Introduction

In everyday life, individuals must often make decisions for themselves and others. Rapid evaluation of decision outcomes for oneself and others is an important function of the cognitive system, which is helpful to guide and improve future decision behavior. In the past 20 years, many studies have focused on neural mechanisms underlying the evaluation of decision outcomes (Gehring & Willoughby, 2002; Gold & Shadlen, 2007; Hajcak, Holroyd, Moser, & Simons., 2005; Holroyd & Coles, 2002; Kahnt, Heinzle, Park, & Romo, 2010; Knutson, Adams, Fong, & Hommer, 2001; Knutson, Westdorp, Kaiser, & Hommer, 2000). Many fMRI studies have shown that several brain areas are associated with reward processing during outcome evaluation, such as the striatum, nucleus accumbens, anterior cingulate cortex, and orbitofrountal cortex (Kahnt, Heinzle, Park, & Romo, 2010; Kirsch, Schienle, Stark, Blecker, & Walter., 2003; Knutson, Adams, Fong, & Hommer, 2001; Knutson, Westdorp, Kaiser, & Hommer, 2000; Preuschoff, Bossaerts, & Quartz, 2006). In particular, the technique of event-related potentials (ERPs) has provided important information for the temporal properties of the processing of outcome evaluation in humans. Researchers have identified two ERP components, the feedback-related negativity (FRN) and the P300, reflecting early and later outcome evaluative processing, respectively (Gehring & Willoughby, 2002; Hu, Xu, Li, & Mai, 2018; Miltner, Braun, & Coles, 1997; Schupp et al., 2000).

The FRN, a negative deflection in the frontocentral region of the scalp, peaks between 200–300 ms following the onset of feedback-related stimuli and is typically larger for negative outcomes (e.g., monetary losses or incorrect responses) than positive outcomes (Gehring & Willoughby, 2002; Hajcak, Holroyd, Moser, & Simons, 2005; Hauser et al., 2014; Holroyd & Coles, 2002; Nieuwenhuis, Holroyd, Mol, & Coles, 2004; Paul & Pourtois, 2017). According to the motivation/affective theory, the FRN reflects the early, rapid evaluation of the affective or motivational impact of outcomes (Gehring & Willoughby, 2002; Yu, Luo, Ye, & Zhou, 2007). However, because FRN amplitude decreases after unpredicted reward and increases after unpredicted nonreward, and dopaminergic neuron firing is related to reward (Hollerman & Schultz, 1998; Schultz, Dayan, & Montague, 1997), the reinforcement learning theory of the FRN proposes that the FRN reflects dopaminergic reward prediction errors (Holroyd & Coles, 2002; Nieuwenhuis, Holroyd, Mol, & Coles, 2004). Some studies also suggest that the FRN is associated with the salience prediction error (Talmi, Atkinson, & El-Deredy, 2013) or expectation (Cao, Gu, Bi, Zhu, & Wu, 2015). Moreover, other recent evidence has suggested that the FRN might reflect a reward positivity that is more positive following a reward compared with a nonreward outcome (Proudfit, & Hajcak, 2015; Wang, Jing, Zhang, Lin, & Valadez, 2017).

The P300 is a centroparietal positive component with a peak in the period of 300–600 ms after stimulus onset, which is larger for reward outcomes or feedback with high arousal levels. Some studies have found that the P300 is sensitive to the valence of the reward, with larger amplitudes for positive outcomes than for negative outcomes (Bellebaum & Daum, 2008; Hajcak, Holroyd, Moser, & Simons, 2005; Hajcak, Moser, Holroyd, & Simons, 2007; Kreussel et al., 2012). However, other studies reported that feedback valence had no impact upon the P300 (Sato et al., 2005; Yeung & Sanfey, 2004). The P300 also is more pronounced for self-related positive feedback than other-related positive feedback (Gray, Ambady, Lowenthal, & Deldin, 2004; Kiat & Cheadle, 2017). Moreover, P300 amplitude can be modulated by some social factors, such as personal responsibility for outcomes and interpersonal relationship (Leng & Zhou, 2010; Li, Han, Lei, Holroyd, & Li, 2011; Ma et al., 2011). Some researchers have posited that the P300 is related to allocation of attentional resources and high-level affective evaluation of outcomes, such as regret or disappointment (Gray, Ambady, Lowenthal, & Deldin, 2004; Hajcak, Moser, Holroyd, & Simons, 2007; Linden, 2005; Yeung & Sanfey, 2004).

