+ All Categories
Home > Documents > Prediction of Elementary School Children's Externalizing Problem Behaviors from Attentional and...

Prediction of Elementary School Children's Externalizing Problem Behaviors from Attentional and...

Date post: 23-Feb-2023
Category:
Upload: independent
View: 0 times
Download: 0 times
Share this document with a friend
17
Child Development, September/October 2000, Volume 71, Number 5, Pages 1367–1382 Prediction of Elementary School Children’s Externalizing Problem Behaviors from Attentional and Behavioral Regulation and Negative Emotionality Nancy Eisenberg, Ivanna K. Guthrie, Richard A. Fabes, Stephanie Shepard, Sandra Losoya, Bridget C. Murphy, Sarah Jones, Rick Poulin, and Mark Reiser The purpose of this study was to examine the moderating role of individual differences in negative emotional- ity in the relations of behavioral and attentional (emotional) regulation to externalizing problem behaviors. Teachers’ and one parent’s reports of children’s regulation (attentional and behavioral), emotionality, and problem behavior were obtained when children were in kindergarten to grade 3 and two years later ( N 5 169; 146 in major analyses); children’s behavioral regulation also was assessed with a measure of persistence. Ac- cording to the best fitting structural equation model, at two ages behavioral dysregulation predicted external- izing behavior problems for children both high and low in negative emotionality, whereas prediction of prob- lem behavior from attentional control was significant only for children prone to negative emotionality. There were unique, additive effects of behavioral and attentional regulation for predicting problem behavior as well as moderating effects of negative emotionality for attentional regulation. INTRODUCTION In recent years there has been a growing interest in the role of temperamental or personality characteris- tics in the development of problem behavior (Caspi, 1998; Rothbart & Bates, 1998). Among those disposi- tional characteristics of most interest have been chil- dren’s regulatory capabilities and their emotionality/ reactivity (e.g., Caspi, Henry, McGee, Moffitt, & Silva, 1995; Colder & Chassin, 1997), which are viewed by some researchers as the core of temperament (Roth- bart & Bates, 1998). Indeed, several researchers have argued that emotionality and/or regulation play a central role in problem behavior and psychopathol- ogy (Eisenberg & Fabes, 1992; Moffitt, 1993; Wein- berger & Schwartz, 1990; also see Rothbart, Posner, & Hershey, 1995). Eisenberg and Fabes (1992) proposed a heuristic model in which individual differences in emotional- ity and regulation often have additive or multiplica- tive (i.e., statistically interacting) effects in regard to the prediction of quality of social behavior. In regard to externalizing problem behavior, they proposed that children who are prone to intense emotion, especially negative emotion, and who are also low in regulation would be expected to be especially high in externaliz- ing problem behavior, particularly externalizing be- havior associated with negative emotion (e.g., reactive or emotionally driven aggression; Dodge, Lochman, Harnish, Bates, & Pettit, 1997). In their heuristic model, Eisenberg and Fabes differ- entiated among several aspects of emotion-relevant regulation, including emotional regulation and atten- tional regulation. Emotion regulation is the process of initiating, maintaining, modulating, or changing the occurrence, intensity, or duration of internal feeling states and emotion-related physiological processes, often in the service of accomplishing one’s goals. This type of regulation, which often includes the regula- tion of attention (attention shifting and focusing) and cognitions (e.g., positive cognitive restructuring), in- volves the regulation of emotion-relevant internal states and processes. In contrast, emotion-related behav- ioral regulation is defined as the process of initiating, maintaining, inhibiting, modulating, or changing the occurrence, form, and duration of behavioral con- comitants of emotion, including observable facial or gestural responses and other behaviors that stem from, or are associated with, internal emotion-related psychological or physiological internal states and goals. This type of regulation (henceforth called be- havioral regulation) can involve the communication of emotion and the inhibition or activation of behav- ior linked to emotion. In studies of temperament and personality, behavioral regulation is often assessed with measures of behavioral inhibition or impulsivity (e.g., Derryberry & Rothbart, 1988; Kochanska, Mur- ray, & Coy, 1997; Rothbart, Ahadi, & Hershey, 1994). The critical difference between what we have labeled emotion regulation and behavioral regulation is the locus of regulation: internal psychological and physi- ological reactions or overt behavior driven by, or as- sociated with, aroused internal states. Emotion regulation and emotion-related behav- ioral regulation obviously are intricately and perhaps © 2000 by the Society for Research in Child Development, Inc. All rights reserved. 0009-3920/2000/7105-0020
Transcript

Child Development, September/October 2000, Volume 71, Number 5, Pages 1367–1382

Prediction of Elementary School Children’s Externalizing Problem Behaviors from Attentional and Behavioral Regulation and Negative Emotionality

Nancy Eisenberg, Ivanna K. Guthrie, Richard A. Fabes, Stephanie Shepard, Sandra Losoya, Bridget C. Murphy, Sarah Jones, Rick Poulin, and Mark Reiser

The purpose of this study was to examine the moderating role of individual differences in negative emotional-ity in the relations of behavioral and attentional (emotional) regulation to externalizing problem behaviors.Teachers’ and one parent’s reports of children’s regulation (attentional and behavioral), emotionality, andproblem behavior were obtained when children were in kindergarten to grade 3 and two years later (

N

5

169;146 in major analyses); children’s behavioral regulation also was assessed with a measure of persistence. Ac-cording to the best fitting structural equation model, at two ages behavioral dysregulation predicted external-izing behavior problems for children both high and low in negative emotionality, whereas prediction of prob-lem behavior from attentional control was significant only for children prone to negative emotionality. Therewere unique, additive effects of behavioral and attentional regulation for predicting problem behavior as wellas moderating effects of negative emotionality for attentional regulation.

INTRODUCTION

In recent years there has been a growing interest inthe role of temperamental or personality characteris-tics in the development of problem behavior (Caspi,1998; Rothbart & Bates, 1998). Among those disposi-tional characteristics of most interest have been chil-dren’s regulatory capabilities and their emotionality/reactivity (e.g., Caspi, Henry, McGee, Moffitt, & Silva,1995; Colder & Chassin, 1997), which are viewed bysome researchers as the core of temperament (Roth-bart & Bates, 1998). Indeed, several researchers haveargued that emotionality and/or regulation play acentral role in problem behavior and psychopathol-ogy (Eisenberg & Fabes, 1992; Moffitt, 1993; Wein-berger & Schwartz, 1990; also see Rothbart, Posner, &Hershey, 1995).

Eisenberg and Fabes (1992) proposed a heuristicmodel in which individual differences in emotional-ity and regulation often have additive or multiplica-tive (i.e., statistically interacting) effects in regard tothe prediction of quality of social behavior. In regard toexternalizing problem behavior, they proposed thatchildren who are prone to intense emotion, especiallynegative emotion, and who are also low in regulationwould be expected to be especially high in externaliz-ing problem behavior, particularly externalizing be-havior associated with negative emotion (e.g., reactiveor emotionally driven aggression; Dodge, Lochman,Harnish, Bates, & Pettit, 1997).

In their heuristic model, Eisenberg and Fabes differ-entiated among several aspects of emotion-relevantregulation, including emotional regulation and atten-tional regulation.

Emotion regulation

is the process of

initiating, maintaining, modulating, or changing theoccurrence, intensity, or duration of internal feelingstates and emotion-related physiological processes,often in the service of accomplishing one’s goals. Thistype of regulation, which often includes the regula-tion of attention (attention shifting and focusing) andcognitions (e.g., positive cognitive restructuring), in-volves the regulation of emotion-relevant internalstates and processes. In contrast,

emotion-related behav-ioral regulation

is defined as the process of initiating,maintaining, inhibiting, modulating, or changing theoccurrence, form, and duration of behavioral con-comitants of emotion, including observable facial orgestural responses and other behaviors that stemfrom, or are associated with, internal emotion-relatedpsychological or physiological internal states andgoals. This type of regulation (henceforth called be-havioral regulation) can involve the communicationof emotion and the inhibition or activation of behav-ior linked to emotion. In studies of temperament andpersonality, behavioral regulation is often assessedwith measures of behavioral inhibition or impulsivity(e.g., Derryberry & Rothbart, 1988; Kochanska, Mur-ray, & Coy, 1997; Rothbart, Ahadi, & Hershey, 1994).The critical difference between what we have labeledemotion regulation and behavioral regulation is thelocus of regulation: internal psychological and physi-ological reactions or overt behavior driven by, or as-sociated with, aroused internal states.

