In multivariate analysis, suppressor variables attenuate the true relationship between predictor and outcome variables. Although suppressor variables relationships were described more than 45 years ago, few examples have been reported in the health care literature. We studied the correspondence between body mass in 111 Mexican-American families. Estimates of exercise suppressed the correlation between body mass for some dyads within these families. We concluded that the suppressor variable relationship may cause underestimates of heritability in multivariate studies.