A growing number of ERP studies have begun to explore how the brain responds to outcomes related to others and found that ERP responses to others’ feedback were similar to those elicited by the outcome of participants themselves (Fukushima & Hiraki, 2006, 2009; Hu, Xu, & Mai, 2017; Liu, Hu, Shi, & Mai, 2018; Yu & Zhou, 2006; Zhu et al., 2018). Yu and Zhou (2006) reported that the FRN was elicited when individuals observed others’ performance. They asked participants to perform a gambling task for their own interest, or merely observe another person’s performance, while the other’s gain or loss having no impact on observer’s own interests. Results showed that the FRN effect was identified in both situations with similar morphology and scalp distribution, suggesting that there are similar neural mechanisms underlying the evaluation processing of outcomes of one’s own and others’ actions. Although the processing of others' outcomes is similar to that of one’s own outcomes, the valence effect of the FRN and P300 amplitude are consistently smaller for others’ outcomes than for one’s own outcomes. (Fukushima & Hiraki, 2006, 2009; Koban, Pourtois, Bediou, & Vuilleumier, 2012; Leng & Zhou 2010; Li et al., 2010; Li, Han, Lei, Holroyd, & Li, 2011; Liu, Hu, Shi, & Mai, 2018; Marco-Pallares, Kramer, Strehl, Schroder, & Munte, 2010).

Marco-Pallares et al. (2010) further explored the temporal processing of outcome evaluation when participants’ interests were relevant to the others’ performance. The gambling task they used included parallel, reverse, and neutral situations, which meant that the other’s performance resulted in the same or opposite income of the participants or had no impact on the participants. They found that the participants in the parallel and neutral situations showed a similar FRN response to the performer’s losses, whereas the participants in the reverse situations showed the FRN for wins of the performer. The authors believe that the neural responses of observers are driven by two evaluative processes: one is related to the gain/loss of oneself, and the other is related to the gain/loss of another person. Leng and Zhou (2014) extended previous findings from passive observing to active performance. They asked participants to complete the gambling task for others’ interests rather than themselves. Results showed that others’ gains and losses also elicited the FRN and P300. These findings suggest that neural responses to the interests of others can be evoked not only in the observation situation but also in the self-participation situation.

Although ERP responses to others’ outcomes have been revealed in some studies, the neural mechanisms underlying the evaluative processing of others’ outcomes have not been well identified. Some researchers believe that the others-ERPs are possibly associated with empathy, an ability of sharing the mental and emotional states of others (Fukushima & Hiraki, 2006, 2009; Leng & Zhou, 2010; Singer, 2006; Liu, Hu, Shi, & Mai, 2018; Ma et al., 2011; Martin, Kwak, Pearson, Woldorff, & Huettel, 2016). Fukushima and Hiraki (2006) used a competitive gambling task, in which two persons alternately played a game, and one player’s gain meant the other’s loss. The results showed that the participants’ subjective ratings of trait empathy were linearly correlated with the amplitude of the others-FRN, indicating the possible association between the others-ERP and empathy. Fukushima and Hiraki (2009) further examined the relationship between the others-FRN and participants’ traits empathy by using two sets of self-report questionnaires: the interpersonal reactivity index (IRI) and the empathy quotient (EQ). They found that only the emotional aspect of trait empathy was correlated with the others-FRN, whereas the cognitive aspect of trait empathy was not related to the others-FRN.

In addition to examining the relationship between the others-FRN and the trait empathy, some researchers suggest that the evaluative processing of other’s outcomes has relationship with state empathy, which could be induced for a particular individual in a particular situation (Liu, Hu, Shi, & Mai, 2018; Ma et al., 2011). Ma et al. (2011) asked the participant to observe two others playing the gambling game in turn, one of whom was a friend and the other was a stranger to the participant. Results showed that the enhanced FRN was elicited when observing friend-related outcomes compared with those related to a stranger, indicating that the context-dependent empathy, which was induced towards socially closer persons, possibly was the key factor modulating the evaluative processing of others’ outcomes. However, another explanation was proposed by Leng and Zhou (2010, 2014), who believe that interpersonal relationship between the participants and others is actually the important factor modulating evaluating processing of others’ outcomes. Because the relationship between participants and friends is closer, the FRN effect induced by friends is greater than that induced by strangers. Until now, it remains unclear which factor modulates the evaluative processing of others’ outcomes: “empathic emotion” or “interpersonal relationship”? Some fMRI studies have demonstrated that when a person’s close friends receiving negative stimuli, the brain areas related to empathy, such as the anterior insula and the anterior cingulate cortex, are activated, which supports the former modulator to some extent (De et al., 2012; Harrison, Singer, Rotshtein, Dolan, & Critchley, 2006; Singer et al., 2004). Moreover, if the hypothesis of interpersonal relationship is true, empathy may lack the power to modulate the processing of outcome evaluation when interpersonal relationship is controlled.

