Predicting US Infants' and Toddlers' TV/Video Viewing Rates: Mothers' Cognitions and Structural Life Circumstances

Abstract

There has been rising international concern over media use with children under two. As little is known about the factors associated with more or less viewing among very young children, this study examines maternal factors predictive of TV/video viewing rates among American infants and toddlers. Guided by the Integrative Model of Behavioral Prediction, this survey study examines relationships between children's rates of TV/video viewing and their mothers' structural life circumstances (e.g., number of children in the home; mother's screen use), and cognitions (e.g., attitudes; norms). Results suggest that mothers' structural circumstances and cognitions respectively contribute independent explanatory power to the prediction of children's TV/video viewing. Influence of structural circumstances is partially mediated through cognitions. Mothers' attitudes as well as their own TV/video viewing behavior were particularly predictive of children's viewing. Implications of these findings for international efforts to understand and reduce infant/toddler TV/video exposure are discussed.

Researchers and policymakers have expressed growing concern over the international rise in the time children spend with screen media. Particular focus has scrutinized infants' and toddlers' escalating rates of consumption, leading governmental and health organizations in several nations to enact new policies and legislation. Health departments from the United States, Australia, and Canada officially discourage screen media use with children under the age of two years (AAP, 2011; ADHA, 2011; Lipnowski & LeBlanc, 2012; Sigman, 2012), and in 2008 the French High Audiovisual Council banned French television channels from airing programming targeting children under age three (Alleyne, 2008).

Despite the growing use of screen media with young children worldwide, research suggests that children glean very little from these sources before their second birthday (Courage & Setliff, 2010; Krcmar, 2010). For example, despite the fact that many videos for infants/toddlers claim to teach language skills to young children (Fenstermacher et al., 2010), most research conducted with commercially available baby videos has evidenced very little or no word learning from these sources among infants and toddlers (e.g., Krcmar, 2011; Robb, Richert, & Wartella, 2009). Furthermore, findings show associations between babies' media use and disruptions in healthy activities such as sleep (Taveras, Rifas-Shiman, Oken, Gunderson, & Gillman, 2008), interaction with caregivers (Christakis et al., 2009; Courage, Murphy, Goulding, & Setliff, 2010), and focused play behavior (Courage et al., 2010; Masur & Flynn, 2008).

Studies conducted in the USA have uncovered differences in infants' and toddlers' foreground viewing rates (that is, TV/video turned on explicitly for the child to watch) based on demographic factors. This research has found differences by child age (viewing increases steadily between 6 and 36 months of age), race/ethnicity (Black children spend more time viewing than their White and Hispanic peers), and parents' education level (children of less educated parents watch more; Anand & Krosnick, 2005; Bank et al., 2012; Certain & Kahn, 2002; Zimmerman, Christakis, & Meltzoff, 2007). Previous research has examined additional predictors, such as maternal depression and parental attitudes toward and restriction of infant/toddler TV/video viewing (e.g., Bank et al., 2012; Barr, Danziger, Hilliard, Andolina, & Ruskis, 2010). However, prior studies have not examined potential mediators operating between demographic predictors and babies' media exposure; that is they have not explored mechanisms which might explain relationships between demographic variables and TV/video-viewing. Nor have they investigated simultaneously a wide range of factors associated with TV/video viewing rates among very young children. This may be due in part to a lack of theory driving the design and interpretation of research in this area.

This study examines the relationships between American infants' and toddlers' TV/ video viewing time and their mothers' demographics (e.g., race/ethnicity), structural life circumstances (e.g., number of children; mother's screen use), and cognitions (e.g., attitudes; norms). An additional goal of this study is to determine the extent to which mothers' cognitions mediate relationships between the structural circumstances of mothers' lives and their infants'/toddlers' estimated TV/video viewing. In doing so, this research evaluates the Integrative Model of Behavioral Prediction and a model of mothers' structural life circumstances as competing explanations for infant/toddler TV/video viewing.

Alternative Models for Explaining Young Children's TV/Video Viewing

Integrative Model

Well-established as a powerful model for predicting behavior in a vast number of fields, the Integrative Model of Behavioral Prediction (IM; Fishbein & Ajzen, 2010) contends that the best way to predict people's behavior is to first understand their intentions to perform or not perform that behavior. Intention, in turn, is determined by an individual's attitudes, perceived social normative pressure, and/or perceived behavioral control (PBC) regarding the behavior. One's attitudes, perceived normative pressure, and perceptions of behavioral control are respectively shaped by his/her underlying beliefs about the expected outcomes from performing the behavior, the perceived expectations of influential social figures regarding the behavior, and the perceived ability or insurmountable obstacles to performing the behavior. As shown in Figure 1, the theory contends that all other “background” factors (e.g., demographics; life circumstances) would only influence individuals' behavior by first impacting underlying beliefs, which impact general cognitions, and so-on through the model.

Figure 1. The integrative model of behavioral prediction (Fishbein & Ajzen, 2010).

