The nature of neurofilament organization within the axonal cytoskeleton has been the subject of controversy for many years. Previous reports have suggested that neurofilaments are randomly distributed in the radial dimension of the myelinated axon. Randomness of distribution implies that there is no interaction between neurofilaments, while order in distribution suggest the presence of forces between neurofilaments. To address the issue of randomness vs. order, we evaluated neurofilament distribution by two different statistical approaches--nearest-neighbor distance and the Poisson tile-counting method. Neurofilament nearest-neighbor distances in a myelinated axon differ from nearest-neighbor distances of a set of random points with similar density (40.6 +/- 7.0 nm vs. 30.7 +/- 12.9 nm, P < 0.0001). The Poisson tile-counting method also indicated that neurofilament distribution is different from a random distribution, under conditions of appropriate tile size and masking of other organelles. To further characterize the distribution of neurofilaments, we compared the relationship between nearest-neighbor distance and density for three sets of data: evenly spaced points, randomly distributed points and measured neurofilament coordinates. Neurofilaments do not conform to either evenly spaced or random distribution models. Instead, neurofilament distribution falls into an intermediate position between evenly spaced and random distributions. This study also demonstrates that the nearest-neighbor distance method of assessing neurofilament distribution offers several technical and theoretical advantages to the Poisson tile-counting method.