Those who consumed 3 or more soft drinks per day also had a BMI z score that was 0.51 higher (95% CI, 0.17 to 0.85; P = .003), had 4.4% more body fat, and were more likely to have a BMI at or above the 85th percentile than those who consumed fewer than 3 soft drinks per day (58.1% vs 33.2%; χ2 = 10.15, P = .006). The BMI z score and percent fat were correlated both with consumption of diet soft drinks (r = 0.19, P = .001; r = 0.18, P = .002, respectively) and regular soft drinks (r = 0.10, P = .08; r = 0.11, P = .06, respectively), so the numbers were combined into the total number of soft drinks consumed per day (Table 4). Boys and girls watched a similar amount of television (1.8 and 1.9 hours per night, respectively) and consumed a similar number of soft drinks (1.4 and 1.2 soft drinks per day), but Latino students watched more television (2.4 hours per night) and consumed more soft drinks (1.6 drinks per day) than either non-Hispanic white students (1.3 hours of television viewing, t263 = 6.46, P<.001; 1.1 soft drinks per day, t278 = 2.92, P = .004) or Asian students (1.3 hours of television, t163 = 3.32, P = .001; 0.7 soft drinks per day, t164 = 2.95, P = .004). Although soft drink consumption and television viewing were correlated (r = 0.27, P<.001), in a logistic regression with BMI at or above the 85th percentile as the dependent variable, 2 or more hours of television watching (odds ratio, 1.50; 95% CI, 1.17 to 1.93) and 3 or more soft drinks per day (odds ratio, 1.61; 95% CI, 1.14 to 2.28) were each significant independent variables when the other variable was controlled for as well as for age and sex (Table 5). Ethnicity accounted for much of the association with BMI at or above the 85th percentile with television viewing. When ethnicity was controlled for (excluding African Americans and Native Americans), television viewing was no longer significantly associated with BMI (odds ratio, 1.55; 95% CI, 0.90 to 2.66). However, soft drink consumption remained significant (Table 5).