To estimate three-dimensional distribution of foliage which affects microclimate in forests and habitat structure of biotic community, we examined the accuracy of the MacArthur-Horn method at various sampling scales and points in a deciduous broad-leaved forest by using a large scaffolding and direct observation. The coefficient of determination of the relationships between estimates and observed values decreased with increasing complexity of foliage profile. The accuracy of estimating the foliage distribution did not change with the sampling scale when the total number of sampling points was constant at 64 points. Therefore, it is necessary to adjust the number of sampling points according to the sampling scale in a forest where foliage height diversity is high. On the other hand, the accuracy increased asymptotically with increasing sampling scale when the number of sampling points per unit area was keep constant (ca. 64 points per 1.8 m × 1.8 m). We can get a suitable sampling area which the accuracy of estimates remains constant. The minimum scale for estimating correct foliage profile was about 80 m 2, and its coefficient of determination was 0.855. Therefore, we will be able to estimate three-dimensional foliage distribution of individual trees or species and the heterogeneity of the foliage profile at any scales in accordance with the aims of study if we design the sampling points and the size of sampling scale into consideration.