A total of 206 students participated: 90 at the public university and 116 at the private university. Complete data were collected from 193 participants. An additional 2 participants were eliminated due to extreme ratings (z-scores > ±3) on some of the food attitude measures. The sample was 70% female and 60% white, non-Hispanic. Participants also included 15% Hispanic, 14% Asian, and 11% black or other ethnicity. Ages ranged from 17 to 44 years (M = 19.1, SD = 2.4).
To assess overweight status, we utilized a measure from the U.S. Youth Risk Behavior survey (Centers for Disease Control, 2004
) and asked participants to indicate their current weight status. According to their own assessment, approximately two-thirds of the sample (68%) indicated that they were normal weight, 17% reported being underweight (slightly or very), and 15% reported being overweight. Overweight status of 4-year college students, as assessed by self-reported height and weight in other studies, ranges from 20 to 30% (Huang, et al., 2003
; Lowry, et al., 2000
). The somewhat lower incidence of overweight in our sample may have been due to inaccurate assessments of weight status and could limit our ability to associate weight status with the other variables in our analysis. As a result, we do not attempt to make conclusions about the relationship between television viewing and obesity.
Food attitudes and diet
Participants reported consuming more healthy foods (on average, 4 to 6 times per week of each healthy category) than unhealthy foods (average of 1 to 3 times per week), t(191) = 9.99, p < .001. Individuals' total consumption of healthy and unhealthy foods, however, were not correlated (r = −.05, p = .49) and indicate that higher consumption of healthy foods did not always indicate lower consumption of unhealthy foods, or vice versa.
As predicted, perceived taste was associated with reported consumption: healthy food consumption was correlated with higher taste ratings for healthy foods, r = .28, p < .001, and lower taste ratings for unhealthy foods, r = −.18, p = .02. In contrast, unhealthy food consumption was highly correlated with taste ratings for unhealthy foods, r = .43, p < .001, but not significantly related to taste ratings for healthy foods, r = −.12, p = .11.
In contrast, assessments of food healthiness were not consistently related to healthy and unhealthy consumption. Healthy food consumption was positively correlated with perceived healthiness of healthy foods, r
= .16, p
= .03, and negatively with perceived healthiness of unhealthy foods, r
= −.15, p
= .05. Perceived healthiness of healthy and unhealthy foods did not, however, predict unhealthy food consumption (r
= −.00, p
= .99; r
= .08, p
= .28, respectively). Higher perceived healthiness of healthy foods was also associated with higher taste ratings for healthy foods (r
= .35, p
< .001). In contrast, healthiness ratings for unhealthy foods (i.e., perceived un
healthiness) was not significantly correlated with unhealthy food taste (r
= .10, p
= .18). Although prior research with children suggests that perceived healthiness of a food may be negatively related to taste (Baranowski, et al., 1993
; Wardle & Huon, 2000
), these findings with young adults demonstrate that perceived taste and healthiness may be positively related for healthy foods, and not at all related for unhealthy foods.
Overall then, the findings largely supported H1a and H1b. The best predictors of healthy and unhealthy diet were perceived taste of healthy and unhealthy foods. In addition, nutrition knowledge (as assessed by higher healthiness ratings for healthy foods and lower ratings for unhealthy foods) was not related to perceived taste or consumption of unhealthy foods. Higher consumption of unhealthy foods was significantly associated only with more positive attitudes about the taste of unhealthy foods. In contrast, nutrition knowledge did predict increased perceived taste and consumption of healthy foods.
Food attitudes and level of advertising
The difference between attitude ratings for foods with more versus less advertising was significant for all levels of healthiness and all types of attitudes. As predicted in H1c, unhealthy foods with higher levels of advertising were perceived as more tasty than unhealthy foods with lower levels of advertising, t(199) = 8.09, p < .001. For healthy foods, however, we found the opposite relationship: Those with lower levels of advertising were rated as more tasty than those with more advertising, t(199) = 4.27, p < .001.
Unexpectedly, we also found that different levels of advertising were associated with differences in perceived healthiness. In the pre-test, we had identified foods with similar levels of healthiness (according to our pre-test sample) that differed only by level of advertising. Among our young adult participants, however, both healthy and unhealthy foods with higher levels of advertising were perceived to be less healthy, (t(199) = 9.78, p < .001 and t(199) = 6.50, p < .001, respectively). We cannot determine from these data whether advertising caused these differing evaluations of taste and healthiness, or whether the differences were due to the specific foods evaluated. The results do suggest, however, that individuals could infer a relationship between advertising, healthiness and taste. For example, they might believe that healthy foods that are advertised do not taste as good as unadvertised healthy foods, or that advertised foods, in general, are not as healthy as unadvertised foods. If confirmed in further research, these findings indicate potential difficulties to effectively promote healthy food consumption through advertising.
