1 Harsh Parenting in Relation to Child Emotion Regulation and Aggression: Lei Chang Chinese University of Hong Kong David Schwartz University of Southern California Kenneth K. Dodge Duke University Catherine A. McBride-Chang Chinese University of Hong Kong Submitted May 2002 This study was supported by a Mainline Research grant (No. 44M2008) from the Chinese University of Hong Kong and a Quality of Education Fund grant (No. EMB/QEF/P1999/2735) from the Hong Kong SAR government. We thank Xinyin Chen for his suggestions and help with the study. Correspondence regarding this manuscript should be addressed to the first author at
2 Harsh Parenting 1 Abstract This study presents a model of harsh parenting that has an indirect effect, as well as a direct effect, on child aggression in schools through the mediating process of child emotion regulation. This model was cast within the family system framework that controlled for marital conflict and parental depression. Tested on a sample of 325 Chinese children and their parents, the model showed adequate goodness of fit. After controlling for emotion regulation, harsh parenting showed a reduced effect on child aggression. Harsh parenting was also found to mediate the family subsystem malfunctioning on child behaviors. Also hypothesized and tested were interaction effects between parents and children s genders. Harsh parenting of fathers affected sons more than daughters, whereas mothers harsh parenting had a similar effect on children of both sexes. The discussion of these results provides cross-cultural glimpses into the processes underlying parent-child relations.
3 Harsh Parenting 2 Harsh Parenting in Relation to Child Emotion Regulation and Aggression Many empirical studies have examined the association between harsh parenting and child aggression (Bierman & Smoot, 1991; Colder, Lochman, & Wells, 1997; Dodge, Bates, & Pettit, 1990; Kochanska, Padavich, & Koenig, 1996; Nix, Pinderhughes, Dodge, Bates, & Mcfadyen-Ketchum, 1999; Patterson, Reid, & Dishion, 1992). A review of these studies suggest that the zero-order correlation between harsh parenting and child aggression ranges between.30 and.40 when parental reports have been used to measure child aggression and between.20 and.30 when teacher ratings or peer nominations have been used. A recent development of this literature seems to have shifted from the main effect to the interaction between harsh parenting and parental warmth in predicting child aggression (Patterson et al., 1992; Pettit, Bates, & Dodge, 1997). This work has been extended by Deater-Deckard, Dodge, Bates, and Pettit (1996, see also Deater-Deckard & Dodge, 1997), who have suggested a non-linear model where the association between harsh parenting and child aggression depends on whether parental disciplinary actions are carried out in an emotionally controlled or an emotionally charged manner. This non-linear relationship implicates the emotional influence of harsh parenting on child aggression independent from the physical act of punishment. In a separate pursuit, Eisenberg and her colleagues have examined emotion regulation on the part of both children and their parents. Their prodigious findings suggest that children s emotion regulation is affected by their parents punitive emotions (e.g., Eisenberg et al., 1999) and, in turn, affects an array of social behaviors (e.g., Eisenberg, Fabes, Guthrie, et al., 1996; Fabes, Eisenberg, & Miller, 1990) including aggression. Putting these two separate developments together, it would be theoretically enriching to examine children s emotion regulation in mediating the effect of harsh parenting on child aggression. Another theoretically important direction through which to pursue harsh parenting research is to examine harsh parenting and child behavior within the family system (Cox & Paley, 1997). In the system approach, the negative effect of harsh parenting is seen as emanating from and disturbing other sub-systems, resulting in spillover effects from system malfunctions. Among the perturbations of family sub-systems, marital conflict and its cause and consequence (Downey & Coyne, 1990), parental depression, have been widely implicated to precede both harsh parenting and child behavior and emotional problems (e.g., Conger et al., 1992). Thus, including marital conflict and parental depression in an investigation of harsh parenting provides a system view of the latter s etiology and malfunctioning.
4 Harsh Parenting 3 A third way to expand the literature is to investigate possible gender effects on the prediction and outcome of harsh parenting. With a few exceptions (e.g., Weiss, Dodge, Bates, & Pettit, 1992; Deater-Deckard & Dodge, 1997), investigators have not focused on the gender of parents or children. Partly because the association between parent variables and child behaviors appears stronger for mothers than for fathers (Rothbaum & Weisz, 1994) and for sons than for daughters (Bronstein, Fitzgerald, Briones, Pieniadz, & D Ari, 1993; MacDonald & Parke, 1984), especially with young children (Eisenberg, Fabes, Murphy, et al., 1996), as Deater-Deckard and Dodge (1997) noted, most published studies on harsh parenting employ mothers and sons rather than fathers and daughters. A study that includes both parents and that separates children by gender will help to reveal possible gender effects of harsh parenting. Finally, despite a growing interest in the study of parenting and other developmental processes in non-western populations (e.g. Developmental Psychology, Vol. 34, Iss. 4, 1998), no study to our knowledge has specifically looked at harsh parenting in Chinese samples. Deater- Deckard et al. (1996) examined harsh parenting across sub-cultures within the United States. A study based on Chinese, who represent one fifth of the world s population, provides yet another important cultural perspective on the processes underlying parent-child relations. In this study, we examined harsh parenting in relation to the child s emotion regulation and aggression in schools and in relation to marital conflict and parental dep ression. These effects were tested in a set of simultaneous equations on 325 Chinese children and their parents. The model that distinguished harsh parenting of mothers from that of fathers was also tested in sons and daughters separately. The Mediating Effect of Emotion Regulation Coinciding with the manifestation of other temperamental characteristics (Rothbart & Ahadi, 1994), emotion regulation comprises a set of competencies to modulate affective states (Shields & Cicchetti, 1998). Examples of emotion regulatory strategies include self-soothing, re-framing upsetting events and provocative stimuli (Schwartz & Proctor, 2000), and inhibiting or initiating emotionally driven behavior (Eisenberg, Gershoff, et al., 2001). These abilities are formed in the family context and transferred to the peer realm (Fabes et al., 1990). Parents shape children s acquisition of regulation skills by coaching and modeling (Carson & Parke, 1996; Gottman Katz, & Hooven, 1997; Kopp, 1982) and simply through parent-child interactions (Gottman & Katz, 1989; Parker & Gottman, 1989; Parke, Cassidy, Burks, Carson, & Boyum, 1992). According to Gottman, Katz, and Hooven (1996), parental coaching in relation to emotion and support of emotion affects
