Similar trajectories of psychological states predict marital satisfaction Open Access

The proverb “Birds of a feather flock together” has long been a guiding principle in research on romantic attraction and marital satisfaction (Watson et al., 2004). A substantial body of literature consistently demonstrates that greater similarity between partners is associated with higher levels of marital satisfaction across a variety of attributes (Epstein & Guttman, 1984). These attributes can be broadly categorized into three domains: demographic variables (e.g., age, religion, education, ethnicity, physical attributes, attractiveness, intelligence quotient, and socioeconomic status), attitudinal domains (e.g., values, social attitudes, and interests), and personality traits (Gattis et al., 2004; Russell & Wells, 1991). However, despite this overwhelming evidence, some studies have introduced nuances to the idea of similarity. For example, research has suggested that in certain cases, perfect similarity may not be ideal. Specifically, in individuals who exhibit low levels of conscientiousness or extraversion, having a partner with higher levels of these traits may lead to greater marital satisfaction. This phenomenon, referred to as the “beneficial compensation effect,” suggests that dissimilarity in certain traits can complement and enhance relationship dynamics (Shiota & Levenson, 2007). Due to these conflicting results, accumulating evidence suggests that personality dimensions are not consistently associated with marital satisfaction (Rammstedt & Schupp, 2008; Watson et al., 2004). These studies primarily focus on comparing partners’ similarity in characteristics measured through behavioral assays or self-report measures. However, the similarity of one’s perceptions, thoughts, and feelings about their surroundings are probably more important (Parkinson et al., 2018). Thus, it is plausible to examine the deeper similarities of couples, which reflects real-time mental responses.

Previous studies have reported that people tend to establish intimate relationships with others who reflect commonalities in perceiving, thinking, and reacting to the world (Clore & Byrne, 1974). Individuals’ interpretations and responses to their environment increase the predictability of their mutual thoughts and actions during social interactions (Berger, 1975). Moreover, when interacting with others who have similar thoughts, people will reinforce their own values, opinions, and interests and make positive affective responses, thus promoting their attraction to each other (Parkinson et al., 2018). Previous studies have demonstrated that the interpersonal synchronization provides high-quality communication (Jiang et al., 2012), and that high predictability allows more enjoyable social interaction (Zheng et al., 2018). Recent functional neuroimaging evidence suggests that the neural activities of two individuals are synchronized when they perform a cooperative or a competitive task and that the level of inter-individual neural synchronization is significantly associated with their similarity in perception and understanding of their surroundings (Cui et al., 2012; Stephens et al., 2010). In addition to complex cognitive tasks, functional neuroimaging data acquired with the naturalistic paradigm provided valuable insights to detect mental processing (Wilson, 2004). This method has several advantages over self-report measures in capturing individual differences in mental processing. It allows real-time observation of mental activities as they occur, eliminates the need for participants to introspect about their mental states, and reduces self-presentation biases (King & Bruner, 2000). Furthermore, self-report questionnaires are typically constrained by a limited number of specific questions. In contrast, functional neuroimaging during the free viewing of audiovisual stimuli allows for the simultaneous measurement of brain activities associated with perceptual, cognitive, and affective processes. Numerous studies have shown that inter-subject correlations (ISCs) of fMRI responses to natural stimuli capture meaningful interpersonal similarities in how people attend to and make sense of what they see and hear. ISCs reflect differences in how participants, when instructed to adopt different mental perspectives, attend to, interpret, and store perceived information (Lahnakoski et al., 2014). Thus, revealing the temporal sequence of neural responses elicited by natural stimuli provides a meaningful window into the unconstrained processing of these stimuli. This processing varies depending on people’s personality traits, goals and values, and preexisting knowledge and assumptions (Berger & Calabrese, 1974). Therefore, intimate partners may be exceptionally similar in how they attend to, interpret, and emotionally react to their surroundings. Indeed, recent findings suggest that social network proximity correlates positively with similarities in the temporal evolution of regional mean neural response amplitudes elicited by natural stimuli, especially in brain regions linked to functions like attention allocation, narrative interpretation, and emotional responses (Parkinson et al., 2018). Modern mate selection probably makes more sense because similarity helps marriage—after all, publicly blinding oneself to another person’s loyalty and forward thinking, not only does a radical change in lifestyle for some, but of course, living day in, day out with the mate requires a certain degree of interaction. Recent studies have demonstrated that brain similarities that reflect real-time mental responses to subjective perceptions, thoughts, and feelings about interpersonal and social interactions are strong predictors of marital satisfaction (Li et al., 2022), suggesting the association between inter-subject synchronization in time series of neural responses and marital satisfaction. However, extensive literature using multi-voxel pattern analysis (MVPA) on fMRI data emphasizes the importance of examining not only univariate response magnitudes but also spatially distributed response topographies (Haxby et al., 2011). Recent studies have integrated ISC and MVPA to investigate the similarity of the trajectory in neural response pattern (Chang et al., 2021; Nastase et al., 2019). If the topology of the distributed responses of brain regions at a given moment can be considered as an approximate index of their mental state, then such an analysis could provide insight into the evolution of a person’s mental state over time. A recent study has shown that the temporal trajectories of MVPs in response to naturalistic stimuli were unusually similar among friends and were associated to social networks (Hyon et al., 2020).

