Obtaining information about the main morphoagronomic and yield components, as well as their interrelationships, is of utmost importance in cowpea breeding programs, aiming at the selection of superior genotypes. The analysis of canonical correlations allows us to predict and describe the relationship between groups of characters, allowing the indirect selection of superior genotypes. The objective of this work was to verify relationships between morphological characters and production components through the analysis of canonical correlations using genotypic values obtained via mixed REML/BLUP models in traditional cowpea varieties. The experiment was conducted in a randomized block design with three replications and 12 treatments. Traditional varieties were characterized and evaluated during the germination, seedling, flowering and harvest stages with data obtained from plants, pods and seeds. Using the genotypic correlation matrix, two sets of characteristics were established, Group I, formed by production components and Group II, formed by morphoagronomic characters. A second correlation analysis was performed to verify associations between morphoagronomic and seed traits (seed length, width and thickness). It was found that dependence between the groups evaluated and intergroup associations can be established, which allows the study of cause-effect and the practice of indirect selection between groups of variables studied. An increase in pod length, hundred-grain mass and seed size is achieved through the selection of early maturing plants, with greater width of the apical leaflet and length of the hypocotyl.