The purpose of the current study was to examine whether situational induced empathy would modulate the evaluative processing of others’ outcomes when interpersonal relationship was controlled. Empathic concern is considered as state empathy, which would be elicited by and congruent with the perceived welfare of a person in need (Batson & Ahmad, 2001; Stocks, Lishner, Waits, & Downum, 2011; Woltin, Corneille, Yzerbyt, & Forster, 2011). In this study, we revised the classical gambling task (Gehring & Willoughby, 2002) and required each participant to perform it under three conditions: gambling for himself or herself (self condition) and for two strangers. One of the strangers was described as an underprivileged student in distress (stranger-in-need condition), whereas the other was depicted as a general student studying in a regular urban school (stranger-not-in-need condition). The interpersonal distance between participants and others was controlled, because both others were strangers to the participants. Based on the empathy–altruism hypothesis, which posits that people feel strong empathic emotions for others who are in need or in distress (Batson & Ahmad, 2001; Batson & Moran, 1999), we considered that the stranger-in-need scenario was the high-empathy condition and the stranger-not-in-need scenario was the low-empathy condition. Given the hypothesis that empathic concern affects evaluations of others’ outcomes, we predicted that the FRN and P300 effects would be larger in the stranger-in-need condition than in the stranger-not-in-need condition.

Methods

Participants

Twenty-four undergraduate and graduate students (12 females; mean age: 22.18 ± 2.05 years) at Renmin University of China participated in this study. All participants received 60 Chinese yuan for participation and a bonus of up to 20 Chinese yuan based on their task performance. All participants were right-handed, had normal or corrected-to-normal vision, and reported no history of neurological or psychiatric disorders. None of them consumed caffeine or alcohol on the day of the experiment. The average of sleep time was more than 7 hours on the day of experiment. Data from one female participant were excluded due to excessive movement artifacts. Thus, the final sample consisted of 23 participants. Written, informed consent was obtained from all participants. The study was approved by the Institutional Review Board of the Department of Psychology at Renmin University of China.

Procedure

At the beginning of the experiment, each participant was instructed to play the gambling game three times for different beneficiaries: himself or herself (self condition) and two strangers. The personal profiles of two strangers were predesigned and read by each participant. In the stranger-in-need condition, the beneficiary was an economically underprivileged student from a school in a remote, impoverished region (high-empathy condition). In the stranger-not-in-need condition, the beneficiary was a general student studying in a normal urban school (low-empathy condition). All participants were informed that they would receive the amount of money earned when playing for themselves; however, when the participants played games for each of the two strangers, the reward money would be donated to the beneficiaries and had nothing to do with participants. All participants were informed of how much money they earned for themselves and strangers after the experiment concluded.

Participants were seated comfortably in front of a computer screen in an electrically isolated room. They were asked to play the gambling game adapted from a task designed by Gehring and Willoughby (2002). As illustrated in Fig. 1, each trial began with a white fixation cross presented for 500 ms on a black background. Then, two gray cards were presented on either side of the fixation point along with a numeral cue (5 or 10), denoting the amount of money in the task (i.e., “5” = 0.5 Chinese yuan; “10” = 1 Chinese yuan). Participants were required to choose between the two alternatives by pressing a corresponding response button (either the F or J key on the keyboard) with their left or right index finger. When the participant responded, the chosen card was highlighted by a thick yellow border for 600–800 ms, and then the outcome behind the chosen card was displayed for 1,000 ms with “+” indicating a gain and “−” indicating a loss. The intertrial interval was 600–800 ms. In order to increase the salience of the valence of the outcome, the chosen card turned red or green to indicate a gain or loss outcome. Half of the participants were assigned red as the gain color and green as the loss color, and half were assigned green as the gain color and red as the loss color. The experiment included 270 trials, divided into 3 blocks of 90 trials per beneficiary condition. At the start of each block, participants were informed of the beneficiary for whom they would be playing and were encouraged to earn as much money as possible. Unbeknownst to the participants, gain/loss feedback was manipulated according to a random sequence, and each participant received equal exposure to each feedback condition. The order of the three blocks was counterbalanced across participants.

