Our brains are essentially massively parallel processing machines.  Even the simple activity of gazing out at the ocean in total bliss requires the coordination of millions of perceptual processes.  When it comes to large-scale goal directed attention or action, however, we struggle to do more than a single thing at once.  A paper published last month in Neuron looked into the brain activity associated with multitasking and attempted to understand why.

A research group at Vanderbilt led by Paul Dux studied the changes that occur when people learn to perform two different tasks–a visual-manual task and an auditory-vocal task–at the same time.  fMRI brain scanning revealed that no areas of the brain respond only when the two activities are undertaken together. In other words, there is no part of the brain explicitly devoted to handling multitasking.  The researchers did find, however, that many regions of the brain were involved in these tasks but that only one, the left inferior frontal junction (IFJ), was more active when they were performed together.

The subjects in this experiment initially found it very difficult to multitask, but they got better with training.  The researchers thus looked at what was happening in the IFJ when these improvements were made to understanding how multitasking works in neural tissue.  They considered three separate hypotheses, each of which made different claims about the neural response to multitasking training.

In the first story, we get better at multitasking because the processing moves away from the slow abstraction of the prefrontal cortex to direct inflexible circuits linking sensory and motor areas.  To test this theory, the researchers looked at the effective connectivity between the regions involved in this task.  Even with training, however, there was no strengthening in the direct circuits between perception and response.  The information was still passing through the IFJ, it was just doing so more quickly.

A second hypothesis, then, was that dedicated circuits formed within the IFJ to segment and accelerate multitasking processes.  Dux and his colleagues performed a pattern classification analysis to evaluate this theory.  Pattern classification works by teaching a computer algorithm to discriminate between the brain response associated with different activities.  According the this second theory, classification performance should improve if the IFJ develops dedicated pipelines to handle the individual requirements of the multitasking procedure.

In fact, however, classification performance slightly decreased with training, indicating that the second hypothesis was also false.  This result does suggest, though, that there were some sort of changes within the IFJ, so the researchers turned to the third hypothesis.  They scanned several additional subjects using high temporal resolution fMRI focused just within the IFJ both before and after multitasking training.

With this new fine-grained data, the researchers were able to reach the conclusion that training leads to gains of efficiency in the central processing module within the IFJ.  The degree of improvement in reaction time corresponded to the acceleration in IFJ processing as revealed by fMRI. This shows that, even though the brain is massively parallel, complicated behaviors must  pass through this bottleneck before they can be executed.