Nitrogen fertilizer rate by timing trials were conducted on bread wheat (Triticum aeslivum L) in peasant farmers' fields during two annual cropping seasons for three years in a drought-prone district of southeastern Ethiopia. Nitrogen rates had an incremental effect on grain yield, but the response to N was economically acceptable only up to 41 kg N ha-1, and then only in the more reliable second growing season. Mean grain yields obtained in different growing seasons were correlated with seasonal rainfall totals, but nitrogen response was not. Early application of N, either all at sowing or split applied between sowing and mid-tillering, resulted in the highest yield increments, and also significantly increased the number of wheat spikes m-2, grains spike-1 grains m-2, plant height, and biomass yield, while decreasing harvest index. Thousand kernel weights and broadleaf weed seedling densities were unaffected by N rate and timing. Split application of N should be recommended as an economic risk-aversion strategy in this marginal rainfall zone.