A heterogeneous cucumber (Cucumis sativus L.) population (mostly gynoecious) was evaluated at five locations for single-plant fruit yield at the mature-fruit stage in 1981. Seeds from the highest-yielding plants were then harvested, combined, and partitioned into five lots. Seeds were combined such that each location received only the superior genotypes from the other four locations. This procedure was continued for an additional four cycles using two types of selection: single-plant selection for fruit number at the mature-fruit stage (1981–82) and half-sib family selection at the once-over harvest stage (1983–84). In 1985, yield improvement from selection was measured by compositing the seeds of the selected plants or families from each of the four cycles and five locations and planting them at the five locations. No progress was made for total, marketable, or early yield. Percentage of culls was reduced an average of 0.7% per cycle. Genotype by environment interaction among the diverse locations may have prevented progress for yield.