[Skip to Content]
Sign In
Individual Sign In
Create an Account
Institutional Sign In
OpenAthens Shibboleth
[Skip to Content Landing]
April 2005

Early Cognitive Stimulation, Emotional Support, and Television Watching as Predictors of Subsequent Bullying Among Grade-School Children

Author Affiliations

Author Affiliations: Department of Health Services (Drs Zimmerman and Christakis), Child Health Institute (Drs Zimmerman and Christakis), Department of Psychiatry and Behavioral Sciences (Drs Glew and Katon), Department of Pediatrics (Drs Glew and Christakis), University of Washington, Seattle.

Arch Pediatr Adolesc Med. 2005;159(4):384-388. doi:10.1001/archpedi.159.4.384

Background  Bullying is a major public health issue, the risk factors for which are poorly understood.

Objective  To determine whether cognitive stimulation, emotional support, and television viewing at age 4 years are independently associated with being a bully at ages 6 through 11 years.

Methods  We used multivariate logistic regression, using data from the National Longitudinal Survey of Youth, to adjust for multiple confounding factors.

Results  Parental cognitive stimulation and emotional support at age 4 years were each independently protective against bullying, with a significant odds ratio of 0.67 for both variables associated with a 1-SD increase (95% confidence interval, 0.54-0.82 for cognitive stimulation and 0.54-0.84 for emotional support). Each hour of television viewed per day at age 4 years was associated with a significant odds ratio of 1.06 (95% confidence interval, 1.02-1.11) for subsequent bullying. These findings persisted when we controlled for bullying behavior at age 4 years in a subsample of children for whom this measure was available.

Conclusion  The early home environment, including cognitive stimulation, emotional support, and exposure to television, has a significant impact on bullying in grade school.

Bullying among children in schools around the world has received emphasis recently as a serious public health problem. Bullying affects between 8% and 50% of children around the world,17 with a recent US estimate of about 30%.8 Despite decades of research, there are still gaps in our understanding of bullying. In particular, we know little about what environmental factors predispose children to become bullies.

Children’s early home environments have been shown in many studies to be strongly associated with the subsequent development of antisocial behavior problems.9 This research has focused on the emotional support that parents provide to young children, which assists the development of empathy, self-regulation, and prosocial skills.913 Very little research has examined the links between the early home environment and subsequent bullying behavior in particular. Existing literature suggests that 3 specific early childhood predictors might play a role in the development of subsequent bullying: cognitive stimulation, emotional support, and television exposure.

Recent theoretical work suggests that bullying might arise out of early cognitive deficits—including language problems, imperfect causal understanding, and poor inhibitory control—that lead to decreased competence with peers, which over time develops into bullying.14,15 A small number of studies provide circumstantial evidence that such a hypothesis might have merit7 : 1 study found a link between poor early cognitive stimulation and (broadly defined) inappropriate school behavior,16 and another found cognitive stimulation at age 3 years to be protective against symptoms of attention-deficit disorder at age 7 years.17 A study of Greek children found that academic self-efficacy and deficits in social cognition were related to bullying behavior.18 A large US national survey found that those who perceive themselves as having average or below-average academic achievement (as opposed to very good achievement) are 50% to 80% more likely to be bullies.8 Yet these studies are based on cross-sectional surveys, with the variables all measured at a single point in time.

The hypothesis that early emotional support is associated with subsequent bullying is also plausible theoretically16 and has received circumstantial empirical support. Parenting style has been associated with contemporaneous bullying in cross-sectional studies,19,20 and maltreatment is associated with bullying.21 Accordingly, the noted cross-sectional association between parenting style and bullying could be biased by reverse causality.

By contrast, a clear consensus in the literature indicates that there is a causal role for television violence in producing aggressive behavior,2226 although only 1 study27 specifically correlated television viewing with bullying behavior and this in a Swiss cross-sectional sample with self-report of both media use and bullying.

