A framework is presented for using threshold data to test hypotheses about the action of a disease, a chemical agent, or a developmental process. A model of the normal rod system, based on models from the physiological and psychophysical literature, is presented. Hypotheses about the alteration of the rod system are specified in this model. The approach is illustrated with a class of hypotheses that places the decrease in sensitivity with retinal disease at the rod receptors and with data from patients with retinitis pigmentosa and congenital stationary night blindness. The implications for models of the normal rod system are considered.