Men say “I love you” before women do: Robust across several countries

Expressing love and the male confession bias

Romantic love and passion are cultural universals (Jankowiak, 2008) and the need to feel belonging within stable social relationships is ubiquitous (Baumeister & Leary, 1995). The feeling of love predicts desire, sympathy, and commitment by facilitating trust and improving how couples resolve conflict (Gonzaga et al., 2001). Commitment, in turn, both predicts and causes forgiveness in relationships (Finkel et al., 2002) and lowers anxiety in response to stress when we feel supported (Collins & Feeney, 2013; Ditzen et al., 2008). Moreover, expressions of love and acts of affection enhance commitment (Joel et al., 2013; Marston et al., 1998) and predict stable marital bonds (Huston et al., 2001). In sum, both the emotion and expression of romantic love contribute to stable long-term relationships and positive health outcomes (see also Cacioppo et al., 2015; Foran et al., 2015; Whisman & South, 2017 for discussion).

While the emotion and expression of romantic love underpins good quality relationships, individuals differ in their proclivity toward romantic love, and such differences may be revealed in speech acts. Saying “I love you” signals commitment to future behavior with a romantic partner (Ackerman et al., 2011). Extensive research on evolutionary approaches to behavior within romantic relationships has revealed differences between men and women in mating-related preferences, cognitions, and behaviors, which may have implications for how heterosexual relationships function at various stages (see Buss & Schmitt, 2019 for a recent review). Many of these studies have used self-report methods to test hypotheses generated from evolutionary theories and observations across diverse cultures and species. Moreover, within this field, robust findings from self-report data converge with experimental tests, among other diverse methods (see Buss & Schmitt, 2019 for a recent review), consistent with the importance of triangulation in science (Munafo & Davey Smith, 2018). Research on sex differences around mating has developed, in part, based on two key theories: parental investment theory of mate choice (Janicke et al., 2016; Todd et al., 2007; Trivers, 1972) and Error Management Theories of human perception and cognition (Haselton & Buss, 2000; Haselton & Nettle, 2006; Johnson et al., 2013). The current paper and the research we attempt to replicate (Ackerman et al., 2011) also draws from these two theories. Error management theory proposes that because decision-making under uncertainty can lead to error, human perception and cognition has evolved to pursue the least costly of two opposite strategies in contexts related to survival or reproductive fitness. While false positive errors are favored (i.e., “optimism”) in contexts where it is less costly to think/act/speak than to not do so, false negative errors (i.e., “caution”) are favored in contexts where it is less costly, from an evolutionary perspective, to avoid that same behavior than to engage in it. Because speech acts such as love confessions are made under a degree of uncertainty and are associated with non-trivial costs and benefits (e.g., backfire in the attempt to escalate the level of commitment), particularly when communicated for the first time, they too can be studied via this framework.

According to biological theories on parental investment, the more investing and “selective” sex would value signals of commitment more than the other, less investing, sex, particularly in light of the long period of parental investment required for our species (see Lawson & Mace, 2011 for discussion). Due to both biological (pregnancy and lactation) and societal factors, women generally invest more heavily in children than men. Therefore, following this theory and empirical evidence for sex differences in mating-related cognition and behaviors (e.g., Haselton & Buss, 2000; Janicke et al., 2016; Todd et al., 2007) that are consistent with an error management framework (i.e., relatively cautious females and optimistic males, Haselton & Buss, 2000), women could avoid the greater costs of a potential poor mate choice by setting a relatively higher threshold than men do in displaying commitment or investment, all else equal. Consistent with this proposal, American men, on average, are more likely than American women to say “I love you” first in a romantic relationship (the “male confession bias”), and are happier than women to hear their partner confess love, particularly if this is communicated before the couple has engaged in sexual intimacy within their relationship (Ackerman et al., 2011). This earlier male confession may function to escalate sexual intimacy within the relationship in light of the stronger relative importance of men communicating commitment to women than vice versa, all else equal (Ackerman et al., 2011).