On the basis of this previous theoretical and empirical work, we hypothesized that (1) early cognitive stimulation, (2) early parental emotional support, and (3) early viewing of television would predict subsequent bullying behavior, controlling for baseline bullying. It is important that these hypotheses be tested simultaneously. Each main predictor is likely to be correlated with the others, as well as potentially with the outcome of bullying, so that each main predictor also functions as a potential confounder of the other 2 relationships.28 The task of this analysis is to determine whether each predictor constitutes an independent risk factor for subsequent bullying. Accordingly, these hypotheses were tested in a multivariate model using survey data from a nationally representative longitudinal data set.

Data source

Data from this study were drawn from the National Longitudinal Survey of Youth 1979 Children and Young Adults (NLSY-Child), an outgrowth of the original National Longitudinal Survey of Youth 1979 (NLSY79). The NLSY79, sponsored by the US Department of Labor, began with a nationally representative sample of almost 12 700 individuals ages 14 through 22 years in 1979 who have been interviewed annually or biennially since.29,30 African American and Latino individuals were oversampled to provide statistical power for analyses involving these important subgroups, and population weights were available to draw valid national inference. Conducted biennially from 1986 through 2000, the NLSY-Child developed an extensive collection of information for more than 11 000 children of the female respondents to the NLSY79 regarding developmental assessment, family background, home environment, and health history.31 Information for the NLSY-Child was obtained from both the mother and child, depending on the child’s age. The records from NLSY79 and NLSY-Child are linkable via the mother’s sample identification number. Data from both the NLSY-Child and NLSY79 were pulled for this study using the CHRR Database Investigator Software (Center for Human Resources Research, Columbus, Ohio).

Our sample consisted of all children who were aged 6 to 11 years (72-143 months) at the time of the 2000 survey interview. The University of Washington (Seattle) institutional review board determined that this research was exempt from review because it used only previously collected, unlinked data.

Outcome measure

Our outcome measure involved characterization of a child as a bully by his or her mother. The statement was, “[name of child] bullies or is cruel or mean to others,” and possible responses were “often true,” “sometimes true,” or “not true.” We grouped together the “sometimes true” and “often true” responses as positive for bully status and “not true” as negative for bully status.

A recent methodological piece on the identification of children who bully suggested that the use of a single-item question is the method of choice for prevalence estimation of bullying.32 The same article discusses the construct validity of the preferred measure by assessing the correlation between bullying and an antisocial measure. Solberg and Olweus32 found that identified bullies have a mean externalizing score that is approximately 1 SD higher than that for nonbullies, and the difference is significant at P<.001. (Externalizing behavior refers to problem behaviors such as antisocial behavior, oppositional behavior, hyperactivity, and so on.) As a check on our own measure of bullying, we performed the same analysis using the short-form Behavior Problems Index, which includes a validated antisocial score for each child.31 (We revised the Antisocial scale by subtracting the values of the answers to the bullying question, which would otherwise contribute to the Antisocial scale.)

In our data, the antisocial score is about 1.3 SDs higher among the bullies than nonbullies, and the difference is significant at P<.001. Of the children in our sample, 13% were identified as bullies, which is similar to a recent national estimate of prevalence of 19%.8 We therefore conclude that the maternal report of bullying is a reasonable and informative assessment of true bullying behavior.

Main predictors

Our first 2 main predictors were the cognitive stimulation and emotional support subscales of the Home Observation for Measurement of the Environment–Short Form, measured at age 4 years. These subscales have been shown to have good psychometric properties and have been used extensively in child development research.31,33,34 They include both maternal-report and interviewer-report items. The cognitive stimulation score generally includes items related to outings, reading, playing, and the parental role in teaching a child. For younger children, the emotional support score consists of elements related to eating meals with both parents, parents talking to the child while working, and spanking. To facilitate interpretation, scores were normalized so that a 1-unit change in the variable represents a 1-SD change.

Our third main predictor variable was the weekly average number of hours of television watched per day. As of 1990, mothers were asked the number of hours of television that the child (younger than 10 years) watches on a typical weekday and on a typical weekend day. When a response indicated no television in the home, television viewing hours were set to 0. When a response indicated more than 16 hours of viewing per day, the viewing was capped at 16 hours. The number of hours per week was computed as 5 times the number of hours watched during a typical weekday plus 2 times the number of hours watched on a typical weekend day. To get a daily average, we divided this number by 7. This computation was performed for the survey year occurring closest to the 4-year birthday of each child.


