Gender differences in cognitive and affective interpersonal emotion regulation in couples: an fNIRS hyperscanning

Abstract

Emotion regulation is vital in maintaining romantic relationships in couples. Although gender differences exist in cognitive and affective strategies during ‘intrapersonal’ emotion regulation, it is unclear how gender differences through affective bonds work in ‘interpersonal’ emotion regulation (IER) in couples. Thirty couple dyads and 30 stranger dyads underwent functional near-infrared spectroscopy hyperscanning recordings when targets complied with their partner’s cognitive engagement (CE) and affective engagement (AE) strategies after viewing sad and neutral videos. Behaviorally, for males, CE was less effective than AE in both groups, but little difference occurred for females between AE and CE. For couples, Granger causality analysis showed that male targets had less neural activity than female targets in CH06, CH13 and CH17 during CE. For inflow and outflow activities on CH06 and CH13 (frontopolar cortex), respectively, male targets had less activity in the CE condition than in the AE condition, while for outflow activities on CH 17 (dorsolateral prefrontal cortex), female targets had more activity in the CE condition than in the AE condition. However, these differences were not observed in strangers. These results suggest gender differences in CE but not in AE and dissociable flow patterns in male and female targets in couples during sadness regulation.

couples, fNIRS, Granger causality, gender difference, hyperscanning, interpersonal emotion regulation

Introduction

Individuals must control their emotions on a regular basis to respond effectively to their environment. The implementation of strategies to control or manage emotion is known as emotion regulation (ER). The process by which emotion is regulated on an individual’s own or through the help of others is defined as intrapersonal or interpersonal emotion regulation, respectively. Interpersonal emotion regulation (IER) plays an important role in a variety of interpersonal relationships, i.e. parent–child (Reindl et al., 2018), teacher–student (Sun et al., 2021) and friend relationships (Kwon and López-Pérez, 2022). Surprisingly, there has been only limited research on IER between couples (Jitaru and Turliuc, 2022). For example, Levy-Gigi and Shamay-Tsoory (2017) examined the ER of couples after they viewed pictures with negative valance. They compared the effectiveness of ER when the target turned to their long-term partner (regulator) to seek comfort and when they downregulated their own negative emotions. They found that the outside perspective of the regulator was more effective in helping to reduce the distress of the target than intrapersonal ER. In another study, Horn et al. (2019) found that positive humor as an IER strategy regulates partners’ momentary emotions by altering their psychological intimacy in couples.

Adaptive IER

An emerging area in the ER literature focuses on ER flexibility (Battaglini et al., 2023). Researchers have divided ER strategies into adaptive and maladaptive strategies according to their consequences (Battaglini et al., 2023). Adaptive ER strategies, which have been found to have more positive effects in families (Shaffer et al., 2012), schools (Salimzadeh et al., 2020) and workplaces (Feinberg et al., 2020), include affective engagement (AE) and cognitive engagement (CE) (Niven et al., 2009). AE refers to the regulation directly engaged with the target’s feelings, while CE refers to the regulation engaged with the target’s cognition, which in turn changes the target’s emotions. Behavioral results indicated that AE had higher perceived efficacy in sadness, whereas CE was more effective for anxiety/stress in healthy adults (López-Pérez, 2018).

Studies on intrapersonal ER have shown that males can implement CE in a more effective and automatic way, while females tend to use AE (Goubet and Chrysikou, 2019), pointing to possible gender differences in CE and AE during IER. Based on attachment theory, there are attachment bonds in romantic relationships that are associated with mentalizing theory and are significant in the IER process (Feldman, 2017). A meta-analysis found that males show lower anxiety and higher avoidance than females in romantic relationships (Del Giudice, 2011), which suggests that attachment bonds in couples might have an impact on gender differences in IER.

Sadness regulation in couples

Sadness is one of the basic human emotions and is an evolutionary and natural response to loss or unfulfilled goals (Beck and Bredemeier, 2016). However, persistent sadness might develop into mental illness, including depression and anxiety (Woody and Gibb, 2015; Kovacs et al., 2016), which is linked to a variety of regulatory implications (Chong and Park, 2017). Sadness is also reflected in social interactions as a response to various social stimuli and is involved in interpersonal exchanges in intimate relationships, i.e. parent–child, friends and couples (Zaid et al., 2021). In couples, sadness is the most predictable negative emotion for oneself and one’s partner, which makes sadness regulation between couples easier to achieve (Jitaru and Turliuc, 2022). When couples were in sad connotation vignettes, males tended to upregulate their emotions through CE and positive problem-focused engagement, while females tended to downregulate their emotions through negative AE and put their own feelings first (Jitaru and Turliuc, 2022). Previous studies examining the regulation of one’s own sadness found that adults showed increased activation in the left dorsolateral prefrontal cortex (DLPFC) (BA9) (McRae et al., 2010; Kanske et al., 2011). However, the neural mechanism of sadness regulation in the IER (e.g. among couples) remains unknown.

