An on-farm study was carried out at 17 different sites within the Western Hills of Nepal during the spring and summer seasons of 1993 and 1994 in order to monitor the plant population density dynamics of the maize (Zea mays L.) crop under farmers' management. Five farmers were selected randomly at each site and a 20-m2 plot was demarcated just after maize sowing. The maize stands within the 20-m2 plots were first recorded 20 d after sowing, and then at 10-d intervals until harvest. Farmers were allowed to perform all cultural practices on their own. On average, the initial plant stand across all locations and over two seasons was 69 100 ± 1800 plants ha−1, which resulted in 37 300 ± 1700 plants ha−1 at harvest, a 46% reduction from the initial stand. At most locations, the population declined by 30–40 × 103 plants ha−1 between emergence and harvesting and in no cases were the recommended plant populations achieved at harvest. Populations varied across sites and seasons. The rate of population decline was higher in the low and mid hills compared to high-altitude areas. Thinning which is carried out between 30 and 60 d after sowing at most sites was the major cause of population reduction, and delay in thinning was intended to increase fodder production from the thinned maize. In addition, the removal of barren plants at the silking stage and of smutted, lodged and damaged plants attributed to about 12% of total reduction in plant population. The larger the initial stand, the larger was the proportion of barren plants to be removed. No clear relationships were observed between soil fertility and final plant stand at harvest.