This paper addresses the issue of genotype by environment (G x E) interaction in relation to the distribution of germplasm from international to national breeding programmes. Theoretical aspects of G x E interaction and of the selection versus the test environment are reviewed. The importance of specific adaptation to maximize yield and yield stability under stress conditions is highlighted in barley and lentil. We conclude that repeated cycles of selection at a few sites have a high probability of reducing the frequency of genotypes specifically adapted to environmental and/or agronomic conditions not represented at those sites. Therefore a wider and earlier devolution of the selection work done by International Agricultural Research Centres (IARCs) to plant breeders in national programmes will increase the chances of exploiting positive G x E interactions. It will also address the needs of small, resource-poor, subsistence farmers and reduce the danger of narrowing genetic diversity within crops and regions. This is an efficient way of making breeding programmes in less developed countries more self-reliant, which is a major objective of International Agricultural Research Centres.