The nature of such sophisticated decision making in the cerebral cortex, which is responsible for high-level processing, has been "poorly studied and little understood," according to Wako Yoshida and Shin Ishii of the Nara Institute of Science and Technology. Now, however, in an article in the June 1, 2006, Neuron, they describe experiments that enabled them to tease apart how different regions of the cerebral cortex process uncertain information and integrate it into decision making.
In particular, their aim was to analyze subjects' navigation through a virtual maze, to explore how different cortical regions function in solving "partially observable decision-making problems."
"In navigation tasks, such as that investigated here, an individual must constantly maintain an estimate as to his/her current location as a guide for deciding the next turn," they wrote, "but in the absence of incontrovertible a priori information, this estimate is best represented by the subject's belief. As information is acquired through observation, this belief may become increasingly convincing or alternatively may be discarded in favor of a new one. This is an intuitive way of making estimations that are appropriate for many real-world behaviors, adopted also by a wide variety of intelligent machines.?" they wrote.
In their experiments, the researchers first taught volunteer subjects the layout of a computer-generated 3D "wire-frame" maze. Then, while the subjects' brains were being scanned using functional magnetic resonance, the researchers "placed" the subjects in different parts of the maze and analyzed activation of cerebral cortical regions as the subjects made a series of decisions to navigate t heir way to a specified goal. Functional MRI involves using harmless magnetic fields and radio waves to image blood flow in brain regions, which reflects activity.
Importantly, Yoshida and Ishii used sophisticated statistical probabilistic analysis of the subjects' movements to overcome a major obstacle to such studies. That obstacle is that the beliefs of the subjects during the experiment could not be determined unequivocally; thus, those beliefs could not be correlated with brain function.
However, the researchers' statistical analysis of the subject's navigation decisions enabled them to infer which of two "cognitive states" the subject was in, to give the researchers insight into which cortical regions were active during the states. One such cognitive state was a belief about where the subject was in the maze, and the other was a set of "operant" states. These operant states were a "proceed or update mode" or a "reevaluate or back-track mode."
Analyzing the brain regions active during these states, Yoshidi and Ishii pinpointed which regions of the subjects' cerebral cortex were active during the different processes involved in "changing their minds." Specifically, the researchers found that "belief maintenance" processes are performed principally by a region called the anterior prefrontal cortex, and "belief back-track" processes occur in the medial prefrontal cortex.
"Our results provide evidence that activity in different regions of the prefrontal cortex reflect critical computational components involved in decision making in uncertain environments," concluded the researchers. "This fits well with the proposed role of these regions in decision making, which is likely to be crucial in complex real-world environments. We also illustrate the utility of statistical model-based inference and regression in delineating key task parameters that may be represented in spatially distinct brain regions," they concluded.