OBJECTIVE. The recent but methodologically limited longitudinal study of the adverse attentional effects of television viewing in early childhood suggests a possible association. The purpose of the present study was to extend this investigation to a more current sample of kindergarten students using structural equation modeling, which allows for the simultaneous evaluation of predictors.
METHODS. Two samples were randomly selected from nationally representative data collected from the Early Childhood Longitudinal Study. A structural equation model was developed positing a relationship between kindergartners' television exposure and subsequent first-grade symptoms of attention-deficit/hyperactivity disorder (ADHD) while controlling for variables related to socioeconomic status and parent involvement. Variables were selected rather than developed and do not include an acceptable measure of ADHD, which limited the scope of the measures used. The model was tested by using the first sample and then cross-validated to the second sample.
RESULTS. Although the adequate fit of the model to the data suggests that children's television exposure during kindergarten was related to symptoms of ADHD during the first grade, the amount of variance accounted for in the ADHD-symptoms variable revealed television exposure as a weak predictor of later ADHD symptoms. Effect sizes for the relationship between television exposure and symptoms of ADHD were close to zero and not statistically significant.
CONCLUSIONS. Methodologic issues, including participant age, the measurement of ADHD symptoms, and evaluation of the importance of variables, may explain the differences between the present study and the results of others who have found television exposure to be related to attention problems. The measurement of ADHD symptoms through the use of longitudinal databases is an important limitation, because only a small number of items can be selected to represent symptoms. Future research is necessary to address these issues.
the american academy of Pediatrics Committee on Public Education1 summarized the risks and benefits of media effects (such as from television viewing) for children and adolescents, warning that such behavior “often displaces involvement in creative, active, or social pursuits.” Despite concerns associated with media exposure ranging from violence to obesity, the committee did not specifically address a concern for young children's subsequent attention. However, the recommendation that parents directly interact with their children to promote healthy brain growth and, thus, cognitive development suggests the possibility that exposure to television at a young age may influence neural development in a negative manner.
Knowledge of how behavior influences neurological development is limited despite considerable advances in functional neuroimaging and in the study of child development.2 Imaging studies show that there are widespread structural and functional differences in the brains of people with attention-deficit/hyperactivity disorder (ADHD) when compared with people without ADHD.3 These findings, combined with those of genetics studies, suggest a neurobiological basis for ADHD.3 The need for additional inquiry into this area is apparent, because evidence exists supporting the influence of early life experiences on the brain, both structurally and functionally.2 Greenough and Black4 suggested that the overly produced synapses that are present in early life are pruned back based on experiences with one's environment. That is, when connections are not formed because of a lack of experience, then synapses weaken and wither away. Thus, the brain and environment interact to assist a child in adaptation.
Christakis et al5 speculated that such an interaction occurs between early television exposure and subsequent attention problems in children, which could possibly be so severe as to result in a diagnosis of ADHD. Using data from the National Longitudinal Survey of Youth 1979 Children and Young Adults, the authors found that early television exposure at age 3 was related to attention problems at age 7 when a number of additional environmental variables were statistically controlled, such as maternal depression, the presence of 2 parents in the household, and home environment stimulation. Results indicated that with an increase of 1 SD in the number of hours of television watched at age 1 year, the child would experience a 28% increase in probability of having attention problems at age 7 years. Christakis et al5 recommended that television viewing be limited during the especially formative years of brain development that occur during early childhood; however, this recommendation may have been too global, because the authors only used a measure of overall television viewing that did not discriminate between educational television and more typical programming. This concern is only one of many that indicate additional research is necessary to implicate television exposure in the development of ADHD symptoms.
