Research Article: Differential Encoding of Factors Influencing Predicted Reward Value in Monkey Rostral Anterior Cingulate Cortex

Date Published: January 18, 2012

Publisher: Public Library of Science

Author(s): Koji Toda, Yasuko Sugase-Miyamoto, Takashi Mizuhiki, Kiyonori Inaba, Barry J. Richmond, Munetaka Shidara, Joy J. Geng.


The value of a predicted reward can be estimated based on the conjunction of both the intrinsic reward value and the length of time to obtain it. The question we addressed is how the two aspects, reward size and proximity to reward, influence the responses of neurons in rostral anterior cingulate cortex (rACC), a brain region thought to play an important role in reward processing.

We recorded from single neurons while two monkeys performed a multi-trial reward schedule task. The monkeys performed 1–4 sequential color discrimination trials to obtain a reward of 1–3 liquid drops. There were two task conditions, a valid cue condition, where the number of trials and reward amount were associated with visual cues, and a random cue condition, where the cue was picked from the cue set at random. In the valid cue condition, the neuronal firing is strongly modulated by the predicted reward proximity during the trials. Information about the predicted reward amount is almost absent at those times. In substantial subpopulations, the neuronal responses decreased or increased gradually through schedule progress to the predicted outcome. These two gradually modulating signals could be used to calculate the effect of time on the perception of reward value. In the random cue condition, little information about the reward proximity or reward amount is encoded during the course of the trial before reward delivery, but when the reward is actually delivered the responses reflect both the reward proximity and reward amount.

Our results suggest that the rACC neurons encode information about reward proximity and amount in a manner that is dependent on utility of reward information. The manner in which the information is represented could be used in the moment-to-moment calculation of the effect of time and amount on predicted outcome value.

Partial Text

A great deal of evidence suggests that the anterior cingulate cortex (ACC) relates reward to motivation, cognition, and action [1]–[3]. Anatomical experiments show that there are dense connections between the ACC and reward-related brain areas, such as midbrain dopamine neurons [4]–[6] and limbic regions [7], [8], whose neurons respond to value of the reward [9]–[14]. Neurons in the ACC are known to respond to reward and error events [15], [16], reward prediction errors [17], reward expectancy [18], [19], reinforcement learning [20], reward-based action selection [21], decision making [22]–[29], and fictive reward learning [30], [31]. All of these suggest that the ACC has a role in processing information about reward value.

We simultaneously manipulated reward proximity and reward amount, two factors that affect how a reward is perceived, that is, the subjective outcome, to examine how rACC neurons encode these two factors. Over 90% of the recorded neurons showed some selective activity in the reward-schedule-amount task. These rACC neurons show different responses about rewards in relation to task context, i.e., valid and random cue conditions. When the cue is providing information about reward proximity and reward amount in the valid cue condition, information about predicted reward proximity is represented strongly whereas information about the predicted reward amount is essentially absent at those times. When the cue does not provide information in the random cue condition, as expected, little information about reward proximity or reward amount was encoded during the course of the trial before reward delivery. However, when the information becomes available, that is, when the reward is actually delivered, the number of neurons and the size of the signal for reward proximity and amount become substantial. Thus, the context on reward information determines what kind of and when information is represented in the neuronal firing within the rACC.