Forty-six healthy participants took part in the experiment. The number of participants was predetermined based on past research (Niedeggen et al. 2014), in which data of 40 participants were analyzed to examine the effect of over-inclusion on ERPs. Four participants were excluded from the analyses due to many EEG artifacts (2 participants) or incorrect performance of the task (2 participants). Thus, the data of 42 participants (17 female, Mage = 22.52, SDage = 2.84; All right-handed except one participant) were analyzed. All participants had a normal or corrected-to-normal vision. They received payment (1250 yen / 1 h) at the end of the experiment. This experiment was approved by the National Institute of Advanced Industrial Science and Technology (AIST) Safety and Ethics committee and was only conducted after each of the participants had given written informed consent.
The visual stimuli were presented on a 17-inch cathode ray tube display (Sony, Trinitron Multiscan G220) with a resolution of 1280 × 1024 pixels, which was controlled by Windows 7, MATLAB (MathWorks Inc.), and Psychophysics Toolbox (Brainard 1997; Kleiner et al. 2007; Pelli 1997). The refresh rate of the display was set to 60 Hz. The viewing distance was approximately 70 cm.
Two visually similar cushions, one soft and the other hard, were used as the objects to be held. Both cushions were covered with white cotton and were 40 × 40 × 10 cm in size. The soft cushion was made of polyester and weighed about 850 g. The hard cushion was made of polyethylene pipes and weighed about 1000 g. We instructed participants to hold each cushion with their left arms and place it on their thighs.
Stimuli and procedure
Following our previous study (Ikeda and Takeda 2019), two computer-generated opponents (black outlined squares, 2.1° × 2.1° of visual angle) appeared at the top left and top right locations of the screen, separated by 6.5° of visual angle. Male and female facial photographs representing co-players were depicted on the screen. The facial photographs were emotionally neutral and selected from a facial expression database (Fujimura and Umemura 2018). A player, controlled by the participant (a black outlined square), appeared at the bottom of the screen. A white filled circle represented the ball (diameter 0.82° of visual angle). To precisely inform participants of the ball movements' timing, 1500 ms before the ball began moving, it flickered for 300 ms (disappearing for 50 ms and appearing for 50 ms, three times). The ball visibly traveled over a distance of 3.25° for 1000 ms until another player received it. If participants received the ball, they were required to press a left or right arrow key on a keyboard with their right index or middle finger to toss the ball to the left or right computer-generated player, respectively. Participants were instructed to toss the ball to the left and right players with approximately equal probability. If the computer players received the ball, they held the ball for a random period lasting between 1000 and 2000 ms, and then they tossed the ball to one of the other players. At the beginning of the experiment, we gave participants a cover story indicating that tossing performance was unimportant because the task was to examine mental visualization skills. We instructed participants that black squares (outlined) indicated players; the lower square corresponded to the participant, and others corresponded to computer players. Participants were aware that the other players did not actually exist (i.e., they were computer-generated players). Note that apparatus, stimuli, and the task were identical to the previous study by Ikeda and Takeda (2019) except that facial photographs were presented on the display, and the size of the ball was small.
Each participant performed 8 blocks. Participants held one type of cushion during the first 4 blocks and the other cushion type during the remaining 4 blocks. The order of soft and hard cushion conditions was randomized among participants (20 participants held a soft cushion first). Participants started the experiment after practicing ten trials of the task without holding a cushion. Each cushion condition consisted of two fair-play blocks and two over-inclusion blocks. In the fair-play block, participants received the ball 20 times (inclusion trials) and then tossed it to one of the computer-generated opponents 20 times. In this condition, the computer-generated opponents also tossed the ball to each other 20 times (micro-rejection trials). Thus, each fair-play block consisted of 60 trials. In the over-inclusion block, the participant received and tossed the ball 28 times (over-inclusion trials), and the computer-generated opponents tossed the ball to each other 4 times. Thus, each over-inclusion block also consisted of 60 trials. At the beginning of each cushion condition, participants performed the fair-play block. In the remaining three blocks, they performed one fair-play block and two over-inclusion blocks in a random order. The fair-play block provides baseline experiences of the Cyberball task. Therefore, this manipulation was expected to result in an unbiased mental set for the task at the beginning of each cushion condition. Participants had a short break after each block.
