Parents' Use of Inductive Discipline: Relationsto Children's Empathy and Prosocial Behavior

Julia KrevansCuyahoga Community College

John C. GibbsOhio State University

KKEVANS, JULIA, and GIBBS, JOHN C. Parents' Use of Inductive Discipline: Relations to Children'sEmpathy and Prosocial Behavior. CHILD DEVELOPMENT, 1996,67,3263-3277. Relations betweenparents' discipline, children's empathic responses, and children's prosocial behavior were exam-ined in order to evaluate Martin Hoffman's claim that children's empathy and empathy-basedguilt mediate the socialization of children's prosocial behavior. 78 sixth and seventh graders(138-172 months in age), their mothers, and teachers completed multiple measures of Hoffman'sconstructs. Results were largely consistent with theory. Parents' use of inductive as opposed topower-assertive discipline was related to children's prosocial behavior. Children of inductiveparents were more empathic; and more empathic children were more prosocial. Moreover, chil-dren's empathy was found to mediate the relation between parents' discipline and children'sprosocial behavior. Few relations were obtained for children's guilt indices, but post hoc analysesyielded theoretically consistent results. Contrary to expectations, parents' use of statements ofdisappointment was the component of the inductive discipline score which was most stronglyrelated to children's prosocial behavior.

Martin Hoffman's (1970, 1982, 1983, such as coercion or threats of punishment1984) theory of the effects of parental disci- are posited to promote self-focused concernspline on children's prosocial behavior has with external consequences, which can inbeen influential for a quarter of a century, turn reduce prosocial behavior.Hoflinan's theory takes as established thatinductive discipline is linked to prosocial The present study evaluated the the-behavior (and more generally, moral inter- ory's central tenet that empathy plays a me-nalization) and accounts for this relation by diating role in the relation between parentalpositing that empathy plays a key role. Spe- discipline and children's prosocial behavior,cifically, because other-oriented inductions Despite a renewal of interest in the media-direct the child to consider how his behavior tors of socialization (e.g., Feldman & Wein-has affected others, their use in the disci- berger, 1994; Kochanska, 1993) and contin-pline encounter is seen as eliciting and culti- ued interest in the determinants andvating empathy and empathy-based guilt, consequences of children's empathy (Eisen-These emotions can in turn motivate proso- berg & Strayer, 1987), studies have not di-cial behavior in subsequent social situations, rectiy evaluated this hypothesis. ComponentAlternative disciplinary styles, love with- relations (discipline-prosocial behavior,drav^al and power assertion, are not posi- discipline-empathy, empathy-prosocial be-tively associated with children's prosocial havior) of Hoffman's hypothesis, however,behavior because they do not elicit chil- have been examined and critiqued in bothdren's empathy. Indeed, power assertions methodological and substantive terms.

This article was based on a doctoral dissertation submitted by the first author to Ohio StateUniversity. Portions of this research were presented at the 1991 biennial meeting ofthe Societyfor Research in Child Development in Seattle. The authors wish to thank teachers, parents, andstudents for their participation as well as the following school administrators for their assistance:James Fodor, Chris Townley, Lynn Artl, Wayne Greene, Jerry Brodsky, Hex Morgan, GordenGerber, James Cipolletti, Sister Lois Peterson, Josephine Dodge, and Richard Koenig. We alsogratefully acknowledge the assistance of Mark Barnett, Jim Burke, Brenda Bryant, Eric Dlugokin-ski, Philip Clark, Nancy Eisenberg, Stephanie Gibbs, Martin Hoffman, Dinah Meyer, JanetStrayer, Gerald Winer, and Carolyn Zahn-Waxler. Requests for reprints should be sent to JuliaKrevans, Cuyahoga Community College, Western Campus, 11000 Pleasant Valley Road ParmaOH 44130.

[Child Development, 1996,67,3263-3277. © 1996 by the Society for Research in Child Development, Inc.AU rights reserved. 0009-3920/96/6706-0031$01.00]

3264 Child Development

Inductive Discipline—Prosocial BehaviorThe use of predominantly inductive dis-

cipline and/or the avoidance of predomi-nantly power assertive discipline have gen-erally been found to be at least moderatelyrelated to children's prosocial behavior (e.g.,Dlugokinski & Firestone, 1974; Hoffman &Saltzstein, 1967; Zahn-Waxler, Radke-Yarrow, & King, 1979), although the ro-bustness of the relation is a matter of somecontroversy. Some reviewers (e.g., Maccoby& Martin, 1983) have been impressed by thecorroboration of findings across varyingmethodologies, whereas others (Brody &Shaffer, 1982; Radke-Yarrow, Zahn-Waxler,& Chapman, 1983) have noted weaknessesand some nonsignificance among the results.Radke-Yarrow et al. (1983) argued that thecontext provided by more general featuresof the parent-child relationship might mod-ify the effect of parental discipline on chil-dren's prosocial behavior. More recently,Grusec and Goodnow (1994) have arguedthat any effects of specific types of parentaldiscipline are variable across a variety ofcontextual factors.

Studies of disciplinary style—prosocialbehavior relations have been hamperedby methodological limitations. Disciplinarystyle and other key constructs have oftenbeen assessed with measures that are ofquestionable reliability and validity (e.g.,Bar-Tal, Nadler, & Blechman, 1980; Hoff-man & Saltzstein, 1967), attributable to theuse of few items and the failure to aggregateacross multiple measures of constructs (seeGook & Goldstein, 1993; Schwartz, Barton-Henry, & Pruzinsky, 1985). These measure-ment limitations might account for at leastsome of the weaknesses observed amongprevious results (see also Brody & Shaffer,1982). However, there are little relevantdata.

Measures of other-oriented inductionhave had particular limitations. Hoffman andSaltzstein's (1967) measure of parental disci-pline specified child misbehavior only interms of parent-child confiict (backtalk, non-compliance, etc.). Ironically, this type ofmisbehavior is least likely to elicit inductivediscipline from parents (Trickett & Kuczyn-ski, 1986). Moreover, insofar as the parentis the distressed—and displeased—other towhom the child's attention is directed by in-ductions used in parent-child confiict, thedistinctiveness of other-oriented inductionfrom love withdrawal is compromised (seealso Dlugokinski & Firestone, 1974; Mac-coby & Martin, 1983; Shoffiett, 1971). Hoff-man (1970) suggests that love withdrawal is

at best weakly related to chidren's mortJ in-temalization. He also links love withdrawalto children's psychoanalytic guilt (anxiety inresponse to unacceptable impulses), whichhe suggests is inhibitory and should there-fore lead to less prosocial behavior. Thus tothe extent that measures of parental induc-tion (such as Hoffnian and Saltzstein's mea-sure) emphasize parent-child confiict, onewould expect correlations with children'sprosocial behavior to be attenuated.

Another limitation is the inclusion (e.g.,Dlugokinski & Firestone, 1974; Hoffman &Saltzstein, 1967) of statements of parentaldisappointment (e.g., "I never would haveexpected you to do that") among exemplarsof other-oriented induction. Parental state-ments of disappointment do meet Hoffman's(1970) crucial theoretical criteria for other-oriented induction: since parents are at leastsecondary victims of the child's misbehav-ior, statements of disappointment orient thechild toward his victims and could be arguedto elicit empathy (for the parents) in thechild. Nonetheless, the literature identifiesdifferences between parental statements ofdisappointment and other disciplinary state-ments classified as other-oriented induction.These might well have consequences for thechild. In an early paper, Hoffman (1963) sug-gested that statements of disappointmentwere distinctive in that they made parents'positive expectations for the child salient.Subsequent commentary (e.g., Shoffiett,1971) focused on the salience of parental dis-approval and suggested that parental state-ments of disappointment should be classi-fied as love withdrawals.

