Wednesday, December 5, 2007

Modularity of perception

Well over 100 years ago, scientists realized that damage to specific brain regions resulted in specific behavioral deficits. For example, damage to left frontal cortex is often associated with impaired language abilities. Observations of this kind led to the hypothesis that each chunk of cortex performs a specific task (often referred to as the ‘modularity’ hypothesis). In contrast, others suggested that different regions of the cortex are not specialized at all – that all regions participate in all aspects of cognition. Time has taught us that both of these extreme views are probably incorrect. We would quickly run out of space in our head if we dedicated a chunk of cortex to each task that we needed to perform. On the other hand, given the knowledge that damage to certain brain regions leads to very specific behavioral deficits, we must acknowledge that some specialization occurs.

How do we reconcile these two points of view? Recently, we investigated the issue of modularity using functional magnetic resonance imaging (or ‘fMRI’), a method that allows us to indirectly measure neural activity in humans (see article linked below). We focused on visual information processing, since we know quite a bit about the parts of the brain that are responsible for sight. Light comes into the eye, where it is converted into a series of electrical impulses by the retina (a process called ‘transduction’). These electrical impulses are then passed from neuron to neuron until they reach a region of cortex at the very back of the head that is referred to as V1. In V1, neurons respond to simple features in the environment, such as the orientation of edges and different colors. Neurons in V1 then pass along information to other visual areas for further processing. There are actually more than 30 visual areas that are involved in the process of analyzing visual inputs, and each one seems to contribute some unique bit of information to support perception. For example, area V4 – which is a few steps up the hierarchy from V1 – registers information about simple shapes, area MT registers the direction of moving objects, and some later regions register the identity of objects (such as faces).

On the surface, this functional specialization seems to support the ‘modular’ account of brain organization; however, no single visual area can support perception without working in concert with other areas. To give an extreme example, suppose the visual system has a module that only processes information about color. Now, suppose someone suffered damage to their eyes and could no longer transduce light coming into their retina. Obviously, this person wouldn’t be able to perceive colors, even though the color module was perfectly intact. Thus, functionally specialized brain regions cannot operate in isolation; some regions convert light into neural activity, some supply information about edge orientations, some about color, some about motion, and so on. Eventually this information is combined to create a coherent perceptual representation of the surrounding environment.

Even though no single area in isolation can give rise to perception, all areas are clearly not created equal. For example, in our study we examined brain activity in area MT while people watched videos of moving objects. Obviously, the ability of MT neurons to respond to motion depends on input provided by the eyes and by earlier visual areas. However, our experiment found that decisions about the perceived direction of motion are based primarily on the activity of MT neurons, even though activity in other areas is necessary to achieve the final overall percept. According to this account, modularity arises primarily when we need to make a judgment about some attribute of our environment. If we need to know about motion, we query the activity of neurons in MT, if we need to know about color, we might query the activity of neurons in V4, and so on. This viewpoint suggests that most cognitive operations rely on neural activity in a series of distinct cortical areas; however, the ultimate output of the process may be largely mediated by a single specialized area of the brain. One important future challenge will be to determine how more complex cognitive operations (beyond judging the direction of a moving object) are carried out and represented in cortex, and if the same organizational principles apply.


No comments: