1. The vertical distribution in the nest of chambers, workers, callow workers, brood and seeds was studied in the harvester ant, Pogonomyrmex badius, in northern florida. On each of four sample dates (May, July, October, January), six to seven colonies, chosen to represent the full range of sizes, were excavated. All chamber contents were collected and counted. Chambers were mapped and measured. In a preliminary study, two nests were excavated after preventing vertical migration by driving barriers into the wall of a pit next to the nest, severing the vertical tunnels. The vertical distribution of these barrier-nests differed little from unrestrained nests, indicating that unrestrained excavation produced a reasonable picture of vertical distributions. 2. Nest depth, chamber number and total area increased with colony size. Chamber area declined sharply with depth, as did chamber number, such that more than half of the total area was found in the upper quarter of the nest. 3. The proportion of dark-coloured (older) workers also declined strongly with depth, but this decline was weaker in the spring, and depended to a modest degree on colony size. Conversely, in the distribution of callow (young) workers, the proportion increased towards the bottom of the nest. Mean worker age was inversely related to the depth at which workers were found. The proportion of the brood also increased towards the bottom of the nest, with worker brood, sexual brood, pupae and larvae all being distributed similarly. 4. By contrast, seeds were stored at a preferred absolute depth between 40 and 100 cm. Colonies shallower than 100 cm stored seeds in their deepest chambers. Larger colonies stored most seeds in the upper third of the nest, but patterns were somewhat erratic because chambers were either filled completely with seeds or were empty. 5. Because chamber area decreased sharply with depth, the densities (individuals cm(-2)) Of all colony members, including dark workers, were lowest near the surface cm and highest in the deepest parts of the nest. Here, worker densities ranged from 2 to 8 cm(-2), and brood from 2 to 25 cm(-2). 6. The regularity of the patterns of distribution suggests that harvester ant colonies have considerable spatial and temporal structure, which serves or is the outcome of important colony processes. A simple mechanism that could generate several of these patterns is discussed. New workers produced deep in the nest move upwards as they age. As they leave the brood zone they change from brood care to general nest duties, including increased nest excavation, leading to the top-heavy pattern of nest area. As they appear at the surface, they change to guarding and foraging. Thus, age polyethism may be partly the result of this upward migration of workers.