In the current study, we utilized individual differences in neural responses to test whether couple-wise similarity in mental states, as indexed by temporal trajectories of MVPs, predicts marital satisfaction. We found that the inter-subject similarity of the trajectory in neural responses was positively correlated with marital satisfaction (highly satisfied married couples exhibited more similar trajectory of psychological state than less satisfied married couples). Additionally, previous studies indicated that humans automatically segment continuous sensory input into discrete events (Baldassano et al., 2017). Therefore, by using modified Hidden Markov Model (HMM), we identified event states in married couples across large-scale brain networks. Compared with high-level satisfaction married couples, the married couples with low-level satisfaction represented fewer/longer events in dorsal attention and default mode networks during viewing movies. The similarity of the trajectory within manually event states was significantly correlated to the marital satisfaction. These findings demonstrate similar trajectories of mental states and provide additional insights into the neurobiological basis of pair-bonding.

This study was approved by the Institutional Review Board of the University of Electronic Science and Technology of China and was conducted in accordance with the Declaration of Helsinki. All the human participants provided written informed consent after the purpose and protocols of the study had been fully explained to them.

Forty-eight Chinese heterosexual couples, totaling 96 participants, were recruited from local communities through flyers or internet advertisements. All participants were right-handed and had an average age of 35.97 ± 6.1 years. The couples had been married for at least 1 year (6.19 ± 5.25), and marriage was the first for both spouses. The initially published paper provided a more detailed description of the data collection procedures (Li et al., 2022). In addition to the fMRI data from couples while they viewed marital and object-related movie clips, resting-state (eyes closed) fMRI data from this cohort of couples were also included in this study. Here, we will describe the aspects of additional data collection and analysis that are most relevant to the new analysis. Following the Chinese Marital Quality Inventory (CMQI) scoring criteria, the couples were categorized into two groups: those with high marital satisfaction (CMQI total scores >60) and those with low-level marital satisfaction (CMQI total scores <60). Marital satisfaction scores for each couple were calculated by averaging the scores of both partners. In the fMRI data from movie-viewing, 12 married couples were excluded due to high levels of head motion. In the fMRI data from resting state, none couples were excluded. Therefore, resting-state fMRI data from all 48 couples and fMRI data from the remaining 36 couples while viewing naturalistic movies were used in further analysis. Participants had no history of psychiatric or neurological disorders and were willing and eligible to participate in the fMRI study. All of them had normal or corrected-to-normal visual acuity. Written informed consent was obtained from all the participants after fully explaining the purpose and protocols of the study. This study was approved by the Institutional Review Board of the University of Electronic Science and Technology of China and was conducted in accordance with the Declaration of Helsinki.

Participants were scanned using a 3T GE DISSOVERY MR750 scanner (General Electric) with an eight-channel prototype quadrature birdcage head coil. An echo-planar sequence (30-ms echo time; 2,000-ms repetition time; 3.75 × 3.75 × 3.2-mm resolution; 64 × 64 matrix size; flip angle = 90°; 240 × 240-mm2 field of view; 43 interleaved transverse slices with no gap; 3.2-mm slice thickness) was used to acquire functional images. The resting-state functional data consisted of 410 dynamic scans, with a total functional data acquisition time of ∼13.6 minutes.

