To describe maternal discipline of children in at-risk families and to determine factors associated with disciplinary practices.
At-risk families in North Carolina followed up in a longitudinal study of child maltreatment.
Maternal caregivers of 7- to 9-year-old children with factors at birth that placed them at risk.
Measurements and Results
A total of 186 maternal caregivers were interviewed. A measure, based on coding parental responses, was used to assess disciplinary practices for 5 different misbehaviors. Limit setting was the most commonly used disciplinary practice for 4 of 5 misbehaviors, with 63% of mothers reporting that this method generally worked best. Spanking was more likely used as a secondary response for each misbehavior, when the primary one had not succeeded. Conversely, teaching or verbal assertion was always less likely as a secondary response. Teaching or verbal assertion was used more commonly for lying than for any other misbehavior, limit setting for disobeying, spanking for stealing, and spanking with an object for being disrespectful. Regression modeling for the 4 most common disciplinary practices showed (P<.05) that black race, lack of Aid to Families With Dependent Children receipt, more-educated mothers, and female sex of child were associated with higher use of teaching or verbal assertion; a biological father in the home was associated with less use of limit setting; and black race and report for child maltreatment were associated with more use of mild spanking.
In this sample, limit setting was the most common disciplinary technique. Disciplinary practices used varied depending on the type of misbehavior and other contextual factors, including child, parent, and family characteristics.
DISCIPLINARY PRACTICES can be important to outcomes such as child self-esteem,1 conscience development,1,2 aggression,1,3,4 behavior problems,5,6 delinquency3,6,7 and subsequent adult criminal behaviors,3 depression, and alcoholism.8 There are notable gaps in our understanding of issues related to discipline. The frequencies of disciplinary practices aside from corporal punishment are not well known, particularly for nonclinical populations of families and children. Measures of discipline have been inadequate to describe the breadth and depth of the construct.9 Little is known about what parents choose as secondary discipline strategies, when the first do not work. Although stress has been shown to be related to parenting practices including discipline,10,11 the disciplinary practices from population-based samples of at-risk families have not been well described.
Physical abuse often starts as discipline that gets out of hand,12,13 so it is plausible that physically abusive parents might have characteristic discipline styles, but this has not been studied extensively. Existing studies suggest that physically abusive families are more likely to use harsh corporal punishment,14 coercion,15 and a negative demeanor.16 Neglectful parents tend to be unresponsive and more negative to their children.16,17 Since some maltreatment occurs when initial attempts at discipline cross a line into psychological or physical abuse, we were interested in whether these at-risk parents typically resorted to more severe strategies for their secondary discipline techniques.
Factors that affect disciplinary practices and their outcomes include child, parent, family, community, and situational characteristics.9 The relationship of some of these factors with corporal punishment has been well described, but for the most part the relationship between contextual factors and other types of disciplinary practices has not been well studied.
In her extensive review of the relationship between gender and disciplinary practices, Steinmetz2 found that discipline depended on the sex of the child. Similarly, Straus and Donnelly4 showed that boys were hit more often, and more often subjected to verbal aggression, than girls.18 Child aggression has been examined both as a precursor to and an outcome of corporal punishment, and a positive association is found between aggression and corporal punishment in both cases.7,19- 23
Several studies have shown that older parents are less likely to use corporal punishment.4,24,25 Multiple studies4,23,24,26,27 have shown that being spanked as a child is one of the strongest predictors of whether one would spank as a parent. In addition, parental sex is found to be associated with disciplinary practice, in that mothers were found to spank more often than fathers,4 though this may be an effect of the greater amount of time that mothers spend with children, as compared with fathers. There is conflicting evidence about disciplinary styles of black families, but a fair amount of evidence suggests that they tend to be more power assertive, punitive, and arbitrary.28- 31 Elder et al10 and Patterson et al32 showed that stressful experiences increase psychologic distress in parents, which in turn increases aversive, coercive, and inconsistent discipline of children. McLoyd and coworkers33 found that unemployed mothers had increased depressive symptoms and that this predicted more frequent punishment of their adolescent children. McLoyd34 linked the disciplinary styles of blacks with the increased stressors they experience—economic and other.