Parental Discipline—EmpathySupport also has been found for the

parental discipline—empathy relation, al-though again with some methodological lim-itations. Hoffman's (1982) theory identifiestwo empathy dimensions as critical featuresof the empathy construct: (a) empathic re-sponsiveness, that is, the frequency and gen-erality of empathic responses; and (b) matu-rity of empathy, that is, the sophistication ofthe cognitions which inform the child's emo-tional response to others. Ghildren's sympa-thy, a relatively mature form of empathy,was related to parents' use of other-orientedsocialization practices across disciplinaryand nondisciplinary contexts, although pa-rental discipline was not assessed separately(Eisenberg et al., 1992; Miller, Eisenberg,Fabes, Shell, & Gular, 1989). White, Walsh,and Gibbs (1988) found parents' use of in-ductive discipline to be related to children'sempathic responsiveness, but the finding

Krevans and Gibbs 3265

could have been attributable to sharedmethod variance (discipline and empathywere both measured through child self-report).

Empathy—Prosocial BehaviorThe literature on the empathy-

prosocial behavior relation provides consis-tent findings. Ghildren's empathic respon-siveness and maturity have been found tobe related to their prosocial orientations(Eisenberg & Miller, 1987; Strayer &Schroeder, 1989). Of particular relevance toHoffman's theory are the few studies whichhave included controls for extrinsic incen-tives. As a theory of moral internalization,Hoffman's theory is most specifically con-cerned with relations between empathy andprosocial behavior in the absence of extrin-sic incentives (i.e., altruism). Although re-searchers obviously cannot control for allpossible extrinsic incentives, they (e.g., Len-non, Eisenberg, & Garroll, 1986) have iden-tified contexts in which the presence of ex-trinsic incentives is minimized (e.g., privatevs. public donations) and have found thatempathy—prosocial behavior relations areenhanced in these contexts.

GuiltAccording to Hoffman, not only empa-

thy, but also empathy-based guilt (whereinthe child not only empathizes with the vic-tim's distress but is aware of his responsibil-ity for it) mediates the relation between pa-rental discipline and children's prosocialbehavior. Empathy-based guilt is distin-guished from psychoanalytic guilt which, asnoted above, Hoffman (1970) describes asanxiety based and inhibitory. Ghildren'sguilt has been found to relate both to par-ents' use of induction (Eisikovits & Sagi,1982; Hoffman & Saltzstein, 1967) and tochildren's helpfulness (Ghapman, Zahn-Waxler, Gooperman, & Iannotti, 1987), butno evidence was provided that empathy-based guilt was specifically measured. Withfew exceptions (e.g., Zahn-Waxler, Kochan-ska, Krupnick, & McKnew, 1990), this dis-tinction with respect to the guilt constructhas not been made in the empirical liter-ature.

Summary and HypothesesIn contrast to previous studies, the pres-

ent study examined all ofthe basic variablesand relations of Hoffman's theory in orderto evaluate the theory's empathy mediationhypothesis. In doing so we hoped to shednew light on how parents teach children tocare for others as well as to advance the as-sessment of Hoffman's theory as an account

ofthe process. To remedy previous method-ological limitations, we used methods andmeasures developed in recent socializationand empathy research. The study's partici-pants were sixth and seventh graders (to en-sure that they had the ability to completeassessments), their mothers, and their teach-ers. Multiple measures of all main variableswere obtained. Ghildren provided severalmeasures of empathy. Ghildren and teachersprovided measures of children's prosocialbehavior which were designed to minimizethe salience of extrinsic incentives. The chil-dren and mothers both provided data on ma-ternal and patemal disciplinary practices.We supplemented parent-child conflict situ-ations on the Hoffman and Saltzstein dis-cipline measure with discipline situationsinvolving transgressions against peers ornonparent adults since the latter are morelikely to elicit parental reasoning (Trickett& Kuczynski, 1986) and thus other-orientedinductions. Additionally, we computed a setof relatively modified or pure other-orientedinduction scores by excluding parental state-ments of disappointment.

From Hoffman's theory, we derived thefollowing hypotheses: (a) that parents' useof inductive as opposed to power assertivediscipline would be positively related tochildren's prosocial behavior; (fo) that par-ents' use of inductive as opposed to powerassertive discipline would also be positivelyrelated to children's empathy, defined interms of empathic responsiveness and matu-rity of empathy; and (c) that the relation be-tween parental discipline practices and chil-dren's prosocial behavior would be reducedwhen the effects of children's empathy werecontrolled. The empathy-mediation hypoth-esis was also studied in terms of the relatedemotion of empathy-based guilt. Further-more, measures of parental discipline wereexamined with the expectation that strongerrelations would emerge when aggregatedmeasures were used. The modified other-oriented induction measure was also ex-pected to yield relatively strong findingssince parental statements of disappointment,possibly a form of love withdrawal, wouldbe excluded.

MethodSubjects

A fiyer which offered families $12 forparticipating in a psychological study wasdistributed at 10 Northeastern Ohio juniorhigh and middle schools. Eighty-four per-cent of families who expressed interest inthe study agreed to participate after learning

3266 Child Development

about the nature of the study. One familylearned the study's hypothesis prior to par-ticipation and was therefore eliminated. Theremaining 78 families were not atypical withrespect to parents' discipline practices orchildren's empathy. Means and variances ofseveral parental discipline and empathy in-dices were compared to statistics from theentire seventh grade of one school district(White et al., 1988), and no differences werefound across samples.

The 34 boy and 44 girl participants werebetween 138 and 172 months in age, with amean of 147.5 months (12 years, 3.5 months).While families ranged from lower to upperclass, as assessed with the Revised Index ofOccupational Status (Stevens & Feath-erman, 1981), most were middle class (75%of the occupational status scores fell be-tween 31 and 51). Most families includedboth biological parents. However, 15% weresingle parent families, and 10% included thechild's biological mother and her new part-ner. Five families did not include a femalecaregiver who had known the child sincebirth. Because children from these familieshad discontinuous experiences and becausethere were not enough of these families toform a subgroup, these families were elimi-nated from analyses which related the par-enting indices to child measures.

The children nominated teachers to ratetheir social behavior. The 57 child-selectedteachers were probably positively "biased.However, since the bias was at least consis-tent for all students rated, it seems likely thatthe effect of teacher bias on analyses wasminimized (see also Gibbs et al., 1986).

ProceduresTwo assessment sessions were held at

the children's school. During the first ses-sion, the parent completed consent forms,then the parenting questionnaires, and fi-nally a demographic fact sheet. In a separateroom, the child completed all assessmentswhich required special equipment or indi-vidual administration. During a second ses-sion, the child completed paper-and-pencilmeasures and participated in a behavioralassessment of prosocial tendencies. Withinthe limits set by practical concerns, childmeasures were administered in randomizedorder.

Teachers received packets of ratingforms after all participants at their schoolhad been tested. Each teacher completedseveral rating measures for from one to fivestudents with the order of rating measures

varied randomly across teachers. Teachersreturned rating forms in a sealed envelopeaddressed to the experimenter and were as-sured that their ratings would be kept con-fidential.

MeasuresParental discipline.—Mothers and chil-

dren completed a modification of Hoffmanand Saltzstein's (1967) parental disciplinequestionnaire. Our modification presentedrespondents with six descriptions of childmisbehavior: two were Hoffman andSaltzstein's (1967) original parent-child con-fiict vignettes and four (adapted from Dlu-gokinski & Firestone, 1974; Hoffman &Saltzstein, 1967; White et al., 1988) de-scribed antisocial behavior directed againstpeers or other adults. Mothers describedhow they would respond if their child (theparticipating child) engaged in each childmisbehavior, first with an open-ended replyand then by using lists of 10-14 disciplineexemplars which followed each misbehaviordescription. The lists of exemplars included(a) other-oriented inductions^ that is, disci-pline which directs the child to attend tohis or her victims' perspectives (e.g., "Pointout how his friend must feel," and "Tellhim I never expected to hear that sort ofthing from him"); (b) power assertions, thatis, discipline which attempts to change thechild's behavior through use of the parents'power over the child (e.g., "Tell him thathe'll be punished for what he's done"); and(c) love withdrawals, that is, disciplinewhich withholds parental approval or atten-tion from the child (e.g., "Ignore him for awhile").

