Developing Intelligence

King of the Cortex: Anterior PFC

As enigmatic as prefrontal function seems to be, the anterior portions of prefrontal cortex (aPFC) are even more mysterious. This results partly from the fact that aPFC is particularly difficult to access and study electrophysiologically in nonhuman primates, as Ramnani and Owen note in their 2004 Nature Reviews Neuroscience article, and so detailed neuroanatomical investigations of aPFC have been conducted only recently. The authors report how this work has led to a breakthrough in the understanding of aPFC’s computations.

Ramnani and Owen review Brodmann’s early analysis of aPFC, also known as Brodmann’s area 10 (BA10). This region occupies most of the frontal pole and shows distinct cytoarchitecture from surrounding regions. Unfortunately there are several problems with this straightforward association of aPFC with BA10. First, this association makes BA10 an impractically large region, so some have argued for the distinction between subregions of BA10. Second, in monkeys, this area is not occupied by BA10, but by BA12. Considering these difficulties, Ramnani & Owen restrict their analysis of aPFC to BA10p, the area of aPFC that lies directly on the frontal pole and includes the most anterior regions of the frontal gyri.

The dendrites of PFC neurons have more spines that neurons found elsewhere in cortex, suggesting that PFC is particularly involved in “integrating” multiple inputs. This interesting fact is even more true of aPFC, which Ramnani & Owen claim has a higher spine density than even other regions of PFC despite having a lower overall number of neurons. Unlike all other PFC areas, which are interconnected with low-level association areas, aPFC is reciprocally interconnected only with an “executive network” of purely multi-modal regions. Tellingly, this includes anterior temporal cortex (thought to be involved in object identity processing), cingulate cortex (thought to be involved in conflict monitoring/detection), and other regions of PFC (broadly involved in the active maintenance of information).

Based on this knowledge of aPFC neuroanatomy Ramnani and Owen analyze several theories of aPFC function, reviewed in turn below.

Internal State Model

This theory suggests that aPFC is important for introspection, which often results in spontaneous, stimulus-indepepdent neural activity. This theory can account for the greater activation of aPFC during baseline fMRI tasks (such as when subjects are simply asked to stare at a screen) relative to tasks that seem to involve more cognitive demands. Likewise, this theory accounts for higher aPFC activity during tasks involving episodic memory relative to recognition only. Ramnani & Owen conclude that while this model may accurately describe some functions of aPFC, other PFC regions are also likely involved in similar processes.

Memory Retrieval Models

The authors also review “memory retrieval” models of aPFC function, which suggest that aPFC generates (“gates”) or evaluates memory search strategies, such as those that are particularly active for source memory (memory for where or under what circumstances a particular item was experienced). Ramnani & Owen conclude that this theory is useful for explaining aPFC activation in some tasks, but is probably too specific to be a good model for the general functions of aPFC.

Prospective Memory Model

Others have suggested that aPFC is important for prospective memory, which is “memory for the future” (i.e., I need to pick up milk on the way home from work today.) Note the similarities between this account and the temporal cascade model of Koechlin et al. described yesterday, in which increasingly anterior regions maintain information for increasingly long periods of time. Unfortunately, Ramnani & Owen conclude that this theory also seems too specific, since tasks that do not clearly utilize prospective memory also activate aPFC.

Cognitive Branching Model

Yet others have suggested that aPFC is important for maintenance of meta-level goals while subgoals are being specified and manipulated. Ramnani & Owen note that this model is very vague, in that every cognitive task can be thought to involve almost any number of subgoals.

Relational Integration Model

This theory holds that aPFC is important for considering how two distinct things relate to one another; Ramnani & Owen review some work that has been successful in distinguishing this from general mental effort. The authors suggest this theory is “off the mark” in that it is both too specific (it cannot explain activation in other tasks that do not seem to involve relational integration) and too broad (dlPFC seems to show a similar response profile).

“Operation Coordination” Model

Finally, Ramnani & Owen propose their own model of aPFC function, which is essentially a reinterpretation of each of the above models according to information processing demand rather than task-specific demands. This “operation coordination” model proposes that aPFC is utilized when a goal requires the use of two or more distinct cognitive operations in order to monitor and integrate their outcomes.

Conclusions

Ramnani & Owen’s “operation coordination” model of aPFC function provided number of predictions that have subsequently been supported in areas as diverse as exploratory behavior, brain-computer interfaces, and traditional n-back working memory tasks. aPFC may also contribute to the “mixing cost” often observed in task-switching paradigms.

Most importantly, this view of aPFC function falls neatly in line with an emerging view about the functional architecture of PFC. PFC can be viewed as a cascading hierarchy of supramodal processors, each recursively connected with lower regions and biasing their representations. aPFC appears to sit at the top of this hierarchy.

As with all theories of PFC function, however, this one can seem a little too fuzzy. For example, how does one distinguish between “cognitve operations”? Is aPFC activity always increased when a task involves 2 relative to 1 cognitive processes (however we might define them), or does PFC perform more dynamic load balancing? Finally, how might non-human primates manifest mixing costs or other putative effects of “operation coordination,” given that their frontal pole contains BA12 rather than BA10?

Related Posts:
Neural Cascades in Prefrontal Cortex
The Anterior Frontier: Prefrontal Cortex