There is not much research regarding parental partner status, education, and religiosity, and their relationships to disciplinary practices, although each plausibly might affect disciplinary practices.
Several studies indicate that lower socioeconomic status is associated with more corporal punishment, but in the research by Straus and Donnelly,4 once the age and ethnic group of the parents were taken into account, there was no longer an association. Several studies have shown larger (or more crowded) families to be more likely to use power-assertive methods.1,31,35
The child abuse literature shows a strong association between stress and lack of social support and physical abuse.13,36 Thus, the practice of extreme forms of corporal punishment might be related to lack of social support and/or higher stress levels.
In previous research, Socolar and Stein24,37 found that the beliefs of parents and physicians about spanking and other types of discipline depend on whether the discipline is in response to dangerous misbehaviors compared with annoying ones. Others38 have also found differences depending on the kind of transgression. It is likely that parents use different disciplinary strategies depending on the type of misbehavior, although this aspect of the context has largely been ignored to date.
In previous research about the relationship between disciplinary practices and contextual factors, only a very limited number of factors have been examined. In this study, we explored a range of contextual factors likely to be associated with disciplinary practices using models that allowed us to control for other variables. In addition, while there are several studies of disciplinary practices in small samples of neglectful and/or abusive families,13 research does not exist that focuses on discipline in larger or population-based samples of at-risk families. Thus, our goals were to examine the relationship of numerous contextual factors to disciplinary practices, to use a new measure of discipline that is more comprehensive and contextually rich, to describe primary and secondary disciplinary techniques, and to learn about disciplinary practices in a statewide sample of at-risk families.
Subjects included 186 maternal caregivers (hereafter referred to as mothers) of children aged 7 to 9 years participating in a longitudinal study of child maltreatment. This sample was drawn from a cohort of 788 mother-infant pairs enrolled in a prospective study39 of child abuse and neglect in 1985-1987. The initial study, "Stress, Social Support, and Abuse and Neglect in High Risk Infants,"39 included mother-infant dyads recruited from North Carolina hospitals in the immediate postpartum period. Eighty-five percent of the infants in that study were selected to be "at risk" for adverse social and medical outcomes. "At-risk" criteria included low birth weight (<2500 g), young maternal age (<18 years), and other major medical or social problems diagnosed at birth (including no prenatal care, significant birth defects, and congenital conditions or serious neonatal illness). The remaining 15% of the subjects in the cohort met none of the "at-risk" criteria. The mother-infant pairs in both groups were recruited from 37 geographically diverse counties in North Carolina. The majority of the mothers in these dyads were black and lived in low-income households.
By 1991, review of the Central Registry on Child Abuse and Neglect, a database maintained by the North Carolina Division of Social Services, revealed that one third of the 788 infants in the initial study had been reported to county child protective service agencies as maltreated. That year, a sample of the cohort was selected for inclusion in a new longitudinal, multisite study of the antecedents and outcomes of child abuse and neglect, the Longitudinal Studies Consortium on Child Abuse and Neglect (LONGSCAN).40 Funding limitations compelled the LONGSCAN/North Carolina site to include in its sample only a portion of the subjects enrolled in the original stress and social support study. These subjects were selected in order from separate, computer-generated randomized lists of reported and nonreported subjects. For each reported child selected, 2 nonreported children matched for sex, race, age, and poverty level were selected as controls. The final LONGSCAN sample includes 243 subjects, 83 of whom had been reported for maltreatment between 6 months of age and their first LONGSCAN interview.
As part of the LONGSCAN protocol, the North Carolina site collected data annually on all subjects. In-depth, face-to-face interviews were completed at age 5, 6, and 8 years. This article is based on data collected for age 8 interviews. Owing to attrition (refusals, lack of cooperation, and inability to locate), 186 of the 243 mothers were interviewed. Statistical comparisons between the original LONGSCAN sample and the age 8 subsample show no significant differences in sociodemographic, risk, or maltreatment variables.
Discipline has been defined in various ways.41- 44 A distinction has been made between reactive discipline, done after the misbehavior, and proactive discipline, done to promote good behavior.9 To narrow the scope of the study and use a definition that was readily operationalized, we defined discipline as the action a parent takes in response to a misbehavior. Thus, we focused only on reactive discipline.