Mothers selected the three disciplineexemplars which they used most frequentlyand ranked them with respect to frequencyof use. Mothers also selected and ranked thediscipline exemplars with respect to theiruse by the child's other caregiver (typicallythe child's father, but sometimes a grand-mother or the mother's friend). The child-report version of the questionnaire did notrequire open-ended descriptions and wasworded from the child's perspective but wasin other respects identical to the mother-report version.

Several types of scores were derivedfrom the mothers' and children's rankings ofdiscipline exemplars. Six-item scores werecomputed by averaging rankings for eachtype of discipline across the six child misbe-haviors. Scores indexed mothers' use ofpower assertion, other-oriented induction.

Krevans and Gibbs 3267

and love withdrawal; and the second care-giver's use of power assertion, other-oriented induction, and love withdrawalboth as reported by mothers and as reportedby children. Twelve-item scores were cre-ated by aggregating across parents (mothervs. other caregiver) to yield three mother-reported discipline scores (power assertion,induction, and love withdrawal) and threechild-reported parental discipline scores.Net scores (Eisikovits & Sagi, 1982) werecreated by subtracting the 12-item power as-sertion scores from the 12-item inductionscores. Net scores indicated the extent towhich parents used inductive as opposed topower-assertive discipline. The use of netscores in the present study was justified bystrong negative correlations between powerassertion and inductive discipline scores,r(70) = - .62, p < .01, for mother-reportscores; r(74) = - .68 , p < .01, for child-report scores. Finally, a set of modified six-and 12-item inductive discipline scoreswere computed: The initial set of other-oriented induction scores (see above) werecomputed in accordance with Hoffman andSaltzstein's procedures. When computingthe modified other-oriented inductionscores, exemplars which expressed peirentaldisappointment (i.e., communications thatthe parent is unfavorably surprised and/orthat more is expected ofthe child, e.g., "Tellhim I never expected to hear that sort ofthing from him") were not scored as exem-plars of other-oriented induction.

The internal consistencies of the six-item scales were fairly low, particularly formothers' reports. The corrected split-half co-efficients ranged from .68 to .40 for scoresbased on mothers' reports and from .81 to.47 for scores based on children's reports.When scores were aggregated across parentsto create 12-item scores, the split-half coef-ficients increased. For example, whereas thereliabilities of six-item child-reported powerassertion scores were .78 and .80, the reli-ability of the 12-item aggregate child-reported power assertion score was .85. Reli-abilities of the 12-item scores ranged firom.85 to .65, with the exception of reliabilitiesfor love withdrawal scores and mother-reported modified induction scores whichwere lower (.47 to .56, and .55, respectively).The reliability coefficients for the 24-itemnet scores were .79 for the mother-report netscore and .89 for the child-report net score.

Data relevant to evaluating the ecologi-cal validity of the discipline questionnairewere obtained by comparing mothers' initial

open-ended responses to each misbehaviorvignette with their questionnaire responses.The open-ended responses were coded forthe occurrence ofthe different types of disci-pline (e.g., power assertion) by an indepen-dent coder. Intercoder reliability was above.84 for all codes except love withdrawal forwhich it was .50, perhaps due to its low fre-quency. We computed the proportions ofthemothers' distinct disciplinary responses thatwere power assertions, love withdrawals,etc. Proportion scores based on mothers'open-ended responses were all positivelyand significantly correlated with the corre-sponding closed-ended questionnairescores, with rs ranging from .28 to .37 (IV =70—72). There were no positive correlationsbetween mothers' proportion scores andnoncorresponding closed-ended scores.

Nurturance.—Mothers also completedHof&nan and Saltzstein's (1967) 17-item self-report measure of parental nurturance(Gronbach's alpha = .65).

Prosocial behavior.—Five measureswere selected with the goal of assessing al-truism as opposed to hedonistically moti-vated prosocial behavior.

measure, shortened version ofHartshorne, May, and Mailer's (1929) por-trait measure, provided teachers with fourverbal portraits of children which variedwith respect to degree of altruistic orienta-tion. The most altruistic portrait described achild who would help regardless ofthe sacri-fice involved, as a result of a sincere desire"to promote the happiness and welfare ofeveryone." The least aJtruistic portrait de-scribed a child who "would expect to berewarded for helping." Hartshorne et al.(1929) reported that five teachers matchedportraits to 129 children with a test-retest re-liability of .84. They also reported positivecorrelations between the portrait measureand several other measures of children's pro-social behavior.

The second measure used teachers' rat-ings of children's helpfulness as the basis forthe assessment. Evidence (Barnett &Thompson, 1985; Eisenberg, Gameron, Pas-ternack, & Tyron, 1988) suggests that, whenteachers assess children's willingness tohelp others, they are infiuenced by the in-trinsic or extrinsic nature of children's moti-vation and not just how often children help.The Barnett and Thompson (1985) prosocialbehavior rating scale described eight op-portunities to help a peer. Teachers usedfive-point rating scales to rate the likeli-

3268 Child Development

hood that the child would help in each situa-tion. The Gronbach's alpha coefficient was.95.

The third and fourth measures weretwo single item scales adapted from Bar-Talet al. (1980). The originals were peer nomi-nation scales which asked students to list theclassmates who would be most likely to (a)share a sandwich with a classmate who for-got to bring lunch and (b) help a sick class-mate with homework. Bar-Tal et al. (1980)reported that these behaviors were good in-dicators of a prosocial orientation. In thepresent study, teachers rated the likelihoodthat children would perform each of the twoprosocial acts along five-point scales.

The fifth measure of prosocial behaviorwas a behavioral assessment (Dlugokinski &Firestone, 1974). Each child was promiseda bonus of $1 and received 10 dimes duringthe second session. At the end ofthe secondsession, the child listened to a story about achild from a disadvantaged country (whichserved as the stimulus for an empathy as-sessment, as will be described below). Thechild was then given an opportunity to do-nate bonus money to UNIGEF, a charitywhich helps children who, like the one inthe story, live in disadvantaged countries. Inorder to reduce extrinsic motives for help-ing, an illusion of anonymity was created.Ghildren were left alone to make their deci-sion and were asked to put a sealed donationenvelope in a collection bag whether or notthey actually made a contribution. Severaldonation envelopes were kept in the collec-tion bag in order to enhance the credibilityof the helping opportunity. The size of thechild's donation served as the index of pro-social behavior.

Empathy.—We measured both maturityof empathy and empathic responsiveness.Two measures assessed the maturity of em-pathy variable.

1. One maturity of empathy measurewas a self-reported sympathy measureadapted from Eisenberg and her colleagues(Eisenberg, Fabes, et al., 1988; Eisenberg,Schaller, et al., 1988). According to Hoffman(1984), sixth and seventh graders are mostlikely to vary in their capacity for the mostmature form of empathy, sympathy for a per-son's life situation. Thus, a stimulus storywas designed to assess sympathy for a per-son's life situation. The story, adapted fromEisenberg, Schaller, et al. (1988), describeda child who w as happy, yet clearly faced adifficult life situation as a result of the pov-

erty of her community. After listening to thestory, respondents described their feelingson 12 rating scales which were organizedinto four factor scores (Eisenberg, Schaller,etal., 1988). Ghildren's scores on the sympa-thetic concern factor served as the matureempathy index. Since the immediate contextofthe story was happy, research participantswho scored high in sympathetic concerndemonstrated empathic sensitivity to the im-plications of the less immediate aspects ofthe story character's life situation. A fillerstory and filler questions were also pre-sented in order to minimize demand effects.

