Relationships at risk: How the perceived risk of ending a romantic relationship influences the intensity of romantic affect and relationship commitment
Drawing on emotional intensity theory (EIT: Brehm in Personality and Social Psychology Review 3:2–22, 1999; Brehm and Miron in Motivation and Emotion 30:13–30, 2006), this experiment (N = 104) shows how the manipulated risk of ending a romantic relationship influences the intensity of romantic affect and commitment. As predicted by EIT, the intensity of both romantic feelings varied as a cubic function of increasing levels of manipulated risk of relationship breakup (risk not mentioned vs. low vs. moderate vs. high). Data additionally showed that the effects of manipulated risk on romantic commitment were fully mediated by feelings of romantic affect. These findings complement and extend prior research on romantic feelings (Miron et al. in Motivation and Emotion 33:261–276, 2009; Miron et al. in Journal of Relationships Research 3:67–80, 2012) (a) by highlighting the barrier-like properties of manipulated risk of relationship breakup and its causal role in shaping romantic feelings, and (b) by suggesting that any obstacle can systematically control—thus, either reduce or enhance—the intensity of romantic feelings to the extent that such obstacles are perceived as ‘risky’ for the fate of the relationship.
KeywordsEmotional intensity Deterrence Perceived risk of relationship breakup Romantic affect Relationship commitment Goals Motivation Emotions Paradoxical effects
Does the perceived risk of ending a romantic relationship influence the intensity of romantic affect and relationship commitment towards the romantic partner? And, if yes, how does it happen? Drawing on motivational and emotional intensity theories (MIT: Brehm 1975, 1999; Brehm and Self 1989; Brehm et al. 1983; EIT: Brehm and Brummett 1998; Brehm and Miron 2006), this article tries (1) to fill an important knowledge gap on the motivational role that perceived risk of romantic relationship breakup may play in romantic relationships, and (2) to complement the prognostic role traditionally attributed to perceived and actual risk factors, by exploring the barrier-like properties of perceived risk, and its paradoxical effects on romantic feelings.
In the context of romantic relationships, research has focused on a multitude of objective risk factors held responsible for relationship breakup, such as pre-engagement cohabitation (Kline et al. 2004), teenage relationships resulting in early marriage (Karney and Bradbury 1995), time spent together (Felmlee et al. 1990), education, years of marriage, or number of children (Gager and Sanchez 2003), and even genetic factors (McGue and Lykken 1992)—all representing actual risk factors for ending a romantic relationship.
Only a few studies have focused, by contrast, on perceived risk in romantic relationships. A prominent example of research and theorizing in this area is represented by the risk regulation model (RRM: Murray et al. 2008, 2006), a theoretical framework that starts with the anticipated fear of being rejected or disregarded by the romantic partner, in a context of mutual interdependence. According to the risk regulation model (RRM), people regulate the risk of being hurt by the partner by adjusting their cognitions and behaviors according to a ‘connectedness’ vs. ‘self-protection’ motive. In this model, special emphasis is given to the role that individual differences (e.g., low ‘self-esteem’, high ‘attachment insecurity’) play in such adjustments (Meuwly and Schoebi 2017, for a review).
In addition, in an attempt to link perceived risk to lessened relationship stability, research in this area has also shown that individuals who fear betrayal or rejection in romantic relationships tend to perceive comparatively poorer relationship quality (Brunell et al. 2007), and also to feel less attracted to their romantic partner (Fishbein et al. 2004). Analogous findings stem from research on the link between perceived conflict—which represents only a potential source of perceived risk—and dissatisfaction with the partner (Brassard et al. 2009; Campbell et al. 2005). Further, Gager & Sanchez (2003) reported an explicit attempt to connect the perceived chances of divorce/separation to actual relationship breakup. This is the only study that explicitly quantifies the perceived risk of ending a romantic relationship.
However, no study has yet investigated the direct effects of perceived risk of relationship breakup on romantic feelings of attraction and commitment in romantic relationships. In our view, a focus on feelings of romantic attraction, rather than almost exclusively on cognitive and behavioral reactions to perceived risk (e.g., the RRM: Murray et al. 2006, 2008), is pivotal for understanding the psychological processes that lead to emotional disengagement and, in turn, to actual relationship breakup. Surprisingly, no research has yet systematically varied—over multiple degrees of intensity—the magnitude of perceived risk of relationship breakup in order to observe the resulting effects on romantic feelings. Further, no study has yet considered, and empirically tested, the hypothesis that the perceived risk of ending a romantic relationship would also produce positive effects on romantic feelings, owing to its barrier-like properties.
The motivational role of barriers in enhancing and reducing romantic affect
Pioneering work by Ach (1910) and Hillgruber (1912), and subsequent work by H. F. Wright (1937), laid the theoretical and empirical foundations for understanding the role of task difficulty, task demand, and barriers in enhancing—rather than reducing—the strength of motivation. Together with further considerations and empirical findings, these early observations and discoveries converged into the theories of motivational (Brehm 1975; Brehm and Self 1989; Brehm et al. 1983; for reviews: Richter et al. 2016) and, later, emotional intensity (Brehm 1999; Brehm and Brummett 1998; Brehm and Miron 2006).
On the basis of a functional analogy established in the way in which motivational, emotional, and affective states operate (Brehm 1999; Brehm and Brummett 1998; Brehm and Miron 2006; Brehm et al. 2009; Fuegen and Brehm 2004; R. A. Wright 2011), emotional intensity theory (EIT) asserts that any obstacle, impediment or, more generally, any potentially obstructing-, resisting-, inhibiting-, or counter-force that interferes either with the experience or with the expression of a given emotion, has barrier-like properties—it systematically modifies (i.e., strengthens or weakens) felt emotional intensity. Such obstacles to emotional intensity are called deterrents (Brehm 1999; Brehm and Brummett 1998; Brehm and Miron 2006; Brehm et al. 2009; see also Silvia and Brehm 2001).
