A comprehensive account of wavelength discrimination and colour saturation discrimination is given in terms of optimum probabilistic signal detection. The theory is a logical deduction from statistical estimation theory of the visual estimate of the spectral parameters of the stimulus. In place of geometrical concepts associated with colour-space geometry, stimulus discriminability is determined by optimum decision rules given by likelihood ratio tests on statistics that are postulated for the trichromatic responses. The classical line element theory and its formulations are deduced to be discriminability measures between signals. The different mathematical forms of classical theory are shown to correspond to different statistical constraints.