Objectives. Recognizing the negative effects of television on children, the American Academy of Pediatrics (AAP) recommends that children 2 years and older watch <2 hours of television per day and that children younger than 2 years watch no television. However, relatively little is known about the amount of television viewed by infants and toddlers. The objective of this study was to describe the prevalence and correlates of television viewing that exceeds the AAP guidelines for 0- to 35-month-olds and to examine the trajectory of a child’s viewing over time.
Methods. Data from the National Longitudinal Survey of Youth, 1990 to 1998, were used to analyze reported television viewing at 0 to 35 months of age and to follow the trajectory of a child’s viewing from infancy through age 6. Logistic regression models were used to determine risk factors associated with greater television viewing at 0 to 35 months and the association of early viewing habits with school-age viewing.
Results. Seventeen percent of 0- to 11-month-olds, 48% of 12- to 23-month-olds, and 41% of 24- to 35-month-olds were reported to watch more television than the AAP recommends. Compared with college graduates, less-educated women were more likely to report that their children watched more television than recommended. Children who watched >2 hours per day at age 2 were more likely to watch >2 hours per day at age 6 (odds ratio: 2.7; 95% confidence interval: 1.8–3.9), controlling for maternal education, race, marital status and employment, household income, and birth order.
Conclusions. A substantial number of children begin watching television at an earlier age and in greater amounts than the AAP recommends. Furthermore, these early viewing patterns persist into childhood. Preventive intervention research on television viewing should consider targeting infants and toddlers and their families.
The substantial amount of television watched by school-aged children1–6 and the associated adverse effects7–26 are increasingly well documented. Although a few researchers highlight the benefits of television,12–14 the majority link increased television viewing with higher rates of violence,15–17 obesity,18–23 and poor school performance.24–26 Moreover, recent randomized, controlled trials have shown that decreasing the amount of television viewed leads to relative decreases in aggression15 and body fat.19,23 Recognizing the adverse health effects of television, the American Academy of Pediatrics (AAP) recommends that children 2 years and older limit their time with entertainment media (television, video games, the Internet) to 2 hours per day and that children younger than 2 watch no television.27,28 Little is known, however, about the amount of television viewed by infants and toddlers. The vast majority of research has remained focused on television viewing in older children.1,29
Delineating patterns of early childhood television viewing may be important for several reasons. First, it may provide insight into the earliest behavioral antecedents of obesity and other health outcomes linked to excessive television viewing. Second, a description of social differences in early childhood television viewing may illuminate the mechanisms by which social disparities in these health outcomes emerge. If school-age television habits begin to develop early in life, then the roots of social disparities in obesity and violence may be explained in part by these early experiences. Finally, if indeed very young children watch substantial amounts of television, then the findings raise important questions about the constraints that parents may face in choosing alternative activities for their children.
The present study had 2 objectives. First, we described the prevalence and correlates of television viewing that exceeded the AAP guidelines for a national sample of 0- to 35-month-olds. Second, we examined the trajectory of a child’s viewing over time. Data came from the National Longitudinal Survey of Youth (NLSY).
Sample and Design
The NLSY began in 1979 as a national sample of young men and women aged 14 to 21, oversampled for blacks, Hispanics, and low-income whites. This original cohort has been surveyed almost every year since 1979, with 84% of the original respondents still in the sample as of 1998.30 Data on children of the women in the cohort have been collected every other year since 1986. This study focused on the 5 child surveys from 1990 to 1998 that included questions about child television viewing.
For determining the prevalence and correlates of television viewing that exceeded the AAP guidelines, a cross-sectional design was used. Our sample for this objective (the “Cross-Sectional Sample”) consisted of children who were 0 to 35 months of age in any survey year from 1990 to 1998 (N = 3556). Because television viewing is strongly correlated with age, this sample was divided into 3 subgroups: 0- to 11-month-olds (“Youngest”), 12- to 23-month-olds (“Middle”), and 24- to 35-month-olds (“Oldest”). Sibling pairs existed within these subgroups, so we randomly chose 1 child for each mother. Five percent of children were missing outcome data. Children who were excluded because of missing data did not differ significantly from those who were included with respect to maternal education level or survey year. The final sample sizes were 1084 for the Youngest, 1254 for the Middle, and 1247 for the Oldest. These are not mutually exclusive samples; for example, a child who was 5 months old at the 1992 survey and 30 months old at the 1994 survey was included in both the Youngest sample and the Oldest sample.