Maternal disciplinary practices were assessed with a new measure designed to prompt respondents for primary and secondary responses usually used for each of 5 specific child behavior problems: disobedience, hitting a younger or smaller child, disrespect, lying, and stealing. Our approach was designed to orient respondents toward thinking about their discipline techniques as specific strategies implemented to address specific problems within the context of the past 6 months. We used this approach because we were interested in determining whether mothers use different strategies for different behavior problems and because we believed it would be easier for mothers to remember and candidly report their disciplinary practices when the context focused on their child's misbehavior. We were also interested in determining the practices mothers are likely to use as secondary strategies when the first are unsuccessful. Finally, we asked what discipline method works best for the index child most of the time. Responses were coded by interviewers to 1 of 13 disciplinary categories that were adapted in part from the work of Trickett and Susman14 and Webster-Stratton and Spitzer45: (1) nothing—avoids dealing with the problem; (2) ignoring—a planned strategy, as opposed to "nothing"; (3) tell someone else, or get someone else to discipline; (4) empathy; (5) teaching or verbal assertion (teaching/verbal assertion)—reasoning, giving explanations, giving alternatives, reminding the child of rules or expectations; (6) limit setting or logical consequences—time-out, withdrawal of privileges or threat of withdrawal of privileges, removing child from situation, making a child return a stolen object, making a child apologize, taking away allowance; (7) verbal or symbolic aggression; (8) guilt induction or embarrassment; (9) mild physical force—spanking with hand; (10) moderate physical force—hitting or spanking with an object, slapping, grabbing, jerking, shaking, throwing an object at child; (11) severe social isolation; (12) terrorizing; and (13) severe physical force or restraint. Interrater reliability was assessed by comparison of original codes with a 10% random sample coded blindly by the consortium coordinator (who supervises training and monitors data quality).
Data on child age and sex were obtained. Manageability was measured by asking mothers, "How easy or difficult would you say it is to manage her/his behavior? Would you say s/he is (1) easy, (2) average, (3) difficult?" (This variable was dichotomized into easy vs average or difficult for analysis.)
The mother's race, age, and education (highest grade completed in school, in years) were assessed. Her history of physical victimization had been assessed previously (1991-1994) in the longitudinal study ("When you were a child or teenager . . . were you ever physically hurt by a parent or someone else—like hit, slapped, beaten, shaken, burned, or anything like that? (spanking did not count); . . . were you ever punished or disciplined by someone in such a way that you were bruised or physically injured?"). Maternal depression was assessed using the depression subscale of the Brief Symptom Inventory.46 The partner status (biological father, other partner, or no partner) and relationship of the mother to the child (biological mother or other caregiver) were assessed with a family chart that ascertained who lived in the home and relationships to the child. Attendance at religious or spiritual services was assessed for the past year (by means of a 6-point Likert scale with responses ranging from "never" to "more than once a week").
Child maltreatment status was determined through a review of the North Carolina Central Registry of Child Abuse and Neglect to determine if there was any report (substantiated or unsubstantiated) of abuse or neglect of the index child. Income as percentage of poverty level was determined by asking about the total income from all sources after taxes and deductions (measured in $5000 increments), and the number of people in the household dependent on this income, and calculating the percentage of the federal poverty level for 1995. Receipt of Aid to Families With Dependent Children (AFDC) was determined in a question about sources of income for the household. Social support was measured with the Social Provisions Scale.47 Neighborhood support was assessed by summing the 4-point Likert responses to 3 questions (whether the neighborhood was one in which people help each other, watch out for each other's children, and can be depended on). The number of siblings in the household also was determined, including full, half, and step siblings).
Frequencies were determined for each item of the discipline measure. Reliability of the measure was assessed with κ statistics. Frequencies also were determined for the child, maternal, and family factors, including summative scales for maternal depression, social support, and neighborhood support.
There were 5 given misbehaviors. Because stealing was reported infrequently, it was not included in regression modeling, leaving 4 possible misbehaviors: disobedience, fighting, disrespect, and lying. A respondent was included in regression modeling if she reported that she had actually dealt with at least 3 of these 4 possible misbehaviors. Logistic regression modeling was performed for each disciplinary practice (that was reported at more than 5%) as the dependent variable, and the child, mother, and family factors listed above as the independent factors. This regression accounted for the correlation of observations within each subject. The resultant odds ratios are the odds of using the particular disciplinary practice for a given misbehavior. Main effects were entered in the model at α=.10, and were eliminated using stepwise elimination at α=.05. All possible 2-way interactions with significant main effects were tested and retained at α=.05.
