ages for children with the focal paren-
and then start to increase as children move into pre- tal reactions measure were 58 months at T1/T2
and early adolescence. (halfway through; SD 5 7), 88 months at T3 (SD 5
Although gender differences generally were not 7), 107 months at T4 (SD 5 7), and 132 months at T5
predicted, one reasonable possibility was that the pat- (SD 5 7). The age range of the children at each time
terns of relations would differ for fathers reports of period was approximately 2 years, although there
boys and girl problem behavior in late childhood. were relatively few children at the extremes in
When children were 68 and 810 years old, relations terms of age.
of negative emotionality and regulation with father- The children participating were predominantly
reported problem behaviors held primarily for boys Caucasian. For example, the sample of 79 children
(whereas relations were found for both sexes for with parental data on reactions at T1/T2 was 89%
mothers reports of problem behavior; Eisenberg et Caucasian; 3% black, 4% of Asian heritage, 1% His-
al., 1995, 1997). It is possible that boys externalizing panic, and 3% of mixed origin. Mean numbers of
negative emotion and unregulated behavior are more years of maternal and paternal education at each of
overt or evocative for adults than are girls. Moreover, the time periods were approximately 16.5 (the range
fathers may be more concerned about, or restrictive of Ms across the four time periods was 16.46 to 17.12)
with, sons than daughters behavior (Power & Shanks, and 17.5 years (the range of Ms across time periods
1989; Rothbart & Maccoby, 1966). In addition, in anal- was 17.65 to 17.69), respectively. Family income
yses with this sample at a younger age, negative emo- ranged from $15,000 to $200,000 (M 5 $70,000, SD 5
Eisenberg et al. 517

$37,000) at the first assessment, and from $12,000 to provided no other data. Teacher questionnaires per-
$200,000 (M 5 $80,000; SD 5 $31,000) at the last as- taining to childrens socially appropriate behavior
sessment (see Eisenberg et al., 1995, 1997, for more in- and problem behavior were sent out near the end of
formation on the sample at T3 and T4).1 the semester at T3, T4, and T5.

Procedure
Measures
As discussed previously, the first data collection
Parents Reactions to Childrens Negative Emotions
for this sample took place over two academic semes-
ters (T1 and T2); this data collection preceded later as- In the present study, parents reports of their nega-
sessments by 2 (T3), 4 (T4), and 6 (T5) years. Measures tive reactions to childrens negative emotion were ob-
of social functioning, regulation, and negative emo- tained. Parents reports, rather than observational
tionality were obtained at all time periods. However, data, were used because it would be quite difficult to
because relations between anger behavior or regula- obtain sufficient observations of real-life PNRs (espe-
tion and parental reactions at T1/T2 were reported pre- cially for families who had moved away). Parental re-
viously (Eisenberg & Fabes, 1994), and the measures of actions to negative emotions probably are expressed
regulation, social functioning, and negative emotion- relatively infrequently, especially in response to emo-
ality differed somewhat between T1/T2 and later as- tions besides anger and when family members are be-
sessments, analyses of the variables of regulation and ing observed. Self-report measures of parental social-
negative emotionality in the present study involved ization appear to be most valid when parents are
primarily T3, T4, and T5 data. T1/T2 data were used asked to respond to questions about their behavior in
only in correlations of T1/T2 parental reactions data specific situations (Holden & Edwards, 1989). This
with problem behaviors at older ages, in correlational was the format of the questionnaire measure used in
analyses of T2 internalizing negative emotion with this study.
parental reactions scores, and when examining con- Parental reactions to their childrens negative emo-
sistency in PNRs over time. tion were assessed with the Coping with Childrens
At T1/T2, parent reporters were mothers. At T3 Negative Emotions Scale (CCNES; Fabes, Eisenberg,
and T4, mothers usually completed the questionnaire & Bernzweig, 1990; see also Eisenberg & Fabes, 1994;
measure of PNRs and the measures of childrens neg- Eisenberg, Fabes, & Murphy, 1996). Parents (usually
ative emotionality, regulation, and problem behavior mothers) were presented with 12 typical situations in
(see Eisenberg et al., 1995, for more detail about T3). which children are described as experiencing distress
However, at T3 and T4, two and one fathers, respec- and negative affect (e.g., being scared of injections,
tively, completed the entire packet of measures. At T5, being nervous about possibly embarrassing him/
questionnaires were mailed to all families and teach- herself in public). All situations involved relatively
ers (with parental permission); all parent respon- normative expressions of negative emotion for chil-
dents except five who provided data on regulation, dren aged 4 to 12; only one vignette pertained to anger
emotionality, and PNRs were mothers. Father data (about being sick and unable to go to a party). For
were kept in the sample because the few fathers who each situation, parents were asked to indicate how
responded generally were the primary care-givers likely (on a 7-point scale from very unlikely to very
(we obtained mothers reports if they were the pri- likely) they would be to react in each of six different
mary care-giving parent but did not come to the lab- ways. The three types of responses included in this
oratory).2 At T3, T4, and T5, both parents were asked study are as follows (alpha coefficients are in paren-
to provide data on problem behaviors, even if they theses): (1) parental distress reactions, reflecting the de-
gree to which parents experience negative emotion
1 T-tests were used to compare study participants who had
when children express negative affect (e.g., feel upset
parental reactions measures at T5 with those who did not (i.e.,
and uncomfortable because of my childs reaction;
had dropped from the sample or had missing data on the target
parent reactions measure) on all of the major variables used in NOT be annoyed with my child for being rude [re-
the analyses. There were no significant differences, although verse coded]; .79, .79, .77, and .73 for T1/T2, T3, T4,
children without data at T5 were marginally higher on teacher- and T5, respectively), (2) punitive responses, indicating
reported internalizing negative emotion at T4, p , .08. the degree to which parents respond with punitive re-
2 T-tests were used to compare the five fathers respondents
actions that decrease their exposure or need to deal
with mother respondents at T5 on all variables. All ts were less
than .84 with the exception that the t for parental minimizing with the negative emotions of their children (tell my
was 1.62, ns. Thus, these fathers did not seem to differ from the child that if he/she starts crying then shell have to go
mother respondents. to her room right away; .78, .75, .68, and .71), and (3)
518 Child Development

minimization responses, reflecting the degree to which work; eight items, as 5 .90, .85, and .89 at T3, T4, and
parents minimize the seriousness of the situation or T5), and aggression (This child says mean things to
devalue the childs problem or distressed reaction peers, such as teasing or name calling; eight items,
(e.g., tell my child that he/she is overreacting; .86, as 5 .93, .94, and .90 at T3, T4, and T5). Items were
.83, .72, and .83). Although parental distress reactions rated from 1 5 never to 7 5 almost always. Given
are not always observed by children, over time it is that the aggression and disruptive subscales were
likely that children sometimes observe or intuit their highly intercorrelated (rs ranged from .69 .82), es-
parents negative emotion; thus, parental distressed/ pecially at T5, they were standardized and aver-
upset reactions can be expected to have some impact aged at each assessment.
on children. Items in the scales were averaged. The aggression/disruptive behavior composite
The CCNES scales used in this study have been was highly negatively correlated with the index of so-
found to have good internal and test-retest reliability cially appropriate behavior at T3, T4, and T5 (rs 5 .76,
(Eisenberg & Fabes, 1994; Eisenberg, Fabes, & Mur- ps , .001, at all times). To further reduce the data,
phy, 1996). Additionally, in regard to concurrent and standardized scores for aggression/disruptive be-
construct validity, the negative responses scales of the havior were subtracted from those for socially appro-
CCNES have been found to correlate in the predicted priate behavior to construct an aggregate index of so-
manner with other conceptually relevant indexes. cial functioning. This score was used in all analyses.
Specifically, nonsupportive parental reactions were
inversely related to parental empathy, and positively
Dispositional Negative Emotionality
related to measures of personal distress, parental anger,
harsh parental control, and parental control of emo- Internalizing negative emotion. At T2, teachers and
tions, usually at r 5 .30 or higher (Poulin, 1997). Fur- mothers (all respondents were mothers at T2) rated
thermore, scores on this questionnaire have been re- childrens emotional intensity with eight items adapted
lated to measures of quality of social functioning in from Larsen and Dieners (1987) Affective Intensity
prior research (Eisenberg & Fabes, 1994; Eisenberg, Measure. Two items pertained primarily to internaliz-
Fabes, Carlo, & Karbon, 1992; Eisenberg, Fabes, & ing emotions (This [my] child tends to get nervous,
Murphy, 1996). tense, or distressed easily and When this [my] child
feels anxiety, it normally is very strong; rated from
1 5 never to 7 5 always). Frequency of negative emo-
Quality of Social Functioning
tionality was assessed with items adapted from
Parents. Mothers and fathers completed the Loch- Derryberry and Rothbarts (1988) scale pertaining to
man et al. (1995) Child Problem Behavior Checklist. autonomic reactivity (four items; e.g., This [my] childs
All 24 items except one pertaining to setting fires palms usually sweat during an important event.),
were included (e.g., teases other children, breaks sadness (five items; e.g., This [my] child rarely feels
things on purpose, defiant toward adults). Items sad, even when things are not going well for him/
were rated on a 4-point scale (1 5 never to 4 5 often). her.), and fear (five items; e.g., This [my] child often
Alphas for mothers and fathers were .86 and .87 at T3, worries about things that turn out to be unimpor-
.91 and .90 at T4, and .91 and .91 at T5. tant.) These 16 internalizing items were combined
Teachers. Teachers reports of social functioning into an internalizing scale (as 5 .90 for teachers and
pertained to socially appropriate, aggressive, or dis- .74 for mothers).
ruptive behavior. These behaviors were chosen be- At T3, T4, and T5, teachers and parents rated chil-
cause they seemed most likely to depend on childrens drens emotional intensity with eight items adapted
regulation of emotion. from Larsen and Diener (1987). Two items pertained
To assess childrens socially appropriate behavior, primarily to internalizing emotion (When this [my]
teachers completed four items (e.g., This child is child gets nervous or distressed, he/she gets very
usually well-behaved versus This child is not well- nervous/upset and When this [my] child feels anx-
behaved) using a 4-point response scale (i.e., se- iety, it normally is very strong; rated from 1 5 never
lected a statement and then indicated if the item was to 7 5 always). In addition, adults reported on chil-
really true or sort of true; as 5 .76, .84, and .86 at drens dispositional affectivity with 15 items slightly
T3, T4, and T5; see Eisenberg et al., 1995, 1997). modified from the PANAS (Positive and Negative Af-
At T3, T4, and T5, teachers completed subscales fect Schedule; Watson, Clark, & Tellegen, 1988). Re-
from Coie, Terry, Dodge, and Underwoods (1993) spondents were asked to report how this (your) child
Teacher Checklist, including disruptive behavior (This feels on the average (1 5 very slightly or not at all, to
child bothers others kids when they are trying to 5 5 extremely). Childrens dispositional internalizing
Eisenberg et al. 519

