Reading Analysis Assignment APA Format

Pergamon Child Abuse & Neglect, Vol. 21, No. I I. pp. 1081-1094, 1997 Copyright O 1997 Elsevier Science Dd Printed in the USA. All rights reserved 0145-2134/97 $17,(}0 + .00 PII S0145-2134(97)00067-7 TOWARD AN INTEGRATED FRAMEWORK FOR UNDERSTANDING CHILD PHYSICAL ABUSE CAROL COOHEY School of Social Work, University of Iowa. Iowa City. IA. USA NORMAN BRAUN Department of Sociology. University of Bern, Bern, Switzerland ABSTRACT Objectives: The article constructed an integrated conceptual framework for understanding child physical abuse based on a content analysis of different theories. Three major determinants of physical abuse were identified: exposure to aggression, exposure to stressors, and access to resources. To test the framework, we included variables from all three major categories in the analyses.

Methodology:

Data from a case-comparison study that included 81 physically abusive and 148 nonabusive mothers was used. Logistic regression determined the relative importance of the predictors. Results: The probability of child physical abuse varied in the expected direction with the number of emotional resources received (listening, help with decision-making, companionship), the number of stressors experienced, and the mothers exposure to physical abuse by their own mothers, current partners, and former partners. The estimation results showed that exposure to aggression in one's own childhood (abused by one's own mother) and exposure to domestic violence in one's adult life (abused by one's previous and current partners) were the most potent factors for predicting whether a mother physically abused her child. Conclusion: The results suggest further research on the interdependencies between domestic violence and child physical abuse. © 1997 Elsevier Science Ltd Key Words--Child abuse, Child maltreatment, Domestic violence, Battered women. INTRODUCTION NUMEROUS THEORIES HAVE been proposed to explain child physical abuse, sexual abuse, and neglect. A comprehensive survey by Tzeng, Jackson, and Karlson (1991) describes 25 differenl theoretical approaches for child physical abuse alone. Most of these theories are characterized by a large set of independent variables. Because variables in one theory overlap with variables in another theory, constructing and testing an integrated theory of child physical abuse should be possible. This article takes a preliminary step toward that goal by identifying, and then examining some of the most frequently occurring etiologic factors across existing theoretical perspectives.

This study was supported by the National Institute of Mental Health (F31 MHI034(t-01) and the New York Community Trust, Fahs-Beck Fund.

Received for publication December 16, 1996; final revision received April 30, 1997: accepted May 5, 1997.

Reprint requests should be addressed to Carol Coohey. University of Iowa, School of Social Work, North Hall. Iowa City, IA 52442. 1081 1082 c. Coohey and N. Braun The first section identifies three broad categories used to organize the most prominent explan- atory factors for physical child abuse. Based on this content analysis, we then construct a conceptual framework to predict child physical abuse. The second section prepares a partial test of the framework by describing the sample, measures, and logistic regression models. The third section reports the findings of the study. The last section summarizes the results and suggests further work with respect to the interdependencies between child physical abuse and domestic violence. INTEGRATED CONCEPTUAL FRAMEWORK A comparison of different theories in Tzeng's and colleagues (199 l) survey suggests that some common factors increase the probability of child physical abuse. The factors were categorized using standard techniques of content analysis (Patton, 1990) into three major groupings: exposure to aggression, exposure to stressors, and access to resources. While all 25 theories included factors that fell into two or three of these categories, each theory emphasized some factors over others. A review of the three categories is presented next. Within each category, we further subdivided factors, depending on whether they occurred inside the parent's personal network, (including the parent), or outside the parent's personal network. Social network nomenclature is adopted as means to integrate the factors and move toward a unified conceptual framework.

Exposure to Aggression Exposure to aggression occurs when an individual views or is the object of another individual's use of physical force. A parent can be exposed to aggressive models either within his or her personal network as a child or as an adult (e.g., between partners or siblings), or outside the personal network (e.g., between gang members or people in the media). A single exposure can vary in frequency (e.g., number of blows) and severity (e.g., hit with or without an object). It can be an enduring part of a person's life or a transient event. A review of the theoretical approaches shows that an abusive parent's own abusive childhood is believed to be a more important factor for predicting child physical abuse than aggressive models outside the parent's personal network. Psychiatric and learning theories, in particular, underscore the direct relationship between being abused, as a child, and abusing one's own children, as an adult, over other etiologic factors. Some theories emphasize the role that members outside the parent's network play in promoting or deterring the use of physical force within the family. While members inside an individual's network can object to or support the use of force, members outside personal networks can officially sanction offending members, or alternatively, reinforce the use of force by failing to act (e.g., exchange theory--Gelles, 1983; systems theory--Straus, 1973). Many theories rely on both sources of aggressive models--those inside and outside the parent's personal network---to explain child physical abuse (e.g., ecological theory--Belsky, 1980, Gar- barino, 1977; offender typology theory--Waiters, 1975; resource theory--Goode, 1971; social psychological theory--Gelles, 1973; social learning theory--Bandura, 1973; systems theory-- Straus, 1973; three factor theory--Lesnik-Oberstein, Cohen, & Koers, 1982). Two additional components were identified in the content analysis: exposure to stressors and access to resources.

