A generalized MDS-choice model of stimulus identification is fitted to four matrices of previously published identification confusion data. The model generalizes the standard MDS-choice model (Nosofsky, 1985; Shepard, 1958) by incorporating stimulus- specific weight and angle parameters. Unlike the standard MDS- choice model, the generalized model allows asymmetric similarity relations, and is not a special case of the classic similarity choice model (Luce, 1963; Shepard, 1957). The generalized model fits all four identification confusion matrices significantly better than the standard MDS-choice model, and is competitive with the similarity choice model and with Ashby and Lee's (1991) recently proposed General Recognition Theory of identification.