Fig. 1.
figure 1

A single trial in the gambling task. Each trial began with a fixation cross. Participants viewed two gray cards with a numeral cue and were required to choose one by pressing the corresponding key. Their choice was then highlighted for 600–800 ms. Thereafter, outcome feedback was presented for 1000 ms

Full size image

Stimuli were presented using the E-prime 2.0 software package (PST, Pittsburgh, PA). The formal experiment began after five practice trials per participant. After finishing the gambling task, participants were asked to complete a simple 5-point scale to rate their subjective motivation to win the game and their feeling about the loss outcome. Specifically, they were asked to rate their willingness to play the game (1 = “not at all” to 5 = “very much”), willingness to win the game (1 = “not at all” to 5 = “very much”), and feeling about loss (1 = “very unhappy” to 5 = “very happy”) for themselves, the stranger in need, and the stranger not in need, respectively. The first item measured participants’ general motivation to exert effort on this task, and the second item measured participants’ motivation to increase the welfare of oneself and others. Scores on these two questions were combined to create a composite measure of the motivation to win on behalf of each beneficiary. The last question measured whether participants felt positive or negative emotions when losing money for themselves or for strangers, which can be an indirect measure of empathic emotion toward others.

EEG recording and data processing

EEGs were recorded via a Neuroscan Synamps 2 amplifier (Compumedics, Charlotte, NC) using an elastic cap with 64 tin electrodes according to the international 10/20 system. Signals were amplified with a band-pass filter of 0.01–100 Hz and continuously sampled at 1,000 Hz/channel for offline analysis. All rows of electrode recordings were referenced to an electrode placed over the left mastoid and re-referenced offline to the average of the left and right mastoids. Vertical and horizontal electrooculograms (EOGs) were collected with electrodes placed on the left supraorbital and infraorbital and on the outer canthi of the left and right eyes, respectively. All interelectrode impedances were less than 5 kΩ.

EEG data were processed offline using Neuroscan 4.5 software. Ocular artifacts were corrected using a regression procedure implemented in Neuroscan software (Semlitsch, Anderer, Schuster, & Presslich, 1986). Raw EEG data were segmented into epochs from 200 ms before to 800 ms after feedback onset; 200 ms preceding the feedback stimulus served as baseline. Epochs containing artifacts exceeding ±75 μV were excluded from analysis. Finally, the number of artifact-free ERP trials was 79.6 (± 5.84), 78.78 (± 5.15), and 76.26 (± 5.97) for the self, stranger-in-need, and stranger-not-in-need condition respectively. Data were digitally low-pass filtered below 30 Hz and then averaged for each condition.

Average ERPs analysis

The analyses focused on the FRN and P300 elicited by outcome feedback. The FRN was measured as the mean amplitudes in the time window of 230–320 ms following feedback presentation. The P300 was defined as the most positive peak in the window of 280–400 ms after onset of feedback stimuli. Based on the topographical distribution of each ERP component and previous research (Leng & Zhou, 2014; Liu, Hu, Shi, & Mai, 2018; Yeung & Sanfey, 2004), the FRN was preliminarily calculated across three electrodes (Fz, FCz, and Cz), and the P300 was quantified across two electrodes (CPz and Pz). Results indicated that the effect of the FRN was greatest at the FCz site, and the effect of the P300 was largest at the CPz site. Hence, we focused on FCz and CPz electrodes for more detailed analyses of where the ERP effects were maximal.

Temporospatial PCA analysis

Due to the possibility that the FRN and the P300 are overlapping in averaged ERP waveforms, it is necessary to use the principal component analysis (PCA) for the decomposition of ERP components (Foti, Weinberg, Dien, & Hajcak, 2011). ERP PCA Toolkit (version 2.47), a Matlab-based software package, was employed to conduct the temporo-spatial PCA in this study (Dien, 2010). The averaged ERPs of each condition of each participant were imported to the toolbox, and then a two-step PCA procedure was performed as in the previous studies (Wang, Zhang, Jing, Valadez, & Simons, 2016; Zhang et al., 2013). In the first step, a temporal PCA with promax rotation was performed on all time points for each participant’s average ERPs. In the second step, a spatial PCA was conducted for each of the resultant temporal factors using all of the recording electrodes with an infomax rotation. Finally, eight temporal factors × one spatial factors were extracted based on the Scree plot, yielding eight temporospatial factor combinations. Two of them were identified as the PCA-FRN and PCA-P300, because they accounted for more than 5% of the variance in the averaged ERPs and shared the time course of the FRN and P300. Based on the visual inspection of the factor combinations, the PCA-FRN and PCA-P300 were measured as the mean amplitudes in the time window of 250–350 ms and 350–450 ms respectively, at the channel Cz where they reached maximum.

