Amer. J. Orthopsychiat., 56(2), April 1986, 253-262.

Returning Children Home: Clinical Decision Making in Cases of Child Abuse and Neglect

Mitchell H. Katz, B. A., Robert L. Hampton, Ph.D., Eli H. Newberger, M.D., Roy T. Bowles, Ph.D., Jane C. Snyder, Ph.D.

Family Development Program, Children’s Hospital Medical Center; Department of Pediatrics, Harvard Medical School, Boston

Factors that influence the decision to remove children from their parents’ care ill cases of abuse and neglect were examined by reviewing hospital records of 185 children. Children with physical injuries were more likely to be placed in a foster home or in residential care if they were from poor families, while those with non-physical injuries were more likely to be removed if their families were more affluent. Implications for clinical decision making are considered.

Cases of child abuse and neglect confront clinicians with the difficult practical and ethical dilemma of whether to initiate action to remove children from their parents’ care. On the one hand, the doctrine to “above all else, do no harm” dictates that they be wary of separating children from their families and engendering emotional trauma. On the other hand, the commitment to protect children from harm precludes returning children thought to be in danger of further injury.

The dilemma is complicated by several other factors. In some cases the child may have sustained no traumatic injury, but his or her condition (e.g., severe neglect) may be a cause for concern. Where an injury has been sustained there is often a lack of clear information about how the injury occurred. Rational clinical judgment based on the child’s condition and family circumstances may be difficult for clinicians because of an absence of systematically assembled data, anger and other emotions they may feel toward unprotecting parents, psychological denial, and culturally conditioned impressions of certain types of families.

Concern about the quality of alternative living arrangements for children removed from their homes further clouds decision making. Recent examinations of foster care have revealed the inadequacies, failures, and high costs of the system.5,10,24 Although one might hope that foster care would provide a satisfactory temporary shelter while the family receives treatment, statistics suggest a different situation. The average length of time spent in foster care has been found to be about five years in some cities.4 Moreover, more than half the children in foster care are moved to a new foster home at least once and are thus deprived of stable, continuous care-giving.15

One way of allowing more children to return home is to provide troubled families with services that will strengthen the family unit. These may include traditional social services such as homemaker services, day care, or counseling. They may also include less official but equally important advocacy services such as working with landlords or welfare agencies on problems such as inadequate housing.17 Providing support services may be a particularly appropriate alternative to removal in cases where the child’s condition is not serious and the child does not appear at great risk. Indeed, several investigators have suggested that many children enter foster care due to problems such as family crisis or inadequate financial resources that could be better addressed by the provision of services.7,8,14,21 Unfortunately such alternatives often are not considered or are unavailable.14,18

Despite the importance of the decision as to whether to remove a child, there has been little research on how such decisions are made. A growing body of research suggests that institutional biases affect decision making in child abuse cases. O’Toole and colleagues20 found that physicians’ judgments about whether abuse occurred in a set of emergency room vignettes were affected by the race and socioeconomic status of the family, as well as by level of injury. In a similar study, McPherson and Garcia16 found that lack of familiarity with a family, but not low socioeconomic status, increased the likelihood of pediatricians diagnosing child abuse. Since minority and poor families are more likely to use emergency room facilities for their children’s health care, and thus be unfamiliar to the physician, their findings are consistent with a greater proportion of poor families losing their children.

A study conducted by Hampton and Newberger11 uncovered bias in the management of child abuse. In a national sample of suspected abuse and neglect cases, they found that hospitals tend to under-report white families to child protection agencies. Ninety-one percent of Hispanic and 74% of black families were reported to the child protection agency, while only 61% of white families were reported.

Studies of how child abuse is handled outside the medical system have also found that class, race, and other family characteristics affect decision making. Ross and Katz,22 in a retrospective study of case records of a protective service agency, found that even after controlling for the nature of abuse, families receiving general welfare, families perceived by the agency as having a family member with a mental health problem, families with a child with behavioral problems, and families with a parent characterized as ineffectual were more likely to have a child removed. Runyan and colleagues23 found that family characteristics such as substance abuse or an employed mother increased the likelihood of foster care placement, but that race and income were not significant predictors of placement. Studies of court decisions have revealed that factors such as substance abuse and the involvement of police may increase the likelihood that parents will lose their children in a court proceeding.1,25 Thus a similar set of premises and biases that might lead professionals to identify and report certain families as abusers at an initial setting, such as a hospital, might subsequently influence decisions made about the fate of the family as it is channeled through the protective service and justice systems.

