Participatory variety selection (PVS) was carried out in two of the major tef-growing woredas (districts) of Ethiopia, Ada and Akaki, in 2003 and 2004. The objectives were to identify farmers' selection criteria, to increase farmers' awareness and their access to improved tef varieties, to enable farmers to assess the performances of improved tef varieties of their choice and to accelerate seed dissemination of farmers' chosen varieties through farmer-to-farmer exchange mechanisms. Seed colour was the overriding selection criterion. Farmers overwhelmingly selected the very white seed variety DZ-01-196 (Magna) for market purposes, indicating that tef is a cash crop. Farmers also selected brown-seeded tef, but for home consumption. There was no evidence of connection between seed colour and desirable agronomic traits, or nutritional quality (protein content). Factors other than seed-colour were of secondary concern to farmers. When market demands become the dominant selection criteria, PVS should not be an end by itself: rather, it complements the formal breeding system. The results imply that tef performance evaluation trials need to be separated on a colour-group basis, and any new successful variety in the two woredas should be superior to DZ-01-196 not only in grain yield but also in seed-colour quality. A faster, more efficient and reliable pure-seed supply system than the traditional farmer-to-farmer exchange mechanism is required in order for farmers to continue planting improved varieties, which might be better achieved through small-scale commercial producers and/or cooperatives.