The broad autism phenotype predicts relationship outcomes in newly formed college roommates

Introduction

The current Diagnostic and Statistical Manual of Mental Disorders (5th ed.; DSM-5) conceptualizes the social impairments and repetitive behaviors associated with autism spectrum disorder (ASD) as varying along a continuum of severity (American Psychiatric Association (APA), 2013). This continuum, however, does not end at the boundary for categorical diagnosis (Ingersoll and Wainer, 2014). The notion that milder but qualitatively similar traits can be observed at subclinical levels dates back to Leo Kanner’s original case studies in which he noted characteristic similarities between parents and their autistic child in certain personality features, including reduced social interest and motivation, differences in communicative ability, and a desire for sameness (Kanner, 1943). Although this “broad autism phenotype” (BAP; Bolton et al., 1994) is found most prominently within family members of an individual with an ASD diagnosis (Bolton et al., 1994; Piven et al., 1994; Sasson et al., 2013a), indicating a strong familial inheritance of autism-related symptoms (Lyall et al., 2014; Ozonoff, et al., 2014; Sasson et al., 2013b), they are also continuously distributed within the general population (Baron-Cohen et al., 2001; Constantino and Todd, 2003, 2005). Here, they are associated with a range of autism-related characteristics, including reduced emotional intelligence (Gökçen et al., 2014), poorer affect recognition (Ingersoll, 2010), and restricted activities and interests (Baron-Cohen et al., 2001).

BAP traits in the general population are modestly predicted by traditional personality measures (e.g. the “Big Five”), with higher levels of overall BAP traits associated with reduced extraversion and increased neuroticism (Murphy et al., 2000). However, traditional personality measures such as the NEO Personality Inventory–Revised (NEO-PI-R) account for just 24% of the variability of BAP traits (Wakabayashi et al., 2006), suggesting that the BAP may constitute a relatively independent component of personality (though for some stronger associations between individual BAP and Big Five traits, see Wainer et al., 2011). Furthermore, unlike in ASD, traits associated with the BAP are often domain-specific and do not necessarily co-occur. Individuals commonly only exhibit a single BAP trait (Sasson et al., 2013b), reduced social motivation, pragmatic language abnormalities, or rigid personality, suggesting that conceptualizing autism-related characteristics as “fractionable” rather than as the unitary concept required for diagnosis may help address the genetic heterogeneity of ASD (Burmeister et al., 2008; Happé and Ronald, 2008). For example, in the general population, social BAP traits, but not rigid personality, are associated with reduced social cognitive performance and interpersonal skill (Sasson et al., 2013c), indicating that the social difficulties characterizing ASD may extend to a lesser degree to those with social BAP traits.

Consistent with this conclusion, multiple studies have reported that subclinical social traits of autism in nonclinical populations are associated with reduced relationship quality. Parents of a child with ASD who express greater autism-related traits report reduced social interest, fewer friendships, and lower quality relationships (Losh et al., 2008; Piven et al., 1997). Furthermore, Pollmann et al. (2010) examined newlyweds and found that responsiveness (i.e. supportive reactions and attending to partners’ needs; Reis et al., 2004) mediated the relationship between newlywed husbands’ overall BAP traits and their relationship satisfaction. In this way, expressions of social BAP traits in certain contexts may correspond with decreased levels of supportive and appropriate responses to social partners. A similar pattern has also been found for nonromantic relationships. Jobe and White (2007) reported that college students with higher levels of overall BAP traits were more likely to experience loneliness and had difficulty maintaining long-term friendships. Collectively, these studies suggest that BAP traits, and perhaps social ones in particular, may relate to the quantity and quality of interpersonal relationships in nonclinical populations.

Although these studies indicate an association between BAP traits and relationship outcomes, how an individual’s BAP traits influence behavior toward a partner and in turn how those traits may elicit behaviors from a partner remains relatively unexplored. One useful way to investigate this question is to use the Actor–Partner Interdependence Model (APIM; Kenny et al., 2006), a statistical tool for dyadic data that estimates intrapersonal and interpersonal effects of predictors on outcomes. Of relevance to the current work, researchers have used this model to investigate the contributions of both romantic partners’ personality traits to their relationship satisfaction (Robins et al., 2000) as well as the implications of personality trait similarity for romantic relationship outcomes (Dyrenforth et al., 2010). Figure 1 details the features of a basic APIM using a single BAP trait, social aloofness, to predict relationship satisfaction among pairs of roommates.