Single-brain fMRI related to IER

A recent review on the neural basis of intrapersonal ER indicated that males tend to activate cognitive control areas because they tend to adopt cognitive strategies, while females activate emotional processing areas because they are more emotionally concerned, and this difference mainly involves the prefrontal cortex (Fridlund, 2022). However, only a few studies have explored the neural mechanism of IER. The first fMRI study, which required healthy participants to moderate their own or others’ emotions, found an overlapping brain network within intrapersonal ER and IER but that the social brain was more engaged in IER than intrapersonal ER, which included the medial prefrontal cortex (BA10) (Hallam et al., 2014). Another fMRI study found that participants could regulate emotions more successfully and activate social cognitive networks more with the help of supportive sentences and pictures provided by best friends than with the help of intrapersonal ER or strangers (Morawetz et al., 2021). These single-brain studies suggested that social closeness promotes IER but did not find possible gender differences in IER during real social interactions. These single-brain studies on IER have focused on white middle-class men. The neural mechanism of how cultural context influences IER still needs more exploration.

Brain-to-brain synchrony and directionality of dynamic interactions

Functional near-infrared spectroscopy (fNIRS) and hyperscanning technology can be combined to simultaneously measure multiple brain activities in real time to better apply in social interaction (Holper et al., 2012; Babiloni and Astolfi, 2014). Following attachment theory (Feldman, 2012, 2017; Carter, 2014), the key feature of human attachment is biobehavioral synchrony, that is, behavioral synchrony and brain-to-brain synchrony. For example, a recent fNIRS study showed that the brain activity of parents and their children was synchronized in the DLPFC (BA9) and frontopolar cortex (BA10), which can predict their cooperative performance in the next trial (Reindl et al., 2018).

Interbrain synchrony (IBS) indexes brain-to-brain correlation but ignores the direction information that is of great significance. Granger causality analysis (GCA) is widely used in fNIRS hyperscanning studies to measure the directionality of dynamic interactions, such as couples (Pan et al., 2017; Duan et al., 2020; Tang et al., 2020), teacher–student interactions (Pan et al., 2018) and strangers (Sciaraffa et al., 2021). For example, a fNIRS hyperscanning study showed stronger directional synchronization from female to male in couples in a cooperative task (Pan et al., 2017). In contrast, two other couple studies found no difference in directional synchrony in a button-press collaborative task (Tang et al., 2020) and a reality-presenting problem task (Duan et al., 2020). Because there are two natural roles in IER, regulator and target, it is interesting to determine how this direction difference occurs between couples during IER.

The current study

As reviewed above, there is a gap in the literature regarding IER in couples, especially in terms of the temporal relation of brain activities between partners. The current study aimed to use fNIRS hyperscanning to investigate gender differences in the underlying neural mechanisms of IER in romantic relationships. To this end, we recruited two groups of participants from universities in China: couple dyads and stranger dyads. In the IER task, the two participants in each pair were designated as a regulator and a target according to their wishes. All participants first watched a video intended to elicit sad or neutral emotion, and then the regulator read to target a passage implementing CE or AE regulation. Next, all participants watched a second video of the same emotion type as that of the first video. Sadness ratings were obtained before and after IER regulation and exposure to the second video. fNIRS recording was implemented throughout the procedure.

For the brain regions of interest, we focused on the DLPFC and frontopolar cortices (BA9 and BA 10), which are believed to be involved in ER (Fridlund, 2022). Based on previous studies (López-Pérez, 2018; Goubet and Chrysikou, 2019), we predicted that AE and CE could be effective in sadness regulation of targets in couples, but gender differences would emerge in the CE condition but not in the AE condition since AE consistently activates attachment bonds in couples. Following previous studies on intrapersonal ER (Goubet and Chrysikou, 2019; Fridlund, 2022), we further predicted that male targets would induce more brain activity in both the inflow and outflow directions in the greater prefrontal region during CE than during AE, while female targets would produce less brain activity in one direction in the DLPFC.