For examining the trajectory of a child’s viewing over time, a longitudinal design was used. The “Longitudinal Sample” was a subset of the Cross-Sectional Sample. Children who were 0 to 23 months of age in 1990 or 1992 were followed for 6 years (until 1996 or 1998); 5% were lost to follow-up. As above, this sample was divided by age into subgroups: children 0 to 11 months of age at baseline (n = 554) and children 12 to 23 months of age at baseline (n = 666).
For both of the designs, the outcome variable was “hours of television per weekday,” assessed by maternal response to the following question: “How much time would you say your child spends watching television on a typical weekday (either in your home or elsewhere)?” A separate question with parallel wording assessed television on a “typical weekend day.” Mothers reported the hours of viewing in whole numbers; watching <1 hour of television per day was considered equivalent to watching no television. Following the AAP guidelines, we dichotomized the outcome variable as 0 versus ≥1 hour per day for children 0 to 23 months of age and 0 to 2 versus ≥3 hours per day for children 24 months and older. The Spearman correlation between weekday and weekend viewing was ≥0.7 for each age group, and the average amount of television on a typical weekend day was the same as on a typical weekday for all ages. Therefore, only the results for weekday viewing are shown. There was no question specific to video viewing, so mothers may have included videos in their estimate of television viewing.
Because of the limited information available on television viewing among infants and toddlers, we chose predictors based on other studies of parenting style and home environment.31–34 We calculated the average annual household income across the 5 surveys and divided the resulting value into quintiles. Children whose household income was missing in 3 or more of the survey years (11%) were given a missing value. Mothers who reported working, going to school, or serving in the active armed forces were considered to be employed “out of the home”; women who were on leave from their jobs, unemployed, out of the labor force, or “keeping house” were considered “in the home”; all others were considered “unknown.” For the paternal variables, we used the data for the person whom the mother identified as both a member of her household and a spouse or partner. The quality of the home environment was determined by the Short Form of the Home Observation for Measurement of the Environment (HOME-SF).35 Both the total standardized score and the 2 standardized subscale scores (cognitive stimulation and emotional support) were analyzed as continuous variables.
Data on child care, maternal depression, and neighborhood quality were available only in selected years; therefore, these variables were analyzed separate from the main analysis.
Because child care information was collected retrospectively, complete data were available for children in 1990 and 1992 only. Mothers were asked about any regular child care during their child’s first, second, and third years of life. We classified child care responses as “none (besides maternal),” “in a (private) home,” and “in a center/preschool.”
Maternal Depressive Symptoms
Maternal depressive symptoms were measured only in 1992 and 1994. In 1992, the NLSY included the full Center for Epidemiologic Studies Depression Scale, a 20-item self-report instrument.36–38 In 1994, the survey included only 7 questions from the Center for Epidemiologic Studies Depression Scale. The scores for each year were divided into quintiles for analysis. Within the 1992 data, the Spearman correlation between the full 20 questions and the 7 questions was 0.9.
Neighborhood quality was assessed in 2 ways, beginning in 1992. First, mothers were asked the following question: “How would you rate your neighborhood as a place to raise children? Would you say it is excellent, very good, good, fair, or poor?” In addition, mothers rated the following problems in their neighborhood: lack of respect for rules and laws; crime and violence; abandoned or run-down buildings; not enough police protection; not enough public transportation; too many unsupervised children; people keep to themselves, don’t care about the neighborhood; and lots of people who can’t find jobs. Mothers rated these problems on a 3-point scale, and the answers were summed into a composite score.