Resting-state fMRI data were preprocessed using the Data Processing and Analysis of Brain Imaging toolbox (v4.3) (https://rfmri.org/dpabi). For each participant, the first 10 volumes were discarded due to instability of the initial MRI signal. Slice-timing correction and head-motion realignment were performed on the remaining 400 volumes, and all 48 couples were included with low levels of head motion during scanning (i.e., head motion of either the husband or wife <3.0-mm translation or 3° rotation). Brain images from these participants were then spatially normalized to a standard template for the Montreal Neurological Institute (MNI) and resampled to 3 × 3 × 3 mm³. The normalized images were then linearly detrended to reduce the effects of signal drifts. Nuisance covariates (24 parameters of head motion, white matter signal, cerebrospinal fluid signal, and global signal) were regressed out from the data (Friston et al., 1996). All the images were smoothed with a 6 × 6 × 6-mm³ full-width at half-maximum Gaussian kernel. Linear detrending and bandpass filtering (0.01–0.1 Hz) were subsequently performed to reduce the effects of low-frequency drift and high-frequency noise. Data scrubbing was used to eliminate potential motion artifacts (Power et al., 2012). Signal outliers whose framewise displacement (FD) >0.5 mm with prior 1 and later 2 volumes were fitted to the clean portion of the time series by using a third-order spline.

In the present study, we used the previously published whole-brain template (resampled to MNI152NLin2009cAsym standard space) with 200 brain regions (Hyon et al., 2020), each of which was associated with one of the following brain networks from Yeo et al. (2011): seven-network parcellation—the visual (VN), somatomotor (SMN), dorsal attention network (DAN), ventral attention (VAN), limbic (LN), frontoparietal task control (CEN), and default model network (DMN) (Yeo et al., 2011).

For both the resting-state and move-viewing fMRI data, in each of the brain regions, the MVPs were extracted for each time point, and the following analyses were performed independently for each of the 200 each region of interests (ROIs) (Fig. 1a). We calculated pairwise Pearson correlations between MVPs at each time point to construct a matrix that captured the trajectory of the MVPs for each participant (Chang et al., 2021; Nastase et al., 2019). In this matrix, elements closer to the main diagonal represent pairwise correlations between MVPs occurring closer in time, while elements further from the main diagonal represent correlations between MVPs that are more temporally separated. In each ROI, Inter-subject trajectory similarity (ISTS) was defined as the Pearson correlation between the vectorized trajectory structures of the husband and wife, or between randomly selected male and female participants. Random male-female pairings are generated by pairing a male with a randomly selected female who is not married to the selected male, without repetition. To illustrate how relative similarities in temporal trajectory in each brain region varied as a function of marital status, the ISTS was normalized (i.e., z-scored) across high-level marital satisfaction, low-level marital satisfaction, and random pairs for each region. The resulting similarity vectors (ISTS-Z scores) for each of the 200 anatomical ROIs were normalized to have a mean of zero and a standard deviation of one (Parkinson et al., 2018).

Whole-brain ISTS measure between married couples was obtained by averaging the ISTS across all 200 ROIs. A 2 × 2 ANOVA with whole-brain ISTS as the dependent variable was used to determine the effects of fMRI state (movies-viewing vs. resting-state), marital status (high-level marital satisfaction vs. low-level marital satisfaction vs. random pairs), and their interaction. Between-group differences (high-level marital satisfaction vs. low-level marital satisfaction vs. random pairs) in the whole-brain ISTS while viewing movies and in the resting state were assessed by using a two-sample t test (two-tailed). Pearson correlation was used to determine the relationship between marital satisfaction and whole-brain trajectory similarity while participants were viewing movies and in the resting state. Two-sample t tests were further performed on the whole-brain ISTS while viewing marital and nonmarital movies between married couples with high-level marital satisfaction and low-level marital satisfaction and random pairs. The significance threshold was set at FDR-corrected P < 0.05 for multiple comparisons.

