+ "Predictive coding hypotheses posit that perception is an active process whereby brain regions predict incoming sensory inputs, against which they are compared by other neural populations. Mismatches between predictions and inputs result in error signals that can then be used to update the predictive model encoded in synaptic weights, thereby driving plasticity. Although increasing amounts of evidence are consistent with the general framework, many different algorithmic implementations have been proposed, requiring further experiments to test specific corollaries of these varied approaches. One important, testable implication distinguishing some current theories involves the coupling strength in L2/3 and L5 pyramidal neurons between distal apical dendrites, which tend to receive top-down inputs that may include sensory prediction data, and their conjoined somata, which are often driven by bottom-up inputs. In particular, error signals are computed or else collocated in apical dendrites in some implementations, resulting in a quiescent subunit when the prediction matches inputs—zero error—and, thus, reduced dendro-somatic coupling during such times. In contrast, a separate proposal implies the opposite: Since many apical dendritic voltage signals can only reach their electrotonically segregated soma when facilitated by bursting induced by concurrent somatic sensory inputs, dendro-somatic coupling would instead be strongest when top-down predictions match bottom-up signals.\n",
0 commit comments