Unfortunately, Christakis et al5 only discussed the prediction of group membership in their logistic-regression models and failed to report the importance of predictors, which seemed to be the analysis that was intended based on the stated purpose of their research. Failing to compare models or the change in odds with and without each predictor, or at least conducting Wald tests, the authors could not confidently report that television viewing played an important role in predicting which group would likely experience attention problems. Also of concern is the authors' collapsing of their outcome measure of attention problems. Because Christakis et al5 were attempting to predict classification, a dichotomous outcome variable was needed rather than a continuous outcome. By using a cut point to separate those children with scores ≥1.2 SD above the mean from those children remaining, important information was lost concerning the effect of television viewing on attention across a variety of attention levels. Therefore, restricting data in this manner can be questionable, especially when an SD of 1.2 is equivalent to the 89th percentile, which would indicate that ADHD has a prevalence rate of 11% in the population. With well-acknowledged prevalence rates of ADHD between 3% and 7%, this estimate is somewhat high.6
Finally, the items collapsed to construct the variable of attention problems in the Christakis et al5 study included symptoms not commonly linked to ADHD. Christakis et al5 included “is easily confused” and “has trouble with obsessions” as 2 of the 5 items used to define attention problems. An examination of the symptoms of ADHD listed in the Diagnostic and Statistical Manual, Fourth Edition, Text Revision6 shows that these are not symptoms of ADHD. Furthermore, an emphasis on attention problems alone is of concern, because many young children display some degree of difficulty with their attention without requiring a formal diagnosis.7
The present study was similar to the work by Christakis et al5 in that a large nationally representative database was used to evaluate the role of early television viewing in later symptoms of ADHD, with the outcome variable composed of items assessing self-control, impulsivity, overactivity, and inattention. Also, the role of additional variables such as limits on watching television, parent involvement, and socioeconomic status was accounted for. However, to address the aforementioned issues, the current data were analyzed by using a structural equation model (SEM) so that regression equations or the relationships of the targeted variables to subsequent symptoms of ADHD could be evaluated simultaneously while also assessing measurement issues. Therefore, the evaluated relationships can be considered to be free of error, because measurement error was estimated and removed.8 Also, “SEM is the only analysis that allows complete and simultaneous tests of all the relationships.”8 As a result, the present analysis would allow not only a test of the adequacy of the developed model but also the amount of variance in the variables accounted for by the factors. In other words, SEM would allow the evaluation of the model while also revealing the importance of television-viewing hours on subsequent symptoms of ADHD in the context of the model or in the presence of the other variables.
Participants were children who were randomly selected from the Early Childhood Longitudinal Study-Kindergarten (ECLS-K) database.9 The ECLS-K is a collaborative project involving the US Department of Agriculture, US Department of Health and Human Services, and US Department of Education that has involved ongoing assessment of 22000 children and families attending >1200 public and private schools to provide data to assist in the investigation of school readiness, elementary school transitions, relationships between the kindergarten experience and subsequent school performance, and growth in cognitive and noncognitive domains. Information has been collected from parents, schools, teachers, and the children themselves. The ECLS-K sample design is nationally representative of kindergartners who started school during the 1998–1999 school year.
Two samples of 2500 children were randomly selected from the ECLS-K database.9 The data used for the present study were selected from 3 data collections (fall 1998, spring 1999, and spring 2000) so that television-viewing time and associated rules for viewing experienced during the kindergarten year could be used to predict symptoms of ADHD during the first-grade school year. Information selected for the current study was collected from parents and teachers.
Two variables were used to assess the amount of time children spent watching television during the spring of their kindergarten school year. Parents were asked the number of hours each school day that their child watched television or videos. Parents did not provide information concerning their children's exposure to video games. Parents were also asked the number of hours their child watched television or videos on Saturdays and Sundays. One variable was used to provide information, albeit limited, concerning the type of programming the children viewed. Parents were asked if their children watched Sesame Street either at home or someplace else for a period of ≥3 months before starting kindergarten.
Limits on Watching Television
Three variables were used to assess whether parents limited the television-viewing hours and viewing content of their children during the spring of the kindergarten school year. Parents responded “yes” or “no” to questions asking whether they had family rules for which television programs the child could watch, how many hours the child could watch television, and how early or late the child could watch television. An estimate of internal reliability using Kuder-Richardson formula 20 was low at .59 for the first sample and .56 for the second.
Parent Involvement With Child
To assess the amount of time parents were involved in activities with their child during the fall of the kindergarten year that did not involve television viewing, a series of variables were used that requested that parents report the amount of time they participated in specific activities, including reading, playing sports, singing, helping with art, completing chores, playing games, building things, and teaching about nature, with their children. The frequency scale used provided 4 options, including not at all, once or twice a week, 3 to 6 times a week, and every day. An internal reliability estimate of Cronbach's α reached .70 for the first sample and .68 for the second.