Exposure to Stressors Stressors are occurrences, chronic or acute, that tend to add stress to our daily lives (general stress theory--Farrington, 1980). The terms "demand," "cues," and "stimuli" are sometimes used Integrated framework for child physical abuse 1083 interchangeably or substituted for the term stressor. Chronic occurrences typically refer to the potentially stressful living conditions of an individual that occur over a longer period (e.g., long term unemployment, poverty). Acute occurrences or life events occur at specific points in time. Different theories focus on different types of stressors when explaining pathways to stress and child physical abuse. Many theories highlight characteristics and needs of the child or character- istics of the parent-child interaction as major sources of stress that trigger physical abuse (e.g., coercion theory--Stringer & LaGreca, 1985; encounter theoryIZimrin, 1984; social interaction theory--Burgess, 1979; symbiosis theory--Justice & Justice, 1976; transactional theoryISamer- off & Chandler, 1975). Other theories focus on a broad range of stressors (e.g., death or poor health of a member in the parent's network, divorce or separation from a partner, birth of a child) that cumulatively affect the parent's behavior (Burgess, 1979; Farrington, 1980). We divide stressors into two categories: Those that occur within the parent (intrapersonal) and those that occur within the parent's network (interpersonal). Within the parent's network, stressors can be further subdivided into inside their personal network and outside their personal network. For example, within the parental network, parent-child and other relationships may be a source of stress. Or, stress may be the product of the parent's interactions with, for example, an employer.

co-workers, neighbors, strangers, caseworkers, and school officials. By dividing stressors into these subcategories, we retain the levels of ecology reflected in some theories (namely, ecological-- Belsky, 1980; Garbarino, 1982; Parke, 1982) and add greater specificity by identifying important exchange relationships. A similar strategy is used to categorize resources.

Access to Resources Resources are the means to accomplish one's goals (e.g., Farrington, 1986; Goode, 1971). They include the knowledge, skills, financial assets, and physical strength possessed by the parent, his or her partner, and other interaction partners. Thus, they may be found within the person (intraper- sonal) or in the parent's network (interpersonal). Access to resources, can be, to use Cicchetti and Rizley's (1981) terms, transient or enduring. For example, network members (e.g., a partner, a sister) may not be available throughout the family's child rearing years.

Most theoretical perspectives suggest that physically abusive parents lack certain psychological or intrapersonal resources (e.g., knowledge of child development, self-esteem, social skills, parenting skills). Goode (1971) and Gelles (1983) have both highlighted physical strength as a resource in their theories. Physical force is an important means for one person to get another person to do what he or she wishes. Resources exchanged between parents and members in their network (namely, family and friends) are frequently called social supports and are emphasized in ecological and resource-oriented theories (e.g., Belsky, 1980; Garbarino, 1977; Goode, 1971). Parents also get resources from members outside their personal networks (social systems theory--Gelardo & Sanford, 1987). Child care workers, social workers, teachers, and lawyers, for example, fall into this category. A few authors (e.g., Garbarino, 1977; Parke, 1982) emphasize variations in the parent's access to service providers depending on the community in which he or she lives. Economic resources are another major category of resources that parents get, usually from an employer or from the state, but sometimes illegally from other network members engaged in black market activities.

Two Additional Components A representation of the integrated conceptual framework is included in Figure 1. To the three components already described, we add two additional components that were infrequently discus- sed by most theorists: the parent's problem solving processes and being abused by a partner. 1084 C. Coohey and N. Braun PERSONAL CHARACTERISTICS AND EXPERIENCES • Access to Intrapersonal Resources • Exposure to Intrapersonal Stressors • Exposure to Aggression as a Child (e.g., parents toward each other, assaults by mother) • Exposure to Aggression by Previous Partner (e.g., partner toward children, assaults by partner) CURRENT ENVIRONMENTAL CIRCUMSTANCES • Access to Resources from Sources Inside and Outside Personal Network (e.g., listening from sister) • Exposure to Stressors from Sources Inside or Outside Personal Network (e.g., conflict with partner) • Exposure to Aggression from Sources Inside and Outside Personal Network (e.g., assault by current partner) PROBLEM SOLVING PROCESS Perceive child as a problem Perceive and evaluate alternative responses to problem(s), given personal and environmental circumstances Choose response (e.g., hit child) Perceive and evaluate results (e.g., reaction from child and others in the environment) Figure 1. Integrated conceptual framework. These additional components, and their relationship to stressors, resources, and aggression are described next.

Problem-solving processes. Some theorists have noted that many abusive parents lack impulse control or make decisions quickly without considering the full range of factors that impinge upon a situation (Azar, 1986; Lesnik-Oberstein et al., 1982; Waiters, 1975). We propose, as suggested by three recent models not included in Tzeng's 1991 review (Crittenden, 1993; Hillson & Kuiper, 1994; Milner, 1993), that a more detailed cognitive or information processing component be added to theories of child abuse. This component might serve as a mechanism for organizing the three major sets of factors discussed so far. Integrated framework for child physical abuse 1085 In Figure l we add a general problem-solving framework to represent the more detailed cognitive models offered by other theorists. Like stress and coping models (e.g., Hillson & Kuiper, 1994), we suggest that a stressor that is perceived as a problem or threat will activate the parent's problem solving capabilities. A parent then determines what resources they have to reduce the problem. For instance, if their child is perceived as the stressor, then a problem, the parent may use their knowledge of child development (an intrapersonal resource) to arrive at a nonviolent solution to the problem, or call on a partner (a potential interpersonal resource) to intervene. Alternatively, when a parent has been exposed to aggression as a child, he or she is likely to have fewer alternative and nonviolent strategies for coping with the child's perceived misbehavior. Consequently, child physical abuse is more likely to occur. In this fiamework, the level of analysis can be shifted up from the parent-child relationship to include other salient relationships within the parent's broader network. For instance, one can imagine a confluence of stressors originating from a range of interpersonal relationships that cumulatively overwhelm the parent's problem-solving capabilities that have the same consequence, child physical abuse. On the other hand, network members could be resources, not stressors. They could have initiated with the parent alternative strategies for coping with the problem that might have prevented an abusive incident.

