Thoughts on multiple pathways for sensory input into the neocortex


I was watching one of Numenta’s videos and heard Jeff mention having sleepless nights over why there are multiple pathways for a single sense leading to the neocortex (one through the thalamus and one directly to the neocortex). I’d like to present a hypothesis that could explain this. First, though, I’d like to say that I just recently started teaching myself about the interregional connections of the brain and the neuronal structure of different areas of the brain so more detailed answers / feedback would be greatly appreciated.

My hypothesis is that both of these pathways are required for sensory attention to manifest. For example, we need to determine what a sound is in order to determine whether or not it requires our attentions. In order to do this, the first pathway that circumvents the thalamus is used to feed sensory input directly into the neocortex. The neocortex determines the source of the sound and potentially (this next part might be determined by converging inputs in the thalamus instead of by the neocortex) whether or not it requires our attention. Once the neocortex has identified the source of the sound, it feeds that information into the thalamus. The thalamus then uses the inhibitory pathway of the thalamic reticular nucleus to prevent that information from continuing up the cortical hierarchy (to parts of the brain of which we are more consciously aware). This allows the thalamus to modulate our attention to sensory input while still allowing the neocortex to do the heavy lifting on the parts at which it’s good.

I think this hypothesis also requires the assumption that conscious attention derives from the level at which you look in the cortical hierarchy and I’m not sure about the efficacy of that assumption.


That’s an interesting hypothesis. There is a long history of literature suggesting the thalamus is involved in attention. As you suggest, others have speculated that the relay cells in the thalamus act like gates, enabling and disabling the passage of information.

Our current working hypothesis as to why there are two pathways is that one pathway sends relatively unprocessed features up the hierarchy and the other passes representation of more complex learned objects. If it can, a region will substitute a recognized complex object but if it can’t recognize the data then it will just pass the basic features. Most of the time you don’t notice the detailed features of objects you recognize. But, via top down feedback, you can pass up the details via attentional gating in the thalamus.


It seems to be a thing for connectome/cortex researchers to ignore the processing that goes on in the “lower” brain structures. These were/are sufficient to run lizards and amphibians. There is absolutely no reason to think that they re not performing much of their original functions inside of the functioning brain.

The lower brain receives a low-resolution image from the eye area “away” from the fovea and directs the eye to point at interesting areas to form identification of an object with a version of the old 20 questions game in the cortex. There the fovea image is constantly updated as the eyes dart around under the control of this primitive and unconscious scanning system. The cortex is presented with these views in a highly stereotyped fashion and parse out this spaciotemporal stream.

The executive control of attention and identification of primitives for emotional response is a well-preserved function in the lower structures. The hippocampus and related structures have a graduated spacial structure that performs a very effective scale invariant coding scheme. Place, direction, borders, and head position are widely touted but the hippocampus codes all kinds of maps and structures. There is ample research to say that all episodic memory is coded with this system. (patient HM for example) This is passed on to the amygdala for evaluation for recognition of what we chose to call emotional content. I might favor archetypical primitives and selection of basic programs with the resulting priming of chemical messengers. [1] This same coding is passed directly to the temporal lobe much the same way that the other more recognized sensory streams are passed to various locations around the occipital lobe.

The raw body sensors (hunger & thirst are prime examples) are also projected to the edges of a lobe - in this case - the frontal lobe. These are elaborated into motor planning including directed attention to bridge the gap between needs and the perceived and remembered environment.

My working model is a dumb boss/smart advisor with the cortex performing a “sharpening” of the perception and connections of various processing regions of the cortex that we describe as “attention and consciousness.” The lower brain is still the boss and incidentally - the source of the whole “unconsciousness” animal nature mystique.

Another way of looking at all this is that the lower brain structures “drags” the cortex along as it does what a lizard brain might do and the cortex responds by projecting back shaping influences to improve this behavior.

[1] Amygdala Response to Emotional Stimuli without Awareness: Facts and Interpretations - Matteo Diano1,2, Alessia Celeghin1,2*, Arianna Bagnis2 and Marco Tamietto1,2,3

What motivates actions?
What are the parts of the lizard brain?
Old versus new neural network
How HTM is supposed to deal with spatial invariance?
Deep Predictive Learning: A Comprehensive Model of Three Visual Streams

This is a great thread. I think all three views (hypothesis) hold some fundamental truth. They are not incompatible concepts, but rather different aspects of the same fundamental mechanism. And, I like Bitking’s evolutionary perspective of the “big picture”. Hope more scientific evidence will help to verify and support this hypothesis.