We include as potential confounders any variable that has been shown to be associated with bullying and has also been shown to be associated with any of the 3 main predictors. Bullying behavior has been shown to vary with the child’s race, age, and sex,7,8,32 as has the amount and type of television viewing.35,36 The association between bullying and socioeconomic status, including parental income and education, has not been explicitly explored, but socioeconomic status has been shown to be strongly associated with externalizing behavior generally.28 Socioeconomic status is also known to influence both television viewing and parenting style.22,36 Model covariates therefore included the child’s sex; race (Hispanic, African American, or non-Hispanic/non–African American); the child’s age when the bullying question was asked in 2000; and the parents’ income and educational levels.

It is of course possible that parents find it more difficult to parent children who for other reasons (eg, genetics) display more antisocial or bullying behavior early on. If so, early bullying might prompt reduced emotional support and cognitive stimulation by the parents and more television viewing. To control for this possibility, we include our measure of bullying at age 4 years, measured contemporaneously with the main predictors (ie, early parental cognitive stimulation, emotional support, and television viewing). Doing so controls for possible confounding in the association between early predictors and subsequent bullying by predisposition to bully. Unfortunately, the early bullying measure was available only on a subsample of children, so this model was conducted as a separate subanalysis, with reduced power.


After examining the univariate characteristics of the independent variables, we ran 3 multivariate logistic regression models. The first model included all children ages 6 through 11 years. The second included only those children for whom age 4 years bullying data were available. All analyses were performed using Intercooled Stata 8.0 (Stata Corp, College Station, Tex).


A total of 1266 6- through 11-year-old children had complete data for the basic set of variables, and 641 children had complete data for the early bullying control variable. Table 1 presents descriptive statistics for all variables. Approximately 49% were female, and 11% were African American; 9% were Native American, Latino, or Asian; and the remainder were white. Approximately 13% of children were reported as bullies by their mothers.

Table 1. 
Descriptive Statistics of Variables by Bullying Status in Grade School*
Descriptive Statistics of Variables by Bullying Status in Grade School*

The cognitive stimulation and emotional support scores at age 4 years were each approximately 0.5 SDs higher among those who were subsequently reported in grade school to be nonbullies than among those who were subsequently reported to be bullies. Children at age 4 years had watched an average of 3.5 hours of television per day, with a mean of 5.0 among subsequent grade-school bullies and 3.2 among subsequent nonbullies. All of these differences were significant at P<.01.

Table 2 presents the results of the logistic regression model for all children, without controlling for early bullying. The odds ratios associated with cognitive support and emotional stimulation at age 4 years were each 0.67 (95% confidence interval [CI], 0.54-0.82, and 95% CI, 0.54-0.84, respectively). The odds ratio associated with each hour of television per day was 1.06 (95% CI, 1.02-1.11). We also controlled for parental income and education and the child’s age, sex, and race or ethnicity. Of these, only being African American was significant, showing an association with decreased bullying (odds ratio, 0.50 [95% CI, 0.29-0.87]) (not reported in the table).

Table 2. 
Regression of Bullying in Grade School on Early Predictors—Full Sample (n = 1266)*
Regression of Bullying in Grade School on Early Predictors—Full Sample (n = 1266)*

In the regression on the subset of children with data available on early bullying (Table 3), early television viewing was significantly predictive of subsequent bullying, with an odds ratio of 1.09 per hour watched per day (95% CI, 1.01-1.17). Emotional support was protective, with an odds ratio of 0.75 per 1-SD change in the subscale (95% CI, 0.56-0.99). Cognitive stimulation was not significant, with an odds ratio of 0.81 (95% CI, 0.62-1.05).

