Townsend et al. (1984) show evidence that human feature perception changes systematically with number of features present. Single feature sensitivity (d' from signal detection theory) and bias (B from signal detection theory) both decrease with number of features present. In addition, there are positive report dependencies (probability of identifying a feature alone compared to the joint identification) and conditional trends. These results can be explained by a descriptive model of processing ( Blobloc, Townsend et al., 1988). This thesis considers a concretization of Blobloc using a mathematical decomposition of visual stimuli (the Fourier Decomposition) to provide a computational basis for the model. Perceptual and decisional dependencies between spatial frequency and orientation are evaluated using General Recognition Theory (Ashby & Townsend, 1986). Confusion data from three identification tasks support that Blobloc be realized as a frequency-orientation selective, resource-reallocating processor on the Fourier space.