Yield progress due to crop breeding for smallholder farmers in developing countries has been slow because of complex genotype-by-environment (GE) interaction and a lack of concordance between selection and target production environments. The objectives of this study were to: (i) assess GE interaction and identify its genotypic and environmental causes for grain yield, (ii) demonstrate the direction of yield progress during the last three decades vis-à-vis high yield and stress environments, and (iii) suggest an appropriate selection strategy to develop full-season food barley ( Hordeum vulgare L.) cultivars in central highlands of Ethiopia. Sixteen barley genotypes were tested in a factorial combination of two levels each of sowing date (at the start of main season rain versus 20 days after) and fertilizers (none versus 41 kg ha −1 N + 20 kg ha −1 P) in 1998, 1999 and 2002 on a Eutric Nitosol at Holetta, Ethiopia. Genotypic sum of squares accounted for 12% and GE interaction for 19% of G + E + GE sum of squares. Genotype-by-year interaction was the largest source of GE interaction. Mean genotype grain yields in 1998 were not correlated with those either in 1999 or in 2002 but the latter two were. Improved food barley genotypes had above average vegetative duration, individual plant weight and leaf width at heading and interacted favorably with environments where season-end moisture stress was low. Farmers’ cultivars that were early maturing with high harvest index, spike number and vegetative vigor interacted positively with environments where season-end moisture stress was high. Baleme, the local cultivar around Holetta, was late maturing but had little contribution to GE interaction. In Ethiopia, yield progressed due to food barley breeding under low season-end moisture stress but declined slightly under intermediate and high season-end moisture stresses until 2001. Nonetheless, yield trends were positive under all the three scenarios when Dimtu, a variety selected under low and high fertilizer inputs and released in 2001, was included. Yield stability is as important as yield potential for subsistence farmers in risk-prone environments such as in Ethiopia. This in part explains the current state of poor adoption of improved barley cultivars and may call for reorientation of food barley breeding strategy to minimize risks while increasing yields. It is suggested that future barley breeding efforts should include season-end drought stress as a target selection environment.