According to EIT, deterrents systematically produce paradoxical effects on emotions (e.g., as when we feel increased sadness when a friend tries to comfort us—or after receiving a gift certificate, as in Brehm et al. 1999), because the greater the deterrent (i.e., the counterforce), the stronger becomes the emotion, up to the point where strength of deterrence outweighs the importance of the reason that originally instigated the emotion, this causing the emotion to drop to a minimum level of intensity. Further, when deterrents are ‘unknown’ (e.g., a person is not aware of their existence and/or oppositional strength), according to EIT emotional intensity will attain its maximum strength. In such a case, the strength of the emotional feeling (actual emotional intensity) will closely mirror the subjective importance (potential emotional intensity) of the instigating event. Emotional intensity will thus be (1) relatively strong in the face of unknown deterrents or counterforces; (2) substantially reduced in the presence of relatively weak deterrents or counterforces; (3) relatively strong in the face of moderately strong deterrents or counterforces; (4) substantially reduced in the presence of too strong (i.e., outweighing) deterrents or counterforces. If considered in quantitative terms, this overall pattern obeys a cubic function. The cubic pattern predicted by EIT has been abundantly substantiated in research on happiness, anxiety, positive and negative (sensory) affect, prejudiced and attitude-related affect, anger, love, romantic interest, enjoyment, strength of affective social identification, collective guilt, and vicarious empathy (Brehm and Miron 2006; Brehm et al. 1999, 2009; Dill 1997; Fuegen and Brehm 2004; Miron et al. 2008, 2011, 2009, 2007; Pantaleo 2011; Pantaleo et al. 2014; Reysen and Katzarska-Miller 2013; Schmitt et al. 2008; Silvia and Brehm 2001). One study by Brehm and colleagues (Brehm et al. 2009—Study 2), in particular, provided strong support for the predicted cubic effect across six distinct adjacent points of deterrence to the intensity of sensory negative affect (including the no-deterrent control condition).
A notable implication of the above reasoning is that, also in the field of romantic relationships, any obstacle to feelings of attraction and/or commitment toward the romantic partner should—by virtue of its barrier-like properties—either strengthen or weaken such feelings. Indeed, a closer look at the research findings in this domain reveals that the pertinent studies have produced a series of mixed findings—objective factors that interfered with the stability of the relationship turned out to be sometimes detrimental and sometimes paradoxically beneficial to the relationship. Whereas general social disapproval, lack of social support, or parental interference have been mostly shown to damage the relationship in the long run (e.g., Felmlee 2001; Sinclair and Ellithorpe 2014; Sinclair et al. 2014; see also Levinger 1999), in other instances the same objective interfering forces seemed, by contrast, to sustain and benefit the couple. A moderate degree of parental interference (Driscoll et al. 1972; see also Driscoll 2014), a certain amount of social disapproval (Felmlee 2001), a certain opposition and lack of support by the family (Parks et al. 1983; Sprecher 2011), as well as romantic involvement in marginalized relationships (e.g., same sex, interracial, or age-gap relationships, Lehmiller and Agnew 2006), were all associated with enhanced, rather than reduced, attraction and commitment toward the romantic partner.
Recently, a series of experiments explicitly designed to test the paradoxical effects that obstacles to romantic feelings may have on such feelings of attraction and commitment, predicted and documented both affect-enhancing and affect-reducing effects (Miron et al. 2009, 2012; Reysen and Katzarska-Miller 2013). In testing EIT in the domain of romantic relationships, these deterrence experiments not only provided further support for the theory, but also a straightforward and coherent theoretical explanation for the above ‘mixed’, unexpected, and—perhaps for some—even somewhat puzzling results (e.g., Driscoll et al. 1972; Driscoll 2014; Felmlee 2001; Lehmiller and Agnew 2006; Sinclair and Ellithorpe 2014): By acting as deterrents, obstacles to romantic feelings seem able to either reduce or enhance, in non-obvious ways, the intensity of attraction and commitment towards the romantic partner.
To the extent that the perceived risk of ending a romantic relationship can be also assumed to have barrier-like properties, then it should likewise act as a deterrent to romantic feelings and, hence, systematically either reduce or enhance feelings of attraction towards the romantic partner (Miron et al. 2009, 2012; see also Reysen and Katzarska-Miller 2013; Wright et al. 1985; Roberson and Wright 1994, on interpersonal attraction as determined by the difficulty of establishing a relationship with a potential date). Further, as romantic affect has also been demonstrated to instigate an urge to stay with the partner (Gonzaga et al. 2001; Miron et al. 2009, 2012), we would similarly expect the same deterrence effects of perceived risk of breakup on romantic commitment, as mediated by strength of romantic affect (Miron et al. 2009, 2012).
The present research
The present research investigated the paradoxical influence of the perceived risk of ending a romantic relationship on the intensity of romantic affect and relationship commitment, by exploring the barrier-like properties of perceived risk of breakup. In so doing, the study intended to fill an important knowledge gap on the motivational role that the (manipulated) perception of risk may play in romantic relationships in general. More specifically, romantic feelings of affect and commitment were first instigated and then deterred, across four distinct and progressively increasing levels of perceived risk of breakup (risk not mentioned vs. low vs. moderate vs. high).