Parents' socioeconomic status was assessed by using 2 composite variables collected during the fall of the children's kindergarten school year. The first variable provided a continuous socioeconomic scale based on parent reports of income, education level, and prestige scores for the parents' occupations. The second variable was a dichotomous classification of whether the family's income was above or below the poverty thresholds based on Census information.
Symptoms of ADHD
To create the latent variable for symptoms of ADHD, 3 variables completed by teachers (approaches to learning, self-control, and externalizing problem behaviors) and one completed by parents (impulsive/overactive) were used. The approaches-to-learning composite evaluated the child's attentiveness, task persistence, eagerness to learn, learning independence, flexibility, and organization. The self-control composite evaluated the child's ability to control behavior through respecting peers' rights, controlling his/her temper, accepting peers' ideas during group activities, and responding appropriately to peer pressure. The externalizing-problem-behaviors composite evaluated the frequency of the child's argumentativeness, fighting, anger outbursts, impulsive actions, and disturbance of ongoing activities. The impulsive/overactive composite evaluated the parents' rating of the frequency of the children's hyperactivity, including fidgeting and impulsive acts.
All composite variables were derived, with permission, from the Social Skills Rating Scale: Elementary Scale A10 to create the Social Rating Scale (SRS).9 The SRS requested that parents and teachers report using a frequency scale (ie, never, sometimes, often, and very often) how often the child demonstrated a described behavior. The purpose of the adaptation was brevity, and exploratory factor analysis and confirmatory factor analysis were used to “confirm the scales.”9 Split-half reliability coefficients for the teacher SRS during the spring of the children's first-grade year were .89 for approaches to learning, .80 for self-control, and .86 for externalizing problem behaviors. Split-half reliability coefficients for the parent SRS impulsive/overactive category was .48 at the same time period. The low estimate for the parents' ratings of impulsivity and overactivity is of concern; however, this likely reflects the problems associated with measuring behavior that can vary considerably across the environments in which parents view their children, such as home, church, community, and so forth.
A theoretical model was developed specifying relationships between the 5 aforementioned latent variables. Television exposure was used to predict symptoms of ADHD. Limits on watching television, parent involvement with child, and socioeconomic status were each linked to symptoms of ADHD to statistically control for these relationships, which were suspected as possible variables related to symptoms of ADHD. Finally, socioeconomic status was used to predict television exposure. Refer to Fig 1 to view the tested model.
LISREL 8.5211 was used to test the goodness of fit of the hypothesized model to the first randomly selected sample of 2500 children. Because modifications were made, the model was then cross-validated using the second randomly selected sample of 2500 children. The assessment of fit through the evaluation of χ2 was not used in the current study because of the extensive amount of criticism this method has received; however, the statistic has been reported. The χ2 value has been criticized for its sensitivity to sample size and lack of robustness to the violation of basic underlying assumptions.8,12 Alternative goodness-of-fit indexes were selected based on the recommendations of Hu and Bentler.13 A 2-index presentation strategy that involves evaluating both the maximum-likelihood–based standardized root-mean-squared residual (SRMR) and the maximum-likelihood–based root-mean-squared error of approximation (RMSEA) was used in the present study. This combinational rule of RMSEA > .05 and SRMR > .06 was used, because Hu and Bentler13 suggested that this strategy “resulted in acceptable type II error rates for simple and complex misspecified models under both robustness and nonrobustness conditions.”
The LISREL 8.52 program11 using the SIMPLIS programming language was used to evaluate the model's fit to the first sample. Maximum-likelihood estimation was used, and parameter estimation matrices were positive definite, with no parameter estimates outside their permissible range. Goodness-of-fit indexes revealed an adequate fit to the data, with RMSEA = .06 and SRMR = .05 (χ2 = 1627.11 ; P < .001). The amount of variance accounted for in symptoms of ADHD by the remaining 4 variables was 10%.