5 Harsh Parenting 4 children s regulatory physiology and ability to regulate their emotions. Parental coaching helps children to develop the ability to inhibit negative affect, to self-sooth, and to focus attention (including attention in social contexts). (Eisenberg et al., 1999, p.514). Parents who exhibit hurtful and hostile negative emotions frequently may model dysregulated behavior for children to imitate (Eisenberg, Gershoff, et al., 2001, p.488). A number of empirical studies bear out the link between the emotion regulatory abilities of parents and their children. For example, during constructive play activities, a mother s display of negative emotion (e.g., negative tone, disapproval, and exasperation toward her child) was significantly and robustly correlated with her child s overall emotion regulation (Melnick & Hinshaw, 2000). Parke et al. (1992) reviewed their own empirical investigations to demonstrate the association between parents emotion expression and regulation and their children s social skills. Carson and Parke (1996) also showed the reciprocal nature of parents and children s emotion dysregulation and how it escalated during experimental play sessions. In another study of similar age group of 4- to 6- year-olds, children s real-life reactions when angered and frustrated coincided with parental negative emotions (Eisenberg & Fabes, 1994). Although these studies do not directly involve harsh or punitive parenting, the latter clearly carries the same negative or punitive emotions (Eisenberg et al., 1999, p.531). Variously described as harsh (Deater-Deckard & Dodge, 1997), overreactive (Arnold, O Leary, Wolff, & Acker, 1993), emotionally negative (Kim, Hetherington, & Reiss, 1999), coercive (Rubin, 1995), overly strict (Vuchinich, Bank, & Patterson, 1992), punitive (Crockenberg, 1987), and controlling and authoritarian (Baumrind & Black, 1967), the specific acts comprising a cluster of harsh parenting behaviors include yelling, frequent negative commands, name calling, overt expressions of anger, and physical threats and aggression (O Leary, Slep, & Reid, 1999). These behavioral descriptions can be summarized into coercive acts and negative emotion expressions. In other words, sometimes parents hit their children when they are angry (Dix, Reinhold, & Zambarano, 1990; Downey, Osatinski, & Pettit, 1993; Patterson, 1982) or enraged and emotionally out of control (Peterson, Ewigman, & Vandiver, 1994; Whiteman, Fanshel, & Grundy, 1987). Indeed, because of the dysregulated emotions in harsh parenting, intervention programs for abus ive parents usually include keeping an anger diary and training in anger management to help parents monitor and regulate their emotions (Wolfe, 1991).
6 Harsh Parenting 5 The available evidence suggests that there is a moderately strong link between parenting styles and chidlren s capacities for emotion regulation. The emotion dysregulation displayed by parents through harsh or punitive parenting affects the ability of their children to regulate their emotions (Eisenberg et al., 1999). Facilitated or even learned from parental socialization of negative emotions (Eisenberg, Losoya, et al., 2001) including harsh parenting (Gottman et al., 1997; Kopp, 1982), children s emotion dysregulation in turn leads to an array of social problems in schools (Eisenberg, Fabes, Guthrie, et al., 1996; Fabes, Lenard, Kupanoff, & Martin, 2001). According to Sroufe and Fleeson (1986), aggressive physical and emotional interactions between parents and a child form the basis for how the child interacts with others. Children transfer some of the negative affect and poor regulation strategies that they have learned from parent-child interactions to their interactions with peers, resulting in incompetent peer relations (Parke et al., 1992). Research has clearly implicated the link between a child s emotion dysregulation and aggression (Coie, 1990; Crick & Dodge, 1996; Pelligrini, 1989; Schwartz & Proctor, 2000). More recent studies by Eisenberg and her colleagues also suggest the mediating effect of children s emotion regulation in channeling the effect of emotion-related negative parenting practices on children s social adjustment (Eisenberg, Spinrad, & Cumberland, 1998; Eisenberg, Gershoof, et al., 2001). For example, mothers expressions of negative emotions were related to children s externalizing behavior problems in schools through the influence of the former on children s emotion regulation (Eisenberg, Gershoof, et al., 2001). Given this literature, we propose that harsh parenting will lead to child aggression, partly through a direct effect of modeling (see Rothbaum & Weisz, 1994, for a review) but also partly as an indirect effect that is mediated by the child s emotion dysregulation. Perturbations in the Family System The above hypothesized effects are viewed in the family system as perturbations resulting from other subsystem malfunctions. According to the family system approach (Cox & Paley, 1997), tension from one subsystem spills over to another subsystem (Almeida, Wethington, & Chandler, 1999; Erel & Burman, 1995). This spillover effect is especially obvious in the tension emanating from a marriage to that involving child rearing (Margolin, Christensen, & John, 1996). Numerous empirical studies have shown a link between self-reported or observed marital conflict and lowered parental warmth (Vandewater & Lansford, 1998; Gonzales, Pitts, Hill, & Roosa, 2000) and heightened parental rejection of and hostility toward children (Harold & Conger, 1997; Harold, Fincham, Osborne, & Conger, 1997; Mann & MacKenzie, 1996). A recent experiment showed that