Statistics

The FRN, P300, and PCA factor data were all subjected to repeated-measures analysis of variance (ANOVA) for two within-subjects factors: beneficiary (self vs. stranger-in-need vs. stranger-not-in-need) and reward valence (gain vs. loss). The significance level was set at 0.05 for all statistical analyses. Bonferroni corrections were performed for post hoc testing of significant main effects, whereas simple effect analysis was used to test significant interactions. Greenhouse-Geisser correction of the ANOVA assumption of sphericity was applied as appropriate. Effect sizes in all ANOVA analyses were reported by partial eta-squared (ηp2), with 0.05 representing a small effect, 0.10 representing a medium effect, and 0.20 representing a large effect (Cohen, 1973). All statistical analyses were performed in SPSS 23.0 (SPSS, Inc., Chicago, IL).

Results

Behavioral results

The mean (±SD) reaction times (RTs) for choice responses in the three conditions were 716 ± 56 ms (self), 659 ± 69 ms (stranger-in-need), and 456 ± 74 ms (stranger-not-in-need). One-way ANOVA revealed a significant effect of the beneficiary on RTs [F(2,66) = 3.76, p < 0.05, ηp2 = 0.10]. Post hoc tests indicated that participants responded more quickly in the stranger-not-in-need condition than in the self and stranger-in-need conditions (ps < 0.05). No differences were observed between the self condition and the stranger-in-need condition (p = 0.20). These results indicated that compared with the low-empathy condition, the high level of empathy may result in increased deliberation times when individuals make decisions for a stranger (Lewinsohn & Mano, 1993; Paulus, 2005).

We defined option “10” as the risky choice (high risk and high reward) and option “5” as the risk-avoidant choice in the gambling task. The ratio of the selection of “10” in the gambling task was 51.32 (±12.92) %, 53.25 (±15.52) %, and 55.61 (±14.21) % for the self, stranger-in-need, and stranger-not-in-need conditions respectively. One-way ANOVA was used to compare the ratio of risky choices from each participant among the three beneficiaries, but no significant difference was found among them [F(2,66) = 0.82, p = 0.4, ηp2 = 0.021].

Subjective ratings

Figure 2 presents subjective ratings of motivation to win and feelings of loss. As the data of subjective ratings were not normally distributed, the differences of subjective winning motivation among three conditions were determined by Friedman test. The results revealed a significant effect of the beneficiary [F(2,66) = 27.78, p < 0.001]. Bonferroni-corrected post hoc tests showed that participants’ motivation to win for the stranger not in need was lower than that for the self and for the stranger in need, (ps < 0.001), whereas the latter two did not differ (p = 0.363).

Fig. 2.
figure 2

Subjective ratings of motivation to win and feelings of loss. Error bars indicate standard error of the mean. ***p < 0.001

Full size image

Friedman test on subjective ratings of participants’ feelings of losing money for the beneficiary revealed a significant effect of the beneficiary [F(2,66) = 18.43, p < 0.001]. Bonferroni-corrected post hoc tests showed that participants felt more empathy toward the underprivileged student than the general student (p < 0.001); a similarly negative feeling was identified when participants lost money for themselves and for the underprivileged student (p = 0.68).

ERP results: conventionally averaged ERPs

FRN

Figure 3a illustrates grand-average ERP waveforms at the FCz site. The mean amplitude of the FRN was analyzed by a 3 (beneficiary: self vs. stranger-in-need vs. stranger-not-in-need) × 2 (reward valence: gain vs. loss) repeated-measures ANOVA. Results showed that the main effect of the beneficiary was significant [F(2, 21) = 11.61, p < 0.01, ηp2 = 0.49], indicating that the FRN amplitude differed among the three beneficiary conditions. The main effect of reward valence was also significant [F(1,22) = 55.15, p < 0.01, ηp2 = 0.72], with a more negative FRN when participants received a loss outcome rather than a gain outcome. More importantly, findings revealed a significant interaction effect of beneficiary × valence [F(2,21) = 6.25, p < 0.01, ηp2 = 0.37], suggesting that the FRN effects of win and loss differed among the beneficiaries. Further simple effect analyses were conducted to investigate this interaction. The results indicated that the valence effect of the FRN was more pronounced for the stranger-in-need condition [F (1, 22) = 57.40, p < 0.001, ηp2 = 0.72] and the self condition [F (1, 22) = 50.09, p < 0.001, ηp2 = 0.69] than for the stranger-not-in-need condition [F (1, 22) = 24.53, p < 0.001, ηp2 = 0.52].