Evaluation of parameter estimates revealed that all were close to zero with the exception of the moderate-positive relationship between socioeconomic status and television exposure, as well as the moderate-negative relationship between socioeconomic status and symptoms of ADHD. The former association indicates that as socioeconomic status increases, symptoms of ADHD decline. The latter is somewhat more complicated in interpretation because of the differences in scales that comprise the television exposure latent variable. This association indicates that as socioeconomic status increases, more Sesame Street is viewed but less television overall is watched. A small and positive association between television exposure and symptoms of ADHD was close to zero and not statistically significant despite the large sample size (t = 1.12). In addition, the small association found between limits on watching television and symptoms of ADHD was not statistically significant (t = −0.56). A small, positive association between parent involvement with child and symptoms of ADHD was statistically significant but was close to zero. Because error variance is very small because of the large sample size, an overestimation of the statistic results. Although statistical adjustments can be used to correct for this problem, this was not deemed necessary, because the parameter estimates, which are estimates of effect size, were clearly low with the exception of those related to socioeconomic status. Parameter estimates are also presented in Fig 1.
One modification was made to the model, because the fit seemed to be adequate rather than good. The limits-on-the-watching-television latent variable was removed, because the relationship between this variable and the variable symptoms of ADHD was not statistically significant. The parameter between television exposure and symptoms of ADHD was not removed despite its lack of statistical significance, because it was the main point of interest for the present study. The model was then cross-validated to a second randomly selected sample. Maximum-likelihood estimation was used, and parameter estimation matrices were positive definite, with no parameter estimates outside their permissible range. Goodness-of-fit indexes still revealed an adequate, although improved, fit to the data, with RMSEA = .05 and SRMR = .05 (χ2 = 940.36 ; P < .001). The amount of variance accounted for in symptoms of ADHD by the remaining 3 variables was 6%. As observed in the first sample, the parameter estimate between television exposure and symptoms of ADHD was very small and not statistically significant (t = −1.75). Refer to Fig 1 for all of the parameter estimates. Also, parent involvement with child seemed to have little predictive ability on symptoms of ADHD. Finally, socioeconomic status seemed to be related to symptoms of ADHD, as well as television exposure in the same manner observed in the initial analysis conducted with the first sample.
As noted by Christakis et al,5 the use of existing databases does not allow for manipulation of variables. In other words, the present study could only investigate relationships rather than causality. This limitation will likely exist for this type of study, because young children should not be purposely exposed to a condition of television viewing that could potentially pose any level of harm. Despite this necessary caution, much clinical research is quasi-experimental in nature because of the aforementioned limitation, and interpretations can be important when presented in the context of the methodology used. Even so, interpretation of the current and future findings should be considered carefully, as well as the variables selected for control.
Similar to the findings of Christakis et al,5 we were able to develop a model of television viewing that adequately fit 2 samples of children randomly selected from a nationally representative database. Although it would initially seem that children's television exposure during their kindergarten school year predicted symptoms of ADHD during their first-grade year, inspection of parameter estimates and amount of variance accounted for in the ADHD variable revealed that the effect of television exposure was weak. Because Christakis et al5 did not evaluate this latter issue or at least failed to report this information, the same phenomenon could be an issue in their study as well. Several other important differences in methodology may have also resulted in the contrast in findings.
Christakis et al5 investigated the television-viewing behavior of 1-year-olds and its association with these students' attention behavior at 7 years, whereas the present study investigated the television-viewing behavior of kindergartners and its relationship to these students' first-grade attention behavior. Although the relationship reported by Christakis et al5 only supported that television exposure and attention problems were related and not that television exposure caused attention problems, the possibility of a causal relationship is the purpose for conducting such a study. Therefore, consideration of the possibilities for the discrepancy of results between the 2 studies includes a causal explanation; however, this discussion is speculative and is not included to imply causality in the relationships evaluated.
First, the discrepancy in results may be related to differences in the ages of participants. Because synaptic activity in the brain typically reaches its peak at around age 3,14 the brains of kindergartners who are ∼5 years of age may not have the malleability of those of 3-year-olds, suggesting that any possible adverse effects of television exposure occur around the age of 3 with later exposure not as harmful. Even so, plasticity still exists at age 5, which could indicate that a period of 1 year is not a long enough period of time for adverse effects to be evident.
