After the Honeymoon: Neural and Genetic Correlates of Romantic Love in Newlywed Marriages

Participants

Participants provided informed consent in accordance with the IRB procedures of the University of California, Santa Barbara (UCSB) and Albert Einstein College of Medicine. Subjects were recruited via advertisements, flyers, listservs, and word of mouth. Before undergoing scanning, all participants were interviewed to assess eligibility criteria such as first-time marriage for both partners, no children for either partner, relatively good health, and fMRI contraindications (right-handedness, good health, no metal in or on the body, no claustrophobia, no pregnancy, and no history of head trauma). The eligibility criteria were selected to reduce variability of the sample since this was the first study to examine the physiology underlying changes in romantic love in newlywed marriages. We selected individuals in first-time marriages with no children to mirror animal studies where monogamous mammals solidify pair-bonds prior to producing offspring. All procedures were described at the time of the interview.

The resulting sample was composed of 19 (11 women and 8 men) healthy, right-handed individuals, ages 21 to 32 (27.21 ± 3.29 years), in established relationships (4.11 ± 3.09 years), without children, and living with their partner about 2 years (1.9 ± 1.6 years). At baseline (T1), some participants were recently married (10 married, 1.9 ± 1.5 months), while others were soon to be married (2.6 ± 1.7 months until the wedding). The sample of participants were mostly college-educated: 11 participants had college degrees and 6 had a master’s degree or higher, while only 2 had a high-school level education. The mean annual household income of the sample was $62,000 (±$28,000, range $16,000 to $110,000).

Thirteen (seven females and six males) of the 19 participants returned for a second scan (T2), approximately 11.3 months (SD ± 1.3, range 9.0–13.5) after the initial scan (T1). Herein, we report findings that were shown at both measurements (T1 and T2) among the group of 13 participants that were scanned twice.

Procedure

Once eligibility was confirmed, participants provided the experimental stimuli: face images of their partner and a highly familiar neutral acquaintance (HFN). The HFN served as a control for facial familiarity and was matched to the partner by age, sex, and length of time known. The partner-HFN face viewing task has been used in prior fMRI studies of romantic love (e.g., Aron et al., 2005; Xu et al., 2011) and was originally developed in a study showing that images of partner faces, compared to other types of stimuli (i.e., songs and scents), elicited the most intense love feelings among individuals in love (Mashek et al., 2000). All photos were digitized according to standard procedures where only the face was presented, and they were shown with Presentation software (Psychological Software Tools, Inc., Pittsburgh, PA, United States) during the scan.

For the fMRI task, participants viewed alternating face images of the partner and the HFN (shown individually) interspersed with a countback task (displayed individually, for 20 s each). For the countback task (used to reduce carry-over effects of viewing the facial images), subjects were asked to mentally count backwards in increments of seven, starting with a random four-digit number displayed on the screen. The entire session lasted for 12 min, and stimuli (images and the countback task) were displayed for 20 s each. At the start of the session, participants were instructed to recall non-sexual events with the person whose face image would be displayed on the screen. After the scan, participants provided affective ratings while still in the scanner to verify that the evoked emotion corresponded to the target image (see Acevedo et al., 2014 for results of the affective ratings). Identical procedures, questionnaires, and photos were used at T1 and T2.

Imaging Data Acquisition and Analysis

A 3.0 T Siemens Trio with a 12-channel phased-array head coil was used for the acquisition of blood oxygenation level dependent (BOLD) responses. A single-shot echo planar imaging sequence sensitive to BOLD contrast was used to acquire 37 slices per repetition time (TR = 2000 ms, 3 mm thickness, 0.5 mm gap), with an echo time (TE) of 30 ms, flip angle of 90 degrees, field of view (FOV) of 192 mm, and 64 × 64 acquisition matrix. Prior to the acquisition of BOLD responses, a high-resolution T1-weighted sagittal sequence image of the whole brain was obtained (TR = 15.0 ms; TE = 4.2 ms; flip angle = 9 degrees, 3D acquisition, FOV = 256 mm; slice thickness = 0.89 mm, acquisition matrix = 256 × 256).

