Psychophysical inference of frequency-following fidelity in the neural substrate for brain stimulation reward

•We test a model of frequency following in the substrate for brain stimulation reward.•The measurement strategy is based on the counter model of reward integration.•We measure current- vs. pulse-frequency trade-off functions in self-stimulating rats.•The psychophysical data are well described by the model.•The asymptotic value was high, implicating fast-firing, non-dopaminergic neurons.•The function has important implications for Shizgal's 3D reward-mountain model.

The rewarding effect of electrical brain stimulation has been studied extensively for 60 years, yet the identity of the underlying neural circuitry remains unknown. Previous experiments have characterized the directly stimulated (“first-stage”) neurons implicated in self-stimulation of the medial forebrain bundle. Their properties are consistent with those of fine, myelinated axons, at least some of which project rostro-caudally. These properties do not match those of dopaminergic neurons. The present psychophysical experiment estimates an additional first-stage characteristic: maximum firing frequency. We test a frequency-following model that maps the experimenter-set pulse frequency into the frequency of firing induced in the directly stimulated neurons. As pulse frequency is increased, firing frequency initially increases at the same rate, then becomes probabilistic, and finally levels off. The frequency-following function is based on the counter model which holds that the rewarding effect of a pulse train is determined by the aggregate spike rate triggered in first-stage neurons during a given interval. In 7 self-stimulating rats, we measured current- vs. pulse-frequency trade-off functions. The trade-off data were well described by the frequency-following model, and its upper asymptote was approached at a median value of 360 Hz (IQR = 46 Hz). This value implies a highly excitable, non-dopaminergic population of first-stage neurons. Incorporating the frequency-following function and parameters in Shizgal's 3-dimensional reward-mountain model improves its accuracy and predictive power.

Graphical abstractDownload high-res image (172KB)Download full-size image