Regional agricultural systems are continually constituted and reconstituted by the strategies of individual farmers as they draw upon, and iteratively change, the resources to which they have access, and the constraints to which they are subject because of their personal experience and their membership of a particular social group. One important factor in regional agricultural change is the propensity of farmers to experiment with and change their agricultural strategies. The range of potential experiments is, however, limited by what the farmer conceives as possible. Moreover, the potential long-term success of such innovation depends upon the acquisition of accurate knowledge of the new crop and the environmental factors to which it is subject.
From 1977 to 1980, the International Potato Center (CIP), in collaboration with Peru's Ministry of Agriculture and the International Maize and Wheat Improvement Center (CIMMYT), conducted a series of farm-level surveys and experiments in the Mantaro Valley of Highland Peru. The lessons of the Mantaro Valley Project can be summarized in 12 points: (1) Ecological conditions and farm type strongly influenced farmers' production methods and technological requirements; (2) Small farmers were eager to improve their methods; (3) The recommended 'technological packages' had many deficiencies; (4) Technical knowledge was available to solve some farmers' problems; (5) Most technologies could not be directly 'transferred' to farmers; (6) Scientific specialization presented several barriers to interdisciplinary research; (7) The project was costly, in terms of operating capital; (8) The project's most valuable result was an institutional innovation not an impact on potato production; (9) Informal surveys and simple on-farm trials had many advantages over more 'respectable', formal methods; (10) Anthropologists played many useful roles; (11) Many results of the on-farm research could be extrapolated; (12) On-farm research was useful for identifying and solving production problems within existing systems, but not for designing entirely new systems.