Table 3. 
Regression of Bullying in Grade School on Early Predictors and Early Bullying (n = 641)*
Regression of Bullying in Grade School on Early Predictors and Early Bullying (n = 641)*

We found that early (age 4 years) cognitive stimulation, emotional support, and exposure to television were each independently associated with subsequent maternal report that the child was considered a bully at grade-school age. This finding was present even while controlling for a number of potential confounding factors, including socioeconomic status and the child’s age, race or ethnicity, and sex. The magnitudes of the emotional support and cognitive stimulation scores were meaningful, with a 1-SD increase in each of these scores at age 4 years associated with a 33% decrease in the odds of being a bully in grade school. The magnitude of the risk associated with television, expressed in our analysis in terms of hours per day of television viewed, is clinically significant, particularly when one considers that the SD of hours per day of television viewing was 3.9. Accordingly, a 1-SD increase in the number of hours of television watched at age 4 years is associated with an approximate 25% increase in the probability of being described as a bully by the child’s mother at ages 6 through 11 years.

When we included a control for bullying behavior at age 4 years (ie, maternal report of whether the child is a bully), the results are similar. This result provides added confidence that the associations identified in this analysis are not confounded by 1 obvious possibility, that of early bullying driving parenting choices. Instead, these results are consistent with a causal role for early emotional support (as a protector) and early television viewing (as a risk factor) in the subsequent development of bullying. Although the effect of early cognitive stimulation did not achieve statistical significance in the model that controls for early bullying, that finding might be partly caused by the reduced sample size.

To our knowledge, this is the first study to test the hypothesis that early emotional support, cognitive stimulation, and television viewing are associated with subsequent bullying. The fact that the data are from a representative longitudinal data set and that the analysis controls for multiple possible confounding variables makes these findings even more compelling.

The fact that early cognitive stimulation is protective against bullying in our data corroborates theoretical work in this area. The benefits of early cognitive stimulation might function through an increase in a child’s comfort level and confidence around academic matters or through the child’s perception that the parents endorse the academic project of schooling. As a result, the child might be less inclined to disrupt the schooling process by bullying others. Clearly, this conjecture needs further research to clarify, but this result is consistent with other research that has found considerable benefits to early childhood education programs such as Head Start. It is also possible that a high score on the cognitive stimulation subscale is a proxy for a broad range of nurturing parental behaviors that are eventually associated with prosocial behavior, such as positive interactions with peers and teachers.

The role of early television viewing is particularly provocative because of the emphasis in the existing literature on television viewing among older children, such as 9-year-olds or adolescents. Most of the literature concluding that television is associated with increased aggression deals only with violent media, not media or television in general. However, approximately 60% of television programs contain violence,37 so the number of television hours watched likely correlates with the overall number of violent television hours viewed. Perhaps some programming that is not explicitly violent also leads to bullying behavior. For example, some programs contain examples of people behaving disrespectfully toward one another, which could serve as a model for children to engage in the kind of verbal abuse that qualifies as bullying in most definitions.1,32 Because watching television can be habit-forming, parents should be encouraged to limit the television viewing of their young children in accord with American Academy of Pediatrics guidelines,38 which recommend no television for children younger than 2 years old and limited television thereafter.

There are several limitations to this study. First, previous research has relied on student self-reporting bullying status, whereas this study used maternal reports. Moreover, in previous work, the definition of the term bully has been provided to those asked to assess whether someone is a bully. Here, by contrast, we did not provide the definition of bully to the mothers filling out the surveys. Not everyone has the same definition of the word. It is possible that mothers would report their children as bullies if they fought a lot, which, although an example of antisocial behavior, might not fit the usual definition of bullying, which must involve some kind of physical, social, or psychological power imbalance. If mothers of children who fight a lot—but in the absence of power imbalances—report their children as bullies, it would accordingly account for at least some of the correlation reported earlier between the antisocial and bullying measures. In this sense, there might be “false positives,” or type II errors, in the bullying measures used.

At the same time, mothers might differ in their willingness to admit that others consider their children to be bullies, especially if they disagree with that label. Hence, we were probably unable to identify all true bullying children in the sample, which would introduce false negatives, or type I errors. The measure of bullying used in this study might accordingly be subject to measurement error. In general, measurement error tends to decrease the ability to detect an effect that is in fact present and therefore introduces a conservative bias.39

Second, it is possible that there are unmeasured characteristics associated with parents who provide low levels of cognitive stimulation or emotional support and who allow their children to watch excessive television and that these unmeasured characteristics account for the relationships between these early variables and subsequent bullying. To mitigate the effects of such a possibility, we included in a second regression a measure of whether the child was a bully at baseline. However, it is possible that this measure undercontrols for the child’s true baseline risk of subsequent bullying behavior.