Surprisingly, no study has yet accomplished such a perceived risk manipulation, and no study has done it across the whole range of risk intensity. The only study that came closest to this is an experimental investigation by Miron et al. (2009, Study 2), in which the researchers intentionally manipulated the perception of the partner’s negative characteristics by varying the severity of the expected consequences of such characteristics on the romantic relationship. Miron and her colleagues informed participants that their relationship was going to have either ‘only minor’ versus ‘moderately serious’ problems because of those negative partner’s characteristics. Though decisive in understanding the general terms of the problem, this study cannot answer, however, our specific question about the cubic effects that perceived risk of relationship breakup is expected to have on romantic feelings on the basis of EIT, for two reasons. For one, the authors did not explicitly manipulate the perceived risk of ending the relationship, as the perception of partner’s negative characteristics may only potentially result in the perceived risk of relationship breakup. Second, they manipulated the perception of partner’s negative characteristics across only three levels of intensity (no information vs. low deterrence vs. moderate deterrence)—i.e., an insufficient number of levels to observe the overall predicted cubic trend.
Our experimental manipulation of the intensity of perceived risk of relationship breakup was accomplished in order to test three hypotheses. Drawing on EIT (Brehm 1999; Brehm and Brummett 1998; Brehm and Miron 2006), we first expected romantic affect to vary as a cubic function of increasing levels of manipulated perceived risk of breakup (risk not mentioned vs. low vs. moderate vs. high). More specifically, the intensity of romantic affect was expected to be (1) relatively strong in the face of an unknown level of risk of relationship breakup; (2) substantially reduced in the presence of a relatively low risk of breakup; (3) relatively strong in the face of a moderate risk of breakup; and (4) considerably reduced in the presence of a too high risk of breakup. Second, we also expected commitment to follow the same cubic pattern of results—i.e., to parallel the effects of perceived risk of relationship breakup on romantic affect. Third, we finally expected romantic affect to mediate the effect of perceived risk of breakup on relationship commitment.
The above hypotheses need some clarification. We predicted a substantial drop in the intensity of romantic feelings both in the ‘low’ and ‘high’ deterrence conditions. This drop is currently predicted and explained by the ‘resource conservation principle’ invoked by MIT (Brehm and Self 1989; see also Richter 2013; Richter et al. 2016; Wright 2008, 2011, 2016) and—by virtue of the functional analogy between motivational, emotional, and affective states (e.g., Brehm et al. 2009)—also by EIT (Brehm 1999; Brehm and Brummett 1998; Brehm and Miron 2006). Note, however, that recent experimental work by Richter and colleagues (Richter 2015; Stanek and Richter 2016) challenged, both theoretically and empirically, the primacy of this principle by documenting that (a) though energy expenditure tends to increase with increasing task demand, actual energy expenditure is systematically higher than strictly theoretically required, and (b) the (relatively high) amount of energy observed in empirical studies in the ‘high difficulty’/‘impossible task’ conditions does not correspond to the substantial disengagement predicted by MIT. Such a challenge, however, does neither deny nor logically contradict MIT’s and, by analogy, EIT’s basic predictions that affect will be comparatively lower in the low and high (vs. control and moderate) deterrence conditions. For any practical research purpose, therefore, our general prediction of an overall cubic trend in the intensity of romantic affect and romantic commitment as a function of perceived risk of relationship breakup (risk not mentioned vs. low vs. moderate vs. high) remains unchanged.
Statistical power analysis and sample size determination
Sample size for this study was determined on the basis of three experiments by Miron and colleagues (2009) that manipulated the recall of partners’ positive and negative characteristics, and assessed their influence on romantic affect.1 More specifically, we examined the effect sizes (ESs) (Cohen’s d) of eight theoretically relevant pairwise comparisons. The magnitude of the corresponding ESs ranged from d = 0.59 to d = 1.36, documenting medium to very large effects. These eight effect size values were meta-analytically combined (Howitt and Cramer 2014) into a value of d = 0.85 (a large effect according to Cohen 1988, 1990). Hence, considering d = 0.85 as a starting point for sample size estimation, we determined the size of the (sub)samples required to ensure enough statistical power to the experiment, with the aid of the computer program G*Power 3.1 (Faul et al. 2007).
Power calculations revealed that, in order to detect average effects at least as large as those documented by Miron et al. (2009) in planned pairwise comparisons, a sample size of no less than 18 participants per cell was required (statistical power 80%; α = 0.05; one-tailed tests; non-centrality parameter δ = 2.55). To be able to detect even somewhat smaller effects of perceived risk of romantic breakup on affect and commitment, we ran a sensitivity analysis and increased the number of participants to no less than 25 per cell, a sample size sufficient to notice also medium-to-large effects (d = 0.71).
Participants, design and procedure
One-hundred and four participants, all students at San Raffaele University of Milan (91.3% females; mean age = 21.55 years, SD = 2.06), expressed their informed consent and volunteered to participate in the study. The study was introduced as investigating—in general terms—participants’ everyday living in a romantic relationship. At the moment of the study, all participants were involved in a relationship; 19.2% of them in a long-distance relationship, and 7.6% of all participants living with their partner. The mean duration of the relationships was 23.54 months (SD = 18.04).