negative emotion was assessed using six emotions used to reduce the burden on mothers). Items were
(afraid, nervous, scared, jittery, sad, and distressed). rated on a 7-point scale. To create a purer scale of be-
Because the emotional intensity and PANAS items havioral regulation, items concerning misbehavior
were rated on different scales, each item was stan- (e.g., breaking rules), clearly negative behaviors in
dardized and then the eight items were combined peer interaction (e.g., disrupting games), and aca-
into a single internalizing scale. Alphas for teachers demic performance were dropped. The remaining 11
and parents were .86 and .85 at T3; .88 and .83 at T4; items pertained to childrens ability to inhibit behav-
and .83 and .80 at T5. ior (e.g., Does your child think before he or she
Externalizing negative emotion. At T3, T4, and T5, acts?, Does your child sit still?), although a few
teachers and parents reports of childrens externaliz- items could pertain to attentional control (as at T4
ing negative emotion were assessed with one item and T5 5 .84).
from the emotional intensity scale (When this [my] As described in Eisenberg et al. (1995, 1997), paren-
child gets angry, it is easy for him/her to still be ratio- tal measures of behavioral inhibition, impulsivity (re-
nale and not overreact.) and three items from the versed), attention focusing, and self-control grouped
PANAS (hostile, irritable, and angry). These items together in principle components factor analyses
were reversed when appropriate, standardized, and with a varimax rotation at T3, T4, and T5, and were
averaged (as 5 .85, .93, and .76 at T3, T4, and T5 for averaged (after standardizing). (Attention shifting
teachers and .75, .80, and .82 for parents). was not used in the parental regulation composite be-
cause it did not factor with the other indexes of regu-
lation [see Eisenberg et al., 1995].)
Regulation
Parental report data on childrens regulation at T3,
T4, and T5 were used in the structural equation mod- RESULTS
els (SEMs). In earlier work with this data set, compos-
Consistency of Reported Parental Reactions
ites of regulation were formed based on correlations
Across Time
and factor analyses (Eisenberg et al., 1995, 1997); these
composites were used in the present study and are Consistency of parental reactions across time
described briefly. was examined with correlational analyses. Because
For parents, the regulation composite included at- fathers occasionally supplied these data, reporters
tention focusing and several measures of behavioral occasionally varied across time, which would be
regulation. As is discussed later, teachers reports of expected to lower consistency slightly. Age at the
regulation did not correlate with PNRs and, thus, time of assessment (within each time period) was
were not tested as mediators in the SEMs. Hence, they not significantly related to any of the variables in
are not presented in the method (see Eisenberg et al., the study; thus, it was not considered further in any
1995, for information on teacher-reported regulation). analyses.
Only as not presented in Eisenberg et al. (1995) are There was considerable consistency in reported re-
provided below. actions across time, even over a 6-year period. Corre-
At T3, T4, and T5, parents rated childrens attention lations for parental reactions ranged from .33 to .75,
focusing on a 7-point scale (from 1 5 extremely untrue with all correlations significant at p , .01 or better
to 7 5 extremely true) with 10 items from the CBQ (see Table 1).
(Rothbart, Ahadi, & Hershey, 1994; e.g., Has difficulty
leaving a project he/she has begun, as 5 .72 and .75 at
Sex Differences and Changes with Age
T4 and T5; see Eisenberg et al., 1995, for T3 as).
To assess aspects of behavioral regulation, at T3, Two (sex) 3 4 (time of assessment) repeated mea-
T4, and T5 parents also were administered the follow- sures multivariate analyses of variance (MANOVAs)
ing 13-item subscales of the CBQ (see Eisenberg et al., were used to examine sex differences and change in
1995, for T3 alphas): impulsivity (e.g, Usually rushes parents reported practices over time. None of the ef-
into an activity without thinking about it; as 5 .83 at fects involving sex approached significance. The multi-
both T4 and T5), and inhibition control (e.g., Can wait variate F was at least marginally significant (at p ,
before entering into new activities if he/she is asked .061 or better) for the effect of time for minimizing
to; as 5 .79 and .77 at T4 and T5). In addition, at T3, and parental distress, Fs(3, 47) 5 2.63 and 6.58, ps ,
T4, and T5, parents rated childrens self-control using .001 and .061, as well as punitive responses, F(3, 46) 5
22 items from Kendall and Wilcoxs (1979) 33-item 5.69, p , .002. According to univariate analyses, dis-
Self-Control Rating Scale (the entire scale was not tress and punitive reactions increased linearly with
520 Child Development

Table 1 Correlations and Trends Across Time for Parental Reactions

Time Periods
Type of
Reaction T1/T2T3 T3T4 T4T5 T1/T2T4 T3T5 T1/T2T5 Trends

Punitive .33** .45*** .61*** .49*** .65*** .42** Lin I; Cubic


Minimizing .35** .59*** .41** .44*** .55*** .41*** Quad
Parental distress .58*** .66*** .70*** .59*** .75*** .68*** Lin I

Note: Lin I 5 an increasing linear trend over time, Quad 5 quadratic trend, Cubic 5 a cubic
trend.
* p , .05; ** p , .01; *** p , .001.

age, Fs(1, 49; 1, 48) 5 17.50 and 11.13, ps , .001 and Relations of Parental Reactions to Adults Reports
.002, respectively (see means in Table 2). There also of Internalizing and Externalizing Emotion
was a cubic effect for punitive reactions, F(1, 48) 5
To determine if childrens negative emotionality
9.41, p , .004. Reports of punitive responses in-
was related to parents reported reactions, parents re-
creased from ages 46 to 68; dropped somewhat
ports of reactions to childrens negative emotion first
from 68 to 810; and then increased again from ages
were correlated with parents and teachers reports of
810 to 1012.
childrens internalizing and externalizing emotion
Only a quadratic effect was obtained for minimiz-
(the latter was assessed at only T3, T4, and T5).
ing reactions, F(1, 49) 5 5.50, p , .023. Parents re-
ported using less minimizing as children aged until
ages 810, followed by a substantial increase in min-
Correlations for Parents Reports of Childrens
imizing reactions at ages 1012. Thus, in general, pu-
Negative Emotionality
nitive and minimizing responses decreased until
sometime in mid-elementary school years; however, Externalizing emotion. There were numerous con-
as children entered pre-adolescence, nonsupportive current and across-time relations between parental
parenting reactions increased. Moreover, parental reactions and parental reports of negative emotion. In
distress reactions increased with the age of the child. regard to externalizing negative emotion, the correla-

Table 2 Means and Standard Deviations for the Major Variables

Variable T1/T2 T3 T4 T5

Parental reactionsa
Punitive 1.77 (.63) 2.02 (.64) 1.92 (.54) 2.16 (.70)
Minimizing 2.22 (.91) 2.14 (.76) 2.11 (.62) 2.36 (.90)
Distress 2.48 (.64) 2.84 (.78) 2.80 (.82) 2.96 (.82)
Social functioning
Problem behavior (M)b 2.11 (.36) 2.04 (.42) 2.00 (.42)
Problem behavior (F)b 2.01 (.34) 1.98 (.38) 1.96 (.42)
Social functioning (T)c .03 (1.82) .06 (1.84) .00 (1.80)
Child emotionalityd
Internalizing (P) 2.03 (.46) 2.01 (.72) .00 (.68) .02 (.70)
Internalizing (T) 2.03 (.64) 2.01 (.72) .04 (.74) .02 (.70)
Externalizing (P) 2.03 (.75) .00 (.79) 2.01 (.82)
Externalizing (T) .00 (.85) .03 (.84) .01 (.78)
Regulation (P) .01 (.80) .00 (.84) 2.02 (.82)

Note: M 5 maternal report, F 5 paternal report, P 5 primary care givers report (usually
mothers), T 5 teacher report.
a Scores ranged from 1 to 7.
b Scores ranged from 1 to 4.
c Scores were ratings for aggression and disruptive behavior (standardized and averaged)

minus scores for socially appropriate behavior (standardized).


d Scores were composed of individual items, standardized within the total sample at each time

period, averaged together.