2. The second maturity of empathymeasure was the Empathy Gontinuum Sys-tem (Strayer, 1989). Five 2—4-min stimulusfilm clips and two filler film clips wereshown. The stimulus film clips portrayed thefear of children approaching an old house,anger between two children, the sadness ofa child after a punishment, the happiness ofa child at the circus, and the challenging lifeof a handicapped child. A semi-structuredinterview was used to elicit children's viewsabout (a) the film characters' emotions, (b)their own emotions in response to the films,and (c) the reason for their emotional re-sponse. Ghildren's answers were recorded,transcribed, and scored. Match scores indi-cated the degree of similarity between thefilm character's emotions and the child'semotional response to the film. When therewas at least a minimal degree of match, alevel score was assigned on the basis ofthenature of the thinking which informed thechild's empathy. The lowest level score wasgiven to responses which did not include anawareness ofthe film character who had elic-ited the emotional response (e.g., "I feltscared because it was a scary night"). Higherscores were given when children wereaware ofthe character's situation, the charac-ter's emotions and—at the highest level—the character's perspective (e.g., "I was sorryfor her, because I would feel terrible if Iwere younger and my dad did that to me").Match scores and level scores were com-bined (see Strayer, 1989) to form the Empa-thy Gontinuum Score which ranged from 0to 19. Interrater agreement was 97% for thematch scores and 84% for the level scores(N = 25).

3. Bryant's (1982) Index of Empathy forGhildren and Adolescents served as a mea-sure of empathic responsiveness. Bryant(1982) reported evidence consistent with thevalidity ofthe scale. However, the scale wascriticized (Eisenberg & Strayer, 1987) for the

Krevans and Gibbs 3269

heterogeneity of its items. Four had moreto do with attitudes toward the display ofemotions than with empathy (e.g., "Peoplewho hug and kiss in public are silly") andwere therefore eliminated in the presentstudy. Gronbach's alpha was .74.

Guilt.—Two measures of guilt wereused. Ghapman et al.'s (1987) guilt measureconsisted of stories, with accompanying pho-tographs, in which a child observed othersin distress (e.g., the protagonist aggressivelyseeks a turn on a friend's bicycle and thatfriend gets hurt). The Hoffman andSaltzstein (1967; Hoffman, n.d.) measurepresented longer stories about a child whohad harmed another (e.g., an older childdoes not help a lost little boy and later learnsthat the boy was struck by a car.) Six of theeight Ghapman et al. stories and one of thetwo Hoffman and Saltzstein stories served asguilt measures in the present study. We alsopresented eight stories which focused onself-esteem and achievement themes in or-der to reduce demand effects. Research par-ticipants were asked to infer the centralcharacter's thoughts and feelings in re-sponse to all stories and were asked to pro-vide endings to the Hoffman and Saltzstein(1967) story and three filler stories. TheGhapman et al. (1987) guilt score was thenumber of times guilt was attributed to acentral character (interrater reliability: r(22)= .94). Hoffman and Saltzstein's (1967) max-imal guilt score was the level ofthe strongestexpression of guilt. The terminal guilt scorewas the level of guilt expressed at the endof the response. The original seven-pointglobal scale was replaced with a four-pointscale and detailed scoring criteria in orderto achieve acceptable levels of interrateragreement (76% for maximal guilt and 72%for terminal guilt, N = 25). The total guiltscore was the sum of maximal and terminalguilt scores (Hoffman & Saltzstein, 1967).

ResultsRelations between Measures of the SameVariable

Gorrelations between different mea-sures of the same variable were examinedfor additional evidence of convergent valid-ity. In addition, significantly correlated mea-sures were aggregated in order to maximizethe reliability and validity of indices for eachvariable (Rushton, Brainerd, & Pressley,1983; Schwartz et al., 1985) and to reducethe number of indices included in the re-gression analyses which w ere used to testthe mediation hypotheses.

Gorrelations between mothers' and chil-dren's reports of parental discipline prac-tices were significant, r(63) = .37, p < .01,for the net score, r(63) = .31, p < .01, forthe inductive discipline score, and r(63) =.38, p < .01, for the power assertive disci-pline score. Thus total scores which re-fiected both mothers' and children's per-spectives on parental discipline werecomputed by standardizing scores and thenaveraging across the two types of respon-dents. The standardized alpha coefficientsfor the three scores were .87 for the total netscore, .75 for the total inductive disciplinescore, and .80 for the total power assertivediscipline score. Love withdrawal scoreswere not aggregated because the correlationbetween mother- and child-report versionsofthe love withdrawal score was not signifi-cant, r(63) = .19, p < .10.

All correlations between pairs of pro-social behavior measures were significant.Gorrelation coefficients ranged from .69 {df= 72, p < .01) to .23 {df = 64, p < .05) andaveraged .46. Thus prosocial behavior scoreswere also standardized and aggregated. Thestandardized alpha coefficient for the totalprosocial behavior scor^ was .85 {N = 66).

Gorrelations among the three measuresof empathy were fairly low but significant,rs = .42-.23, df = 75-73, p < .05. The threemeasures of empathy were standardized andaggregated following the argument detailedin Rushton et al. (1983) that the aggregationof even weakly correlated measures of thesame variable improves validity. The stan-dardized alpha was .61.

Correlations between Parental Disciplineand Children's Prosocial Behaviorand Empathy

Parental discipline indices were sig-nificantly related to parents' education,child's age, and child's sex. However, onlychild's sex was also significantly related todependent variables, specifically, children'sprosocial behavior and children's empathy(girls evidenced greater prosocial behaviorand empathy, r(73) = .28, p < .05, and r(75)= .28, p < .05, respectively). Thus controlsfor sex differences were included in analy-ses which related parental discipline prac-tices to children's prosocial behavior andchildren's empathy. The partial correlationsare presented in Table 1.

The total parental discipline net scorewas positively related to children's total pro-social behavior score, r(67) = .34, p < .01.Thus, parents who used inductive as op-

3270 Child Development

TABLE 1

CORKELATIONS BETWEEN PARENTAL DISCIPLINE SCORES AND CHILD MEASURES

CHILDREN'S PROSOCIAL AFFECT AND BEHAVIOR

PARENTAL DISCIPLINE

Total net scoreM-R* net scoreC-R"* net scoreTotal power assertion ....M-R love withdrawal ....C-R love withdrawalTotal inductionM-R modified inductionC-R modified inductionTotal disappointment ....M-R disappointmentC-R disappointmentM-R nurturance

TotalEmpathy"

.41**

.44**

.30**-.33**- .15

.02

.41**

.32**

.25*

.33**

.27*

.26*

.37**

Hoffman andSaltzstein

Guilt''

- .07-.09

.02

.12-.13

.10- .01

.00

.05-.06

.03-.08

.16

Chapman et al.Guilt''

.14

.22*

.00-.10

.02

.15

.01

.13

.07

.06

.19-.12- .01

TotalProsocialBehavior"

.34**

.38**

.18- .31**-.10-.06

.31**

.15

.18

.34**

.38**

.12

.15

NOTE.—iV = 63-73.•Correlations are partial correlations with sex controlled.•"Zero-order correlations since sex is unrelated to children's guilt.''M-R refers to mothers' report."•C-R refers to child's report.*p < .05, one-tailed.**p < .01, one-tailed.

posed to power-assertive discipline had chil-dren who were more prosocial. Parents' useof inductive discipline, as indexed by thetotal induction score, was also positively re-lated to their children's prosocial behavior,r(67) = .31, p < .01; and parents' use ofpower-assertive discipline was negativelyrelated to their children's prosocial behav-ior, r(67) = —.31, p < .01. Parents' use oflove withdrawal was not related to chil-dren's prosocial behavior nor to any of theother child variables (see Table 1). Thus, re-sults of the correlation of love withdrawalwith child variables will not be discussedfurther.