To assess the hypothesized influence of manipulated risk perception on romantic affect and relationship commitment, participants were randomly assigned to one of four deterrence conditions (control vs. low vs. moderate vs. high risk of ending the relationship). Upon assignment to conditions, participants received a short paper-and-pencil questionnaire, consisting of four sections. The first section requested basic demographic data (age, gender), and asked participants whether they were involved in a long-distance relationship (no/yes), if they lived with their partner (no/yes), and the duration of their relationship. The second section required them to think about central positive aspects of their romantic relationship, such as important time spent together, the well-being experienced in the relationship, and their sense of completion and reciprocity with the romantic partner (instigation of romantic affect—e.g., see Brehm et al. 2009; Miron et al. 2007, 2008). The third section introduced the experimental manipulation of the risk of ending their romantic relationship (deterrence of romantic affect and relationship commitment, on the basis of false feedback information). The fourth and final section contained some questions intended to measure felt romantic affect and relationship commitment (the dependent measures). Participants completed the questionnaires individually and anonymously. Then, they were extensively debriefed and thanked for participation.
Manipulation of the risk of ending the relationship
Participants were randomly assigned to one of four conditions (control vs. low vs. moderate vs. high risk of breakup). The risk of relationship breakup was manipulated by giving participants false feedback information about the chances of ending their own romantic relationship.
The manipulation consisted of three steps. In the first step, participants assigned to the low (vs. moderate vs. high) risk condition read a short passage on romantic relationships. Specifically, the passage stated that, after 2 years from the beginning of a romantic relationship, about 86 (vs. 52 vs. 18%) of young couples were typically still together, despite their experiences of stressful situations. The strategic purpose of this first step was to provide a general interpretive background, so that participants could later (more easily) believe the false feedback information. In the second step, participants indicated on the questionnaire how often, within a typical week, they experienced troubling discussions and/or minor quarrels with their romantic partner. Then, they handed the first three parts of the questionnaire back to the experimenter. In the third and decisive step, the experimenter ostensibly read participants’ answers on quarrels and discussions and, irrespective of participants’ actual responses, commented aloud that their own risk of ending the relationship was low (vs. moderate vs. high). This comment constituted the false feedback. Of course, steps 1 and 3 were always coordinated, such that the background information given at step 1 was always consistent with the feedback given at step 3.
In the control condition, participants read a short excerpt about organic food and related nutrition practices. Then they indicated how often, always within a typical week, they ate organic food with their romantic partner. In this condition, participants received no feedback information. Also, no reference of any sort was made to troubling and/or stressful episodes and the related risk of ending the romantic relationship.
We measured both romantic affect and commitment to the romantic relationship on bipolar visual-analogic scales (VAS) ranging from 0 to 12.50 cm (scale neutral midpoint = 6.25 cm). Strength of romantic affect was measured by averaging participants’ evaluations of the following three statements: “Right now, from a pure emotional point of view, I feel…” (Really bad with my romantic partner—Really good with my romantic partner), (Completely unmotivated towards my romantic partner—Completely motivated towards my romantic partner)2, (Totally unsatisfied by the relationship with my romantic partner—Completely satisfied by the relationship with my romantic partner) (three items, Cronbach’s α = 0.93).
Following Miron et al. (2009, 2012), strength of relationship commitment was measured with one single question, asking participants to what extent they agreed with the statement: “Right now, I feel…” (Not committed at all to my romantic relationship—Completely committed to my romantic relationship). Our items are very similar in content to items typically used in research on romantic relationships to measure romantic affect and commitment (e.g., Lehmiller and Agnew 2006; Rusbult et al. 1998).
Results and discussion
After examining the predicted overall cubic effect of risk manipulation (deterrence) on romantic affect and commitment, we ran polynomial contrasts with a pooled error term to test for the significance of the single legs in planned pairwise comparisons. Contrast weights were specified for each analysis. Following the lead of Richter (2016), we limited our significance testing to the assessment of whether each contrast was statistically significant or not, without testing also the significance of the residuals. Further, in agreement with the standards outlined by Wilkinson and the APA Task Force on Statistical Inference (1999), recently reminded and strongly advised also by Gendolla and Wright (2016), given the directional, theory-driven nature of our pairwise planned comparisons, we used one-tailed focused contrast tests in our hypotheses testing (Rosenthal and Rosnow 1985; see also Miron et al. 2008, p. 331, footnote 2, for practical instances in the same research area, and the complete rationale). All other significance tests were two-tailed.