Descriptive Statistics

Descriptive statistics are reported in Table 1. The mean change in Eros scores from T1 to T2 (M = −0.13 ± 0.89, range = −3.00 to +0.50) was not statistically significant. Specifically, 75% of the sample showed increases of less than a point, 25% showed no change, and 25% showed decreases of less than a point in Eros scores. Thus, the majority of the sample reported romantic love maintenance. Only one participant showed a steep decrease (−3.00 points) in romantic love over the first year of marriage. Thus, we examined the data without the outlier. However, the brain imaging correlations did not change significantly, including the OXTR, AVPR1a, DRD4, and COMT interactions with romantic love maintenance (Eros T2-T1). Additionally, activation of the VTA in response to images of the partner remained positive, but in some cases became non-significant, when the outlier was excluded. Thus, we proceeded with analyses including the outlier because variable values make these results more generalizable to the population. That is, it is expected that some couples will experience steep decreases in romantic love in the early stages of marriage as shown by research reporting “honeymoon effects” (e.g., Huston et al., 2001).

Correlations Among Variables

Correlations among self-report measures are reported in Table 2. At each time point, romantic love was significantly correlated with frequency of sexual activity (T1: r = 0.55, p < 0.05; T2: r = 0.50, p < 0.05). Romantic love was also strongly correlated with relationship satisfaction: at T1 relationship satisfaction predicted romantic love at T2 (r = 0.65, p < 0.05), and at T2 relationship satisfaction was correlated with romantic love at T2 (r = 0.73, p < 0.01). Gene correlations showed that AVPR rs3 (long alleles) was significantly correlated with relationship satisfaction at T1 (r = 0.58, p < 0.05) and with romantic love at T2 (r = 0.46, p < 0.05). Also, both dopamine polymorphisms, DRD4-7R (r = −0.81, p < 0.01) and COMT rs4680 (r = −0.50, p < 0.10), were negatively correlated with romantic love scores at T1.

Neuroimaging Results

Sex and Romantic Love Maintenance

Other notable findings shown in the present group of newlyweds in association with romantic love maintenance emerged in regions important for sex and sensory processes (the PCL and sensory cortex). Interestingly, the PCL is the genital sensorimotor region activated in women during orgasm and clitoral stimulation (Wise et al., 2016, 2017). Activation of the PCL in the present study is interesting because the primary sensory cortex (SI) usually requires direct tactile stimulation to activate it. There was no stimulation of the genitals in this study, and participants were instructed not to think about sexual memories. The traditional textbook understanding of SI function does not include memory, emotion, or person representation, only sensory processing features like pressure. However, studies in recent years suggest that SI may contain memory capacity and a genetically controlled mechanism for cortical memory (Bancroft et al., 2014; Kragel and LaBar, 2016; Muckli and Petro, 2017; Galvez-Pol et al., 2018; Kim et al., 2018). Finally, there is substantial evidence that a memory code for persons and traits is active in the ventral medial prefrontal cortex, while other cortical areas are also involved in the mental representation of a person (Heleven and Van Overwalle, 2016; Thornton and Mitchell, 2018). We speculate that looking at the face of the marriage partner and thinking romantic thoughts might activate the mental representation of that person, as faces have in other studies (e.g., Thornton and Mitchell, 2018). Additionally, it’s likely that that engaging in sexual acts with the same partner over time would activate genital sensory cortex memory storage mechanisms that importantly become part of the mental representation of the partner.

Although many therapists have suggested an important role for sexual activity in maintaining a marital relationship, this is the first time a cortical brain region associated with direct sexual stimulation has been correlated with self-reports of romantic love in marriages while simply thinking (and viewing face images) of a spouse. Further support for the importance of sex emerged from the robust correlations between romantic love scores and sexual frequency and satisfaction ratings at both time points (see Table 2).

AVPR1a and Romantic Love Maintenance

Interestingly, the present sample of newlyweds showed significant interactions with AVPR1a rs3 and romantic love maintenance, at both time points, in the right VTA (Figures 2A–C), the PAG, posterior hippocampus, occipital cortex and the superior temporal gyrus (STG)— regions important for reward, attachment, memory, and visual and sensory processing (Nagy et al., 2012; for a meta-analysis, see Phan et al., 2002; Schultz et al., 2003; Wager et al., 2003). Most of these regions have appeared in the context of long-term romantic love and maternal love (e.g., Acevedo et al., 2011; for review see Bartels and Zeki, 2004; Acevedo, 2015), highlighting the role of attachment in sustained romantic love among newlywed pair-bonders. They are also consistent with research implicating AVPR1a in pair-bonding (Walum et al., 2008) and suggest the diversity of the pair-bonding system through its engagement of reward, memory, sensory, visual, and auditory functions.

