Assumptions concerning orthogonality in threshold-scaled versus cone-contrast colour spaces

A recent article by D'Zmura and Knoblauch (1998) argued from experiments using sectored noise that chromatic detection is subserved by multiple post-receptoral mechanisms, and not solely by mechanisms tuned to specific cone-opponent 'red – green' and 'blue– yellow' directions as has been proposed previously. An important feature of their analysis is their use of angles within a threshold-scaled cardinal colour space to determine the response of a mechanism to a given stimulus. In this letter, we point out that the angle between the colour direction of two stimuli depends on the space in which the stimuli are represented. Moreover, the choice of each space embodies specific assumptions about chromatic detection. Specifically, the use of threshold-scaled cardinal space in the sectored noise study assumes that the visual response arises from the linear combination of postreceptoral outputs. In this letter, we use a more standard assumption of chromatic detection , that of linear summation at the cone level, to demonstrate that their data are consistent with earlier findings that chromatic detection is served by two fundamental mechanisms. This letter, as a whole, addresses an important issue regarding the appropriate use of colour spaces in quantitative analysis and its dependence on underlying assumptions concerning chromatic detection mechanisms. D'Zmura and Knoblauch's sectored-noise study follows from previous experiments that use to investigate the mechanisms subserving chromatic detection (e.g. In all these experiments, the detection threshold of a fixed test is measured in the presence of a variable chromatic noise mask. In the sectored-noise approach of D'Zmura and Knoblauch, the noise mask consists of two orthogonal vector components: a fixed component N T coincident in direction with the test vector, and a variable component N P perpendicular to this direction in the chosen colour space (Fig. 1). They demonstrate that the variable noise component N P has no masking effect on the test, and therefore correctly conclude that the underlying detection mechanism is orthogonal to the chromatic direction of this noise component. Adopting a threshold-scaled cardinal space, they argue that this orthogonal mechanism direction always coincides with the direction of the test, and not with fixed red–green and blue–yellow axes as observed in previous studies Firstly, it is clear, that establishment of an orthogonal direction in a given colour space is dependent on the choice of colour space itself. This is illustrated in Fig. 2. In Fig. 2(a), the vectors u and 7 are set in …