The partial correlation between the totalparental discipline net score and children'stotal empathy was significant, r(69) = .41, p< .01. Thus, parents who used inductive asopposed to power-assertive discipline hadchildren who were relatively empathic. Thecorrelation was still obtained when the pa-rental discipline scores and children's empa-thy scores were each derived from differentinformants; mother-report based disciplinenet scores were significantly related to chil-dren's empathy, r(63) = .44, p < .01. Rela-tions between parental discipline measuresand the individual measures of children'sempathy were consistent with the mainfindings. Empathy Continuum scores, self-

reported sympathy scores, and Bryant's em-pathic responsiveness scores were each pos-itively and significantly related to totalparental discipline net scores, r(69) = .27, p= .01, r(68) = .34, p < .01, and r(68) = .33,p < .01, respectively. Further, as with themain findings, the individual empathyscores were positively related to parentaldiscipline net scores which were basedsolely on mothers' reports, rs(65—73) =.44-.25, p < .05.

Relations between Children's Empathyand Prosocial Behavior

A third set of partial correlations wasused to detennine whether children whowere prone to experience empathy weremore likely to behave prosocially. Again,children's sex was controlled. Children's to-tal empathy scores were positively relatedto their total prosocial behavior scores, r(71)= .40, p < .01, Partial correlations betweenindividual measures of empathy and chil-dren's prosocial behavior were all also posi-tive and significant, rs(74-72) = .37-.23, p< .05.

The Mediation HypothesesThe hypothesis that children's empathy

would mediate the relation between paren-tal discipline practices and children's proso-cial behavior was tested with a series of re-

Krevans and Gibbs 3271

TABLE 2

REGRESSION ANALYSES: TEST OF THE MEDIATION HYPOTHESIS

Regression Equation Beta t p

Predicting children's empathy from parental discipline:Total empathy = Parental discipline+ Child's sex H- Constant.

Parental discipline Beta 407 3.66 <.O1Predicting children's prosocial behavior from parental discipline and children'sempathy:

Total prosocial behavior = Parentaldiscipline" + Child's sex + Constant

Parental discipline Beta 305 2.57 <.O5Total prosocial behavior score =Total empathy + Parental discipline" +Child's sex + Constant

Total empathy Beta 367 2.96 <.O1Parental discipline Beta 155 1.26 n.s.

NOTE.—N = 69."The total net score (inductions minus power assertions) was used.

gression analyses (see Table 2). Children'ssex served as a control variable in each anal-ysis. In the first analysis, the total parentaldiscipline net score was the independentvariable, and children's total empathy wasthe dependent variable. The parental disci-pline score predicted children's empathy,beta = .407, t{66) = 3.66, p < .01. Children'sprosocial behavior was the dependent vari-able in the second analysis. The parentaldiscipline net score also predicted chil-dren's total prosocial behavior scores, beta= .305, t{66) = 2.57, p < .05. The third anal-ysis included both total parental disciplinenet scores and children's total empathyscores as predictors of children's total proso-cial behavior scores. Consistent with the me-diation hypothesis (see Barron & Kenny,1986), the relation between parental disci-pline and children's prosocial behavior wasreduced when the effect of children's empa-thy was controlled. When both the total pa-rental discipline net score and the children'stotal empathy score were included as pre-dictors of children's prosocial behavior, thebeta weight associated with children's em-pathy differed significantly from zero, beta= .367, t{65) = 2.96, p < .01, but the betaweight associated with the parental disci-pline net score was reduced and was nolonger significant, beta = .155, i(65) = 1.26,N.S.

Correlates of Children's GuiltChildren's Hoffman and Saltzstein

(1967) guilt scores and their Chapman et al.(1987) guilt scores were each correlated withparental discipline scores, children's total

prosocial behavior scores, children's totalempathy scores, and children's scores on theindividual measures of empathy. SinceChapman et al. guilt scores were highlytruncated in range (most scores were either0 or 1), the scale was dichotomized (0 = ascore of 0, 1 = a score of 1 or greater) andpoint biserial correlations were computed.

Children's Chapman et al. guilt scoreswere positively related to children's proso-cial behavior, r(73) = .32, p < .01, and toparents' use of inductive as opposed topower-assertive discipline, as indexed bythe mother-report version of the parentaldiscipline net score, r(65) = .22, p < .05.However, correlations between children'sChapman et al. guilt scores and the otherparental discipline indices were not signifi-cant (see Table 1). Similarly, although chil-dren's Chapman et al. guilt scores were posi-tively correlated with children's scores onStrayer's (1989) Empathy Continuum, r(72)= .26, p < .05, the guilt scores were notrelated to children's scores on the other twoempathy measures nor to children's aggre-gate empathy scores. The Hoffman andSaltzstein (1967) measure had no correlates.

It was possible that findings consistentwith Hoffman's theory would have emergedhad the Chapman et al. (1987) measure al-lowed us to discriminate between children'sempathy-based guilt and other sorts of guiltexperienced by children. Post hoc analyseswere designed to circumvent this measure-ment limitation through the identification ofsubsamples of children for whom highChapman et al. scores were especially likely

3272 Child Development

to reflect empathy-based guilt and childrenwhose high guilt scores refiected other sortsof guilt. First, we examined relations be-tween guilt and prosocial behavior for highversus low empathy children. Variation inChapman et al. guilt scores in the context ofhigh empathy quite possibly reflected varia-tion in empathy-based guilt, whereas varia-tion in guilt scores in the context of lowempathy could not reflect variation in empa-thy-based guilt. Thus if it was empathy-based guilt which motivated children's pro-social behavior, one would expect acorrelation between children's guilt and pro-social behavior for empathic children only.The data were consistent with this expecta-tion. The effect of the interaction betweenchildren's Chapman et al. guilt scores andtheir total empathy scores on their prosocialbehavior was examined. The incremental R-squared associated with the interaction termwas significant, R^ = .045, F(l, 62) = 4.09,p < .05. The sample was then divided intothirds (Cohen & Cohen, 1983) on the basis ofchildren's total empathy scores. The partialcorrelation between children's guilt andtheir prosocial behavior was only significantfor children who were high in empathy,r(21) = .60, p < .01, r(22) = .06, N.S., andr(22) = .20, N.S., for high, moderate, andlow empathy groups, respectively.

Our second analysis examined the rela-tion between parental discipline and empa-thy-based guilt using a strategy suggested byZahn-Waxler et al. (1990). We divided chil-dren into three groups on the basis of thesort of parental discipline that they experi-enced. Hoffman's theory predicts that, forthose children who experience a predomi-nance of inductive parental discipline, highguilt scores will reflect the high levels ofempathy-based guilt promoted by other-oriented inductions. Thus one would expectguilt and empathy to covary among thesechildren. In contrast, when parents use rela-tively little other-oriented induction, chil-dren, if they experience guilt, should not ex-perience empathy-based guilt. Thus onewould expect little correlation between em-pathy and guilt for children of less inductiveparents. The data were consistent with theseexpectations. The incremental R-squared as-sociated with the effect ofthe interaction be-tween parental discipline net scores andchildren's total empathy scores on children'sChapman et al. guilt scores was signiflcant,fi2 = .060, F(l, 65) = 4.30, p < .05. Partialcorrelations between children's total empa-thy scores and Chapman et al. guilt scores

(with sex controlled) were significant onlywhen parents had high parental disciplinenet scores, that is, when parents used ahighly inductive approach to discipline,r(23) = .49, p < .01. Correlations were notsignificant for children whose parents ob-tained moderate or low net scores, that is,less inductive parents, r(21) = .03, N.S.,r(22) = .04., N.S., respectively.

