The optic chiasm with its complex fiber micro-structure is a challenge for diffusion tensor models and tractography methods. Likewise, it is an ideal candidate for evaluation of diffusion tensor imaging tractography approaches in resolving inter-regional connectivity because the macroscopic connectivity of the optic chiasm is well known. Here, high-resolution (156 pm in-plane) diffusion tensor imaging of the human optic chiasm was performed ex vivo at ultra-high field (9.4 T). Estimated diffusion tensors at this high resolution were able to capture complex fiber configurations such as sharp curves, and convergence and divergence of tracts, but were unable to resolve directions at sites of crossing fibers. Despite the complex microstructure of the fiber paths through the optic chiasm, all known connections could be tracked by a line propagation algorithm. However, fibers crossing from the optic nerve to the contralateral tract were heavily underrepresented, whereas ipsilateral nerve-to-tract connections, as well as tract-to-tract connections, were overrepresented, and erroneous nerve-to-nerve connections were tracked. The effects of spatial resolution and the varying degrees of partial volume averaging of complex fiber architecture on the performance of these methods could be investigated. Errors made by the tractography algorithm at high resolution were shown to increase at lower resolutions closer to those used in vivo. This study shows that increases in resolution, made possible by higher field strengths, improve the accuracy of DTI-based tractography. More generally, post-mortem investigation of fixed tissue samples with diffusion imaging at high field strengths is important in the evaluation of MR-based diffusion models and tractography algorithms. (C) 2007 Elsevier Inc. All rights reserved.