http://dx.doi.org/10.5007/1980-0037.2008v10n3p243 The purpose of this study was to identify which skinfold thicknesses best predict the Body Mass Index (BMI) of children aged 6 to 10 years. The sample consisted of 188 schoolchildren (99 boys and 89 girls) from the private Education System of the city of Ponta Grossa, PR, Brazil. Anthropometric variables were measured: body mass and height (for computing BMI) and skinfolds (biceps, triceps, subscapular, oblique midaxillary, oblique suprailiac, vertical abdominal and medial calf). The relationships between skinfolds and BMI were analyzed using Pearson’s correlation. Stepwise multiple regression analyses were used to determine the independence and mutual contribution of skinfolds to prediction of BMI, with p <0.05. For boys, the abdominal skinfold (model 1) was the strongest predictor of BMI, explaining 72.3% of the variance, while including the subscapular skinfold (model 2) produced a small increase to 73.7%. For girls, the results indicated that the suprailiac skinfold was responsible for 82% of the variance in BMI (model 1), while the addition of the triceps skinfold (model 2) increased to 85% of variance. A third model, which incorporated the subscapular skinfold (model 3), caused a further increment in explained variance to 86%. The findings of this study indicate a moderate relationship between BMI and abdominal and suprailiac skinfolds, for boys and girls, respectively. It was shown that, for our sample, skinfolds representing central fat distribution were the best predictors of changes in BMI.