How do human brains represent things?


The question I have is how do brains build internal representations:

  1. go from general to specific
  2. specific to general

Using dog as an example
dog to legs/tail/head/trunk to tail has tip, legs come in front and back and have orientation and paws on end, head has many features, this can go on tongue is long and wet and warm,…


tail tip/paws/nose leads us to tail/leg/head leads us to dog

I am not asking about the sensor system rather the internal mental representation. I note that little kids start drawing people as stick figures with simple legs and arms and head. Sort of the opposite approach used by sensor processing. As the child grows/learns and their representation grows it adds finer and finer features.

How do human brains represent things?

That is an open question at the core of Numenta’s theory research. We will release a paper when we think we have a clear theory. You’ll certainly want to read this if you have not already:


Funny you should ask that.
To answer your immediate question: tail and other furry bits build up to the big shaggy dog, with all the doggy bits coexisting and stacking up as splotches of activations across a grid pattern. The doggy does not stand alone, you sort him out of the overall scene so there are other patterns stacking up in the same grid representation all at the same time. [1][2][3]

This doggy grid pattern in the association area is the key to recall the associated doggy related info from other parts of the brain. I don’t expect to find any direct connections to this information at the lower-sensory areas.

Note that a doggy in the other parts of the brain primes recognition in the sensory areas, presumably through forming the same general grid patterns. [4]

See also, the Feb 2 hackers hang out.

[1] Understanding mid-level representations in visual processing.

[2] Probing intermediate stages of shape processing

[3] Processing context: Asymmetric interference of visual form and texture
in object and scene interactions

[4] Feedforward and Feedback Processes in Vision