Methods

Participants

We recruited 142 healthy participants, consisting of 31 heterosexual couple dyads (31 males and 31 females; 19.13 ± 1.47 years) and 40 stranger dyads (40 males and 40 females; 19.03 ± 1.23 years), through advertising in universities in Hengyang, China. Our sample size was determined by a power analysis. In our case, the minimum sample size we needed was 56 with a significance criterion of α = 0.05 and power = 0.95, and we recruited 71 pairs of participants to ensure the reliability of the experiment. We found no differences in their ages (t (118) = 0.76, P = 0.44). In addition, all participants had either normal or corrected normal vision and normal hearing function and had not previously participated in any similar experiments. In stranger dyads, the two participants in each dyad did not know each other before the experiment, and their relationship types were determined by the measurement of perceived relationship quality components (PRQC, Fletcher et al., 2016). We found a significant difference in the closeness of couple dyads and stranger dyads, which was consistent with their self-reports (t (118) = 19.19, P < 0.001). In total, data from 11 dyads of all participants were not included in the final analysis, of which six dyads were used for preprocessing and data from the other five dyads were unavailable due to equipment failures, program errors, etc. Detailed information on all participants is provided in Table 1. All participants gave written informed consent prior to the experiment and received a participation fee of RMB30 after the experiment. The study protocol was approved by the Ethics Committee of Hengyang Normal University.

Table 1.

Participant demographics in the whole sample, including 30 couple dyads and 30 stranger dyads

	Whole sample (n = 120)	Couple dyad (n = 60)	Stranger dyad (n = 60)	P
Gender (target)
Male	25	11	14	–
Female	35	19	16	–
Handedness
Left	2	1	1	–
Right	118	59	59	–
Age (years)	19.14 (1.32)	19.23 (1.37)	19.05 (1.27)	=0.448
Relationship length (days)	–	432.60 (450.14)	–	–
Measurement
PRQC	25.92 (13.48)	37.58 (4.02)	14.27 (8.52)	<0.001
RESE	42.62 (5.69)	43.47 (5.78)	41.77 (552)	=0.102
IERQ	64.43 (11.22)	64.05 (11.97)	64.80 (10.50)	=0.716

Note. Mean (Std.). PRQC: the scale of perceived relationship quality components; RESE: regulatory emotional self-efficacy scale; IERQ: interpersonal emotion regulation questionnaire.

Measurements

Prior to the experiment, all participants were asked to complete three scales. To specifically estimate relationships between dyads, we used the intimacy and trust subscales of PRQC (Fletcher et al., 2016; Wu, 2017). In this study, Cronbach’s α was 0.97. Then, we used the Regulatory Emotional Self-Efficacy scale (RESE) to assess their sense of perceived self-efficacy in managing negative emotions and expressing positive affect, which included subscales of expressing positive affect (POS), managing despondency distress (DES) and perceived self-efficacy in managing anger (ANG) (Cronbach’s α = 0.78; Caprara et al., 2008). The Interpersonal Emotion Regulation Questionnaire (IERQ) has been used to assess their IER preference, including subscales of enhancing positive affect (EPA), perspective taking (PT), soothing (S) and social modeling (SM) (Cronbach’s α = 0.88; Hofmann et al., 2016).

Materials

Video materials

First, we determined the video material based on previous literature, such as Aftershock, City of love and death and Planet Earth. Then, we produced 24 video clips, including 12 negative videos and 12 neutral videos, with similar length, intelligibility and discreteness (Gross and Levenson, 1995; Xu et al., 2010). Then, 40 healthy Chinese university students were asked to rate the arousal and valence of each video using the scale of emotional arousal and the Positive and Negative Affect Scale (PANAS), respectively (Gross and Levenson, 1995; Xu et al., 2010). During this process, we used 13 similar positive videos to ensure that the participants were emotionally stable. Finally, we selected 16 videos (8 sad videos and 8 neutral videos) in which the same type of video was highly similar to sadness arousal (sad videos: F (7, 384) = 0.64, P = 0.74; neutral videos: F (7, 384) = 1.60, P = 0.12), while a significant difference was found in sadness arousal between sad and neutral videos (F (1, 774) = 517.53, P < 0.001). All videos were matched highly in terms of valence (sad videos: Cronbach’s α = 0.93; neutral videos: Cronbach’s α = 0.94; all videos: Cronbach’s α = 0.93).

