Germplasm collections underpin national and international breeding activities for all food crops. Effective and representative germplasm collections, however, tend to be very large and are difficult to maintain and organise efficiently. Recent development in molecular marker technology have allowed us to address problems such as the identification of duplicate material within such collections and the assembling of 'core collections' which represent a high percentage of all variation available within a single species. This work has been carried out using the large rice germplasm collection held at IRRI as a model system. The collection now comprises more than 100,000 samples of rice and is a major supplier of rice germplasm to research groups and breeders throughout the world. The germplasm has made important contribution to the production of new rice varieties. Pressure on germplasm distribution will further increase over the next 30 years as plant scientists strive to meet the demands for increased rice production. In this programme we have carried out research which will enhance the efficiency with which rice germplasm can be selected for use in breeding programmes and contribute to the anticipated gain of 1.5% per annum in rice yield defined as a DFID target. We have developed a range of molecular marker techniques that allow the rapid estimation of genetic relationships between rice samples. These techniques have been compared and decisions made about the most appropriate methods for use with rice germplasm. Using these molecular marker methods, protocols for the identification of duplicate accessions and the development of core collections have been developed. The use of markers unmapped and mapped markers for such purposes have been compared and the limitations and advantages of these classes of marker have been documented. A molecular marker laboratory has been set up in the International Rice Germplasm Collection at IRRI. Associations between performance for quantitative traits and the presence/absence of specific markers have been established allowing prediction of rice field performance using marker data. Key AFLP markers have been sequenced and converted to single-locus PCR markers which are currently being used at IRRI for trial predictions of performance of rice accessions. The results of the research have been widely disseminated during visits between collaborators between the three groups involved (at Birmingham, JIC and IRRI), at conferences and lecture tours, and in a series of publications in international journals.