It is likely—as the IM would contend, that mothers' use of TV/videos with their young children is influenced by some combination of their attitudes, perceived normative pressure, and PBC. International research using this theory has shown that mothers' cognitions impact other aspects of their childcare, including decisions regarding breastfeeding and children's diets (Khoury et al., 2005; McMillan et al., 2009; Zhang, Shi, Chen, Wang, & Wang, 2009). For example, Khoury et al. (2005) found that mothers with stronger attitudes that breast-feeding was healthier for their baby and not unpleasant were more likely to breast-feed their own baby, as were those who perceived support from medical professionals and family members.

A mother's attitude about whether young children's media use is mostly good or mostly bad may be formed innumerable ways, such as her own experiences growing up with media, information encountered from doctors or news stories, or marketing messages from media producers. As she interacts with family members, friends, and others, she may perceive support or disapproval of media use from these sources, a form of normative pressure named “injunctive normative pressure” within the IM. Contact with other mothers likely provides her with a sense of the extent to which others like her are using TV/videos with their young children, labeled “descriptive normative pressure” in the context of the IM. Finally, a mother's consideration of her skills, abilities, and life circumstances likely impacts her sense of control over the time her child spends with TV/videos PBC. However, it is difficult to predict a priori which of these cognitive constructs will be most predictive of children's viewing time (Fishbein & Ajzen, 2010).

Research Question 1: Which of mothers' cognitions (i.e., attitudes, perceived normative pressure or PBC) will be most predictive of children's TV/video viewing?

Structural Life Circumstances

In the case of young children's TV/video viewing, it is possible that this principle of “theoretical sufficiency”—that the influence of background factors on behavior should be fully mediated through cognitions—is unfounded. It may be that the daily milieu of their lives ultimately determines the extent of children's viewing, regardless of mothers' beliefs about that viewing. That is, some structural aspects of parents' lives may directly constrain or encourage opportunities for children to view TV/videos.

In particular, structural circumstances may boost or decrease young children's viewing time via several means. The first is by increasing families' access to TV/video-viewing sources or providing alternative means for entertaining the child. For example, high levels of access to screen media and relatively low access to toys could foster greater TV/video use with young children, regardless of a parent's attitudes, perceived norms, or feelings of control over that viewing. Conversely, those with more rooms in their homes and no television set in the child's bedroom may find it easier to entertain their infant/ toddler away from the television set. Mothers who spend a lot of time viewing their own TV/video programming may not leave as much time to put on child-directed programming for their infant/toddler.

Structural life circumstances may also impact infant/toddler TV/video-viewing by alleviating or placing additional burden on mothers' time and energy demands. Mothers parent their young children in a variety of structural circumstances that may limit or boost the time and energy they have available. Those who are single-parenting or have multiple children may face greater demands that make it difficult to keep from using TV/video with their child. Mothers who work may also have more limited time and energy to devote to TV/ video-free activities with the child; or, conversely, may have childcare and household help that alleviate some parenting burdens and free up time that they can spend with their children. Finally, those with older toddlers may find it more exhausting and difficult to keep them entertained and occupied, compared to younger infants.

In sum, various realities of mothers' lives may impact their access to TV/video-viewing sources, alternative means for entertaining their children, and the time and resources they have available to devote to their children, thereby constituting either barriers to avoiding TV/video use with the child or providing alternatives to media use. As such, TV/video-use cognitions and TV/video-use behaviors may be inconsistent among some mothers due to the structural realities of their lives.

Research Question 2: Which variables regarding mothers' structural life circumstances are associated with infants' and toddlers' TV/video viewing?

Research Question 3: Are mothers' structural life circumstances directly associated with children's TV/video viewing, or are relationships mediated through the IM constructs?

Methods

This study consists of a cross-sectional survey of mothers with children between 3 and 27 months old. The survey was conducted online with an instrument largely reflecting the integrative model survey design outlined by Fishbein and Ajzen (2010), with additional items to measure mothers' structural life circumstances and other items not reported here. Survey construction was based on results of an elicitation interview study with a separate sample of mothers (Vaala, 2011).

Participants

Table 1 contains descriptive information about the sample of mothers (N = 698). While more white/non-Hispanic (67.9 per cent) and married mothers (74.8 per cent) participated, the sample was quite diverse in education and household income. On average, mothers watched 18.4 hours of TV/videos weekly (SD = 12.3). Their target children were 14.6 months old on average (SD = 6.11). Half of the target children were girls (50.6 per cent), and 42.7 per cent were first-born. One-fifth of children spent time in childcare (19.6 per cent). Of these children, 34.3 per cent spent 20 hours a week or less in childcare, 20.4 per cent spent 21– 30 hours a week, and 45.3 per cent spent 31 hours or more a week in childcare.

Table 1. Characteristics of the survey sample.

