The Input Canonicalization Theory of Layer IV Development: A Unifying Principle for Cortical Type Variation and Motor...

Layer IV (L4) is the primary target of thalamic input to the neocortex. Its development varies from the thick L4 of primary sensory cortex to the thin L4 of primary motor cortex, which Nissl staining does not resolve. Why the gradient takes the values it does has remained unexplained. I propose the Input Canonicalization (IC) theory: L4 development is set by the demand for IC - the computation converting inputs into the canonical form downstream intracortical processing presupposes. IC demand is the cross-condition variability of the dominant input, and falls as Canonicalization Fixation proceeds: input statistics stabilize through development and learning. The theory asks what sets the gradient of laminar differentiation that the Structural Model takes as given. With the LPP Conservation Assumption - the laminar architecture of corticocortical projections is preserved between cortical types - I derive the motor areas' projection patterns (aFF/aFB). For termination layers it agrees with the Structural Model and macaque data; for origin layers it accounts for the loss of feedforward-feedback asymmetry in those data, and reverses the hierarchical assignment drawn from them. The postnatal course of L4 in area 4, and its persistence in cerebral palsy, fit the prediction. Active Inference has been offered for reduced L4 in motor cortex; IC theory is broader, and explains features that accounts of cortical type leave open, among them the three separate islands of most differentiated cortex in adult primates. IC demand can be estimated independently of L4, and I state the theory's failure conditions.