Neural Representation Of Vestibular-visual Implicit Causal Inference In The Ventral Intraparietal Cortex In Rhesus Monkeys

In order to survive and reproduce successfully in nature,organisms must have a good Perception of self-motion information,which is called Self-motion Perception.However,self-motion perception requires the participation of various sensory factors,such as vision,vestibular and somatosensory.Therefore,in order to perceive self-motion information as accurately as possible,organisms often need to properly integrate and separate information of multiple sensory modes:when the modal sensory information from the same source is consistent or different,individuals will integrate these multi-sensory information,so that the subject can better perceive the state of self-motion;However,when sensory information is inconsistent with large differences,different sensory information will be distinguished or separated to avoid illusion or discomfort.Such integration and separation is considered as a Multisensory Causal Inference process.Through the dual-task module design,we constructed a behavioral paradigm of implicit causal inference on macaques,that is,subjects did not explicitly report whether different sensory information was homologous,but only made perceptual estimation of one of the two sensory stimulus information,and detected how this estimation was influenced by the other sensory information to lead to integration and separation.Specifically,we presented rhesus monkeys with visual-vestibular bimodal stimuli of different course angles（0°,±20°,±40°,or±60°,respectively）,and then measured the rhesus monkeys’perceived direction of vestibular heading under the bimodal stimulus.It should be noted that macaques are trained to make course decisions based on vestibular signals before they are subjected to causal reasoning experiments.Behavioral experiments found that under the Disparity of bimodal stimuli,the disparity of vestibular perception Bias was first up and then down,consistent with the typical characteristics of implicit causal inference.Moreover,these Causal Inference behaviors of macaques could be well fitted by Hierarchical Bayesian Causal Inference model（HBCI）.HBCI model believes that the final estimate of implicit causal inference is obtained by adding the integrated component and the separated component.By analyzing and calculating the changes of component values with different stimulus difference conditions,We find that the change in the overall behavior estimation deviation is mainly caused by two sub-components of the former,namely the probability of integration p(C=1∣x₀),x₄))and the estimated s?₀,C=1) under the integration condition.We also recorded the neurons in the ventral region of the medial parietal sulci（VIP）during the execution of the behavior paradigm in macaque monkeys.Among the recorded neurons,we found three types of typical neuronal responses:one type of neuron had a higher normalized firing rate when the disparity was 0°,and the normalized firing rate gradually decreased with the increase of the disparity.Another kind of neurons had the lowest normalized firing rate when the stimulus disparity was-60°,and the normalized firing rate increased gradually when the disparity reached60°.In addition,there is a class of neurons with the highest normalized firing rate when the stimulus disparity is-60°,and the normalized firing rate decreases gradually as the disparity reaches 60°.Through decoding analysis and correlation analysis of these neuronal activities,we found that there was a strong correlation between the responses of these three types of neurons and the two computational subcomponents that caused the bias change in the above behavioral HBCI model,namely,the possibility of integration and the estimate of integration.Specifically,there was a significant positive correlation between the firing rate of the first type of neurons and the probability of integration(r=0.88,p=8.3*10^-3).The second and third neuronal firing were positively correlated(r=0.98,p=1.18*10^-4)and negatively correlated(r=-0.94,p=1.5*10^-3)with the estimation of vestibular Angle under integration,respectively.We hypothesize that the first type of neurons represent the degree or probability of the integration of the two types of stimulus information,and the latter two types of neurons jointly contribute to the bias generated by the integration of stimulus information.In conclusion,we successfully constructed the visual-vestibular recessive causal reasoning paradigm in macaque monkeys,and found that VIP brain region has neurons strongly related to the two main computational components of causal reasoning behavior.These results will be helpful for further research on the complete neural mechanism of causal reasoning.