The current study examined the records of 185 suspected abused or neglected children seen at Children’s Hospital, Boston, in an attempt to reveal how demographic characteristics, family history, family stress, the nature of the injury, and aspects of the medical encounter influence the outcome of the case.


A coding instrument was developed and pretested to provide a standardized approach for reviewing the hospital records of children referred to the hospital’s interdisciplinary child abuse consultation team, the Trauma X team. Data on the child, family, medical condition, and discharge disposition were obtained for each case. Each record was independently reviewed by two coders. Differences between coders were discussed in team meetings and records were subsequently recoded to reflect the consensus of the study group.

Two hundred and eighty cases referred to the Trauma X team over a three- year period (1978-1981) were coded. This represented approximately 80% of the cases about which the Trauma X team was consulted and which fell within the official team review. The remaining cases were not coded because records were either incomplete or unavailable. Cases about which a Trauma X team member was consulted but which were not officially reviewed by the team were not coded because reliable data could not be obtained. (These tended to be emergency cases or unofficial consultations where all the decisions were made before the Trauma X team could meet.)

To be included in this analysis the child had to be living at home before hospitalization. Obviously, discharging a child to a foster home has a different meaning if the child was initially living in a foster home. In addition, only cases in which the child had sustained a physical injury (excluding animal bites) or in which there was a suspicion of neglect, failure-to-thrive, or poisoning were included in the analysis. Excluded from the analysis were cases in which the child had suffered a fatal injury or had been discharged to another hospital. (In this latter group of cases there was no analogous custody disposition.) These selection criteria resulted in a sample of 185 cases.

Family stress was measured using a checklist based in part on The Social Readjustment Rating Scale.12 Stresses included on the checklist were death of a spouse, divorce, marital separation, jail term of parent or other household member, death of a close family member, pregnancy, death of a close friend, son or daughter leaving home, a handicapped child, unemployment, violence between spouses, and other significant stresses (up to two) present within a year of the current hospitalization. Items were added and grouped into two categories: families with no or only one stress (low stress) and families with more than one stress (high stress).

Two variables were used to characterize the nature of the injury. Severity of condition refers to the harm to the child’s body or function. On a four-point scale the injuries were coded as: 1) life threatening, death imminent without medical intervention; 2) serious, death unlikely but further deterioration of function highly probable without medical intervention; 3) moderate, death or deterioration of function unlikely but the condition serious enough to interfere with usual function and treatment of some type necessary to hasten reversal of the injurious process; 4) minimal, possibility of slight loss of function, injury can resolve with or without medical intervention. Physical injury was coded as present if the child had an old or current physical injury (e.g., burn, laceration) or was suspected of being a victim of physical or sexual abuse. Cases of neglect, poisoning, and failure-to-thrive were coded as non-physical injury. (Cases of physical injury with a nonphysical injury also present were coded as having a physical injury.)

Three case outcomes were distinguished: child returned home without services; child returned home with services; and child placed in foster home or institution. In cases in which the child was returned home (N=37) the decision was made that the family did not require further systematic intervention. These children were in most cases receiving medical follow-up through the hospital or other clinic, however. In cases in which the child was sent home with services (N=110), services were provided on an ongoing basis; services included day care, homemaker, visiting nurse, therapy for child or family, or protective service involvement. Out of home placements (N=38) included foster care with relatives or with non-relatives, and residential placement. Some of the children in this group will subsequently be returned home, but even temporary removal may have a major impact on children and parents. Young children, in particular, may have difficulty understanding a “temporary” removal. For parents, a removal can be equally traumatic and undermine their sense of competence in caring for their children.

In many cases, actions by the protective service agency and the courts influence hospital management and discharge disposition. However, the opinions of professional reporters such as medical personnel weigh heavily in the actions taken by protective service agencies2 and the court. Moreover, to the extent that hospital clinicians make initial decisions regarding whether to file a report with the protective service agency or whether to seek court approval for custody, they choose the cast of other professionals involved in the case. It should also be noted that some of the placements of children outside the home may be “voluntary” by the parents.