7 Harsh Parenting 6 conflictual exchanges between the parents were more likely to elicit disrupted and less responsive parenting behaviors (Kitzmann, 2000). Equally clear is the link between harsh parenting and child problem behavior, especially of externalizing nature (e.g., Rothbaum & Weisz, 1994). However, somewhat controversial is the effect of marital conflict on child behaviors. Several studies suggest a direct effect of marital conflict on child behaviors (e.g., Katz & Gottman, 1993; Jekielek, 1998; Gryc h, 1998). Other studies have included parenting or parent-child relation and found the latter to mediate the negative effect of marital conflict on child behaviors (Fauber, Forehand, Thomas, & Wierson, 1990; Mann & MacKenzie, 1996; Osborne & Fincham, 1996; Vandewater & Lansford, 1998). For example, in a set of structural equation models, Margolin and John (1997) showed that, for both daughters and sons, parenting power assertion significantly mediated the positive effect of marital aggression on child hostility. For both genders, power assertion, as well as positive parenting, significantly attenuated the direct link between parental aggression and child hostility, whereas the mediating effect remained intact with or without the estimation of the direct effect. The same mediating effect has been observed in ethnic samples (Gonzales et al., 2000). A closer look at this body of research suggests several characteristics distinguishing between the putative direct versus indirect effects. First, many of the studies identifying a direct effect of inter-parental conflict on child behavior did not include parenting or parent-child relations as a mediating variable (e.g., Katz & Gottman, 1993; Jekielek, 1998; Jouriles, Murphy, & O Leary, 1989; Wierson, Forehand, & McCombs, 1988). When the latter mediation is incorporated, the direct effect seems to dissipate or attenuate (e.g., Fauber et al., 1990; Harrist & Ainslie, 1998; Margolin & John, 1997). Second, much of the direct effect seems to emanate from a specific inter-adult conflict, such as verbal abuse and insult (e.g., Davies, Myers, & Cummings, 1996; Cummings, Ballard, El-Sheikh, & Lake, 1991) or physical aggression (Cummings, Vogel, Cummings, & El-Sheikh, 1989; Cummings, Simpson, & Wilson, 1993), often derived from laboratory experiments (e.g., Cummings et al., 1993). The indirect effect, on the other hand, has been observed primarily in questionnaire data tapping general marital functioning or satisfaction (Gonzales et al., 2000; Mann & MacKenzie, 1996; Holden & Ritchie, 1991; Tschann, Johnson, Kline, & Wallerstein, 1989). Third, the direct effect of inter-adult conflict on children has been primarily internalizing leading to fear (El-Sheikh, 1997), anxiety (Katz & Gottman, 1993), or depression (Jekielek, 1998). The indirect effect mediated by parenting practice, on the other hand, is of both internalizing and externalizing with the latter seemingly outnumbering the former. Finally, related to the last characteristic, the guiding theory
8 Harsh Parenting 7 underlying the causal mechanism of a direct effect suggests that children must have the cognitive capacity, e.g., causal inference and attribution and goal setting and attainment, to experience distress and threat to security (Davies & Cummings, 1994; Grych & Fincham, 1990). Subsequently, most of the studies discerning a direct effect have been based on school-age children or older (e.g., Hennessy, Rabideau, Cicchetti, & Cummings, 1994; Cummings et al., 1991). Given these characteristics and the features of the present study, self-reports of general martial conflict is hypothesized to affect young children s emotion regulation and aggression indirectly through the mediating role of harsh parenting. This hypothesis, however, is tested against an alternative model that stipulates both indirect and direct effects of marital conflict. In the same family system framework, harsh parenting is similarly hypothesized to mediate the negative effect of parental depression on child aggression and emotion regulation. Parental depressio n has often been studied as a comorbid of marital conflict in affecting child rearing practice (Conger et al., 1992; Kitzmann, 2000). Related to marital conflict both as a cause and as a consequence (Downey & Coyne, 1990; Fendrich, Warner, & Weissman, 1990), parental depression has shown similar effects on parenting and child behaviors. A number of studies suggest a direct link between parental depressive moods and child externalizing behaviors (Dumas, Gibson, & Albin, 1989; Hammen, Burge, & Stansbury, 1990; Harnish, Dodge, Valente, & Conduct Problems Prevention Research Group, 1995; Weissman et al., 1984). Most of these studies, however, used mother reports to measure child behavior. Because depressed mothers tend to perceive the behavior of their children negatively (Brody & Forehand, 1986; Forehand, Wells, McMahon, Griest, & Rogers, 1982), the direct link between maternal depression and child externalizing obtained from mother reports is confounded by the same source of depressive symptomatology. Other researchers have theorized an indirect effect mediated by perturbations in parenting practice and have received consistent empirical support (e.g., Emery, Weintraub, & Neale, 1982; Susman, Trickett, Iannotti, & Hollenbeck, & Zahn-Waxler, 1985). Mother-child interaction measured through home-visit observations was found to significantly mediate the effect of maternal depressive symptoms and child externalizing behavior, measured by teacher reports, in a large non-clinical sample of parents and their children of about six years old (Harnish et al., 1995). In another non-clinical sample (Miller, Cowan, Cowan, Hetherington, & Clingempeel, 1993), parental depression was found to have no direct effect on child externalizing, which was obtained from multiple informants including parents, but only an indirect effect through the mediating channel of harsh parenting. The mediating effect
