Experiments to assess the yield and economic performance of a maize-groundnut rotation compared with continuous maize (both when inorganic fertilizer was applied to maize and when not), were conducted under management by smallholder farmers in Zimbabwe over six years. The experiment was planted on-station near Harare and at six smallholder sites in northeast Zimbabwe, predominantly on sandy soils. Fertilizer rates and practices were those used by farmers, as described in surveys. On-farm grain yields from continuous maize without fertilizer were generally in the range 0.5–0.8 t ha−1 over five years. Maize yield responses to inorganic fertilizer on smallholder farms were highly variable, but moderate (up to 29 kg grain per kg N) with adequate rainfall. With no inorganic fertilizer applied to maize, the on-station groundnut crop (producing 0.260–0.355 t ha−1 shelled grain) almost doubled the grain yield of the following maize crop (in 1995–96), increasing output from 2.46 t ha−1 to 4.61 t ha−1. Where inorganic fertilizer was applied to maize, the rotation produced even more additional maize grain (an increase of 2.93 t ha−1). Up to 0.50 t ha−1 extra grain was obtained in the second year of maize following groundnut (1996–97). With inorganic fertilizer, groundnut improved the grain yield of following maize crops at only two of five on-farm sites. Without fertilizer, the groundnut rotation increased maize grain yields at five on-farm sites by an average of only 0.28 t ha−1. For the on-station groundnut and two subsequent years with maize, discounted net benefits (DNBs) over cash costs (seed and fertilizer) were greater for the rotation than for continuous maize, irrespective of whether or not inorganic fertilizer was applied. When labour costs were added, continuous maize plus fertilizer showed better returns than did the rotation, while the returns for the rotation and continuous maize without fertilizer were almost the same. On-farm the rotation was far less profitable. At only two sites, DNBs over cash costs were higher for the rotation whereas DNBs over all costs (including labour valued at a local casual-worker wage) were always negative or close to zero. At three sites, it was far more profitable to grow continuous maize, especially with fertilizer. These findings of low groundnut yield, marginal to zero profitability, and high labour cost of groundnut-maize rotations, support and explain the general trend by smallholder farmers to reduce groundnut area in Zimbabwe.