Eisenberg et al. 521

Table 3 Relations of Parental Reactions to Parental Ratings of tion and minimizing reactions for the total sample.
Externalizing and Internalizing Emotion Although correlations for boys and girls were signifi-
Parental Externalizing Emotion cantly different in only two cases (probably due in
Reaction part to small sample sizes), it should be noted that for
and Age T3 T4 T5 boys only, reported externalizing emotion at T4 was
positively related to minimizing at T3, T4, and T5,
Punitive rs(32, 35, 25) 5 .35, .32, and .41, ps , .04, .053, and .037.
46 (T1/T2) .231 .221 .21
68 (T3) .17 .11a .29*
Moreover, boys externalizing emotion and parental
810 (T4) .28* .20*b .47*** minimizing at T5 were correlated, r(25) 5 .46, p ,
1012 (T5) .241 .21c .49***d .016. Thus, there was some relation between parental
Minimizing reports of externalizing emotion and minimizing, but
46 (T1/T2) .25* .231 .231 primarily for boys, and this relation was less consis-
68 (T3) .09 .08e .00 tent than for punitive or distress reactions.
810 (T4) .16 .11 .211
Internalizing emotion. In regard to parental reports
1012 (T5) .10 .03f .241
of internalizing emotion, parental reports of dis-
Distress
46 (T1/T2) .00 .10 .231
tressed (including upset) reactions and childrens in-
68 (T3) .16 .231 .35** ternalizing emotion generally were positively related
810 (T4) .24* .25* .41*** both concurrently and across time (see Table 3). In
1012 (T5) .231 .32* .51*** contrast, there were no significant correlations be-
tween internalizing emotion and either punitive or
Internalizing Emotion
minimizing reactions for the total sample. However,
T1/T2 T3 T4 T5 for girls, reported internalizing negative emotion at
Distress T3 was significantly positively related to parental pu-
46 (T1/T2) .34** .201 .221 .31* nitive reactions at T4 and T5, rs(32, 29) 5 .35 and .37,
68 (T3) 2.02 .27* .31** .231
ps , .045 and .041, and internalizing emotion at T5
810 (T4) .26* .34* .25* .221g
1012 (T5) .18 .241 .30* .33** was positively related to punitive reactions at T4,
r(28) 5 .38, p , .041. Correlations were not significant
a The correlations for boys, r(35) 5 .33, p , .049, was marginally for boys (and the sex difference in the correlations was
significantly larger than that for girls, r(31) 5 2.12, ns, z 5 1.85, significant for the relation between T3 internalizing
p , .10. emotion and T5 punitive reactions, z 5 2.01, p , .05).
b The correlation for boys, r(35) 5 .44, p , .006, was marginally sig-

nificantly larger than that for girls, r(32) 5 .04, ns, z 5 1.76, p , .10.
c The correlation for boys, r(25) 5 .46, p , .006, was marginally sig-
Correlations for Teachers Reports of Childrens
nificantly larger than that for girls, r(27) 5 2.05, ns, z 5 1.93, p , .10.
d The correlation for boys, r(25) 5 .67, p , .001, was marginally Negative Emotionality
significantly different than that for girls, r(29) 5 .31, p , .09, z 5 There were relatively few findings for teacher-
1.76, p , .10.
e The correlation for boys, r(35) 5 .32. p , .053, was significantly reported negative emotion, although those that were
different than that for girls, r(39) 5 2.16, ns, z 5 1.96, p , .05. significant generally were consistent with expectations.
f The correlation for boys, r(25) 5 .40. p , .037, was significantly Externalizing emotion. Teacher-reported externaliz-
different than that for girls, r(27) 5 2.28, ns, z 5 2.51, p , .05. ing negative emotion at T3 was positively related to
g The correlations for boys, r(27) 5 2.02, ns, was marginally sig-
punitive reactions at T4 and T5, rs(67, 54) 5 .26 and
nificantly smaller than that for girls, r(28) 5 .47, p , .008, z 5
1.93, p , .10.
.29, ps , .03, and T4 externalizing emotion predicted
1 p , .10; * p , .05; ** p , .01. T5 punitive reactions, r(55) 5 .27, p , .042. Given
that 12 correlations were computed between teacher-
reported externalizing emotion and parental punitive
reactions, the number of significant correlations was
tions were primarily for parental punitive (especially more than would be expected by chance. This pattern
for boys) and distress reactions (see Table 3), although of relations was consistent with the possibility of child
the numbers of correlations that were at least margin- externalizing emotion eliciting parental punitiveness.
ally significant were greater than chance even for min- Teachers reports of externalizing behavior were
imizing reactions. For the punitive and distress reac- not significantly related to parental minimizing reac-
tions, the pattern of positive associations appeared to tions for the total sample, although for boys, T3 and
become somewhat more consistent with age. T4 reports of externalizing were associated with
In contrast, there were fewer significant correla- parental minimizing at T5 and T3, respectively,
tions between parental reports of externalizing emo- rs(24, 36) 5 .39 and .40, ps , .05 and .014. Even
522 Child Development

though these correlations are consistent with those (SEM) can only test the plausibility of specific causal
for parent-reports of externalizing emotion, they may relations over time (and cannot prove causality), SEM
be chance findings. Moreover, teachers reports of ex- can be used to indicate whether the pattern of rela-
ternalizing emotion at T4 were positively correlated tions in a given data set supports one set of causal re-
at p , .06 or better with parental report of distress at lations versus another, or bidirectionality of relations.
T3 and T5, rs(71, 55) 5 .22 and .28, ps , .058 and .032 Ideally, all the variables in the study would be tested
(.7 correlations would be expected to be significant by in a single model but we could not do so due to the
chance at p , .06 or better). relatively small sample size.
Internalizing emotion. There were relatively few Models were tested only when there were numer-
significant correlations between teacher-reported in- ous correlations between childrens negative emo-
ternalizing emotion and parental punitive and dis- tionality and a given parental reaction. Thus, models
tress reactions. T4 reports of internalizing emotion were tested for the relations of parent-reported exter-
were positively correlated with T5 punitive parental nalizing negative emotion with parental punitive and
reactions for the total sample, r(55) 5 .30, p , .024 (as distress reactions, as well as for the relation between
well as with T4 punitive reactions for boys only, parental reports of childrens internalizing negative
r(33) 5 .34, p , .049). The only significant correlation emotion and parental distress reactions. No models
for parental distress reactions was between internal- were computed for (1) relations of parental minimiz-
izing emotion at T3 and parental distress at T5, r(54) 5 ing reactions with parental reports of externalizing or
.27, p , .047. These findings may be due to chance. internalizing emotion, or (2) teachers reports of chil-
A few more findings were obtained for parental drens negative emotions. For these sets of data, cor-
minimizing reactions, although the pattern of find- relations were limited in number and tended to be
ings was mixed and relations held primarily for boys. confined to a few time periods or one sex.
For the total sample, a positive correlation was ob- In the three models that were tested, paths from
tained between T4 internalizing emotion and T5 pa- PNRs to negative emotion at the next assessment
rental minimizing, r(55) 5 .36, p , .005. However, a were included, as were paths from negative emotion-
negative correlation was found between T1/T2 pa- ality to PNRs at the next assessment (see Figure 1).
rental minimizing reactions and T1/2 and T5 inter- Data from T3, T4, and T5 were included. Data from
nalizing emotion, rs(76, 51) 5 2.22 and 2.29, ps , T1/T2 were not for two reasons: (1) externalizing
.033 and .053. For boys, although the relation between negative emotion was not assessed at T1/T2, and (2)
T1/T2 internalizing emotion and minimizing was neg- the measure of internalizing negative emotion at T1/
ative, r(39) 5 2.35, p , .024, T4 teacher-reported inter- T2 differed from those at later assessments. In addi-
nalizing emotion was positively related to parental tion to the cross-lagged effects, contemporaneous re-
minimizing reactions at T3, T4, and T5, rs(36, 34, 26) 5 lations between PNRs and childrens negative emo-
.45, .38, and .52, ps , .005, .021, and .006. None of tionality were included in the model at the time points
these correlations was significant for girls, although where they could be an alternative to prediction from
T3 internalizing emotion was positively related with cross-lagged effects.
girls T4 minimizing, r(31) 5 .44, p , .01. Thus, there Composite scale scores of parental socialization
was some evidence that by mid- to late elementary and child emotionality were used as indicators of the
school, internalizing emotion was associated with pa- latent constructs in the analyses. Thus, there were
rental minimizing over a period of time, but primarily single indicators of each construct at each time period
for boys. However, minimizing reactions when the (i.e., the scales for parental punitive, minimizing, or
children were young seemed to be linked to low inter- parental distress was used, as well as the scales of child
nalizing. Although more findings were significant internalizing or externalizing emotion at each time
than would be expected by chance, the findings on period). Errors for the indicators and the latent con-
parental minimizing and childrens internalizing emo- structs were estimated by the program. In all the
tion should be interpreted with caution until replicated. models, paths between negative emotion from T3 to
T4 and from T4 to T5 (as well as their error terms) ini-
tially were constrained to be equal (i.e., the same), as
Structural Equation Modeling Analyses
were the analogous across time paths for PNRs.
Bentlers (1989) EQS structural equation modeling Moreover, the measurement errors for measures of a
program was used to examine the direction of the re- given construct (i.e., PNR or negative emotionality)
lations over time between childrens negative emo- were constrained to be equal at all three assessments
tionality and parental reactions to childrens negative (this is typical of a Markov longitudinal model). The
emotion. Although structural equation modeling contemporaneous correlations between PNRs and
Eisenberg et al. 523

Figure 1 Relation of parental punitive and distress reactions to parents reports of externalizing emotion: Dots represent hypoth-
esized but nonsignificant paths. Unstandardized coefficients for punitive reactions are not in parentheses; unstandardized coef-
ficients for parental distress are in parentheses. Values on curved lines with double-headed arrows are unstandardized covari-
ances between the two connected constructs. Numbers in the circles are ages in years. p , .10; *p , .05; **p , .01; ***p , .001.