Domestic violence.

Finally, to the framework we add being abused by a partner. In Gayford's 11975), O'Keefe's (1995), and Stark and Flitcraft's (1988) studies, over one-third of the battered mothers had physically abused their children. Although a relationship was found, it is not clear from these studies why being hit by a male partner should increase the likelihood that a mother would hit her child. Therefore, it is not clear where this factor should be situated in Figure 1.

Although being hit is an exposure to aggression, it differs from, for example, being hit as a child because it has occurred in the near past. We suggest that being abused by a partner is primarily a stressor that is likely to be perceived as a problem. To retain its prominence as a potential predictor of child abuse, we include exposure to aggression as an adult as a separate component in the Figure despite conceptualizing it as a stressor. HYPOTHESES As a preliminary step toward testing the framework, we include variables from all three major categories derived from the content analysis: exposure to aggression, exposure to stressors, and access to resources. Within each of these categories, we focus on the mother's interactions within her personal network. That is, we place greater importance on the effect of being hit by members within the mother's personal network, stressors that originate within her personal network, and resources available within her personal network. This approach departs from some ecological theories that emphasize multiple or higher levels of analysis (i.e., individual, family, community, and society). We restrict attention to members inside the mother's personal network because these role-types (close family members and friends), compared with other role-types (e.g., neighbors and co-workers), have been found to interact more frequently, live in proximity, be enduring sources of emotional and instrumental support and stress (see Laireiter & Baumann, 1992, and Wellman, 1991 for reviewst and are more likely to have hit the mother. Accordingly, we hypothesize that when the mother was assaulted by members in her personal network (the mother's own mother and father), is exposed to more stressors, (including being physically abused by a previous and current partner), and has fewer interpersonal resources, then the probability of child physical abuse will increase. The mother's problem-solving processes and other intrapersonal resources are not included in the analyses. 1086 C. Coohey and N. Braun Table 1. Demographic Characteristics (N = 229) Physical Abuse (n - 81) Comparison (n = 148) African-American ~' (%) Below 100% of the Federal Poverty Line (%) Below 120% of the Poverty Line (%) Mother's Age (Mean) Number of Children (Mean) Highest Grade Completed (Mean) 70.4 66.2 65.8 59.0 73.7 71.2 30.2 31.6 3.3 3.1 I 1.4 l 1.9 ~' All other mothers were Caucasian. METHOD Study Design and Sample We used a case-comparison design (Schlesselman, 1982) to collect data from physically abusive and nonabusive mothers. The case group included mothers that child protective services (CPS) had classified as physically abusive and neglectful, or physically abusive. They required these mothers to attend a parenting class and referred them to classes in the Chicago area. More than 96% of the CPS-classified mothers taking the classes agreed to participate in the study and completed the anonymous questionnaire during their regularly scheduled classes.

To find comparison mothers who would be comparable to the case mothers on race and income, we selected public schools in the same communities where the case mothers were attending parenting classes. We recruited mothers from several public schools on Report Card Day--a day where they require that all parents pick up their children's report cards. The mothers learned about the study from flyers handed out in the schools entrances on the same day they completed the anonymous questionnaires. Of the mothers who asked about the study, 99% agreed to participate.

Both the CPS-classified mothers and public school mothers completed the questionnaire in small groups. The administration of the questionnaire began with a presentation on how to complete the questionnaire. After orienting the mothers to the questionnaire and reading aloud the informed consent form, the mothers were given the option to participate in the study. A public school mother had to meet two criteria before she was assigned to the comparison group: She had no previous involvement with child protective services, and she had not severely assaulted her children in the last year. To determine whether a mother had severely assaulted her children, we used Straus's (1988) Conflict Tactics Scales (Form R). We chose two of their severe assault items to establish a criterion for child physical abuse. The two items were slightly altered to reflect one type of actual (versus attempted) assault. The item "kicked, bit, or hit with a fist" was changed to "hit them with a closed hand." "Hit them or tried to hit with something" was rephrased to "hit them with something." We weighted these items with the frequency of assault to create a conservative criterion of physical abuse: If a potential comparison mother hit her child with a closed hand at least once a month or hit her child with an object at least once a week within the last year, she was excluded from the study. This procedure eliminated 23 (or 13%) of the potential comparison mothers; it may have also left some abusive mothers in the comparison group. We chose to err on the side of leaving some abusive mothers in the comparison group, as opposed to creating an atypical comparison group. The final sample consisted of 229 mothers: 81 were physically abusive, and 148 were, presum- ably, not physically abusive. As shown in Table 1, physically abusive and comparison mothers were similar on level of poverty, race, number of children (SD = 1.81 and 1.67, respectively), education (SD = 1.76 and 1.73, respectively), and age (SD = 6.57 and 7.33, respectively). Integrated framework fl)r child physical abuse 1087 Measures Exposure to aggression.