Replication of this effect across cultures is important as it allows us to infer whether prior observations are universal or generalizable (see, e.g., Pollet & Saxton, 2019 for a recent discussion). Romantic love is a cultural universal (Jankowiak, 2008), but cultures differ in their romantic practices (Hatfield & Rapson, 2005) and gender norms. Moreover, little is known about love confessions as they are normally a private act, which may partly explain why folk beliefs and perceptions differ from observed behaviors when examining who confesses love first (Ackerman et al., 2011). Thus, cross-cultural research on love confessions is needed and can shed light on the utility of evolutionary explanations for this topic if the previously observed sex differences are robust. Indeed, cross-cultural research is central to evolutionary approaches to human behavior, as this field is interested both in patterns that occur at the species level and cultural variations that can be predicted by evolutionary theory. Thus, the current study attempted to replicate and generalize Ackerman and colleagues’ findings for sex differences in confessing love (who confesses first), and other sex differences related to a love confession (happiness at hearing a love confession, days into relationship before: thinking about confessing love, and confessing love), across a larger global sample, while also examining whether this sex difference is consistent across each nation surveyed (Hypothesis #1).

Cultural differences and the male confession bias

Consistency in behaviors across diverse cultures does not rule out systematic variation between cultures in those same behaviors, such as the extent to which they are expressed, driven by the characteristics and demands of a given environment. Thus, we also tested for cultural differences in responses to our survey items, examining whether the timing of a love confession was predicted by the national sex ratio of males to females. Sex ratio theory, corroborated by correlational evidence, suggests that proxies for male investment and female promiscuity are observed, respectively, in ecologies with male-biased versus female-biased sex ratios (reviewed in Del Giudice, 2012). For example, historical data from North America demonstrates that men invested more in pair bonds and children in male-biased regions where women had more “bargaining power” as women were more likely to find a partner than men were (reviewed in Schacht & Smith, 2017). Conversely, female-biased sex ratios are related to greater promiscuity according to psychometric tests in non-western communities (Schacht & Borgerhoff Mulder, 2015). Complementary experimental work demonstrates that men and women shift from sex-typical mating strategies (female monogamy and male promiscuity, Schmitt, 2005) toward the strategy of the opposite-sex when the opposite-sex is scarce, as the latter has more bargaining power (Moss & Maner, 2016). Here, we integrate this theory with the earlier logic on love confessions to make predictions about the extent to which the sex ratio alters the costs versus benefits of confessing love, regardless of the veracity of the speech act (i.e., all else equal). Within the general hypothesis that the male confession bias varies according to the national sex ratio (Hypothesis #2) we can test two alternate hypotheses, examining the direction of this variation. Evidence that men confess love earlier than women in countries with a more male-biased sex ratio (Hypothesis #2a) would support the proposal that men confess love as part of an unconscious strategy to signal investment potential in light of their environment, because male–male competition is more intense, and investment potential is valued by women, who have greater choice in male-biased environments. Alternatively, evidence that men confess love earlier than women in countries with a more female-biased sex ratio (Hypothesis #2b) would support the proposal that men confess love as part of an unconscious strategy to escalate intimacy (see Ackerman et al., 2011) when their environment permits greater mating opportunities. Tests of these alternate hypotheses can shed initial light on the potential evolutionary dynamics of love confessions in different environments.

Attachment styles and individual differences in the male confession bias

In addition to sex differences in the expression of and emotional response to romantic love, there are some individual differences. Indeed, some people worry a lot about the security of their relationships and others avoid aspects of romantic intimacy, reflecting two insecure attachment styles of anxiousness and avoidance, respectively. The final aim of our cross-cultural study was to incorporate attachment styles into the study of love confessions. Developmental theories of attachment (Bowlby, 1973; see Groh et al., 2017 for a recent review) have been applied extensively to the study of romantic relationship functioning (e.g., Hazan & Shaver, 1987; Mikulincer & Shaver, 2007) both within and across cultures (see, e.g., Schmitt et al., 2004). Attachment styles predict trust in romantic contexts (Fitzpatrick & Lafontaine, 2017; see also Bartz et al., 2011), closeness (Mikulincer et al., 2010), relationship quality (Noftle & Shaver, 2006), care provision and support seeking (Collins & Feeney, 2000; Feeney & Collins, 2003), and how we respond to emotional warmth from others (Philipp-Muller & MacDonald, 2017), and close contact from our partner (Kim et al., 2018). Attachment styles are also related to the expression of concern for the good of our partner (i.e., compassionate love) in both younger and older adult cohorts (Sabey & Rauer, 2017; Sprecher & Fehr, 2011). Collectively, attachment styles are related to a variety of behaviors within romantic relationships.