In addition, we were interested in examining what each mother chose first as a disciplinary response to a misbehavior, ie, the primary response, and what mothers chose to do when the primary response did not work, ie, the secondary response. To examine factors associated with different primary responses relative to secondary responses, we divided the disciplinary techniques into 3 categories: (1) corporal punishment (codes 9 and higher); (2) teaching/verbal assertion or limit setting (codes 5 and 6); and (3) other (codes 1-4, 7, and 8). We then looked at 4 primary to secondary response groups (since the frequency for the "other" code was small): corporal punishment (primary) to corporal punishment (secondary) (CP to CP); corporal punishment to teaching/verbal assertion or limit setting (CP to T/LS); teaching/verbal assertion or limit setting to corporal punishment (T/LS to CP); and teaching/verbal assertion or limit setting to teaching/verbal assertion or limit-setting (T/LS to T/LS). The T/LS to CP response group starts with nonphysical techniques and moves on to physical means if the other methods do not change behavior. The CP to T/LS response group starts with physical methods and then backs off to less intense methods. When the primary and the secondary responses are the same (ie, CP to CP and T/LS to T/LS), this might be thought of as a gauge of the intensity of use of that type of discipline. Since the practice is not changing contingent on the child's behavior, it could potentially be a gauge of noncontingent disciplinary practices as well.
We examined the frequencies for each misbehavior for these 4 primary to secondary response groups when subjects reported them for at least 3 of the 4 given misbehaviors. We examined 4 regression models—one each with the 4 primary to secondary response groups as the dependent variable, and the same array of independent variables as above, but found 2 models not to be robust due to small frequencies. Thus, we report bivariate analysis for all independent variables found to be significant at α < .05, for the primary to secondary response groups. Dichotomous variables are reported using odds ratios, and continuous variables using means for 1 or more misbehaviors vs never used.
Data were analyzed using SAS software (SAS Institute, Cary, NC) on a personal computer. The study was approved by the University of North Carolina Institutional Review Board for the School of Public Health, and informed consent was obtained from all participants.
A total of 186 mothers were interviewed. For our sample, risk factors present at birth included no prenatal care (15%), congenital condition (21%), birth weight less than 2500 g (32%), and maternal age younger than 18 years (33%). Multiple demographic and social measures show these families to be disadvantaged (Table 1).
Of the 13 discipline codes, only 4 were reported as primary or secondary responses by more than 5% of respondents: teaching/verbal assertion, limit setting, mild physical force, and moderate physical force (Table 2). The agreement between coders was excellent, with an overall κ of 0.91 for questions about disciplinary practice (95% confidence interval, 0.87-0.96; P<.001).
Limit setting was the most commonly used disciplinary practice for 4 of 5 misbehaviors. Sixty-three percent of mothers reported that limit setting worked best with their children most of the time. Teaching/verbal assertion was used more commonly for lying than for any other misbehavior, limit setting for disobeying, mild physical force for stealing, and moderate physical force for being disrespectful. Mild physical force was more likely as a secondary response when the primary one had not succeeded for each misbehavior, whereas teaching/verbal assertion was always less likely as a secondary response (Table 3). Thirteen percent of respondents reported CP to CP for at least 1 misbehavior; 33% reported CP to T/LS; 58% reported T/LS to CP, and 73% reported T/LS to T/LS for at least 1 misbehavior.