Sentence materials of interpersonal emotion regulation

On the basis of the IER strategy classification (Niven et al., 2009) and the IER examples provided in previous literature (López-Pérez et al., 2016; López-Pérez, 2018), we produced eight IER sentences in Chinese (four AE sentences and four CE sentences) and adapted them appropriately so that they could be used in all sad videos. Then, six dyads of participants were enrolled in a preexperiment to ensure the efficacy of the sentences. These IER sentences showed a significant effect of regulating sadness (AE: t (47) = 4.08, P < 0.001; CE: t (47) = 2.83, P = 0.007). Our sentence materials are provided in the supplementary materials.

Procedure and tasks

First, participants were asked to complete preexperimental tests when they arrived in the laboratory. Then, the two persons in the pair were directed to sit in front of a computer screen, 30 cm apart and with an opaque cardboard in the middle of the computer screen to keep them from seeing each other’s screen. Before the experience, 30 min are given to participants of stranger dyads to introduce themselves and interact with each other.

During the experiment, the instructions about the whole experiment were first displayed on the screen, while the experimenter was also present to explain. Once both participants indicated that they understood the procedure, they would perform a practice trial to familiarize themselves with it. The formal experiment was then started and consisted of two blocks of 16 trials each. There was a 1-min break between blocks, during which participants could close their eyes and relax their brains.

In each trial (Figure 1A), a fixation cross was first displayed in the middle of the two halves of the screen for 4000 ms, and then both participants watched a video together. Next, the target was required to evaluate his or her sadness using a 7-point Likert scale with a range from 1 (‘not at all’) to 7 (‘very much’). After the assessment, the regulator was asked to read out loud the IER sentence presented on his or her half of the screen for 5000 ms. Then, they watched the same video again, and the target was then needed to evaluate sadness again. At the end, a blank interval was presented for 7000 ms to allow for communication. In each trial, the video clip and IER sentence were presented randomly, not paired. In the same block, each clip and IER sentence were displayed four times. During the experiment, no physical touch was allowed.

Fig. 1.

The experimental design. (A) The flow of a trial. Each trial began with a fixation of 4 s, and then both the regulator (red) and target (blue) watched a video at the same time. After watching the video, the target needed to evaluate his or her sadness, and sooner, the regulator read the IER sentence. Subsequently, they watched the same video, the target evaluated sadness again, and they had communication for 7 s at the end. (B) The montage provided by NIRx Medical Technologies. Sources are depicted as red dots, detectors as green dots.

fNIRS data acquisition

With a sample rate of 7.8125 Hz, an fNIRS device (NIRScout, NIRx Medical Technologies, Berlin, Germany) was used to record raw data at two wavelengths (760 and 850 nm). We placed two sets of customized optodes (8 sources and 8 detectors, 20 measurement channels) on the PFC of two participants per montage by NIRx (Figure 1B). One was used for the male participants and the other for the female participants. The distance between the source and detector was set at 30 mm. Each optode set was placed on the participant’s PFC in line with the 10–20 system (Ramstrand et al., 2020; Dybvik and Steinert, 2021). This montage encompassed the anterior frontal lobe, specifically BA 8, 9, 10, 11, 44, 45, 46 and 47 (Ramstrand et al., 2020; Dybvik and Steinert, 2021).

fNIRS data analysis

Preprocessing

To evaluate the signal-to-noise performance of channels, we computed the relative coefficient of variation (CV, in %) of the raw data. Channels with CV values above 15% were rejected, and the remaining channels with trial CV values <5% were retained for future analyses (Schmitz et al., 2005; Schneider et al., 2011; Piper et al., 2014). The signals were then preprocessed via the functions of Homer2 based on MATLAB (The MathWorks, Inc., Natick, MA, USA) (Huppert et al., 2009). We converted intensity to optical density (Huppert et al., 2009) and performed a cubic spline correction to remove motion artifacts (Scholkmann et al., 2010). In addition, we applied a bandpass filter (0.01–0.2 Hz) in the data and converted them to concentrations using the modified Beer–Lambert law (Sassaroli and Fantini, 2004). Finally, we performed group averaging of oxygenated and deoxygenated hemoglobin concentrations by using the Homer2 block averaging functions hmrBlockAvg, with tRange: [−5, 20], synchronized by markers placed in the data at the start time of each video. In our study, we concentrated on oxygenated hemoglobin (HbO) signals due to their high sensitivity to change (Hoshi, 2007) and wide use in fNIRS hyperscanning studies (Cui et al., 2012; Baker et al., 2016; Nozawa et al., 2016; Feng et al., 2020). Following Chen et al. (2020), we calculated IBS before performing GCA (see the supplementary methods and results for IBS in details). However, 2 × 2 × 2 ANOVA did not find any significant IBS results to support our hypothesis.