Attachment styles also bias perceptions of our relationship in the present and future (see Dykas & Cassidy, 2011 for a review). For example, attachment styles bias the affective content and written interpretation of events within our relationship (Collins, 1996), how we evaluate our partner’s empathy (Simpson et al., 2011), and support from our partner under stress (Collins & Feeney, 2004), and perceive our partner as responsive to our needs (or lack thereof) for trust, intimacy, and independence (Ren et al., 2017). Indeed, anxiously attached individuals require more time, affection, and self-disclosure than secure individuals do before they judge their relationship as “close,” while avoidant individuals require less time, affection, and self-disclosure than secure individuals do before they judge their relationship as “close,” which suggests that perceptual mechanisms motivate different approach-avoidance behaviors among insecurely attached individuals (Hudson & Fraley, 2017). Collectively, attachment styles underpin how individuals navigate and appraise a romantic relationship, which may explain relationship outcomes (see also Karantzas et al., 2014).

In the current study, we examined whether the male confession bias (the tendency to say “I love you” first/earlier) is moderated by attachment style. Men, on average, have a more avoidant and less anxious attachment style than women (see Del Giudice, 2011 for a meta-analytic review across different nations). Both biological sex and attachment style may therefore play a role in love confessions, as the latter is important in how we appraise our relationship over time. Considering the evolutionary logic behind earlier male confession, we therefore examined whether attachment style could moderate Ackerman and colleagues’ previously observed sex difference. Because attachment styles motivate vigilance to relationship maintenance in different ways (Barbaro et al., 2016), they could alter the perceived costs and benefits of confessing love (escalating or maintaining a relationship) and responding to a confession (i.e., optimistically or cautiously, Haselton & Buss, 2000; Haselton & Nettle, 2006), interacting with the different mating strategies of men and women outlined in Hypothesis #1. Here, we examined whether men confess love earlier than their partners do if the respondent in our study has an insecure attachment style (i.e., anxious or avoidant, Hypothesis #3). We predict that both insecure attachment styles are related to an earlier confession among male respondents. Anxiously attached men would be expected to confess love earlier than their partners do to feel secure about retaining their romantic partner, as anxious attachment is related to hypervigilance to partner rejection (Barbaro et al., 2016), and parental investment theories emphasize the importance of sexual access to selective females for male reproductive fitness. Given the theorized role of love confessions in escalating sexual intimacy (Ackerman et al., 2011), earlier love confession would also represent a strategy for avoidant men to escalate intimacy, as avoidant people require less time to perceive their relationship as “close” (Hudson & Fraley, 2017). Finally, as error management theories predict greater female skepticism of signals of male commitment (Haselton & Buss, 2000; Haselton & Nettle, 2006; Johnson et al., 2013), anxiously attached women would be predicted to delay a love confession, as anxious people require more time to perceive their relationship as “close” (Hudson & Fraley, 2017), and skepticism of partner commitment (see Ackerman et al., 2011; Johnson et al., 2013 for discussion) may be heightened when women are anxiously attached. We have no a priori directional hypothesis for female avoidance and the male confession bias.

Thus, in sum, our study tested the following pre-registered hypotheses ¹ :

Hypothesis # 1: We will attempt to replicate previously reported sex differences in confessing love, and other behaviors related to a love confession, across a larger global sample, and test for consistency in the “male confession bias” across each nation surveyed via our convenience sampling strategy (see Table 1): Men confess love first in a relationship more often than women do; men are happier than women to hear “I love you”; men think about, and confess love earlier in a relationship than women do.

OPEN IN VIEWER

Table 1.

Cross-cultural analyses showing the proportion of each national sample in which men confessed love first in their current/most recent romantic relationship.

Country	N	M	t	p	Effect size (r)	95% CI (of ES)
Australia	63	0.62	1.93	=.06	.53	[.42, .65]
Brazil	70	0.79	5.78	<.001	.69	[.60, .79]
Chile	69	0.81	6.57	<.001	.72	[.62, .82]
Colombia	130	0.68	4.51	<.001	.59	[.51, .66]
France	59	0.59	1.44	=.15	.51	[.37, .65]
Poland	54	0.76	4.41	<.001	.66	[.54, .78]
United Kingdom	525	0.65	7.07	<.001	.56	[.52, .60]

Hypothesis # 2: People who live in countries with a more male-biased sex ratio will report that men confess love earlier than women (Hypothesis #2a). People who live in countries with a more female-biased sex ratio will report that men confess love earlier than women (Hypothesis #2b).

Hypothesis # 3: If the respondent in our study has a relatively more insecure attachment style (i.e., anxious or avoidant ² ), they will be more likely to report that the male in the relationship confessed love earlier than the female did.

