Outcome—behavioral link

To evaluate whether neural response to outcome affects future behavior, we analyzed each of the four outcome conditions separately. For each condition, we focused on neurons with a significant increase in FR following its outcome. For these neurons we divided outcome trials into trials with a neural response (neural firing between 200 and 800 ms following outcome) and trials without a neural response. Neurons with a high FR which resulted in less than one trial in which the neuron did not fire and followed by an approach choice or less than one trial in which the neuron did not fire and followed by an avoidance choice were omitted from this analysis. Next, for each trial, we evaluated whether the subsequent coin trial resulted in approach or avoidance behavior. Thus we could compare, for each condition, the effect of neural response following outcome on subsequent behaviors. Only outcome trials with a subsequent HGC coin trial (with more than one ball on the way to the coin) were included as LGC trials almost always resulted in approach behavior (above 94%).

To evaluate the complex interaction of neural firing, behavior- and paradigm-related variables, we performed six GLMM (binomial) with behavior in subsequent HGC trials as the dependent variable. GLMM test 1–2: for each HGC trial, we evaluated the previous punishment outcome by calculating the following variables: (1) a binary index indicating whether a temporal/frontal neuron fired in the time range 200–800 ms following punishment outcome (as before, only neurons that significantly increased FR following punishment outcome were evaluated); (2) normalized total movement ±1 s of outcome time; (3) time delay between the punishment outcome and subsequent HGC trial. The analysis was done separately for temporal (GLMM test 1: 12 neurons, 953 total trials) and frontal neurons (GLMM test 2: 29 neurons, 2187 total trials) with neuron as the grouping variable. We used these three variables for both fixed and random effects including fixed and random intercepts grouping trials by neurons.

GLMM tests 3–4: for each HGC trial, we evaluated the previous HGC trial by calculating the following variables: (1) a binary index indicating whether a temporal/frontal neuron fired in the time range 200–800 ms following reward outcome (as before, only neurons that significantly increased FR following reward outcome were evaluated); (2) outcome—a binary variable indication whether the coin was caught or missed; (3) normalized total movement between reward appearance and disappearance either by avatar catching or missing; (4) number of ball hits on the way to the coin; and (5) behavioral decision (to approach or not). The analysis was done separately for temporal (GLMM test 3: 14 neurons, 1023 total trials) and frontal neurons (GLMM test 4: 19 neurons, 748 total trials) with neuron as the grouping variable.

To evaluate the connection between frontal responsivity to controlled punishment on one hand and the subsequent behavioral effect of temporal neurons to punishment on the other hand, we concentrated on four sessions (from patients 3, 4 and two sessions from patient 7) that had neurons with punishment-related FR increase in both the temporal and frontal lobes. For each HGC trial, we evaluated previous punishment calculating the following variables: (1) average firing of temporal neurons in the time range 200–800 ms following punishment outcome; (2) average firing of frontal neurons in the time range 200–800 ms following punishment outcome; (3) interaction between the previous variables; (4) normalized total movement ±1 s of outcome time; (5) time delay between the punishment outcome and subsequent HGC trial.

Similarly, for reward trials, we concentrated on three sessions (from patients 3, 7, and 10) that had neurons with reward-related FR increase in both the temporal and frontal lobes. For each HGC trial, we evaluated the previous HGC trial by calculating the following variables: (1) average firing of temporal neurons in the time range 200–800 ms following HGC reward outcome; (2) average firing of frontal neurons in the time range 200–800 ms following HGC reward outcome; (3) interaction between the previous variables; (4) normalized total movement between reward appearance and disappearance either by avatar catching or missing; (5) number of ball hits on the way to the coin (we did not add the behavior variable since all previous HGC trials in this case turned out to be approach trials).