Second, the present results may have differed from those of Christakis et al5 because of their use of a cut point to determine attention problems and the current study's used of a continuous variable of attention problems. A possibility exists that television viewing does adversely affect the attention of those children who may be predisposed to experiencing a greater degree of symptomatology, which would most likely be evident when these children are separated out from those who are more typical, as in the Christakis et al5 study. However, when considering the effects of television exposure on more typical children, which was the design of the present study, television exposure may not be strongly related to attention problems. In other words, Christakis et al5 may have been more likely to capture children with attention problems severe enough to indicate an ADHD diagnosis, and such a diagnosis may have interacted with television-viewing hours to produce a more negative response.
Others have found that ADHD interacts with the environment to influence attention during the viewing of television programming.15,16 For example, Landau et al15 found that boys diagnosed with ADHD did not differ in their television-viewing attention in comparison with typical boys in the control group unless toys were present in the environment. When toys were present, the boys diagnosed with ADHD spent about half the time as boys in the control group attending to the television. This effect was most pronounced with younger children of both groups. Future research is necessary to determine whether harmful effects on attention based on television-viewing hours depends on whether or not the young child is diagnosed with ADHD.
Neither the Christakis et al study5 nor the present study actually evaluated children with ADHD, which is a significant limitation when attempting to interpret the findings in the context of this disorder. Although the current analysis included data from both teachers and parents to develop a latent construct composed of ADHD symptoms beyond attention, a social skills rating scale was used to collect this information, which differs considerably from measures designed to investigate the presence of an ADHD disorder. Social difficulties are often experienced by children diagnosed with ADHD,7 which has led some to evaluate the effectiveness of the original Social Skills Rating System (SSRS) in identifying preschoolers who require intervention for ADHD.17 Bain and Pelletier17 used cluster analysis to group preschool children based on shared behavioral characteristics related to ADHD using the Conners' Teacher Rating Scales-28 and then evaluated whether the children in these groups reflecting average, moderate, and high problems differed on the SSRS. The authors found that SSRS Social Skills and Problem Behavior scores differed as expected across the 3 groups, suggesting that these related problems could be used to identify children requiring remediation, because this strategy might allow for the avoidance of misdiagnosis associated with the greater report of self-control and activity problems in preschool age children that has been identified by Barkley.7 These findings suggest that the symptoms of the ADHD variable in the present study may be related to ADHD; however, caution in the interpretation of findings is still warranted, because the degree of children's impairment was not included and is necessary when considering an ADHD diagnosis.
Future research is clearly warranted in investigating the aforementioned discrepancies and measurement issues. Although the use of national databases is advantageous because of their national representativeness and sample size, the use of such databases limits researchers in selecting the age of participants and items for variable creation. This latter concern extends beyond the creation of the symptoms of the ADHD variable to the construct of television exposure. A distinction should be made between educational television viewing and other types of programming, because some types of television programming have been linked to positive outcomes18; however, this information was not recorded in the database, and a more general variable was constructed. The accuracy of general or global estimates of television exposure provided by parents has been questioned. Anderson et al19 investigated the validity associated with data yielded from the question, “How many hours would you estimate your child watches TV during the following times at this time of year?” The item was found to have a statistically significant moderate correlation (.60) with a parent diary estimate that accurately reflected the actual time 5-year-olds spent with the television as assessed through videotaped observations. These results led to the authors' conclusion that “parents can be good observers; the more specific the behavior and time of report required, the better the observation.”19 Although global parental measures may not provide the best assessment of television viewing, they do seem to be adequate if not “good.”
Anderson et al19 additionally discussed the differences between visual attention to the television and time spent with the television. In other words, children may be in front of the television but not actually attending to it, because they are engaged in other activities. Anderson et al19 found that the percentage of children's visual attention to the television did not significantly correlate with the time that they spent with the television. As a result, the authors warned that descriptions of television viewing extend beyond time spent with and/or looking at the television. Therefore, parent responses to the questions used in the present study that evaluated how many hours children “watched” television likely include periods of time in which the children were not actively attending to the television. Even so, simply being in front of the television and experiencing background exposure would likely have some affect on children's cognitive processes.