9 Harsh Parenting 8 has also been found in ethnic samples (e.g., McLoyd, Jayaratne, Celballo, & Borquez, 1994). Given this literature and the earlier theorizing about marital conflict, the same modeling approach concerning marital conflict is taken with parental depression. In summarizing the above, Figure 1 contains our model of harsh parenting. Six sets of simultaneous equations comprise the model: 1) harsh parenting is predicted by marital conflict and parental depression; 2) harsh parenting predicts child aggression; 3) child emotion regulation predicts child aggression; 4) harsh parenting predicts child emotion regulation; 5) child emotion regulation partially mediates the effect of harsh parenting on child aggression; and 6) harsh parenting mediates the effects of marital conflict and parental depression, respectively, on child emotion regulation and aggression. This model is also tested against an alternative model that includes the direct effects of marital conflict and parental depression on child variables. (Insert Figure 1 here.) Genders of Parents and Child ren The effect of gender in the parenting literature is somewhat perplexing, partly because gender is sometimes conceptualized as a nuisance variable but not a conscious target of investigation. In reviewing this literature, we have found inconsistencies in many of the reported normative differences between the two sexes, except for the observation that boys are more aggressive than girls (e.g., Rubin, Hastings, Chen, Stewart, & McNichol, 1998). For example, in some studies, mothers have been found to engage in more disciplinary actions with their children than fathers (Patterson, 1982) and, at times, to administer more physical punishment than fathers (Mulhern & Passman, 1981). Others have found no difference in harsh parenting (Feldman & Wentzel, 1990) or the expressed parenting styles between mothers and fathers (Lytton & Romney, 1991; Rubin, Nelson, Hastings, Asendorpf, 1999). The common belief that parents use more physical punishment on their sons than on their daughters is also not fully corroborated (Holden, Coleman, & Schmidt, 1995; Lytton & Romney, 1991). These mixed normative findings point to the importance of including gender as an independent variable in parenting investigations but offer little guidance in hypothesizing the direction of any normative difference. Instead, we focused on the processes rather than the outcomes of gender. From this perspective, we looked for gender effects in the parent-child relations and tried to discern patterns to guide our investigation. One pattern that emerges from the literature is that the effect of harsh parenting on child aggression seems to be stronger when the parent and child are of the same gender.
10 Harsh Parenting 9 For example, in Deater-Deckard and Dodge s study (1997), the average correlation between harsh parenting and child aggression across several age groups was.24 for mothers with daughters but.14 for mothers with sons. It was.28 for fathers with sons but.19 for fathers with daughters. In another study, the emotion regulation of fathers but not mothers was found to correlate with that of their ADHD sons (Melnick & Hinshaw, 2000). According to Bacon and Ashmore (1985), parents in general see aggression from the same sex children as more problematic than that from the opposite sex children. An early study also shows that fathers reported more frequent use of physical punishment and tangible rewards with sons than daughters (Baumrind & Black, 1967). This pattern of gender differences is consistent with the general observation that parents, especially fathers, interact with children of the same gender more often than with those of the opposite gender (Stoneman & Brody, 1981; Bell, Johnson, McGillicuddy-Delisi, & Sigel, 1981). This pattern of gendered findings is consistent with social learning theory (e.g., Bandura & Walters, 1959) which postulates that the role modeling effect is facilitated by gender identification. Early studies on socialization and gender identification (Maccoby & Jacklin, 1974) suggest that children have a preference for imitating adults of the same rather than opposite gender. This gender identification effect was observed in experimental studies (Ochman, 1996; Rothbaum, Zigler, & Hyson, 1981; Wolf, 1973; 1975) as well as in natural observation in the classroom (Stake & Granger, 1978; Stake & Noonan, 1985) or the family setting (Munsinger & Rabin, 1978), especially in modeling antisocial behaviors (Jankowski, Leitenberg, Henning, & Coffey, 1999). One nuance of this gender identification literature also suggests, however, that girls do not identify as strongly as boys with same-sex role models (Brown, 1957; Bussey & Perry, 1982; Seegmiller, 1980). Similar gender differential effect has been observed in Asian children. For example, in an experiment involving Japanese kindergarteners, boys showed a higher preference for toys that had previously been model-played by same-sex adults than girls (Minami, 1978). Thus, we predict the effect of harsh parenting on child aggression and emotion regulation to be stronger when parents and child are of the same gender, and we predict a stronger gender identification effect among sons and fathers than daughters and mothers. The two separate sets of parental variables also allowed an examination of possible gender differences in family system functioning or malfunctioning. Although parental depression seems to affect the discipline practice of both mothers and fathers (Stoneman, Brody, & Burke, 1989), the effect is more clearly shown in mothers (Conger, Patterson, & Ge, 1995). For example, it was
11 Harsh Parenting 10 associated with mother-son relations (Osborne & Fincham, 1996), mother-adolescent conflict (Smith & Forehand, 1986), and mother-child play (Roopnarine, Church, & Levy, 1990) and involvement (Fishman & Meyers, 2000). In these studies, the same effect on father-child interactions was either non-existent or less robust. Thus, we expected parental depression to have stronger effects on the harsh parenting of mothers than of fathers. However, the literature on marital conflict suggests little gender difference in relation to mothers and fathers parenting practice. For example, the magnitude of the positive effect of marital dissatisfaction was almost identical on fathers and mothers rejection of children (Mann & MacKenzie, 1996). In another study, marital conflict also affected parental hostility for mothers and fathers equally (Harold et al., 1997). Thus, we anticipate a similar effect of marital conflict on mothers and fathers. Harsh Parenting in Cultural Contexts Despite a growing interest in learning psychological processes from diverse cultural samples, the Chinese and other Asian populations are understudied in the social developmental domain. The scant literature that exists also shows a lack of balance, both in terms of the focus of the studies and in terms of the developmental ages these studies cover. For example, many of the existing parenting studies on Chinese children focus on cognitive development and academic achievement (e.g., Chao, 1994; Leung & Kwan, 1998; Leung, Lau, & Lam, 1998). A few studies that do include social variables treat these variables primarily as facilitators of academic development (e.g., Chen, Dong, & Zhou, 1997; McBride-Chang & Chang, 1998). These primarily draw on samples from late primary school to middle school years. Few studies have