Granger causality analysis

Following prior fNIRS hyperscanning studies, GCA is a reliable method to estimate the directionality between fNIRS time series by using vector autoregressive models (Granger, 1969; Chen et al., 2020). The HbO signals were first downsampled to 1 Hz to reduce the computing time, and then we used HERMES (http://hermes.ctb.upm.es/) to perform GCA for all channels. Our GCA was carried out in HERMES with a model order of 3 based on a selection of 100 random vectors. The same model order is used for all data because they have all undergone the same processing. We computed Granger causality for two groups (couple dyads & stranger dyads) of both participants’ directions (regulator→target and target→regulator). In this analysis, we did not take into account any time delays. Finally, we converted the GC values into Fisher z-statistics and used the mean GC values to replace them in bad channels for further study.

Statistical analysis

The entire statistical analysis was conducted using SPSS 22.0 (IBM, New York, NY, USA) with the alpha level set at 0.05 (two-tailed). For self-report measurements, we first ran a reliability analysis to check the quality of the tests in our study. We then conducted independent-sample t tests to confirm that there were no significant differences between couple dyads and stranger dyads, apart from closeness. To simplify the analysis, behavioral performance and GC data were defined as the ‘sadness change score’, which was calculated by subtracting the second score (after regulation) from the first score (before regulation) of the same trial and then the ‘neutral’ result from the ‘sad’ result.

For both behavioral and GC data, we conducted a 2 × 2 × 2 analysis of variance (ANOVA) with GROUP (couple dyad and stranger dyad) and GENDER (the gender of the target, male and female) as between-participant factors and REGULATION (AE and CE) as a within-participant factor to verify their effects on sadness change score. We then performed Pearson correlation analyses on all data to reveal the relationship between behavioral and neural data.

Results

Subjective measurements

The results of subjective measurements revealed no differences between the different types of participants in terms of RESE, t (118) = 1.65, P = 0.102, and IER preference, t (118) = −0.36, P = 0.716, except for their closeness, t (118) = 19.16, P < 0.001, which meant they were well matched (Table 1). When the target was male or female, there were no differences in RESE, t (118) = 0.37, P = 0.710, IER preference, t (118) =−0.72, P = 0.473 and closeness, t (118) = −1.28, P = 0.202.

Behavioral performance

The 2 × 2 × 2 ANOVA revealed a significant interaction of REGULATION × GENDER, F (1, 56) = 4.22, P = 0.045, |$\eta _p^2$|= 0.15. Subsequent simple effect analysis found that the sadness change score in CE (M = 0.03, SE = 0.11) was obviously lower than that in AE (M = 0.22, SE = 0.11), P = 0.046 when the target was male, while no difference was found in the situation of female target, P = 0.461 (all Bonferroni corrected) (see Figure 2).

Fig. 2.

Behavioral results. (A) Descriptive statistical results of the sadness difference score. (B) The interaction effect of REGULATION × GENDER on sadness difference score in couple dyads when the target was male or female. (C) The correlation between sadness difference score and enhancing positive affect (EPA) abilities of participants. AE: affective engagement; CE: cognitive engagement. C–C dyad: couple dyad; S–S dyad: stranger dyad. Error bars indicate standard errors. * P < 0.05 (Bonferroni corrected).

Directional coupling in fNIRS

For our analysis, we focused on the following channels: CH06, CH13 and CH17 within BA9 and BA10, following previous studies (Acevedo et al., 2014; Hallam et al., 2014; Reindl et al., 2018; Fridlund, 2022). A 2 × 2 × 2 (REGULATION × GENDER × GROUP) ANOVA was conducted for these channels (see Figure 3).

Fig. 3.

ANOVA results of directional coupling. (A) The interaction effect of REGULATION × GENDER × GROUP on the GC change in CH06 in the inflow direction in different dyads when the target was male or female. (B) The interaction effect of REGULATION × GENDER × GROUP on GC change in CH13 in the outflow direction in different dyads when the target was male or female. (C) The interaction effect of REGULATION × GENDER × GROUP on GC change in CH17 in the outflow direction in different dyads when the target was male or female. AE: affective engagement; CE: cognitive engagement. Inflow: from regulator to target; outflow: from target to regulator. C–C dyad: couple dyad; S–S dyad: stranger dyad. Error bars indicate standard errors. * P < 0.05 (FDR corrected).