Methodological Issues: Parental DisciplineIndices

Regarding the effects of aggregation,we found that the average partial correlationbetween six-item inductive discipline scoresand children's total prosocial behavior scorewas .24, and the average partial correlationbetween six-item power assertion scores andchildren's prosocial behavior was —.23,whereas the average partial correlation be-tween parental discipline aggregates whichwere based on 24 or more items and chil-dren's total prosocial behavior scores (whencomputed without regard to the direction ofthe correlation) was .31. Further, whereas63% of the partial correlations which in-volved six-item scores were signiflcant, 80%of the correlations which involved aggregatescores were significant.

We next examined the correlations be-tween the modified other-oriented inductionscores (parental expressions of disappoint-ment deleted) and child measures. Correla-tions between the modified other-orientedinduction score and children's empathy (seeTable 1) were no stronger than those ob-tained when the original other-oriented in-duction scores were used. Correlations withchildren's total prosocial behavior scoreswere relatively weak, r(61) = .15, N.S., forthe mother-report score and r(65) = .18, p <.10, for the child-report score. Since mother-and child-report scores were not signifi-cantly related, r(62) = .16, p < .10, an aggre-gate modified induction score was computedfor comparative purposes only. Not even theaggregate score was signiflcantly related tochildren's prosocial behavior, r(67) = .19, p< .10.

In order to investigate why parents'expressions of disappointment enhancedrather than reduced relations between pa-rental induction and child prosocial behav-ior, we computed disappointment scores.The scores were derived from mothers' andchildren's rankings of response optionswhich described the parent as expressingdisappointment with the child. The indexprovided a comparison for the modified

Krevans and Gibbs 3273

TABLE 3

CORRELATIONS BETWEEN PARENTING MEASURES

Parental Discipline 1 2 3 4 5 6 7

1. Net score -.88** - .19 .92** .80** .36** .34**2. Power assertion -.90** . . . - .14 -.62** -.52** -.28** - .25*3. Love withdrawal - .19* - .11 . . . .-.44** -.36** -.24* -.34*4. Induction 93** -.68** - .41** . . . .89** .33** .35*5. Modified induction 88** -.64** -.40** .95** . . . - .11 .27*6. Disappointment 32* -.24* - .10 .33** .02 . . . .25*7. Matemal nurturance" . . . . . . . . . . . . . . . . . .

NOTE.—IV = 70—76. Correlations above the diagonal relate scores based on mothers' reports. Correlations belowthe diagonal relate scores based on children's reports.

'Children did not report on matemal nurturance.*p< .05, one-tailed.**p < .01, one-tailed.

other-oriented induction index. The disap-pointment indices measured parents' use ofstatements of disappointment only; the mod-ified other-oriented induction indices mea-sured parents' use of more obviously other-oriented inductions only.

The scores were correlated with othermeasures of parenting as well as with childmeasures. Results (Table 3) indicated thatparents' tendency to express disappointmentwas negatively (and not positively) relatedto their use of love withdrawal, r(70) =— .24, p < .05, for mothers' reports and r(74)= —.10, N.S. for children's reports. In thisand other respects, disappointment scoresbehaved similarly to the modified other-oriented induction scores. For example,with reference to parent measures, both re-lated negatively to power assertion and posi-tively to parents' nurturance (Table 3). Fur-thermore, both related positively tochildren's empathy (Table 1). However, un-like the modifled other-oriented inductionscore (see above), disappointment scoreswere significantly correlated with children'sprosocial behavior, at least when assessedby mothers' reports, r(67) = .34, p < .01,r(63) = .38, p < .01, r(65) = .12, N.S., foraggregate, mother-report, and child-reportscores, respectively.

DiscussionThe findings of the present study were

largely consistent with Hoffman's socializa-tion theory and provide important new sup-port for it. The results of previous researchwere replicated and extended: (a) parentswho used predominantly inductive disci-pline as opposed to power assertion hadchildren who were relatively prosocial; {b}children's empathy predicted their prosocial

behavior; and (c) parents who relied on in-duction as opposed to power assertion hadchildren who were relatively empathic.Most important, the present study extendedsupport for Hoffman's theory through a testof his mediation hypothesis. Our regressionfindings were consistent with Hoffman'sclaim that an inductive as opposed to power-assertive approach to discipline promotesprosocial behavior in children because itpromotes the development of children's em-pathy. These results are likely to be replica-ble and not attributable to common methodvariance insofar as the variables were as-sessed with multiple data sources (e.g., pa-rental discipline scores based on mothers'reports were related to children's empathyscores based on children's reports).

The results of the present study did notinitially support theoretically derived pre-dictions concerning children's guilt. Ac-cording to Hoffman's theory, it is specificallychildren's empathy-based guilt which medi-ates the relation between parental disciplineand children's prosocial behavior. Thusmeasures that discriminate empathy-basedguilt are necessary to test Hoffman's theory.The results ofthe present study suggest thatneither Hoffman's own measure of chil-dren's guilt (Hoffman & Saltzstein, 1967),nor Chapman et al.'s (1987) measure of chil-dren's guilt are adequate measures of empa-thy-based guilt. Post hoc analysis did yieldfindings consistent with Hoffman's theory.When parents relied on inductive as op-posed to power-assertive discipline, theirchildren's scores on the Chapman etal. guiltmeasure were correlated with empathyscores. Further, it was only among childrenhigh in empathy, and therefore capable ofempathy-based guilt, that guilt predicted

3274 Child Development

children's prosocial behavior. This post hocsupport must be considered suggestive. Anumber of artifacts can complicate the inter-pretation of moderator effects such as these(Barron & Kenny, 1986). However, the guiltfindings, taken together, do point to the im-portance of Hoffman's distinction betweentypes of guilt and to the need to develop andvalidate measures of empathy-based guilt.

Improved measures of guilt would alsohelp clarify the effects of parents' use of lovewithdrawal. Because parental love with-drawal plays a minor and ambiguous role inHoffman's theory, we made no speciflc pre-dictions about its relation to child variables.Nonetheless, it is of interest that parentallove withdrawal was unrelated to our mea-sures of children's guilt. It seems likely thatthe guilt measures were not only poor mea-sures of empathy-based guilt but were alsopoor measures of psychoanalytic guilt (seeZahn-Waxler et al., 1990). Better guilt mea-sures are needed to investigate relations be-tween love withdrawal and children's psy-choanalytic guilt.

The present results also point to theneed for renewed attention to the assess-ment of parental discipline practices. Theinternal consistency of our six-item mea-sures—measures similar to those used inprevious research (e.g., Dlugokinski & Fire-stone, 1974; Hoffman & Saltzstein, 1967)—was moderate and even low in some cases.The reliability of measurement was im-proved through the aggregation of six-itemmeasures across parent (mother and othercaregiver) and respondent (mother andchild). The inclusion of fathers' perspectiveson parental discipline in addition to moth-ers' and children's perspectives might wellresult in further increases in reliability (seeSchwartz et al., 1985). In addition to improv-ing reliability, the aggregation of measuresincreased the magnitude and consistency ofcorrelations between parental discipline andchildren's prosocial behavior. Continuedimprovement of parental discipline mea-sures seems likely to lead to further in-creases in the robustness of socializationfindings.

Conversely, the lack of robustness ob-served in previous flndings would seem tobe at least partially attributable to the lim-ited reliability of parental discipline mea-sures and not exclusively to variation indiscipline effectiveness across context. Re-search has identifled child variables thatmoderate the effectiveness of discipline

practices (Kochanska, 1994). Co-occurringparental practices may also moderate disci-pline effectiveness. Of special interest isHoffman's (1983; cf Damon, 1995) ownclaim that effective induction presupposesthat the parent has made optimal use ofpower assertion and love withdrawal(enough to ensure that the child is attendingto and taking the parents' message seriously,but not so much as to engender interferingemotions). Other moderator variables havebeen suggested (e.g., Crusec & Coodnow,1994; Radke-Yarrow et al., 1983), and furtherresearch may demonstrate their influence ondiscipline effectiveness. However, it is haz-ardous to speculate about moderator effectson the basis of weaknesses in the parentaldiscipline literature inasmuch as method-ological limitations might well have attenu-ated findings (cf Brody & Shaffer, 1982).