Following the previous studies (Ikeda and Takeda 2019; Kawamoto et al. 2013), during the short breaks between the blocks, participants were required to provide subjective ratings of their social pain by responding to each of the following questions: “I felt liked,” “I felt rejected,” "I felt invisible,” and “I felt powerful” (in Japanese). These questions corresponded to evaluations of self-esteem, belongingness, meaningfulness, and control, respectively, and were rated on a scale ranging from 1 (not at all) to 9 (very much). Two questions, “I felt liked” and “I felt powerful,” were reverse-scored, such that higher scores for each question indicated a greater level of social pain. We averaged these four questions as the subjective rating of social pain scores. Debriefing followed the end of the eighth block. The total experiment time was about 50 min.
The electroencephalographic (EEG) signals were acquired with a digital amplifier (Brain Products, BrainAmp standard system). The silver-silver chloride electrodes were placed at 27 scalp sites: Fp1, Fp2, F7, F3, Fz, F4, F8, FCz, T3, C3, Cz, C4, T4, CPz, T5, P3, Pz, P4, T6, PO7, PO3, POz, PO4, PO8, O1, Oz, and O2, according to the extended international 10–20 system, with AFz as the ground electrode. The EEGs were re-referenced to mathematically averaged earlobes (A1–A2) offline. To monitor blinks and eye movements, vertical and horizontal electrooculograms (EOGs) were also acquired using electrodes placed above and below the right eye and the outer left and right canthi, respectively. The impedance of all electrodes was kept below 10 kΩ. The EEGs and EOGs were digitized at a sampling rate of 1000 Hz and the time constant was set at 10 s. All EEG and EOG signals were low-pass-filtered at 30 Hz with a second-order Butterworth filter to compute the CNV, whereas the 0.1–30 Hz bandpass filter was adopted to compute the P3.
The time epochs were set between − 2000 ms and 1000 ms and between − 200 ms and 600 ms relative to the onset of ball movement (i.e., the ball’s appearance at the intermediate position between players) to compute CNV and P3, respectively. We examined the epochs in which the computer-generated players threw the ball, after excluding epochs in which the participants threw the ball. We adopted an independent component analysis using EEGLAB version 14.1.1b (Delorme and Makeig 2004) to remove eye-blink-related components. We excluded epochs in which signal changes exceeded ± 50 μV on any of the EEGs from the analysis. We averaged 69.0 and 68.7 CNV epochs in the soft and hard cushion conditions of the fair-play block. We also averaged 52.2 and 53.7 CNV epochs of the soft and hard cushion conditions. Moreover, in the inclusion trials (i.e., the ball tossed to participants in the fair-play block), we averaged 39.57 and 39.45 P3 epochs of the soft and hard cushion conditions. Furthermore, in the micro-rejection trials (i.e., the ball tossed between others in the fair-play block), we averaged 38.86 epochs and 38.57 epochs of the soft and hard cushion conditions. Finally, in the over-inclusion trials (i.e., the ball tossed to the participants in the over-inclusion block), we averaged 54.19 epochs and 53.88 epochs of the soft and hard cushion conditions.
The amplitudes of ERPs were evaluated relative to baseline (the mean amplitude of the − 2000 ms to − 1500 ms window in the CNV; the mean amplitude of the − 200 ms to 0 ms in the P3). In each condition, the CNV was estimated by the mean amplitude at the centroparietal site (average of FCz, Cz, CPz and Pz) between -1000 and 0 ms. The P3 was estimated by the mean amplitude at the centrooccipital site (average of Cz, CPz, Pz and POz) between 320 and 400 ms. These time windows and electrodes were determined from visual inspection of the grand mean waveforms and topographical maps.
