domain, of small farmers in the Amazon Basin of Brazil (Singh, 1990). A number of biophysical environments (land type) exist in the area represented by whether the land was cleared from primary or secondary forest (PF or SF) and whether it was in first or second year of use (the subscripts). WL symbolizes land cleared by bulldozer when the communities were established and is essentially waste land. Four treatments were included: the farmers' local practice (FP), which had essentially no, or very low levels of fertilization; a "full dose" of triple super phosphate (TSP); chicken manure (CM) plus a half dose of TSP; and processed city waste (PCW) from Manaus, plus a half dose of TSP. Two criteria represent several which might be relevant in the area. The criterion Mg ha-1 is more often used by researchers and extension personnel than by farmers. When cash is a scarce resource, as is the case in these communities, kg $-' of cash cost is a common criterion used by resource-poor, small farmers. Other criteria also could be included. The result of using MSA and on-farm research is to enable the creation of multiple extension (recommendation) messages tailored to specific field environments and the different criteria which farmers may use to evaluate new technology, Figure 6. Farmers with fields from all the environments included in on-farm research in a research domain can benefit from this kind of a research-extension program. Following the first year's research program, extension demonstration trials (which also can serve the purpose of verification trials for research and farmer purposes) can be set up so that individual farmers can evaluate the specific recommendations against their current practices. Farmers, extension personnel and researchers all benefit from this additional effort. Table 1. Summary of the recommendation domains and the technology recommended for maize, Rio Preto da Eva, Amazonas, Brazil, based on environmental factors and evaluation criteria. Land Type Evaluation criterion Mg ha-1 kg $- PFI TSP FP SFi CM FP PF2 CM CM SF2 CM CM WL CM Source: Singh, 1990