Sample Characteristics

Children ranged in age from one month to 16 years, with the sample skewed toward younger children (median=16.3 months). Forty percent of the children were less than one year old and 70.8% were under three years of age. There were approximately equal numbers of girls (51.9%) and boys (48.1%). The sample is composed predominately of lower-income families. Almost two-thirds of the families (65.4%) were eligible for Medicaid. Forty-two percent of the families were white, 36.4% black, 17.9% Hispanic, and 3.2% other (N=184). (The N is reported for all results where N is not equal to 185 due to missing data or a selected sample.) Over half of the families (52.8%) were female-headed households (N=178). The mean age of mothers was 26.3 (SD=7.2; N=168) and of fathers 29.7 (SD=8.3;N=112). Families had, on average, 2.3 children (SD= 1.3) with 83.6% of families having no more than three children (N=183).

For a large number of families this was not their first contact with an agency due to concern about child maltreatment. Over one-fourth of the families (29.2%) were the subjects of reports to child protective services agencies either for the index child or for someone else in the family. Nineteen percent of children had a record of a previous accident (N=182).

Commonly noted stresses included unemployment (47.0%), pregnancy (44.9%), and marital separation (23.8%). Seventy-one percent of families had more than one stress noted. This finding is in keeping with other research indicating a close association between family stress and child accidents and maltreatment.3,6,13

Eight percent of the conditions were rated as life-threatening, 35.9% as serious, 47.8% as moderate, and 8.2% as minimal (N=184). These figures indicate that, while a substantial number of children sustain severe injury, the majority of cases are not severe. The wide range in severity of injury underscores the importance of having flexible responses to maltreatment cases.

In 73.5% of the cases, the child suffered a physical injury. The mother was suspected of maltreating or permitting maltreatment of the child in 44.3% of the cases in which a mother was present in the household (N=183). In families in which a father was present, the father was suspected of being involved in the maltreatment in 40.7% of the cases (N=86).

Fourty-four percent of cases were seen through the medical emergency room and 39.2% through the surgical emergency room (N=176). In almost all of the cases (97.3%), the child was admitted to the hospital; however, five cases (2.7%) were sent home after evaluation (N=184). Most of the children were cared for by a medical service (45.4%) or a surgical service (including neurosurgery, plastic surgery, and orthopedics) (50.2%), while the remaining patients were seen by the psychosomatic, outpatient, or other hospital clinic or service. The duration of stay for inpatients ranged from one to 169 days, with a median stay of 9.1 days (N=179).


We tested the influence of each independent variable on the outcome measure using x2 analyses. Results are presented in TABLE 1. Families who were Medicaid-eligible and those with a previous report of suspected child maltreatment were more likely to have their children removed. Minority families were not, however, more likely to lose their children.

Table 1
Medicaid Eligibility      
Yes (65.4%)
No (34.6%)
x2=6.31; df=2; p=.04; (N=185)      
Previous Filing      
Yes (29.2%)
No (70.8%)
x2=20.7; df=2; p<.0001; (N=185)      
White (43.8%)
Black (37.6%)
Hispanic (18.5%)
x2=3.88; df=4; p=.42; (N=178)      
Severity of Condition      
Life-threatening (8.2%)
Serious (35.9%)
Moderate (47.8%)
Minimal (8.2%)
x2=4.05; df=6; p=.67; (N=184)      
Mother Involved in Maltreatment      
Yes (44.3%)
No (55.7%)
x2=9.51; df=2; p=.009; (N=183)      
Father Involved in Maltreatment      
Yes (40.7%)
No (59.3%)
x2=3.87; df=2; p=.14; (N=86)      
Physical Injury:      
Physical injury (73.5%)
Non-physical inj. (26.5%)
x2=6.49; df=2; p=.04; (N=185)      
Hospital Service      
Medical (47.5%)
Surgical (52.5%)
x2=7.81; df=2; p=.02; (N=177)      
Emergency Room      
Medical ER (44.3%)
Surgical ER (39.2%)
No (16.5%)
x2=10.2; df=4; p=.04; (N=176)      
Family Stress      
Low stress (29.2%)
High stress (70.8%)
x2=9.78; df=2; p=.008; (N=185)      
Child’s Age      
Under six (83.8%)
Six or over (16.2%)
x2=8.99; df=2; p=.01; (N=185)      
Previous Accident      
Yes (19.2%)
No (80.8%)
x2=8.83; df=2; p=.01; (N=182)      
1 Child’s sex, mother’s age, father’s mage, single mother, others present at home, and number of siblings at home were not significantly associated with discharge disposition.