Partial least squares (PLS) regression was used to determine the relationship between marital satisfaction and the regional ISTS in each of the 200 brain regions between couples. The ISTS of all brain regions were used as predictor variables of marital satisfaction in the PLS regression (Abdi, 2010). The PLS-1 was the linear combination of ISTS across regions that was the most strongly correlated with marital satisfaction. Permutation testing (5,000 times) was used to test the null hypothesis that PLS-1 explained no more covariance between ISTS and marital satisfaction than expected by chance. Bootstrapping (bootstrap samples = 500) was used to estimate the variability of each regional ISTS’s weight in PLS-1. The ratio of regional trajectory similarity to bootstrap SE was used to calculate z values. The confidence intervals (CIs) for the ISTS in each brain region were calculated using z values. Brain regions that reliably contributed to PLS-1 were subsequently identified after FDR correction (P < 0.05) for multiple comparisons.

Previous research has shown that during realistic continuous perception, participants progress through a sequence of discrete event representations (hidden states) and that each event has a distinct (observable) signature (a multi-voxel fMRI pattern) that is present throughout the event (Baldassano et al., 2017). Notably, the optimal number and boundary of events varied across the cortex. A data-driven event segmentation model based on an HMM was used and fitted to the fMRI data while the participants were viewing movies. This model temporally divided the time into “events” with stable patterns of activity punctuated by “event boundaries,” where the activity pattern rapidly transitions to a new stable pattern. All HMM analyses were performed using the Brainiak toolbox function, Event Segmentation with Hidden Markov Models (Kumar et al., 2020). Each Hidden Markov Model (HMM) state was characterized by a distinct mean activity pattern across all brain regions within the large-scale network parcellation. As indicated by prior studies, this analytical framework incorporates the restriction that the HMM cannot return to a state once it has transitioned away from it. Put differently, each subsequent neural pattern is assigned to either the same state as the preceding time step or to a new state that has not yet been visited (Baldassano et al., 2017). HMMs were trained on a version of this Brainiak function that provides better fits when event states are uneven in length (“split_merge=True”’ in Brainiak v0.10). To train the HMMs within each brain functional network, we initially determined the optimal number of states using a nested cross-validation approach. In each iteration of the outer loop of this procedure, a single test subject was selected, while the remaining 35 subjects were divided into 27 for training and 8 for validation. The data from the 27 training subjects were averaged, as were the data from the 8 validation subjects. Within the inner loop of the cross-validation process, various versions of the model were tested, ranging from 1 to 100 HMM states. For each inner loop fold, the model was trained on the averaged data from the 27 training subjects, applied to the averaged data from the 8 validation subjects, and the log-likelihood of the fit to the validation set was computed. Based on the results of the inner loop, we selected the number of states that maximized the log-likelihood of the fit to the validation set. Subsequently, a new HMM was fitted to the withheld “test” subject using this optimal number of states. After fitting the model to the fMRI data collected from couples while viewing movies, we determined the optimal numbers and boundaries of events for seven large-scale brain networks (Fig. 1b). In order to investigate whether the perceived sensitivity to marital-related events is related to marital satisfaction, fitting models were estimated for two groups of married couples with high-level marital satisfaction and low-level marital satisfaction, respectively. We used this method to determine the optimal number of events for each functional network. In these analyses, the neural activity time courses of the different brain regions in each network were used to determine the event boundaries of the activity patterns of that network.

For each subject, we extracted the MVPs at each time point. Event segmentation model results indicated that neural activity patterns in the DMN of married couples with high- and low-level marital satisfaction have different sensitivities to the perception of marital-related events. Therefore, temporal trajectory matrixes of MVPs were then calculated across time points within each of the corresponding defined events of the DMN. For each married couple, we calculated Pearson correlations between subjects’ temporal trajectory matrixes corresponding to the same events (Fig. 1c). Correlation coefficients were then averaged across events for each couple, resulting in a single within-event ISTS measure for each married couple for each brain region. In addition, these analyses were repeated for each of the movie-clips of fMRI stimuli to obtain the within-movie-clips ISTS. Between-group differences (high-level marital satisfaction vs. low-level marital satisfaction) of the whole-brain ISTS in manually defined events and movie clips while viewing movies were assessed by using a two-sample t test (two-tailed). PLS and bootstrapping-based significance tests, as described above, were performed to explore whether and which of the within-events and within-movie-clips ISTS were predictive of marital satisfaction.