Emotion regulation and emotion-related behav-ioral regulation obviously are intricately and perhaps

© 2000 by the Society for Research in Child Development, Inc.All rights reserved. 0009-3920/2000/7105-0020

1368 Child Development

sometimes inextricably associated, particularly in in-fancy. Not only is behavioral regulation affected byinternal emotion-related processes, but the conse-quences of behavioral regulation often may influencethe course of internal emotion-related processes andstates and may modify future emotion-related cogni-tive or physiological processes. Moreover, it wouldappear that some processes, such as the ability to vol-untarily inhibit behavior, can serve as a mechanismfor both emotional regulation or behavioral regula-tion. For example, inhibition control—the ability tovoluntarily cease behavior—most likely is usuallyused to curb the expression of behavior stemmingfrom or associated with emotion. It also could beused, however, to inhibit approach toward, and expo-sure to, a potentially distressing situation—whichlikely will affect the internal experience of emotion.Moreover, although we believe that attentional con-trol, especially attention shifting, most often is used tomodulate internal emotion-related processing, atten-tional control (especially attention focusing) some-times may be used to manage overt behavior associ-ated with emotion when the emotion itself is notsufficiently regulated. Nonetheless, attentional regu-lation is likely used primarily for managing internalemotion-related states, whereas voluntary behavioralcontrol is used primarily to manage the expression ofemotion-related behavior. Moreover, even if the twodo overlap in terms of their involvement in regulationof internal states versus overt behavior, they are dif-ferent in that one is an internal process and the otheris related to overt behavior.

There is initial evidence that children’s lack of reg-ulation is associated with externalizing behavior. Forexample, Hart, Hofmann, Edelstein, and Keller(1997), Newman, Caspi, Moffitt, and Silva (1997), andRobins, John, Caspi, Moffitt, and Stouthamer-Loeber(1996), identified groups of undercontrolled Icelan-dic, European American and African American, andNew Zealand children or young adults, respectively.In all three studies, undercontrolled individuals wereprone to problem behaviors concurrently or in lateradolescence or adulthood. In addition, regulation onbehavioral measures (e.g., persistence on a task or de-lay of gratification; Eisenberg, Fabes, Guthrie, et al.,1996; Krueger, Caspi, Moffitt, White, & Stouthamer-Loeber, 1996; Oosterlaan & Sergeant, 1996) has beenlinked to low levels of externalizing kinds of problems.Rothbart and Bates (1998), in their review of the temper-ament literature, concluded that early unmanageabilityis especially linked to later externalizing problems. Al-though there is less evidence regarding attentional reg-ulation, low attentional regulation and associated ex-ecutive control functions appear to be related to

proneness to anger (Derryberry & Rothbart, 1988), ex-ternalizing problem behavior and conduct disorders(Eisenberg, Fabes, Guthrie, et al., 1996; Moffitt, 1993),aggression (Hart, Keller, Edelstein, & Hofmann, 1998),and psychopathy (O’Brien & Frick, 1996; Patterson &Newman, 1993).

Negative emotionality also has been linked to ex-ternalizing (e.g., Rothbart et al., 1994; Eisenberg,Fabes, et al., 1997; Stice & Gonzales, 1998) as well asinternalizing (Clark, Watson, & Mineka, 1994; Teglasi& MacMahon, 1990) problem behaviors. In manystudies, however, measures of emotionality are com-bined with assessments of regulation (Block, Block, &Keyes, 1988; Caspi et al., 1995; Newman et al., 1997;Pulkkinen & Hamalainen, 1995), so much less is knownabout the role of negative emotionality in externalizingproblem behavior.

Whether negative emotionality is associated withexternalizing problem behaviors may depend on theindividual’s level of regulation. Even if children areprone to anger or frustration, they may not exhibituncontrolled, emotional behavior if they can inhibitand control their behavior. Furthermore, it is reason-able to hypothesize that the regulation of internalstates is a more important predictor of externalizingproblem behavior for children predisposed to experi-ence negative emotions than for other children. Emo-tionally reactive children who can modulate the in-tensity and duration of their experience of negativeemotion through processes such as attention shiftingand focusing may be better able to manage the emo-tion internally so that they are relatively unlikely toexpress the negative emotion in their behavior. Con-sistent with the view that attentional control is linked tomodulation of negative emotion, infants and childrenwho can regulate their attention tend to be relativelylow in observed or other-rated negative emotionality(Eisenberg et al., 1993; Rothbart, Ziaie, & O’Boyle, 1992;Ungerer et al., 1990).

Because children who are relatively low in emo-tional reactivity (especially in negative emotionality)do not experience as much or as intense negative emo-tion as other children, attentional control—which wehypothesize serves primarily to modulate internalemotion-related processes—may be a poorer predictorof externalizing behavior for children low in negativeemotionality than for emotionally reactive children.Behavioral regulation may, however, be linked to ex-ternalizing problems regardless of the level of thechild’s emotionality because behaviorally regulated in-dividuals would be expected to be low in externalizingeven if they were emotional. Because covert externaliz-ing behaviors such as stealing or lying often may beperformed primarily as a result of the desire to obtain a

Eisenberg et al. 1369

goal, even if one is not emotionally aroused, low be-havioral regulation would be expected to predict exter-nalizing problems for children who are low as well ashigh in emotion reactivity. Similarly, researchers havenoted that some aggression—proactive aggression—is organized, unemotional, and aimed at fulfilling aneed or desire (Dodge et al., 1997). Children who arenot impulsive and who can inhibit inappropriate be-havior are likely to be low in these forms of externaliz-ing problems, regardless of their level of dispositionalemotionality, whereas those low in behavioral regula-tion likely are prone to such behavior. In short, becausebehavioral regulation acts primarily to inhibit the ex-ternalizing behavior, be it the result of emotion or not,it is likely a predictor of low externalizing behavior formost children. Nonetheless, it is possible that the rela-tion of behavioral regulation to externalizing problemsis greater for children prone to negative emotionality ifthey are predisposed to engage in more (rather thanmerely different) types of externalizing behavior thanare less emotional children.

On the basis of this line of reasoning, we hypothe-sized that children’s externalizing behavior is pre-dicted by the multiplicative effects of individual dif-ferences in emotionality and regulation and thatattentional control, and perhaps behavioral regula-tion, would be a stronger predictor of externalizingproblem behaviors for children high versus low indispositional negative emotionality. Initial researchsupports the notion that emotionality and regulationinteract in predicting externalizing behavior, al-though to our knowledge no one has differentiatedbetween various types of regulation in this research.Research on the topic is scarce; in their review of theliterature on Temperament

3

Temperament interac-tions used to predict problem behavior, Rothbart andBates (1998, p. 158) concluded that there was, as yet,little evidence for this type of interaction, although“there are indications that some interactive effectswill be discovered.” They emphasized the need forreplication in the demonstration of Trait

3

Trait inter-actions as predictors of socially relevant outcomes.

In the research on substance abuse/alcohol risk,there is some evidence of additive and interaction effectsof emotionality and regulation. Colder and Chassin(1997) found that negative emotionality and impul-sivity provided unique and additive prediction of al-cohol use, even when parental alcoholism and posi-tive emotionality were controlled in the regressionanalysis (impulsivity but not negative affectivity pre-dicted alcohol-related impairment). Impulsivity mod-erated the effects of positive affectivity on alcohol useand alcohol-related impairment. Impulsive adoles-cents who also were prone to low levels of positive af-

fectivity were higher in alcohol use and impairment.Negative emotionality did not moderate the relation ofimpulsivity to alcohol use and impairment, although itis impossible to know if there would have been moder-ation if positive emotionality had not been controlledin the analysis. Moreover, depressed, impulsive ado-lescents have been found to drink more heavily thandepressed, nonimpulsive adolescents or nonde-pressed adolescents (Hussong & Chassin, 1994). Inanother sample of adolescents, however, the interac-tion of impulsivity and anger was not a significantpredictor of substance use (Colder & Stice, 1998).

Pandina, Johnson, and LaBouvie (1992) found amodest number of interactional effects when predictingadolescent substance abuse from negative emotionalityand regulation. They assessed variables labeled “nega-tive emotionality” (including a variety of negative emo-tions) and “arousal,” with arousal reflecting disinhibi-tion, impulsivity, and experience seeking. Thus, theirindex of arousal reflected primarily lack of regulation.For individuals who were stable in arousal (lack of reg-ulation) and negative affectivity over a 3-year period,moderating effects were obtained. Adolescents high indisinhibition (labeled high in arousal) and high in neg-ative affect exhibited the most alcohol-related motiva-tional problems, whereas those low in both negativeaffect and arousal (regulated) exhibited the least prob-lems with marijuana (at least for one sex). Interactionsbetween arousal and negative emotionality were rarefor adolescents who were not stable in both predictors,perhaps because these adolescents were not as extremein dispositional regulation or emotionality as adoles-cents who were stable on these characteristics.

In regard to other types of externalizing behaviors,we know of only two studies that have examined theissue of Trait

3

Trait interactions. Colder and Stice(1998) found that adolescents’ delinquency was pre-dicted by the additive and interactive effects of angerand impulsivity. High levels of anger and impulsivitywere both related to concurrent delinquency, and an-ger was associated with high levels of delinquencyonly for adolescents high in impulsivity. When, how-ever, anger and impulsivity were used to predict de-linquency 9 months prospectively (controlling for ini-tial delinquency), only anger predicted delinquencyand there was no multiplicative effect.

