In the present study, we examined how well adolescents (12-13 years) are able to select the correct dietary aid portion sizes after having been shown different food items. We also evaluated the effectiveness of two-dimensional life-size drawings and three-dimensional food models, used as dietary aids in this process. Fifty black children and 42 white children from Johannesburg participated in the study (N = 92). Trained interviewers individually tested each child following a prescribed sequence, throughout. Each participant was shown a plate of actual food of a pre-determined weight. The participant was required to select a two-dimensional drawing, and thereafter a three-dimensional food model, which most closely resembled the real food portion. In this manner, portion size estimation was evaluated with respect to 11 different food items. Correlations between nutrients calculated from actual weight of food portions and estimates ranged from 0.842 to 0.994 (P < 0.0001), indicating a significant positive linear association between the actual and estimated nutrients, using either of the dietary aids. However, findings also suggest that the drawings provided a better estimate of actual energy, fat and carbohydrates than did the food models (with respect to lying within the limits of agreement). On the other hand, the food models were more frequently selected correctly than the drawings. Hence, both methods had advantages and disadvantages. Overall, it was found that there were no gender differences (P < 0.05) when using either the models or drawings to estimate portion size, however, there were significant ethnic differences (P < 0.05). With two exceptions, black children selected the correct aids (drawings and models), more often compared with white children. It is recommended that in dietary interviews undertaken in black children in urban areas one could use either aid; while in white adolescents the use of the food models is recommended.