This study included 96 participants, comprising 48 heterosexual married couples, all of whom had been married for at least 1 year. The data reported here were also used in our previously published study and were reanalyzed in this study. The initially published paper provided a more detailed description of the data collection procedures (Li et al., 2022). In addition to the fMRI data from couples while they viewed marital and object-related movie clips, resting-state (eyes closed) fMRI data from this cohort of couples were also included in this study. In the fMRI data from movie-viewing, 12 married couples were excluded due to high levels of head motion. In the fMRI data from resting state, none couples were excluded. Therefore, resting-state fMRI data from all 48 couples and fMRI data from the remaining 36 couples while viewing naturalistic movies were used in further analysis.

The originally published paper utilized an ISC approach, which focused on fluctuations in response amplitude and ignored the information contained in distributed spatial patterns of neural activity. In contrast, the present paper focused specifically on spatially distributed response patterns (regardless of their overall magnitude) and, in particular, on individual differences in the evolution of such patterns over time during natural stimulation. For both the resting-state and move-viewing fMRI data, in each of the 200 ROIs, the MVPs were extracted for each time point (Fig. 1a). In each ROI, we calculated pairwise Pearson correlations between MVPs at each time point to construct a time-point-to-time-point matrix that captured the trajectory of the MVPs over time in each subject (Chang et al., 2021; Nastase et al., 2019). Then, ISTS was defined as the Pearson correlation between the vectorized full-pattern trajectory structures of the husband and wife or between randomly selected male and female participants. Random male–female pairs were generated by pairing each male with a randomly selected female (who was not married to him), ensuring no repetitions. To illustrate how the relative similarities in temporal trajectory in each brain region varied as a function of marital status, the ISTS was normalized (i.e., z scored across high-level marital satisfaction vs. low-level marital satisfaction vs. random pairs for each region). The resulting similarity vectors (ISTS-Z scores) for each of the 200 anatomical ROIs were normalized to have a mean of zero and an SD of 1 (Parkinson et al., 2018).

We first examined the relationship between marital satisfaction and trajectories of MVPs throughout the entire study. ANOVA revealed a main effect of marital status (High-level satisfaction couples vs. low-level satisfaction couples vs. random pairs), with stronger ISTS-Z scores in married couples (F = 15.71, P < 0.001). A significant interaction effect was found between fMRI state and marital status (F = 14.23, P < 0.001). Compared with those of random pairs, the ISTS-Z scores were greater in married couples while viewing movies (t = 57.42, P < 0.001, Cohen’s d = 2.37 [married couple with high-level satisfaction]; t = 29.21, P < 0.001, Cohen’s d = 1.41 [married couple with low-level satisfaction]; FDR corrected) and in the resting state (t = 69.16, P < 0.001, Cohen’s d = 2.56 [married couple with high-level satisfaction]; t = 60.41, P < 0.001, Cohen’s d = 2.36 [married couple with low-level satisfaction]; FDR corrected). Moreover, compared with couples with low-level marital satisfaction, married couples with high-level marital satisfaction had higher ISTS-Z scores when viewing movies (t = 21.56, P < 0.001, Cohen’s d = 0.63, FDR-corrected) but not during resting state (t = 1.05, P = 0.15, Cohen’s d = 0.13) (Fig. 2).

Next, we examined whether the ISTS at the whole-brain level was related to individual differences in marital satisfaction. We averaged the ISTS across all 200 ROIs to obtain a whole-brain ISTS in married couples and investigated its relationship with marital satisfaction. We found that the whole-brain ISTS was significantly correlated with marital satisfaction while viewing movie clips (r = 0.42, P = 0.01) (Fig. 3a) but not during the resting state (r = -0.08, P = 0.42) (Fig. 3b). These results demonstrate that a higher-level inter-subject similarity in temporal trajectory is associated with increased marital satisfaction while viewing movies in married couples. Additionally, we found that marriage duration during the resting state was significantly related to ISTS (Spearman correlation, r = 0.31, P = 0.01) (Fig. 3c), while there was no such correlation when viewing the movies (Spearman correlation, r = 0.10, P = 0.54) (Fig. 3d).