A 2-way repeated measure analyses of variance (ANOVA) with cushion condition (soft vs. hard) and block (fair-play vs. over-inclusion) were conducted on subjective ratings of social pain and CNV amplitudes. Also, a 2-way repeated measure ANOVA with cushion (soft vs. hard) and trials (inclusion vs. micro-rejection vs. over-inclusion) was performed to assess the P3. Three-way mixed ANOVAs with cushion (soft vs. hard), block (fair-play vs. over-inclusion), and order was conducted when there were significant effects and/or cushion condition interactions, to rule out possible contamination by effects of cushion condition order (first holding a soft cushion vs. first holding a hard cushion). The significance level was set at 5%.
Subjective ratings of social pain
Figure 1 depicts subjective ratings of social pain in each condition. A 2-way ANOVA was conducted on subjective ratings of social pain to test whether holding a soft compared a hard cushion reduced social pain scores. Results indicated significant main effects of the block suggesting significantly greater social pain in fair-play than over-inclusion blocks (F (1, 41) = 43.83, p < 0.001, ηp2 = 0.52). Moreover, the main effect of the cushion was marginally significant (F (1, 41) = 2.92, p = 0.095, ηp2 = 0.07), indicating that social pain scores in the soft cushion condition were less than in the hard cushion condition. There was no significant interaction between the cushion and social pain conditions (p = 0.11).
Figure 2 shows grand-averaged CNVs time-locked to ball movements, and mean CNV amplitudes. A 2-way ANOVA was conducted on the CNV amplitudes to test whether holding a soft cushion increased expectations more than a hard cushion. Results indicated a significant cushion × block interaction (F (1, 41) = 8.93, p < 0.01, ηp2 = 0.18) for CNV amplitudes. Post-hoc analyses with the Ryan method indicated that CNV amplitudes were significantly larger (more negative) in the soft than the hard cushion condition for the over-inclusion block (M = − 1.27 µV vs. M = − 0.65 µV, p < 0.01), but not for the fair-play block (M = − 0.49 µV vs. M = − 0.72 µV, p = 0.30). Furthermore, the CNV amplitude was significantly larger (more negative) for the over-inclusion block than for the fair-play block in the soft cushion (M = − 1.27 µV vs. − 0.49 µV, p < 0.001), but not in the hard cushion condition (M = − 0.65 µV vs. − 0.72 µV, p = 0.76). A significant main effect of block was also observed (F (1, 41) = 4.82, p < 0.05, ηp2 = 0.11), indicating that CNV amplitude was significantly larger (more negative) in the over-inclusion block than in the fair-play block. However, the main effect of cushion was not significant (p = 0.23).
Three-way ANOVAs were conducted on the CNV amplitude to examine the effects of the cushion holding order. A cushion (soft vs. hard) × block (fair-play vs. over-inclusion) × order (first holding a soft cushion vs. first holding a hard cushion) ANOVA confirmed the significant cushion × block interaction and the main effect of the block observed in the 2-way ANOVA. Notably, there were no other interactions (ps > 0.58) or main effects (ps > 0.23), which indicated that the order of holding the cushions had no effect on the CNV amplitude.
Figure 3 shows grand-averaged P3s time-locked to ball movements and mean P3 amplitudes. A cushion (soft vs. hard) × trials (inclusion vs. micro-rejection vs. over-inclusion) ANOVA was conducted on the P3 amplitudes to examine the processes of evaluating the presented stimulus. Results indicated a significant main effect of trial (F (2, 82) = 31.80, p < 0.001, ηp2 = 0.44). Post-hoc analyses with the Ryan method indicated that the P3 amplitude was significantly larger (more positive) in inclusion trials than in over-inclusion trials; in micro-rejection trials than in over-inclusion trials, and in micro-rejection trials than in inclusion trials (ps < 0.001). However, there was neither a main effect of cushion (F (1. 41) = 2.12, p = 0.15, ηp2 = 0.05) nor a significant cushion × trials interaction (F (2. 82) = 0.25, p = 0.78, ηp2 = 0.01).