Following our conceptual framework, we include both a history of childhood and partner abuse to assess the likelihood that a mother will be abusive. In determining a criterion for physical abuse, we again relied on the work of Straus and Gelles (1988). Three severe tactics were used: "Was hit with something" and "'Was hit with a closed hand." The third item, also taken from the Conflict Tactics Scales, was changed from "Was beat up" to "'Was hit a lot at one time." Mothers rated the frequency of the three types of severe assault perpetrated against them by four network members: their own mother /as a child), their own father (as a child), an ex-partner, and their current partner. The network members received a score of zero if they used no severe assaults against the mother and a score of one if they used any severe assaults.

Exposure to stressors.

Because of the large number of variables included in this study, {,nly 10 major life events occurring within the last year were included. Most of the items were adapted from Holmes and Rahe's (1967) Stressful Life Events Scale: "was separated or diw)rced,'" "'was pregnant or gave birth," "death of someone close," "had a serious health problem," "had a serious financial problem," "'moved," "stopped or started working," and "'stopped or started going to school." We added two additional items: "had a problem with welfare:" and "'partner has a drug or alcohol problem." Access to resources.

Mothers were asked whether they received six different types of resources within the last 30 days from up to 13 family members and friends from their personal networks:

"money or loans," "'help with housework~chores~errands," "'baby-sitting," "'really listens." "'help with making decisions," and "companionship (someone to do things with)." The last three items refer to emotional resources received (listening, decision-making, and companionship), whereas the first three items refer to transfers of instrumental resources (money, housework, and baby-~,itting).

Data Analysis Techniques Bivariate relationships between the predictor variables and the dichotomous dependent variable were calculated using chi-square and independent t-tests. These relationships may be of interest, since they include, for example, the percentage of mothers who abuse their children and are also abused by their partners. Our major interest, however, was in predicting the partial efl~cts of the independent variables on the probability that the dichotomous dependent variable (physically abusive or not abusive) takes on a certain value. Consequently, we used the simplest version of logistic regression analysis, the binary logit model (Aldrich & Nelson, 1984: DeMaris, 1992). To specify the regression models, we examined the pairwise interrelations between our measures. The bivariate analyses for all pairs of independent variables revealed just one significant association, a nonnegligible positive relation between the total number of stressors experienced during 1he last year and being severely assaulted by a former partner (Contingency Coefficient C = .30./~ < .02 t.

This relationship, between stressors and being assaulted, may reflect the inclusion of two partner- related stressors (being separated or divorced from a partner within the last year and having a partner with a drug or alcohol problem), since they may relate both to being physically assaulted by an ex-partner. Because of this association, we modeled the mothers' tendency for physical abuse in two different ways. In the first approach we used all of the st,'essors as an explanatory ~ariable and excluded being physically abused by an ex-partner. In the second model we created a measure for stressors that excluded the two partner related stressor items and included the additional regressor, being physically abused by an ex-partner. Since the results for both models were very similar (for the overall fit, Pseudo R:, statistical significance of the estimates, rankings of probability effects), we included results from the second model only. In both models, the effect of 1088 C. Coohey and N. Braun Table 2. Bivariate Relationships Physical Abuse (n = 81) Comparison (n = 148) Exposure to Aggression Severely assaulted by own mother (%)b Severely assaulted by own father (%)~ Severely assaulted by current partner (%)a Severely assaulted by ex-partner (%)b Access to Resources Instrumental resources received (Mean) Emotional resources received (Means) ~ Exposure to Stressors (Mean) c 89.7 62.6 53.1 36.7 24.1 13.0 54.8 27.9 9.44 9.72 11.66 15.35 2.58 2.15 a Chi-square test: p < .05. b Chi-square test: p < .0001. t-test: p < .05. o t-test: p < .0001.

being severely assaulted by one's father was not significant, and therefore, was eliminated. The technical appendix provides additional information on the statistical model and Petersen's (1985) method for calculating probability effects for logit models with binary and continuous predictors. RESULTS Table 2 includes results from the bivariate analyses. Except for instrumental resources, all relationships examined were significant at the p < .05 level. Physically abusive mothers were exposed to more stressors in the past year compared to mothers who had not physically abused their children. They also received fewer emotional resources than nonabusive mothers. The overwhelm- ing majority of physically abusive mothers had been severely assaulted (i.e., hit with something, a closed hand, and/or a lot at one time) by their own mothers, although many nonabusive mothers had also been abused. More than 50% of the abusive mothers had been severely assaulted by their own fathers, compared to 37% of the nonabusive mothers. Partners, too, were frequent perpetrators of assault against the physically abusive mothers. More than 50% of their previous partners had severely assaulted them in the past, and about 25% of their current partners had severely assaulted them within the last year.