The χ2 test and t test were used in the cross-sectional analysis to examine the associations between the independent variables and television viewing. Any predictors that were significantly (P < .05) correlated with television viewing in the bivariate analyses were included in multivariate logistic regression models that examined the odds of viewing ≥1 hour/d for the Youngest and Middle groups and ≥3 hours/d for the Oldest group. In the longitudinal analyses, a logistic regression model was used to examine the odds of watching ≥3 hours of television per day at age 6. The predictor of interest was television viewing as an infant or toddler, but models also controlled for maternal education, marital status, employment, race, household income, child age and birth order.
Weighting and Design Effects
As recommended when combining data across survey years,30 we did not use sample weights to make our results nationally representative of children born to mothers who were 14 to 22 years of age in 1979.30 However, an exploratory comparison of weighted results to our unweighted results for any given individual year showed no difference in the amount of television viewed at each age.
The data were collected through a multistage stratified cluster random sampling procedure, and consequently the standard errors may be underestimated. The information required to adjust for such design effects is not in the public use data; however, respondent movement out of their original sampling units has reduced such design effects.30 We present significance at the 95% confidence level, but a conservative approach is to focus on findings significant at P < .01.39 All analyses were conducted using SAS 8.1 for Windows 95 (SAS Institute, Inc, Cary, NC).
Approximately 25% of the mothers were black, 20% were Hispanic, 75% were married, 10% had not finished high school, and 20% to 25% had finished 4 years of college (Table 1). The median household income was $41 000/y.
Seventeen percent of the Youngest sample were reported to exceed AAP television viewing guidelines, watching at least 1 hour of television on a typical weekday (Fig 1). Among the Middle sample, 48% were reported to watch at least 1 hour per day, and 22% were reported to watch 3 or more hours per day. In the Oldest sample, 41% were reported to watch 3 or more hours per day, and 16% were reported to watch 5 or more hours per day.
In bivariate analyses, the most consistent correlates of increased television viewing were black maternal race, lower maternal education, and having an unmarried mother (Table 2). For example, 51% of mothers who had not graduated from high school reported that their 2-year-olds watched at least 3 hours of television on a typical weekday, compared with only 27% of college graduates (P < .0001). Although maternal education was associated with television viewing, paternal education was not.
Lower HOME scores were associated with increased television viewing in some children (data not shown). In the Oldest group, the average HOME score for children who exceeded AAP television viewing guidelines was a third of a standard deviation worse than the average score for children who were within the guidelines (95.0 vs 99.4; P < .0001); in the Youngest group, the average score for those who exceeded the guidelines was a quarter standard deviation better (100.6 vs 97.0; P = .007). There were no differences in HOME scores for the Middle group (97.0 vs 97.5; P = .6).
Of the variables that were available only in selected years, none were consistent correlates of television viewing. Child care was significantly associated with television viewing for the Oldest sample only; children in center-based child care were the least likely to watch more than the AAP recommends (P = .02). Similarly, increased maternal depressive symptoms were significantly associated with increased television viewing for the Oldest sample only (P = .02). Poor neighborhood quality was significantly associated with increased television viewing in the Youngest and Oldest samples (P < .05 for both neighborhood assessments).
In logistic regression models, maternal race, maternal education, and child age were consistent predictors of high television viewing (Table 3). A black mother was twice as likely as a white/other mother to report that her 2-year-old watched at least 3 hours of television per day (odds ratio [OR]: 2.0; 95% confidence interval [CI]: 1.4–2.8). A woman who had not graduated from high school was almost 4 times as likely as a woman who had graduated from college to report that her 0- to 11-month-old watched at least 1 hour of television per day (OR: 3.7; 95% CI: 1.7–7.7). Survey year was also a significant predictor of child television viewing, particularly for the Youngest sample. Infants in 1998 were more likely to watch television than infants in preceding years. The HOME score was significant for the Youngest children only; higher total and subscale scores were associated with increased viewing.