Severity of condition was not significantly associated with outcome. Cases in which the mother was suspected of being involved in the maltreatment of the child were more likely to result in removal. (There was a similar trend when the father was suspected of being involved in maltreatment but the sample was small and the result was not statistically significant.)

The presence of a physical injury decreased the likelihood of a child being placed outside of the home. Over two-thirds of cases of physical injury (67.7%) were seen on the surgical service and 53.1 % were seen in the surgical emergency room. It is not surprising, therefore, that cases involved with a surgical service, as well as those involved with the surgical emergency room, were also less likely to have a child removed.

Families in which the child was under the age of six and families that were experiencing more than one stress at the time of hospitalization were more likely to have their child sent home with services. Cases in which there was a history of a previous accident (which some clinicians might interpret as a possible abuse incident or sign that the family is under stress) were less likely to be sent home with services (more likely to be sent home without services or removed from their homes).

A severe limitation of bivariate analysis is that it does not tell us whether the predictor variables are independently associated with the dependent variable. To resolve this problem we employed log-linear analysis, which is a multivariate procedure for analyzing categorical data. We tested the influence of each key independent variable on the dependent variable, controlling for those independent variables that were significantly associated with the same discharge disposition in bivariate analyses as the key independent variable in question.

We found that the effect of class is not independent of a history of a previous report of child maltreatment, service type, or a mother’s role in maltreatment. This finding does not mean that class is not important. Rather, it appears that more than a single effect is operating. There is a group of poor families that are more likely to have a history of a previous report (x2=5.96; df=1;p=.01), be seen by the medical service (x2=7.16; df=1;p=.007), and have the mother involved in the maltreatment (x2=10.6; df=1;p=.001). Since these variables are strongly associated with poverty it is impossible to determine precisely why this group of cases is at risk for removal.

While we were unable to demonstrate an independent effect of class on discharge disposition, we found that there was a significant three-way interaction among class, presence of a physical injury, and discharge disposition (TABLE 2). Specifically, families that were Medicaid-eligible were more likely to have their child removed than were more affluent families in cases of physical injury and less likely to have their child removed in cases of non-physical injury.

Table 2
Non-physical Injury      
Medicaid eligible: Yes
Medicaid eligible: No
Physical Injury      
Medicaid eligible: Yes
Medicaid eligible: No
G2=7.84*; df=2; p=.02; (N=185)      
Non-physical Injury      
Low stress
High stress
Physical Injury      
Low stress
High stress
G2=11.5*; df=2; p=.003; (N=185)      
1 Residuals should be interpreted as deviations from the bivariate results (see Table 1). Positive numbers indicate a greater number of cases than expected in the cell, while negative numbers indicate fewer than expected cases in the cell.* Results verified by random sampling.

The history of a previous child abuse report was a significant determinant of placement outside the home, even after controlling for other independent variables. In terms of those variables found to be associated with sending children home with special services, we found that the statistical effects of high stress and having preschool children are redundant. That is, either of the variables predicts to outcome, but neither variable has a unique statistical association with discharge disposition. This is consistent with the finding that families in which the child was of preschool age were under greater stress (x2=4.33; df=1; p=.04). There is a significant three-way interaction among family stress, physical injury, and discharge disposition (TABLE 2). While high stress increased the likelihood of going home with services in cases of physical injury, high stress increased the likelihood of removal in cases of non-physical injury. Also, there is a marginally significant three-way interaction among family stress, previous report of child abuse, and discharge disposition (G2=5.59; df=2; p=.06), indicating that in low stress families a previous report increases the likelihood of the child being returned home with services instead of home without services. High stress marginally increases the likelihood of a child returning home with services even after controlling for a history of a previous accident (G2=8.21; df=4; p=.08). Conversely, a history of a previous accident marginally decreases the likelihood of a child being returned home with services after controlling for family stress (G2=8.07; df=4; p=.09). A three-way interaction among a previous accident, child’s age, and discharge disposition (G2=10.84; df=2; p=.004) indicates that preschool children with a previous accident are more likely to be removed, while older children with a previous accident are less likely to be removed.


