The texture of 64 mangoes of different commercial varieties was analyzed by a trained sensory panel. The intensities of the characters "firmness," "crunchiness," "melting" and "juiciness" were noted from 0 to 10. Soluble solids content, visible/ NIR spectrum and texture of the same fruits were determined using instrumental methods. Texture was determined by means of mechanical and acoustic tests. Four sensory classes were identified by agglomerative hierarchical clustering from among the fruits analyzed andwere characterized in terms of the measured physicochemical parameters. The relationships between the sensory and instrumental texture variables were visualized using a multiple factorial analysis. The prediction of sensory attributes from instrumental data was better for firmness, crunchiness and melting than for juiciness. Power-law or logarithmic models were proposed to predict firmness (r = 0.89), crunchiness (r = 0.89) and melting (r = 0.88). Multivariate models (multiple linear regression, partial least squares regression) were also proposed.