In separate regression models for the variables that were available only in selected years, child care and neighborhood retained significance but maternal depression did not. Compared with children in center-based child care, 2-year-olds with no formal child care were more likely to watch more than the AAP recommends (OR: 1.6; 95% CI: 1.0–2.6; P = .04), as were 2-year-olds who were cared for in a private home (OR: 1.6; 95% CI: 1.0–2.6; P = .05). Compared with a mother who rated her neighborhood as “excellent” for raising children, a mother in a “poor” neighborhood was more likely to report that her infant watched at least 1 hour of television per day (OR: 3.6; 95% CI: 1.5–8.3; P = .003), adjusting for covariates. Using the composite scale, infants in the worst quartile of neighborhoods were twice as likely as infants in the top quartile to watch at least 1 hour per day (OR: 2.2; 95% CI: 1.1–4.5; P = .03).
To determine whether neighborhood quality confounded the relationship between television viewing and race or education, we looked for changes in the β-coefficients for race and education on entering neighborhood quality in the model. Including neighborhood quality did not substantially change the coefficients of either, suggesting that perceived neighborhood quality was not a confounder of these associations.
The trajectory of television viewing with age is shown in Fig 2. Daily television viewing increased by roughly 1 hour per year during the first 3 years of life, then leveled off. Differences between children of less-educated mothers and children of well-educated mothers were significant at an early age and became more pronounced as the children got older. By age 4, children of less-educated mothers were watching an additional 2 hours of television per day, on average, a nearly 2-fold difference.
Given the varying trajectories for different subgroups, we investigated whether early factors, particularly early television viewing, were associated with television viewing at school age (age 6). We found that television viewing at 24 to 35 months predicted school-age television viewing, but television viewing at 0 to 11 months did not, adjusting for maternal education, race, income, marital status, and employment. Children who watched at least 3 hours of television per day at age 2 were almost 3 times as likely as other children to watch at least 3 hours per day at age 6 (OR: 2.7; 95% CI: 1.8–3.9; P < .0001). Maternal education was also a significant predictor of television viewing at age 6, controlling for viewing at age 2. Children of high school graduates were more than twice as likely as children of college graduates to watch more television than the AAP recommends (≥3 hours per day) at age 6 (OR: 2.3; 95% CI: 1.4–3.9; P = .002).
This study presents the first national data on the prevalence, correlates, and trajectory of television viewing habits for infants and toddlers. Seventeen percent of 0- to 11-month-olds, 48% of 12- to 23-month-olds, and 41% of 24- to 35-month-olds were reported by their mothers to watch more television than the AAP recommends. Less-educated mothers reported that their children viewed more television; 29% of mothers with <12 years of education reported that their infants watched television, compared with only 14% of college graduates. These differences, apparent at an early age, increased as the children grew older. Both early television viewing and maternal education had significant, independent effects on television viewing at school age.
Children in our study were reported to watch slightly more than those in a national cross-sectional study, in which the average 1-year-old watched 6 hours per week and the average 3-year-old watched 13 hours per week.6 Children in our study watched slightly less than children surveyed in 2 pediatric clinics.40 These differences are probably attributable to differences in sample composition, data collection methods, and the inclusion of videos. For example, the national study used diaries to determine the amount of time spent with television, and the clinic study separated television from video viewing.
Children of black and less-educated mothers watched more television at all ages. These social gradients in television viewing were our most persistent findings and were consistent with studies of older children1,18 and adults.29 Because it is unlikely that race and education directly influence television viewing, there must be other contributing factors. For example, it has been suggested that concerns about safety might present a barrier to children’s going outside to play and that blacks are more likely to live in neighborhoods perceived as unsafe.18 Surprisingly, adding perceived neighborhood quality to our regression models did not change the estimates for race or maternal education, perhaps because the questions in the NLSY did not capture the relevant neighborhood characteristics.41,42 In addition to neighborhood quality, other factors potentially related both to television viewing and to race or education need to be considered, such as residential stability,42 wealth,43 and the accessibility of playgrounds, museums, and libraries. Our results can speak only indirectly to the many constraints that disadvantaged families and their children may face in pursuing beneficial alternatives to television.
