1. Numerical methods were used to simulate the voltage responses to an intrasomatic current step of neuronal models that incorporated tapering dendrites, dendrites of unequal electrotonic length, nonlinear membrane properties, and regional differences in specific membrane resistivity (Rm). A "peeling" technique was used to estimate the time constants (tau 0 and tau 1) and coefficients (a0 and a1) of the first two exponential terms of the series of exponential terms whose sum represented the slope of the voltage response. 2. The electrotonic structure of models with a uniform Rm was calculated using equations derived by Rall or Johnston or Brown et al. The adequacy of these methods were tested using a wide variety of models that conformed to the equivalent cylinder approximation of Rall. Johnston's method provided the most reliable estimate of electrotonic length (L) and the ratio of the dendritic conductance to the somatic conductance (rho). However, if L exceeded 2 and rho was eight or larger, the equations derived by Johnston could frequently not be solved due to small errors in the peeled values of tau 0, tau 1, a0, and a1. Although the method suggested by Brown et al. could be applied to all models, this method invariably underestimated L and rho. These errors were particularly large for model neurons with L values of 1.5 or larger and rho values of four or larger. Estimates of L using Rall's method were only reliable if rho was large and L was two or less. 3. Changing the geometry of the dendritic tree (dendritic tapering or dendrites of unequal L) or the addition of a time- and voltage-dependent conductance designed to mimic a sag process commonly seen in spinal motoneurons caused systematic changes in tau 0, tau 1, a0, and a1. The sag process always led to an underestimate of tau 0 even after applying a correction procedure. On the other hand, the ratio, tau 0/tau 1, was not affected by the sag process or dendritic tapering.(ABSTRACT TRUNCATED AT 400 WORDS)