childrens emotionality at T4 and T5 also were con- poraneous correlations between PNRs and child neg-
strained to be equal to one another.3 LaGrange Multi- ative emotionality in the model. Especially given the
plier Tests for releasing constraints were examined to relatively small sample size in this study, it is difficult
determine if these constraints should be removed and for cross-lagged paths to obtain statistical significance.
they were not problematic. For externalizing negative emotionality, the mod-
It should be noted from the onset that a model such els for both parental punitive and distress reactions fit
as this is a very stringent test of the across time (cross well, x2s(2, N 5 31) 5 5.15 and 3.64, ps , .08 and .16,
lagged) relations between PNRs and child negative CFIs (comparative fit index) 5 .973 and .988. As can
emotionality. Any across time relation between PNR be seen in Figure 1, in general, PNRs were predicted
and child negative emotion (or vice versa) must hold by PNRs at the prior assessment whereas childrens
even when the effects of consistency over time in each externalizing emotion similarly was predicted by
construct are taken into account. Moreover, cross- such emotion 2 years earlier. This pattern was to be
lagged effects are, in a sense, pitted against contem- expected given the relatively strong consistency of
ratings of PNRs and childrens emotionality over
3 T3 punitive reactions were not significantly correlated with
time. Moreover, childrens externalizing negative
T3 externalizing negative emotion, although T3 distress reac-
emotion at ages 68 predicted parental punitive reac-
tions were positively correlated with T3 internalizing emotion.
Because adding the correlation between parental negative PNRs tions at ages 810, and parental report of punitive re-
and child negative emotion sometimes resulted in obviously er- actions at T4 (ages 810) predicted externalizing emo-
roneous results (e.g., reversal of positive relations of a construct tion at T5 (ages 1012). Thus, punitive parental
over time) or the model would fail to converge in 30 iterations, reactions and children externalizing negative emotion
the T3 contemporaneous correlation was fixed at 0 if it did not
appeared to influence one another across time. For
approach significance. However, because T3 internalizing nega-
tive emotion was significantly correlated related to parental dis- parental distress reactions, there also was a marginally
tress reactions at T3, the T3 contemporaneous correlation was significant path from externalizing negative emotion
included in the model for internalizing emotion. at ages 68 to parental distress reactions at ages 810;
524 Child Development

however, the path from parental reactions to external- lation of PNRs to parental report of problem behavior
izing emotion was not significant. It is not surprising (or vice versa).
that the R2s in the models were high given the fact
that each construct was predicted by multiple vari-
Parents Reports of Problem Behavior
ables, including the same construct at an earlier period,
and most of the data came from the same reporter.4 Correlations. As presented in Table 4, parental re-
A similar model with the same constraints was ports of punitive reactions generally were positively
tested for the relation between parental reports of related to mothers reports of problem behavior. In
childrens internalizing negative emotion and paren- addition, they were positively correlated with fa-
tal distress reactions (except the correlation between thers reports of problem behavior, particularly at
internalizing emotion and parental distress was in- older ages and for boys. Parental reports of distress
cluded at ages 68 because the zero-order correlation reactions also tended to be associated with both
was significant). This model would not produce rea- mothers and fathers reports of problem behavior, al-
sonable values with all paths in it (e.g., the path from beit primarily for problem behaviors reported at ages
parental distress at T3 to T4 became significant and 810 and older.
negative). Even without the contemporaneous corre- There were relatively few significant correlations
lations, the model did not fit until, based on the Wald for parental minimizing reactions, albeit more than
Test statistic (which indicates paths that could be would be expected by chance. Parental reports of
dropped), the path from T4 internalizing emotion to minimizing reactions at ages 46 (all reports were
T5 parental distress was dropped, x2(3, N 5 51) 5 from mothers) often predicted mothers and fathers
5.29, p , .14, CFI 5 .98. Even then, however, there was reports of problem behavior at later assessments.
no evidence of significant cross-lag correlations; the Moreover, reports of minimizing at ages 1012 were
only paths that were significant were those paths associated with fathers reports of boys problem be-
within a construct over time (i.e., prior parental dis- haviors, concurrently and at T3. There were no signif-
tress predicted parental distress two years later and icant relations between minimizing reactions at T3 or
prior child internalizing predicted later internaliz- T4 and problem behavior.
ing). Thus, there was no evidence that the two con- SEM with parental reports of problem behavior. The
structs causally affected one another in the longitudi- general structure of the models tested can be seen in
nal model, even though parental distress and childrens Figure 2. As for the models for negative emotionality,
internalizing negative emotion often were signifi- data from T3, T4, and T5 were used in the analyses.
cantly correlated within and across time. Single indicators of each latent construct were avail-
able. Cross-lagged paths were included between the
given PNR and childrens regulation, and contempo-
Relations of Parental Reactions to Social Functioning
raneous correlations between parental PNRs and chil-
The relations of PNRs with parental reports of drens regulation were included at the times being
problem behavior and teachers reports of socially ap- predicted (i.e., T4 and T5; contemporaneous correla-
propriate behavior versus aggressive/disruptive be- tions were not significant at T3 and did not compete
havior (henceforth labeled socially appropriate behav- with the cross-lagged paths). Moreover, T5 PNR (pu-
ior) initially were examined with correlational analyses. nitive or distress) and regulation were predictors of
In all cases, the numbers of correlations that were sig- T5 mother- or father-reported problem behavior. T3
nificant or marginally significant were considerably mother-reported problem behavior also was included
greater than would be expected by chance. If numer- in the model to control for the effects of parents (al-
ous correlations were obtained, SEM was used to ex- most always mothers) perceptions of problem be-
amine whether regulation may have mediated the re- havior at ages 68 on contemporaneous perceptions
of regulation and PNRs at the youngest age in the
4 Occasionally the program would not compute the R2 for a
model, as well as to assess the relation of such percep-
construct because the disturbance (error for a latent construct) tions to later problem behavior. In all models, T3 re-
had been set to 0 by the program. In these cases, the disturbance ports of problem behavior were assumed to be corre-
was estimated with the value of the same construct 2 years be- lated with T3 PNRs or regulation (so paths for
fore or after. Then the program was rerun to ascertain that the covariation of predictors at T3 were included). Be-
path values were highly similar to the run in which the program
cause we were interested in the relations of fathers
estimated the disturbance. Assuming this was the case, the R2
value (but not the path coefficients) was taken from program reports of problem behavior with PNRs reported by
output in which the disturbance was estimated in the aforemen- mothers, different models were computed for moth-
tioned manner. ers and fathers reports of problem behavior at T5
Eisenberg et al. 525

Table 4 Concurrent and Across Time Relations of Parental Reactions to Maternal and
Paternal Report of Childrens Problem Behavior: Zero-Order and Partial Correlations

Report of Problem Behavior

Age 6 8 (T3) Age 810 (T4) Age 1012 (T5)


Type of Parental
Reaction Mother Father Mother Father Mother Father

Punitive
46 (T1/T2) .29* .14 .35**a .07 .241 .17
68 (T3) .23* .02 .13 .11 .35** .241
8 0 (T4) .28* .07 .31** .06b .45***c .31*
1012 (T5) .21 .13d .14e .10 .41** .18f
Minimizing
46 (T1/T2) .32** .21 .42*** .30* .23 .30*
68 (T3) .09 .09 .02 .02 2.01 .08
8 10 (T4) .06 .02 .12 .06 .231 .19
10 12 (T5) .21 .14g .04h .04 .13 .17i
Parental distress
46 (T1/T2) .15 .26* .221 .29* .18 .34*
68 (T3) .25* .32* .29* .42*** .40** .38**
810 (T4) .14 .21 .33** .31* .53*** .34*
1012 (T5) .21 .20 .29* .38** .39** .33*

a The correlation for boys, r(32) 5 .50, p , .002, was greater than that for girls, r(26) 5 .02, ns,
z 5 1.99, p , .05.
b The correlation for boys, r(32) 5 .29, p , .094, was marginally greater than the correlation for

girls, r(30) 5 2.14, ns, z 5 1.69, p , .10.


c The correlation for boys, r(25) 5 .62, p , .001, was marginally greater than that for girls, r(25) 5

.23, ns, z 5 1.74, p , .10.


d The correlation for boys, r(21) 5 .40, p , .059, was marginally greater than that for girls, r(22) 5

2.09, p , .10.
e The correlation for boys, r(25) 5 .36, p , .066, was marginally greater than that for girls, r(27) 5

2.10, ns, z 5 1.65, p , .10.


f The correlation for boys, r(23) 5 .50, p , .011, was marginally greater than that for girls,

r(25) 5 2.02, ns, z 5 1.93, p , .06.


g The correlation for boys, r(21) 5 .52, p , .011, was greater than that for girls, r(22) 5 2.05, ns,

z 5 2.00, p , .05.
h The correlation for boys, r(25) 5 .36, p , .066, was greater than that for girls, r(27) 5 2.22, ns,

z 5 2.18, p , .05.
i The correlation for boys, r(24) 5 .45, p , .02, was marginally greater than that for girls, r(25) 5

.02, ns, z 5 1.76, p , .10.


1 p , .10; * p , .05; ** p , .01; *** p , .001.