Eisenberg, Fabes, et al. (1995, 1997) examined the in-teractions between externalizing problems and emo-tionality and found few significant interactions (onefor parent reports only at one age). Their sample size,however, was relatively small, so their power to detectstatistical interactions was quite limited. In a largerstudy, Eisenberg, Fabes, Guthrie, et al. (1996) examinedthe concurrent prediction of externalizing problems in

1370 Child Development

kindergarten to third graders from negative emotion-ality (intensity and frequency combined) and a com-posite score of behavioral and attentional regulation.They found both additive and multiplicative effects:children prone to externalizing problems were high innegative emotionality and low in regulation and, ac-cording to the interaction effect, regulation was a stron-ger predictor of problem behavior for children prone tonegative emotionality. There also is research indicatingthat individual differences in emotionality and regula-tion interact when predicting social competence, pro-social behavior, and sympathy. In general, regulation ismore strongly associated with social competence (e.g.,peer status, socially appropriate or prosocial behavior)for children high in negative emotionality (Eisenberg,Fabes, et al., 1995, 1997; Eisenberg, Guthrie, et al.,1997). Moreover, sympathetic children tend to be thosehigh in both general emotional intensity (i.e., intensityof emotions, regardless of valence) and regulation(Eisenberg, Fabes, Murphy, et al., 1996; Eisenberg,Fabes, Shepard, et al., 1998). These findings supportthe notion that individual differences in emotionalityand regulation often interact when predicting qualityof social functioning.

The present study is a 2-year longitudinal follow-upof the children in the Eisenberg, Fabes, Guthrie, et al.(1996) study. In this follow-up, reports of both regula-tion and emotionality were obtained from the primarycaregiving parent (usually mother) and the child’steacher. In addition, children played a puzzle box gamein which their persistence was assessed as an index ofbehavioral regulation. Attentional regulation and be-havioral regulation were treated as separate latent con-structs in the structural equation model, and relationsbetween earlier and later measures of constructs couldbe controlled in the analysis. We hypothesized that reg-ulation would predict externalizing problem behaviorat the initial assessment as well as 2 years later and thatsuch prediction would be stronger for children prone tonegative emotionality. Moreover, we expected atten-tional regulation (and perhaps behavioral regulation) tobe an especially important predictor of externalizingproblems for children prone to negative emotionality. Inaddition, we hypothesized that the consistency acrosstime would be carried through consistency in children’sdispositional emotionality and regulation rather thanthrough mere consistency of problem behavior.

METHOD

Participants

Participants were involved in a longitudinal studythat included two assessments. The initial sample

consisted of 199 children (97 girls, 102 boys) in kin-dergarten, first, second, and third grades (henceforthcalled T1; Eisenberg, Fabes, Guthrie, et al., 1996).These children were followed up 2 years later (hence-forth called T2). At this time, the sample included 169children (86 girls, 83 boys).

Only children who completed the puzzle box taskand with parental and teacher ratings of attentionalcontrol, behavioral regulation, and problem behav-iors could be included in the structural equation anal-yses in the present study. Children with these datawere 73 girls and 73 boys who had data both at T1 andT2. The mean ages for these children were 88.62months,

SD

5

13.71, at T1 and 112.62 months,

SD

5

13.70, at T2. The age range of the children at each timepoint was approximately 2 years.

The children were predominantly EuropeanAmerican. Demographics on the 146 children withcomplete data at T2 were 82% European American,11% Hispanic, 1% African American, 2% NativeAmerican, 1% Asian, and 3% of mixed origin. Meannumber of years of maternal and paternal educationat T2 were 14.91 (

range

5

11–20) and 14.93 years at T2(

range

5

8–20), respectively. Family income rangedfrom $8,000 to $150,000 (

M

5

47,690,

SD

5

25,030, me-dian

5

$45,000) at T1 and from $4,000 to $175,000(

M

5

$49,007,

SD

5

28,340, median

5

$50,000) at T2.Based on

t

-tests and

x

2

(and ANOVAs), there weresome differences between individuals who attritedfrom T1 to T2 and those used to compute the modelsin this paper. Families who did not return tended tobe disproportionally from minority groups, particu-larly African American,

x

2

(5,

N

5

174

)

5

15.29,

p

,

.009; attrited families were lower in maternal and pa-ternal education,

t

s(172, 159)

5

2

2.69 and

2

2.70,

p

s

,

.008, and were marginally lower in income,

t

(165)

52

1.68,

p

,

.10. Attrited children were viewed byteachers as less attentionally regulated,

t

s(173)

5

2

2.19,

p

s

,

.029, and were rated as higher in problem behav-ior by teachers and lower by parents,

t

s(174)

5

1.95and

2

2.14,

p

s

,

.053 and .034. When the 146 childrenin the major analysis were compared with the attritedchildren and those with insufficient data to be in-cluded in the models, the only differences were forrace,

x

2

(5,

N

5

199

)

5

11.34,

p

,

.045, and parents’ re-ports of behavioral regulation,

t

(197)

5

2

2.06,

p

,

.04(with fewer minorities and unregulated children inthe retained sample).

Procedures

The first data collection occurred over two aca-demic semesters of one school year (T1); childrenwere followed up 2 years later, again over two aca-

Eisenberg et al. 1371

demic semesters (T2). Measures of attentional control,behavioral regulation, and problem behavior wereobtained at both time points.

Usually mothers completed parent report mea-sures at the laboratory; however, four fathers com-pleted these measures at T1 and six fathers at T2. Aspart of the laboratory session, children participated ina regulation task (the puzzle box). Teacher measurespertaining to children’s attentional control, behav-ioral regulation, and problem behaviors were sent outnear the end of each semester at T1 and T2.

Measures

Regulation

Regulation measures included the following: (1) abehavioral measure of regulation (the puzzle box), (2)parents’ and teachers’ reports of children’s ego con-trol; and (3) parents’ and teachers’ reports of temper-amental voluntary attentional regulation. See Eisen-berg, Fabes, Guthrie, et al. (1996) for descriptions and

a

s of T1 measures. The puzzle task and the measures ofego control are viewed as indexes of behavioral regula-tion, whereas adults’ reports of attentional control areviewed as primarily measures of emotion regulation.

Regulation puzzle box task.

At T1 and T2, children’sregulation was assessed with a puzzle box task. Chil-dren were asked to try to assemble a wooden puzzlein a large box without looking at it. A cloth coveredthe front; children slipped their arms through sleevesto get into the box. The cloth could be lifted up so thata child could cheat by looking. Children were toldthat if they finished the puzzle within 5 min, theywould receive an attractive prize and that they couldcall the experimenter back when finished by ringing abell. A timer was set so that the children could ascer-tain how much time had elapsed. The seconds a childpersisted without cheating on the puzzle box whenalone were timed by means of a hidden camera; inter-rater reliability based on 76 children was .93. Becausesome children called for the experimenter before 5min elapsed, proportion of time working was com-puted by dividing the number of seconds working bythe total amount of time with the puzzle.

Cheating on the task also was timed (interrater reli-ability

5

.93). Cheating on this task was significantly re-lated to parents’ and teachers’ reports of problem be-havior,

r

s(144)

5

.26 and .32,

p

s

,

.001, which is evidenceof the validity of adults’ reports of problem behavior.

Ego control.

At T1, parents and teachers completedthe Block and Block Q-sort (Block & Block, 1969,1980). They sorted the cards into 9 unequal piles (witha normal distribution) on a nine-point scale (1

5

most

undescriptive to 9

5

most descriptive). Because theQ-sort took so long, at T2 we constructed question-naires using the Q-sort items for ego control (and neg-ative emotionality) and adults rated children on thesame nine-point scale used for the Q-sort procedure.Some items were simplified slightly by addingphrases from Caspi et al. (1992).

To construct the ego control scale, at T1 we usedthe list of items identified by Block and Block (1969;personal communication, 1992) as being highly repre-sentative of ego control (i.e., were rated by expertswith a value of 7.7 or higher, or 2.3 or lower, on a nine-point scale). Then items that were deemed by consen-sus of the first three authors as reflecting specific socialskills, moral behaviors, problem behaviors, or overtemotional responding were eliminated. As a conse-quence, our assessment of the relations of regulationto problem behavior was relatively uncontaminatedby the problem of overlapping items (see Sanson,Prior, & Kyrios, 1990).

This procedure resulted in a 19-item ego-controlscale (e.g., “Is inhibited and constricted”; see Eisen-berg, Fabes, Guthrie, et al., 1996, for item numbers).The

a

s for parents and teachers were .80 and .84 at T2,respectively. All items except one pertained primarilyto behavioral regulation; one item could refer to bothattentional regulation (”is attentive and able to con-centrate”) and the ability to sit still and work.

Attention shifting and focusing.