Next, we sought to identify the brain regions driving the relationship between the ISTS and marital satisfaction. Partial Least Squares Regression (PLS) regression was used to determine the relationship between marital satisfaction and the ISTS in 200 ROIs during the viewing of movies. The first component of the PLS (PLS-1) was defined as the spatial map that captured the greatest fraction of the total ISTS (P_permutation < 0.001). After bootstrapping, FDR corrections were used to correct for multiple comparisons across brain regions. We found that the ISTS in multiple brain regions contributed to marital satisfaction while viewing movies (P < 0.01; two-tailed, FDR-corrected). These regions were involved in the DAN (Fig. 3e). There were no brain regions in which the ISTS was significantly correlated with marital satisfaction in the resting state.

During fMRI scanning, each participant was presented with two types of movie clips: six depicting scenes related to marital life, and six featuring non-marital natural object-related scenes. Compared with that of random pairs, the ISTS-Z scores were higher in married couples when viewing marital movies (t = 21.16, P < 0.001, Cohen’s d = 2.07 [married couple with high-level satisfaction]; t = 12.35, P < 0.001, Cohen’s d = 1.25 [married couple with low-level satisfaction]; FDR-corrected), and viewing non-marital movies (t = 34.82, P < 0.001, Cohen’s d = 3.26 [married couple with high-level satisfaction]; t = 27.78, P < 0.001, Cohen’s d = 2.61 [married couple with low-level satisfaction]; FDR-corrected). Moreover, compared with couples with low-level marital satisfaction, married couples with high-level marital satisfaction had higher ISTS-Z scores when viewing marital movies (t = 5.78, P < 0.001, Cohen’s d = 0.88, FDR-corrected) but not when viewing non-marital movies (t = 1.46, P = 0.15, Cohen’s d = 0.38) (Fig. 4). These results indicated that married couples with high-level satisfaction showed higher levels of brain-wide neural trajectory similarity than those with low-level satisfaction. Furthermore, these findings demonstrated that this differentiation is specific to marital movies.

Given that married couples process the movie in a similar way to each other, we next investigated whether their neural activity contains evidence of event segmentation. We tested whether the perceived sensitivity to marital-related events was related to marital satisfaction and captured signatures of the segmentation of continuous experience in neural patterns during viewing of movies. A specialized HMM was employed to partition the network-level neural response patterns into a reduced set of continuous, stable activity patterns that align with event structures within the narrative (Baldassano et al., 2017). The number of events that yielded the largest log-likelihood was identified as the optimal number of events for each functional network. The neural patterns were defined from all the regional ROIs in each functional network. We found that the optimal number of events differed across the functional networks. Overall, we replicated previous work showing that visual regions exhibit more events than higher-order associative regions (Yates et al., 2022). However, our results indicated that, compared with couples with low-level satisfaction, couples with high-level satisfaction had more or shorter events in the DAN and DMN (Fig. 5a). Qualitative inspection revealed a notable association between the boundaries in the DAN and multiple types of visual changes in the movie, whereas boundaries in the DMN were specific to marital-related cues. Specifically, compared to couples with low-level marital satisfaction, those with high-level marital satisfaction exhibited more events in the DAN, regardless of whether they were viewing scenes related to marital or non-marital content. However, in regard to the DMN, couples with high-level marital satisfaction experienced more events only when viewing scenes related to marriage (Fig. 5b–c).

The optimal number of events in the DMN differs across married couples with high- and low-level satisfaction, but this does not necessarily mean that the patterns of neural activity are unrelated. The coarser event structure in married couples with low-level satisfaction may still be present in married couples with low-level satisfaction. It may be that some additional additive effect of higher marital satisfaction makes married couples segment these longer events more finely. We thus investigated whether within HMM-denfied events, the ISTS in married individuals was associated with satisfaction. HMM event segmentation models were applied to DMN activity patterns in all married couples. This method was employed to ascertain whether a specific pair of TRs should be categorized as “within-event” (i.e., if both TRs belong to the same event). Then, we examined within-HMM-event ISTS in each of the 200 ROIs for each married couple (Section 2). When analyzing the trajectory structures defined by HMM within events, we observed a pattern of results similar to those in our main analyses. More specifically, compared with couples with low-level marital satisfaction, married couples with high-level marital satisfaction had greater within-event ISTS scores when viewing movies (t = 2.96, P = 0.006) (Fig. 6a). In fact, the fMRI stimuli consisted of 12 movie clips. Each clip has a theme, involving married life, sex, children, food, objects, and architecture. Hence, we delineated each natural segmentation of movie clips as an event and proceeded with an additional exploratory analysis, as outlined in the Section 2, investigating the intra-movie-clips ISTS for each married couple. Compared with couples with low-level marital satisfaction, married couples with high-level marital satisfaction also had greater within-movie-clips ISTS scores when viewing movies (t = 2.02, P = 0.05) (Fig. 6c). However, compared with the within-movie-clips ISTS (Fig. 6d), the within-event ISTS in more extensive brain regions contributed to marital satisfaction while viewing movies (Fig. 6b).