We also examined potential effects of self-rated attractiveness in one exploratory analysis, as a proxy for “mate quality.” Attractiveness may partly explain the timing of a love confession, if more attractive individuals can afford to be choosier/selective in light of our positive orientation toward them (see Maestripieri et al., 2017 for general discussion). Thus, such individuals may be more likely to delay a love confession.

Participants

A total of 3109 participants (M_age = 31.90 years, SD = 11.60 years, 71% women, 26% men, 1% non-binary, 3% did not disclose; 85% heterosexual, 6% homosexual, 5% bisexual, 1% other, 3% did not disclose) were recruited to an online study by all authors in their respective countries, via campuses and the wider community, research participant pools, word of mouth, Twitter, academic groups on social media and a press release from the lead author’s communications department. The press-release informed readers that we were conducting a global study into romantic expression but did not mention that we were measuring sex differences or attachment styles. The survey platform did not permit duplicate responses from the same device, and participants were not compensated for their time.

All procedures for testing and recruitment were approved via the lead author’s Ethics Committee, with our introduction and method sections pre-registered via the Open Science Framework after data collection but before data analysis (https://osf.io/hsvx9/). Participants provided informed consent after reading an information sheet describing the contents of the survey. We excluded participants who (i) reported being less than 18 years old, (ii) did not report their sex as male or female or identify as heterosexual, or, for cross-national analyses, and (iii) if their IP address did not match their reported country of residence. After applying exclusion criteria (see Data analysis), we analyzed data from the full eligible sample (N = 1428 participants, 336 men, 1092 women. M _age = 32.90 years, SD = 11.59 years), and for analyses comparing nations (seven countries from three continents), included countries with data from at least 50 respondents (N = 970 participants, 251 men, 719 women. M _age = 34.16 years, SD = 12.28 years), who reside in the same country as their birth (following Watkins et al., 2019), which exceeds 80% power to detect moderate effects (Lakens & Evers, 2014).

Measures

Data analysis

First, chi square tests on the binary male confession bias variable examined whether men were more likely to admit love first than women were, with data analyzed across the sample and for each sampled nation separately. We also tested for sex differences in (i) days into the relationship before thinking and (ii) confessing love, and (iii) happiness at hearing “I love you” using t-tests.

Then, generalized linear mixed effects models (GLMMs) were run to test for relationships between national sex ratio and male confession bias (both binary and continuous variables), nested within the higher-level variable of country (random intercept).

Finally, a linear mixed effects model on the continuous male confession bias variable was run with the between subjects’ factor participant sex, the covariates anxious attachment score and avoidant attachment score and each covariate entered separately as a two-way interaction term with the between subjects’ factor, and with a random intercept for each country. ³ Models were re-run with different outcome variables: First thought about confessing love (days into the relationship); First love confession (days into the relationship, i.e., in absolute terms instead of relative to partner); and Happiness at hearing “I love you.” Two additional non-pre-registered models were run on Partner’s love confession (days into the relationship in which their partner confessed love) and the binary male confession bias variable.

Data on number of days before love confessions were analyzed if the respondent gave a definitive numerical answer (with an average taken if the participant estimated a range within 10 days) that was logical (i.e., thinking about confessing love before saying it; days before confessing love was less than the participant’s age), complete (i.e., provided data on when love was confessed/reciprocated and who confessed first), and consistent with their other responses to these questions (i.e., on who confessed first). Details of further robustness checks are provided in the results. This included repeated tests after the exclusion of outliers, to confirm whether outliers exert any undue influence on the conclusions derived from our models, given that extreme values could still represent genuine responses. Of the total sample recruited, 4% were dropped for not being eligible for broad/initial analyses of the sample as described in the participants section, with a further 12% dropped for not being heterosexual. Forty-nine percent of the recruited sample did not meet the criteria for cross-cultural analyses detailed in the participant’s section. Four percent were dropped for not completing the love confession questions correctly, and six percent were dropped for analyses involving attachment as they did not complete all items on the questionnaire. Four percent were dropped for analyses involving both attachment and cultural differences due to missing data on these items.

The male confession bias across cultures (Hypothesis # 1)

Participants reported that men confessed love first in the relationship at levels greater than chance (χ²(1) = 170.89; p < .001). Interestingly, a greater proportion of women than men reported that men confessed love first in the relationship (χ²(1) = 19.40; p < .001), a reporting or memory bias observed previously (Ackerman et al., 2011). However, the male confession bias effect was still significant when chi square tests were split by respondent sex (both χ²(1) > 7.44, both p < .007). Men confessed love first in six of the seven countries with sufficient data (see Table 1), representing a mean weighted effect size (r) across nations of 0.59 (95% CI [0.55, 0.63], as calculated in Watkins et al. (2019). Of note, while there was no significant sex difference in France, a one-tailed prediction would be valid for Australia given the directional pre-registered hypothesis, and as the confidence intervals for the effect size suggest a moderate effect.