(correlations between mothers and fathers reports fit well and two changes were made based on the
ranged from .50 at T3 to .56 at T5). The same constraints Wald test (which indicates paths that can be dropped)
were included in these models as in the externalizing and Lagrange Multiplier test (which provides infor-
emotionality models (in regard to constraining paths, mation on paths that would improve the model if
measurement errors, and contemporaneous correla- added) statistics. First, a path was added from T3 pu-
tions to be the same across time). Models were tested nitive reactions to T5 punitive reactions (see Figure 2).
for parental punitive and distress reactions; they were In addition, the constraint that the path from T3 to T4
not computed for minimizing reactions due to the rel- regulation had to be equal to the path from T4 to T5
atively low numbers of correlations between such re- regulation did not hold and was removed. After these
actions and problem behavior at T3 and T4. changes, the model fit well, x2(10, N 5 50) 5 14.19,
The first model was for the prediction of T5 p , .17; the CFI (fit) was .983. As in previous models
mother-reported problem behavior from regulation (and in all other models to be presented), punitive re-
and punitive reactions (usually based on mothers re- actions were predicted by earlier punitive reactions
ports; see Methods), as well as T3 mother-reported and regulation was predicted by earlier regulation
problem behavior; N 5 50). The initial model did not (see Figure 2). Moreover, T3 problem behavior and
526 Child Development

Figure 2 Prediction of mother- or father-reported (for boys only) problem behavior at T5 from primary care-giving parents re-
ports of punitive reactions and regulation, as well as mothers reports of problem behavior at ages 68. Note: Dotted lines repre-
sent hypothesized paths that were nonsignificant at p . .10 for both mothers and fathers reports of problem behavior. Unstand-
ardized coefficients for mother-report of problem behavior data are not in parentheses; unstandardized coefficients for father-
report of problem behavior data at T5 for boys are in parentheses. Values on curved lines with double-headed arrows are unstand-
ardized covariances between the connected constructs. NA 5 not applicable because the path was not necessary. Ages are 6 8, 8
10, and 10 12. p , .10; *p , .05; **p , .01; ***p , .001.

regulation were significantly correlated at T3. T5 p , .06, CFI 5 .96. However, in this model, only moth-
problem behavior was significantly predicted by low ers earlier reports of problem behavior predicted T5
regulation at T5 and by T3 problem behavior, and was problem behavior (path coefficient 5 .434, p , .05).
marginally, directly predicted by T5 punitive reac- None of the cross-lag paths for prediction between
tions. Of the most interest, regulation at ages 68 pre- punitive reactions and regulation was significant.
dicted parental punitive reactions at ages 810, which Thus, for father-report of problem behavior at ages
in turn predicted regulation at ages 1012. Thus, the 1012 for the combined sample of girls and boys,
pattern was consistent with the view that there was there was no evidence of cross-lagged prediction or of
bidirectional causality between childrens regulation PNRs predicting childrens problem behavior.
and parental punitive reactions, and that the relation The fact that punitive reactions did not predict T5
of punitive reactions to T5 problem behavior was par- father-reported problem behavior was to be expected
tially mediated by childrens level of regulation (and because punitive reactions predicted fathers reports
perhaps, to a limited degree, vice versa). Mediation of problem behavior at T5 only for boys. Thus, the
could be assumed because punitive reactions at T3 model was recomputed using only boys in this model
were significantly correlated with T5 problem behavior. (the constraint that the T3/T4 and T4/T5 paths were
The same model was tested for the relation of pa- equal was dropped in the prior models did not need
rental punitive reactions and regulation to T5 father- to be dropped in this model). Although the sample
reported problem behavior (N 5 50). In this model, it size was small (N 5 25), the model converged and fit
was not necessary to add a path from T3 to T5 puni- well, x2(11, N 5 25) 5 14.79, p , .19, CFI 5 .978. Gen-
tive reactions. The model fit, x2(11, N 5 50) 5 19.063, erally a model will not converge easily or the param-
Eisenberg et al. 527

eter estimates are unstable or cannot be computed if The models for punitive reactions were then com-
the sample is too small to test the model. With a small puted replacing parental punitive reactions with pa-
sample, the x2 is affected (i.e., with a small sample, there rental distress reactions. In neither model was it nec-
is low power for the x2 test). However, the CFI is not essary to make the two modifications described above
affected much by small (or large) sample size in punitive reactions model for mother-reported prob-
(Bentler, 1990) and, due to lack of power, it is more lem behavior (i.e., adding a path from T3 to T5 dis-
difficult for path coefficients to attain significance in tress reactions and releasing the constraint that the
small than large samples. two paths between measures of regulation be equal).
As can be seen in Figure 2, in this model, both T5 pu- The fit was good for both models using the total sam-
nitive reactions and mothers reports of T3 problem be- ple of boys and girls, x2s(11, 12, Ns 5 50) 5 19.30 and
havior positively predicted boys T5 father-reported 8.61, ps , .056 and .73, CFIs 5 .97 and 1.00, for
problem behavior. In addition, there was a significant mother-report and father-report of T5 problem be-
negative path between regulation at age 68 and puni- havior models, respectively. As can be seen in Figure
tive reactions at age 810. Thus, as in the mother-report 3, for both mother-report and father-report of prob-
of problem behavior model, there was evidence that lem behavior at T5, the path from parental distress re-
regulation affected later parental punitive reactions. Ef- actions at ages 68 to childrens regulation at ages 8
fects of regulation on problem behavior may not have 10 was at least marginally significant. Moreover, the
been significant because the correlation between father- paths from T5 distress reactions and T3 problem be-
reported T5 problem behavior and regulation for the havior to T5 problem behavior were significant and
boys in this model was not significant (the two con- positive for both models. Regulation negatively pre-
structs were significantly correlated for the sample of dicted mother-reported, but not father-reported,
both boys and girls at T5, r(38) 5 .40, p , .008). problem behavior. Thus, much of the relation of

Figure 3 Prediction of mother- or father-reported problem behavior at ages 10 12 from primary care-giving parents reports of
parental distress reactions and childrens regulation, as well as mothers reports of problem behavior at ages 6 8. Note: Dotted
lines represent hypothesized paths that were nonsignificant for both models. Unstandardized coefficients for mother-report of
problem behavior at ages 10 12 are not in parentheses; unstandardized coefficients for father-report of problem behavior at ages
10 12 are in parentheses. Values on curved lines with double-headed arrows are unstandardized covariances between the two
connected constructs. Ages are 6 8, 8 10, and 10 12. See text for findings on model including fathers reports of problem behavior
at age 68. p , .10; *p , .05; **p , .01; ***p , .001.
528 Child Development

parental distress reactions to problem behavior ap- Table 5 Concurrent and Across Time Relations of Parental Re-
actions to Teachers Reports of Childrens Socially Appropriate
peared to be independent of regulation. Behavior
In summary, mothers reports of problem behavior
at ages 1012 were predicted by both childrens low Report of Social Functioning
regulation and by parental punitive and distress reac-
Type of Age 68 Age 810 Age 1012
tions (albeit marginally for the former). For punitive Parental
reactions and mothers reports of problem behavior, Reaction Girls Boys Girls Boys Girls Boys
there was clear evidence of a bidirectional relation be-
tween regulation and punitive reactions, with part of Punitive
the relation between punitive reactions and T5 prob- 46 (T1/T2) .00 .00 .22 2.05 .25 2.40*b
68 (T3) 2.27 2.17 2.11 2.18 2.06 2.351
lem behavior being indirect through regulation (even 810 (T4) 2.24 2.23 .14 2.26 .29 2.43*a
when the effects of contemporaneous correlations be- 1012 (T5) 2.341 2.42* 2.44* 2.361 2.25 2.50*
tween reports of punitive reactions and child regula- Minimizing
tion were controlled). In the model for T5 father- 46 (T1/T2) 2.28 .01 .01 2.07 .20 2.50**a
reported problem behavior for the total sample, only 68 (T3) 2.35* 2.19 .02 2.39* .04 2.19
earlier maternal reports of problem behavior pre- 810 (T4) 2.26 2.16 .01 2.22 .24 2.351b
1012 (T5) 2.321 2.25 2.27 2.45* 2.04 2.401
dicted T5 father reports of problem behavior. How-
ever, parental punitive reactions did predict father- Distress
46 (T1/T2) .21 2.22 .26 2.281b .371 2.40*a
reported problem behavior for boys and in this model, 68 (T3) 2.12 2.01 2.25 2.17 2.14 2.38*
any effect of regulation on T5 problem behavior was 810 (T4) 2.10 2.04 2.21 2.04 .09 2.33
through parental punitive reactions. Finally, parental 1012 (T5) 2.15 2.341 2.341 2.45* 2.03 2.381
reports of distress reactions predicted mothers and
fathers reports of T5 problem behavior and there was Note: High scores on the index of socially appropriate behavior in-
dicate low aggressive and disruptive behavior and high levels of
evidence that parental distress reactions at ages 68
socially appropriate behavior. Ns ranged from 2537 for girls and
predicted low regulation at ages 810. Because regu- from 2542 for boys.
lation was significantly related to T5 problem behav- a The correlation for girls and boys differed at p , .01.

ior only for mothers reports, there may have been in- b The correlation for girls and boys differed at p , .05.
1 p , .10; * p , .05; ** p , .01; *** p , .001.
direct effects of parental distress on later problem
behavior for mothers, but not fathers, reports of
problem behavior.
ings for girls: Punitive reactions reported at ages 10
12 were associated with girls (and boys) low social
Correlations with Teachers Reports of Social Behavior
functioning at younger ages (at p , .10 or better), sug-
As would be expected, parental reactions to chil- gesting that children who exhibit problematic behav-
drens negative emotions were less frequently related ior at school elicit punitive reactions from parents by
to teachers than to parents reports of social behavior. ages 1012.
Because of clear sex differences in the pattern of find- Parental reports of minimizing childrens negative
ings (including a number of significant sex differ- emotion generally predicted low levels of socially ap-
ences), the data are presented by sex (see Table 5). Re- propriate behavior for boys at ages 810 and 1012.
call that different teachers provided information on There were few significant findings for girls, although
quality of social functioning at each time period, so a girls who were low in social functioning at school at
pattern of correlations involving several time periods ages 68 had parents who reported minimizing their
is likely to be quite reliable. In all cases, the numbers daughters negative emotion at ages 68.
of correlations significant at p , .10 or better were Correlations of parental distressed reactions with
considerably greater than would be expected by chance. teacher-reported social functioning held primarily for
Parental punitive reactions frequently were associ- older boys. Parents who reported distressed reactions
ated with low quality of boys social functioning, es- to their sons negative emotion at younger ages had
pecially for teachers reports of socially appropriate sons who were relatively low in quality of social func-
behavior at ages 1012. For example, parental reports tioning by age 1012. The pattern of association ap-
of punitive reactions at all ages were at least margin- peared to strengthen with age because distressed re-
ally related to low socially appropriate behavior at actions at earlier ages were not significantly related
ages 1012; this pattern is consistent with the notion with social functioning at ages 68. Moreover, boys
that parental punitive reactions have an effect over who were relatively low in socially appropriate be-
time on boys social functioning. There were few find- havior at T3, T4, or T5 tended (at p , .10 or better) to
Eisenberg et al. 529