Our prior research examined neural responses to naturalistic stimuli by focusing on fluctuations in overall response magnitude among married couples (Li et al., 2022). The findings from the present study indicate that trajectories of spatially distributed response patterns offer another perspective on individual variations in response to these stimuli. We found that married couples have more similar trajectories of MVPs than randomly selected pairs of couples when viewing movies. Higher ISTS between married couples was associated with greater marital satisfaction. Furthermore, the ISTS of MVPs within manually defined events was predictive of marital satisfaction above and beyond the effects of the ISTS within inherent movie clips. These results suggest that married couples with high levels of marital satisfaction may exhibit significant similarity in the evolution of their psychological states, particularly in relation to social processing and endogenous attention. Our findings underscore that the temporal trajectories of MVPs capture unique individual differences in neural responses, distinct from those captured by temporal fluctuations in regional signals. These results provide strong evidence for the link between couple similarity in terms of trajectories of psychological states and marital satisfaction and extend our understanding of couple similarity in mate selection and relationship outcome.

MVP trajectories effectively disregard the spatial layout of each subject’s data, as individuals are compared based on the relationships between response patterns across time points rather than the direct response patterns themselves (Kriegeskorte & Bandettini, 2007). The current approach, thus, considers spatial distribution of response patterns and how these patterns change over time. In the present study, we found that married couples exhibited greater similarities in the trajectories of MVPs compared to random pairs. Multiple analyses confirmed that couples with high levels of marital satisfaction had higher ISTS. Notably, this effect was specific to dynamic, time-varying response patterns elicited by exposure to naturalistic stimuli. Our results also revealed a significant correlation between marriage duration and ISTS during the resting state, while no such significance was observed during movie viewing. Previous research suggests that cultural contexts modulate resting-state neural similarity, providing a plausible explanation for these findings (Xu et al., 2023). Specifically, the higher resting-state similarity observed in married couples may stem from shared socio-cultural backgrounds and the cumulative effects of long-term social interactions. In contrast, the heightened neural similarity observed during movie viewing in couples with high marital satisfaction likely reflects mate preferences and shared emotional or cognitive resonance in response to marital-themed stimuli. These findings collectively highlight the complex interplay between socio-cultural factors, long-term interactions, and individual preferences in shaping neural synchrony within close relationships.

In addition to ISTS measures at the whole-brain level, our analyses identified specific brain areas predictive of marital satisfaction. The most significant effects were localized in regions of the DAN, which is associated with endogenously driven shifts in attention. DAN areas are also linked with episodic memory retrieval, especially when attention is allocated to memories (Hutchinson et al., 2009). Given that all participants were exposed to the same stimuli, the way in which exogenously determined attentional states fluctuate over time may not differ significantly. Therefore, our main findings may reflect how married couples allocate their attention to specific stimuli or scenes based on their similar goals or memories when viewing video clips. These results suggest that comparing perceivers based on changes in the trajectory of MVPs over time, rather than solely on fluctuations in response amplitude, may offer a promising approach for capturing similarities between subjects (Haxby et al., 2014; Norman et al., 2006).

Next, we investigated neural event segmentation using a data-driven, computational approach in married couples with high and low levels of marital satisfaction while they viewed movies. We found synchronous processing of the movie and a reliable event structure in both groups. However, married couples with low marital satisfaction exhibited coarse neural event structures across regions in the DAN and DMN. In contrast, compared to those with low marital satisfaction, movies evoked more and shorter events in the DMN among couples with high marital satisfaction. This could be due to a tendency to segment events according to goals during encoding (Pérez et al., 2021).

