What this means is that a plant which has adapted to a specific, often harsh, environment is better able to adapt itself. It will therefore make the transition to a high-yielding environment better than an improved, high-yielding variety will make it to a poor one. We proved this for ourselves. Between 1985/86 and 1993/94, we took the 5% highest-yielding barley genotypes in low-yielding environments, and the 5% highest-yielding in high-yielding environments. Those selected from poor sites gave two-and-a-half times the yield when grown in good ones. In fact, they managed four-fifths of the yield from the high-yielding ones in a good environment. By contrast, yield of the varieties selected from good sites dropped by four-fifths when they were grown in more difficult ones. So, for a resource-poor farmer, adopting plants developed under good conditions on a research station could be a disaster. Instead, the strategy used by Third World farmers is to mix both different crops, and different varieties of the same crop, in their fields. There is thus a fantastic on-farm genetic diversity available in their fields.
        In the early 1980s, we set out to exploit it in Syria. We tested a large collection of landraces (farmers' varieties) collected by Eva Weltzien in Jordan and Syria in 1981. She was then doing her PhD thesis at ICARDA (she is now doing similar research on pearl millet at ICRISAT, ICARDA's sister Center in India).
        In 70 fields--60 in Syria, 10 in Jordan--Dr Weltzien collected 100 heads (spikes) per field. This gave 7,000 individual spikes in 7,000 envelopes; given 20 seeds per spike, this was 140,000 seeds! These envelopes had been kept as separate accessions, not bulked together, and that proved to be the key. Individual testing revealed very broad diversity in growth characteristics and pest and disease resistance. In the end, we hit upon three especially promising accessions, and these emerged as ICARDA lines that were subsequently released for farmers' use. They were Tadmor, Zanbaka and Arta--three among millions of barley spikes in farmers' fields that turned out to be what we were looking for.
        Arta in particular is good news. It was developed from a spike Dr Weltzien collected in a field near Sweda, about 100 km east of Damascus. Oddly, the village was Um Zeitoun, meaning Mother of Olives. It seems we had the mother of barley instead! Arta has been consistently outperforming existing landraces in farmers' fields all over Syria. And it has been doing so because it was developed from a landrace in the first place, selected in the target environment.
        There is a danger here. Arta was developed from a single spike. This type of breeding--from a pure selection--could lead to a new genetic homogeneity that could take us back to where we started. In fact, what we have seen of farmers' adoption patterns in the region suggests that this is not so great a danger as it might be, but we should avoid it. In the long term, ICARDA strategy will be to breed-in characteristics to landraces from landraces in order to preserve genetic diversity in the farmers' fields.
        There is also a practical difficulty. We said at the beginning that we were trying to get away from varieties that were not bred for a specific target environment. If we exploit biodiversity we can end up with the huge variety of varieties that we need for the different environments, but how do we get them to farmers? Centralized seed production and distribution would be uneconomic.
        We postulate that the way to do it is to bring the farmer into the breeding process. If the new lines are evaluated in his field in the first place, he already has what he needs at the end of the process; it will then be distributed to his neighbors. ICARDA is already working on this (see Caravan No. 1).
        But the varieties are not going to spread in this way from Sweda to (say) Tunisia. If they did, they might not prove to be the right ones for that target environment. ICARDA's own testing of barley lines simultaneously in Syria and the Maghreb has demonstrated this.
        So the process must also be decentralized to national programs. Better for Syria and Tunisia to swap lines for testing, and then to do that testing both on the research station and in farmers' fields. Then we really will have got away from the "top-down" approach to plant breeding--and the seeds will be in the farmers' hands. It means a move away from the traditional pattern of varietal testing and release on a national scale, but that procedure was designed for a different set of circumstances, anyway.
        A final thought. We started this article with the premise that biodiversity must be preserved in order to produce the food the pople in the dry areas will need in the 21st century. But it is a two-way process. The decentralized, farmer-based breeding methods just mentioned will not only exploit biodiversity, but, by vastly increasing the lines being grown in different places, they will also preserve it in themselves.