Age mean ± SD, years	28.5 ± 6.6
Race/ethnicity, n (%)
White/non-Hispanic	474 (67.9)
White/Hispanic	35 (5.0)
Black/African American	95 (13.6)
Asian	27 (3.9)
Othera	60 (8.6)
Refused/missing	7 (1.0)
Marital status, n (%)
Married/living as married	522 (74.8)
Separated/divorced/single	168 (24.1)
Refused/missing	8 (1.1)
Employment, n (%)
Full-time	134 (19.2)
Part-time	88 (12.6)
Homemaker	315 (45.1)
Student	49 (7.0)
Retired/disabled/unemployed	104 (14.9)
Refused/missing	8 (1.1)
Education, n (%)
No high school diploma	31 (4.4)
High school diploma/GED	190 (27.2)
Some college/Associate's	288 (41.2)
Four-year college degree	137 (19.6)
Graduate school	44 (6.3)
Refused/missing	8 (1.1)
Income, n (%)
Less than $10,000	74 (10.6)
$10,000-$29,000	192 (27.5)
$30,000-$49,000	179 (25.6)
$50,000-$74,000	113 (16.2)
$75,000 or more	97 (13.9)
Refused/missing	43 (6.2)
Target child age, n (%)
Less than 12 months	264 (37.8)
12 to 18.9 months	228 (32.7)
19 months or older	204 (29.2)

N = 698;

^aincludes participants of mixed race.

Procedure

Following Institutional Review Board study approval, participants were recruited through Survey Sampling International (SSI), which has a national panel of nearly one million US members. SSI recruits its members through various techniques online (e.g., banner ads), and provides compensation for study completion in the form of lottery drawings or points which can be cashed in for money. SSI sent recruitment emails to panel members who fit the criteria for study participation (i.e., women over age 18 living in the USA and parenting children between 3 and 24 months of age). Those who wished to participate clicked on a link in the email and were directed to the online survey. Data collection occurred over seven consecutive days in mid March, 2011.

Measures

Participants were asked how many children they had between 3 months and 24 months of age.¹ Those with more than one child in this age range were prompted to think of the child “whose name comes first in the alphabet.” All participants were asked to type the target child's first name into a given space so the computer could generate the name into subsequent questions (to encourage respondents to answer questions in regards to only the target child).

Demographics

Respondents gave demographic information, including their age, race/ethnicity, last grade or degree completed in school, household income, and marital status. They also described the number of rooms in their home (i.e., from 1–2 to 11 rooms or more).

Child's TV/video viewing

The following statement was displayed on the screen before the TV/video exposure items:

The following questions are about your child's television/video viewing – that is, television programs and videos made for children that you or someone else turn on with the intention that your child will watch it at least a little. Your child may watch these programs or videos on any type of a screen- such as a television, computer or portable DVD player.

Respondents reported the number of weekdays (0–5) the child typically watches at least some television or videos. Mothers of children that viewed on at least one day each week were asked to indicate how much time in a typical weekday the child spends viewing, choosing one of five response options (broken up in 2-hour increments between “less than 2 hours” and “8 hours or more”). They then chose one of four response categories to indicate a more detailed range of viewing time in a typical day (e.g., “less than 30 minutes;” “at least 30 minutes but less than 1 hours”). This series of questions was repeated to measure weekend viewing. The two-item structure was used to boost report accuracy and enable the report of specific time quantities in a less unwieldy way than would be possible with one item. Mothers were asked to estimate time their children viewed programming “that you or someone else turn on” in order to prompt them to include estimated time the child spent viewing while in the care of others as well.

The number of week days that the child watches was multiplied by the midpoint of the specific category of daily viewing (i.e., 45 minutes for the category “at least 30 minutes but less than 1 hours”). This step was repeated for weekend viewing. These two figures were summed to represent the child's estimated weekly TV/video viewing.

IM cognitive constructs²

Attitude

Mothers reported their attitudes toward letting their child watch TV/videos for more than an hour a day on at least several days each week on three dimensions: bad/ good; foolish/wise; harmful/beneficial. Each item was on a 7-point scale, with a 7 representing the most pro-TV/video attitude.

Perceived descriptive norms

Two items measured perceived descriptive norms (mothers' perceptions of how normative the behavior is among peers): (1) Most people like me with children 2 and under let their children watch television or videos for more than an hour a day on at least several days each week (7-point scale from “unlikely” to “likely”); (2) How many of the people who are most similar to you with children 2 and under let their children watch television or videos for more than an hour a day on at least several days each week? (5-point scale from “None or very few” to “Almost all or all”).

Perceived injunctive norms

Perceived injunctive norms (perceptions of whether important social referents support or disapprove of the behavior) were assessed through two items, including: (1) Most people who are important to me think I should let [child's name] watch television programs or videos for more than an hour a day on at least several days a week during the next month” (7-point scale from “true” to “false”); and (2) “Most people whose opinions I value think that I should let [child's name] watch television programs or videos for more than an hour a day on at least several days a week during the next month” (unlikely/likely).

Perceived behavioral control

Two items measured PBC over children's TV/video viewing: (1) “I am confident that I can control how much television- and video-watching [child's name] does during the next month” (7-point scale from “false” to “true”); and (2) “The amount my child watches television and videos during the next month is under my control” (7-point scale from “not at all” to “completely”).

Structural circumstance variables

Child age

Participants reported the target children's date of birth, from which age in months was calculated.

Family structure

Items assessed target children's birth orders and the number of additional children and adults in the home.

Home environment and media access

Respondents reported how many rooms contained television sets, whether the target child had his/her own bedroom, whether there was a television in the target child's bedroom, and how often a television was on during the day “even if no one is actually watching it.” Four additional questions inquired about children's access to non-traditional screens, including a screen built into the family car; a computer screen; a cellular telephone; and a portable DVD player. The final question in this section asked whether anyone ever recorded programs for the target child to watch via DVR or TiVo.

