Über die Detektion visueller Geschwindigkeitsänderungen in Primaten : Neuronale Repräsentation im Areal MT des Makaken und Modellierung der perzeptuellen Verhaltensleistung des Menschen

Abstract

Neurons in the mediotemporal area of the macaque (MT) respond to accelerations of visual stimuli with a high transient increase in firing rate that decays to a markedly lower reponse plateau after a few hundred milliseconds. The latency of this transient response correlates with reaction time in tasks requiring the detection of such changes. However, it remains unknown how the sign, the amplitude and the latency of transients in MT generally depend on the sign and the size of a speed change, and how this interacts with a neuron's tuning to absolute speeds. Moreover, it has never been adressed whether such transients may explain speed change detection at a behavioral level. By characterizing the representation of a wide range of positive and negative speed changes in MT of awake and behaving macaques, this thesis obtains three major findings: First, neuronal transients do not only depend on the absolute speed after a speed change, but are scaled in such a way that their relative amplitudes systematically reflect the size of the speed change itself. For each neuron, this scaling increases with the distance of a speed change from its preferred speed, and depends on its transient-sustained-ratio as a measure of the extent to which it is subject to short-term adpatation within the local ciruitry of MT. Second, simulations in the framework of signal-detection-theory demonstrate that the detection rates of human observers under similar conditions can be reproduced by a physiological detection model that applies an upper and lower threshold, respectively, to positive and negative population transients of a moderate number of MT neurons. Third, the latency of the transient responses is found to mirror the pattern of reaction times in detection tasks, and fullfills predictions of a psychophysical model of motion detection by its dependence on the the absolute value of a speed change and an incomplete normalization for higher base speeds. Because of this close relation between the neuronal representation and measures of detection behavior, this thesis draws the conclusion that neuronal transients in MT are systematically shaped to facilitate the detection of speed changes by the application of a simple threshold mechanism to the population response. It highlights yet another functional purpose of adaptation and presents a framework for future analyses of the detection of changes in other motion or stimulus domains.