Squid giant axons were used to obtain axonal cytoskeletons that had been separated from the confines of their plasma membranes. To remove the plasma membrane, axoplasm was extruded from the giant axon directly into an artificial axoplasm solution (AAS). This procedure produces a smooth axoplasmic cylinder in which neurofilaments (NFs) are the most prevalent cytological elements. The NFs scatter light strongly and thus dark-field light microscopy can be used to quantify the volume occupied by these polymers. Measurements of the widths of the dark-field images of the axoplasmic cylinders showed that the cross-sectional area of the NF population increased by 60-110% (n = 8) between 1-100 min after plasma membrane removal, and then continued to increase more slowly for many hours. After 1,000 min, the cross-sectional area was 75-160% (n = 8) larger than at 1 min. These light microscopic measurements of axoplasm suggest that the NF population disperses to occupy a continuously increasing volume after removal of the plasma membrane and immersion in AAS. This inference was confirmed by quantitative ultrastructural studies of NFs in axoplasmic cross-sections, which demonstrated that the spacing between the NFs increased between 1-1,000 min after plasma membrane removal. Comparison of the NF density distribution after 1,000 min with a theoretical distribution calculated using the Poisson theorem indicated that the NFs dispersed randomly. These studies on NFs in isolated axoplasm suggest that ordinary thermal forces of Brownian motion are sufficient to move axonal NFs apart independently and thereby to disperse them. We propose that, in the intact axon, the dispersive movements of the NFs spread the NF cytoskeleton radially and expansively to fill out the cylindrical space contained by the axonal plasma membrane and its surrounding connective tissue elements.