We have developed a simple method for the direct analysis of two-dimensional trabecular structure. Using this method, we have compared the structure of the left and right iliac crest and the second lumbar vertebra and studied the relationship between two-dimensional trabecular structure and the ultimate compressive stress of bone samples from the second lumbar vertebra. Transverse cores were taken from the left and right iliac crests and second lumbar vertebrae of 35 subjects (12 females) who died suddenly. The trabecular structure was subdivided into nodes (node count) and free ends (free end count), which were expressed per square millimeter of cancellous space. A node was defined as the point of junction between two or more trabeculae; a free end was defined as the end of a trabecula that was unconnected in the plane of the section to any other trabecular element. The following struts were defined by drawing a line between the nodes and the free ends: node to node, node to free end, free end to free end, and cortex-derived struts. Strut lengths were measured and expressed as a percentage of the total strut length and per square millimeter of cancellous space. Intra-observer variation was generally low, with coefficients of variation for repeat measurements in the range 3.9–17.8%. There was no statistically significant difference between left and right iliac crests for any of the variables measured. With the exception of cortex-derived struts, there was no statistically significant correlation between crests for any variable. When compared to the second lumbar vertebra, the iliac crest had a greater node count and node-to-node strut length, indicating a higher degree of connectedness. There was a significant correlation between the lumbar vertebra and iliac crest for node count and node-to-node strut length. Variables measured in the second lumbar vertebra correlated well with ultimate compressive stress, the strongest correlations being with free end-to-free end strut length (r = -0.75, p < 0.0005) and node-to-free end strut length (r = 0.76, p < 0.0002). In both the second lumbar vertebra and iliac crest, node count, node-to-node strut length, and free end-to-free end strut length correlated significantly with age. The method reported here requires inexpensive equipment and is reproducible. Two-dimensional trabecular structure in the iliac crest has greater connectedness compared to that in the second lumbar vertebra, although there is a relationship between the two sites. The structural variables measured in the lumbar vertebra are indicative of ultimate compressive stress at that site. Similar relationships between the structural variables and age are seen in the iliac crest and second lumbar vertebra. This type of analysis should provide important quantitative information concerning the changes in trabecular structure with age and in disease, as well as the ability of different drug treatments to improve trabecular structure.