White matter tractography using diffusion tensor MR images is a promising method for estimating the pathways of white matter tracts in the human brain. The success of this method ultimately depends upon the accuracy of the white matter tractography algorithms. In this study, a Monte Carlo simulation was used to investigate the impact of SNR, tensor anisotropy, and diffusion tensor encoding directions on the accuracy of six tractography algorithms. The accuracy was assessed in straight and curved tracts and tract geometries with divergence properties. In general, the tract dispersion increased with distance and decreased with SNR and anisotropy. The tract orientation with respect to the encoding scheme also influenced tract dispersion. Divergent tract geometries increased tract dispersion, whereas convergent tract geometries reduced dispersion. Analytic models of tract dispersion were constructed as a function of the tract distance, SNR, eigenvalues of the tracts, voxel size, and the relationship between the tract direction and the diffusion tensor encoding directions. In certain cases, the mean tract trajectory was found to deviate from the ideal pathway for curved trajectories. Analytical models of mean displacement were constructed as a function of the curvature, tract distance, step size, and tensor eigenvalues. These models may be used in future studies to assess the level of confidence associated with a tractography result.