examined preschool age and no study we are aware of has examined harsh parenting in relation to children s social and emotional outcomes. The preschool age on which we choose to study harsh parenting and child aggression is also well suited with existing developmental theories underlying the mechanisms of these behaviors. For example, for this age group, the outcome of negative parenting is primarily externalizing, as opposed to internalizing (Dodge, Pettit, & Bates, 1994), and negative parenting takes the form of punitive and harsh discipline in contrast to psychological control (Barber, 1996; also see Pettit, Laird, Dodge, Bates, & Criss, 2001). There is no evidence suggesting cultural differences in this age-related characteristic of parent-child relations. In studying the effect of harsh parenting on Chinese preschoolers, we take a phylogenetic approach, against a tradition that, when a study is conducted based on non-western samples, its focus has to be the identification of cultural or ethnic differences.
12 Harsh Parenting 11 In many areas of psychology this approach is acceptable, but it is particularly appropriate in the study of harsh parenting because we believe that both the outcome and the process of disciplinary parenting behaviors share many cultural similarities. In commenting on early cross-cultural surveys, Rohner (1975) reached a similar conclusion that the belief, practice, and function of disciplinary parenting are universal. He also pointed out three universal dimensions of parenting. These are warmth, rejection, and neglect (Rohner, 1975; 1986). This pan-cultural view on parenting is echoed by another cultural theorist, whose work has otherwise been cited often to support East-West crosscultural differences. Ho (1986) acknowledged, in a review of Chinese socialization processes, that parental warmth and control have much cross-cultural common ground. The pan-cultural view also finds support in several empirical studies on Chinese parenting (Chen et al., 1997; Ho & Kang, 1984; Lau, Lew, Hau, Cheung, & Berndt, 1990; McBride-Chang & Chang, 1998). These studies suggest that parental warmth and control underlie similar psychological processes in both Chinese and Western cultures. Other studies point to East-West cultural similarities in parenting on the behavioral level. For example, American and Japanese mothers both used rule based strategies to gain compliance from their children (Conroy, Hess, Azuma, & Kashiwagi, 1980). Disobedience shown by children was a cause for concern to both Caucasian and immigrant Chinese mothers in the U.S. (Kelley & Tseng, 1992). Chinese, immigrant Chinese and Caucasian American mothers equally used open expressions of affection towards their children (Lin & Fu, 1990), while Quebecois, Haitians and Vietnamese mothers living in Montreal showed no difference in their attitudes towards child care or in their use of physical punishment (Pomerleau, Malcuit, & Sabatier, 1991). Finally, there seem to be few East- West differences in the use of physical punishment. For example, a survey of American and Indian college students showed that 93% and 91%, respectively, of males and 91% and 86%, respectively, of females reported having received corporal punishment during childhood (Graziano, Lindquist, Kunce & Munjal, 1992). A national family violence survey conducted in the U.S. in 1985 showed that 90% of the parents of 3-year olds used physical punishment (Straus & Donnelly, 1994). According to a similar survey conducted in Hong Kong during the same period, 95% of the respondents reported using physical punishment in their home (Samuda, 1988). Straus and Donnelly s survey also indicated that American mothers and fathers engaged in equal amounts of disciplinary parenting. The same finding was obtained from a study of Taipei mothers (Stevenson, Chen, & Lee, 1992). Although there is also literature indicating main effect differences in parenting strategies across
13 Harsh Parenting 12 cultures (e.g., Chao, 1994), no theory has postulated that the processes within the child that affecta the child s behavioral response to parenting will vary across cultures. In fact, recent studies on Chinese children s social-emotional functioning suggest the same psychological processes as found in the Western literature (Eisenberg, Liew, & Pidada, 2001; Schwartz, Chang, & Farver, 2001; Schwartz, Farver, Chang, & Lee-Shin, 2002). The purpose of our study is to use a non-western sample to glimpse the processes of harsh parenting within its cultural context. Such a study should help better understand harsh parenting in the same way a study based on Western subjects would. Method Samples and Procedures Samples were taken from two kindergartens in a southern Chinese city, which is one of the best-developed urban areas in China. The two kindergartens were randomly selected from those operated by the municipal pre-school education system. These kindergartens enjoy a higher academic reputation than those run by independent organizations such as various workplaces. In general, the family social-economic status of the children enrolled in the government-run kindergartens lies between that of the independents and that of a small but growing number of more expensive private schools. Parents of the participating children completed a set of questionnaires, at home, about themselves and their child. About six months after the initial questionnaire collection from the parents, teachers filled out a behavioral checklist on the children in their classes. In these kindergartens, each class had one teacher and two assistant teachers. All three adults independently rated each child in their class. Thus, each child had three independent teacher ratings, for which the teachers were paid. Parental consent was obtained from all the participating children at each stage of the research. Like most schools and kindergartens in China, these two kindergartens act in guardianship of the children during the course of school day. These kindergartens do not obtain written parental permission before involving children in special activities. To respect school practice, verbal consent was sought from the parents on multiple occasions during parent meetings frequently held within classes. In addition, reassurance of confidentiality and voluntary participation was stated in other written communications with the parents. Parents were fully informed that the study was conducted by researchers independent from their child s school and that, if they did not want their child to be included in the study, they could so inform the contacting researcher.