These rates of assault are higher than in most studies using the CTS. Our data collection procedure or our modification of the CTS could have affected these findings. First, these data were collected from parents attending groups that supported the disclosure of parenting behaviors, including abusive behaviors. Second, changes in the wording of the CTS could have encouraged mothers to report higher rates. For instance, "Was beat up" was changed to "Was hit a lot at one time," a presumably less jarring description of the same event. Results from the multivariate analysis are included in Table 3 and Table 4. Inspection of Table 3 reveals significant contributions of all the independent variables in the expected direction and an acceptable overall fit of the model (Aldrich & Nelson's, 1984, Pseudo-R 2) suggests the importance of the predictors. All independent variables make significant contributions in the expected direc- tion.

Using the information in Table 3, we determine the effect of each independent variable on the probability of child physical abuse. To compute such effects, we need to select a reference case.

A theoretically meaningful and statistically plausible reference category is based on the modes of each dichotomous variable and the means of the continuous variables (see technical appendix for procedure). Table 4 summarizes the results. Integrated framework for child physical abuse Table 3. Logistic Regression of Child Physical Abuse on Emotional Resources, Exposure to Stressors, and Exposure to Severe Assaults Variable Coefficient Estimate t value Severely Assaulted by Own Mother Severely Assaulted by Current Partner Severely Assaulted by Ex-Partner Emotional Resources Received Exposure to Stressors Constant 1.990 3.999 '~ 0.870 1.890 t~ 0.75t) 2.119 ~ 0.085 3.498 "~ 0.207 1.74 I" 1.93q - 3.289 '~ N = 192: Model chi-square ¢ = 48.486: dt' 5: p < .00Ill: Pseudo-R ~ 0.20.

~' 17 < .08. h p < .06.

• p < .05.

a p < .00 I. 1089 The effects of our predictors correspond with our expectations. Consider first the shift of the variable being severely assaulted by one's mother as a child. An isolated variation from its modal value ("at least one severe assault by own mother") to its alternate category ("no severe assault by own mother") decreases the probability of an average mother's physically abusive behavior toward her child by .270. This remarkably strong effect of being abused and then abusing clearly supports theories that emphasize the intergenerational transmission of violence. A similar conclusion can be drawn about the effect of having been abused by one's current partner over the last year. According to Table 4, the probability of child physical abuse by the average mother increases .21 1 if, all other variables being equal, she now has a physically abusive partner. Due to the lack of information on the timing of the assaults, we do not know whether the partner's assaultive behavior contributes to the mother's abusixeness or vice versa. Theretore, the large probability effect suggests a plausible positive relationship between the presence of a physically abusive partner and the mother being physically abusive toward her children.

Having been severely assaulted by a previous partner also has a significant effect. A partial change from its modal category ("no severe assault by ex-partner") to the other category ("at least one severe assault by ex-partner") raises the probability of child assault by. 182 relmive to that of Table 4. Effects of Emotional Resources, Stressors, and Severe Assault on the Probability of Child Physical Abus& (N = 1921 Variation of Probability Probability Effecl Severely Assaulted by Own Mother 0.064 Severely Assaulted by Current Partner 1t.545 Severely Assaulted by Ex-Partner 0.516 Emotional Resources Received b 0.207 Exposure to Stressors ~ 0.405 .... 0.270 0.21 I 0.182 0.127 0.071 ~' Reference case: Average mother (Emotional resources 14.326, stressors 2.30, severely assaulted by own mother 1, severely assaulted by partner 0, severely assaulted by ex-partner = 0) with a probabilily of child physical abuse of 0.334. h Calculated for an increase of one standard deviation t Emotional resources : 22.024, stressors = 2.30, severely assaulted by own mother = I, severel) as saulted by partner = 0, severely assaulted by ex-partner - 0/. Calculated for an increase of one standard deviation IEmotional resources 14.326, stressors = 3.77, severely assaulted by own mother = 1, severely as- saulted by partner = 0, severely assaulted by ex-partner 0). I090 c. Coohey and N. Braun the average mother. Physical abuse by an ex-partner therefore seems to positively effect the mother's tendency to be abusive.

The relationship between being battered by a partner and battering one's children could be misleading, if, as suggested by Bowker, Arbitel, and McFerron (1988) and Stark and Flitcraft (1988), CPS falsely accused the battered mothers of physical abuse when in fact their partners abused their children. Data were available to verily that battered mothers had in fact abused their children, Mothers were asked if they severely assaulted their children in the last year. More than 94% (17) of the mothers who were severely assaulted by their current partner and more than 97% (38) of the mothers who were severely assaulted by ex-partners said they hit their children with an object, a closed hand, and/or a lot at one time within the last year. The assaults were not only severe; they were frequent. Fifty percent of the battered mothers said they hit their children at least once in the last month with a closed hand. Over one-fourth of the battered mothers said they hit their children a lot at one time in the last month. Therefore, it is unlikely that the battered mothers in this sample were falsely accused. All other variables being equal, an increase of one standard deviation for emotional resources corresponds to a reduction in the probability of physical abuse (compared to that of the average mother) by. 127. Thus, greater emotional support seems to reduce the likelihood for child physical abuse. Here, too, the causal relationship between the two variables is not clear. Physically abusive behavior toward one's children may decrease network members' emotional support, or lower emotional support from network members may increase the tendency to physically abuse.

