Anterior prefrontal cortex contributes to action selection through tracking of recent reward trends

Christopher K. Kovach, Nathaniel D. Daw, David Rudrauf, Daniel Trane, John P. O'Doherty, Ralph Adolphs

Research output: Contribution to journalArticlepeer-review

81 Scopus citations


The functions of prefrontal cortex remain enigmatic, especially for its anterior sectors, putatively ranging from planning to self-initiated behavior, social cognition, task switching, and memory. A predominant current theory regarding the most anterior sector, the frontopolar cortex (FPC), is that it is involved in exploring alternative courses of action, but the detailed causal mechanisms remain unknown. Here we investigated this issue using the lesion method, together with a novel model-based analysis. Eight patients with anterior prefrontal brain lesions including the FPC performed a four-armed bandit task known from neuroimaging studies to activate the FPC. Model-based analyses of learning demonstrated a selective deficit in the ability to extrapolate the most recent trend, despite an intact general ability to learn from past rewards. Whereas both brain-damaged and healthy controls used comparisons between the two most recent choice outcomes to infer trends that influenced their decision about the next choice, the group with anterior prefrontal lesions showed a complete absence of this component and instead based their choice entirely on the cumulative reward history. Given that the FPC is thought to be the most evolutionarily recent expansion of primate prefrontal cortex, we suggest that its function may reflect uniquely human adaptations to select and update models of reward contingency in dynamic environments.

Original languageEnglish (US)
Pages (from-to)8434-8442
Number of pages9
JournalJournal of Neuroscience
Issue number25
StatePublished - Jun 20 2012
Externally publishedYes

All Science Journal Classification (ASJC) codes

  • General Neuroscience


Dive into the research topics of 'Anterior prefrontal cortex contributes to action selection through tracking of recent reward trends'. Together they form a unique fingerprint.

Cite this