Efforts to improve Hoffman andSaltzstein's (1967) measure of parental in-duction were less successful. Modified in-duction scores were computed in which par-ents' statements of disappointment were notscored as inductions; yet the improvedmatch between measure and theory led to aweaker rather than stronger correlation withchildren's prosocial behavior. This weakereffect may have been an artifact ofthe fewerresponse options for the modified inductionmeasure. Interestingly, even though themodification procedure reduced the reliabil-ity of the modified as compared to originalscores, correlations between the modifledother-oriented induction scores and chil-dren's prosocial behavior approached sig-nificance. However, it is important to notethat our index of parents' use of disappoint-ment, although not any more reliable thanthe modified induction index, was signifl-cantly, and substantially, correlated withchildren's prosocial behavior. In otherwords, parents' statements of disappoint-ment made a stronger contribution to chil-dren's prosocial behavior than did more ex-plicitly other-oriented inductions.

This flnding raises questions about thenature of parental disappointment and its re-lation to other types of psychological disci-pline. Our results were not consistent withan initial expectation (cf. Maccoby & Martin,1983; Shoffiett, 1971) that parents' state-ments of disappointment, as parent-orientedinductions, might be classifled as love with-drawals. Rather, disappointment is moresimilar to other-oriented induction andshould probably be considered a part of aninductive approach to discipline. The two

Krevans and Gibbs 3275

types of discipline correlate similarly withother parenting variables (e.g., negativelywith power assertion and positively withnurturance). Further, they are similar in thatboth seem to cultivate the child's empathy.Parents' statements of disappointment, likeother-oriented inductions, include refer-ences to another's hurt feelings and, also likeother-oriented inductions, are associatedwith relatively high levels of empathyamong children.

Nonetheless, the present results suggestthat parents' statements of disappointmentmerit distinct theoretical attention. State-ments of disappointment go beyond other-oriented induction to communicate the par-ents' confldence in the child's capacity forbetter behavior. The relatively strong corre-lation between parental disappointment andthe child's prosocial behavior could reflectthe effects of an additional mediational pro-cess. Speciflcaiiy, parental disappointmentmay not only cultivate the child's empathybut may also induce a prosocial self-concept.Consistent with this interpretation are re-search flndings (Barnett, 1987; Larrieu &Mussen, 1986) that link parental practicesthat enhance children's self-esteem to chil-dren's prosocial tendencies.

Along these lines, it is of historical inter-est that Hoffman (1963) initially identifiedparental statements of disappointment, notother-oriented inductions, as the type of psy-chological discipline which promotes a pro-social type of moral internalization. Accordingto Hoffman, statements of disappointmentcommunicated that the child was "capableof living up to an ideal" (1963, p. 311) andthus provided children with an internal ba-sis for good conduct—a positive ego ideal.He contrasted parental disappointment witha second type of psychological disciplinewhich he termed "ego-attacks." Hoffenanlater (1970) abandoned the distinction be-tween disappointment and ego attacks, in-cluded parents' statements of disappoint-ment with other-oriented inductions, andcontrasted other-oriented inductions withlove withdrawals as the key antecedent ofempathy-based moral internalization. Dur-ing the ensuing decades, reviewers (e.g.,Crusec & Coodnow, 1994; Maccoby & Mar-tin, 1983) have repeatedly called for morerefined distinctions between types of psy-chological discipline. The results of thepresent study suggest a refinement of psy-chological discipline into three categories;love withdrawal, other-oriented induction,and parental disappointment. This refine-

ment might well provide new insights intothe socialization process.

Finally, an important qualificationshould be noted: relations established in thepresent study might well refiect repeated re-ciprocal transactions between parent andchild. Child-effects interpretations are ofspecial relevance in the current contextsince Bell's (1968) introduction ofthe childeffects critique was directed in part at theresearch support for Hoffman's theory.There is no speciflc child-effects theorywhich provides an alternative to Hoffman'stheory as an interpretation of the presentflndings. Speculatively, however, it may bethat parents are more likely to use inductivediscipline in response to prosocial childrenand that it is this relation which is mediatedby the child's empathy. From this perspec-tive, the mediation results would have to beinterpreted as meaning that children's em-pathy elicits inductive discipline from par-ents and also promotes children's prosocialbehavior. Thus, the parents' disciplinechoice would not be a response to the child'sprosocial behavior per se, but to the child'sunderlying empathy. Although this directionof effect could be claimed to be the primaryone, both parent-to-child and child-to-parenteffects are likely to contribute to contempo-raneous correlations (Hoffman, 1975, 1994;Wahler, 1990).

In conclusion, the signiflcance of thestudy's results was twofold. First, inductivediscipline as historically measured wasfound to include distinct components of dis-cipline, other-oriented induction and disap-pointment, that have distinct consequencesfor children. Second, provisional but criticalsupport was found for Hoffman's empathy-mediation hypothesis. A flnding that the re-lation between parents' discipline practicesand children's prosocial behavior was un-changed when empathy indices were heldconstant would have discouraged further re-search stimulated by Hoffnian's theory. Itwould have meant, whatever the reasons forthe relation between parents' inductive styleof discipline and children's prosocial behav-ior, the relation could not be attributed toparents' promotion of empathy as Hoffmanclaims. Thus, the present study establisheda precondition for further attention to Hoff-man's theory of empathy as a key mediatorin moral internalization.

ReferencesBamett, M. A. (1987). Empathy and related re-

sponses in children. In N. Eisenberg & J.

3276 Child Development

Strayer (Eds.), Empathy and its development(pp. 146-163). New York: Cambridge Univer-sity Press.

Bamett, M. A., & Thompson, S. (1985). The role ofperspective-taking and empathy in children'sMachiavellianism, prosocial behavior andmotive for helping. Journal of Genetic Psy-chology, 146, 295-305.

Barron, R. M., & Kenny, D. A. (1986). The modera-tor-mediator variable distinction in social psy-chological research: Conceptual, strategic,and statistical considerations. Journal of Per-sonality and Social Psychology, 51, 1173-1182.

Bar-Tal, D., Nadler, A., & Blechman, N. (1980).The relationship between Israeli children'shelping behavior and their perception of par-ents' socialization practices./ourna/ of SocialPsychology, 111, 159-167.

Bell, R. Q. (1968). A reinterpretation ofthe direc-tion of effects in studies of socialization. Psy-chological Review, 75, 81-95.

Brody, C. H., & Shaffer, D. R. (1982). Contribu-tions of parents and peers to children's moralsocialization. Developmental Review, 3,31-75.

Bryant, B. K. (1982). An index of empathy for chil-dren and adolescents. Child Development,53, 413-425.

Chapman, M., Zahn-Waxler, C , Cooperman, G.,& Iannotti, R. (1987). Empathy and responsi-bility in the motivation of children's helping.Developmental Psychology, 23, 140-145.

Cohen, J., & Cohen, P. (1983). Applied multipleregression analysis for the behavioral sci-ences, second edition, Hillsdale, NJ; Erl-baum.

Cook, W. L., & Coldstein, M. J. (1993). Multipleperspectives on family relationships: A latentvariables model. Child Development, 64,1377-1388.

Damon, W. (1995). Greater expectations: Over-coming the culture of indulgence in Amer-ica's homes and schools. New York: FreePress.

