The brain is often considered a parallel processor, which is why many cite that it is superior to a silicon chip. However, as research in neuroscience continues, a progressively better understanding of the brain reveals increasingly more commonalities between the seemingly different two. For example, new experiments performed by Ruthruff, Pashler, and Hazeltine suggest that the brain fundamentally operates serially.
Clearly, at the macroscopic level, brains must operate in parallel. Otherwise, the bodies they inhabit would be unable to perform even the simplest task. Since neurons have a maximum data transfer rate of approximately ten spikes per second, serial neural processing would require at least several seconds to complete a single operation. Because such a long delay would be disastrous, computations must occur in parallel. Indeed, there is direct evidence of this: rods stuck into a monkey brain detect multiple simultaneous electrical impulses. Thus, a better question to pose is "does the brain operate in parallel at all levels?"
Current research actually suggests that the brain operates in serial for some tasks. Previously, neurologists believed that participants carrying out dual tasks were "limited by a central bottleneck." That is, only a single central operation, such as response selection, memory retrieval/consolidation/rotation, or lexical processing, can actively run (Ruthruff, et al 1). Fortunately, new research confirms that the brain executes dual tasks serially. Hence, only one task can run at a time.
By having the subjects perform a combination of single and dual tasks, Ruthruff, Pashler, and Hazeltine virtually eliminate response grouping, thus making the data better fit either the bottleneck theory (serial execution) or the overlapping central processing theory (parallel execution). Previous experiments using this mixed presentation of single and dual tasks indicate that participants, after several practice sessions, sometimes learned to perform dual tasks without interference. Additionally, this design is believed to be "conducive to the overlap of central processing" (Ruthruff, et al 6). Therefore, if parallel processing occurs at all, this experiment would emphasize it.
However, results actually confirm the opposite: the brain operates serially during response selection. Subjects devoted nearly all of their resources to completing the central operation of their first task, before moving on to the second. Although there are other studies concluding that central bottlenecking can be avoided, parallel processing of central operations occurs only after devoting much practice towards that task. In such a case, automatic response activation proceeds through an alternative, specialized processing route, thus avoiding the central bottleneck (Ruthruff, et al 8). Without this practice, though, response selection executes serially.
The dual parallel and serial nature of the brain mirrors the multitasking behavior of operating systems. Modern computer systems operate in parallel at the macroscopic level. Fundamentally, however, the operating system is only switching very quickly between all tasks, thus making everything appear to run in parallel. Similarly, the brain's gross behavior appears to be parallel. However, most computations actually run serially. This serializing within a processor and brain occurs because both are limited to performing only one computation at a time.
Despite this, the computer and brain both essentially run in parallel. Assuming more experiments confirm this, the holy grail of computer science slowly creeps within reach. Of course, this ultimate goal of which I am referring is artificial intelligence. When we finally understand how to duplicate human thought in a machine, a whole new revolution will occur.
Ruthruff, Eric, Pashler, Harold E., and Hazeltine, Eliot. Dual-Task Interference with Equal Task Emphasis:
Graded Capacity-Sharing or Central Postponement?. [online] Available at
http://www.cs.rochester.edu/users/faculty/dana/Ruthruff_r.pdf