Fig. 3.
figure 3

(a) Grand-average ERP waveforms at FCz electrode site; gray areas indicate the time window of the FRN (230–320 ms) used for statistical analysis. (b) Difference waveforms of loss minus gain; gray areas indicate the time window of the d-FRN (230–320 ms) used for statistical analysis. (c) Topographic maps of different waveforms (loss minus gain) in the 230–320 ms time window for self, stranger-in-need, and stranger-not-in-need conditions

Full size image

Figure 3b and c depict loss-minus-gain difference waveforms and topographic maps in the time window of 230–320 ms, in which we can more intuitively observe the difference of the FRN valence effect among three beneficiaries. We measured the mean amplitude of the FRN on the difference waves (d-FRN). Pearson correlation analysis was conducted between the d-FRN of three beneficiary conditions and task-related state measurements (motivation and state empathy) globally. Results indicated that the d-FRN was negatively correlated with subjective scores of motivation (r = −0.308, p < 0.01) and marginally negatively correlated with scores of empathic concern (r = −0.141, p = 0.053).

P300

Figure 4a shows grand-average ERP waveforms at the CPz site. The peak amplitude of the P300 at CPz also was analyzed by a 3 (beneficiary: self vs. stranger-in-need vs. stranger-not-in-need) × 2 (reward valence: gain vs. loss) repeated-measures ANOVA. The main effect of reward valence was significant [F(1,22) = 6.45, p < 0.05, ηp2 = 0.226], indicating that the P300 was larger for gain than for loss. The main effect of the beneficiary was also significant [F(2,21) = 7.28, p < 0.05, ηp2 = 0.409]. Bonferroni-corrected pairwise comparison1s showed that the P300 was larger in the self condition than in both stranger conditions (ps < 0.05), whereas there was no difference between the stranger-in-need and stranger-not-in need conditions (p = 0.092). The interaction effect of beneficiary and reward valence was not significant [F(2,21) = 0.583, p = 0.57, ηp2 = 0.053]. Pearson correlation analysis also was conducted between the P300 for gain trails of three beneficiary conditions and task-related state measurements, as well as between the P300 for loss trails of three beneficiary conditions and task-related state measurements. However, neither the P300 of gain trails nor the P300 of loss trails was correlated with the subjective scores of motivation (rwin = 0.181, p = 0.12; rloss = 0.099, p = 0.41) or empathic concern (rwin = −0.109, p = 0.17; rloss = −0.139, p = 0.25).

Fig. 4.
figure 4

(a) Grand-average ERP waveforms from CPz electrode site; gray areas indicate the time window of the P300 (320–400 ms) in which peak amplitude was measured. (b) Topographic maps of the P300 for the self, stranger-in-need, and stranger-not-in-need conditions

Full size image

ERP results: temporospatial PCA factors

Figure 5a shows grand-average waveforms of the PCA-FRN at the Cz site. For the amplitude of the PCA-FRN, the main effect of the beneficiary was significant [F(2, 21) = 9.17, p < 0.01, ηp2 = 0.46], indicating that the PCA-FRN differed among the three beneficiary conditions. The main effect of reward valence was significant [F(1, 22) = 45.93, p < 0.01, ηp2 = 0.67], with a more negative PCA-FRN for loss than for gain. Importantly, there was a significant interaction effect of beneficiary × valence [F(2, 21) = 5.36, p < 0.05, ηp2 = 0.33]. Further simple effect analyses were conducted to investigate this interaction. The results indicated that the valence effect of the PCA-FRN was smaller in the strangers-not-in-need condition [F(1, 22) = 23.07, p < 0.01, ηp2 = 0.51] than in the stranger-in-need condition [F(1, 22) = 49.15, p < 0.01, ηp2 = 0.69] and the self condition [F(1, 22) = 40.86, p < 0.01, ηp2 = 0.65].