At T1 and T2, parentsand teachers completed 11-item attention shifting andattention focusing subscales from Rothbart’s ChildBehavior Questionnaire, slightly adapted for teachers(Rothbart et al., 1994; e.g., “Has an easy time leavingplay to come inside for school work” and “Whenpicking up toys or other tasks, usually keeps at thetask until it’s done,” respectively). It is important tonote that both subscales tap the ability to voluntarilycontrol attention as needed in particular situations.The

a

s at T2 for attention shifting and attention focus-ing were .74 and .81 for parents and .85 and .88 forteachers). Parents and teachers rated how true itemswere for children on a seven-point scale (ranging from1

5

“extremely untrue” to 7

5

“extremely true”).At T1, reports for attention shifting and attention

focusing were significantly correlated for both par-ents and teachers,

r

s(165, 157)

5

.34 and .46,

p

s

,

.001.Thus, attention shifting and attention focusing rat-ings were standardized and averaged in all subse-quent analyses (henceforth called attentional control).

Negative Emotionality

Children’s negative emotional intensity was rated1

5

“never” to 7

5

“always” by parents and teachers

1372 Child Development

by using an adaptation of Larsen and Diener’s (1987)Affect Intensity Scale (Eisenberg et al., 1995). Themeasure included five items pertaining to negativeemotionality (e.g., “When my child experiences anxi-ety, it normally is very strong”; T2

a

s

5

.74 and .85 forparents and teachers, respectively.

Adults also rated 11 items from Block and Block’s(1969, 1980) Q-sort pertaining to children’s negativeemotionality (e.g., “Is fearful and anxious [ner-vous]”). T2

a

s for parents and teachers were .84 and.86, respectively.

For parents and teachers, ratings on negative emo-tionality and the Q-sort emotionality composite weresubstantially correlated at T2,

r

s(160)

5

.65 and .79,

p

s

,

.001, respectively. Thus, the two measures werestandardized and averaged to construct a negativeemotionality composite score.

Externalizing Problem Behavior

At T1 and T2, parents and teachers rated children’sexternalizing problem behavior (1

5

“never” to 4

5

“often”) by using the 24-item screening measure ofLochman and Conduct Problems Prevention Re-search Group (1995). This measure was designed toscreen for externalizing problem behaviors; it in-cluded items drawn from existing behavior check-lists. All but one item (“sets fires”) were included(e.g., “argues,” “lies,” “aggressive to adults”; “setsfires” was dropped because we believed it would bevery infrequent and might offend parents because ofits severity). The

a

s for parents and teachers at T2were .91 and .97, respectively.

RESULTS

To examine the hypothesized relations, we used struc-tural equation modeling because it allows estimationof both the measurement and structural models to es-timate the relations among latent variables withoutconfounding effects of measurement error; allows es-timation of lagged and contemporaneous effects; andincludes procedures for multiple group analysis thatallow systematic determination of whether or notspecified linkages are similar for different subgroups(i.e., children high and low in negative emotionality;see

LISREL 7

, 1988). Although structural equationmodeling cannot prove causal relations among vari-ables, it can assess whether inferences about causalityare consistent with the data. Further, it allows one todetermine whether one model fits the data better thananother model. Note that age was not significantly re-lated to any variables except persistence on the puz-zle task at both T1 and T2,

r

s(144)

5

.26 and .28,

p

s

,

.001. In addition, at T1 girls were higher in parent-and teacher-rated attentional control and teacher-reported behavioral regulation and lower in mothers’and teachers’ reports of externalizing problem behav-ior,

t

s(144)

5

2.49, 3.14, 2.44,

2

3.09,

2

3.45,

p

s , .014,.002, .016, .002, and .001, respectively. At T2, girls werehigher than boys in teacher reports of attentional con-trol and behavioral regulation and lower in teachers’and mothers’ reports of negative emotionality, ts(144) 53.91, 3.69, 23.44, 25.09, ps , .001.

The raw data were analyzed by using Bentler’s(1989) EQS program to obtain estimates. Only partici-pants who had data for all variables were included inthe analyses. Children were split into high and lownegative emotionality groups on the basis of a mediansplit on the T2 negative emotionality composite score.The moderational effects of negative emotionality onregulation were then examined to determine whetherthe same structural equation model was consistentacross both groups. Refer to Table 1 for means and stan-dard deviations and Table 2 for correlations among allobserved (i.e., measured) variables.

Moderational Models

The initial model was derived from regressionanalyses with T1 data (see Eisenberg, Fabes, Guthrie,et al., 1996). These findings supported the moderatingrole of negative emotionality on the prediction ofproblem behavior from regulation (attentional con-trol and behavioral regulation combined). That is,regulation was most important for predicting prob-lem behavior for children high versus low in negativeemotionality.

The primary aim of this study was to examine themoderating effect of negative emotionality on regula-tion and problem behavior longitudinally. The mea-surement model consisted of six latent variables: at-tentional control at T1 and T2, behavioral regulationat T1 and T2, and problem behaviors at T1 and T2.Parents’ and teachers’ composite ratings of atten-tional control were indicators of the attentional controllatent variables at T1 and T2. Parents’ and teachers’ rat-ings of ego control and the behavioral regulation boxtask administered to the child were indicators of thebehavioral regulation latent variables at T1 and T2.The latent variables of problem behaviors comprisedtwo indicators: parents’ and teachers’ ratings of prob-lem behaviors at T1 and T2.

It was expected that the paths from attentional con-trol and behavioral regulation to problem behaviorsat both T1 and T2 would be moderated by negativeemotionality. To test this, initially these paths wereconstrained to be equal across groups and across time

Eisenberg et al. 1373

(e.g., the magnitude of the bs for the paths betweenattentional control and problem behaviors across T1and T2 and for children high in negative emotionalityand children low in negative emotionality were con-strained to be equal). The one exception (besides thefact that factor loadings of measures on a constructwere not constrained to be equal if doing so was aproblem; see those released in Figure 1) was that thelagged (i.e., across time) path of behavioral regulationfrom T1 to T2 was constrained to be equal acrossgroups, whereas the lagged path of attentional con-trol from T1 to T2 was allowed to vary across groups.This was done because the Lagrange Multiplier Teststatistics for a preliminary fully constrained model in-dicated that the constraint for the path from T1 and T2for attentional control across groups was not war-ranted. The Lagrange Multiplier Test statistics (calledmodification indices in LISREL) indicate whether therestriction given in the model is appropriate for thedata. Each degree of freedom for testing the model isobtained from an overidentifying restriction on a pa-rameter. Further, the Lagrange Multiplier Test statisticssuggest specific parameters, previously fixed in themodel, for which removal of the constraints acrossgroups would result in a model that better representsthe data.

If the parameters were found to be equivalentacross groups (i.e., if the same model fit the data forchildren high and low in negative emotionality across

T1 and T2), then moderation is not present (Baron &Kenny, 1986; Bollen, 1989). This nearly fully con-strained model had a x2(128) 5 169.845, p , .008 anda Comparative Fit Index of .950.

Because negative emotionality had been found tomoderate the relations between regulation and prob-lem behaviors at T1 in the regression analyses (Eisen-berg, Fabes, Guthrie, et al., 1996), in the second model(Figure 1) the paths from attentional control to prob-lem behaviors and behavioral regulation to problembehaviors were allowed to vary across groups andtime. The exception was for the path from attentionalcontrol to problem behavior for children high in neg-ative emotionality; initial models run separately forchildren high and low in negative emotionality indi-cated that children high in negative emotionality didnot differ across time for this path. Thus, this pathwas constrained to be equal at T1 and T2 for childrenhigh in negative emotionality. This model fit the datafairly well with a x2(123, N 5 146) 5 156.116, p , .02,with a Comparative Fix Index of .961. Further, thismodel represents an improvement over the fully con-strained model with a difference of 13.729 on 5 de-grees of freedom for the likelihood ratio statistic,which is significant, p , .03. As Figure 1 indicates, thepaths from attentional control and behavioral regula-tion to problem behaviors are moderated by negativeemotionality. Specifically, the paths from attentionalcontrol to problem behaviors appear more important

Table 1 Means and Standard Deviations of Model Variables

High NegativeEmotionality

Low NegativeEmotionality

Variable M SD M SD

Parent: attentional control T1 2.17 .76 .21 .63Teacher: attentional control T1 2.10 .89 .22 .82Parent: ego control T1 4.65 .84 5.13 .58b

Teacher: ego control T1 4.91 1.02 5.71 .92Behavior regulation box T1 .64 .28 .79 .20a

Parent: problem behavior T1 2.32 .39 2.04 .33Teacher: problem behavior T1 1.84 .70 1.39 .45b

Parent: attentional control T2 4.33 .71 4.82 .57Teacher: attentional control T2 4.32 .92 5.02 .74Parent: ego control T2 4.47 1.01 5.14 .79Teacher: ego control T2 4.76 1.15 5.62 .92Behavior regulation box T2 .67 .29 .81 .25Parent: problem behavior T2 2.32 .38 1.90 .36Teacher: problem behavior T2 2.14 .73 1.37 .35b

Note: No variances (SDs are presented above) were significantly different between the two groups un-less otherwise noted. All means differed significantly across groups at p , .05 using a two-tailed t testunless otherwise noted.a Variances significantly different at p , .05.b Variances significantly different at p , .01.