Dr Salvatore Ceccarelli is senior barley breeder, and Dr Stefania Grando barley breeder, at ICARDA. Ir Joop van Leur is a barley pathologist with ICARDA and is currently based in Ethiopia, where he is working with Ethiopian scientists to identify barley lines for yield stability in low-input agriculture.The authors would like to acknowledge the valuable support of OPEC and the Government of Italy for this work; also to thank the German development organization BMZ, which is assisting farmer-participatory breeding work in Syria.

ne of the clarion calls of the 1992 United Nations Conference on Environment and Development (UNCED), now generally known as Rio after its venue, was for the preservation of genetic diversity--biodiversity. For many people around the world, this was the first time the phrase had been heard, and they took it to mean the protection of some of the world's vanishing animal species.
        They were not wrong about this, but there is an even more pressing need to preserve plant biodiversity if we are to feed the world's growing population in the 21st century and beyond--and avoid the drastic fluctuations in agricultural production that lead to poverty and famine. Nowhere is this need more urgent than in the world's drylands. In these harsh environments, biodiversity is the key to better yields, and thus to food security and poverty alleviation.
       One of the clarion calls of the 1992 United Nations Conference on Environment and Development (UNCED), now generally known as Rio after its venue, was for the preservation of genetic diversity--biodiversity. For many people around the world, this was the first time the phrase had been heard, and they took it to mean the protection of some of the world's vanishing animal species.
        They were not wrong about this, but there is an even more pressing need to preserve plant biodiversity if we are to feed the world's growing population in the 21st century and beyond--and avoid the drastic fluctuations in agricultural production that lead to poverty and famine. Nowhere is this need more urgent than in the world's drylands. In these harsh environments, biodiversity is the key to better yields, and thus to food security and poverty alleviation.
        To understand why, it is necessary to know roughly how crops have hitherto been bred for supply to farmers. Under the methods normally used today, to make plant breeding and seed distribution economic, breeders need to come up with a product that can be grown under the widest possible variety of climatic and environmental conditions. This has two consequences.
        First, they select raw genetic material with moderate genotype (G) by environment (E) interaction. This is a complex area of plant genetics, but G x E interaction is basically the extent to which a plant's behavior is affected by its environment. What the conventional breeder is looking for is a plant that is not too specifically adapted--because s/he wants to grow it over a broad, diverse area. If s/he can't, then, to put it bluntly, it won't pay. In fact, uniformity and broad adaptation are very useful in enabling large-scale, cetralized seed production; so useful that one wonders whether breeding of this sort was designed for seed companies rather than small farmers.
        The second consequence is that s/he will breed for high-input agriculture. Inputs may be defined as external factors introduced by the farmer to make his crop grow, or grow better. Inputs, such as fertilizer, pesticides, or irrigation, tend to make all environments similar; a process of adapting the world to the plants instead of the other way around!
        What the breeder gets from this process is a variety that will grow over a diverse area, provided it receives plenty of inputs. For developed countries with large-scale farmers who can afford such inputs, that's fine. We are not arguing that conventional breeding is bad; for the appropriate environment, it is not, and indeed it is now a highly-developed technique in which much has been learned over the years.
        But ICARDA does not breed only for that sort of farmer. We are very happy when we do produce a genotype that is useful in the developed world (for example, the lentil germplasm we supplied to Australia, and the sources of resistance we found to stripe rust in barley; see Caravan No. 2). But our most important target group is the subsistence farmer, who accounts for 60% of the farmers in the world and grows about 15-20% of its food, mainly in developing countries.
        S/he doesn't use many (or, often, any) inputs. This is only partly because they are expensive or not available. It is also because, in harsh environments like the those in the West Asia and North Africa (WANA) region, farming is a chancy business. In these low-yielding areas, why spend scarce money on fertilizer for a crop that will never pay for them, and may even fail? These farmers therefore do not adopt modern high-yielding varieties. Conventional plant breeders are all too ready to blame this on weak seed-production systems, poor extension services and conservative, uninformed farmers.
        Actually, the farmers are all too well informed! They know not only that varieties bred for high inputs are useless to them, but also that material originating in a high-yielding environment may actually do far worse than their own 'old-fashioned' landraces when grown in a low-yielding site. This is because of a phenomenon called crossover G x E interaction.