have parents who reported distressed reactions by 12. Thus, as children approached early adolescence, it
the time their sons were 1012 years of age. appeared that parental nonsupportive reactions to
SEMs were not computed for teachers reports of childrens expression of negative emotion were associ-
socially appropriate behavior because parental re- ated with, or contributed to, expressions of external-
ports of regulation seldom were significantly related izing emotion in children. The rise in negative emo-
to teachers reports of socially appropriate behavior.5 tion (Larson & Richards, 1991; Murphy et al., 1998)
and in parent-child conflict in early adolescence (Col-
lins, 1990) noted in the literature may be particularly
DISCUSSION
evident for families in which parents are nonsupport-
In this study, we obtained some evidence consistent ive and intolerant of childrens expression of negative
with the hypothesis that parents reported reactions emotion (at least when its expression is not highly in-
to childrens negative emotions, especially punitive appropriate). In addition, the path from externalizing
reactions, affect children regulation and externalizing emotion at age 68 to later parental reactions sug-
negative emotion. Moreover, when children were rel- gests that childrens externalizing negative emotion
atively young, their regulation and externalizing neg- affected parental reactions at younger ages. Because
ative emotion (as perceived by parents) predicted pa- we do not have data from the earliest years, it is im-
rental report of punitive negative reactions. Although possible to know who influenced whom first, but it
SEM cannot provide conclusive evidence of causality, appears that there is a bidirectional relation between
the findings from this longitudinal sample provide parental reactions and childrens temperament, as
considerable support for bidirectional parent-child least as viewed and reported by parents. The fact that
influences. contemporaneous correlations between parents re-
It was predicted that there would be correlational ports of their own reactions and childrens externaliz-
associations, as well as bidirectional relations over ing emotion were controlled in the analyses is consis-
time, between childrens negative emotionality and tent with the view that these across time relations did
parents negative reactions to childrens negative not occur primarily because parents who report puni-
emotion, especially for childrens externalizing emo- tive reactions also tend to report high levels of exter-
tion. There were numerous significant positive corre- nalizing emotion in their children.
lations of both parental punitive and distress reac- Only parental distress reactions were correlated
tions with parental reports of childrens externalizing consistently with childrens parent-reported internal-
emotion. Fewer significant correlations were obtained izing emotion. Childrens internalizing reactions, be-
for parental minimizing reactions, although there were cause they often are subtle and less aversive to others
some positive relations, especially for boys. Moreover, than externalizing emotions, may be relatively un-
although there were fewer significant correlations be- likely to elicit punitive responses from parents. More-
tween teachers (than parents) reports of childrens over, it is likely that teachers often do not notice chil-
externalizing emotion and PNRs, the pattern of cor- drens sadness and anxiety, which could account for
relations, especially for punitive reactions, was sim- the modest numbers of relations between teacher-
ilar in direction to that for parent-reported external- reported internalizing emotion and PNRs. Based on
izing emotion. the model for internalizing emotion, it did not appear
In models for both parental punitive and distress that the consistent pattern of association between pa-
reactions, childrens reported externalizing emotion rental distress reactions and childrens internalizing
at age 6 8 at least marginally predicted parental emotion was due to one of these variables having a
negative reactions at age 8 10. In addition, punitive causal influence on the other. Rather, the findings are
reactions at age 8 10 predicted childrens parent- consistent with the view that a third variable was re-
reported externalizing negative emotion at age 10 sponsible for the association between childrens inter-
nalizing negative emotion and parental distress. Two
5 Moreover, parent and teacher reports of quality of social
possible scenarios readily come to mind. First, it is
functioning were not combined in analyses because the two likely that the association is due, in part, to tempera-
were weakly related, rs ranged from .05 (for teachers and fathers mental tendencies toward negative emotionality, es-
at T3) to .28, p , .035 (for mothers at T3). Only one of the six cor- pecially internalizing emotionality, in both parents
relations was significant. Furthermore, only one correlation was and children. There appears to be a genetic compo-
significant when parents reports of problem behavior were cor-
nent to negative emotionality (Plomin & Stocker, 1989),
related with teachers reports of aggression/disruptive behavior
(not combined with socially appropriate behavior). This pattern and this genetic tendency could underlie the parent-
is not surprising given the lack of consistency across reporters in child similarity. In addition, parents who are prone to
other work (e.g., Fagot, 1995). distress may be more aware of, and more likely to re-
530 Child Development

port, internalizing emotions in themselves and their punitive parental reactions and externalizing nega-
children. For whatever reason, the pattern of relations tive emotion. Child characteristics may elicit punitive
between childrens internalizing emotion and paren- parenting in the early school years, but by age 1012,
tal distress differed from that for externalizing emo- punitive parental reactions appear to elicit externaliz-
tion in regard to evidence of causal relations. ing behavior and emotion from children. Because it is
The results of the correlations and structural equa- very difficult in SEM models to get paths in both di-
tion models based on the longitudinal data provide rections at the same time, especially with a relatively
interesting insights regarding the relations of PNRs small sample, it is quite possible that there are bidi-
with childrens socially externalizing problem behav- rectional effects at all ages, but that child characteris-
ior (rather than externalizing emotion) and socially tics have the stronger influence in the early school
appropriate behavior. In the correlations, parents who years, whereas punitive parenting behavior and be-
reported punitive or distress reactions tended to have liefs have a substantial impact later.
children with problem behaviors by age 810 or 10 In the models including parental distress there was
12, even when mothers usually provided data on evidence that parental distress reactions at age 68
their reactions and fathers provided information on predicted childrens regulation at age 810 and, for
problem behaviors (although there were somewhat mothers reports of problem behavior, that some of
fewer significant correlations between fathers than the effect of early parental distress on problem behav-
mothers reports of problem behavior and punitive ior was indirect through regulation. Parental distress
reactions). Findings for parental minimizing reac- could have a negative impact on childrens regulation
tions were in the same direction, but fewer in number. because parents who easily become distressed or upset
Consistent with findings for the parent-report data, by their childs negative emotion may tend to disen-
nonsupportive parental reactions were correlated gage from the distressing interaction rather than help
with low social functioning at school (problem behav- their children learn to manage emotions. However,
ior and low socially appropriate behavior), although the findings for parental distress generally were less
the findings for teachers reports of social functioning consistent than those for reports of parental punitive
were primarily for boys (especially older boys). reactions. The effects of disengaging from the child
In the structural equation models for punitive and likely differ from punitive responses.
distress reactions, maternal report of problem behav- The models for parents reports of problem behav-
ior at age 1012 was at least marginally predicted by ior provide several important insights. First, PNRs
regulation and parental punitive or distress reactions, tended to be associated with problem behaviors at
as well as by mothers reports of childrens problem age 1012, even when the effects of early problem
behavior at age 68. In the model for punitive reac- behavior and childrens regulation were controlled.
tions, fathers reports of boys (but not combined Second, the findings provide some support for the bi-
boys and girls) problem behavior were predicted by directionality of relations between PNRs and chil-
parental punitive reactions and earlier problem be- drens regulation (at least as perceived by parents)
havior. Parental distress reactions (as well as earlier over time, and for the notion that there often are indi-
problem behavior) also predicted fathers reports of rect or mediated effects of each for the other when
problem behavior for the total sample. predicting problem behavior. The findings in the pu-
In the punitive reaction models, there also was ev- nitive models (including both externalizing emotion
idence of across-time relations between punitive reac- and regulation) are consistent with prior work on
tions and childrens regulation. Regulation at age 68 conduct disorders (Lytton, 1990) in suggesting that
predicted punitive reactions at age 810 (for mother- child characteristics and behaviors often elicit non-
report of problem behavior for the total sample and supportive, negative reactions from parents, which
for boys for fathers reports of problem behavior), then predict quality of childrens social functioning.
which in turn predicted childrens regulation at age Moreover, it is likely that the bidirectional relation be-
10 12 for the model including mother-report of tween parental punitive reactions and childrens
problem behavior at T5. Thus, for the model predict- regulationthat predicts mothers reports of prob-
ing mother-report of problem behavior, there was ev- lem behaviorreflects a history of interactions simi-
idence of bidirectional causality between punitive re- lar to those reported by Patterson (1982), in which co-
actions and regulation. For both mothers and fathers ercive maternal reactions to sons behaviors lead to a
reports of problem behavior, there was evidence that cycle of ineffective parenting and inappropriate be-
some of the effect of childrens regulation on problem havior on the part of the child.
behavior was indirect through punitive reactions. The finding that parental punitive reactions pre-
This pattern of findings mirrors that for the model of dicted childrens regulation is consistent with theory
Eisenberg et al. 531