1374 Child Development

for children high in negative emotionality acrosstime, whereas the paths from behavioral regulation toproblem behaviors appear more important for chil-dren low in negative emotionality.

To further examine whether the important moder-ating path was either from attentional control or be-havioral regulation to problem behaviors, two addi-tional models were computed. First, a model (Figure2) was tested that constrained the path from behav-ioral regulation to problem behaviors across childrenhigh and low in negative emotionality and across T1and T2 (so there was no moderation effect). As in theprior model, the path from attentional control toproblem behaviors was allowed to vary for childrenlow in negative emotionality across time, whereasthis path was constrained to be equal across time forchildren high in negative emotionality. This model (inFigure 2) also fit the data fairly well, x2(126, N 5 146) 5160.671, p , .02, with a Comparative Fit Index of.959. This model, like the second model, represents asignificant improvement over the fully constrainedmodel with a difference of 9.174 on 2 degrees of free-dom for the likelihood ratio statistic, p , .03. Thus,this model suggests that the paths from attentionalcontrol to problem behaviors across time are moder-ated by negative emotionality and that attentionalcontrol appears more important for children high ver-sus low in negative emotionality. In contrast, moder-ation was not necessary for the path from behavioralregulation to problem behavior. This model is moreparsimonious than the model presented in Figure 1.Adding a path from externalizing problem behaviorsat T1 to T2 in Model 2 resulted in the exact same CFIand a higher AIC (289.42700, which indicates a lesserfit), and the added path was not significant. Thus, theeffects over time were not affected through consis-tency in problem behavior.

The fourth model (Figure 3) differed from the thirdmodel as follows: (1) the paths from attentional con-trol to problem behaviors were constrained from T1to T2 and across emotionality groups, and (2) thepaths from behavioral regulation to problem behav-iors were allowed to vary across emotionality groupsand time. This model had a x2(125, N 5 146) 5 163.541,p , .012, with a Comparative Fit Index of .954. In thismodel, the paths from behavioral regulation to prob-lem behaviors were higher for children high versus lowin negative emotionality. In addition, the paths from at-tentional control to problem behavior were marginallysignificant. This model represented a marginally signif-icant improvement over the fully constrained modelwith a difference of 6.304 on three degrees of freedom,p , .10 for the likelihood ratio statistic.

Further, examination of Table 3 reveals that theTab

le 2

Cor

rela

tion

Mat

rice

s of

Mod

el V

aria

ble

s

Var

iabl

es1

23

45

67

89

1011

1213

14

1.Pa

rent

: att

enti

onal

con

trol

T1

.28*

.38*

**.2

7*.1

62

.30*

*2

.22

.68*

**.3

2**

.27*

.26*

.12

2.1

82

.22

2.T

each

er: a

tten

tion

al c

ontr

ol T

1.3

0*.4

2***

.45*

**.2

4*2

.21

2.4

9***

.37*

**.5

4***

.40*

**.3

9***

.18

2.2

7*2

.32*

*3.

Pare

nt: e

go c

ontr

ol T

1.2

4*.1

1.4

9***

.29*

2.3

7***

2.3

5**

.36*

*.4

8***

.77*

**.4

7***

.22

2.2

9*2

.35*

*4.

Tea

cher

: ego

con

trol

T1

.11

.39*

**.4

0***

.28*

2.2

22

.54*

**.3

4**

.45*

**.4

6***

.55*

**.2

12

.27*

2.3

3**

5.B

ehav

ior

regu

lati

on b

ox T

1.0

3.0

4.1

1.1

62

.13

2.2

0.2

1.2

7*.2

8*.2

7*.4

6***

2.1

62

.20

6.Pa

rent

: pro

blem

beh

avio

r T

12

.30*

2.1

22

.25*

2.1

62

.09

.27*

2.1

82

.24*

2.2

22

.21

2.1

0.4

4***

.17

7.T

each

er: p

robl

em b

ehav

ior

T1

2.1

52

.46*

**2

.21

2.6

3***

2.0

5.2

02

.28*

2.3

6**

2.3

3**

2.3

2**

2.1

5.2

1.4

8***

8.Pa

rent

: att

enti

onal

con

trol

T2

.38*

*.2

8*.0

7.1

0.0

32

.08

2.1

4*.5

7***

.51*

**.4

1***

.19

2.4

9***

2.3

6**

9.T

each

er: a

tten

tion

al c

ontr

ol T

2.2

9*.3

9***

.11

.15

.04

2.1

12

.20

.39*

**.5

6***

.58*

**.2

5*2

.40*

**2

.61*

**10

.Pa

rent

: ego

con

trol

T2

.07

.09

.58*

**.3

2**

.09

2.0

92

.17

.34*

*.1

7.6

6***

.31*

*2

.51*

**2

.45*

**11

.T

each

er: e

go c

ontr

ol T

2.1

3.1

7.4

4***

.45*

**.1

82

.18

2.3

3**

.23

.46*

**.5

0***

.30*

*2

.36*

*2

.59*

**12

.B

ehav

iora

l reg

ulat

ion

box

T2

2.0

12

.01

2.0

22

.03

.40*

**.0

1.0

22

.01

2.0

22

.03

2.0

52

.17

2.2

013

.Pa

rent

: pro

blem

beh

avio

r T

22

.03

2.0

42

.10

2.1

42

.04

.51*

**.0

82

.28*

2.0

82

.21*

2.2

3.0

1.2

8*14

.T

each

er: p

robl

em b

ehav

ior

T2

2.0

82

.11

2.2

5*2

.35*

*2

.10

.10

.32*

*2

.14

2.5

6***

2.2

9*2

.63*

**.0

3.1

3

Not

e:C

orre

lati

ons

abov

e th

e d

iago

nal a

re f

or c

hild

ren

high

in n

egat

ive

emot

iona

lity;

thos

e be

low

the

dia

gona

l are

for

chi

ldre

n lo

w in

neg

ativ

e em

otio

nalit

y. N

s fo

r ch

ildre

n hi

gh a

ndlo

w in

neg

ativ

e em

otio

nalit

y w

ere

74 a

nd 7

2, r

espe

ctiv

ely.

*p ,

.05;

**p

, .0

1; *

**p

, .0

01.

Eisenberg et al. 1375

third model in Figure 2 (in which the path from be-havioral regulation to problem behaviors was con-strained across groups and time and the path from at-tentional control to problem behaviors was allowedto vary) was the best fitting model according to theAkaike Information Criteria (AIC). The AIC (Akaike,1987) evaluates the likelihood function and adds apenalty that depends on the number of parameters inthe model with the lowest value representing themost preferable model. Specifically, it evaluates the“true” likelihood distribution to the expected likeli-hood distribution. Further, it provides a deviation ofthe expected distribution from the “true” distribu-tion. Thus, it appears that the difference between chil-dren high and low in negative emotionality occurs inthe paths from attentional control to problem behav-iors rather than in the paths from behavioral regula-tion to problem behaviors.

Finally, to test the alternative hypothesis that prob-lem behaviors have a causal effect on attentional con-

trol and behavioral regulation, we recomputed theoptimal model in Figure 2 with one change: We re-versed the direction of the paths between problembehavior and attentional and behavioral regulation.The model fit was considerably poorer than it was forthe original model in Figure 2, x2(126, N 5 146) 5200.973, p , .001, CFI 5 .911, AIC 5 251.973. Thus,the direction of effects did not appear to be from prob-lem behavior to regulation.

Because sex was related to numerous variables, weexamined sex as a moderator of the model. The sam-ple size was too small to examine two moderators(sex and negative emotionality) simultaneously, sonegative emotionality was not included in this model.First a model was computed in which the paths wereconstrained to be equal from behavioral regulationand attentional control to problem behavior acrossboys and girls (some correlations between constructsor loadings of measures on latent constructs were al-lowed to vary if the model was improved by doing so).

Figure 1 The relations of attentional control and behavioral regulation to problem behavior: Unconstrained model.

1376 Child Development

This model fit moderately well, x2(131, N 5 146) 5181.422, p , .002, CFI 5 .949, AIC 5 280.51844. Bothbehavioral and attentional regulation were nega-tively related to problem behavior at both T1 and T2.According to the Lagrange Multiplier Statistics, themodel was not improved by allowing the paths fromattentional or behavioral regulation to problem be-havior to vary across sex. Thus, the relations betweenregulation and problem behavior did not differ acrossthe sexes and sex did not appear to be a proxy for neg-ative emotionality in regard to moderating the relationbetween regulation and problem behavior. We alsotried to construct models in which we tested whetherego resiliency, as assessed with Block and Block’s(1980) Q-sort items, mediated the relation between at-tentional or behavioral regulation and problem behav-ior (as was found for social competence in Eisenberg etal., 1997). We could not obtain a reasonable fit in themodel with resiliency as a mediator.