Regression models were evaluated using the 4 most reported disciplinary practices as dependent variables and child, mother, and family characteristics as the independent variables (Table 4). Independent factors that were predictive of disciplinary technique in multiple models included race, AFDC receipt, maternal education, and report of maltreatment. Black race was associated with teaching/verbal assertion and mild physical force, and race was part of an interactive effect with maternal age for limit setting. Those receiving AFDC were less likely to use teaching/verbal assertion and more likely to use moderate or severe physical force. Mothers who were more educated reported significantly more use of teaching/verbal assertion and less use of moderate or severe physical force. Mothers who had been reported for maltreatment of the child were more likely to use mild physical punishment but less likely to use moderate or severe physical punishment. Limit setting was used less often in homes with biological fathers than in homes with no father present, whereas limit setting was used more with a nonbiological father in the home compared with homes with no father. Girls were disciplined with teaching/verbal assertion more often than boys. Mothers with a history of victimization reported that they used mild physical force less often than nonvictimized mothers.
Bivariate analysis of factors associated with the 4 primary to secondary response groups showed that the poorer families use corporal punishment as both a primary and a secondary response (Table 5). Factors associated with using corporal punishment first and teaching/verbal assertion or limit setting as a secondary response were black race, receipt of AFDC, the absence of the biological father, and less social support. There was an association between mothers with more education and more religious service attendance and use of teaching/verbal assertion or limit setting as a primary response and corporal punishment as a secondary response. The use of teaching/verbal assertion or limit setting as primary and secondary responses was more likely for female children. The use of corporal punishment as primary and secondary responses was more likely for poorer families.
Factors that we studied that were not associated with any disciplinary type or response group in the analyses performed included child manageability, maternal depression, relationship of the mother to the child (biological or other caregiver), and the number of siblings of the child.
In this sample, limit setting was the most common disciplinary technique reported by mothers. It would be interesting to know to what extent this is generalizable to other families. Patterson et al48 found that the most frequent parental discipline techniques in their sample of antisocial boys were ignoring (18%), commanding or requesting (15%), giving a time-out (10%), and scolding or nattering (8%). It is difficult to compare disparate classification systems for parental discipline, but likely that commanding or requesting and time-out would fall under limit setting, making it the most common discipline in the sample of Patterson et al as well. There is very little literature describing disciplinary practices, other than corporal punishment, for general populations. Such descriptive research would be helpful as a first step toward learning about the effects of various disciplinary practices in the general population.
We found that disciplinary practices depend on contextual factors including child, parent, family, and situational characteristics. Increasingly we see that it does not make sense to ask simply, "What kind of discipline does a parent use?" But rather, "What kind of discipline does a given parent use for a given child, in a specific family for a particular misbehavior?" It is notable that black mothers in this sample tended to use teaching or verbal assertion more and to spank more than white mothers. The finding that they spank more is not new, but that they use teaching or verbal assertion more is notable. Black mothers may emphasize their role as teachers even when they use corporal punishment. It will be interesting to explore this finding in more detail in the future.
Aversive stimuli are known to stimulate aggression in animals and humans.49,50 The most aggressive parental behavior included in this study, spanking with an object, was used for the misbehavior of disrespect. This may suggest that parents found disrespect to be the most aversive misbehavior. In addition, it raises the question of whether physical punishment is often a parental reflexive response to aversive stimuli rather than a planned strategy or a response to a child's need for learning. There is evidence that parents use mild physical punishment as a planned strategy,24 but it is not clear if this is the case for more severe forms of physical punishment.
The factors that we found associated with the 4 primary to secondary response groups must be considered as exploratory, since they are bivariate analyses, and do not control for other independent variables. Using corporal punishment as a primary response and teaching/verbal assertion or limit setting as a secondary response implies a coercive and potentially illogical contingency pattern for parental response to child behavior. The association of black race, AFDC receipt, the absence of the biological father, and less social support with using this pattern of response for 1 or more misbehaviors needs to be explored further to determine the motivation and efficacy associated with this response. Perhaps these mothers are trying to use corporal punishment to stop the misbehavior and get the child's attention, before going on to discuss it.
The new measure of discipline used in this study was helpful in exploring the relationship between the type of misbehavior and type of discipline, and allowed the extension of the existing body of research about discipline that focuses on situational context. We also gained new knowledge about some of the patterns mothers exhibit as they follow-up on primary strategies—namely, that mild physical force is used more often and teaching/verbal assertion less often as a backup than as a primary response to misbehavior. On the other hand, this measure indicated that fewer than 5% of mothers reported verbal aggression. It is likely that the true incidence of yelling is higher than this.18 Possibly, yelling is remembered as "talking." Sometimes there may be a fine distinction between verbal assertion and verbal aggression. Finally, the method of open-ended response with coding used in this new measure elicits a richer description of actual parental practices than closed-ended structured surveys. It would be useful to learn more about whether an open-ended approach increases the validity of measurement of parental discipline compared with a closed-ended approach.