(e.g., Buck, 1984; Gottman et al., 1996) and research sample size made it difficult to obtain significant
(Gottman et al., 1996; Patterson, 1982) discussed pre- cross-lag correlations, so those that were significant oc-
viously. However, there has been little direct empiri- curred despite a lack of power. It also should be em-
cal evidence that parental socialization of emotion phasized that it is very difficult to obtain two cross-
has an effect on childrens regulatory capacities. In time paths between PNRs and child variables in the
addition, parental distress reactions, which some- same direction. For example, to obtain two significant
times are similar to personal distress reactions to an- cross-lags from parental punitive reactions to chil-
others negative emotion in the empathy literature drens regulation (from T3 punitive reactions to T4
(Batson, 1991; Eisenberg, Fabes, Murphy, et al., 1994), regulation and from T4 punitive reactions to T5 regu-
often may result in parents attempts to avoid the child lation) would require that there be substantial change
and in nonsupportive behavior. These behaviors would in parental reactions from T3 to T4 and that this
be expected to undermine childrens sense of security change would have an effect on regulation above and
and regulation (Davies & Cummings, 1994). beyond the effect of parental punitive reactions at T3.
Unlike in the punitive reactions model, there was Moreover, this effect must be stronger than the con-
no evidence of childrens regulation affecting paren- temporaneous correlation at T4. Thus, out of the four
tal distress reactions. The lack of a path from regula- theoretically possible cross-lags, one would not ex-
tion to PNRs in the distress model may be because pa- pect more than one or two lags at most to be signifi-
rental distress reactions reflect, to a greater degree cant, unless the time between assessments was large
than parental punitive reactions, parental characteris- and the sample size was quite substantial.
tics such as temperamental or personality-based neg- Although parental reports of minimizing were less
ative emotionality and reactivity. Thus, parental dis- frequently related to childrens problem behavior
tress may be somewhat more intra-individually than were punitive and distress reactions, there were
stable than parental punitive reactions (an assump- some findings of interest. Parental minimizing at ages
tion consistent with the pattern of correlations in Ta- 46 predicted maternal and/or paternal reports of
ble 1) and vary less as a function of characteristics of problem behavior even 4 or 6 years later. There also
the child (e.g., the childs regulation). was some evidence of a relation between earlier child
Fathers reports of problem behavior at T5 were re- problem behavior and parental minimizing reactions
lated to reports of punitive parental reactions only for at ages 1012, but only for boys. Moreover, parental
boys. This is not surprising given that at T3 and T4, fa- minimizing often predicted low social functioning
thers reports of problem behavior were correlated years later (or concurrently) at school for boys aged
with childrens regulation and emotionality primarily 810 and 1012. Because minimizing likely is a more
for boys. As discussed in previous papers (Eisenberg benign parental reaction than punitive reactions, and
et al., 1995; Eisenberg et al., 1997), it is possible that fa- often may be meant to be helpful by the parent, per-
thers attend more to boys than girls problem behav- haps it is less likely than some other nonsupportive
ior. However, parental distress reactions did predict parental reactions to undermine the quality of chil-
fathers reports of problem behaviors at age 1012 for drens social behavior. It is unclear why the findings
the total sample of boys and girls (and correlations for minimizing were primarily for boys; however, in
between T5 parental distress and father-reported another study, parental emphasis on childrens con-
problem behaviors did not differ markedly across trol or denial of their emotions was related to physi-
sex). The father-report data provide some evidence ological arousal in an emotion-eliciting situation
that the links between problem behavior at age 1012 only for boys (Eisenberg, Fabes, Schaller, Carlo, &
and parental reports of their reactions were not due Miller, 1991). Perhaps, due to gender stereotypes in
solely to the same parent providing information on regard to the expression of emotion, parental verbal-
both problem behaviors and parental reactions. izations that urge children to ignore or suppress
The cross-lagged effects that were obtained are no- negative emotions have more impact on boys.
table given the fact that contemporaneous correla- There was considerable consistency, even over 6
tions were accounted for in the model, as were effects years, in the degree to which parents viewed them-
from consistency of measures over time. Because of selves as reacting in the various ways to childrens
the relatively small sample size in this study, models negative emotions (i.e., in correlations across time).
fit relatively easily (i.e., models were relatively un- Thus, parents perceptions of their own reactions to
likely to be rejected based on the x2). However, the their childrens negative emotions were relatively
high fit index in most models suggests that the stable. This is not altogether surprising given the con-
models are reliable because the CFI is not affected sistency in some other aspects of self-reported parent-
much by sample size. Moreover, the relatively small ing (e.g., McNally, Eisenberg, & Harris, 1991). The fact
532 Child Development

that parental reactions to childrens negative emo- assess causal relations, although the fact that the
tions sometimes were related to childrens problem study was longitudinal with repeated measures of
behavior years later, even as reported by adults critical variables enables one to make some tentative
who did not provide information on parental reac- inferences about causal relations.
tions (i.e., fathers or teachers), suggests that the differ- Despite these limitations, the data from the present
ences among parents in reported PNRs have some study are consistent with the conclusion that parents
basis in reality. perceptions of their own reactions to childrens nega-
There also were age-related changes in mean levels tive emotions are fairly stable and enduring over time
of parents reports of their reactions. Punitive and and predict significant social outcomes. Thus, paren-
minimizing reactions dropped in frequency from age tal reactions to childrens negative emotions and their
68 to age 810, and then increased somewhat from beliefs about these actions may be an important as-
age 810 to 1012 (punitive responses also increased pect of the socialization of socioemotional compe-
from age 46 to 68, showing a cubic trend). Parental tence. Moreover, because the sample in the present
distress reactions increased in a linear fashion with study included relatively few children with serious
age. The increase in nonsupportive reactions to chil- problem behaviors, it is reasonable to predict that
drens negative emotions in early adolescence is con- more and stronger relations would be obtained with
sistent with decreases in positive maternal expres- heterogeneous samples including children with rela-
siveness with children as they enter early adolescence tively extreme behavioral problems.
(McNally et al., 1991), as well as with research indicat-
ing that there is an increase in negative conflictual in-
ACKNOWLEDGMENTS
teractions between parents and children in early ado-
lescence (Collins, 1990). Moreover, because there is This research was supported by grants from the Na-
some evidence that children experience and exhibit tional Science Foundation (DBS-9208375) and the
an increase in negative emotion in early adolescence National Institutes of Mental Health (1 R01 HH55052)
(Eccles, Wigfield, & Schiefele, 1998; Larson & Rich- to the first two authors and by Research Scientist De-
ards, 1991; Murphy et al., 1998), parents may feel that velopment and Research Scientist Awards from the
they need to use more coercive or negative, rather National Institute of Mental Health (K02 MH00903
than supportive, reactions to control their childrens and K05 M801321) to Nancy Eisenberg. The authors
expressions of negative emotion. wish to thank the many students that assisted in this
There are several limitations of the present study. study (particularly Sarah Jones, Pat Maszk, and
As in any longitudinal work, it is possible that histor- Rick Poulin), the parents and children involved, and
ical changes could contribute to age-related changes the principals and teachers in the Tempe, Kyrene,
in parental reports of their reactions to childrens neg- Mesa, and Scottsdale School Districts. We also thank
ative emotions. For example, exposure to ideas through Sandra Losoya for assistance with the analyses.
the mass media (e.g., a show on child rearing) could
affect parents actual or reported reactions to their
ADDRESSES AND AFFILIATIONS
childrens negative emotion. However, one might ex-
pect exposure to information on child rearing to re- Corresponding author: Nancy Eisenberg, Psychology,
sult in less report of nonsupportive child rearing, a Arizona State University, Tempe, AZ 85287-1104; e-mail:
pattern that generally was not obtained. Nonetheless, nancy.eisenberg@asu.edu. Richard A. Fabes, Stepha-
it is impossible to determine if historical factors con- nie A. Shepard, Ivanna K. Guthrie, and Mark Reiser
tributed to the pattern of findings in some way. are also at Arizona State University. Bridget C. Mur-
Moreover, most children in the sample were Cau- phy was at Arizona State University during the
casian and from relatively affluent families; thus, the course of this work, and is currently at the Univer-
findings may not be generalizable to other popula- sity of Oklahoma.
tions. In addition, the measure of parental reactions
was self-report and data were obtained primarily
from mothers. It is important to keep in mind that pa- REFERENCES
rental reports represent their beliefs about their reac-
Albrecht, T. L., Burleson, B. R., & Goldsmith, D. (1994). Sup-
tions to childrens negative emotions and may not be portive communication. In M. L. Knapp & G. R. Miller
identical to their actions. Observational data and data (Eds.), Handbook of interpersonal communication (2nd ed.,
on fathers reactions to their childrens negative emo- pp. 419449). Thousand Oaks, CA: Sage.
tions would be useful in future work. In addition, be- Batson, C. D. (1991). The altruism question: Toward a social-
cause the data were correlational, it was impossible to psychological answer. Hillsdale, NJ: Erlbaum.
Eisenberg et al. 533