Effects of manipulated risk of ending the relationship on romantic affect
To the extent that the perceived risk of relationship breakup can be assumed to act as a counterforce to the motivating aspects of the relationship, it should represent a deterrent to the intensity of romantic affect. Accordingly, on the basis of emotional intensity theory, we predicted systematic cubic variations in strength of romantic affect3 as a function of manipulated risk of ending the relationship. As shown in Fig. 1, a one-way ANOVA revealed that the predicted overall cubic effect of manipulated risk of relationship breakup on romantic affect was significant, F(1, 100) = 10.92, p = .001, MSE = 3.21, η2 = 0.10.4 Planned polynomial contrasts further revealed that, as predicted, romantic affect decreased from the control (M = 10.81, SD = 1.28, bootstrap 95% CI [10.30–11.29]) to low risk condition (M = 9.40, SD = 2.14, bootstrap 95% CI [8.57–10.16]), t(100) = 2.88, p = .002, d = 0.80 (contrast weights + 1 − 1 0 0), increased from the low to moderate condition (M = 10.74, SD = 1.73, bootstrap 95% CI [10.02–11.39]), t(100) = 2.71, p = .004, d = 0.69 (contrast weights 0 − 1 + 1 0), and decreased from the moderate to high condition (M = 9.66, SD = 1.88, bootstrap 95% CI [8.90–10.40]), t(100) = 2.14, p = .017, d = 0.60 (contrast weights 0 0 + 1 − 1). In all pairwise comparisons, each group’s mean fell outside the other group’s 95% CI. As predicted, neither the control vs. moderate risk condition, nor the low vs. high risk condition significantly differed from each other, ts < 0.52, ps > 0.60. Always as predicted, the control condition significantly differed from the high risk condition, t(100) = 2.30, p = .012 (contrast weights + 1 0 0 − 1). Table 1 reports the overall pattern of results for romantic affect. This pattern indicates that the perception of an increasingly higher manipulated risk of ending the romantic relationship was effective in modulating the intensity of romantic affect, according to the predicted cubic function. Within this general configuration, a first non-obvious finding is that, if compared with the control condition, in which no reference was made to the risk of relationship breakup, the perception of a low risk of ending the relationship significantly reduced the intensity of romantic affect. This finding is especially relevant here, because it is essentially counter-intuitive but, at the same time, fully predictable—and thus explainable—on the basis of emotional intensity theory (Brehm 1999; Brehm and Miron 2006). Notably, we further observed a significant drop from the moderate to the high risk condition, showing that, when faced with a ‘too high’ risk of ending the relationship, participants clearly reduced the intensity of their positive feelings towards the romantic partner. This finding is noteworthy for at least three reasons. First, within the field of romantic relationships, the only existing study that explicitly employed a false feedback procedure to experimentally manipulate deterrence to romantic affect (Miron et al. 2009, Study 2) did not include the high deterrence condition, so that no drop could be observed. Second, and more broadly speaking, the drop from the ‘moderate’ to the ‘high’ deterrence condition is generally difficult to document, as testified to by a certain number of studies, explicitly designed to test emotional intensity theory, which report mixed evidence on this specific point (e.g., Brehm et al. 2009; Brummett 1996; Dill 1997; Fuegen and Brehm 2004; Miron et al. 2007, 2008, 2009, 2011, 2012; Reysen and Katzarska-Miller 2013; Silvia and Brehm 2001). Third, and perhaps most importantly, on a visual-analogic bipolar scale (VAS) ranging from 0 to 12.50 cm (scale neutral midpoint = 6.25 cm), the absolute mean value of affect intensity we observed in the high risk condition was 9.66 (SD = 1.88). This value is significanlty higher than the scale neutral midpoint (one-sample t-test: t (24) = 9.07, p < .001). Thus, despite the drop we predicted and documented on the basis of EIT, affect intensity remained relatively strong in the high risk condition. This result seems to contradict the energy conservation principle invoked both by MIT and EIT and, actually, to substantiate Richter’s (2015) and Stanek and Richter’s (2016) recent argument that challenges the primacy of energy conservation in determining the strength of motivational (and emotional) responses.
The effect of perceived risk of ending the relationship (not mentioned vs. low vs. moderate vs. high) on romantic affect and commitment
Bootstrap 95% CIs
Bootstrap 95% CIs
Effects of manipulated risk of ending the relationship on commitment
Also in the case of commitment, we predicted systematic cubic variations in strength of relationship commitment as a function of risk manipulation. A one-way ANOVA confirmed the expected overall cubic effect of deterrence on commitment, F(1, 100) = 7.16, p = .009, MSE = 3.14, η2 = 0.07. As predicted, pairwise planned contrasts revealed that romantic affect decreased from the control (M = 10.91, SD = 1.32, bootstrap 95% CI [10.39–11.40]) to low risk condition (M = 9.68, SD = 2.00, bootstrap 95% CI [8.92–10.41]), t(100) = 2.55, p = .006, d = 0.73 (contrast weights + 1 − 1 0 0), increased from the low to moderate condition (M = 10.78, SD = 1.81, bootstrap 95% CI [10.03–11.46]), t(100) = 2.25, p = .013, d = 0.58 (contrast weights 0 − 1 + 1 0), and decreased from the moderate to high condition (M = 10.07, SD = 1.87, bootstrap 95% CI [9.35–10.79]), t(100) = 1.41, p = .080, d = 0.39 (contrast weights 0 0 + 1–1), though this effect attained only marginal significance—a result nearing prior work, in which the drop did not attain statistical significance (Miron et al. 2009, Study 1). As predicted, neither the control versus moderate risk conditions, nor the low versus high risk conditions significantly differed from each other, ts < 0.80, ps > 0.43. Always as expected, the control condition significantly differed from the high risk condition, t(100) = 1.69, p = .047 (contrast weights + 1 0 0 − 1). Taken together, these results show that, as predicted, the intensity of relationship commitment is a cubic function of manipulated risk perception. This novel finding is especially noteworthy, as it has been obtained by systematically manipulating, through false feedback information, the intensity of risk perception across four degrees of risk severity, so that we were able to observe, for the first time, the effects of distinct levels of perceived risk on relationship commitment. Most importantly, we observed the non-obvious drop in the intensity of commitment from the control to the low risk condition, together with a marginally significant decrease from the moderate to the high risk condition. However, though reduced, felt commitment in the high risk condition remained intense (M = 10.07, SD = 1.87), i.e. significantly higher than the scale neutral midpoint (6.25 cm), one-sample t-test: t (24) = 10.24, p < .001. This drop would not seem to correspond to the complete emotional disengagement one would expect on the basis of MIT and EIT in the high risk condition. As in the case of romantic affect, this finding adds to the argument that energy investment—as reflected in the intensity of felt commitment—may not be entirely driven by energy conservation concerns (Richter 2015; Stanek and Richter 2016). For the moment, however, this remains an open question that invites further research for clarification.