OXTR and Romantic Love Maintenance

The pattern of replicated interactions for OXTR rs53576 with romantic love maintenance were different from AVPR1a effects, appearing in the septum (bilaterally) and the left (L) VTA. Activation of L VTA has mostly appeared in studies of facial attractiveness, specifically showing response to smiling and supportive faces (Vrticka et al., 2008). Also, L VTA activation was shown in a study of males given intranasal OT in response to viewing facial images of their female partner (Scheele et al., 2013). Interestingly, when given OT males rated their partners as more attractive, but OT did not affect attractiveness ratings for a familiar matched control. These findings suggest that OT-related effects are partner-specific, thus facilitating attachment and pair-bond solidification. It should be noted that although we did not test for sex differences, sex may influence how OT affects mate choice and pair-bonding (Xu et al., 2020). Also, individual differences, such as personality and attachment style, may influence how OT interacts with pair-bonding choices (Pearce et al., 2019).

Activation of the septum—which is rich in binding sites for OT and, to a lesser extent, AVP—is consistent with animal studies showing that the septum is critical for pair-bond establishment (Liu et al., 2001). In humans, activation of the septum has been implicated in early-stage and long-term romantic love (Aron et al., 2005; Acevedo et al., 2011), and it was specifically associated with obsession-related items of the Passionate Love Scale (Hatfield and Sprecher, 1986) among long-term pair-bonders. We add to this body of work, showing OT’s effects in romantic love maintenance.

Dopamine Gene Polymorphisms (DRD4 and COMT) and Romantic Love Maintenance

Robust neural activations were positively correlated with romantic love maintenance and dopamine polymorphisms (DRD4-7R and COMT rs4680) in the L VTA/SN region and the posterior insular cortex at both time points. As noted above for the OXTR findings, the L VTA is specifically activated in response to facial attractiveness (e.g., Aron et al., 2005; Liang et al., 2010). Interestingly, the DRD4-7R genetic polymorphism, which is associated with reduced binding for dopamine and greater novelty seeking, was negatively correlated with romantic love scores at T1. Individuals with the 7R allele show higher rates of promiscuity and novelty seeking (He et al., 2018, meta-analysis; Munafo et al., 2008, meta-analysis; Garcia et al., 2010). Thus, it is not surprising that in the present study, individuals with the DRD4-7R variant showed lower romantic love scores but higher activation in the L VTA, where facial attractiveness promotes activation (Aharon et al., 2001; Winston et al., 2007). Dopamine-related gene expression (COMT and DRD4) in the L VTA suggests that facial attractiveness, reward, and more generally attraction mechanisms may be fruitful areas of examination for future research on sustaining romantic love in marriages.

COMT and DRD4 also showed significant interactions in the insular cortex which is involved in a variety of functions including reward, emotion, social bonding, sensory processing, and self-awareness (for review see Gogolla, 2017). Specifically, the posterior insular cortex area where DRD4-effects were shown, is implicated in social support networks in elderly individuals (Cotton et al., 2019), making this an interesting region for future investigations of relationships. The human insula has also become a target for treatment in a variety of disorders such as substance abuse, depression, anxiety disorders, schizophrenia, and autism. Specifically, dopamine and opioid receptors in the insular cortex are thought to influence addiction (Ibrahim et al., 2019). Our findings highlight the role of attachment in sustained romantic love and are consistent with theories suggesting that romantic love is a “natural addiction” (Fischer et al., 2016).

Collectively, activation of the insula and other regions identified here (e.g., the STG, occipital area, hippocampus, PCL involved in sensory processing) are consistent with the idea that romantic love is an emergent property of pair-bonding whereby multi-sensory information is translated into processes such as communication, empathy, and decision-making as well as complex cognitive processes such as imagining a future together (Walum and Young, 2018). Thus, basic reward, sensory and higher-order cortical processes and their intersections, as exemplified herein, are critical for the maintenance of romantic love in established pair-bonds.

Deactivations Associated With Romantic Love Maintenance

Deactivations emerged in association with romantic love maintenance, at both time points, in the inferior frontal gyrus (IFG) and the temporal gyrus. The IFG plays an evaluative role in multisensory stimuli and may be deactivated when evaluative processes are not engaged (Ethofer et al., 2006; Schirmer and Kotz, 2006). These results are consistent with previous brain imaging studies suggesting that in romantic love, suspension of negative judgment occurs, coinciding with deactivation in the temporal lobe (Zeki, 2007; Xu et al., 2011; Wang et al., 2020). IFG deactivation has also been associated with impairment in stopping a task once initiated (Chambers et al., 2006), consistent with the persistence of romantic love in the present sample.

