||This paper analyses the results from on-farm experiments on tillage, variety, fertiliser, and weed control on rainfed wheat production in the high and medium rainfall zones of Pakistan. The experiments were conducted from 1982 to 1988 by the Wheat Programme of the National Agricultural Research Center (NARC) in collaboration with the International Maize and Wheat Improvement Center (CIMMYT). During that period, a total of 230 mostly researcher-managed trials were layed out to obtain quantitative data on potential yields in the area. Fields representing typical land types (mera and lepara) and cropping patterns (wheat after fallow or after maize) were selected. "The wheat growing seasons differed substantially in rainfall and temperature. The years 1983-84 and 1984-85 were generally hot and dry, but drought stress occurred only at the vegetative and grain filling stages during the former, and at the flowering stage during the latter. No moisture stress was experienced at all during the 1982-83 and 1985-86 cycles. Total rainfall during 1986-87 and 1987-88 was above average, but some stress occurred at the early flowering stage in 1986-87, and there was severe moisture stress at the early crop stages of 1987-88, seriously affecting germination and tillering. Average annual yield in the experiments was highly correlated with annual rabi season rainfall. From 1982 to 1987 deep moldboard tillage in the early kharif season before planting wheat led to average annual increases of 25% over yields obtained using conventional cultivator tillage. This effect was more pronounced during the dry years, and was chiefly attributed to better rooting associated with the breakup of the compacted soil below the plough layer. Deep tillage also resulted in more moisture, less weeds and less foot rot disease, thus contributing to improved yields. There was evidence of a positive residual effect on subsequent crops from deep tillage for wheat, making this practice an even more valuable contribution to increased productivity in the barani areas. Deep, moldboard ploughing for the maize crop followed by minimum tillage for wheat also increased the productivity of the maize-wheat system. Moldboard ploughing was no more expensive, and often cheaper, than traditional practices. Adoption of this technology, however, will depend on the availability of moldboard ploughs and training in their proper use. A nitrogen and phosphorus incomplete factorial experiment provided response curve data for calculation of economic recommendations for high rainfall barani areas. Responses varied significantly between years. In 1983-84, when drought occurred during the vegetative period, nitrogen response was low and was not profitable at an MRR equal to one. Phosphorus response was higher in 1983-84 but recommended rates were low (20 kg/P/ha for low wheat prices and 36 kg for a higher price). In 1984-85, a dry year, both N and P responses were significant. However, a land type by P interaction reduced the P recommendation for lepara land. At a MRR equal to one and low wheat prices 47-56 kg N-P was economic. In the wetter 1985-86, and 1986-87 years, N-response was greater resulting in more economic returns from N-fertiliser. In 1985-86 there was also a land type by P interaction that reduced the P recommendation on lepara land. The recommendations for mera and lepara land were 98-82 and 98-8 kg/ha N-P (at an MRR= 1 and low wheat prices). Land type by N and previous crop by N interactions complicated the recommendations in 1986-87. Maize grown before wheat significantly reduced wheat yields (-1535 kg) and needed more N for wheat. Use of farmyard manure increased wheat yields by 691 kg and N recommendations could be reduced on lands receiving this input. No phosphorus response was detected in this year. Recommendations ranged from 31 kg/ha of N on lepara land after fallow to 149 kg/ha of N on mera land after maize at a MRR= 1. Combined data analysis over the 4 years showed a significant year effect and year by N interaction. More nitrogen was needed in wet years than dry years with phosphorus levels remaining the same between years. The overall recommendation was 124-34 kg/ha of N-P for wet years versus 21-34 kg/ha for dry years at a MRR=l. It was recommended that some N and all P be applied basally and further N be top-dressed if rainfall is favourable to the end of tillering. In varietal trials, Pak-81, S-19 (Junco "S") and Barani-83 provided higher yields and greater rust resistance than the predominant farmer variety, Lyallpur-73. Weeds were not a serious problem. Broadleaf types were easily controlled by the relatively cheap phenoxy herbicides 2,4-D, MCPP or MCPA, or benzoic acid compounds like Dicamba. However, the common local practice of intercropping wheat and mustard rules out chemical weed control for most farmers. Recommendations in this report for barani wheat growing in the high rainfall areas should allow farmers to easily obtain an average yield of 4.0-4.5 tons/ha, in favourable years. Suggestions for further research are also given.