Respondents reported the number of toys in various categories to which target children had access, including: soft/cuddly toys; electronic toys; children's books; push/pull/ ride on toys; toys that make noise; stackable/insertable toys; children's videos; and videos made specifically for babies. These questions had seven response options, ranging from “none” to “more than 20.”

Mother's media use

In the same manner as children's exposure, participants reported the number of weekdays and weekend days they usually watched some TV/videos, and how much time on a typical weekday and weekend day they usually spent watching.

Employment status and childcare

Respondents reported their employment status (employed, unemployed, or homemaker), and whether the target child was “currently in any type of childcare.” Those who responded that their child was in childcare were asked whether the child ever watched TV/videos while in childcare.

Results

Analytic Plan

Following individual item and scale analyses, ordinary least squares (OLS) regression models were constructed to examine the research questions. The approach involved adding the structural life circumstance variables and IM constructs as sets of predictors in varying order to determine the extent of independent predictive power of each set in explaining variance in children's estimated TV/video-viewing time. The adjusted R ² values were compared to determine the extent of variance each set predicted, and individual standardized beta coefficients were compared to determine which constructs were particularly predictive. Finally, bootstrapping tests of mediation were used to assess the extent and statistical significance of mediation of each structural circumstance variable through the four IM cognitive constructs.

Home Environment and Media Access

Table 2 contains descriptive information about participants' homes. Few children reportedly watched video content on a cell phone (14.2 per cent) or TV mounted in the car (15.6 per cent). More than a fifth of children watched video content on a portable DVD player (21.9 per cent), while over a third viewed content on the computer (31.9 per cent). Nearly 40 per cent of mothers reported that their child watched content recorded via DVR or TiVo (38.3 per cent). A summative index was created of the number of reported sources for viewing video content available to the child described above (0–5 sources), representing children's access to non-traditional sources for viewing video content. The mean score on this index was 1.21 sources (SD = 1.35).

Table 2. Participants' home and media environments.

Number of rooms in home, n (%)
1–2	87 (12.5)
3–4	312 (44.7)
5–6	179 (25.6)
7 or more	120 (17.2)
Number of rooms with a TV, n (%)
0	9 (1.3)
1	134 (19.2)
2	243 (34.8)
3	193 (27.7)
4 or more	119 (17.0)
Child bedroom arrangement, n (%)
Own room	275 (39.4)
With parent(s)/guardian(s)	333 (47.7)
With sibling(s)	90 (12.9)
Child has bedroom TV, n (%)	238 (34.1)

Toy ownership items were recoded to represent the midpoint of the range of toys a given response (e.g., “5–10” = 7.5) to create interval-level variables, which were summed to create one toy index. The mean index score was 43.95 toys (SD = 22.36).

Children's TV/Video Viewing

The estimates of children's weekly TV/video viewing ranged from 0 to 68.25 hours per week, with a mean of 8.82 hours (SD = 10.86), and a median of 4.50 hours per week. Due to the lack of normality and the high skew (i.e., skew = 2.12, SE = 0.09) of the estimates, this variable was transformed by adding 1 and then taking the square root for subsequent analyses.

IM Constructs

The skewness and kurtosis coefficients were assessed for individual integrative model items. The two items measuring PBC were particularly skewed toward high perceived control (skew = −2.03 and −2.04, respectively). In keeping with the IM and its appropriate analysis, these items were not transformed despite deviations from normality.

Next, internal consistencies for the items intended to form integrative model scales were tested. The three attitude items had a Cronbach's alpha of 0.94. They were averaged together to create an estimate of participants' general attitudes toward letting the target children watch more than an hour of TV/videos a day for at least several days each week (higher values indicate more favorable attitudes). This scale had a mean value of 3.93 (SD = 1.51) and a median of 4.00 (on a 7-point response scale).

The two injunctive normative pressure items were correlated at r = .87 (p < .01). They were averaged together to form an estimate of participants' perceived injunctive normative pressure (higher values indicate greater pressure to not let child view TV/videos). The mean of this scale was 3.40 (SD = 1.85; 7-point scale) and the median was 3.50.

The two descriptive normative pressure items were correlated at r = .74 (p < .01). These items were standardized due to varying response scales (i.e., 5-point and 7-point), and then averaged together to form a single estimate of descriptive normative pressure (higher values represent greater perception that other mothers let their children view TV/ video). This scale had a mean value of 0 (SD = 0.93) and the median was 0.25.

Finally, the two items that assessed mothers' PBC were correlated at r = .78 (p < .01). They were averaged together to create an estimate of mothers' perceived control over their children's viewing (higher values represent greater perceived control). The scale mean was 6.40 (SD = 1.02; 7-point scale) and the median 7.00.

Research Question 1: Predictive Value of Cognitions

The transformed estimate of children's TV/video viewing was significantly related to mothers' attitudes (r = 0.44, p < .01), injunctive norms (r = 0.37, p < .01), descriptive norms (r = 0.28, p < .01), and PBC (r =– 0.11, p < .01).

