The breakdown of hard seed, germination and establishment of heat-treated and untreated seed was compared for Stylosanthes humilis, S. hamata cv. Verano and S. scabra cv. Fitzroy at Katherine, N.T. Heat treatment significantly increased yield and seedling establishment for S. hamata and S, scabra. Recovery of seed samples at the end of the wet season showed that no decrease in hard seed occurred after the first wave of seedling establishment at the onset of the wet season. This result indicates that high surface temperatures are required for seed softening, and that hard seed which has not broken by the onset of the wet season is unlikely to contribute to the sward in the season of sowing. The lack of a significant heat-treatment effect for S. humdis was attributed to the high soft-seed content (29%) in the untreated seed at the time of sowing and the loss of seedlings from the heat-treated seeds owing to rapid germination on rainfall insufficient for seedling survival. In a second experiment, the changes in seed characteristics of heat-treated seed sown in the late dry season were studied. Seed was recovered before and after each major rainfall event. Standard germination tests were carried out at near optimal conditions of water availability, light and temperature to measure potential germination speed and changes in viability. Large increases in potential germination speed occurred over 4 weeks in the field-from < 10% germination in 18 h to 70-90% in 18 h. As a result, greater than 80% of germination occurred on isolated rainfall events with less than 18-h surface moisture and were followed by rapid seedling death. We found that slow potential germination speed of sown seed did not confer permanent protection against false-start germination, since large increases in potential germination speed are likely to occur under field conditions.