To test whether romantic affect indeed mediated the effect of experimentally manipulated risk of ending the relationship on commitment, we conducted a mediation analysis. For this analysis, we recoded manipulated risk (i.e., deterrence) as a variable with the following levels: + 1 − 1 + 1 − 1. This allowed us to contrast the control/moderate risk conditions to the low/high risk conditions, as we made comparable predictions for the collapsed conditions. Thus, the two levels of the new variable were + 1 (control/moderate manipulated risk) versus − 1 (low/high manipulated risk). Using a bias-corrected bootstrapping procedure (5000 resamples) with SPSS code (Hayes 2013; Preacher and Hayes 2004), we found that romantic affect fully mediated the effect of manipulated risk perception on relationship commitment, as the initial total effect of risk perception on commitment (c = 0.49, 95% CI [0.15–0.83]) became non-significant after romantic affect was included in the analysis (c’ = − 0.03, 95% CI [− .23, 0.17]), while the indirect effect of risk on commitment through romantic affect was nonzero (ab = 0.52, 95% CI [0.26–0.82]) (Fig. 2).5 The results of the mediation analysis thus confirm the hypothesized role of romantic affect in furthering relationship commitment (e.g., Gonzaga et al. 2001), and furthermore clearly substantiate and also complement Miron et al.'s (2009, 2012) findings on the mediational role of romantic affect—in our case, on the basis of manipulated risk perception.
Drawing on emotional intensity theory (Brehm 1999; Brehm and Brummett 1998; Brehm and Miron 2006), the main purpose of the present research was to test whether—and to show how—the manipulated risk of ending a relationship influenced the intensity of romantic affect and commitment towards the partner. The intensity of romantic feelings of affect and commitment varied as a cubic function of increasing levels of manipulated risk of breakup (risk not mentioned vs. low vs. moderate vs. high).6 All three study’s hypotheses were supported. First, we found that the intensity of romantic affect was strong when the risk of breakup was not mentioned; substantially reduced when the risk was low; strong when the risk was moderate; and, again, significantly reduced when the risk was high. Second, the intensity of commitment followed exactly the same cubic pattern. Third, we also showed that the effects of manipulated risk on romantic commitment were fully mediated by feelings of romantic affect.
Our results conceptually confirm and extend prior work by Miron and colleagues on the intensity of romantic feelings (Miron et al. 2009, 2012). In our study, the manipulation of risk was responsible for the cubic variations in the intensity of romantic feelings, well beyond participants’ objective reports of the actual number of disputes they had with their romantic partner (cf. footnote 4). Miron et al.’s (2009) manipulation of partner’s flaws and negative characteristics, clearly referred to obstacles that only potentially qualified as ‘risky’ for the relationship. In our study, we brought this line of thought a step further. We directly manipulated the risk of breakup, in the conviction that any challenge to the stability of a romantic relationship would systematically strengthen or reduce the intensity of romantic feelings, as long as it affected the perceived risk of relationship breakup. Any potentially stressful event, any negative partner’s characteristic, or any actual risk factor can act as a deterrent to the extent that it is perceived as ‘risky’ for the fate of the relationship.
A direct implication of the above reasoning is that the perceived risk of breakup may itself mediate the effects of actual obstacles—whether internal or external to the relationship (e.g., Neff and Karney 2004)—on romantic feelings. The perceived risk of breakup would thus represent a critical element directly linking the effects of actual obstacles to variations in strength of romantic feelings. This insight could be profitably tested in future research.
Though not explicitly designed to test the energy conservation principle at the core of MIT (Brehm and Self 1989; for a recent review: Richter et al. 2016), the present research clearly documented that in the ‘low’ and ‘high’ deterrence conditions the intensity of the reported romantic feelings was reduced if compared with the ‘control’ and ‘moderate’ deterrence conditions. In the ‘high’ deterrence condition, however, the strength of the reported affect did not seem to drop to the lowest possible level—i.e., to the level of total energy disinvestment—as one would expect on the basis of MIT and EIT. Our findings seem thus to substantiate and complement—also from the perspective of emotional intensity theory—the recent argument by Richter (2015) and Stanek and Richter (2016) that the energy conservation principle might not be the only principle at work for the fine-tuning of energy investment (Brehm 1975; see also Richter et al. 2016, for a review).
Aside from what principles may actually drive energy investment and strength of romantic feelings, we would encourage researchers to continue trying to document the drop from the ‘moderate’ to the ‘high’ deterrence condition, even if such a drop does not correspond to total energy withdrawal. This remains an important task to be accomplished to exclude reactance and/or dissonance alternative explanations of the effects of deterrence on affect intensity (e.g., Fuegen and Brehm 2004). Of course, the difficulty will still be that of finding an effective operational definition of ‘strong deterrence’ that, in the eyes of participants, really constitutes a credible counter-force to the ongoing emotion (Brehm et al. 2009; Stanek and Richter 2016).
Implications for the study of motivational and emotional processes in romantic relationships
In the field of romantic relationships, scholars have traditionally attributed to perceived and actual risk factors a negative prognostic role for the fate of the relationship. Obstacles and barriers have been almost exclusively conceptualized as costs and hindrances (e.g., Levinger 1999; Rusbult 1980). For instance, a notable implication of Rusbult’s investment model (Rusbult 1980, 1983; Rusbult et al. 2012; Rusbult and Buunk 1993) is that strength of romantic commitment is directly affected by three (negative) addictive factors: low levels of satisfaction, the presence of viable alternatives to the relationship, and few investments in the relationship. In Rusbult's model, these factors represent counter-forces to the stability of the relationship and, as such, predict a linear decrease in the intensity of romantic commitment. Our results on the cubic effects of perceived risk of breakup on strength of romantic commitment, by contrast, contradict the linear predictions of the investment model, and show the positive effects that obstacles may have on romantic feelings.
