Joint contribution of adaptation and neuronal population recruitment to response level in visual area MT: a computational model

Abstract

Adaptation is a form of short-term plasticity triggered by prolonged exposure to a stimulus, often resulting in altered perceptual sensitivity to stimulus features through a reduction in neuronal firing rates. Experimental studies have explored adaptation to bistable stimuli, specifically a stimulus comprising inward-moving plaids that can be perceived as either a grating moving coherently downward or two plaids moving incoherently through each other. Functional magnetic resonance imaging (fMRI) recordings have shown higher activity during incoherent perception and lower activity during coherent stimulus perception. There are two potential explanations for the underlying neural mechanisms: a weaker coherent stimulus response may result from stronger adaptation to coherent versus incoherent motion, or a stronger incoherent stimulus response could stem from the involvement of more neural populations to represent motion in more directions. Here, we employ a computational model of visual neurons with and without firing rate adaptation to test these hypotheses. By simulating the mean activity of a network of thirty-two columnar populations of visual area MT, each tuned to one direction of motion, we investigate the impact of firing rate adaptation on the blood-oxygen-level-dependent (BOLD) signal generated by the network in response to coherent and incoherent stimuli. Our results replicate the experimental curves both during and after stimulus presentation only when the model includes adaptation, highlighting the importance of this mechanism. However, our findings reveal that the response to incoherent motion is larger than the response to coherent motion for a wide variety of stimulus parameters and adaptation regimes, suggesting that the observed reduced response to coherent stimuli is most likely due to the activation of smaller neuronal populations, in alignment with the second hypothesis. Hence, adaptation and differential neuronal recruitment work together to give rise to the observed hemodynamic responses. This computational work sheds light on experimental results and enriches our understanding of the mechanisms involved in neural adaptation, particularly in the context of heterogeneous neuronal populations.