Finally, as in Eisenberg et al. (1996) for the T1 data,we also examined whether there was a quadratic re-lation between behavioral regulation and externaliz-ing behavior. Regressions were computed in whichthe main effect of behavioral regulation was enteredon the first step, followed by the multiplicative termfor the quadratic effect on the second step of the equa-tion. Separate regressions were computed for paren-tal and teacher data. At T2, the quadratic term forteachers’, but not mothers’, reports of both behavioraland attentional control were nearly significant or sig-nificant, F(1, 143) 5 3.52 and 5.80, ps , .063 and .018,as were the linear effects, Fs(1, 144) 5 124.05 and104.41, ps , .001. Behavioral and attentional regula-tion were linearly, inversely related to externalizingproblem behavior. In addition, when the data for thequadratic were mapped using points using the proce-dures of Aiken and West (1991), behavioral or atten-tional regulation generally was associated with a

Figure 2 The relations of attentional control and behavioral regulation to problem behavior: Constrained model testing moder-ation from attentional control to problem behavior.

Eisenberg et al. 1377

drop in problem behavior. However, the drop inproblem behavior between high and average regula-tion was greater than that between average and lowregulation (plot points for low, average, and high reg-ulation were 2.22, 1.69, and 1.31 for behavioral regu-lation and 2.19, 1.66, and 1.34 for attentional control).This pattern of findings was virtually identical whenthe effect of age was controlled on the first step of theregressions. There were no significant interactions be-tween sex and the quadratic terms.

Exploratory Analyses

Because the best fitting model indicated that therewas no moderation of the effects of behavioral regula-tion by negative emotionality, we did some exploratoryanalyses to see if attentional and behavioral regulationrelated differently to different types of externalizing be-havior for individuals low in negative emotionality.

Our measure of externalizing behavior was not suit-able for examining relations with proactive versus re-active externalizing problems because many of theitems (e.g., starts fights with other children, aggres-sive to adults) could reflect either type of response.The measure did, however, include some of the itemsdesignated by Loeber and Schmaling (1985) as overt(e.g., argues, stubborn, yells at others, talks back/sasses) and covert (e.g., sneaky, takes things that be-long to others, lies) externalizing problems. When wesplit the sample on the median into high and low neg-ative emotionality and examined the correlations ofcomposites of covert versus overt externalizing prob-lems with both attentional and behavioral regulation(in low and high negative emotionality groups), therewere not obvious differences in patterns for the twotypes of regulation (in either emotionality group).Thus, although our items and procedures were notideal for testing relations between regulation and

Figure 3 The relations of attentional control and behavioral regulation to problem behavior: Constrained model testing moder-ation from behavioral regulation to problem behavior.

1378 Child Development

overt versus covert externalizing behavior, there waslittle evidence of the overt/covert distinction explain-ing the pattern of findings.

Given that there was some suggestion that nega-tive emotionality moderated the relation of behav-ioral regulation to problem behavior in some models(albeit not the best fitting), it is possible that there wasa moderation effect for behavioral regulation but thatit was just somewhat weak and thus dissipated whenthe effects pertaining to attentional control were con-trolled in the model. To explore this possibility, wedid auxiliary regression analyses for behavioral regu-lation at T2, the assessment at which the pattern ofmoderation seemed to be the reverse of that hypothe-sized in the initial unconstrained model (see Figure1). Parents’ and teachers’ reports of behavioral regu-lation were standardized and averaged, as were theirreports of negative emotionality. Then parent-reported problem behavior was predicted from theinteraction of regulation and negative emotionality inone regression, whereas teacher-reported problembehavior was the criterion variable in a second regres-sion. In addition, two analogous regressions werecomputed to test the interaction of negative emotion-ality with the measure of persistence on the box puz-zle task (an index of regulation). All children withdata were used in each regression analysis.

After the main effects were entered, the interactionterm was significant for two regressions predictingteacher-reported problem behavior, Fs for R2 change(3, 156 and 3, 144) 5 4.83 and 5.14, ps , .029 and .024,R2 change 5 .02 and .02, for the analyses involvingadult-reported and behavioral regulation, respec-tively. The interactions were mapped by using theprocedures of Aiken and West (1991). For adult-reported behavioral regulation, regulation predictedlower problem behavior for children low, average,and high in negative emotionality, but the relationwas much stronger for the children average or high innegative emotionality, slopes 5 2.21, 2.30, and 2.39,ts(156) 5 22.93, 26.00, and 26.64, ps , .01, .001, and

.001, respectively. When persistence was used as theindex of regulation, the negative relation betweenregulation and problem behavior was significant onlyfor children average and high in negative emotionality,slopes 5 2.38 and 2.70, ts(144) 5 22.31 and 23.38,ps , .05 and .01, respectively, and not for children lowin negative emotionality, slope 5 2.05, t(144) 5 2.22.Thus, for teachers’ reports of problem behavior, mod-eration was obtained in the predicted direction suchthat the negative relation between behavioral regula-tion and problem behavior was stronger for childrenprone to negative emotionality. A quite similar pat-tern of findings was obtained at T1 (although one ofthe two interactions for teachers’ reports of problem be-havior was even stronger than at T2). Thus, it appearsthat the moderational relation was not significant in thebest SEM model either as a result of controlling for othereffects or because moderation was apparent for onlyteachers’, and not parents’, reports of problem behavior.

DISCUSSION

There were several interesting findings in the struc-tural equation modeling. First, in the best fittingmodel, the path from attentional control to problembehavior was moderated such that at both T1 and T2,attentional regulation predicted externalizing behav-ior but did so primarily for children prone to frequentand intense negative emotionality (the path for lownegative emotionality children also was marginally sig-nificant at T1). In contrast, behavioral regulation wasimportant for children both high and low in negativeemotionality, although regression analyses suggestedthat even this relation was moderated for teachers’ re-ports of problem behavior. Similar findings were ob-tained at both assessments.

The fact that the relation of attentional control toproblem behavior was more significant for high neg-ative emotionality than low emotionality childrenprobably was not the result solely of differences invariability in the high and low negative emotionalitygroups. There were not differences in variability in at-tentional control at T1 or T2 between the two groups.Teachers’ (but not mothers’) reports of problem be-havior were more variable for the high than low neg-ative emotionality group. On the basis of differencesin the mean values of problem behavior (see Table 1),what seemed to be happening was that children lowin negative emotionality simply were low in external-izing problem behaviors, so individual differences inattentional control did not predict externalizing behav-ior for this group of children. Children high in negativeemotionality, however, exhibited greater levels of prob-lem behavior, and there was more variability among

Table 3 Akaike Information Criteria (AIC) for All ReportedModels

Model AIC

Fully constrained model 286.15482Fully released model (Figure 1) 289.88399Model releasing paths from attentional control

to problem behaviors (Figure 2) 291.32945Model releasing paths from behavioral regulation

to problem behaviors (Figure 3) 286.45855

Note: Lower values indicate a better fit.

Eisenberg et al. 1379

these children in the degree of externalizing prob-lems. For these children, those who were prone to ex-ternalizing problems lacked attentional control. Be-cause attentional regulation likely plays a role in theregulation of the internal experience of emotion, it ap-pears that a deficit in emotional regulation accountedfor the difference in externalizing behavior betweenchildren prone versus not prone to intense and fre-quent negative emotions.

As noted previously, the best fitting model indi-cated that there was not significant moderation of thepath from behavioral regulation to problem behavior.Moreover, although in one nonoptimal model (in Fig-ure 1) it appeared that there may be a somewhatstronger relation between behavioral regulation andproblem behavior for children low in negative emo-tionality, the zero-order correlations tended to behigher between behavioral regulation and problembehavior for the high negative emotionality group(see Table 2). Moreover, behavioral regulation wasnot more predictive for low negative emotionalitychildren in either of the more parsimonious models(Figures 2 and 3) and regression analyses demonstratedthat there was the predicted moderational effect for be-havioral regulation albeit only for teachers’ reports ofproblem behaviors. Controlling for the relation be-tween attentional control and behavior problems mayhave had a slight suppression effect on the relationbetween behavioral regulation and problem behav-iors in the model in Figure 1. In any case, the bestmodel was the one in which the path from behavioralregulation to problem behaviors was not moderated.Thus, behavioral regulation appeared to be an impor-tant predictor of externalizing problem behavior forchildren both low and high in negative emotionality(although it is possible that the relation actually isweaker for low emotionality children).

It is likely that negative emotionality did not mod-erate the effects of behavioral regulation in the bestfitting model because there were moderational effectsfor teachers’, but not parents’, reports of problem be-havior or because of the effects of controlling for rela-tions involving attentional regulation. The lack ofmoderation in the model did not appear to be due tobehavioral regulation, but not attentional regulation,predicting covert types of externalizing behavior forchildren low in negative emotionality. We could nottest the possibility that behavioral regulation, in com-parison with attentional regulation, was more consis-tently related to organized, unemotional, and instru-mental types of externalizing behavior (e.g., proactiveaggression; Dodge et al., 1997). Thus, it is possiblethat behavioral regulation is an important predictorof externalizing problems for children who are not

prone to negative emotion because of its role in sup-pressing proactive externalizing problem behaviors.Because the relation of behavioral regulation to exter-nalizing behavior and its potential moderation bynegative emotionality has been examined in only onesample (albeit at two time periods), it will be impor-tant to replicate the pattern of findings and to exam-ine whether behavioral regulation is a better predic-tor of proactive, unemotional modes of externalizingbehavior for children low in negative emotionality.All we can say is that the moderational effect was lessconsistent and clear for behavioral than attentionalregulation.