14 Harsh Parenting 13 The sample, which had complete data from two parents and three teachers, consisted of 325 children of which 45% were female. This represents a return rate of above 95%. A few cases of single parents were not included in the sample. One case where grandparents who were legal guardians of the child filled out the questionnaires were not separated out. The age of the children (Mean = 4.6; SD = 0.97) ranged from 3 to 6 years with the majority (65%) falling between 4 and 5. The average age of the parents was (SD = 3.96) for fathers and (SD = 3.34) for mothers. Sixty percent of the fathers and 50% of the mothers had a college education. The remainder had a high school (30% of fathers and 40% of mothers) or middle school (close to 10%) education with three parents having just an elementary school education. On all of the demographic variables and other variables used in the study, there were no statistical differences between children from the two kindergartens. Measures Parental Depression. Both parents filled out 19 depression items from the Chinese Personality Assessment Inventory (CPAI, Cheung et al., 1996). CPAI is an indigenous Chinese personality test that has been validated in Chinese populations. Sample depression items included I m pessimistic about the future ; I feel depressed ; I m happy with what I have (the last being subject to reverse coding). Parents were asked to indicate whether or not each item represented their current state of mind. The internal consistency was.82 for fathers ratings and.80 for mothers responses. Marital Conflict. Both parents were also asked four questions regarding their marriage. The parents were asked, How often do you and your spouse have fights, feel annoyed by each other, resent each other, or consider divorce? They were asked to respond to the questions on a 6-point scale ranging from 1 = never to 6 = all the time. Internal consistency of the four marriage conflict items was.57 for fathers and.66 for mothers. These items originally written in Chinese have been used previously (e.g., Shek, 1995). Harsh Parenting. Both parents filled out the Chinese translation of the Parental Acceptance Rejection Questionnaire (Rohner, 1986). This translated form of the questionnaire has previously been used with Chinese parents (e.g., Chen, Rubin, & Li, 1997). Reported in this study were 11 items measuring harsh parenting. Sample items included, When my child does not behave, I will scold, kick, hit, get really mad with, or humiliate him/her. The items were presented on a 4-point
15 Harsh Parenting 14 scale ranging from rarely to always. Internal consistency was.66 and.71 for fathers and mothers responses respectively. Child Emotion Regulation. Mothers filled out the Emotion Regulation Checklist (Shields & Cicchetti, 1998). The scale was translated into Chinese by a psychology professor and proof read by two bilingual graduate students in psychology. The translated scale was piloted with an independent group of parents before its use in this study. The scale contains 24 items measuring the abilities and skills of a child to emotionally regulate and self-sooth. Sample items included, My child responds angrily to limit-setting by adults, can say when she/he is feeling sad, angry or mad, or is easily frustrated. Items were rated on a 7-point scale from 1 = never true of child to 7 = almost always true of child. Consistent with existing studies (e.g., Schwartz & Proctor, 2000), some items were reverse coded to form a single scale, with higher numbers indicating emotion dysregulation. Internal consistency reliability was.69. Child Aggression in School. One teacher and two assistant teachers of each child independently rated the child s school behaviors. The rating items were derived from the literature (e.g., Schwartz, McFadyen-Ketchum, Dodge, Pettit, & Bates, 1998) and have also been piloted with Chinese children of a similar background (Schwartz et al., 2001). In this study, seven ratings of bullying and aggressive behavior were used. The items are the child bullies, makes fun of, pushes and hits, starts fights with, takes things away from, yells at other kids and calls other kids bad names. The items were rated on a 5-point scale ranging from 1 = not at all true of child to 5 = very true of child. Internal consistency of the seven items was.90. For all the scales described above, mean scores instead of sums were computed. For the depression scale, where items were dichotomized, the mean score represents the proportion of the depressive responses. Results Means, standard deviations, and correlation coefficients of the variables used in the study are reported in Table 1. Table 2 presents the same information for sons and daughters separately. Boys scored significantly higher on school aggression and emotion disregulation than girls. There were no significant differences on parental variables between mothers and fathers. Two variables, child aggression in school and parental depression, had a positive skew of above one standard deviation. Log transformation was tried on these variables to normalize the data. However, results based on the
16 Harsh Parenting 15 transformed variables were almost identical to those based on the original variables. Therefore, all the analyses reported here are based on the original measurements. (Insert Tables 1 and 2 here) To test the model in Figure 1, individual items were first made into parcels to achieve an acceptable sample-size-to-variable ratio. This approach which has been shown to have satisfactory statistical properties (MacCallum, Widaman, Zhang, & Hong, 1999) is widely used in structural equation modeling studies (Bandalos & Finney, 2001). For the School Aggression construct, ratings on the seven items were averaged for each teacher to derive three composite ratings. Thus, this variable had three indicators representing the independent evaluations of the three teachers. For other constructs, two to four items were randomly combined to form item parcels. After parceling, three indicators represented each construct except for the two parental depression scales which each had two indicators. The resulting sample size to parameter-estimates ratio was 8, which is adequate for stable estimation (Bentler & Chou, 1987). Means, standard deviations, and correlation coefficients of the item parcels are reported in Table 3. (Insert Table 3 here) Tested using LISREL 7, the model had an adequate fit of data (χ 2 (191) = , p =.10; Goodness of Fit =.94; Adjusted Goodness of Fit =.92; Root Mean Square Residual =.05). The alternative model that included the estimation of eight direct effects involving marital conflict and parental depression did not improve the model fit (χ 2 (183) = , p =.07; Goodness of Fit =.94; Adjusted Goodness of Fit =.92; Root Mean Square Residual =.05; R 2 =.68). With a reduction of eight degrees of freedom, the expected chi-square reduction should exceed 40 to justify the additional parameter estimates. The resulting chi-square and other statistics, however, remained almost the same. None of the eight direct effects was statistically significant. Two of the strongest were fathers depression on child emotion regulatio n (β = , t = -0.73) and mothers marital conflict on child aggression (β = 0.093, t = -0.53). These results and the non-significant chi-square provide strong support of the hypothesized model. In addition, we did not allow measurement errors to correlate. While substantially improving the goodness of fit of any model, allowing measurement errors to correlate violates some of the basic psychometric assumptions. Parameter estimates of the model are presented in Figure 2. To facilitate interpretation, standardized regression coefficients are reported. Throughout the discussion, statistically significant effects have p-values of.05 or lower. (Insert Figure 2 here.)