No sex differences were observed in days in which respondents first thought about confessing love (M_men = 69.87 days, 95% CI [60.04, 79.69], M_women = 76.99 days, 95% CI [71.19, 82.79], absolute t(1411) = 1.19; p = .23), days into the relationship in which love was confessed (M_men = 107.76 days, 95% CI [90.13, 125.39], M_women = 122.61 days, 95% CI [104.08, 141.15], absolute t(1426) = 0.84; p = .40), or happiness at hearing “I love you” (M_men = 87.01, 95% CI [85.29, 88.74], M_women = 88.73, 95% CI [87.81, 89.66], absolute t(1421) = 1.76; p = .08).

The male confession bias between cultures (Hypotheses # 2a and #2b)

A GLMM using the logit link function on the binary variable male confession bias (Full model: Male confession bias ∼ National Sex Ratio [nested within country] + Participant Sex + Participant age + Relationship status + Self-rated attractiveness) revealed a negative effect of National Sex Ratio, before and after outliers were excluded and with and without control variables (Participant sex was the only significant control variable in this model, Est(b) = 0.46, SE = 0.15, t = 2.95; p < .01). Men were more likely to confess love first if they lived in a nation with a relatively more female-biased sex ratio (Est(b) = −10.51, SE = 4.85, t = −2.16, p = .04 for the full model). Of note, we chose participants’ sex, age and relationship status as the control variables to include in our models as a robustness check because these variables may have an effect on the outcome variable and their distributions might differ between countries in light of our convenience sampling strategy.

Running this same model (but with the identity link function) on our continuous male confession bias variable revealed effects of Participant sex (Est(b) = 23.17, SE = 5.20, t = 4.46; p < .001) and Relationship status (Est(b) = 13.98, SE = 5.90, t = 2.37; p = .02) only after controlling for outliers (+/− 3SD). No effects of the National Sex Ratio were observed in the simple model (National Sex Ratio only), after controlling for outliers, or when including the above demographic characteristics as a robustness check (all other absolute t < 1.51, all other p > .14). We should note here that the continuous male confession bias variable used in these models is less informative than the binary variable used in the previous set of models. We have more confidence in the results of these first models, as most people responded to a love confession on the same day as their partner (see discussion).

Individual differences in the male confession bias (Hypothesis #3)

To test Hypothesis # 3, which involves combinations of several continuous and categorical variables plus their interactions, separate LMMs were run (Outcome variable ∼ Participant sex + Anxious attachment style + Avoidant attachment style + [Anxious attachment style * Participant sex] + [Avoidant attachment style * Participant sex] + [1 | Country]) on the following outcome variables: continuous male confession bias; First thought about confessing love; First love confession; Happiness at hearing “I love you”; Partner’s love confession; and a GLMM using the logit link function was run on the binary male confession bias variable. These analyses revealed no significant main effects or interactions (all p > .07), except for a negative relationship between avoidant attachment style and Happiness at hearing “I love you” (Est(b) = −5.83, SE = 0.96, t = −6.06; p < .01) and a positive relationship between anxious attachment style and Happiness at hearing “I love you” (Est(b) = 2.13, SE = 0.78, t = 2.72; p < .01), which remained significant after excluding outliers (see supplementary materials and Figure 1 for effect sizes). Excluding outliers resulted in significant interactions between participant sex and both avoidant attachment style and, separately, anxious attachment style for the model on continuous male confession bias (see supplementary materials). These interactions were driven by women’s, but not men’s, attachment style, such that more avoidant and less anxious women took longer to confess love relative to their partner. As this was not observed in the full dataset, we do not interpret these interactions further. Similar non-robust results were observed for an additional model using the binary, instead of the continuous, male confession bias variable (see supplementary materials). Of general note, we observed no sex differences across the sample in avoidant (M_men= 2.45, M_women= 2.44, 95% CI [−0.11, 0.13], absolute t(1233) = 0.17; p = .87) or anxious attachment styles (M _men= 3.50, M _women= 3.61, 95% CI [−0.26, 0.04], absolute t(1233) = 1.50; p = .13), while anxious and avoidant styles were correlated in both men (r(282) = .17; p = .004) and women (r(949) = .17; p < .001).OPEN IN VIEWER