Fig. 5.
figure 5

Grand-average waveforms of the PCA-FRN (a) and PCA-P300 (b) at Cz electrode site. Gray areas indicate the time window of the PCA-FRN (250–350 ms) and the PCA-P300 (350–450 ms) used for statistical analysis

Full size image

Figure 5b illustrates grand-average waveforms of the PCA-P300 at the Cz site. For the amplitude of the PCA-P300, the main effect of reward valence was significant [F(1, 22) = 4.61, p < 0.05, ηp2 = 0.17], with a larger PCA-P300 for gain than for loss. The main effect of the beneficiary also was significant [F(2, 21) = 6.07, p < 0.01, ηp2 = 0.36]. Bonferroni-corrected pairwise comparisons showed that the PCA-P300 was larger in the self condition than in both stranger conditions (ps < 0.05), whereas there was no difference between the stranger-in-need condition and stranger-not-in-need condition (p = 0.29). However, the interaction between beneficiary and reward valence did not reach significance [F(2, 21) = 1.76, p = 0.19, ηp2 = 0.14].

To confirm the component identification, Pearson correlations between the amplitude of PCA components and conventionally averaged ERP components were computed. Results indicated the significant correlations between the FRN and the PCA-FRN (r = 0.97, p < 0.01), as well as between the P300 and the PCA-P300 (r = 0.94, p < 0.01).

Discussion

In this study, using a gambling task in which participants earned money for themselves and for two strangers, we examined the neural mechanism of empathic concern modulating the evaluation processes of others’ outcomes when the interpersonal relationship was controlled. The ERP results showed that a larger FRN effect emerged in the stranger-in-need condition (high empathic concern) than in the stranger-not-in-need condition (low empathic concern), but no difference was found in the FRN effect between the self condition and the stranger-in-need condition. In addition, the P300 was larger in the self condition than in the two stranger conditions, but no difference emerged between the stranger-in-need and stranger-not-in-need conditions. Moreover, the PCA was applied to decompose ERP components, which were possibly overlapping in grand-averaged waveforms. The PCA factors showed a similar FRN and P300 pattern, which repeated our ERP results. These findings suggest that empathic concern can modulate the early stage of outcome processing. When gambling for strangers in need, individuals’ neural responses to outcomes demonstrated an enhanced FRN effect.

The FRN was more negative-going to loss than to gain when gambling for each beneficiary, replicating the results of previous studies (Gehring & Willoughby, 2002; Kang, Hirsh, & Chasteen, 2010; Leng & Zhou, 2010, 2014; Ma et al., 2011; Zhu et al., 2016). These findings indicated that the FRN can be elicited by one’s own outcomes and others’ outcomes. More importantly, as expected, the FRN effect (loss minus win) was larger for the stranger in need than for the stranger not in need, suggesting that outcome feedback elicited a larger FRN effect in the high-empathy condition than in the low-empathy condition. Furthermore, subjective ratings of emotional empathy revealed marginally negative correlations with the FRN effect. Together, these results imply that the level of empathy modulated evaluative processing of others’ outcomes. Based on the affective/motivational theory that the FRN reflects the early, rapid evaluation of the affective or motivational impact of outcomes (Gehring & Willoughby, 2002; Yu, Luo, Ye, & Zhou, 2007), a larger FRN effect in the high-empathy condition suggests stronger empathic emotional/altruistic motivational impacts of others’ outcomes.

Some scholars have argued that the modulator of others’ outcomes should be interpreted as an interpersonal factor (Kang, Hirsh, & Chasteen, 2010; Leng & Zhou, 2014; Ma et al., 2011; Zhu et al., 2016). Leng and Zhou (2014) reported increased FRN amplitude when participants played a gambling game on behalf of their best friends than for a stranger. Zhu et al. (2016) found similar results when participants made decisions for their mothers and for strangers. However, in the present study, we still observed the FRN effect after controlling for interpersonal factors. Larger FRN effects emerged when participants played the gambling game for a stranger in need than for one not in need. In addition, the study has shown that empathic responses can be modulated by the context of an interpersonal interaction (Dewall & Baumeister, 2006). Thus, it is possible that interpersonal factors modulate individuals’ sense of empathy first, and then the empathy level directly affects evaluative processing of others’ outcomes.