Dlugokinski, E. L., & Firestone, I. J. (1974).Other-centeredness and susceptibility tocharitable appeals: Effects of perceived disci-pline. Developmental Psychology, 15,128-137.

Eisenberg, N., Cameron, E., Pasternack, J., &Tyron, K. (1988). Behavioral and sociocogni-tive correlates of ratings of prosocial behaviorand sociometric status. Journal of GeneticPsychology, 149, 5-15.

Eisenberg, N., Fabes, R. A., Bustamante, D., Ma-thy, R. M., Miller, P. A., & Lindholm, E.(1988). Differentiation of vicariously inducedemotional reactions in children. Develop-mental Psychology, 24, 237-246.

Eisenberg, N., Fabes, R. A., Carlo, G., Troyer, D.,Speer, A. L., Karbon, M., & Switzer, G. (1992).The relations of matemal practices and char-acteristics to children's vicarious emotionalresponsiveness. Child Development, 63,583-602.

Eisenberg, N., & Miller, P. A. (1987). The relationof empathy to prosocial and related behaviors.Psychological Bulletin, 101, 91-119.

Eisenberg, N., Schaller, M., Fabes, R. A., Busta-mante, D., Mathy, R., & Rhodes, K. (1988).The differentiation of personal distress andsympathy in children and adults. Develop-mental Psychology, 24, 766-775.

Eisenberg, N., & Strayer, J. (1987). Critical issuesin the study of empathy. In N. Eisenberg & J.Strayer (Eds.), Empathy and its development(pp. 3-13). New York: Cambridge UniversityPress.

Eisikovits, Z., & Sagi, A. (1982). Moral develop-ment and discipline encounters in delinquentand nondelinquent adolescents. Journal ofYouth and Adolescence, 11, 217-232.

Feldman, S. S., & Weinberger, D. A. (1994). Self-restraint as a mediator of family inHuences onboys' delinquent behavior. Child Develop-ment, 65, 195-211.

Gibbs, J. C , Clark, P. M., Joseph, J. A., Green,J. L., Goodrick, T. S., & Makowski, D. C.(1986). Relations between moral judgment,moral courage, and field independence. ChildDevelopment, 57, 185-193.

Grusec, J. E., & Goodnow, J. J. (1994). Impact ofparental discipline methods on the child's in-ternalization of values: A reconceptualizationof current points of view. Developmental Psy-chology, 30, 4-19.

Hartshorne, H., May, M. A., & Mailer, J. B. (1929).Studies in the nature of character: Vol, 2.Studies in service and self-control. New York:Macmillan.

Hoffman, M. L. (1963). Childrearing practices andmoral development: Generalizations from re-search. Child Development, 34, 295-318.

Hoffman, M. L. (1970). Moral development. InP. H. Mussen (Ed.), Carmichael's manual ofchild psychology (Vol. 2, pp. 457—557). NewYork: Wiley.

Hoffman, M. L. (1975). Moral internalization, par-ent power and the nature of parent-childinteraction. Developmental Psychology, 11,228-239.

Hoffman, M. L. (1982). Development of prosocialmotivation: Empathy and guilt. In N. Eisen-berg (Ed)., The development of prosocial be-havior (pp. 218-231). New York: AcademicPress.

Hoffman, M. L. (1983). Affective and cognitiveprocesses in moral internalization: An infor-mation processing approach. In E. T. Higgins,

D. Ruble, & W. Hartup (Eds.), Social cogni-tion and social development: A socioculturalperspective (pp. 236-274). New York: Cam-bridge University Press.

HofEman, M. L. (1984). Empathy, its limitationsand its role in a comprehensive moral theory.In W. M. Kurdnes & J. L. Gerwirtz (Eds.),Morality, moral behavior and moral develop-ment (pp. 283-302). New York: Wiley.

Hoffinan, M. L. (1994). Discipline and internaliza-tion. Developmental Psychology, 30, 26-28.

Hof&nan, M. L. (n.d.) Criteria for scoring guiltseverity. Unpublished manuscript. New YorkUniversity.

Hoffman, M. L., & Saltzstein, H. D. (1967). Parentdiscipline and the child's moral development.Journal of Personality and Social Psychology5, 45-57.

Kochanska, G. (1993). Toward a synthesis of pa-rental socialization and child temperament.Child Development, 64, 325-347.

Kochanska, G. (1994). Children's temperament,mothers' discipline, and security of attach-ment: Multiple pathways to emerging inter-nalization. Child Development, 66, 597-615.

Larrieu, J., & Mussen, P. (1986). Some personalityand motivational correlates of children's pro-social behavior. Journal of Genetic Psychol-ogy, 147, 529-543.

Lennon, R., Eisenberg, N., & Carroll, J. (1986).The relations between nonverbal indices ofempathy and preschoolers' prosocial behav-ior. Journal of Applied Developmental Psy-chology, 7, 219-224.

Maccoby, E. E., & Martin, J. A. (1983). Socializa-tion in the context ofthe family: Parent-childinteraction. In P. Mussen (Ed.), Handbook ofchild psychology (Vol. 4, pp. 1-101). NewYork: Wiley.

Miller, P. A., Eisenberg, N., Fabes, R. A., Shell,R., & Gular, S. (1989). Mothers' emotionalarousal as a moderator in the socialization ofchildren's empathy. In N. Eisenberg (Ed.),New directions for child development (No.44, pp. 65-83). San Francisco: Jossey-Bass.

Radke-Yarrow, M., Zahn-Waxler, C , & Chapman,M. (1983). Children's prosocial dispositionsand behavior. In P. H. Mussen (Ed.), Hand-book of child psychology (Vol. 4, pp. 469-546). New York: Wiley.

Krevans and Gibbs 3277

Rushton, J. P., Brainerd, C. J., & Pressley, M.(1983). Behavioral development and con-struct validity: The principal of aggregation.Psychological Bulletin, 94, 462-469.

Schwartz, J. G., Barton-Henry, M. L., & Pruzinsky,T. (1985). Assessing child-rearing behaviors:A comparison of ratings made by mother, fa-ther, child and sibling on the CRPBI. ChildDevelopment, 56, 462-469.

ShofEett, G. P. (1971). The moral development ofchildren as a function of parental moraljudgements and child-rearing practices. Un-published doctoral dissertation, George Pea-body Gollege for Teachers.

Stevens, G., & Featherman, D. L. (1981). A re-vised socioeconomic index of occupationalstatus. Social Science Research, 10, 364-395.

Strayer, J. (1989). What children know and feel inresponse to witnessing affective events. In G.Saarni & P. Harris (Eds.), Children's under-standing of emotions (pp. 259-289). NewYork: Gambridge University Press.

Strayer, J., & Schroeder, M. (1989). Ghildren'shelping strategies: Influences of emotion, em-pathy, and age. In N. Eisenberg (Ed.), Newdirections for child development (No. 44, pp.85-105). San Francisco: Jossey-Bass.

Trickett, P. K., & Kuczynski, L. (1986). Ghildren'smisbehaviors and parental discipline strate-gies in abusive and nonabusive families. De-velopmental Psychology, 22, 115-123.

Wahler, R. G. (1990). Who is driving the interac-tions? A commentary on "Ghild and parenteffects in boy's conduct disorders." Develop-mental Psychology, 26, 702-704.

White, M. M., Walsh, B. W., & Gibbs, J. G. (1988).Relations between parents' discipline tech-niques and empathy in preadolescents. Un-published manuscript, Ohio State University,Columbus.

Zahn-Waxler, C , Kochanska, G., Krupnick, J., &McKnew, D. (1990). Patterns of guilt in chil-dren of depressed and well mothers. Develop-mental Psychology, 26, 51-59.

Zahn-Waxler, G., Radke-Yarrow, M., & King, R.(1979). Ghild-rearing and children's prosocialinitiations toward victims of distress. ChildDevelopment, 50, 319-330.