Finally, we have no data on the content of the television watched by these children.


Our results have some important implications. First, we have provided some empirical support to theories that suggest that bullying might arise out of cognitive deficits as well as emotional ones. Second, we have added bullying to the list of potential negative consequences of excessive television viewing along with obesity, inattention, and other types of aggression.17,4044 Third, our findings suggest some steps that can be taken with children to potentially help prevent bullying. Maximizing cognitive stimulation and limiting television watching in the early years of development might reduce children’s subsequent risk of becoming bullies. The early childhood period is an important formative time for the subsequent behavior of children. Future research will be valuable in reproducing and refining these findings and in identifying the precise causal mechanisms involved. In particular, research on how particular television content might affect bullying is essential.

Correspondence: Frederick J. Zimmerman, PhD, Department of Health Services and Child Health Institute, University of Washington, 6200 NE 74th St, Suite 210, Seattle, WA 98115-8160 (fzimmer@u.washington.edu).

Back to top
Article Information

Accepted for Publication: December 13, 2004.

Glew  GRivara  FFeudtner  C Bullying: children hurting children.  Pediatr Rev 2000;21183- 190PubMedCrossref
Olweus  D Bullying at School: What We Know and What We Can Do.  Malden, Mass Blackwell Publishers1993;
Whitney  ISmith  PK A survey of the nature and extent of bullying in junior/middle and secondary schools.  Educ Res 1993;353- 25Crossref
Hoover  JJuul  K Bullying in Europe and the United States.  J Emotional Behav Problems 1993;225- 29
Perry  DGKusel  SJPerry  LC Victims of peer aggression.  Dev Psychol 1988;24807- 814Crossref
Gofin  RPalti  HMandel  M Fighting among Jerusalem adolescents: personal and school-related factors.  J Adolesc Health 2000;27218- 223PubMedCrossref
Dake  JAPrice  JHTelljohann  SK The nature and extent of bullying at school.  J Sch Health 2003;73173- 180PubMedCrossref
Nansel  TROverpeck  MPilla  RSRuan  WJSimons-Morton  BScheidt  P Bullying behaviors among US youth: prevalence and association with psychosocial adjustment.  JAMA 2001;2852094- 2100PubMedCrossref
Simons  RLRobertson  JFDowns  WR The nature of the association between parental rejection and delinquent behavior.  J Youth Adolesc 1989;18297- 310Crossref
Gottfredson  MRHirschi  T A General Theory of Crime.  Stanford, Calif Stanford University Press1990;
Conger  RDConger  KJElder  GHLorenz  FO A family process model of economic hardship and adjustment of early adolescent boys.  Child Dev 1992;63526- 541PubMedCrossref
Deater-Deckard  K Within family variability in parental negativity and control.  J Appl Dev Psychol 1996;17407- 422Crossref
MacKinnon-Lewis  CVolling  BLLamb  MEDechman  KRabiner  DCurtner  ME A cross-contextual analysis of boys' social competence from family to school.  Dev Psychol 1994;30325- 333Crossref
Hay  DFPayne  AChadwick  A Peer relations in childhood.  J Child Psychol Psychiatry 2004;4584- 108PubMedCrossref
Dodge  KAPettit  GS A biopsychosocial model of the development of chronic conduct problems in adolescence.  Dev Psychol 2003;39349- 371PubMedCrossref
Dodge  KAPettit  GSBates  JE Socialization mediators of the relation between socioeconomic status and child conduct problems.  Child Dev 1994;65649- 665PubMedCrossref
Christakis  DAZimmerman  FJDiGiuseppe  DLMcCarty  CA Early television exposure and subsequent attentional problems in children.  Pediatrics 2004;113708- 713PubMedCrossref
Andreou  EMetallidou  P The relationship of academic and social cognition to behavior in bullying situations among Greek primary school children.  Educ Psychol 2004;2427- 41Crossref
