Associations between primary and secondary components of fruit production in sweet-pepper genotypes were evaluated by means of canonical correlation analysis. Additionally, a study of data fitness for this analysis was done as a result of observing multicoUinearity in the correlation matrices used. The experiment was carried out in the municipality of Viçosa (MG), during the period from October 6th 1993 to March 19th 1994. The mean annual relative air humidity of the region is 80%, the mean annual temperature is 20°C and the mean annual precipitation is 1340 mm The experimental design used was complete randomized blocks with three replications. Each plot had a row with six plants, spaced 0.40 m, with a space of 0.9 m between lines. The four central plants represented the useful plot The study of correlations and of canonic pairs revealed that there was a significant dependence (P > 0.01), by the chi-square test, between morphological and agronomic characters (peduncle and stem diameters, and plant and first bifurcation heights) and fruit (fruit length, width and length/width relation and pulp thickness) with the primary yield components (fruit average weight and total number). The most expressive association between these components was established with fruit characters. For these characters, fruit length, width and pulp thickness should be considered simultaneously to obtain gains in total number of fruits and, mainly, in fruit average weight. For morphological and agronomic characters, peduncle diameter stands out as it one for morphological and agronomic characters, peduncle diameter stands out as it is the one character correlated with fruit average weight, correlated with fruit average weight.
Capsicum annuum L.; multicoUinearity; fruit average weight; fruit characters.