In Benin, average cotton yields have stagnated for several years. Although the main factors are known to bc decreasing levels of inputs by the farmers, yield improvement has become a research priorityfor cotton research as a whole, including plant breeding. Selecting variables with low heritability like yield requires specific experimental field design, more demanding than those usedfor less complex characters. The design in which single crosses are evaluated through the performance of an F3, 30 lines saniple as described in this paper intends to meet this requirement. The lines are grown in cropping conditions that are similar to those of commercial fields. After examining the general parameters status under which the design is most relevant, examples show how to interpret the data: (1) from the statistical analysis, the crosses are compared and their lines ranked in each cross; then (2) with an appropriate quantitative genetics model, the genetic variability of the traits under selection can be described more precisely. Finally, the cost-effectiveness of this design is discussed and compared to more common designs.