The measurement of root length density (L-V) with soil depth is tedious and laborious. Yet the information is necessary in order to model root nutrient uptake. Measuring the number of roots exiting the face of a soil trench is less laborious and has previously been shown to be five times faster than the soil core-break method. Based on geometric theory, the theoretical relationship between the number of roots exiting the face of a three-dimensional block of soil (N) is related to L-V by the equation L-V = 2N, if roots are isotropic and the soil volume is small. The coefficient of 2 presumes roots to be randomly oriented in three-dimensional space (i.e. no preference for vertical or horizontal orientation). The equation is appropriate to anisotropic root distributions, if root density is evaluated in three mutually perpendicular planes. In the case of non-random or preferential root orientation, the coefficient will be different than 2. If roots are anisotropic, the equation becomes L-V = 2N(AVG), where N-AVG is the mean of the N values measured on the faces of a soil block in all three-dimensions. The purpose of thus study was to test this relationship with roots of Melaleuca quinquenervia growing on a sandy Spodosol, Florida and Bactris gasipaes (peach palm) growing on a clayey Oxisol in Amazonas, Brazil, and to evaluate a simple sampling method using soil trenches that would estimate L-V with soil depth. Roots of Melaleuca were anisotropic with 50% fewer roots exiting the basal face compared to the lateral faces. Roots of peach palm were isotropic. The relationship of L-V with N and NAVG does not fit the theoretical coefficient of 2, suggesting that the most likely problem was either that the method was not accounting for small roots exiting the soil volume or that the method was using a soil volume greater than required to fit the theoretical relationship. If the first explanation is correct, then the field method overlooked 66-74% of the Melaleuca roots and 36-82% of the peach palm roots exiting the face of a soil block. Correlations between N and L-V showed that N explained between 80 and 87% of the variability in L-V, indicating that this can be a useful method to predict L-V distributions with depth. (C) 2002 Elsevier Science B.V All rights reserved.