There are limitations in this study. First, discipline was measured by parental report, rather than actual observation. Survey methodology has the advantage that it can be used more readily for large samples, and has the potential to be valid since parents know about their own behavior. However, Patterson et al48 found that parental report was not as accurate as direct observation for most aspects of discipline. In particular, there could be a social acceptability reporting bias against the use of physical discipline. In fact, we found a paradox that could be due to bias in the willingness of mothers to report about moderate or severe physical punishment. We found that families that had been reported to state child protection authorities for abuse or neglect said that they used mild physical punishment more and moderate or severe physical punishment less than other families. The latter may be due to fear of reporting this kind of punishment when a family has been in trouble with state authorities in the past. On the other hand, this finding may be true because both neglectful and abusive families are included in the maltreatment report category, and neglectful families may actually use moderate or severe physical punishment less. If physical punishment is underreported, then the corollary is that other types of disciplinary practices are likely to be overreported. Second, it is difficult to know to what population these findings can be generalized based on the sample selection process. Third, 20% of mothers were nonbiological when the child was 8 years old, and 11% were nonbiological when the child was aged 4 years. It is likely that many important factors have changed over time for a significant fraction of these families, including the identity of the primary caregiver. Thus, this cross-sectional view at 8 years does not embody the ways in which the at-risk status may have changed over time. Fourth, we have described disciplinary practices, but not outcomes that may be related to those practices. It would be useful to know more about short- and long-term outcomes of these practices, eg, which strategies are more effective in stopping the misbehavior, and what are the ultimate effects of these strategies in terms of how the child is socialized? Finally, we have described relationships between contextual factors and disciplinary practices, but for these to be useful, it will be important to learn in more detail about the nature of these relationships. For example, why do families receiving AFDC teach or use verbal assertion less and use moderate or severe physical force more?
Recently, there has been much interest about the effects of fathers on family interactions and child outcomes.51 In our study, the presence of a biological father was associated with the group who used teaching/verbal assertion or limit setting as a primary strategy and used corporal punishment as a secondary strategy. In addition, it is interesting to note that our data suggest that it may not be just the presence or absence of a father, but whether the father is biological that makes a difference in the extent to which limit setting is used to discipline children. By extension, it seems likely that for some child-rearing behaviors the role of the father within the household and the biological ties he has to the family may be more important than whether a father is present.
As we continue to learn about discipline, it will be important to learn more about what is effective in the short-term, and the ways in which disciplinary practices affect long-term child outcomes such as prosocial behavior, self-esteem, conscience, and aggression. We will also need to understand more about discipline as an interactive process through time, between parent and child within a contextual frame. As we learn about the short- and long-term effects of disciplinary practices, and the processes involved in those practices, we will need to introduce interventions to help parents discipline in the ways that bring about the results they desire.
Accepted for publication March 4, 1999.
The Longitudinal Studies Consortium on Child Abuse and Neglect has been supported by grants 90CA1433 and 90CA1467 from the National Center on Child Abuse and Neglect, Washington, DC. The analyses presented herein were supported by grant UO1HD30945 from the National Institute of Child Health and Human Development, Bethesda, Md.
We thank Vincent Dufort for valuable input at several stages of the analyses, Dot Browne for editing, and Denise Craig for help in preparation of the manuscript.
Editor's Note: I wonder what the results would be if this study were repeated focusing on fathers. Would "disrespect" be the stimulus to elicit the most severe response?—Catherine D. DeAngelis, MD
Corresponding author: Rebecca R. S. Socolar, MD, MPH, CB 7225, Community Pediatrics, University of North Carolina, Chapel Hill, NC 27599-7225 (e-mail: email@example.com).
Socolar RRS, Winsor J, Hunter WM, Catellier D, Kotch JB. Maternal Disciplinary Practices in an At-Risk Population. Arch Pediatr Adolesc Med. 1999;153(9):927-934. doi:10.1001/archpedi.153.9.927