It is known that the striatum plays a role in value-based decision making, but little is known about how the various elements of the striatum contribute to this cognitive ability. The presentation of choices in an environment which changes quickly or demands novel decision requirements may require a different neural substrate for decision making compared to a familiar environment or known sets of environmental cues. Wunderlich et al have now performed an fMRI study of striatal activity during a type of minimax decision task. In this context, minimax decisions involve the participant planning ahead and assessing all possible choices in the task in order to maximize the minimum possible gain. The authors found that forward planning (involving novel decisions) is represented in the anterior caudate nucleus, while decisions needed in extensively trained tasks are represented in the putamen. The ventromedial prefrontal cortex appears to be functionally linked to both of these striatal areas during decision making, and may act as a value comparator when a choice is made involving extensively trained and novel tasks.
The decision tasks were structured into 4 different conditions, including: pure planning trials (P trials) in which subjects navigated through a tree-shaped maze of 3 levels ending with trial-unique labeled reward rooms; extensively trained trials (E trials) in which subjects navigated tree shaped mazes with fixed color-coded reward rooms; mixed trials in which the tree structure consisted of both planned and extensively trained branches (PE trials); and trials in which the tree structure offered a choice between extensively trained branches (EE trials). The pure planning trial trees consisted of 3 layers (8 terminal rooms) in which the first and last transition choices were made by the participant, but the second choice was made by a computer agent in a known deterministic manner to minimize the reward value. This forced the participants to use a tree search strategy for planning in an attempt to maximally engage the patient in forward planning for each branch of the tree. The extensively trained trials only involved planning trees of 2 layers, with fixed color-coded reward rooms which the subjects learned over the course of 3 prior days of training trials.
Twenty subjects participated in the complete study, which included concurrent fMRI imaging with intermixed E and P trials. The fMRI imaging demonstrated distinct neuro-anatomic correlates for the E and P trials. P trials were associated with increased blood oxygen level dependent (BOLD) responses in the medial caudate, thalamus, bilateral anterior insula, dorsomedial prefrontal cortex, bilateral medial frontal gyrus, and bilateral precuneus. E trials on the other hand were associated with increased BOLD responses in the lateral posterior putamen, and the posterior insula extending to the medial temporal gyrus and somatosensory cortex. The subjects were also given trials involving trees with E branches and uncorrelated P branches. In this situation, BOLD signal in the caudate was consistently correlated with the difference in value taken between the eventual planned target, and an alternative option in the planning tree branch, demonstrating the function of the caudate in performing value difference computations for planning. The imaging studies also demonstrated strong functional coupling between ventro-medial prefrontal cortex (vmPFC) and both the caudate and putamen during specifically mixed trials. A modeling study used by the authors supports the hypothesis that the vmPFC resolves the competition between the caudate and putamen-supplied choice values.
In summary, Wunderlich et al have examined 2 different types of choice trials, 1 involving on the fly or impromptu forward planning and 1 in which the subject had been extensively trained. These 2 trials utilized distinct neuroanatomic substrates as demonstrated by fMRI (the caudate nucleus was active during trials with significant planning, while the putamen was active for extensively trained trials). Additionally, the authors also provide evidence for the role of the ventro-medial prefrontal cortex acting downstream from these subcortical elements, acting as a value comparator or output generator for choice given the supplied caudate and putamen input signals. This study is salient for neurosurgeons given recent attempts at treating a new and evolving set of cognitive problems involving the basal ganglia and subcortical structures such as major depression, obsessive compulsive disorder, and even memory loss. A better understanding of their dynamic functions and interactions will allow us to improve our efforts at identifying neuromodulation targets and the effects of therapeutic or diagnostic stimulation.