The effect of potentially stressful life events in the last year is comparatively small. An increase of one standard deviation in the average number of stressors raises the probability of child physical abuse by .071. This relatively weak effect is probably due to shortcomings in our measurement approach. For example, it may be the specific characteristics of the child or the daily hassles a mother experiences--particularly those hassles related to parenting--that most influence her behavior (Kanner, Coyne, Schafer, & Lazarus, 1981). Nevertheless, stressors seem to have a positive influence on the probability of abuse even after controlling for other variables. CONCLUSIONS Based on a content analysis of existing theories, we presented a conceptual framework for understanding child physical abuse. This framework was partially tested with logistic regression using data from a case-comparison study of physically abusive and nonabusive mothers who were comparable on levels of poverty, race, number of children, education, and age. We identified the simultaneous effects of different abuse determinants and found that exposure to stressors within the last year, compared with all other predictors, had the weakest effect on the tendency toward physical abuse. Next, we found that the mother's tendency toward physical abuse was inversely related to the number of members who gave her emotional resources. Finally, we found that assaults by three types of prominent network members (her own mother, a previous partner, and current partner) increased the probability that a mother would physically abuse her children. Being abused by one's own mother dominated the effects of all other predictors. The violence begets violence effect is an intuitively plausible explanation for why an abused girl may, as an adult, abuse her own children; she learns a pattern of behavior that she replicates with her own children (i.e., learning from a role specific model--Seltzer & Kalmuss, 1988). The second most important probability effect was a current abusive partner. It exceeded in magnitude the probability effect of the mother's exposure to an assaultive former partner. The ranking of probability effects with respect to domestic violence at different points in time suggest more recent events may play a more important role than similar but more distant events. Integrated framework for child physical abuse 1~')91 In contrast to the transmission of abuse across consecutive mother-child relationships, it is theoretically less clear why being hit as an adult by a male partner should increase the likelihood of hitting one's children. Here, additional research is needed. What is clear is that the study of domestic violence and child physical abuse begins with at least three actors: two partners and at least one child. The three family members are likely to share a common set of interpersonal stressors and a common set of resources, but they also will have their own socialization experiences, stressors, and intra and interpersonal resources. How each of these factors affect each member's individual problem-solving process is a complex research question.

How each member's problem solving processes affect and are affected by the other's problem solving processes, given varying levels of stressors and resources, is an even more challenging question. There may be, for example, spillover processes such that aggressive conflicts between adults facilitate more conflictual responses by children, and vice versa, so that lower thresholds for the use of physical force eventually characterize how stronger family member's respond to adverse environmental circumstances. The simultaneous study of abuse by multiple family members will present many challenges to researchers, including identifying enough families where there is both domestic violence and child physical abuse. A few states (e.g., Florida, Iowa, Massachusetts) have begun to experiment with new child protection protocols that systematically identify abused children who also have battered mothers. These experiments are likely to create opportunities l'~r researchers. They are also likely to usher in a new set of issues related to culpability. One lear expressed by some feminist writers is that CPS will punish mothers because ~)f their inability to leave their abusive partners (Bograd, 1990: McKay, 1994: Stark & Flitcraft, 1988). This study shows that child physical abuse is more likely to occur when a mother is embedded in an environment that includes more stressors and fewer network members to turn to lot emotional support. In over one-half of the cases in this study, an abusive mother is also being abused by her partner or has been in the past. It seems reasonable then, to suggest that these circum,.tances, especially being battered, may limit an abusive mother's alternatives and actions. Instead of punishing mothers tbr their partners abusiveness, family advocates need to understand the c~m~plex decisions that battered mothers face and the resources that they will need to reexamine their relationships--not just with their children but also with their partners.

Ack.owledgeme.t--The authors would like to thank Susan Murty, Salome Raheim, and Susan Schecter for thcir helpful conlments. REFERENCES Azar, S. T. (1986). A framework for understanding child maltreatment: An integration of cognitive, behavioral and developmental perspectives. Canadian Journal of Behavioral Science, 18, 340-355. Aldrich, J. H., & Nelson, F. D. (1984). Linear probabiliO', logit, and probit models. Beverly Hills. CA: Sage. Bandura, A. (1973). Aggression: A social learning anah,sis. Englewood Cliffs. NJ: Prentice-Hall. Belsky, J. (1980). Child maltreatment: An ecological integration. American P~vehologist, 35, 320 335. Bograd, M. (1990). Why we need gender to understand human violence. Journal qf b~terpersomd Violence. 5, 32 135. Bowker. L., Arbitel. M., & McFerron, J. (1988). On the relationship between wife beating and child abuse. In K. YIIo & M. Bograd (Eds.), Feminist perspectives on wi/i" abuse (pp. 29-39). Newbury Park. CA: Sage. Burgess, R. L (1979). Project interact: A study of patterns of interaction in abusive, neglectful and control families. Chiht Abuse & Neglect, 3, 781-791. Cicchetti, D., & Rizley, R. (1981). Developmental perspectives on the etiology, intergenerational transmission, and sequela of child maltreatment. New Directions./or Child Development, 11, 31-55. Crittenden, P. M. (1993). An information-processing perspective on the behaxior of neglectful parents. Criminal Justice Behavior, 20. 27-48. DeMaris. A. (1992). Logit modeling. Newbury Park. CA: Sage. 1092 C. Coohey and N. Braun Farrington, K. (1980). Stress and family violence. In M. A. Straus & G. Hotaling (Eds.), Social causes (~/'husband-w~/i~ violence (pp. 94-114). Minneapolis, MN: University of Minnesota Press. Garbarino, J. (1977). The human ecology of child maltreatment.

