2002, Brazilian Society of Plant Breeding 171 Crop Breeding and Applied Biotechnology, v. 2, n. 2, p. 171-178, 2002 INTRODUCTION Tandem selection and selection based either on independent culling levels or on an index are used for the selection of superior genotypes by considering several traits. Hazel and Lush (1942) and Young (1961) compared the relative efficiency of these methods in terms of genetic gain and concluded that, theoretically, selection based on an index was never less efficient than the other two methods. Selection indices combine genetic and economic information in a regression equation for the prediction of the value of a genotype. In addition, they combine the information for many traits into a single value (Bridgwater et al., 1983). Despite its relevance, selection index theory presents certain limitations both of a practical and a methodological order. The major limitations described by Lin (1978) are: 1) imprecise estimates of variances and covariances, 2) possibility of changes occurring in the parameters with selection, and 3) difficulty in establishing the relative importance of the traits. However, these complicating factors apply to the general problem of multitrait selection and do not necessarily weigh more heavily against the selection index than against other approaches to that problem (Martin et al., 1982). Selection based on an index may tend to choose genotypes that express the highest phenotypic values for the traits targeted to obtain positive genetic gains and the lowest phenotypic values for traits targeted to obtain negative genetic gains. However, in some cases, the phenotypic value considered optimum is in the range between the maximum and minimum Selection based on distances from ideotype Claudio Guilherme Portela de Carvalho 1 ; Cosme Damião Cruz* 1 ; José Marcelo Soriano Viana 1 and Derly José Henriques da Silva 2 1 Departamento de Biologia Geral, Universidade Federal de Viçosa, CEP 36571-000, Viçosa, MG, Brazil; 2 Departamento de Fitotecnia, Universidade Federal de Viçosa, CEP 36571-000 Viçosa, MG, Brazil. (* Corresponding Author. E-mail: cdcruz@mail.ufv.br) ABSTRACT Two selection methods based on distance from ideotype using selection index and principal components are proposed. They took into consideration the existence of an optimum range for each trait, the difference in scale of the measurements and the relative importance of the traits. The index was also based on mean Euclidean distance. Analysis performed by considering four eggplant traits and a 35,3% selection pressure showed that the two selection methods presented the same predicted genetic gains. KEY WORDS: Eggplant, euclidean distance, principal components, selection index. values expressed by the genotypes. An optimum range can be established for these traits, outside of which the product undergoes devaluation on the consumer market or is curtailed in production. In eggplants, for example, fruit diameter and length are traits included in the first case and plant height is included in the second case. In selecting for height, very tall or very short plants are considered undesirable because of harvesting difficulties. The phenotypic values used to obtain a selection index are located outside the optimum range should therefore be depreciated. Depreciation refers to the increase in the distance of the phenotypic value from the value considered optimum. Harding et al. (1991) elaborated an index denoted weighted Euclidean distance index to classify phenotypes in relation to an optimum phenotype, which may be intermediate in value. Tourjee et al. (1995) applied this index to select genotypes expressing phenotypic values that minimize the mean distance in relation to an optimum value for flower colour in the Davis population of Gerbera. Although genotype scores are obtained by calculating the difference between the reference point and the observation for each trait, this index does not take into consideration the difference in scale of the measurements or the existence of an optimum range for each trait. The purpose of this report was to present two selection methods based on distance from ideotype: a) use of the selection index, and b) analysis of graphic dispersion on axes established by principal components. The index was constructed and the study by principal components was conducted taking into