For CH06 in the regulator→target (inflow) direction, two statistically significant interactions were identified: REGULATION × GENDER, F (1, 56) = 4.96, P = 0.030, |$\eta _p^2$| = 0.14 and REGULATION × GENDER × GROUP, F (1, 56) = 4.13, P = 0.047, |$\eta _p^2$| = 0.12. The simple effect analyses showed the following: First, in couple dyads, the ‘GC change’ of CE was lower in male targets (M = −0.88, SE = 0.50) than in female targets (M = 0.52, SE = 0.35), P = 0.025, but this difference was not observed when using AE, P = 0.215. Second, in couple dyads, the ‘GC change’ of the male target was lower in CE (M = −0.88, SE = 0.50) than in AE (M = 0.45, SE = 0.42), P = 0.023; however, this difference did not appear in stranger dyads, P = 0.871 (all FDR corrected). In conclusion, in couples, males showed less inflow brain activity of CH06 after regulation than females in the CE condition and showed less inflow activity of CH06 after regulation in the CE condition than in the AE condition.

For CH13 in the target→regulator (outflow) direction, there was a pronounced interaction of REGULATION × GENDER × GROUP, F (1, 56) = 4.36, P = 0.041, |$\eta _p^2$| = 0.17. Its simple effect analysis indicated that the ‘GC change’ in the male target (M = −0.92, SE = 0.55) was lower than that in the female target (M = 0.46, SE = 0.39), P = 0.048, when implementing CE on couple dyads, but this difference did not occur in the AE situation, P = 0.143. In addition, when the target was male, the ‘GC change’ of couple dyads in CE (M = −0.91, SE = 0.55) was lower than in AE (M = 0.98, SE = 0.61), P = 0.034; in contrast, no difference was observed in stranger dyads, P = 0.357 (all FDR corrected). In brief, after regulation, males in romantic relationships showed less outflow brain activity of CH13 than their female partners in the CE condition and showed less outflow activity of CH13 in the CE condition than in the AE condition.

For CH17 in the target→regulator (outflow) direction, there were two significant interactions: REGULATION × GENDER, F (1, 56) = 4.37, P = 0.041, |$\eta _p^2$| = 0.16 and REGULATION × GENDER × GROUP, F (1, 56) = 4.44, P = 0.040, |$\eta _p^2$| = 0.13. The results of the simple effect analyses are listed below. First, when applying CE to participants, the ‘GC change’ when the target was male (M = −0.65, SE = 0.32) was significantly lower than when it was female (M = 0.33, SE = 0.26), P = 0.021. Nevertheless, no difference was observed in the AE condition, P = 0.448. Furthermore, the ‘GC change’ in the male target (M = −1.08, SE = 0.49) was lower than that in the female target (M = 0.78, SE = 0.35), P = 0.003, during the application of CE in couple dyads, and no similar difference was found in the AE situation, P = 0.236. Furthermore, when the target was female, the ‘GC change’ of couple dyads in CE (M = 0.78, SE = 0.35) was higher than in AE (M = −0.46, SE = 0.44), P = 0.030; in contrast, no difference was found when it was male, P = 0.055 (all FDR corrected). In short, when applying regulation in couples, we found less outflow activities in CH17 of male participants than females in the CE situation and more outflow activities in CH17 of females in the CE situation than in the AE situation after regulation.

The association between the brain and behavior

For all targets, first, their sadness change score of CE was positively correlated with the EPA ability (subscale of IERQ) (r = 0.31, P = 0.016) (see Figure 2C). In addition, the ‘GC change’ of AE at CH13 in the target→regulator (outflow) direction was directly correlated with managing despondency distress ability (DES, subscale of RESE) (r = 0.26, P = 0.046) and (see Figure 4B). Furthermore, the ‘GC change’ of CE at CH17 in the target→regulator (outflow) direction was negatively correlated with perceived self-efficacy in managing anger (ANG, subscale of RESE) (r = −0.27, P = 0.041) see Figure 4C and RESE scores (r = −0.27, P = 0.040). More correlation results are provided in the supplementary material (Supplementary Figure S1).

Fig. 4.

Correlation results of directional coupling. (A) Direction of Granger causality (GC) values at CH06, CH13 and CH17 in the inflow and outflow directions. (B) The correlation between GC change in AE at CH13 in the outflow direction and managing despondency distress (DES) scores. (C) The correlation between GC change in CE at CH17 in the outflow direction and perceived self-efficacy in managing anger (ANG) scores. AE: affective engagement; CE: cognitive engagement. Inflow: from regulator to target; outflow: from target to regulator. * P < 0.05 (FDR corrected).

