Value Computation in the Prefrontal Cortex in a Probabilistic Reasoning Task


A recent study published in PNAS demonstrated that the orbitofrontal cortex (OFC) and the dorsal lateral prefrontal cortex (DLPFC) contribute differently in value-based decision making. This work was performed by researchers in Dr. YANG Tianming’s Lab at the Institute of Neuroscience, CAS Key Laboratory of Primate Neurobiology, Center for Excellence in Brain Science and Intelligence Technology of the Chinese Academy of Sciences. This work has successfully found that during decision making based on multiple pieces of information, the OFC neurons only encode the value associated with single pieces of evidence in the stimulus domain transiently while the DLPFC neuronal activities reflected the information accumulation process leading to the chosen actions. The study further broadened our understanding of the different roles of the prefrontal subregions in value-based decisions.    

A key step in decision making is to evaluate the value associated with each option. Previous studies have found that the prefrontal area, especially the OFC, plays an important role in value-based decision making. However, it is unclear how different prefrontal subregions work together to compute value during decision making. In addition, neuroscientists hold different opinions on whether there is an intermediate stage in the brain where value computation is independent from any motor contingencies.   

In order to address these questions, the researchers trained two monkeys to perform a sophisticated decision-making task. The monkeys needed to accumulate evidence associated with four sequentially presented shapes displayed on a computer screen and use eye movements to choose between a red and a green target also displayed on the screen. Each shape provided a piece of evidence on which colored target would lead to a juice reward. The monkeys had to accumulate the information from these four shapes to find out which target was more likely to be rewarded and choose it accordingly. Furthermore, the colored targets were randomly located on the left or right side of the screen. (Fig. A). Thus, the value information associated with colors had to be transformed to be spatial coordinates for the motor circuitry to carry out the eventual choice actions. The key question was how value information was accumulated in the prefrontal cortex and whether there was an accumulation stage independent from the actions or not in the prefrontal cortex (Fig. B). 

The authors recorded single-unit activities from the OFC and DLPFC areas in the two monkeys. Using various regression models, they demonstrated that the neurons in the two areas encoded value information differently and played distinct roles during decision making. In particular, the OFC neurons encoded the information provided by individual pictures for computing the red or green target’s value, regardless of where they were. On the other hand, the DLPFC neurons not only encoded the value information associated with individual pictures but also combined information from multiple pictures. The accumulated evidence represented in the DLPFC indicated which choice action, regardless whether it was for choosing the red or the green target, was more valuable (Fig. C). The transformation of evidence from the color domain into the action domain occurred in the DLPFC and before the evidence accumulation. There was not an intermediate stage where the evidence was accumulated independently from the actions in either the OFC or the DLFPC. 

The study provided solid evidence against a popular view in the field that the OFC computes action-independent value during value-based decision making. The findings of the distinct response patterns of the OFC and the DLPFC indicate how they may work together to compute the value in decision makings.  

This work entitled “Evidence Accumulation for Value Computation in the Prefrontal Cortex During Decision Making” was published online in the Proceedings of the National Academy of Sciences, USA, on November 17, 2020. LIN Zhongqiao, NIE Chechang , and ZHANG Yuanfeng are the first authors with equal contributions. This work was supported by CAS, STCSM and NSFC. 


Figure legend: (A) Behavioral paradigm. (B) Schematic hypotheses. The computation of value could be computed in the color domain (good-based) or the action domain (action-based). The transformation of the value from the color domain to the action domain could occur at the stage of single evidence or the summed evidence. (C) Representations of accumulated value information in the action domain in the DLPFC. (Image by CEBSIT)


YANG Tianming 

Institute of Neuroscience, CAS Key Laboratory of Primate Neurobiology, Center for Excellence in Brain Science and Intelligence Technology of the Chinese Academy of Sciences