Regression analysis examined the predictive values of the IM constructs in accounting for children's TV/video viewing estimates. The model was significant and accounted for 22 per cent of the variance children's viewing estimates (F(4, 685) = 28.39, p < .01; adj. R ² = 0.22). More positive attitudes among mothers predicted higher viewing estimates (β = 0.35, p < .01). Lower-perceived behavioral control was related to more viewing (β = –0.14, p < .01). Descriptive norms had a significant positive relationship with children's viewing (β = 0.09, p < .05), and injunctive normative pressure was a marginally significant positive predictor (β = 0.08, p = .09).

Covariates

Correlations were used to test associations with the four continuous or ordinal-level variables: (1) mother's education level; (2) annual household income (3) mother's age; and (4) number of rooms in the home. Only mother's education was significantly associated with children's TV/video viewing (r = −0.08, p = 0.05).

Individual OLS regressions were used to test for differences in children's viewing based on nominal-level demographic variables, including: (1) mother's race/ethnicity (i.e., dummy variables for Black/non-Hispanic and “other,” compared to White/non-Hispanic) (2) child's gender (i.e., dummy variable for female). The results indicated no differences by race/ ethnicity (F(2, 689) = 1.47, p = 0.23), or child's gender F(1, 696) = 0.01, p = 0.98).

A preliminary OLS regression was conducted containing all potential demographic variables as predictors of the transformed viewing estimate. This was done to ensure that no significant predictors were omitted due to intercorrelations suppressing bivariate relationships. The model was marginally significant and accounted for 1 per cent of variance in viewing estimates (adjusted R ² = 0.01; F(8, 652) = 1.88, p = .06). Two variables were significant predictors: mother's education (β = −0.12, p < .01) and mother's age (β = 0.10, p < .05). The number of rooms in the home was a marginally significant negative predictor (β = −0.08, p = 0.06). These three variables were used in subsequent models as covariates.

Research Question 2: Predictive Value of Structural Circumstances

Correlational analyses were conducted between children's estimated TV/video viewing (transformed) and the seven continuous or ordinal structural circumstance variables as well as the 10 nominal-level structural circumstance variables. As shown in Table 3, eleven relationships were significant.

Table 3. Zero-order correlations between structural circumstance variables and children's TV/video viewing estimates.

Structural Circumstance Variable	Association with Transformed Viewing Estimate (r)
Continuous/ordinal
Toy index	0.16**
Number of rooms with TVs	0.11**
Non-traditional sources of video	0.25**
Child's age	0.19**
Mother's own TV/video viewing time	0.27**
Number of children in the home	0.07†
Number of adults in the home	0.05
Nominal
Use of childcare	0.08*
Use of childcare with TV	0.22**
TV set in child's bedroom	0.17**
Mother is employeda	0.10**
Mother is unemployeda	0.12**
> 1 child 3–14 months old	0.10**
At least one other adult in the home	0.06
Mother is single	0.06
Child has own bedroom	0.02
Child is first-born	0.02

Note. values represent bivariate Pearson correlation coefficients unless otherwise noted.

^avalues were obtained by entering both dummy variables in a multiple regression analysis predicting children's transformed viewing estimate (homemaker was omitted as the comparison category).

^**p < .01;

^*p < .05;

^†p < .10.

A linear regression analysis was conducted to determine the structural circumstance variables significantly predictive of children's TV/video viewing when all were included, and to generate an estimate of the predictive power of all structural variables as a set. Together the structural circumstance variables accounted for 25 per cent of the variance in children's estimated TV/video viewing (adjusted R ² = 0.25; F(16, 677) = 14.86, p < .01). Seven variables had significant positive relationships with viewing estimates, including the toy index (β = 0.12, p < .01); number of non-traditional sources for the child's video-viewing (β = 0.16, p < .01); having a television in the child's bedroom (β = 0.12, p < .01); mother's unemployment (β = 0.11, p < .01)³; target child age (β = 0.23, p < .01); mother's own TV/video viewing (β = 0.27, p < .01); and childcare TV/videos use (β = 0.25, p < .01). Having another child 3–24 months of age was also marginally associated with higher TV/video viewing (β = 0.07, p = 0.07). Only the use of childcare was significantly associated with less weekly TV/video viewing for target children (β = −0.17, p < .01).⁴ These nine structural variables were included in subsequent models as predictors, as was mother's status as employed (so that unemployment status would continue to be compared with homemakers, the omitted nominal-level category).

Next, a hierarchical OLS regression analysis tested the extent of variance in the transformed estimate of children's TV/video viewing accounted for by the demographic and structural circumstance variables. The three demographic covariates were entered together in the first step, followed by the ten significant and marginally significant structural circumstance variables in the second step.

The standardized and unstandardized coefficients obtained from the analysis are displayed in Models 1 and 2a within Table 4. The addition of the structural circumstance variables significantly increased the variance accounted for by the model (ΔR ² = 0.25; p < .01). Each of the structural circumstance variables was a significant or marginally significant predictor. Mother's own time spent viewing TV/videos was the strongest predictor of children's viewing (β = 0.26, p < .01).

Table 4. Mothers' structural life circumstances as predictors of children's estimated TV/video viewing.