Future Directions and Limitations

Although this is the first study to provide evidence of the neural and genetic correlates of romantic love maintenance in a sample of newlyweds, it is important to recognize some limitations. The major limitation of this study is the small sample size. Although Friston et al. (2013) argued that small samples have advantages, small sample sizes and low statistical power may contribute to inflated effect sizes. Many of the effect sizes reported here were moderate to large, but with a larger sample in future studies effect sizes may be smaller. However, we relied on region-of-interest analyses and predicted/planned comparisons, which reduces possible statistical errors. Most importantly, many of our key findings were replicated. Replication is the most important statistical procedure for reliability of a result, and many of our key findings were shown at both time points. Also, a strength and a limitation of the study is the homogeneity of the sample.

Constraining the variables in the sample is important for the reproducibility of the results, but renders the results applicable to a limited population. Thus, generalizability of results is another limitation, as the participants in the present study were mostly well-adjusted, in-love, and highly satisfied with their relationship partner at T1. Nevertheless, this group of newlyweds experienced common marital concerns including the balancing of dual careers, managing domestic chores, and financial issues (Lavner and Bradbury, 2010). Also, in line with theories of “honeymoon effects,” one participant showed a steep decline in marital satisfaction and romantic love over the first year of marriage. Thus, it will be critical for future research to recruit larger and more diverse samples to capture the full range of relationship trajectories and pair-bonding strategies.

Another issue was that although we replicated many key findings at T1 and T2, other effects emerged separately for T1 and T2. For example, romantic love at T1 showed significant correlations in regions that are rich in serotonin (raphe and pons), while at T2 the patterns of neural activation were more robust in regions associated with emotion processing and rich in opioid receptors (amygdala and GP). Indeed, there were differences after 1 year of marriage that might indicate changes in attitude toward the partner, additional experiences with partner, envisioning a future together with children, conflict, and general life experiences. However, we refrained from speculating on what these differences in activation (and deactivation) might represent, but we did include the results so that future studies with larger samples (that may use different statistical approaches) may form hypotheses and determine if the neural mechanisms underlying romantic love, and its maintenance, change in consistent ways as a function of time.

To this point, it will be important for future studies on the biological basis of pair-bonding and romantic love to recruit couples with diverse levels of relationship quality. Relationship studies are often biased with positive couples because distressed/conflicted couples are more difficult to recruit as romantic partners often feel uncomfortable disclosing negative thoughts, sentiments, and doubts about their relationship. Additionally, social desirability effects may be especially strong around the time of the wedding; thus, appropriate measurement and objective markers are important for capturing couples that may be particularly vulnerable to conflict and sharp decreases in relationship quality.

In the current study we focused on four genes that have been implicated in social behaviors, including pair-bonding. There is a strong empirical basis for examining the particular genetic polymorphisms. For example, the dopamine receptor variant DRD4-7R, which we assessed in the present study, is associated with reproductive sexual behaviors (Eisenberg et al., 2007), desire for a wider variety of sexual behaviors (Halley et al., 2016), and higher rates of promiscuous behavior and infidelity (Garcia et al., 2010). However, there are other possibilities to explore. It will be critical for future research to examine a wider array of genetic polymorphisms underlying pair-bonding with larger samples, both with genome-wide association studies (GWAS) and more directed approaches with predicted polymorphisms. It has been shown that in many cases single genes have very small effect sizes (for review, see Fox and Beevers, 2016). However, GWAS studies are limited in that they require very large sample sizes (Landefeld et al., 2018). Other important genetic variants may also be examined in future studies, for example, the 5-HTTLPR VNTR of the serotonin transporter gene that has been associated with differences in life history strategy and risk-acceptance in mating competition (Minkov and Bond, 2015). Such findings are linked to a broader framework of life history theory than we investigated here, but they are relevant to variation in human mating and pair-bonding strategies (Minkov and Bond, 2015; Pearce et al., 2019).

Also, it is important to note that although identifying biological markers for pair-bonding in group studies is helpful, individual differences must be accounted for. For example, in recent years OT has received significant attention for strengthening pair-bonds (e.g., Quintana et al., 2019). However, responses to OT may vary according to some oxytocin genetic polymorphisms and gender (e.g., Pearce et al., 2019; Xu et al., 2020). Also, results from brain imaging studies indicate that oxytocin genetic variants may influence couples’ sociosexual feelings, sexual behaviors, and intimacy (e.g., Acevedo et al., 2019a, b; Pearce et al., 2019).

Finally, although this study is the first to report neural and genetic mechanisms underlying changes in romantic love in first-time newlyweds, it would be beneficial for future neurobiological studies to expand measurements beyond the first year of marriage. This might capture important changes that occur over marital development such as the addition of offspring, career transitions, and increased interdependence that is an inevitable aspect of marriage.