Inter-relationship between television experience, parental mediation and diet
To test H2 and H3, we used structural equation modeling (SEM) to identify the relative contribution of television exposure and parental mediation on perceived food taste and relative healthy and unhealthy diet. Correlational analyses revealed that television exposure and parenting factors did not exhibit similar relationships to healthy and unhealthy food ratings and diet (see ). Therefore, to simplify the analysis, we tested separate healthy and unhealthy models. In addition, we chose to include taste, but not perceived healthiness, ratings in our models. As discussed earlier, healthiness ratings were not significantly related to healthy or unhealthy consumption, nor were they significantly correlated with most television viewing and parental communication measures (see ). As a result, when included in the models, most paths to and from healthiness ratings were not significant, and we chose to exclude them from the analysis.
Zero-order correlations between television viewing, parental influence, food ratings, and diet (N=191)
A latent variable was used to represent television viewing experience that included childhood and high school television viewing and positive mediation. Independent variables included food rules, television viewing restrictions, and parental critical viewing. As the viewing restrictions scale, displayed unacceptable kurtosis, data for that variable were transformed (logX / (1-X)) to achieve a normal distribution. Modeling utilized maximum likelihood estimates and estimated the covariance matrix.
Our first model specification required two minor adjustments to the originally hypothesized relationships between parental influence and television viewing variables. First, the original path from parental restrictions to television viewing experience was changed to predict only childhood television viewing. As predicted in H3c, viewing restrictions was negatively correlated with childhood television viewing; however, the hypothesized relationship between viewing restrictions and other television experience variables (i.e., high school viewing and positive mediation) was not significant (see ). It appears that parents who do not approve of television may be more effective at limiting their younger children’s television viewing than that of their teenagers. We also added a path from critical viewing to positive mediation to improve the model fit. Although not predicted, this relationship is in line with other analyses that have found positive mediation and critical viewing to be correlated; they both may reflect an underlying measure of parental discussion about television (Austin & Pinkleton, 2001
). As expected, critical viewing, viewing restrictions and food rules were all related to each other and, together, may indicate an overall parenting style to discuss expectations and limits.
Unhealthy diet model
To examine the potential unhealthy influence of television food advertising, we first specified the model to predict taste of unhealthy, highly advertised foods and unhealthy diet (see ). Goodness-of-fit indicators showed that the data fit the model, χ2(13, N = 191) = 12.2, p = .51; RMSEA = .00 (90% CI = .00 to .07); CFI = 1.00. Multivariate normality was acceptable (Mardia's normalized coefficient = 4.56). All standardized regression coefficients indicated the expected relationships between variables, and the only insignificant path led from food rules to taste of unhealthy, advertised foods.
As predicted in H2a and H3b, greater television viewing experience was associated with a higher aggregate taste rating for unhealthy advertised foods which, in turn, was associated with a more unhealthy diet. Television experience was also directly related to a more unhealthy diet, however, further analysis revealed that taste partially mediated this relationship: An alternative model without the taste variable resulted in a stronger direct relationship between viewing experience and unhealthy diet (β = .29, p < .01 for the direct relationship vs. β = .18, p < .05 when taste was included as a mediator). Food rules and parental critical viewing also contributed independently to unhealthy diet, and critical viewing was related to lower perceived taste of unhealthy advertised foods. These findings partially support H3a: parents who are critical of the messages presented in food advertising can reduce television influence. Food rules did not, however, directly predict perceived taste of unhealthy, highly advertised foods, as hypothesized.
To further examine the potential influence of television food advertising on unhealthy diet we tested two alternative versions of the unhealthy diet model. In the first alternative, we assessed whether taste ratings for unhealthy foods that are not heavily advertised also partially mediated the relationship between television viewing experience and unhealthy diet. This model also fit the data, χ2(13, N = 191) = 10.50, p = .65; RMSEA = .00 (90% CI = .00 to .06); CFI = 1.00. As in the original model, the taste of unhealthy foods with less advertising was related to both television viewing experience (β = .20, p < .05) and unhealthy diet (β = .26, p < .01). In this model, however, the direct relationship between television viewing experience and unhealthy diet did not differ from the version without taste rating as a mediator (β = .29, p < .01 in the unmediated model). In addition, parental critical viewing was not related to perceived taste of unhealthy foods with less advertising (β = −.03, ns). These results further support our hypothesis that exposure to advertising messages on television directly contribute to unhealthy diet by increasing perceived taste of the unhealthy foods advertised, and that parents who question television messages can moderate this influence. Television viewing experience, however, also appears to directly predict a more unhealthy diet, beyond its influence on perceived taste of advertised foods.