17 Harsh Parenting 16 Almost all of the parameter estimates were in line with our hypotheses. Harsh parenting from both mothers (β =.39) and fathers (.33) affected child emotion regulation, which, in turn, led to school aggression (.28). These effects were significant. Fathers harsh parenting also had a significant effect (. 19) on child aggression. For mothers, however, the direct effect (-.06) on child aggression was not significant. Following Sobel (1988), we conducted significance test on the mediating effects. Estimates of the mediating effects of harsh parenting through emotion regulation and their asymptotic standard errors (in parenthesis) are.102 (.048) for fathers and.113 (.052) for mothers. Both effects were significant. In terms of fathers harsh parenting, the mediating effect through child emotion regulation represents 35% of its total effect ( ) on child aggression. The effect of harsh parenting of mothers on child aggression was entirely mediated by child emotion regulation. Also as hypothesized, harsh parenting mediated the negative effects of marital conflict and parental depression on child variables. For mothers, through the mediation of harsh parenting, the indirect effect on child emotion disregulation from parental depression was.098 (.032) and from marital conflict was.109 (.033). Both were significant. The indirect effect on child aggression was not computed because there was no direct effect between mothers harsh parenting and child aggression. Mediated through fathers harsh parenting, the indirect effect of marital conflict was.083 (.031) on emotion disregulation and.067 (.029) on child aggression. (The last effect was computed from a separate LISREL analysis that excluded the mediating effect of emotion regulation on child aggression.) Both were significant. The indirect effect of parental depression was not computed because there was no direct effect between parental depression and fathers harsh parenting. In line with our gender hypotheses, fathers were not affected by depressive mood (.02), whereas mothers were (.25). Also as hypothesized, marital conflict affected mothers (.28) and fathers (.25) equally. Both were significant. Similar results were obtained when sons and daughters were analyzed separately as presented below. To test the effect of genders of both parents and children, separate path analyses were conducted for boys and girls, respectively. Path analysis excluding the measurement model was used to accommodate the reduced sample size of each gender. Because the model was still under-identified, the same set of goodness of fit statistics was available to test the model. Correlation matrices, reported in Table 2, were used to conduct the path analyses. The model had adequate fit for each gender. For the female sample, χ 2 (12) = 10.16, p =.60; Goodness of Fit =.98; Adjusted Goodness of Fit =.95; Root Mean Square Residual =.01; for the male sample, χ 2 (12)
18 Harsh Parenting 17 = 6.95, p =.86; Goodness of fit =.99; Adjusted Goodness of Fit =.95; Root Mean Square Residual =.01. Standardized path coefficients are reported in Figure 3. An invariance test was also conducted. The results support structural invariance across the two genders (χ 2 (33) = 27.70, p =.73). (Insert Figure 3 here.) Somewhat consistent with our hypotheses, fathers harsh parenting affected sons more than daughters, whereas the gender identification effect was non-existent involving mothers and daughters. For fathers, the direct link between harsh parenting and child aggression was significant for sons (β =.20) but not for daughters (.07). Its effect on emotion regulation was also stronger when involving sons (.21) than daughters (.12). The former was significant but the latter was not. These gender differences, however, were not shown in mothers harsh parenting. The link between mothers harsh parenting and child emotion regulation was similar for daughters (.29) and sons (.27). These effects were statistically significant. The direct effect of harsh parenting on child aggression was equally weak for daughters (-.02) and sons (.10). In both cases, the estimates were not statistically different from zero. Discussion The major contribution of this study is the finding that harsh parenting affects the child s development of aggressive behavior through an effect on the child s regulation of emotions. As mentioned earlier, the zero-order correlation reported in the literature between harsh parenting and child aggression ranges between.20 and.30 when teacher reports have been used to measure child aggression. In our study the correlation was close to.30 for fathers and.20 for mothers. However, including emotion regulation, the link between harsh parenting and child aggression was significantly reduced, especially for mothers. For mothers, this effect was not statistically significant in relation to either sons or daughters. For fathers, it was significant for sons but not for daughters. Apparently, some of the total effect of harsh parenting on child aggression occurs indirectly, through emotion regulation. Our focus on the emotional influence of harsh parenting has been motivated by the research on emotion regulation (e.g., Carson & Parke, 1996; Boyum & Parke, 1995; Eisenberg et al., 1999; Fabes et al., 1999; Parke et al., 1992). These researchers have examined in detail the effect of affect communication in the family on the child s social behavior outside of the family. Their results suggest that children transfer negative emotion response strategies they have acquired from parental punitive emotions to other contexts, resulting in incompetent social behaviors. Viewed from the
