Fundamental Flaws In The Derivation Of Stevens'Law For Taste Within Norwich's Entropy Theoryof Perception

摘要

Norwich's Entropy Theory of Perception (1975-present) is a general theory of perception, based on Shannon's Information Theory. Among many bold claims, the Entropy Theory presents a truly astounding result: that Stevens' Law with an Index of 1, an empirical power relation of direct proportionality between perceived taste intensity and stimulus concentration, arises from theory alone. Norwich's theorizing starts with several extraordinary hypotheses. First, "multiple, parallel receptor-neuron units" without collaterals "carry essentially the same message to the brain", i.e. the rate-level curves are identical. Second, sensation is proportional to firing rate. Third, firing rate is proportional to the taste receptor's "resolvable uncertainty". Fourth, the "resolvable uncertainty" is obtained from Shannon's Information Theory. Finally, "resolvable uncertainty" also depends upon the microscopic thermodynamic density fluctuation of the tasted solute. Norwich proves that density fluctuation is density variance, which is proportional to solute concentration, all based on the theory of fluctuations in fluid composition from Tolman's classic physics text, "The Principles of Statistical Mechanics". Altogether, according to Norwich, perceived taste intensity is theoretically proportional to solute concentration. Such a universal rule for taste, one that is independent of solute identity, personal physiological differences, and psychophysical task, is truly remarkable and is well-deserving of scrutiny. Norwich's crucial step was the derivation of density variance. That step was meticulously reconstructed here. It transpires that the appropriate fluctuation is Tolman's mean-square fractional density fluctuation, not density variance as used by Norwich. Tolman's algebra yields a "Stevens Index" of -1 rather than 1. As "Stevens Index" empirically always exceeds zero, the Index of -1 suggests that it is risky to infer psychophysical laws of sensory response from information theory and stimulus physics while ignoring empirical biological transformations, such as sensory transduction. Indeed, it raises doubts as to whether the Entropy Theory actually describes psychophysical laws at all.