Haynie  DLNansel  TEitel  P  et al.  Bullies, victims, and bully/victims: distinct groups of at-risk youth.  J Early Adolesc 2001;2129- 49Crossref
Baldry  ACFarrington  DP Bullies and delinquents: personal characteristics and parental styles.  J Community Appl Soc Psychol 2000;1017- 31Crossref
Shields  ACicchetti  D Parental maltreatment and emotion dysregulation as risk factors for bullying and victimization in middle childhood.  J Clin Child Psychol 2001;30349- 363PubMedCrossref
Anderson  DRHuston  ACSchmitt  KLLinebarger  DLWright  JC Early Childhood Television Viewing and Adolescent Behavior.  Boston, Mass Basil Blackwell2001;
American Academy of Pediatrics, Media violence: Committee on Public Education.  Pediatrics 2001;1081222- 1226PubMedCrossref
Strasburger  VCDonnerstein  E Children, adolescents, and the media: issues and solutions.  Pediatrics 1999;103129- 139PubMedCrossref
Anderson  CABushman  BJ Psychology: the effects of media violence on society.  Science 2002;2952377- 2379PubMedCrossref
Villani  S Impact of media on children and adolescents: a 10-year review of the research.  J Am Acad Child Adolesc Psychiatry 2001;40392- 401PubMedCrossref
Kuntsche  EN Hostility among adolescents in Switzerland: multivariate relations between excessive media use and forms of violence.  J Adolesc Health 2004;34230- 236PubMedCrossref
Deater-Deckard  KDodge  KABates  JEPettit  GS Multiple risk factors in the development of externalizing behavior problems: group and individual differences.  Dev Psychopathol 1998;10469- 493PubMedCrossref
US Department of Labor, NLS Handbook 2000.  Washington, DC Bureau of Labor Statistics2000;
 The NLSY79 User's Guide.  Columbus, Ohio Center for Human Resource Research1999;
Baker  PCKeck  CKMott  FLQuinlan  SV NLSY Child Handbook, Revised Edition.  Columbus, Ohio Center for Human Resource Research1993;
Solberg  MEOlweus  D Prevalence estimation of school bullying with the Olweus Bully/Victim Questionnaire.  Aggress Behav 2003;29239- 268Crossref
Caldwell  BMBradley  RH Home Observation for Measurement of the Environment.  Little Rock University of Arkansas at Little Rock1984;
Caldwell  BMBradley  RH Environmental issues in developmental follow-up research. Friedman  SLHaywood  HCeds. Developmental Follow-up: Concepts, Domains, and Methods San Diego, Calif Academic Press1994;235- 256
Wright  JCHuston  ACVandewater  EA  et al.  American children's use of electronic media in 1997: a national survey.  J Appl Dev Psychol 2001;2231- 47Crossref
Certain  LKKahn  RS Prevalence, correlates, and trajectory of television viewing among infants and toddlers.  Pediatrics 2002;109634- 642PubMedCrossref
Center for Communication and Social Policy, National Television Violence Study. 3 Santa Barbara University of California1998;
American Academy of Pediatrics, Television and the family. Available at:http://www.aap.org/family/tv1.htmAccessed on January 18, 2005
Clayton  DHills  M Statistical Models in Epidemiology.  Oxford, England Oxford University Press1993;
Crespo  CJSmit  ETroiano  RPBartlett  SJMacera  CAAndersen  RE Television watching, energy intake, and obesity in US children: results from the third National Health and Nutrition Examination Survey, 1988-1994.  Arch Pediatr Adolesc Med 2001;155360- 365PubMedCrossref
Ma  GSLi  YPHu  XQMa  WJWu  J Effect of television viewing on pediatric obesity.  Biomed Environ Sci 2002;15291- 297PubMed
Hu  FBLi  TYColditz  GAWillett  WCManson  JE Television watching and other sedentary behaviors in relation to risk of obesity and type 2 diabetes mellitus in women.  JAMA 2003;2891785- 1791PubMedCrossref
Janz  KFLevy  SMBurns  TLTorner  JCWilling  MCWarren  JJ Fatness, physical activity, and television viewing in children during the adiposity rebound period: the Iowa Bone Development Study.  Prev Med 2002;35563- 571PubMedCrossref
Robinson  TN Television viewing and childhood obesity.  Pediatr Clin North Am 2001;481017- 1025PubMedCrossref