Natural conditions and major resource differences such as farm size are commonly used in on-farm research to define homogeneous 'recommendation domains' within which new technologies are tested. Successful trial results within these domains may not necessarily imply uniform acceptance across farmers, because an array of socio-economic factors influences the extent of the benefits accruing to individual farmers. Analysis of on-farm trial results from Zimbabwe and Swaziland is used to show that variations in the socio-economic circumstances of small farmers operating in similar natural environments substantially alter the benefits of new technologies across farmers in these domains. These findings have implications for how extension messages are developed and delivered.22-26