Bell, R. (1977). Socialization findings reexamined. In R. Bell Eisenberg, N., & Fabes, R. A. (1992). Emotion, regulation,
& R. Harper (Eds.), Child effects on adults. Hillsdale, NJ: and the development of social competence. In M. S.
Erlbaum. Clark (Ed.), Review of personality and social psychology: Vol.
Belsky, J., Fish, M., & Isabella, R. (1991). Continuity and dis- 14. Emotion and social behavior (pp. 119150). Newbury
continuity in infant negative and positive emotionality: Park, CA: Sage.
Family antecedents and attachment consequences. De- Eisenberg, N., & Fabes, R. A. (1994). Mothers reactions to
velopmental Psychology, 27, 421431. childrens negative emotions: Relations to childrens
Bentler, P. M. (1989). EQS: Structural equations program man- temperament and anger behavior. Merrill-Palmer Quar-
ual. Los Angeles: BMDP Statistical Software. terly, 40, 138156.
Bentler, P. M. (1990). Comparative fit indexes in structural Eisenberg, N., Fabes, R. A., Carlo, G., & Karbon, M. (1992).
models. Psychological Bulletin, 107, 238246. Emotional responsivity to others: Behavioral correlates
Bryant, B. K. (1987). Mental health, temperament, family, and socialization antecedents. In N. Eisenberg & R. A.
and friends: Perspectives on childrens empathy and so- Fabes (Eds.), Emotion and its regulation in early develop-
cial perspective taking. In N. Eisenberg & J. Strayer ment: New directions in child development (pp. 5774). San
(Eds.), Empathy and its development (pp. 245270). Cam- Francisco: Jossey-Bass.
bridge, UK: Cambridge University Press. Eisenberg, N., Fabes, R. A., & Murphy, B. C. (1996). Parents
Buck, R. (1984). The communication of emotion. New York: reactions to childrens negative emotions: Relations to
Guilford Press. childrens social competence and comforting behavior.
Bugental, D. B., & Goodnow, J. J. (1998). Socialization pro- Child Development, 67, 22272247.
cesses. In N. Eisenberg (Ed.), W. Damon (Series Ed.), Hand- Eisenberg, N., Fabes, R. A., Murphy, B., Karbon, M., Maszk,
book of child psychology: Vol. 3. Social, emotionality, and per- P., Smith, M., OBoyle, C., & Suh, K. (1994). The relations
sonality development (5th ed., pp. 389462). New York: of emotionality and regulation to dispositional and situ-
Wiley. ational empathy-related responding. Journal of Personal-
Carson, J. L., & Parke, R. D. (1996). Reciprocal negative af- ity and Social Psychology, 66, 776797.
fect in parent-child interactions and childrens peer com- Eisenberg, N., Fabes, R. A., Murphy, B., Karbon, M., Smith,
petency. Child Development, 67, 22172226. M., & Maszk, P. (1996). The relations of childrens dispo-
Coie, J. D., Terry, R., Dodge, K. A., & Underwood, M. K. sitional empathy-related responding to their emotional-
(1993). A teacher checklist of childrens social behavior. Un- ity, regulation, and social functioning. Developmental
published manuscript, Duke University. Psychology, 32, 195209.
Collins, W. A. (1990). Parent-child relationships in the transi- Eisenberg, N., Fabes, R. A., Murphy, M., Maszk, P., Smith,
tion to adolescence: Continuity and change in interaction, M., & Karbon, M. (1995). The role of emotionality and
affect, and cognition. In R. Montemayor, G. R. Adams, & regulation in childrens social functioning: A longitudi-
T. P. Gullotta (Eds.), From childhood to adolescence: A transi- nal study. Child Development, 66, 12391261.
tional period? (pp. 85106). Newbury Park, CA: Sage. Eisenberg, N., Fabes, R. A., Nyman, M., Bernzweig, J., &
Cook, W. L., Kenny, D. A., & Goldstein, M. J. (1991). Parental Pinuelas, A. (1994). The relations of emotionality and
affective style risk and the family system: A social rela- regulation to childrens anger-related reactions. Child
tions model analysis. Journal of Abnormal Psychology, 100, Development, 65, 109128.
492501. Eisenberg, N., Fabes, R. A., Schaller, M., Carlo, G., & Miller,
Davies, P. T., & Cummings, E. M. (1994). Marital conflict P. A. (1991). The relations of parental characteristics and
and child adjustment: An emotional security hypothesis. practices to childrens vicarious emotional responding.
Psychological Bulletin, 116, 387411. Child Development, 62, 13931408.
Denham, S. A. (1993). Maternal emotional responsiveness Eisenberg, N., Fabes, R. A., Shepard, S. A., Murphy, B. C.,
and toddlers social-emotional competence. Journal of Guthrie, I. K., Jones, S., Friedman, J., Poulin, R., &
Child Psychology and Psychiatry, 34, 715728. Maszk, P. (1997). Contemporaneous and longitudinal
Derryberry, D., & Rothbart, M. K. (1988). Arousal, affect, prediction of childrens social functioning from regula-
and attention as components of temperament. Journal of tion and emotionality. Child Development, 68, 642664.
Personality and Social Psychology, 55, 958966. Emde, R. N., Plomin, R., Robinson, J., Corley, R., DeFries, J.,
Diener, M., Mangelsdorf, S., Fosnot, K., & Kienstra, M. Fulker, D. W., Reznick, J. S., Campos, J., Kagan, J., &
(1997, April). Effects of maternal involvement on toddlers Zahn-Waxler, C. (1992). Temperament, emotion, and
emotion regulation strategies. Paper presented at the bien- cognition at fourteen months: The MacArthur Longitu-
nial meeting of the Society for Research in Child Devel- dinal Twin Study. Child Development, 63, 14371455.
opment, Washington DC. Fabes, R. A., Eisenberg, N., & Bernzweig, J. (1990). The Cop-
Dunn, J., & Brown, J. (1994). Affect expression in the family, ing with Childrens Negative Emotions Scale: Description
childrens understanding of emotions, and their interac- and Scoring. Unpublished scale, Department of Family
tions with others. Merrill-Palmer Quarterly, 40, 120137. Resources and Human Development, Arizona State
Eccles, J. S., Wigfield, A., & Schiefele, U. (1998). Motivation University.
to succeed. In N. Eisenberg (Ed.), W. Damon (Series Ed.), Fabes, R. A., Eisenberg, N., Karbon, M., Bernzweig, J., Speer,
Handbook of child psychology: Vol. 3. Social, emotional, and A. L., & Carlo, G. (1994). Socialization of childrens vicar-
personality development 10171095. New York: Wiley. ious emotional responding and prosocial behavior: Rela-
534 Child Development

tions with mothers perceptions of childrens emotional Matheny, A. P., Jr. (1986). Stability and change of infant tem-
reactivity. Developmental Psychology, 30, 4455. perament: Contributions from the infant, mother, and
Fagot, B. I. (1995). Classification of problem behaviors of family environment. In G. Kohnstamm (Ed.), Temperament
young children: A comparison of four systems. Journal of discussed (pp. 4958). Berwyn, PA: Swets North America.
Applied Developmental Psychology, 16, 95106. McNally, S., Eisenberg, N., & Harris, J. D. (1991). Consis-
Garner, P. W., Jones, D. C., & Miner, J. L. (1994). Social com- tency and change in maternal child-rearing practices
petence among low-income preschoolers: Emotion so- and values: A longitudinal study. Child Development, 62,
cialization practices and social cognitive correlates. Child 190189.
Development, 65, 622637. Murphy, B. C., Eisenberg, N., Fabes, R. A., Shepard, S., &
Gottman, J. M., Katz, L. F., & Hooven, C. (1996). Parental Guthrie, I. K. (in press). Consistency and change in chil-
meta-emotion philosophy and the emotional life of fam- drens emotionality and regulation: A longitudinal study.
ilies: Theoretical models and preliminary data. Journal of Merrill-Palmer Quarterly.
Family Psychology, 10, 243268. Parke, R. D., & Buriel, R. (1998). Socialization in the family:
Gottman, J. M., Katz, L. F., & Hooven, C. (1997). Metaemo- Ethnic and ecological perspectives. In N. Eisenberg
tion: How families communicate emotionally. Mahwah, NJ: (Ed.), W. Damon (Series Ed.), Handbook of child psychol-
Erlbaum. ogy: Vol. 3. Social, emotionality, and personality develop-
Hoffman, M. L. (1983). Affective and cognitive processes in ment (5th ed., pp. 463 552). New York: Wiley.
moral internalization. In E. T. Higgins, D. N. Ruble & Parke, R. D., Cassidy, J., Burks, V. M., Carson, J. L., &
W. W. Hartup (Eds.), Social cognition and social develop- Boyum, L. (1992). Familial contributions to peer compe-
ment: A sociocultural perspective (pp. 236274). Cam- tence among young children: The role of interactive and
bridge, MA: Cambridge University Press. affective processes. In R. D. Parke & G. W. Ladd (Eds.),
Holden, G. W., & Edwards, I. A. (1989). Parental attitudes Family-peer relationships: Modes of linkage (pp. 107134).
toward child rearing: Instruments, issues, and implica- Hillsdale, NJ: Erlbaum.
tions. Psychological Bulletin, 106, 2958. Patterson, G. R. (1982). Coercive family processes. Eugene, OR:
Kendall, P. C., & Wilcox, L. E. (1979). Self-control in chil- Castilla Press.
dren: The development of a rating scale. Journal of Con- Plomin, R., & Stocker, C. (1989). Behavioral genetics and
sulting and Clinical Psychology, 47, 10201030. emotionality. In J. S. Reznick (Ed.), Perspectives on behav-
Kliewer, W., Fearnow, M. D., & Miller, P. A. (1996). Coping ioral inhibition (pp. 219240). Chicago: University of Chi-
socialization in middle childhood: Tests of maternal cago Press.
and paternal influences. Child Development, 67, 2339 Poulin, R. (1997). Psychometric properties of the Coping with
2357. Childrens Negative Emotions Scale. Unpublished mas-
Larsen, R. J., & Diener, E. (1987). Affect intensity as an indi- ters thesis, Arizona State University, Tempe.
vidual difference characteristic: A review. Journal of Re- Power, T. G., & Shanks, J. A. (1989). Parents as socializers:
search in Personality, 21, 139. Maternal and paternal views. Youth and Youth and Adoles-
Larson, R., & Richards, M. H. (1991). Daily companionship cence, 18, 203220.
in late childhood and early adolescence: Changing de- Rothbart, M. K., Ahadi, S. A., & Hershey, K. L. (1994). Tem-
velopmental contexts. Child Development, 62, 284300. perament and social behavior in childhood. Merrill-
Lochman, J. E., & Conduct Problems Prevention Research Palmer Quarterly, 40, 2139.
Group. (1995). Screening of child behavior problems for Rothbart, M. K., & Maccoby, E. E. (1966). Parents differen-
prevention programs at school entry. Journal of Consult- tial reactions to sons and daughters. Journal of Personality
ing and Clinical Psychology, 63, 549559. and Social Psychology, 4, 237243.
Lytton, H. (1990). Child and parent effects in boys conduct Thomas, A. (1984). Temperament research: Where we are,
disorder: A reinterpretation. Developmental Psychology, where we are going. Merrill-Palmer Quarterly, 30, 103109.
26, 683697. Tomkins, S. S. (1963). Affect, imagery, consciousness: Vol. 2.
Maccoby, E. E., & Martin, J. A. (1983). Socialization in the Negative affects. New York: Springer.
context of the family: Parent-child interaction. In E. M. Watson, D., Clark, L. A., & Tellegen, A. (1988). Development
Hetherington (Ed.), P. H. Mussen (Series Ed.), Handbook and validation of brief measures of positive and nega-
of child psychology: Vol. 4. Socialization, personality, and so- tive affect: The PANAS Scales. Journal of Personality and
cial development (pp. 1101). New York: Wiley. Social Psychology, 54, 10631070.

Potrebbero piacerti anche