It is important to note that adding a path fromproblem behavior at T1 to problem behavior at T2 didnot improve the model and that this path was not sig-nificant. This indicates that the effects over time werethe result of dispositional regulation (sometimesmoderated by negative emotionality) rather thanthrough consistency of problem behavior across time.Thus, temperamental/personality differences in chil-dren seem to play an integral role in the prediction ofproblem behaviors over time. Moreover, reversingthe paths from problem behavior to regulation resultsin a model with a much poorer fit; thus, the data arenot consistent with the notion that children who getinto trouble become less regulated.

The results of this study support a growing body ofresearch that documents the importance of childhoodindividual differences in emotionality and regulation inchildren’s and adults’ problem behavior (e.g., Caspi etal., 1995; Caspi, Moffitt, Newman, & Silva, 1996; Chas-sin, Pillow, Curran, Molina, & Barrera, 1993; Eisenberg,Fabes, et al., 1996; Rothbart et al., 1994). The data alsosupport the utility of differentiating between behav-ioral and attentional regulation; our model indicatesthat they have unique (additive) effects, especially forchildren prone to negative emotions. In addition, thefindings support the view that Temperament 3 Tem-perament (or Personality 3 Personality) interactionsare important predictors of problem behavior. Thedata in this study provide the kind of evidence forTrait 3 Trait interactions that Rothbart and Bates(1998) predicted would be forthcoming. Moreover,consistent with Rothbart and Bates’ (1998) call forreplication of interaction effects, the results of thisstudy provide a within-sample longitudinal replica-tion of the interaction effect. Moreover, the findingsare consistent with recent findings with adolescents inwhich emotionality (sometimes positive emotionality;Colder & Chassin, 1997) and impulsivity interacted inpredicting substance abuse (often considered an exter-nalizing problem; Colder & Chassin, 1997; Pandina etal., 1992) or delinquency (Colder & Stice, 1998). Thus,

1380 Child Development

it appears that prediction of quality of children’s so-cial functioning may be enhanced by attention to themultiplicative effects of various components of tem-perament or personality.

There are several limitations to this study. First, thesample was mostly White, although the proportion ofminority children was not trivial. Unfortunately, themajority of the Black children as well as some of theleast regulated children were lost over the 2-year timeperiod. Thus, because of attrition, the sample at thefollow-up was not as diverse in regard to tempera-ment (or socioeconomic status) as it had initiallybeen. In addition, because this sample was drawnfrom the public schools, the rate of externalizing be-havior was moderate; thus, the findings in this studymay or may not generalize to high-risk groups inwhich externalizing behavior is very common, evenin elementary school. Thus, replication of these find-ings with a more diverse sample, as well as with othermeasures of predictors and outcomes, is desirable.

Additionally, the fact that measures of emotional-ity and regulation taken from temperament scalespredicted problem behavior provides little insightinto the origins of the externalizing problems. Tem-perament, including temperamental emotionality orregulation, likely has both biological and experientialbases (Rothbart & Bates, 1998). There does seem to bea substantial genetic predisposition to both negativeemotionality (Plomin & Stocker, 1989) and externaliz-ing behaviors such as aggression (Miles & Carey,1997), although estimates of heritability vary with themeasure of aggression (Miles & Carey, 1997) and find-ings in regard to heritability of emotionality varygreatly across adoption and twin studies (Plomin,Corley, Caspi, Fulker, & DeFries, 1998). Nonetheless,environmental factors such as familial interactionshave been implicated in the development of external-izing behavior both in studies of heritability in child-hood (Miles & Carey, 1997) and in studies of interven-tion programs with parents (Patterson, Reid, &Dishion, 1992). Moreover, there is evidence that tem-perament interacts with environmental influences inpredicting externalizing problem behaviors (Rothbart& Bates, 1998); thus, it is likely that environmentalfactors that were not assessed in this study contrib-uted to the observed relations. For example, Bates,Pettit, Dodge, and Ridge (1998) found that tempera-mental resistance to control in infancy and toddler-hood predicted externalizing behavior at ages 7 to 10more strongly for children with mothers low in re-strictive control. Findings such as these support theassertion that the interaction of biological and envi-ronmental factors best predicts the development ofexternalizing problems.

ACKNOWLEDGMENTS

This research was supported by grants from the Na-tional Science Foundation (DBS-9208375) and the Na-tional Institutes of Mental Health (1 R01 HH55052and 1 R01 MH60838) to the first two authors and Re-search Scientist Development and Research ScientistAwards from the National Institute of Mental Health(K02 MH00903 and K05 M801321) to Nancy Eisen-berg. The authors thank the many students who as-sisted in this study, the parents and children in-volved, and the principals and teachers in the Tempe,Kyrene, Mesa, and Scottsdale School Districts.

ADDRESSES AND AFFILIATIONS

Corresponding author: Nancy Eisenberg, Psychol-ogy, Arizona State University, Tempe, AZ 85287-1104;e-mail: [email protected]; Ivanna K. Guthrie,Richard A. Fabes, Sandra Losoya, Rick Poulin, and MarkReiser are also at Arizona State University; StephanieShepard and Bridget Murphy were at Arizona State Uni-versity at the time the study was conducted.

REFERENCES

Aiken, L. S., & West, S. G. (1991). Multiple regression: Testingand interpreting interactions. Newbury Park, CA: Sage.

Akaike, H. (1987). Factor analysis and AIC. Psychometrika,52, 317–332.

Baron, R. M., & Kenny, D. A. (1986). The moderator-media-tor variable distinction in social psychological research:Conceptual, strategic, and statistical considerations.Journal of Personality and Social Psychology, 51, 1173–1182.

Bates, J. E., Pettit, G. S., Dodge, K. A., & Ridge, B. (1998). In-teraction of temperamental resistance to control and re-strictive parenting in the development of externalizingbehavior. Developmental Psychology, 34, 982–995.

Bentler, P. M. (1989). EQS: Structural equations program man-ual. Los Angeles: BMDP Statistical Software.

Block, J., & Block, J. H. (1969). The California Child Q-Set.Berkeley, CA: Department of Psychology, University ofCalifornia at Berkeley.

Block, J. H., & Block, J. (1980). The role of ego-control andego-resiliency in the organization of behavior. In W. A.Collins (Ed.), The Minnesota Symposia on Child Psychology:Vol. 13. Development of cognition, affect, and social relations.(pp. 39–101). Hillsdale, NJ: Erlbaum.

Block, J., Block, J. H., & Keyes, S. (1988). Longitudinallyforetelling drug usage in adolescence: Early childhoodpersonality and environmental precursors. Child Devel-opment, 59, 336–355.

Bollen, K. A. (1989). Structural equations with latent variables.New York: Wiley.

Caspi, A. (1998). Personality development across the lifecourse. In N. Eisenberg (Ed.), W. Damon (Series Ed.),

Eisenberg et al. 1381

Handbook of child psychology: Vol. 3. Social, emotional, andpersonality development (pp. 311–388). New York: Wiley.

Caspi, A., Block, J., Block, J. H., Klopp, B., Lynam, D., Moffitt,T. E., & Stouthamer-Loeber, M. (1992). A “common-lan-guage” version of the California Child Q-Set for personal-ity assessment. Psychological Assessment, 4, 512–523.

Caspi, A., Henry, B., McGee, R. O., Moffitt, T. E., & Silva, P. A.(1995). Temperamental origins of child and adolescentbehavior problems: From age 3 to age 15. Child Develop-ment, 66, 55–68.

Caspi, A., Moffitt, T. E., Newman, D. L., & Silva, P. A. (1996).Behavioral observations at age 3 years predict adult psy-chiatric disorders. Archives of General Psychiatry, 53,1033–1039.

Chassin, L., Pillow, D. R., Curran, P. J., Molina, B. S. G., &Barrera, M., Jr. (1993). Relations of parental alcoholism toearly adolescent substance use: A test of three mediatingmechanisms. Journal of Abnormal Psychology, 102, 3–19.

Clark, L. A., Watson, D., & Mineka, S. (1994). Temperament,personality, and the mood and anxiety disorders. Journalof Abnormal Psychology, 103, 103–116.

Colder, C. R., & Chassin, L. (1997). Affectivity and impulsiv-ity: Temperament risk for adolescent alcohol involve-ment. Psychology of Addictive Behaviors, 11, 83–97.

Colder, C. R., & Stice, E. (1998). The moderating effect of im-pulsivity on the relationship between anger and adoles-cent problem behavior: Cross-sectional and prospectivefindings. Journal of Youth and Adolescence, 27, 255–274.