The FRN effect showed no difference between the self condition and stranger-in-need condition in this study, indicating that in the early processing stage of outcomes, neural responses to outcome feedback for oneself were similar to those for others upon inducing empathic concern. This result might support the notion of shared neural representations in empathy, which suggests that either observing or imagining affective states in others activates neural networks involved in first-hand experiences of these states (Cacioppo & Decety, 2009; Cui, Abdelgabar, Keysers, & Gazzola, 2015; Lamm, Decety, & Singer, 2011; Marsh et al., 2014; Singer et al., 2004). Singer et al. (2004) asked volunteers to observe their loved ones suffering from painful stimuli. Findings revealed that relevant brain areas, such as the anterior insula and dorsal-anterior midcingulate cortex, were activated in direct pain and vicarious pain situations, and the neural response to another’s pain was as strong as that to one’s own pain. Hence, the “shared representative network” was markedly activated in the vicarious pain condition when empathy was induced for others. Further evidence has substantiated the extent of empathy for basic human emotions, such as fear (Gelder et al., 2004), anger (De et al., 2012), sadness (Harrison, Singer, Rotshtein, Dolan, & Critchley, 2006), and disgust (Benuzzi, Lui, Duzzi, Nichelli, & Porro, 2008). Mobbs et al. (2009) extended pain-related empathy to social emotions by contrasting neural responses to earning rewards for socially desirable others versus earning money for oneself. The authors identified a similar neural mechanism in both situations, confirming that the shared representation network could apply to complex social emotions elicited by favorable or unfavorable outcomes. Accordingly, the shared neural network could be evoked in outcome processing when individuals expressed high empathic concern, making them more likely to experience another’s feelings. Therefore, in the stranger-in-need condition of this study, individuals may feel more empathic concern for underprivileged students and express a strong altruistic motivation to help them, which could influence evaluative processing of others’ outcomes.

Findings related to the P300 showed that the amplitude was larger for monetary wins than for losses when gambling for each beneficiary, consistent with previous studies (Goyer, Woldorff, & Huettel, 2008; Hajcak, Moser, Holroyd, & Simons, 2007; Leng & Zhou, 2010; Wu & Zhou, 2009). Interestingly, the P300 amplitude was larger in the self condition than in the two stranger conditions. It is consistent with the previous finding that the P300 was larger for the self-related outcomes than for the others’ outcomes (such as friends or strangers) (Ma et al., 2011). Research also has found the P300 to be larger for self-relevant stimuli relative to control stimuli (Gray, Ambady, Lowenthal, & Deldin, 2004; Tacikowski & Nowicka, 2010). The P300 has been thought to be related to attentional allocation as well (Gray, Ambady, Lowenthal, & Deldin, 2004; Linden, 2005). Our findings indicate that participants allocated more attentional resources to their own outcome feedback than to that of others.

In contrast to the FRN, the P300 showed no difference between the high-empathy and low-empathy conditions, indicating that empathic concern did not modulate the late evaluative processing of others’ outcomes. Moreover, subjective scores of empathic emotion were only related to the FRN, not the P300. These results indicate that the allocation of attentional resources reflected by the P300 cannot be modulated by empathic concern. The possible explanation is that the empathic emotion elicited in the present study is not strong enough to affect the allocation of attentional resources. The FRN could be modulated by the empathic concern, indicating that the empathic emotion toward others is indeed induced in the stranger-in-need condition. This further supports the affective/motivational theory of the FRN (Gehring & Willoughby, 2002). Leng and Zhou (2010) reported that the P300 was larger when gambling for friends than for strangers; therefore, the P300 could likely be modulated by an interpersonal relationship but not by empathic concern. As the others in the high-empathy and low-empathy conditions were strangers to the participants in this study, it is unsurprising that the P300 showed no differences between these conditions. However, because little work has considered this effect thus far, we must treat such speculative conclusions cautiously.

In summary, we found that empathic concern could modulate the early evaluative processing of others’ outcomes. A stronger FRN effect in the stranger-in-need condition than in the stranger-not-in-need condition implies stronger empathic emotional/altruistic motivational impacts of outcomes for strangers in need. Despite our findings, this study had several limitations. For instance, we manipulated the level of empathic concern based on vulnerable strangers, which could activate altruistic motivations; thus, it was difficult to exclude the influence of motivation in the evaluative processing of others’ outcomes. In future studies, it would be worthwhile to separate the factors of altruistic motivation and empathic concern and delineate their influences on evaluations of others’ outcomes. In addition, accumulating evidence has revealed differences in women’s and men’s capacity for empathy (Christov-Moore et al., 2014; Gardner, Sorhus, Edmonds, & Potts, 2012; Schirmer, Simpson, & Escoffier, 2007) as well as among individuals with different social value orientations (Declerck & Bogaert, 2008). Fukushima and Hiraki (2006) found that discernable medial frontal negativity to opponents’ outcomes emerged only for female participants but not for males. Therefore, individual differences in how empathy modulates outcome evaluations warrant further investigation.