Our longitudinal analysis indicates that greater television viewing in early childhood is associated with greater viewing at school age. The persistence of this behavior pattern may reflect continuing environmental influences, the development of child preferences or habits, or, most likely, an interaction between the 2. Regardless, prevention research directed at much younger children and their families is warranted, especially because successful preventive intervention studies by Robinson and colleagues15,19 and Gortmaker et al23,44 strongly suggest a causal relationship between school-age viewing and obesity and aggression. Our analyses also indicate that, like television viewing in general, the social gradients in television viewing emerge at an early age, increase, and persist into childhood. These viewing gradients may be among the earliest antecedents of social disparities in health, offering potential insight into “predisease pathways”45 and suggesting earlier opportunities to address these disparities.
Surprisingly, child care, maternal employment, and marital status were not among the strongest independent predictors of increased television viewing, which runs counter to the notion that television often serves as a “babysitter” for busy parents trying to juggle jobs, children, and taking care of a home. Similarly, one would expect a measure of the home environment to capture parenting style and therefore to correlate with the amount of television watched. However, the HOME score was not a strong predictor of television viewing. This could be attributable to 2 factors: first, the HOME-SF may be a less accurate measure of parenting style for children younger than 346; second, overall parenting style may not correlate with parental approaches to child television viewing. If a parent believes that television is beneficial, then letting an infant or toddler watch television may reflect a desire to do what is best for the child. The surprising positive association between HOME scores and viewing in the Youngest group may be reasonably explained by such parental beliefs. Indeed, 1 study found that the majority of parents of 0- to 35-month-olds believed that television could improve a child’s vocabulary,40 highlighting the need to study parental knowledge about and attitudes toward television. Future studies should focus on untangling the interactive effects of the home environment, parenting style, and child preference on child television viewing.
An important limitation of this study is the reliance on maternal response to single questions regarding weekday and weekend day television viewing. Studies comparing parental estimates with diaries or with direct observation (video) suggest that parents may overestimate their children’s time with television.47–49 However, when we analyzed children 2 to 11 years of age in 1998 and compared our (weighted) findings to data collected using the Nielsen People Meter in 1999,29 NLSY mothers reported only a half-hour more per day, on average, a difference that may be explained by NLSY mothers’ inclusion of videos. In addition, parental estimates of television viewing in older children were used in the randomized, controlled trials that linked television with obesity,19 indicating that parental estimates have some predictive validity.
Another limitation is that we do not know how mothers defined “watching television” for their infants and toddlers. This is particularly an issue for the Youngest sample. Although 6-month-olds will attend to television roughly half the time that it is on,50 qualitative observation of infants in a small study suggested that they look at television continuously for only short periods of time.51 By 15 months, children can imitate what they see,52,53 and by 18 months attention to television can last as long as 30 minutes,51 but more information on the early development of television viewing is needed. The maternal estimates are further limited by the fact that viewing was reported in whole numbers. Any child who views <1 hour a day was reported as watching no television, and older children who watch television between 2 and 3 hours per day may have been reported as 2 hours per day. Both could result in some underestimation relative to the AAP viewing guidelines. Additional work, based on both detailed diaries and direct observation, is clearly needed to confirm our results.
A final limitation is that the NLSY sample design makes it difficult to generalize the findings to all young children. The survey oversampled disadvantaged groups, and the requirements for our sample meant that no mothers were younger than 22 years at delivery, and most were older than 30. It is not clear whether or in what direction these sample characteristics might bias the results.
This study could not address the content of children’s viewing. The relative consumption by and effects on infants and toddlers of educational versus entertainment television are not known. A study of the relationships between preschool television viewing and adolescent achievement, behavior, and attitudes found that the effects of television depended on the content of the programs viewed.54 Studies of elementary school children, however, have shown positive effects of reducing television without reference to the quality of programs viewed.15,19,23
A substantial number of children begin watching television at an earlier age and in greater amounts than the AAP recommends. Furthermore, these viewing patterns persist into childhood, when the direct adverse effects of television are better documented. Important research questions remain regarding television program content and possible direct effects on infants and toddlers. Nevertheless, these findings should encourage parents and pediatricians to discuss young children’s television viewing (and beneficial alternatives) and should alert researchers to the potential window of opportunity for preventive interventions before age 2.