In the second alternative model, we examined whether the relationship between television experience and unhealthy diet was mediated by an underlying parenting style to communicate limits and expectations that influenced both outcomes. In this version, we added a latent variable that included food rules, critical viewing and viewing restrictions. This model did not fit the data, χ2(16, N = 191) = 36.71, p < .01; RMSEA = .08 (90% CI = .05 to .12); CFI = .94. Therefore, this alternative explanation for the relationship between television viewing and unhealthy eating was not supported by the data, as predicted by H2b.
Model to predict healthy diet
We then applied the unhealthy diet model to predict healthy diet, replacing taste of unhealthy foods with taste of healthy foods (see ). This model also fit the data, χ2(13, N = 191) = 9.60, p =.73; RMSEA =.00 (90% CI = .00 to .05); CFI = 1.00. As in the unhealthy diet model, taste of healthy foods and television viewing experience directly predicted healthy diet. In this model, however, the relationship between television experience and taste of healthy foods was near zero. In addition, both food rules and critical viewing predicted taste of healthy foods, but did not directly predict healthy diet. These findings suggest that a more healthy diet is related to television viewing experience, but not through its effect on perceived taste of healthy foods. In contrast, parental communication about food and television appear to contribute to healthy diet through their influence on perceived taste of healthy foods. The strong negative relationship found between television viewing and healthy diet appears to be due to some factor not measured in this model, for example, perceived importance of a healthy lifestyle.
Overall, these findings support our hypothesis that healthy and unhealthy diets are most directly related to the perceived taste of healthy and unhealthy foods. In addition, as expected, nutrition knowledge was not related to unhealthy diet, although it was indirectly related to healthy diet: Participants' ratings of the healthiness of healthy foods predicted greater perceived taste of those foods.
The results also support our hypothesis that the relationship between early television viewing and unhealthy eating found in previous research with children and adolescents continues into early adulthood. Past and present television viewing were correlated with both unhealthy and healthy diet in our college sample. In fact, eating behaviors were more highly correlated with childhood and adolescent viewing than current television viewing, although these results could be due to either greater influence from early television viewing or to unusual television viewing patterns while attending college.
The SEM analyses provide insights into potential reasons for the relationship between television viewing and diet. Both healthy and unhealthy diets were directly related to television viewing experience and may represent an underlying healthy (or unhealthy) lifestyle. The relationship between television viewing and unhealthy diet, however, was partially mediated by perceived taste of unhealthy, highly advertised foods, but not unhealthy foods with less advertising. In fact, the most significant path from television experience to unhealthy diet included taste of unhealthy, highly advertised foods. In contrast, the relationship between television viewing experience and the taste of healthy foods was near zero. These findings support our prediction that television viewing experience is associated with endorsement of the messages presented in children's food advertising (i.e., the unhealthy foods advertised taste great). We were surprised, however, to find no significant relationships between television viewing and attitudes about healthy food or ratings of food healthiness. The evidence consistently supports the conclusion that individuals who watch more television simply like the taste of unhealthy foods more, especially those that are highly advertised.
Also as predicted, parental influence measures moderated the relationship between television viewing and diet. Viewing restrictions and positive mediation were related to television viewing experience. Viewing restrictions, however, was only related to childhood viewing, which suggests the importance of limiting children's viewing early on, when it can be done effectively. In addition, lower viewing in childhood was associated with continued lower viewing in high school and college; we found no evidence of a rebound effect from early viewing restrictions.
Parental efforts to educate their children about food and television (i.e., food rules and critical viewing) also influenced healthy and unhealthy diet, but through different paths. Critical viewing was related to higher taste ratings for healthy foods and lower taste ratings for unhealthy, highly advertised foods. In contrast, food rules predicted higher taste ratings for healthy foods, but not unhealthy foods. We were surprised, however, that parental influence measures were not significantly related to nutrition knowledge. Overall, these findings support expectancy theory and suggest that parental efforts to counteract the unhealthy messages on television and teach healthy food preferences can influence their children's evaluations of food taste, and thus affect their healthy and unhealthy diet into adulthood. Both critical viewing and food rules were also directly related to unhealthy, but not healthy diet (in addition to their influence on taste preferences). Although we did not predict this relationship, perhaps parental communication can also teach children how to refrain from unhealthy eating, in spite of taste preferences for unhealthy foods.
Finally, the data did not support the alternative hypothesis, suggested by food industry proponents, that television viewing itself has no direct effect on unhealthy diet and that the relationship between television viewing and diet can be explained by a more permissive parenting style that allows children to watch less television and eat more unhealthy foods.