The data help us to understand those factors that do and do not influence discharge disposition of cases of abuse and neglect. We found that children with non-physical injuries were more likely to be removed. One explanation for this result is that non-physical injuries, which include failure-to-thrive and neglect, may be perceived by clinicians as evidence of chronic family problems rather than as a single mishap. Second, the decision to admit a child who does not have a physical injury (and therefore has more limited treatment possibilities) may itself indicate consideration of removal. A third possibility is that clinicians on the surgical services (which see the majority of children with physical injuries) are more likely to send children home after treatment than are clinicians on the medical services (which see the majority of non-physical injuries).

While somewhat surprising, the fact that severity of condition was not associated with placement outside of the home is consistent with the findings of Hampton and Newberger.11 It may be that, in considering whether to place children after hospitalization, other factors such as the perceived risk of reinjury to the child weigh more heavily in decision making. Alternatively, our scale of severity of injury may not be a sensitive measure. One important factor our scale does not take into account is the injurer’s intention. A sharp object thrown at a young child may result in very different injuries depending on whether it just hits or just misses the child’s eye. Yet the resulting injuries may be viewed similarly by clinicians who focus on the injurer’s intention.

Social class was not found to have an independent effect on discharge disposition in the sample as a whole. However, low-income families were more likely to lose their children in cases of physical injury. With physical injuries to young children it is difficult to establish whether the injury was inflicted or accidental. Indeed, several studies have suggested that accidents and abuse may have similar etiologies.9,19 In clinical practice, however, physical injuries are open to two very different interpretations: abuse or accident. Our findings suggest that physical injuries may more frequently be diagnosed as “abuse” in poor families and more frequently characterized as “accidents” in more affluent families. The fact that more affluent families are more likely to lose their children in cases of non-physical injury suggests that a negative evaluation is made of families who appear to neglect their children despite adequate financial resources.

Although, overall, clinicians took note of families that were under stress and provided them with services to maintain their integrity, multivariate analysis showed that this relationship was not true of cases of non-physical injury in which high stress made removal more likely. One possible explanation for this finding is that families with chronic conditions, such as failure-to-thrive or neglect, and high stress are perceived as too overwhelmed to care for their child even with services. More intensive services, in addition to traditional protective service casework, such as freely available day care and homemakers, might allow more children to return home from the hospital.

A history of a previous child abuse report was an important determinant of placement outside the home. It may be that clinicians view these families as “repeat offenders,” unable to protect their children. This finding should, however, remind us of the inherent dangers of labeling families as “child abusers” as occurs when there is a note in the child’s medical record stating that the family was previously reported for abuse. Such a note may make further referrals to child protection agencies, as well as removals, more likely even when past reports may be unsubstantiated.

Two significant limitations of this study should be noted. First, it is an entirely hospital-based sample. Nonetheless, hospitalized children may be particularly at risk for removal because their injuries are generally more serious. Also, this sample reflects only those cases in which the Trauma X team was consulted, rather than all children entering the hospital who may have been abused or neglected. There was an initial” screening” of these cases before they reached the Trauma X team. This study cannot reveal the factors influencing the initial recognition of child abuse and neglect. While no bias against minority children emerged in this study, as it has in previous research,11 white and minority children with similar conditions may not be referred to the Trauma X team in similar proportions.


Based on the findings of this study, as well as on the clinical experience of the child abuse team, we offer the following four recommendations.

1. Formalize decision making. Emotional reactions to cases are much more likely to affect case management if decisions are made by a single clinician on an ad-hoc basis. This is true regardless of the talents or sensitivities of any individual clinician. A multidisciplinary group (including pediatrician, psychiatrist, psychologist, social worker, nurse, and lawyer) offers an opportunity for involved clinicians to organize their observations about a particular case and receive feedback from a variety of perspectives.

2. Include members of class and racial minorities in all decision making groups. One of the best ways to avoid bias in decision making is to insure that there are members of the group who will represent the position of poor and minority families and who are especially sensitive to cultural differences in child rearing and family structure.

3. Establish systematic linkages with social agencies. The decisions made by hospital clinicians frequently require the support of other agencies, and particularly the state child protection agency. Close ties with social agencies will insure the best service for the child.

4. Act as advocates. Clinicians must recognize and accept the important role they can playas advocates for their patients, both on a case-by-case and a community-wide basis. Clinicians must help families to obtain needed services so that their children can be safely returned home. Moreover, clinicians must also be advocates for governmental provision of services to troubled families so as to insure that clinical judgments are not determined by the scarcity of services.