	Model 1		Model 2a		Model : 2b		Model 3
	B(SE B)	β	B(SE B)	β	B(SE B)	B	B(SE B)	β
Mother's education	−0.10 (0.04)	−0.08†	−0.06 (0.04)	−0.06	0.001 (0.04)	0.001	0.002 (0.04)	0.002
Mother's age	0.02 (0.01)	0.09*	0.004 (0.01)	0.02	0.01 (0.01)	0.02	−0.01 (0.01)	−0.02
Number of rooms in the home	−0.10 (0.06)	−0.07†	−0.12 (0.06)	−0.08*	−0.05 (0.05)	−0.03	−0.09 (0.05)	−0.06†
Child's age			0.06(0.01)	0.23**			0.06(0.01)	0.23**
Mother is unemployeda(dummy)			0.43 (0.16)	0.10**			0.35 (0.15)	0.08*
Mother is employeda (dummy)			0.23 (0.12)	0.07†			0.09 (0.12)	0.03
>1 child 3–24 months old (dummy)			0.31 (0.18)	0.06†			0.28(0.17)	0.05†
Child is in childcare (dummy)			−0.61 (0.19)	−0.16**			−0.50 (0.17)	−0.13**
Non-traditional video source index			0.1 9 (0.04)	0.16**			0.09 (0.04)	0.08*
Toy index			0.01 (0.002)	0.14**			0.01 (0.002)	0.14**
Child is in childcare with TV (dummy)			1.23 (0.23)	0.24**			0.94 (0.22)	0.19**
Child has a bedroom TV (dummy)			0.27(0.11)	0.08*			0.18(0.11)	0.05
Mother's TV/video time			0.03 (0.004)	0.26**			0.03 (0.004)	0.20**
Attitude					0.34 (0.05)	0.34**	0.28 (0.05)	0.28**
Injunctive norms					0.07 (0.04)	0.09†	0.02 (0.04)	0.02
Descriptive norms					0.16 (0.07)	0.10*	0.14(0.06)	0.09*
Perceived behavioral control					−0.20 (0.05)	−0.13**	−0.17 (0.05)	−0.11**
R	0.12		0.52		0.48		0.61
Adj. R2	0.01		0.25		0.22		0.35

N = 685. ΔR² for Step 2a = 0.25 (p < .01); for Step 2b = 0.22 (p < .01); for Step 3 = 0.10 (p < .01).

^aHomemakers omitted as comparison category.

^**p < .01;

^*p < .05;

^†p < .10.

Research Question 3: Mediation of Structural Circumstances

The next set of analyses assessed how much predictive power the set of structural circumstance variables might add to the IM variables. First, a hierarchical OLS regression was conducted predicting the transformed viewing estimate. The first model step contained the three covariates. In the second step the four IM constructs were added, followed by the ten structural life circumstance variables in the third step.

The regression coefficients and R and R ² values from this analysis are displayed in Models 2b and 3 within Table 4 (although the coefficients in Model 3 were in a different order per the final regression analysis below, the coefficients from the third step of this analysis and the third step below are identical as they contain the same variables). The four IM constructs accounted for an additional 22 per cent of variance in the estimates of children's viewing (ΔR ² = 0.22, p < .01; see Model 2b). The full model was significant (F(17, 684) = 22.66, p < .01). The structural circumstance variables in the third step added an additional 14 per cent of variance accounted for by the model (i.e., full model adj. R ² = 0.35; step 3 ΔR ² = 0.14, p < .01).

A final hierarchical regression analysis examined how fully the integrative model constructs mediate associations between mothers' structural life circumstances and children's viewing estimates. In this analysis, the last two steps from the above analyses were reversed: first the structural variables were entered, then the IM variables. This permitted a clearer assessment of how much of the influence of the structural variables was mediated by the IM variables and how much was independent of them.

As shown in Model 3 within Table 4, the coefficients for seven structural circumstance variables were moderately weaker after adding the IM constructs. In Model 2a the structural variables account for 25 per cent of the variance in viewing without the IM variables included. As reported above, they account for an add-on 14 per cent of the variance when the IM variables were already included in the model. Thus, crudely, (1.00–14/25) or 44 per cent of the association of the structural variables with child viewing was mediated by the four IM variables.⁵

Bootstrapping analyses tested the significance of indirect paths from each circumstance variable to viewing estimates through the IM constructs (see Preacher & Hayes, 2008).⁶ Each analysis tested the indirect path of a structural circumstance variable through the four IM constructs, controlling for the covariates and other circumstance variables. Table 5 contains the indirect point estimates for the structural circumstance variables through each cognitive construct, as well as the estimate of mediation through the four IM constructs combined. Each point estimate was divided by the original unstandardized regression coefficient from Table 4 (Model 2a). The resultant ratios represent the estimated proportion of the total relationship between each circumstance variable and estimated TV/video viewing that is mediated by the given construct. The confidence intervals around the point estimates were examined to determine which indirect paths were significantly different from zero.

Table 5. Indirect paths between structural circumstance variables and children's TV/video viewing through mothers' cognitions.

