Sunday, November 21, 2010

The Binding Problem


The binding problem is an issue in theoretical neuroscience that arises because of the lack of clarity about how two features of the same object are bound together in the brain leading to a single, coherent perception. Most neuroscientists in this area assume that individual features of sensory stimuli correspond to tiny assemblies of cells in the brain. These neural assemblies specialize in representing specific features and will become active if you perceive this feature, or even if you simply imagine it. For example, if one sees a blue, wrinkled glove, neural assemblies that correspond to blue, wrinkled and glove would all be activated simultaneously. The coactivation of these assemblies, and all of the other assemblies responsible for other features of the glove, would create a unified picture of a single object. The question is though, how do these assemblies take discrete, individual features and combine them to make one perception?

It is thought that when assemblies that correspond to the same perception are coactivated they send electric messages back and forth. They also recruit other areas that get involved in these informational oscillations. When these assemblies oscillate in synchrony they are thought to bind features such as contour, shape, motion, color, depth and other aspects into a composite image. These electrical oscillations alone cannot adequately explain how conscious perception arises in the brain but they give clues about how we perceive multiple objects. 

Another very similar problem in feature binding is the superposition catastrophe. This is where you have multiple objects, each with their own features, and you are able to keep track of the features of each object. Imagine that you are looking at a green raccoon and a yellow hippo. If all of the assemblies that correspond to these animals and colors must fire simultaneously, then how can you discern that the raccoon is green and the hippo is yellow? How can you tell that the raccoon is not yellow? One way of solving this problem is to assume that we have dedicated assemblies for green raccoons and yellow hippos and that these can become activated at the same time without confusing their contents. This is an unreasonable hypothesis though because most of us have never seen or imagined these animals. If we had dedicated assemblies for every combination of features in our memory the total number of necessary assemblies would be astronomical, and would require far more neurons than we can hold in our heads. Neuroscientists are starting to think that perhaps the assemblies for green and raccoon are activated simultaneously, the assemblies yellow and hippo as well, but that the two pairs are activated out of sync with each other. In other words, the assemblies for green, those for raccoon, and all of the other assemblies and networks associated with these, but not with yellow hippos, fire neural messages back and forth with each other at their own tempo that is distinct from the timing of other activities that are going on in the brain. Such amalgams of associated assemblies are often called modules in the literature. The ability of the brain to compartmentalize features into modules allows the segregation of information which in turn allows the perception of multiple objects, each with their own features. 

Modules of neural assemblies are thought to contribute to conscious perceptions and are thought to interact with other modules to create behavior. If they could be visualized these modules would not appear as neat, uniform structures in the brain, but would probably be messy, wiry structures that criss-cross between a variety of brain areas. It is not as easy to visualize the neural assemblies. Some may involve neurons that are clustered very close together yet others may be more distributed. It is thought that assemblies have to fire back and forth very fast (around 40 Hz or 40 times per second) to bind together to form a module. Experiments have shown that when someone is experiencing an optical illusion, that illusorily associated features bind together. Interestingly, it has also been shown that when someone cannot perceive a hidden image, that the features necessary to see the image are either not activated or not bound. In other words, in order to notice something you have to bind its features together and when we are fooled we bind the wrong features.

A Partial Solution to the Binding Problem:

I don't believe that there is a superposition catastrophe. To posit one you must assume that in higher brain centers, the physical location and orientation of the two, differently colored objects are lost. I think that the higher-order associative areas may not code this information but that the primary and secondary visual areas (which the association areas continually interact with) do because they are organized retinotopically and would be able to keep the objects, and their respective features (such as color), consistent without mixing things up. This is akin to having a TV in your head, as long as you are watching the TV, you cannot forget the colors of different objects. Certainly, this occipital "TV" is erased every 250 ms due to the fact that its outputs amount to transient "sensory memory" but if the higher association areas bind to the right elements, they can keep them active on this TV for as long as they are either visually rehearsed or committed to memory.



Read the full article that I wrote on this topic here:

http://www.sciencedirect.com/science/article/pii/S0031938416308289

No comments:

Post a Comment