19 Harsh Parenting 18 emotion regulation literature, our findings provide additional empirical evidence supporting the transfer of emotional incompetence. From the perspective of harsh parenting research, our findings further clarify the paths through which the negative effect of harsh parenting is channeled to children. Bridging these two corpora of research, our work provides a new perspective on harsh parenting. In addition to seeing it as a form of behavior, harsh parenting can also be seen as a form of affect communication. Its effect on children occurs both directly through behavioral modeling (e.g, Rothbaum & Weisz, 1994) and indirectly via emotion dysregulation. This conception is consistent with the common experience that when a parent hits or scolds a child, she/he communicates anger. Our findings suggest that the expression of anger, coldness, or hatred that accompanies the physical act of parental aggression could be more detrimental than the act of aggression. This perspective is especially relevant when viewed within the family system where parental punitive emotion and behavior are spillovers from other dysfunctional subsystems such as marital conflict and parental depression. Our results are consistent with the literature that marital conflict and associated depressive moods do not affect children s behaviors directly but indirectly through perturbations in parenting (e.g., Gonzales et al., 2000; Harnish et al., 1995). These mediated outcomes are especially expected of young children of the age sampled in this study. In early childhood, because compliance issues are most salient, parental hostility in gaining compliance is most likely to result in resisting behaviors and emotions from children (Pettit et al., 1997; Pettit et al, 2001). In gaining compliance of this age group, incompetent parenting strategies primarily take the form of punitive and harsh disciplines (Pettit et al., 2001), as in contrast to psychological control (Barber, 1996). As shown in this study, when parents felt depressed over their marital problems, they were more likely to deal with child compliance by way of harsh parenting, to which children responded with disregulated emotions and aggression. There is clearly a chain of spillover effects. It is important to understand and intervene into these perturbations in the family system early on. As children get older, escalation of these family malfunctions may lead to new problems in parent-child relations. Pettit et al. (2001) showed that as young children reached adolescence, punitive discipline was preceded by psychological control to which young adolescents responded with withdrawal and other internalizing adjustment problems. Implicit in the models of family systems (Steinglass, 1987) and family ecology (Belsky, 1984), the effect of parenting on child behavior is best studied by including both parents. Related to this ecological approach is the effect of gender on parent-child relations. By including in this study
20 Harsh Parenting 19 separate measures for both parents and for children of both sexes, we were able to identify several gender interaction effects. First, it was found that, controlling for emotion regulation, mothers harsh parenting had no effect on child aggression whereas fathers did. One speculation, consistent with earlier theorizing, is that the total effect of mothers harsh parenting is primarily emotional, whereas, for fathers, it also has a behavioral effect. Because of the differential role specializations of the two parents (Russell & Saebel, 1997), the nature and function of harsh parenting from the two parents may overlap but also differ. For example, since mothers spend more time with their children (Lamb, Ketterlinus, & Fracasso, 1992; Campos, Lamb, Goldsmith, & Stenberg, 1983), they are more likely to confront a child s incompliance issues than fathers. Harsh parenting from a mother is thus more likely to be a spontaneous and immediate reaction driven by her negative emotions than the harsh parenting of a father. On the other hand, harsh parenting from fathers is more likely to be delayed or even planned to serve the instrumental function of child discipline than the harsh parenting of mothers. This is also consistent with our finding that mothers were more responsive to their own moods than fathers in engaging in harsh parenting. This latter finding is consistent with the existing literature (e.g., Fishman & Meyers, 2000; Osborne & Fincham, 1996; Stoneman et al., 1989). Second, the hypothesized gender identification effect was found in fathers but not in mothers. That is, whereas fathers harsh parenting had a stronger effect on sons than daughters, mothers had a similar effect on children of both sexes. This finding is in line with research suggesting a higher level of gender role differentiation by fathers than mothers (Lytton & Romney, 1991; Siegal, 1987). For example, fathers treat sons and daughters differently, whereas the differential gender treatment is small to non-existent in mothers (Kerig, Cowan, & Cowan, 1993). The stronger effect of fathers gender differential treatment seems especially obvious in negative parenting. In 6 to 10 year olds, fathers punitive emotions were associated only with son s externalizing negative emotions (Eisenberg et al., 1999), whereas the same association involving mothers was held for both sons and daughters (Eisenberg et al., 1995; 1997). Early studies also suggest that fathers were more concerned with and restrictive about sons than daughters behavior (Power & Shanks, 1989; Rothbart & Maccoby, 1966) and used more physical punishment and tangible rewards with sons than daughters (Baumrind & Black, 1967). Finally, boys also seem to be more responsive to same-sex influences than girls (e.g., Minami, 1978). Our finding of a stronger gender identification effect in fathers and sons may help to untangle some potentially confounding findings in the literature. For example, separating the parents but not the child s gender, one may reach the conclusion that fathers harsh