Marital movie scenes are rich, engaging, and energetic, aligning well with the mental processes involved in partners’ daily interactions, such as communication, sex, attitudes toward relatives, conflict resolution, financial disputes, and shared values (Hasson et al., 2010; Li et al., 2022). Consequently, event boundaries detected in the brain may provide more “objective” insights into the similarity of couples’ mental processes as they experience and react to the world around them. We then compared the neural response patterns of psychological states to those of corresponding events across groups. We conducted an ISTS analysis, where ISTS were summarized by HMM-identified events in the DMN rather than by TRs. This finding contrasts with results from similar analyses relating marital satisfaction to ISTS across the whole experiment. Compared to ISTS within inherent segmentation of movie clips, ISTS within HMM-identified events were more predictive of marital satisfaction. This could be attributed to the association between marital satisfaction and similarities in spouses’ personality traits and previous marital experiences. This association may lead couples to perceive stimuli with similar goals, knowledge, memories, and expectations, thereby influencing which aspects of the stimuli are deemed relevant or interesting at any given time (Hyon et al., 2020).

Notably, the DMN is a functionally heterogeneous, large-scale brain network involved in a range of internal mental processes, including internally directed attention, self-referential thought, and social cognition (Menon, 2023). It also plays a crucial role in predicting individuals’ current and future mental states (Thornton et al., 2019). Innovative studies employing naturalistic stimuli have demonstrated that DMN activity synchronizes when two individuals process a shared narrative, with this synchronization being particularly sensitive to social and communicative cues (Yeshurun et al., 2021). Our previous research also suggested that neural synchronization within the DMN is associated with marital satisfaction in couples (Li et al., 2022). In this context, inter-subject synchronization within the DMN may reflect similar mental representations, such as shared values, beliefs, and comparable perceptual “styles” of the world (Menon, 2023). Previous work has proposed that these shared representations within the DMN may arise from dynamic social interactions between the self and others. Through reciprocal and interactive exchanges, the brain responses of one individual can influence those of another, and vice versa, thereby creating a shared cognitive reality (Yeshurun et al., 2021). For married couples, high neural alignment within the DMN may thus result from ongoing social interaction and coordination during day-to-day activities, such as communication, sexual interactions, managing financial disputes, resolving conflicts, and discussing shared values and attitudes toward relatives. These interactions foster continuous synchronization of mental states over time. The current findings build upon and extend these observations, suggesting that married couples not only share neural synchrony but also exhibit temporal similarities in their psychological states.

In summary, by analyzing the trajectory of distributed response patterns within brain regions, rather than considering only the overall response magnitude, a wealth of information about mental states would be detected. We investigated the association between ISTS to naturalistic audiovisual stimuli and marital satisfaction in married couples. Compared to randomly selected pairs of couples, married couples showed significantly higher levels of ISTS during viewing of movies, and ISTS in DMN/DAN between married couples predicted higher levels of marital satisfaction. Therefore, married couples may undergo a comparable ebb and flow of internally driven mental states throughout naturalistic stimulation. This phenomenon could stem from similarities in individuals’ expectations, interests, and predispositions related to marital life, which subsequently shape their allocation of attention to various elements of external stimuli and their internal values, beliefs, and memories during the viewing of naturalistic stimuli.

However, we cannot draw conclusions from the current data on whether neural similarity is a cause or consequence of successful marriage. Thus, longitudinal studies should be adopted in the future to measure the important role of couple similarity in the neural response in mate selection and relationship maintenance and outcomes. Moreover, although the naturalistic neuroimaging paradigm of this study has many advantages, a more detailed understanding of the cognitive and emotional processes underlying these effects may require additional future studies involving behavioral measurements and more constrained experimental paradigms.

Anonymized fMRI, demographic and behavioral data, and code used for analyses have been deposited in a Zenodo under (https://zenodo.org/records/11377487).

L.L., H.C., and X.D. designed research; L.L. and X.H. performed research; L.L., X.H., J.X., Q.Z., X.S., H.C., and X.D. contributed new reagents/analytic tools; L.L. analyzed data; L.L. and X.D. wrote the paper; and H.C. and X.D. lead the project.

The authors declare no competing interest.

This study was supported by the National Natural Science Foundation of China (62273076, 82121003, and 62036003), Fundamental Research Funds for Central Universities (ZYGX2019Z017), Innovation Team and Talents Cultivation Program of National Administration of Traditional Chinese Medicine (ZYYCXTD-D-202003), and the National Social Science Foundation of China (20&ZD296).