In a similar vein, our results have also implications for the risk regulation model (RRM) (Murray et al. 2006, 2008). The model predicts that people will regulate the perceived risk of being rejected by the partner by adjusting their cognitions and behaviors according to a ‘connectedness’ versus ‘self-protection’ motive (i.e., either by intensifying or reducing the psychological distance from their partner). According to the RRM, the perceived risk of rejection should impair the strength of romantic cognitions and behaviors in the romantic relationship. This leads to the prediction of a (linear) negative relationship between perceived risk and romantic involvement, in which low(er) levels of perceived risk correspond to high(er) levels of involvement. Therefore, the RRM cannot explain the drop in the intensity of romantic feelings we systematically observed in the ‘low’ risk condition—a drop that, by contrast, (a) is currently justified by the energy conservation principle (Brehm and Self 1989; Richter et al. 2016; but see also: Richter 2015; Stanek and Richter 2016), and (b) must be observed in the data, from the perspective of EIT (Brehm 1999; Brehm and Brummett 1998; Brehm and Miron 2006).
In our research, the manipulated risk of romantic breakup was treated as the causal factor for the observed variations in the intensity of affect and commitment. A better understanding of the causal role played by perceived risk in romantic relationships would enable professionals to determine and regulate the intensity of romantic feelings by intervening, directly, on such ubiquitous causal factor. From this perspective, a successful intervention on risk perception would thus directly ameliorate romantic affect and, in turn, relationship commitment.
Within the broader field of interpersonal relationships, an accurate assessment and regulation of commitment could be quite useful to people who make decisions (e.g., psychologists, clinicians, counselors), because commitment has been repeatedly shown to affect human interaction in many important respects. Reduced relationship commitment, for instance, leads to dissolution considerations and, thereby, to actual relationship breakup (VanderDrift et al. 2009). Relationship breakup, in turn, plays a critical role in the onset of depression (Monroe et al. 1999), psychological distress, and reduced life satisfaction (Rhoades et al. 2011).
This research started to fill an important knowledge gap on the motivational role that perceived risk of relationship breakup plays in romantic relationships. We accomplished this by focusing on the barrier-like properties of the perceived risk of breakup, and on the resulting paradoxical effects on romantic feelings. Perhaps most importantly, our findings revealed a complex and articulated picture that, by contrast, can be explained by a single, straightforward, coherent, and parsimonious theory (Brehm 1999; Brehm and Brummett 1998; Brehm and Miron 2006), whose reach goes well beyond the field of romantic relationships.
We did not consider more recent work by Miron et al. (2012) for sample size estimation, because two of the three studies reported by Miron and her colleagues were correlational in nature, whereas the third study—a true experiment—was designed to evaluate a more articulated research question and, also, did not find any main effect of the experimental manipulation on romantic affect.
In the original Italian wording of this item, the expression “From a pure emotional point of view, I feel completely motivated towards my romantic partner” [“Da un punto di vista puramente emotivo, mi sento completamente motivato nei confronti del mio partner”] conveys a strong sense of emotional involvement, that corresponds to a strong feeling of leaning towards the partner and feeling good with her/him, and also to a manifest sense of comfort and attraction towards the partner/relationship. This fact is reflected in the high value of the Cronbach’s alpha coefficient for the three items measuring romantic affect (α = 0.93).
A preliminary one-sample t-test revealed that the mean ratings of romantic affect in the control condition (M = 10.81, SD = 1.28; scale range 0–12.50 cm) were, on the average, significantly higher than the scale neutral midpoint (= 6.25), t (25) = 18.18, p < .001. This result may be taken to suggest that, at the beginning of the study, a relatively high degree of romantic affect was successfully induced in participants (see ‘Participants, design and procedure’ section above; cf. Brehm et al. 2009; Miron et al. 2007, 2008). As suggested by emotional intensity theory, in order to deter an affective state, that state must first be present with a certain degree of intensity (Brehm 1999). An analogous one-sample t-test run with respect to participants’ commitment ratings in the control condition (M = 10.91, SD = 1.32; scale range 0–12.50 cm) revealed that relationship commitment was significantly higher, on the average, than the corresponding scale neutral midpoint (= 6.25), t (25) = 18.01, p < .001. Thus, also a certain degree of romantic commitment was presumably present before the risk (i.e., deterrence) manipulation.
Importantly, a further one-way ANOVA indicated that, as expected, the mean number of discussions and/or minor quarrels that participants reported to have typically had, per week, with their romantic partner (Mtotal = 1.35, SDtotal = 1.54) did not differ among low (M = 1.29, SD = 1.24) vs. moderate (M = 1.54, SD = 2.19) vs. high (M = 1.22, SD = 1.01) risk conditions, F (2, 75) = 0.30, p = 0.743. It was therefore the false feedback in itself—not the actual number of recalled troublesome exchanges between partners—that systematically deterred romantic affect. In the control condition, of course, no reference to disputes between partners was made, nor participants were asked to recall such potentially stressful episodes.
We conducted also a traditional Baron and Kenny (1986) mediation analysis, in which (1) the risk of breakup predicted both the intensity of romantic commitment (β = 0.27, t = 2.841, p = .005) and (2) of romantic affect (β = 0.34, t = 3.594, p = .001); (3) the intensity of romantic affect predicted the intensity of commitment (β = 0.85, t = 16.161, p < .001); and, finally, (4) the risk of breakup no longer predicted the intensity of commitment when romantic affect was entered into the equation (β = − 0.02, t = − 0.271, p = 0.79)—this signaling complete mediation. The indirect path was significant at the Sobel (1982) test (test statistic = 3.52, Se = 0.15, p < .001).