Derryberry, D., & Rothbart, M. K. (1988). Arousal, affect,and attention as components of temperament. Journal ofPersonality and Social Psychology, 55, 958–966.

Dodge, K. A., Lochman, J. E., Harnish, J. D., Bates, J. E, &Pettit, G. S. (1997). Reactive and proactive aggression inschool children and psychiatrically impaired chronicallyassaultive youth. Journal of Abnormal Psychology, 106, 37–51.

Eisenberg, N., & Fabes, R. A. (1992). Emotion, regulation,and the development of social competence. In M. S.Clark (Ed.), Review of personality and social psychology: Vol.14. Emotion and social behavior (pp. 119–150). NewburyPark, CA: Sage.

Eisenberg, N., Fabes, R. A., Bernzweig, J., Karbon, M., Pou-lin, R., & Hanish, L. (1993). The relations of emotionalityand regulation to preschoolers’ social skills and socio-metric status. Child Development, 64, 1418–1438.

Eisenberg, N., Fabes, R. A., Guthrie, I. K., Murphy, B. C.,Maszk, P., Holmgren, R., & Suh, K. (1996). The relationsof regulation and emotionality to problem behavior inelementary school children. Development and Psychopa-thology, 8, 141–162.

Eisenberg, N., Fabes, R. A., Murphy, B., Karbon, M., Smith,M., & Maszk, P. (1996). The relations of children’s dispo-sitional empathy-related responding to their emotional-ity, regulation, and social functioning. DevelopmentalPsychology, 32, 195–209.

Eisenberg, N., Fabes, R. A., Murphy, B., Maszk, P., Smith,M., & Karbon, M. (1995). The role of emotionality andregulation in children’s social functioning: A longitudi-nal study. Child Development, 66, 1360–1384.

Eisenberg, N., Fabes, R. A., Shepard, S. A., Murphy, B. C.,Guthrie, I. K., Jones, S., Friedman, J., Poulin, R., &Maszk, P. (1997). Contemporaneous and longitudinalprediction of children’s social functioning from regula-tion and emotionality. Child Development, 68, 642–664.

Eisenberg, N., Fabes, R. A., Shepard, S. A., Murphy, B. C.,Jones, J., & Guthrie, I. K. (1998). Contemporaneous andlongitudinal prediction of children’s sympathy from dis-positional regulation and emotionality. DevelopmentalPsychology, 34, 910–924.

Eisenberg, N., Guthrie, I. K., Fabes, R. A., Reiser, M., Mur-phy, B. C., Holmgren, R., Maszk, P., & Losoya, S. (1997).The relations of regulation and emotionality to resiliencyand competent social functioning in elementary schoolchildren. Child Development, 68, 367–383.

Hart, D., Hofmann, V., Edelstein, W., & Keller, M. (1997).The relation of childhood personality types to adolescentbehavior and development: A longitudinal study of Ice-landic children. Developmental Psychology, 33, 195–205.

Hart, D., Keller, M., Edelstein, W., & Hofmann, V. (1998).Childhood personality influences on social-cognitive de-velopment: A longitudinal study. Journal of Personalityand Social Psychology, 74, 1278–1289.

Hussong, A. M., & Chassin, L. (1994). The stress-negativeaffect model of adolescent alcohol use: Disaggregatingnegative affect. Journal of Studies in Alcohol, 55, 707–718.

Kochanska, G., Murray, K., & Coy, K. C. (1997). Inhibitorycontrol as a contributor to conscience in childhood: Fromtoddler to early school age. Child Development, 68, 263–277.

Krueger, R. F., Caspi, A., Moffitt, T. E., White, J., &Stouthamer-Loeber, M. (1996). Delay of gratification,psychopathology, and personality: Is low self-controlspecific to externalizing problems? Journal of Personality,64, 107–129.

Larsen, R. J., & Diener, E. (1987). Affect intensity as an indi-vidual difference characteristic: A review. Journal of Re-search in Personality, 21, 1–39.

LISREL 7 (2nd ed.). (1988). Chicago: SPSS.Lochman, J. E., & Conduct Problems Prevention Research

Group (1995). Screening of child behavior problems forprevention programs at school entry. Journal of Consult-ing and Clinical Psychology, 63, 549–559.

Loeber, R., & Schmaling, K. B. (1985). Empirical evidencefor overt and covert patterns of antisocial conduct prob-lems: A metaanalysis. Journal of Abnormal Child Psychol-ogy, 13, 337–352.

Miles, D. R., & Carey, G. (1997). Genetic and environmentalarchitecture of human aggression. Journal of Personalityand Social Psychology, 72, 207–217.

Moffitt, T. E. (1993). The neuropsychology of conduct disor-der. Development and Psychopathology, 5, 135–151.

Newman, D. L., Caspi, A., Moffitt, T. E., & Silva, P. A. (1997).Antecedents of adult interpersonal functioning: Effectsof individual differences in age 3 temperament. Develop-mental Psychology, 33, 206–217.

O’Brien, B. S., & Frick, P. J. (1996). Reward dominance: As-sociations with anxiety, conduct problems, and psychop-athy in children. Journal of Abnormal Child Psychology, 24,223–239.

1382 Child Development

Oosterlaan, J., & Sergeant, J. A. (1996). Inhibition in ADHD,aggressive, and anxious children: A biologically basedmodel of child psychopathology. Journal of AbnormalChild Psychology, 24, 19–36.

Pandina, R. J., Johnson, V., & LaBouvie, E. W. (1992). Affec-tivity: A central mechanism in the development of drugdependence. In M. Glantz & R. Pickens (Eds.), Vulnera-bility to drug abuse (pp. 179–209). Washington, DC:American Psychological Association.

Patterson, C. M., & Newman, J. P. (1993). Reflectivity andlearning from aversive events: Toward a psychologicalmechanism for the syndromes of disinhibition. Psycho-logical Review, 100, 716–736.

Patterson, G. R., Reid, J. B., & Dishion, T. J. (1992). A sociallearning approach: Vol. 4. Antisocial boys. Eugene, OR:Castalia Press.

Plomin, R., Corley, R., Caspi, A., Fulker, D. W., & DeFries, J.(1998). Adoption results for self-reported personality:Evidence for nonadditive genetic effects? Journal of Per-sonality and Social Psychology, 75, 211–218.

Plomin, R., & Stocker, C. (1989). Behavioral genetics andemotionality. In J. S. Reznick (Ed.), Perspectives on behav-ioral inhibition (pp. 219–240). Chicago: University of Chi-cago Press.

Pulkkinen, L., & Hamalainen, M. (1995). Low self-control as aprecursor to crime and accidents in a Finnish longitudinalstudy. Criminal Behaviour and Mental Health, 5, 424–438.

Robins, R. W., John, O. P., Caspi, A., Moffitt, T. E., &Stouthamer-Loeber, M. (1996). Resilient, overcontrolled,and undercontrolled boys: Three replicable personalitytypes. Journal of Personality and Social Psychology, 70, 157–171.

Rothbart, M. K., Ahadi, S. A., & Hershey, K. L. (1994). Tem-perament and social behavior in childhood. Merrill-Palmer Quarterly, 40, 21–39.

Rothbart, M. K., & Bates, J. E. (1998). Temperament. N.Eisenberg (Ed.), W. Damon (Series Ed.), Handbook of childpsychology: Vol. 3. Social, emotional, and personality develop-ment (pp. 105–176). New York: Wiley.

Rothbart, M. K., Posner, M. I., & Hershey, K. L. (1995). Tem-perament, attention and developmental psychopathology.In D. Cicchetti & D. Cohen (Eds.), Manual of developmentalpsychopathology (Vol. 1, pp. 315–340). New York: Wiley.

Rothbart, M. K., Ziaie, H., & O’Boyle, C. G. (1992). Self-regulation and emotion in infancy. New Directions inChild Development, 55, 7–23.

Sanson, A., Prior, M., & Kyrios, M. (1990). Contamination ofmeasures in temperament research. Merrill-Palmer Quar-terly, 36, 179–192.

Stice, E., & Gonzales, N. (1998). Adolescent temperamentmoderates the relation of parenting to antisocial behav-ior and substance use. Journal of Adolescent Research, 13,5–31.

Teglasi, H., & MacMahon, B. H. (1990). Temperament andcommon problem behaviors of children. Journal of Ap-plied Developmental Psychology, 11, 331–349.

Weinberger, D. A., & Schwartz, G. E. (1990). Distress and re-straint as superordinate dimensions of self-reported ad-justment: A typological perspective. Journal of Personal-ity, 58, 381–417.

Ungerer, J. A., Dolby, R., Waters, B., Barnett, B., Kelk, N., &Lewin, V. (1990). The early development of empathy:Self-regulation and individual differences in the firstyear. Motivation and Emotion, 14, 93–106.

This document is a scanned copy of a printed document. No warranty is given about the accuracy of the copy.

Users should refer to the original published version of the material.


Recommended