The rapid appraisal and farmer selection procedures, preceding the establishment of over fifty agroforestry research trials on farms in Talamanca, Costa Rica are described. The highest probability of sucess and impact of these long-term collaborative trials is obtained by selecting innovative, experienced, motivated and locally respected farmers. The methodological lessons learned from implementing three types of researcher managed trials (shade-cacao; black pepper on living support posts; timber production in boundary lines) are discussed. The importance of fitting farmer selection criteria to project objectives, experimental designs to on-farm limitations, and project goals to farmers' goals, are emphasized.