Dedicated Representation of Others in the Macaque Frontal Cortex: From Action Monitoring and Prediction to Outcome Evaluation

Abstract

Social neurophysiology has increasingly addressed how several aspects of self and other are distinctly represented in the brain. In social interactions, the self-other distinction is fundamental for discriminating one's own actions, intentions, and outcomes from those that originate in the external world. In this paper, we review neurophysiological experiments using nonhuman primates that shed light on the importance of the self-other distinction, focusing mainly on the frontal cortex. We start by examining how the findings are impacted by the experimental paradigms that are used, such as the type of social partner or whether a passive or active interaction is required. Next, we describe the 2 sociocognitive systems: mirror and mentalizing. Finally, we discuss how the self-other distinction can occur in different domains to process different aspects of social information: the observation and prediction of others' actions and the monitoring of others' rewards.

Keywords: action observation; frontal cortex; monkey; outcome; prediction; social.

Figure 1

(A) Experimental conditions in the food-grab task used by Fujii et al. (2007). The circles indicate the possible locations where food could be placed on the table; colors in the circles represent the percentage of success for food retrieval for subjects M1 (green) and M2 (red). In position (A) (noncompetitive; left panel), the monkeys did not share any locations, for example, they could not reach food located in the spaces in front of the other monkey. In positions (B) and (C) (competitive; middle, and right panels), the monkeys shared one location where both of them could reach the food. (B) Percentage of neurons in M2 modulated by actor and action across the 3 task conditions. Nsp bars represent neurons modulated by motion without actor or action specificity. The proportion of other-left-responding neurons was higher in competitive position C than in the noncompetitive position (red bars indicate significantly different proportions between positions A and C). Modified from Fujii et al. (2007).

Figure 2

(A) Sequence of task events in the role-reversal task used by Yoshida et al. (2011). In this task, the monkeys were required to press one of 2 buttons (green or yellow) to receive a reward. Pressing one button led to reward delivery while pressing the other did not. (B) The correct response was associated with pressing a button of a specific color for blocks of a variable number of trials (5–17). Blocks could switch unpredictably, alternating between green and yellow. (C) Recording sites. (D) Example of a “partner-type” neuron, which showed a higher firing rate during actions performed by the partner than by the self. Neural activity was aligned to the button press. (E) Partner-type neuron population activity (in the PMv and the medial prefrontal cortex [MPFC]) reported in Ninomiya et al. (2020) in the animate-partner condition (RA, real agent; top panel) and the inanimate-partner condition (FM, filmed monkey; middle panel), and the difference between the conditions (bottom panel). Pink bars indicate periods during which the difference between the partner-action trials in the 2 conditions was greater than 0. Modified from Yoshida et al. (2011) and from Ninomiya et al. (2020).

Figure 3

(A) The prisoner’s dilemma task used by Haroush and Williams (2015). The monkeys chose separately whether to cooperate or defect, and then their choices were shown to both monkeys and the reward was delivered accordingly. (B) Matrix of the possible outcomes based on the choice to cooperate or defect. (C) Top row: example of a neuron encoding the monkey’s own (left) but not the other’s choice (right) to cooperate or defect. Bottom row: example of a neuron encoding the other’s (left) but not the monkey’s own choice (right) to cooperate or defect. Neural activity was aligned to the monkey’s own choice before the choice of the other agent was revealed (the period indicated in gray). Red represents cooperation trials whereas blue represents defection trials. Modified from Haroush and Williams (2015).

Figure 4

(A) Top: Sequence of task events in the spatial version of the nonmatch-to-goal task used by Falcone et al. (2017). The target stimulus represented by a gray square is presented in 2 out of 4 possible positions (top left, top right, bottom left, and bottom right). In each trial, one of the two target stimuli presented on the screen was the correct target stimulus from the previous trial, and the other was either a new one or the one not previously selected. Bottom: the monkey and the human performed the trials, switching roles. Both the human and the monkey followed the same rule: choose the target that is in a different position from that of the target chosen in the previous trial. For example, in the first human trial in the figure the human agent should not select the bottom left target, because it was chosen in the previous trial by the monkey. At the end of a trial performed by the monkey, the human agent could perform the next trial, with the monkey observing his choices and actions, or he could allow the monkey to perform another trial. (B) Example of a “Human-only” neuron exhibiting left-target selectivity only in the human trials. (C) Example of a “Both-agent” neuron exhibiting incongruent target selectivity between the monkey and human trials (right preference in monkey trials and left preference in human trials). In both (B) and (C) the neural activity is aligned to the presentation of the targets (coinciding with the beginning of the delay period). Vertical green bars represent the go signal. Modified from Falcone et al. (2017).

Figure 5

(A) Sequence of task events during the center–out reaching task used by Cisek and Kalaska (2004). Two spatial cues of different colors were placed in 2 out of 8 possible positions arranged in a circle around the center of the screen. After the disappearance of the spatial cues, a color cue of the same color as one of the two previous spatial cues was presented in the center of the screen. (B) Population activity during performance (top) and observation (bottom) conditions, aligned to the presentation of the spatial cue (S on the horizontal axis), the color cue (C on the horizontal axis), and the go signal (G on the horizontal axis). Blue traces represent trials for the preferred direction of each cell. Red and green traces represent trials in the opposite and orthogonal directions to the preferred direction, respectively. The population activity within the dorsal premotor cortex exhibited similar patterns of activation under performance and observation conditions. Modified from Cisek and Kalaska (2004).

Figure 6

Mean population firing rate for the populations of Monkey-only, Human-only, and Both-agent neurons recorded in Cirillo et al. (2018). Neural activity is aligned to the delay onset. The gray shaded areas indicate the period of analysis (0.4–0.8 s of the delay period). Error bars indicate ± the standard error of the mean. (A) Top: Mean firing rates for Monkey-only and Human neurons in monkey and human trials, respectively. The rank that identified the preferred location was assigned to each cell individually by comparing mean firing rates in right and left trials for each agent. Bottom: Mean firing rates for Monkey-only and Human-only neurons in human and monkey trials, respectively. For each group, the activity was assigned the preferred and nonpreferred locations derived from the original trials. (B) Top: Mean firing rates for Both-agent neurons in monkey and human trials. Bottom: Mean firing rates for Both-agent neurons with the rank inverted. Monkey-only and Human-only neurons did not have the same spatial tuning in trials performed by different agents, and Both-agent neurons did not show agent-specific spatial tuning, since the majority exhibited congruent spatial preference between agents. Modified from Cirillo et al. (2018).