For example, Pool et al20 found that the performance of eighth-grade students completing academic tasks in front of a television playing soap operas was significantly lower than that of students completing the same tasks in a quiet environment. The authors noted that the students used time to look at the television screen and suggested that students' cognitive resources were split between the 2 tasks, resulting in poorer outcomes on the academic tasks. Interestingly, students in an auditory-only condition that involved the presentation of the soap-opera dialogue but no visual information did not experience significantly lower performance when compared with the students in the control condition. These findings suggest that background exposure to television is a problem because it distracts children, resulting in periodic shifts in attention to the television.
Although the children in the present study may have simply been in front of the television and perhaps involved in another task, it is likely that they were attending to the television at times as well. The estimate of their viewing time may have been somewhat inflated, but they were still viewing television. This is certainly an issue when considering the measurement of this variable.
The results of the present study do not indicate the presence of an important relationship between television exposure and subsequent attention problems, which is in contrast to the recent results of Christakis et al.5 Differences in the methodology of the 2 studies may be the cause of this discrepancy or possibly developmental issues are implicated, with future research warranted to resolve these concerns. Furthermore, it should be noted that ADHD, although identified by other names, has been recognized as a disorder of childhood well before children had television to watch.21 Many studies have examined various aspects of parenting to identify causes of ADHD. No support has been found for the idea that parenting causes ADHD,21 although environmental factors have been found to contribute to some cases of ADHD.22 For example, prenatal, postnatal, and perinatal trauma have been linked in some studies to ADHD,23 as has exposure to environmental toxins.22 Strong support has been found for a genetic basis of ADHD.22 Researchers have learned that much of child development is reciprocal, with characteristics of a child influencing the way that child is parented in addition to parenting influencing characteristics of a child.24 It may be that exhausted parents of very active and inattentive children resort to using the television as a “babysitter” more commonly than do parents of less active and more attentive children. Thus, the relationship between early television viewing and later attention problems may be linked to child temperament as much as or more than television causing children to be inattentive.
- Accepted July 22, 2005.
- Address correspondence to Tara Stevens, EdD, Department of Educational Psychology and Leadership, Box 41071, Texas Tech University, Lubbock, TX 79409-1071. E-mail:
The authors have indicated they have no financial relationships relevant to this article to disclose.
- ↵American Academy of Pediatrics. American Academy of Pediatrics Committee on Public Education: media education. Pediatrics.1999;104:341–343
- ↵Greenough WT, Black JE. Induction of brain structure by experience: substrates for cognitive development. In: Gunnar MR, Nelson CA, eds. Minnesota Symposia on Child Psychology: Developmental Neuroscience. Hillsdale, NJ: Erlbaum; 1992:155–200
- ↵Christakis DA, Zimmerman FJ, DiGiuseppe DL, McCarty C. Early television exposure and subsequent attentional problems in children. Pediatrics.2004;113:708–713
- ↵American Psychiartric Association. Diagnostic and Statistical Manual of Mental Disorders. 4th ed, Text Revision. Washington, DC: American Psychiatric Association; 2000
- ↵Barkley RA. Attention Deficit Hyperactivity Disorder: A Handbook for Diagnosis and Treatment. New York, NY: Guilford Press; 1990
- ↵Tabachnick BG, Fidell LS. Using Multivariate Statistics. 4th ed. New York, NY: Harper Collins; 2001
- ↵National Center for Education Statistics. Early Childhood Longitudinal Study, Kindergarten Class of 1998-99: First-Grade Public-Use Data Files User's Manual. Washington, DC: National Center for Education Statistics; 2002
- ↵Gresham FM, Elliott SN. Social Skills Rating System. Circle Pines, MN: AGS; 1990
- ↵Joreskog, K, Sorbom, D. LISREL 8: Structural Equation Modeling With the SIMPLIS Command Language. Chicago, IL: Scientific Software International; 1993
- ↵Shore N. Rethinking the Brain: New Insights into Early Development. New York, NY: Families and Work Institute; 1997
- Copyright © 2006 by the American Academy of Pediatrics