1. ABER, J. 1980. The involuntary child placement decision: Solomon’s Dilemma revisited. In Child Abuse: An Agenda for Action, G. Gerbner, C. Ross and E. Zigler, eds. Oxford University Press, New York.

2. CARR, A. AND GELLES, R. 1978. Reporting Child Maltreatment in Florida: The Operation of Public Child Protective Service Systems. Report submitted to the National Center on Child Abuse and Neglect.

3. DANIEL, J., HAMPTON, R. AND NEWBERGER, E. 1983. Child abuse and accidents in black families: a controlled comparative study. Amer. J. Orthopsychiat. 53:645-653.

4. FANSHEL, D. 1981. Decision-making under uncertainty: foster care for abused or neglected children? Amer. J. Pub. Hlth 71:685-686.

5. FANSHEL, D. AND SHINN, E. 1978. Children in Foster Care: A Longitudinal Investigation. Columbia University Press, New York.

6. OIL, D. 1970. Violence Against Children. Harvard University Press, Cambridge, Mass.

7. GOLDSTEIN, J., FREUD, A. AND SOLNIT, A 1973. Before the Best Interests of the Child. Free Press, New York.

8. GOLDSTEIN, J., FREUD, A. AND SOLNIT, A. 1979. Beyond the Best Interests of the Child. Free Press, New York.

9. GREGG, G. AND ELMER, E. 1969. Infant injuries: accident or abuse? Pediatrics 44:434-439.

10. GRUBER, A. 1978. Children in Foster Care. Human Sciences Press, New York.

11. HAMPTON, R. AND NEWBERGER, E. 1985. Child abuse incidence and reporting by hospitals: significance of severity, class and race. Amer. J. Pub. Hlth 75:56-60.

12. HOLMES, T. AND RAHE, R. 1967. The social readjustment rating scale. J. Psychosomat. Res. 11:213-218.

13. HOLTER, J. AND FRIEDMAN, S. 1968. Child abuse: early case finding in the emergency department. Pediatrics 42:128-138.

14. KENISTON, K. 1978. All Our Children. Harcourt, Brace, Jovanovich, New York.

15. KNITZER, J. AND ALLEN, M. 1973. Children Without Homes: An Examination of Public Responsibility to Children in Out-of-Home Care. Children’s Defense Fund, Washington. D.C.

16. MC PHERSON, K. AND GARCIA, L. 1983. Effects of social class and familiarity on pediatricians’ responses to child abuse. Child Welfare 62:387-393.

17. MORSE N. ET AL. 1977. Environmental correlates of pediatric social illness: preventive implications of an advocacy approach. Amer. J. Pub. Hlth 67:612-615.

18. NEWBERGER, E. AND DANIEL, J. 1976. Knowledge and epidemiology of child abuse: a critical review of concepts. Pediat. Ann. 5: 15-26.

19. NEWBERGER, E. ET AL. 1977. Pediatric social illness: toward an etiologic classification. Pediatrics 60:178-185.

20. O’TOOLE, R., TURBETT P. AND NALEPKA c. 1983. Theories, professional knowledge, and diagnosis of child abuse. In The Dark Side of Families: Current Family Violence Research, D. Finkhelhor et al., eds. Sage, Beverly Hills. Calif.

21. ROSS, C. 1980. The lessons of the past: defining and controlling child abuse in the United States. In Child Abuse: An Agenda for Action, G. Gerbner, C. Ross and E. Zigler, eds. Oxford University Press, New York.

22. ROSS, C. AND KATZ, M. 1983. Decision Making in a Child Protection Agency. Unpublished manuscript, Yale University.

23. RUNYAN, D. ET AL. 1981. Determinants of foster care placement for the maltreated child. Amer. J. Pub. Hlth 71:706-711.

24. SCHOR, E. 1982. The foster care system and health status of foster children. Pediatrics 69:521-528.

25. WEINBERGER, P. AND SMITH, P. 1970. The disposition of child neglect cases referred by caseworkers to a juvenile court. In Child Welfare Services, A Kadushin, ed. Macmillan, New York.

Submitted to the Journal in March 1985. Research was supported in part by grants from the National Center on Child Abuse and Neglect (90CA891), National Institute of Mental Health (TO1-MH155117-03-CD), Department of Health and Human Services, and Rockefeller Foundation Postdoctoral Fellowship Program (RLH).