An outline is given of the designs and procedures of statistical analysis followed in coffee experiments at the Instituto Agronômico de Campinas. Four groups of experiments are considered: 1) fertilizer tests; 2) varietal trials; 3) progeny tests; and 4) miscellaneous experiments. Under 1) the writers discuss old experiments with systematic layouts and more recent ones with randomized designs. A factorial experiment supplied conclusive results within a few years. The problem of changing some of the treatments arose in one of the experiments, and the solution proposed by the writers is discussed. Under 2) the treatment of data used by W. L. Stevens (1949) in the analysis of a systematic experiment comparing coffee varieties is commented in detail. The evolution in designs used for progeny comparisons is discussed under 3). In early tests progenies were compared on the basis of rows of 20 plants without replications; later the designs were changed to replicated plots of 4 plants per plot, and more recently to plots of a single plant with a higher number of replications. The results of spacing trials and of tests comparing several methods of planting-coffee are discussed under 4.) The plot size for coffee experiments is also discussed,, based on individual yield records from a planting of the Bourbon variety.