Journal (~fMarriage and the Family, 39, 721-736. Garbarino, J. (1982).

Children and families in the social environment.

New York: Aldine. Gayford, J. J. (1975). Wife battering: A preliminary survey of 100 cases.

British Medical Journal, 25, 194-197. Gelardo, M. S., & Sanford, E. E. (1987). Child abuse and neglect: A review of the literature.

School Psychology Review, 16, 137-155. Gelles, R. J. (1973). Child abuse and psychopathology: A sociological critique and retkwmulation.

American Journal ~?/' Orthopsychiatry, 43, 611- 621. Gelles, R. J. (1983). An exchange/social control theory. In D. Finkelhor, R. J. Gelles, G. T. Hotaling, & M. A. Straus (Eds.), The dark side of families: Current family violence research (pp. 151-165). Beverly Hills, CA: Sage. Goode, W. J. (1971). Force and violence in the family.

Journal (~\]'Marriage and the Family, 33, 624 636. Hillson, J. M. C., & Kuiper, N. A. (1994). A stress and coping model of child maltreatment.

Child Psychology Review, 14, 261-285. Holmes, T. H., & Rahe, R. H. (1967). The social readjustment scale.

Journal ~/' Psychosomatic Research, 11, 213-218. Justice, B., & Justice, R. (1976).

The abusing family.

New York: Human Sciences. Kanner, A. D., Coyne, J. C., Schafer, C., & Lazarus, R. S. (1981). Comparison of two modes of stress measurement: Daily hassles and uplifts versus major life events.

Journal of Behuviorul Medicine, 4, 1 39. Laireiter, A., & Baumann, U. (1992). Network structures and support functions: Theoretical and empirical analysis. In H. Veiel & U. Baumann (Eds.), The meaning and measurement r~fsocial support (pp. 33-55). New York: Hemisphere. Lesnik-Oberstein, M., Cohen, L., & Koers, A. J. (1982). Research in the Netherlands on a theory of child abuse: A preliminary report.

Child Abuse & Neglect, 6, 199-206. McKay, M. M. (1994). The link between domestic violence and child abuse: Assessment and treatment considerations.

Child Welfare, 73, 29-39. Milner, J. S. (1993). Social information processing and physical child abuse.

Clinical Psychology Review, 13, 275-294. O'Keefe, M. (1995). Predictors of child abuse in maritally violent families.

Journal (~/'h~terpersonal Violence, 10, 3-25. Parke, R. D. (1982). Child abuse: An overview of alternative theoretical models.

Journal ~\[ Pediatric Psychology, 3, 9-13.

Patton, M. Q. (1990).

Qualitative evaluation and research methods.

Newbury Park, CA: Sage. Petersen, T. (1985). A comment on presenting results from logit and probit models.

American Sociological Review, 50, 130-131. Sameroff, A. J., & Chandler, M. J. (1975). Reproductive risk and the continuum of caretaking casualty. In F. D. Horowitz, E. M. Hetherington, S. Scarr-Salapatek, & G. Siegel (Eds.), Review (~\['child development research (Vol. 4) (pp. 187-244). Chicago, IL: University of Chicago Press. Schlesselman, J. J. (1982).

Case-control studies: Design, conduct, analysis.

New York: Oxford University Press. Seltzer, J. A., & Kalmuss, D. (1988). Socialization and stress explanations for spouse abuse.

Social Force.~, 67, 473-491. Stark, E., & Flitcraft, A. (1988). Women and children at risk: A feminist perspective on child abuse, hTternational Journal of Health Sciences, 18, 97-118. Straus, M. A. (1973). A general systems theory approach to a theory of violence between family members.

Social Science Information, 12, 105-125. Straus, M. A. (1988).

Measuring psychological and physical abuse of children with the Col!¢lict Tactics Scales.

Durham, NH: University of New Hampshire, Family Research Laboratory. Straus, M. A., & Gelles, R. J. (1988). How violent are American families'? Estimates from the National Family Violence Resurvey and other studies. In B. Hotaling, D. Finkelhor, J. Kirkpatrick, & M. A. Straus (Eds.), Family abuse and its consequences: New directions,\[or research (pp. 14-36). Newbury Park, CA: Sage. Stringer, S. A., & LaGreca, A. M. (1985). Correlates of child abuse potential.

Journal r~/'Abnormal Child Psychology, 13, 217-226. Tzeng, O. C., Jackson, J. W., & Karlson, H. C. (1991).

Theories (~\]'child abuse and neglect.

New York: Praeger. Walters, D. R. (1975).

Physical and sexual abuse o\[ children: Causes and treatment.

Bloomington, IN: Indiana University Press. Wellman, B. (1991).

Which type of ties and networks provide what kinds ~/'social support?

Toronto, Canada: University of Toronto, Center for Urban and Community Studies. Zimrin, H. (1984). Child abuse: A dynamic process of encounter between needs and personality traits within the family.

