Pearl millet varieties for on-farm testing in Namibia were selected from onstation testing during the 1990/91 and 1991/92 seasons. The on-farm trials were conducted at up to 14 locations in Owambolands and Kavango during the 1992/93 and 1993/94 seasons. During both seasons research-managed farmerimplemented (RMFI) trials were conducted. Farmer-managed farmer-implemented (FMFI) trials were also conducted at five locations during the 1993/94 season. Under RMFI trials, Okashana 1 gave significantly higher grain yields over the farmers' local landrace variety (LLV) in Owambolands. In Kavango the performance of Okashana 1 and the farmers' LLV control was the same over the last 2 years of testing. Under farmer management all the varieties produced similar yields to the farmers' LLV. However, in the RMFI trial, varieties responded to fertilizer, and Okashana 1 was significantly superior to the farmers' LLV in respect of grain yield. Without any improvement in crop management practices, there seemed to be a little yield gain from cultivar change. None of the improved varieties under test produced higher grain yield than Okashana 1. The development of pearl millet cultivars needs to be targeted to zones currently broadly defined as Owambo and Kavango.