||On the basis of a diagnostic survey in 1984 and results of previous on farm experiments, a series of twelve verification trials were planted in the Swat Valley (average altitude of 1000 m.a.s.l.) in 1985. These trials were completely managed by the farmer and compared three "improved" technological components, (1) the recommended variety (Azam), (2) phosphorous use and (3) early thinning (target density of 60,000 - 70,000 plants/ha at three weeks after emergence), to the farmers' normal practice. The same set of trials were also conducted in rainfed mountain terraces at an average altitude of 1700 m.a.s.l. These were regarded as exploratory, given the lack of prior information on the mountain areas. Three months after harvest, all cooperating farmers were interviewed to obtain their assessment of trial results. In the irrigated valley, yields were generally high averaging 5.0 t/ha even though farmers' non-recommended practices of broadcast planting at a high seed rate with continuous thinning throughout the season for fodder was employed. The only treatment giving a consistent and significant yield response was Azam variety which yielded 600-800 kg/ha more than the farmers' variety. There was no overall yield response to phosphorous use or early thinning. Other factors explaining yield differences were shaftal as a previous crop (positive), grain moisture percent at harvest and lodging (negative). Initial experiences from the rainfed mountain terraces indicate that this zone is a distinct recommendation domain. Compared to the irrigated valley, variability between fields and within-fields was very high, the incidence of drought, hail, insect and disease problems was much greater and farmers used quite different maize production systems. As on the plains, Azam variety outyielded the farmers' variety but suffered serious smut disease attack. There was response to the other technological yields were early season insect attack, altitude and grain moisture percent ar harvest (all negative effects). In the irrigated valley, famers' assessment of the yield response was generally consistent with trial results and most farmers' intended to plant Azam variety in the nex cycle. However, in the mountains, farmers' assessment bore little relation to trial results, possibly because of lack of measures of land area, and high variability. Implications for further research and extension are identified, particularly the high payoffs to making good seed available to farmers. It is also concluded that farmer managed verification trials offer many advantages over traditional experimental methods in developing and verifying appropriate production recommendations for farmers.