Estimating the number of repetitions of data obtained from several experiments performed ingiven location is important to minimize human and financial resources and leverage the existinginformation. The objective of this research was to determine how many evaluations (replications)were necessary to predict productive and morphologic characters of soybean genotypes (Glycinemax L. Merril) and to verify its variability in the number of replications among characters. Twenty-eightsoybean genotypes were evaluated in five experiments conducted in Jaboticabal, São Paulo State,Brazil, from 2002 to 2005. Randomized blocks design with three repetitions was installed to evaluatethe following productive characters (seed yield, number of pods per plant and mass of one hundredgrains) and morphologic (plant height at maturity, height of first pod insertion, number of nodesper plant and number of branches per plant). It was performed analysis of variance, estimated therepeatability coefficient and calculated the number of repetitions necessary to predict the realvalue of the genotypes. It was found variability in the number of replications among the productiveand morphologic characters. Experiments with three replicates identify soybean genotypes withhigher 17.52% (height of first pod insertion) to 61.07% (number of nodes per plant) of real accuracy.To improve the accuracy of the evaluation of characters in soybean plants, you need to use agreater number of repetitions.A estimativa do número de repetições obtida de dados de vários experimentos realizados num determinado local é importante por minimizar recursos humanos e financeiros e aproveitar as informações existentes. O objetivo deste trabalho foi determinar o número de medições (repetições) necessários para avaliar caracteres de produtividade e de morfologia em genótipos de soja (Glycine max L. Merril) e avaliar a variabilidade do número de repetições entre os caracteres. O estudo constituiu-se na avaliação de 28 genótipos de soja por meio de cinco experimentos conduzidos em Jaboticabal, São Paulo, Brasil, no período de 2002 a 2005. Utilizou-se o delineamento de blocos ao acaso, com três repetições. Avaliaram-se os caracteres de produtividade (produtividade de grãos, número de vagens por planta e massa de cem grãos) e de morfologia (altura de planta na maturação e de inserção de primeira vagem, número de nós por planta e de ramos por planta). Foi realizada a análise de variância, estimado o coeficiente de repetibilidade e calculado o número de repetições necessário na predição do valor real dos genótipos. Há variabilidade da estimativa do número de repetições entre os caracteres de produtividade e de morfologia. Experimentos com três repetições identificam genótipos de soja superiores com 17,52% (altura de inserção de primeira vagem) a 61,07% (número de nós por planta) de exatidão no prognóstico de seu valor real. Para melhorar a precisão da avaliação de caracteres em plantas de soja, necessita-se utilizar maior número de repetições. Estimated number of replications for evaluation of productive andmorphologic characters in soybean genotypesAbstractEstimating the number of repetitions of data obtained from several experiments performed ingiven location is important to minimize human and financial resources and leverage the existinginformation. The objective of this research was to determine how many evaluations (replications)were necessary to predict productive and morphologic characters of soybean genotypes (Glycinemax L. Merril) and to verify its variability in the number of replications among characters. Twenty-eightsoybean genotypes were evaluated in five experiments conducted in Jaboticabal, São Paulo State,Brazil, from 2002 to 2005. Randomized blocks design with three repetitions was installed to evaluatethe following productive characters (seed yield, number of pods per plant and mass of one hundredgrains) and morphologic (plant height at maturity, height of first pod insertion, number of nodesper plant and number of branches per plant). It was performed analysis of variance, estimated therepeatability coefficient and calculated the number of repetitions necessary to predict the realvalue of the genotypes. It was found variability in the number of replications among the productiveand morphologic characters. Experiments with three replicates identify soybean genotypes withhigher 17.52% (height of first pod insertion) to 61.07% (number of nodes per plant) of real accuracy.To improve the accuracy of the evaluation of characters in soybean plants, you need to use agreater number of repetitions.