Studies about phenotypic diversity are common in germplasm banks. The inference about this variability must be realized through several methods. The multidimensional scaling is a multivariate technique that has not yet been well explored in plant breeding programs. The objective of this study was to evaluate the phenotypic dispersion of landrace lima bean varieties using the non-metric multidimensional scaling technique (nMDS) based on seed morphology. Seeds of 25 lima bean accessions were characterized based on the morphological descriptors proposed by the International Plant Genetic Resources Institute. Distance matrices between the accessions were estimated based on the qualitative and quantitative variables, in addition to simultaneous analysis of the qualitative and quantitative data, using the Mahalanobis and Gower distances. The distances were represented by non-metric multidimensional scaling. The adjustment level of the nMDS mapping was calculated using Kruskal’s Stress. The scaling based on the quantitative and mixed data was efficient to represent the distances of the lima bean accessions in the bidimensional plane presenting Stress less than 20%. Divergent accessions, such as 11, 13, 17 and 25 were identified. The inclusion of qualitative characters provided the best discrimination of the accessions, confirming the importance of the simultaneous character analysis. The nMDS must be used as a complementary technique to those commonly employed in studies of phenotypic diversity in lima beans.