American Journal of Family Therapy, 12, 37-47. RESUMI~ Objectif:

Cet article prEsente un schcme conceptuel pour expliquer les mauvais traitements physiques, s'appuyant sur une analyse de diverses th6ories. On a postul6 trois facteurs pouvant affecter la probabilit6 de mauvais traitements: avoir 6t6 expos6 h une agressivit~ quelconque, etre expos6 h des conditions stressantes et avoir acc~s h des ressources. Pour chacun des trois groupes de facteurs, les auteurs ont identifi6 des variables. M~thode: On s'est bas6 sur des donn6es comparatives d'6tudes portant sur 81 m6res abusives et 148 m6res non abusives et on a identifi6 l'importance relative des facteurs pouvant pr6dire les mauvais traitements.

R6sultats:

Tel que pr~vu, la probabilit6 de comportements abusifs variait selon que la m6re avait b6n6fici6 d'appuis Integrated framework for child physical abuse 1093 psychologiques (l'ecoute, l'aide dans la prise de decisions, avoir de la compagnie), qu'elle vivait des situations stressantes ou qu'elle avait elle-m~me 6t6 maltraitee par son conjoint ou des conjoints anterieurs ou par sa propre mere durant son enfance. On a conclu que les facteurs les plus decisifs pour predire I'agression de la mere sont ceux de la violence maternelle en enfance ou celle du ou des conjoints. Conclusions: Les resultats proposent des recherches plus poussees pour mieux comprendre les liens reciproques entre la violence en enfance et la violence par son conjoint. RESUMEN Objetivo: Este articulo construy6 un marco tedrico para comprender el abuso ffsico contra los nifios basado en un antilisis de contenido de diferentes teorias. Se determinaron tres determinantes principales del abuso ffsico: Exposici6n a la agresion, exposici6n a estresantes, y acceso a los recursos. Para poner a prueba el marco te6rico, en el anfilisis inclufinos variables de las tres principales categorfas. Methodologia: Se utilizaron los datos de un estudio comparatiw) que inclufa 81 madres que abusaban ffsicamente y 148 madres no abusivas. La importancia relativa de los predictores se determin6 por regresi6n logfstica.

Resultados:

La probabilidad del abuso ffsico contra los nifios variaba en la direcci6n esperada con el nfimero de recursos emocionales recibidos (escuchar, ayuda en la toma de decisiones, compafiia), el nt~mero de estresores experimentados, y la exposici6n de la madre al abuso ffsico de sus propias madres, parejas actuales, y pasadas parejas. Los resultados estimados demostraron que la exposici6n a la agresi6n en su propia nifiez (abusado por su propia madre) y exposicion a la violencia domdstica en la vida adulta (abusado por la actual pareja o por la anterior) fueron los factores naris potentes para predecir si una madre abusaba ffsicamente de su hijo. Conclusidn: Los resultados sugieren realizar mils invesfigaciones sobre las interdependencias entre la violencia domdstica y el abuso ffsico contra los nifios. APPENDIX Statistical Model Following our conceptual framework, the statistical model for the probability of child physical abuse is written: q=eL/(1 +e L) with L=b~ +bI, E+bMM+bi, P~ bvX+bsS where q = Probability that the dependent variable takes on the value I (0 = not physically abusive and I = physically abusive), e = Base of natural logarithm, L ---- Logit of the dependent variable (i.e., the logarithm of the ratio between the probability of physically abusive behavior and nonphysically abusive behaviorL E = "Number of emotional resources received during the last 30 days." M = Dichotomous variable "physically abused by own mothers" with values 0 (no severe assault by own mothers) and 1 (at least one severe assault by own mothers), p = Dichotomous variable "physically abused by current partner" with values 0 (no severe assault by current partner) and 1 (at least one severe assault by current partner), X = Dichotomous variable "physically abused by ex-partner" with values (1 (no severe assault by ex-partner) and 1 (at least one severe assault by ex-partner), and S ---- "Number of stressors or life events experienced" within the last year that are not directly related to a partner.

The coefficients b,,, br, bM, etc., are the parameters to be estimated. An iterative maximum likelihood estimation procedure yields the results reported in Table 3.

Procedure jbr Determining Reference Case Using Petersen's (19851 method for the calculation of probability effects in logit models with binary predictors, we calculate first a baseline or reference probability for child physical abuse by combining the parameter estimates fiom Table 3 with the values of the independent variables typical for the average mother (Table 4). Table 5 shows that the reference case is a mother with the average emotional resources received and the average number of stressors who was severely assaulted by her mother, but was not severely assaulted by her current partner or her ex-partner. Beginning from the reference case, we calculate the probability of child physical abuse for each independent variable separately, holding all other variables at their average value. More precisely, we switch the value of any binary variable (M, P, and X) from its modal category to the alternative category and increase the mean of each metric variable (E and S) by exactly one standard deviation. The probability effect of each independent variable, then, is the difference between the probability of a mother abusing her child and the baseline probability of the average or reference mother abusing her child. 1094 Table 5. C. Coohey and N. Braun Distributions of Dichotomous and Continuous Independent Variables Variable Frequency Percent Mean Standard Deviation Assault by Own Mother No severe assault 60 27.6 -- Severe assault 157 72.4 -- Assault by Partner No severe assault 187 83. I -- -- Severe assault 38 16.9 -- Assault by Ex-Partner No severe assault 131 62.7 -- Severe assault 78 37.3 -- Emotional Resources Received 229 -- 14.326 7.698 Exposure to Stressors 225 -- 2.300 1.470