Discussion

This study used fNIRS hyperscanning to explore the neural mechanisms that are involved in AE and CE to regulate partners’ sadness in couples with a special focus on gender differences. At the behavioral level, for male targets in both groups, CE was less effective than AE, while for female targets, there was little difference between CE and AE. At the neural level, CE is more effective on male targets than on female targets during the IER of couples. Additionally, the ‘GC change’ of male targets in the inflow direction of CH06 (BA10, frontopolar area) and outflow direction of CH13 (BA10) in the CE condition was significantly lower than that in the AE condition when implementing regulation in couples. In contrast, female targets’ ‘GC change’ in the outflow direction of CH17 (BA9, DLPFC) was higher in the CE situation than in the AE situation in couples. In contrast, no correspondent differences were observed in strangers. In addition, our correlation results were able to distinguish regulators and targets and showed that the key to the IER process lies in targets. These results suggest that dissociable behavioral and brain patterns of male and female targets exist during the IER process, which provides important implications for understanding gender differences in IER in couples.

Gender differences in CE but not in AE during sadness regulation in couples

Neural results showed that the ‘GC change’ of male targets at BA9 and BA10 was lower than the ‘GC change’ of female targets when CE was applied to couples. To be consistent with our behavioral results, our ‘GC change’ was calculated by subtracting the second GC value (after the regulation) from the first GC value (before the regulation). This means that male targets have greater neural activities in both BA9 and BA10 than female targets when their sadness is regulated by CE. Generally, BA9 and BA10 are associated with higher-order cognitive functions, such as cognitive control of affect, mentalizing and self-knowledge (Zelazo and Cunningham, 2006). This is consistent with our correlation results between ‘GC change’ of CH13 in the outflow direction and despondency distress ability and between ‘GC change’ of CH17 in the outflow direction and perceived self-efficacy in managing anger. The key factor contributing to gender differences is the close relationship between regulators and targets of couples because gender differences did not exist in strangers, which is consistent with the studies of Levy-Gigi and Shamay-Tsoory (2017).

On the one hand, one possible explanation is that couples in our study were mainly college students at a time when most of them were in the early phase of their romantic relationships. At this stage, males take on their role, displaying acceptable characteristics and showing the ability to achieve sexual contact (Pan et al., 2017). Therefore, male targets’ automatic cognitive regulation pattern may be interrupted by their partners’ regulation so that they need more cognitive effort to regulate their own sadness. However, we found no correlation between the results and intimacy or trust scores, probably because our sample size was small. Future exploration needs to consider more elements, e.g. attachment style (Furman et al., 2002), relationship stage (Kenrick et al., 1990) and affiliation (Furman et al., 2014).

On the other hand, previous studies found that females usually focus on emotion-focused strategies and tend to activate emotion-associated areas but can apply most IER strategies flexibly and tend to seek useful support (Mak et al., 2009; Goubet and Chrysikou, 2019). In contrast, targets in our study could not choose how to regulate their emotions but accept what their partners said. Although they are not good at CE during intrapersonal ER (Mak et al., 2009; Goubet and Chrysikou, 2019), female targets in a romantic relationship may reduce cognitive control more actively and apply their partner’s regulation more flexibly when they are provided with CE strategies. Therefore, female targets in couples exhibit lower activation at BA9 and BA10 during sadness regulation of CE in IER.

Additionally, we found no gender differences during the implementation of AE in couples. There are two main explanations. One is that no true gender difference occurs at the neural level when applying AE. Attachment bonds rely on higher-order cognitive processes such as emotional control, mentalizing and self-knowledge (Feldman, 2017). Since males and females display secure attachment patterns in supportive ecologies (Del Giudice, 2009), we proposed that AE might automatically activate attachment bonds that induce male and female targets in secure affective romantic states. The other explanation is that gender-related brain areas triggered by AE are not in the brain regions we selected. This still needs further clarification using neuroimaging tools covering the whole brain.

