About 25 years ago, Crick and Koch wrote a very thought-provoking paper,

https://www.nature.com/articles/34584

It tries to answer the question, why do feedforward (FF) and feedback (FB) projections have such different properties in terms of terminating in different cortical layers and physiological impact — something that's ignored in many recurrent network models of cortex.

i.o.w.: Why do higher areas talk so differently to lower areas than lower areas talk to higher areas?