Structural Variable a (Original Effect)	Total Indirect Paths Point Estimateb (Proportion of B)c	Attitudes Point Estimateb (Proportion of B)c	Injunctive Norms Point Estimateb (Proportion of B)c	Descriptive Norms Point Estimateb (Proportion of B)c	Perceived Control Point Estimateb (Proportion of B)c
Child's age (0.06)	0.006 (0.10)	0.003 (0.05)	0.0003 (0.005)	0.001 (0.02)	0.002 (0.03)
Mother is unemployed (0.43)	0.023 (0.05)	0.041 (0.10)	−0.003 (0.01)	−0.018 (0.04)	0.003 (0.01)
Mother is employed (0.23)	0.116 (0.50)	0.043 (0.19)	0.005 (0.02)	0.026 (0.11)	0.041 (0.18)
>1 child between 3–24 months (0.31)	0.034 (0.11)	0.021 (0.07)	0.005 (0.02)	0.038 (0.12)	−0.030 (0.10)
Child is in childcare (−0.61)	−0.111 (0.18)	−0.071 (0.12)	−0.008 (0.01)	−0.029 (0.05)	−0.003 (0.005)
Non-traditional video source index (0.19)	0.099 (0.52)	0.077 (0.41)	0.007 (0.04)	0.005 (0.03)	0.010 (0.05)
Toy index (0.01)	0.000 (0.00)	−0.0003 (0.03)	0.0001 (0.01)	0.0005 (0.05)	−0.0004(0.04)
Child is in childcare with TV (1.23)	0.290 (0.24)	0.165 (0.13)	0.013 (0.01)	0.026 (0.02)	0.087 (0.07)
Child has a bedroom TV (0.27)	0.091 (0.33)	0.097 (0.36)	0.005 (0.02)	0.012 (0.04)	−0.022 (0.08)
Mother's TV/video time (0.03)	0.007 (0.23)	0.007 (0.23)	0.0004 (0.01)	0.001 (0.03)	−0.001 (0.03)

N = 685.

^aValues represent B values for structural circumstance variables in Table 4 Model 2a.

^bValues represent indirect point estimates from bootstrapping analyses with 1,000 samples, controlling for other structural circumstance variables and covariates.

^cValues represent the ratio of indirect relationship point estimates to the original B value (i.e., proportion of total relationship that is mediated). Bold values indicate confidence intervals that do not contain zero.

The structural circumstance variables most strongly mediated by the IM variables were the index of non-traditional sources for viewing video content (estimated mediation = 52 per cent) and mother's status as employed (50 per cent). Three other variables were moderately mediated, including having a television set in the child's bedroom (33 per cent), having childcare arrangements that use TV/videos (24 per cent), and mothers' TV/video viewing time (23 per cent). Most of the strongest indirect paths were through attitude.

Discussion

This study examined the predictive value of mothers' structural life circumstances and cognitions in accounting for their infants' and toddlers' TV/video viewing rates, as well as the degree to which those cognitions mediate the role of structural life circumstances. The results indicate that mothers' structural life circumstances account for nearly as much variance in young children's TV/video viewing as do the cognitive constructs of the integrative model. Furthermore, while there is some evidence of mediation through the cognitive constructs, the findings suggest that numerous structural circumstance factors also directly impact young children's TV/video viewing rates.

The results suggest that while the integrative model of behavioral prediction does operate relatively well in predicting young children's TV/video viewing from their mothers' cognitions, the model's constructs are not sufficient for predicting viewing rates. In particular, the results do not support the model's “principle of theoretical sufficiency,” which contends that the impact of exogenous factors on behavior is mediated fully through cognitions (e.g., see Ajzen & Albarracin, 2007; Hennessy et al., 2010). However, one possible concern is that the IM measures address cognitions about the mother's behavior (e.g., attitude towards permitting the child to watch) but the outcome in the analysis is her report of how much watching the child does. Her actions and the child's watching quantity may not overlap completely.⁷

In this study, children's viewing rates did not vary strongly across demographic lines, particularly when structural life circumstance or IM cognitive constructs were included in analyses. Number of rooms in the home did retain marginal significance as a predictor across models. It may be that this variable is an artifact of differences by socioeconomic status, or that it may be more accurately considered a structural circumstance. It is also possible that race/ethnicity would have predictive value among a more diverse sample of mothers, as other research has found differences in children's viewing based on parents' race/ethnicity (e.g., Anand & Krosnick, 2005; Certain & Kahn, 2002).

Mothers' attitudes regarding their children's TV/video viewing were the strongest predictors of children's concurrent weekly TV/video viewing. That is, the more they felt children's TV/video use was good, wise, and beneficial, the higher their reports of children's actual viewing. The principles of the IM contend that attitudes are driven by discrete beliefs regarding the likelihood of various favorable and unfavorable outcomes associated with performing the behavior. Future research should examine the range and predictive value of these discrete beliefs.

Mothers generally expressed high perceptions of their own control over children's TV/video viewing. This was surprising given the range of challenging structural circumstances many mothers negotiate in the context of their parenting. Despite stunted variance in PBC among mothers in this sample, this construct was significantly related to children's viewing estimates. Because these data are cross-sectional, the extent to which relationships between children's viewing and mothers' cognitions are truly causal cannot be determined. As it is possible that mothers' feelings of personal control over their children's TV/video viewing impact their viewing rates, the possibility of a direct effect between mothers' perceived control and infants' and toddlers' time spent viewing TV/videos should be investigated in future research.