Technically, we cannot attribute, univocally, the observed effects to the influence of the manipulation at step 1 (perceived likelihood of general relationship failure after two years), instead to the influence of the manipulation at step 3 (perceived likelihood of personal relationship failure after two years), or to a combination of both. However, we considered the complete three-step procedure necessary to guarantee the credibility of the personalized feedback information to be given at step 3. Each of the two steps (step 1 and 3), if implemented alone, could have not been sufficient to effectively manipulate the independent variable. Further, both steps 1 and 3 were intended to operationalize the same theoretical construct in different but converging ways. Thus, both steps shared the same operative goal—inducing in participants a sense of being at risk.
- Ach, N. (1910). Über den Willensakt und das Temperament [About the act of will and temperament]. Leipzig: Quelle und Meyer.Google Scholar
- Brehm, J. W. (1975). Research on motivational suppression [Grant Proposal]. Lawrence: University of Kansas.Google Scholar
- Brummett, B. H. (1996). The intensity of anger. Unpublished doctoral dissertation, University of Kansas.Google Scholar
- Cohen, J. (1988). Statistical power analysis for the behavioral sciences. Hillsdale: Lawrence Erlbaum Associates.Google Scholar
- Dill, J. (1997). Paradoxical anger: Investigations into the emotional and physiological predictions of Brehm’s theory of emotional intensity. Unpublished doctoral dissertation, University of Missouri, Columbia.Google Scholar
- Fuegen, K., & Brehm, J. W. (2004). The intensity of affect and resistance to social influence. In E. S. Knowles & J. A. Linn (Eds.), Resistance and persuasion (pp. 39–63). Mahwah: Lawrence Erlbaum Associates.Google Scholar
- Hayes, A. F. (2013). Introduction to mediation, moderation, and conditional process analysis: A regression-based approach. New York: Guilford Press.Google Scholar
- Hillgruber, A. (1912). Fortlaufende Arbeit und Willensbetätigung [Continuos work and will activity]. Leipzig: Quelle und Meyer.Google Scholar
- Howitt, D., & Cramer, D. (2014). Introduction to statistics in psychology (6th ed.). London: Pearson Education Ltd..Google Scholar
- Kline, G. H., Stanley, S. M., Markman, H. J., Olmos-Gallo, P. A., St. Peters, M., Whitton, S. W., & Prado, L. M. (2004). Timing is everything: Pre-engagement cohabitation and increased risk for poor marital outcomes. Journal of Family Psychology, 18, 311–318. doi:10.1037/0893-318.104.22.1681.CrossRefPubMedGoogle Scholar
- Monroe, S. M., Rohde, P., Seeley, J. R., & Lewinsohn, P. M. (1999). Life events and depression in adolescence: Relationship loss as a prospective risk factor for first onset of major depressive disorder. Journal of Abnormal Psychology, 108, 606–614. doi:10.1037/0021-843X.108.4.606.CrossRefPubMedGoogle Scholar
- Murray, S. L., Derrick, J. L., Leder, S., & Holmes, J. G. (2008). Balancing connectedness and self-protection goals in close relationships: A levels-of-processing perspective on risk regulation. Journal of Personality and Social Psychology, 94, 429–459. doi:10.1037/0022-3522.214.171.1249.CrossRefPubMedGoogle Scholar
- Pantaleo, G. (2011). Enjoying multiplicity: From familiarity to ‘multiple perspectives’. In M. Cadinu, S. Galdi, & A. Maass (Eds.), Social perception, cognition, and language in honour of Arcuri (pp. 51–65). Padua: Cleup.Google Scholar
- Parks, M. R., Stan, C. M., & Eggert, L. L. (1983). Romantic involvement and social network involvement. Social Psychology Quarterly, 46, 116–131. http://www.jstor.org/stable/3033848.
- Rhoades, G. K., Kamp Dush, C. M., Atkins, D. C., Stanley, S. M., & Markman, H. J. (2011). Breaking up is hard to do: The impact of unmarried relationship dissolution on mental health and life satisfaction. Journal of Family Psychology, 25, 366–374. doi:10.1037/a0023627.CrossRefPubMedPubMedCentralGoogle Scholar
- Rosenthal, R., & Rosnow, R.L. (1985). Contrast analysis: Focused comparisons in the analysis of variance. New York: Cambridge University Press.Google Scholar
- Sinclair, H. C., & Ellithorpe, C. N. (2014). The new story of Romeo and Juliet. In C. R. Agnew (Ed.). Social influences on romantic relationships: Beyond the dyad (Advances in Personal Relationships) (pp. 148–170). Cambridge: Cambridge University Press. doi:10.1017/CBo9781139333640.010.CrossRefGoogle Scholar
- VanderDrift, L. E., Agnew, C., & Wilson, J. E. (2009). Non-marital romantic relationship commitment and leave behavior: The mediating role of dissolution consideration. Department of Psychological Sciences Faculty Publications. Paper 25. doi:10.1177/0146167209337543.
- Wright, H. F. (1937). The influence of barriers upon strength of motivation. Durham: Duke University Press.Google Scholar
- Wright, R. A. (2011). Motivational when motivational wasn’t cool. Chapter. In R. M. Arkin (Ed.), Most underappreciated: 50 prominent social psychologists describe their most unloved work. New York: Oxford University Press.Google Scholar