- ↵Roberts DF, Foehr UG, Rideout VJ, Brodie M. Kids & Media @ the New Millennium. Menlo Park, CA: The Henry J. Kaiser Family Foundation;1999
- Bernard-Bonnin A-C, Gilbert S, Rousseau E, Masson P, Maheux B. Television and the 3- to 10-year-old child. Pediatrics.1991;88 :48– 54
- ↵Hofferth SL. Healthy Environments, Healthy Children: Children in Families. Ann Arbor, MI: Institute for Social Research;1998
- ↵Strasburger VC, Donnerstein E. Children, adolescents, and the media: issues and solutions. Pediatrics.1999;103 :129– 139
- Van Evra JP. Television and Child Development. 2nd ed. Mahwah, NJ: Lawrence Erlbaum Associates;1998
- Zuckerman DM, Zuckerman BS. Television’s impact on children. Pediatrics.1985;75 :233– 240
- ↵Singer JL, Singer DG. Barney & Friends as entertainment and education: evaluating the quality and effectiveness of a television series for preschool children. In: Asamen JK, Berry GL, eds. Research Paradigms, Television, and Social Behavior. Thousand Oaks, CA: Sage Publications;1998:305– 367
- ↵Centerwall BS. Television and the development of the superego: pathways to violence. In: Chiland C, Young JG, eds. Children and Violence. Northvale, NJ: Jason Aronson;1994:178– 197
- ↵Andersen RE, Crespo CJ, Bartlett SJ, Cheskin LJ, Pratt M. Relationship of physical activity and television watching with body weight and level of fatness among children: results from the Third National Health and Nutrition Examination Survey. JAMA.1998;279 :938– 942
- ↵Clarke AT, Kutz-Costes B. Television viewing, educational quality of the home environment, and school readiness. J Educ Res.1997;90 :279– 285
- ↵American Academy of Pediatrics, Committee on Public Education. Media education. Pediatrics.1999;104 :341– 342
- ↵American Academy of Pediatrics, Committee on Public Education. Children, adolescents, and television. Pediatrics.2001;107 :423– 426
- ↵2000 Report on Television: The First Fifty Years. New York, NY: Nielsen Media Research;2000
- ↵Research CfHR. NLSY79 User’s Guide: A Guide to the 1979–1998 National Longitudinal Survey of Youth. Columbus, OH: Ohio State University;1999
- ↵Weissman MM, Sholomskas D, Pottenger M, Prusoff BA, Locke BZ. Assessing depressive symptoms in five psychiatric populations: a validation study. Am J Epidemiol.1977;106 :203– 214
- ↵Dalzell VP, Msall ME, High PC. Parental attitudes of television and videocassette viewing of children aged birth to 36 months. J Dev Behav Pediatr.2000;21 :390
- ↵Sampson RJ, Raudenbush SW, Earls F. Neighborhoods and violent crime: a multilevel study of collective efficacy. Science.1997;277 :918– 924
- ↵National Research Council. Predisease pathways. In: Singer BH, Ryff CD, eds. New Horizons in Health: An Integrative Approach. Washington, DC: National Academy Press;2001:25– 44
- ↵Mariner CL, Zaslow MJ, Sugland B. Factor Structure and Predictive Validity of the HOME-Short Form for Three Racial/Ethnic Groups in the National Longitudinal Survey of Youth-Child Supplement. Methods Working Paper #98.2. Washington, DC: Child Trends, Inc;1998
- ↵Borzekowski DLG, Robinson TN. Viewing the viewers: ten video cases of children’s television viewing behaviors. J Broadcasting Electronic Media.1999;43 :506– 528
- ↵Lemish D. Viewers in diapers: the early development of television viewing. In: Lindlof TR, ed. Natural Audiences: Qualitative Research of Media Uses and Effects. Norwood, NJ: Ablex;1987:33– 57
- Copyright © 2002 by the American Academy of Pediatrics