Mothers' perceived descriptive norms were also significantly predictive of children's viewing, such that those who perceived that many mothers like themselves use TV/videos with their children tended to have children who viewed more. Notably, the other normative dimension, perceived injunctive normative pressure, was not significantly predictive of children's TV/video viewing. Thus, when deciding the appropriate TV/video diet for their infants/toddlers, mothers seem to be more influenced by what other mothers are doing, rather than how others in their lives want them to act. Given that these data are correlational, however, it is also possible that these relationships are not causal.

Additional findings of note in this study are the mediation patterns among mothers' structural life circumstances. Though none of the variables was fully mediated, there was evidence of moderate partial mediation for some. Mothers' employment status was most strongly mediated by PBC and descriptive normative pressure. This makes sense as, for many families, the child must spend time in the care of others (e.g., spouse; daycare) while the mother is at work. Mothers would likely feel less control over their children's TV/video viewing while the children are not in their direct care. In addition, working mothers may also be busier than homemakers and unemployed mothers and feel that they need to use TV/videos to occupy the child in order to accomplish all of their responsibilities. Furthermore, employed mothers may have coworkers and friends in similar situations as working mothers. In their circle of peers, then, television and video use with young children may be considered a normative behavior, causing the indirect relationship through descriptive norms.

Conversely, the relationships with four factors reflecting children's access to TV/videos were most strongly mediated by mothers' attitudes. Mothers may develop attitudes that are in-line with their behavior to some extent. For example, having increased access to non-traditional viewing sources may help keep young children quiet and occupied in a greater array of settings. These favorable experiences may boost mothers' positive attitudes toward the use of TV/videos with their infants/toddlers in-turn.

Mediation of mothers' own TV/video-viewing time through attitudes is similarly logical. It is reasonable to expect that a mother's attitude toward her child's viewing is likely correlated with her attitude towards her own viewing and a major influence on her own viewing. One possibility with regards to childcare is that mothers may be told or infer from childcare representatives that viewing TV/video programs can be helpful for infants and toddlers, thereby boosting their attitudes. It is also possible that mediation relationships occur in reverse order, and that mothers' cognitions regarding infant/toddler TV/video viewing influence various structural circumstances over which they have control. Longitudinal data is needed to establish the temporal order of relationships.

This study offers important methodological insights to future efforts to understand and reduce infants'/toddlers' screen media exposure. These findings illustrate the importance of incorporating measures of home and family contexts when seeking to understand children's media exposure. Among these US infants and toddlers, structural life circumstances together with mothers' cognitions about TV/video viewing accounted for 35 per cent of variance in children's estimated viewing rates. Children do not encounter screen media in a vacuum; rather their viewing is a part of the larger family ecology in which youth behavior is shaped by their environment (see also Jordan et al., 2010).

Due to the cross-sectional nature of the data it is not possible to confirm the causality or temporal order of relationships. Furthermore, there may be important determinant factors boosting or reducing infant/toddler viewing that were not included here, such as maternal depression, child temperament, or additional structural factors. Finally, these predictors may differently predict children's exposure to different categories of programming (e.g., educational; entertainment), though content differences are not examined in this study. In fact, the content that children view has been found to differ as children transition from infancy (Barr et al., 2010), and may impact the outcomes associated with their viewing. Structural and cognitive predictors may differently predict children's exposure to different categories of programming (e.g., educational; entertainment).

Still, this study suggests that the integrative model of behavioral prediction is a useful theory for approaching the examination and reduction of young children's screen media consumption, though these findings do not provide a universal roadmap. While the present results offer specific insights regarding US mothers with infants/toddlers, the theory contends that each combination of discrete population and behavior must be studied on its own terms (Fishbein & Ajzen, 2010). The predictive constructs uncovered here may suggest appropriate targets of an intervention aimed at American mothers' TV/video use with their infants and toddlers, but may not reflect the most predictive constructs for American fathers or parents from other countries. In light of the international interests in reducing young children's screen time, future work should explore the cognitive and structural facets most predictive of this behavior among additional populations.

Biographies

Sarah Vaala (author to whom correspondence should be addressed), PhD is a Martin Fishbein Post-Doctoral Fellow at the Annenberg Public Policy Center at the University of Pennsylvania. From 2011-2012 she was the Research Fellow at the Joan Ganz Cooney Center at Sesame Workshop. Sarah is interested in the educational and health implications of screen media in the lives of children and adolescents, as well as the ways caregivers perceive and make decisions about their children's media use. Annenberg School for Communication, University of Pennsylvania, 202 S. 36th Street, Philadelphia, PA 19104. Tel: 215-746-0043. E-mail: svaala@asc.upenn.edu

Robert Hornik, PhD is Wilbur Schramm Professor of Communication and Health Policy at the Annenberg School for Communication, University of Pennsylvania. He is also Director of the Center of Excellence in Cancer Communication Research. His major research interests include health effects of mass media exposure and evaluation of health communication campaigns. E-mail: rhornik@asc.upenn.edu