The objective of this paper was to present a methodology for the analysis of experiments in plant pathology that considers the comparison of disease progress curves in the presence of a large number of treatments by cluster analysis. Forty-two accessions were grown from the Germoplasma Vegetable Bank (BGH), of Universidade Federal de Viçosa (UFV). The exponential model was fitted to the data of late blight severity percentage, and the obtained parameter estimates obtained on the initial incidence of the disease (yo) and rate of disease progression (r) - were submitted to the multivariate analysis of variance (MANOVA). The adjusted means were submitted to the cluster analysis. An optimal number of six distinct groups was observed.