Abstract Sensory processing is continuously shaped by internal bodily states, yet the neural mechanisms underlying this interoceptive-exteroceptive integration remain poorly understood. Predictive processing theories propose that bodily states modulate perceptual inference through precision-weighting (the contextual adjustment of prediction error gain according to sensory reliability), but empirical validation with neurobiologically realistic models has been lacking. We address this gap by combining cardiac phase-locked magnetoencephalography with systematic dynamic causal modelling to test competing mechanistic hypotheses about systolic-induced sensory attenuation. Using an auditory oddball paradigm, we observed selective suppression of deviant responses during cardiac systole (200-250ms post-stimulus), affecting prediction errors from unexpected tones whilst sparing expected tones. To identify the underlying synaptic mechanisms, we implemented three methodological innovations: (1) systematic comparison across 20 architectures representing precision-weighting (via intrinsic gains and/or modulatory connections), sensory gating (via forward connections), and predictive suppression (via backward connections) hypotheses; (2) Bayesian model reduction testing all 256 parameter configurations within the winning architecture to handle distributed model evidence; and (3) sensitivity analysis quantifying both direct effects and second-order interactions across the cortical hierarchy. Model comparison decisively favoured precision-weighting implementations (>99.99% posterior probability), with Bayesian model averaging revealing distributed gain control: superficial pyramidal self-inhibition in primary auditory cortex (94%) and inferior frontal gyrus (100%), inhibitory interneuron modulation in superior temporal gyrus (99%), and top-down modulatory connections from superior temporal to primary auditory cortex (96%). Critically, sensitivity analysis demonstrated that intrinsic inhibitory mechanisms exerted order-of-magnitude larger effects than hierarchical modulatory connections, with superior temporal gyrus emerging as an integration nexus showing extensive parameter interactions. These findings provide the first empirical validation of precision-weighting mechanisms during cardiac-sensory integration, establishing that systolic attenuation operates primarily through coordinated local inhibitory gain control rather than hierarchical (expected) attentional modulation. This modelling framework bridges computational theories of interoceptive-exteroceptive integration with laminar-specific cortical mechanisms, offering a generalisable methodology for testing predictive coding hypotheses about embodied perception.
Discussion(0)
No comments yet. Be the first to comment.