To what extent do brains (e.g. of humans) contain recurrent connections?

I am studying artificial neuronal networks and frequently encountered the statement, that recurrent neural networks are closer to biological neuronal networks than for example feed forward networks. But I didn't find information on the question if feed forward or recurrent architectures dominate the brain architecture.


The recurrent patterns of connections in a network are known as network motifs.

You can check this paper out. They have identified common network motifs in different types of real networks including neural networks (of C.elegans though).

Apart from feed forward loops, bi-fan is also a common network motif in neural networks.

Perhaps the connectome project has data for human brain which you can analyze for identifying network motifs. You would have to request for the data, though.


Ok, let's talk about mammalian neocortex rather than about the entire central nervous system.

The vast majority of synapses within the cortex are formed between neurons within the same cortical area (Binzegger et al 2004). Although most of these synapses will not be self-connections (from a single neuron back to itself), they are recurrent in the sense that there is a high degree of interconnectivity within a cortical area. In visual cortex, connections within the superficial layers of cortex (layer 2/3 <-> layer 2/3) make up around 70% of all excitatory synapses in visual cortex!

In contrast, the "feed-forward" pathway into visual cortex makes up less than 1% of the excitatory synapses in visual cortex (Costa & Martin 2011).

If you change your scope to look only at inter-cortical-regional projections, then there is comparatively less recurrence (i.e. strong bi-directional connections between two cortical areas), although it still exists. Connections between cortical areas seem to be correlated with physical distance across the cortex (Vezoli et al. 2004). Connections between adjacent areas (V1 and V2, for example) are strong in both directions, and appear to project to similar cortical laminae in both directions. In contrast, projections over longer distances are comparatively weaker, and target different cortical laminae in either direction.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.