Lütkenhöner’s ‘Intensity dependence of auditory responses’: an instructional example in how not to do computational neurobiology

In "Threshold and beyond: modeling the intensity dependence of auditory responses" (JARO 9, 2008, pp. 102-121), Dr. Bernd Lütkenhöner introduces us to a puzzle, namely, that responses to stimuli of low intensity appear to climb as the logarithm of the intensity, contrary to the expectation that they be linear with intensity. Towards solving that puzzle, Dr. Lütkenhöner has produced the latest in a long line of attempts by various authors to model "the gross response of the population of auditory nerve fibers" in response to a pure tone (Lütkenhöner, p. 102). From his model, Dr. Lütkenhöner predicts linearity with intensity at low intensities, changing to linearity with log-intensity at higher stimulus intensities. The present author is one of those who have contributed to this particular field of modeling (Nizami & Schneider, 1997; Nizami 2001, 2002, 2005a, 2005b), and was intrigued by what new insights Dr. Lütkenhöner might have to offer. However, substantial omissions and mistakes were found, which greatly reduce the value of Dr. Lütkenhöner’s contribution. In particular, Lütkenhöner omits to mention the many earlier contributions towards modelling the firing of the whole auditory nerve or some portion thereof, so that his model escapes comparison to those of others. His computations use data taken from a variety of species, whose firing-rate characteristics are known to differ, such that his computations may describe no real species at all. Further, Dr. Lütkenhöner’s equations do not account for the known variance of dynamic range across neurons, a crucial component of any model of mass neural firing. Indeed, he encourages the use of a discredited equation that cannot actually account for dynamic range variation, while ignoring a proven equation than can. Lütkenhöner also perpetuates the absurd notion of an infinitely low detection threshold for auditory stimuli. Finally, Lütkenhöner’s derivations provide an equation of a form known to arise from circular logic. All of these errors are incontrovertible and render Dr. Lütkenhöner’s work null and void. Altogether, Lütkenhöner’s errors serve as a warning to those who attempt to compute the average neuronal response available from a mass of responding neurons, but serve the larger purpose of illustrating the folly of accepting conventional wisdom and frequently-cited papers on face value without delving deeply enough into the literature to achieve a professional understanding of a problem.