Color vision impairment was examined in patients with type 2 diabetes mellitus (DM2) without retinopathy. We assessed the type and degree of distortions of individual color spaces. DM2 patients (n = 32), and age-matched controls (n = 20) were tested using the Farnsworth D-15 and the Lanthony D-15d tests. In addition, subsets of caps from both tests were employed in a triadic procedure (Bimler & Kirkland, 2004). Matrices of inter-cap subjective dissimilarities were estimated from each subject's "odd-one-out" choices, and processed using non-metric multidimensional scaling. Two-dimensional color spaces, individual and group (DM2 patients; controls), were reconstructed, with the axes interpreted as the R/G and B/Y perceptual opponent systems. Compared to controls, patient results were not significant for the D-15 and D-15d. In contrast, in the triadic procedure the residual distances were significantly different compared to controls: right eye, P = 0.021, and left eye, P = 0.022. Color space configurations for the DM2 patients were compressed along the B/Y and R/G dimensions. The present findings agree with earlier studies demonstrating diffuse losses in early stages of DM2. The proposed method of testing uses color spaces to represent discrimination and provides more differentiated quantitative diagnosis, which may be interpreted as the perceptual color system affected. In addition, it enables the detection of very mild color vision impairment that is not captured by the D-15d test. Along with fundoscopy, individual color spaces may serve for monitoring early functional changes and thereby to support a treatment strategy.