Dissociable patterns of inflow and outflow for male and female targets in couples

When male targets in couples were regulated, their ‘GC change’ in the inflow and outflow directions at BA10 was lower in the AE condition than in the CE condition. This means that male targets can actively accept more information from their partners and pass on more information to their partners in the CE condition than in the AE condition. One possible explanation is that the attachment bonds between couples can recruit both sides of male targets’ BA10 in similar patterns when applying CE to them. According to intrapersonal ER results, males are more likely to use avoidance and expression suppression to regulate their own emotions, which is not consistent with our study (Fridlund, 2022). Perhaps this kind of attachment bond can make males in this stage focus more on their own emotions or their partners’ emotions (Pan et al., 2017). In addition, we discovered that male targets received information on the left side and sent information on the right side according to our neural results. Thus, male targets’ brain activity, which is caused by CE relative to AE, has symmetry in BA10.

For females, there were more ‘GC changes’ in the outflow direction in the left BA9 when applying CE to female targets than when applying AE, which means that females sent less information during the regulation of CE. This may be because attachment bonds can make females more likely to follow their partners and lower their output of information when they are using CE strategies that they are not good at (Mak et al., 2009; Goubet and Chrysikou, 2019). Female targets in romantic relationships may think that their partners know their thoughts and decrease the output of information. On the other hand, all participants in our study came from China, where females are required to be good wives and wise mothers (xianqiliangmu) and follow their husbands during their growth (Wang, 2012). More cross-cultural studies are needed to further clarify this possibility. As female targets’ information transmission seems to be one-way, we proposed that there is an asymmetry pattern in the CE-related neural activity of female targets. However, unlike trait-based function lateralization in females (Mak et al., 2009; Fridlund, 2022), the information flow in female targets in our study may be simply an interpersonal state-based lateralization pattern.

Clinical implications

Past research on intrapersonal ER has found gender differences in PFC using different regulation strategies (Mak et al., 2009; Goubet and Chrysikou, 2019), which is partly inconsistent with our study. Our study found gender differences in CE but no gender differences in AE in couples’ regulation of sadness. This could provide a better understanding of the IER process of sadness regulation. A recent review indicated that low-quality evidence limits couple therapy to be as effective as individual therapy in improving depressive symptoms (Barbato et al., 2018). Therefore, our study could also be more informative for the treatment of depression between couples at the neural level and provide psychologists and counselors with further information about cognitive-behavioral couple therapy (Epstein and Zheng, 2017).

We discovered dissociable brain patterns during the application of IER in couples. When they were regulated by CE, similar patterns occurred in BA10 of male targets in the inflow direction and outflow direction, while lower information flows in right BA9 of female targets in the outflow direction. These results show that affective bonds are important for IER in couples and that gender differences are significant for functional lateralization. This can provide a new perspective for understanding women’s greater focus on emotions (Mak et al., 2009; Goubet and Chrysikou, 2019), which can further combine transcranial magnetic stimulation and task state to help treat affective disorders in women, such as depression.

Limitations and future directions

There are multiple limitations of this study. First, our sample size was small. Although all data that failed Mauchly’s test of sphericity were corrected with the Greenhouse–Geisser method, it was highly unlikely for us to guarantee the statistical power and normality of the data. Second, all the couples in our study had been together for at least 1 month and were distinguished by the measurement of trust and intimacy. Nevertheless, this approach might be too simple, and a more precise definition of couples will be needed in the future. Taken together, this study provides a noteworthy future direction for the field of IER. This study is the first to explore the neural mechanism of IER in couples. Future studies can also explore how IER works neurally in social relationships in general, e.g. the workplace. These might have a more positive impact on our social relationships and help us live better.

Conclusions

IER is very important in interpersonal situations and can play a greater role in romantic relationships. Our study is the first hyperscanning study that focuses on couples during the IER process. It reveals gender differences during the IER process and functional lateralization in female targets, which can provide a new perspective for understanding IER regulation of sadness in couples. These observations provide a cognitive neuroscience basis for nontraumatic brain stimulation to treat depression in couples and help people maintain a high-quality romantic relationship or even social relationship in reality.

Supplementary data

Supplementary data is available at SCAN online.

Data availability

All data generated or analyzed during this study are included in this article. Further enquiries can be directed to the corresponding author.

Contribution statement

W.Z. and L.Q. designed the experiment and wrote the first manuscript. L.Q. collected and analyzed the data. F.T. reviewed and supervised the study. H.J.S. advised on the data analysis and revised the manuscript.

Funding

This study was granted by the Hunan Natural Science Foundation of China (2022JJ30099) and the Key Project of Teaching Reform Research in Hunan Province (HNJG-2022-0227).

Conflict of interest

The authors declared that they had no conflict of interest with respect to their authorship or the publication of this article.