OPEN IN VIEWER

Figure 1.

Basic Actor–Partner Interdependence Model using aloofness to predict relationship satisfaction. Actor effects (i.e. paths a1 and a2) estimate the association between participants’ levels of Aloofness and their own satisfaction with the roommate relationship, and partner effects (i.e. paths p1 and p2) estimate the association between participants’ levels of Aloofness and their roommate’s satisfaction with the relationship. Designating a person as roommate 1 or roommate 2 is arbitrary, and thus actor and partner effects are pooled when estimated in the model. Interaction terms between actor and partner effects are used to determine whether links between certain BAP traits of participants and relationship outcomes depend upon the BAP traits of their partner. The correlation between the predictors (r_x) denotes degree of similarity on Aloofness at the beginning of the relationship, and the error terms (E₁, E₂) are correlated (r_e) to provide an estimate of whether, after partialling out the effects of both participants’ Aloofness scores, roommates’ reports of relationship satisfaction are still similar to one another.

Wainer et al. (2013) used the APIM to assess how the BAP is expressed and experienced within close friendships and found that participants with higher overall BAP traits engaged in less positive interpersonal interactions with their friends and reported lower levels of friendship satisfaction. Individuals who viewed their friends as higher on overall BAP characteristics also perceived their friendship as less satisfying and engaged in more negative behaviors toward them. These findings suggest that both an individual’s BAP traits and a friend’s perception of those traits are associated with reduced relationship quality. Wainer et al. (2013) also found that overall BAP traits among close friends were modestly correlated, suggesting that these personality features may overlap within friendships. Whether this occurs because similarities drive friendship formation (e.g. the “assortative mating” hypothesis for romantic partners exhibiting the BAP; Constantino and Todd, 2005; Losh et al., 2008) or because individuals begin behaving more similarly over time (Lakin et al., 2003) is not clear. Understanding the direction of this effect would necessitate assessing the presence of these traits prior to relationship formation and tracking their influence on the development of the relationship over time. Given that the social features of autism can impede the formation and maintenance of relationships (Mundy and Sigman, 1989), such information may illuminate whether these difficulties extend to individuals with certain BAP traits, and if so, help identify the process by which they influence relationship quality and longevity.

In an effort to explore possible associations between autism-related traits and future relationship outcomes, this study assessed the self-reported BAP traits of pairs of college roommates with minimal familiarity prior to cohabitation at the beginning of the semester, as well as their satisfaction with, and commitment to, the roommate relationship after 9–10 weeks of cohabitation. By examining the associations between BAP traits and future relationship outcomes within newly formed relationships, this design enabled us to avoid the “direction of effect” problems inherent in the study of BAP traits within preformed friendships or romantic partnerships. Furthermore, we chose to move beyond simply assessing the association between BAP traits and relationship outcomes to also measure factors that may underlie how these associations may develop. According to Back et al. (2011), the effects of individuals’ broad and stable personality dispositions (e.g. BAP traits) on specific relationship outcomes (e.g. relationship satisfaction) should be mediated by social behaviors and related perceptions. Because interpersonal theorists and researchers have cogently argued that the dimensions of warmth (i.e. behaviors that foster connectedness and closeness with another person) and dominance (i.e. behaviors that maintain independence and control in a relationship) can be used to succinctly describe social behavior (Horowitz et al., 2006; Leary, 1957; Pincus, 2005), we investigated the degree to which the expression and elicitation of these behaviors within the roommate relationship can help to explain associations between BAP traits and relationship quality. Warmth or dominance within an interpersonal interaction typically invites predictable and complementary behavioral responses (e.g. warmth is met by warmth; Sadler and Woody, 2003), with negative affect often occurring as a result of noncomplementary reactions, such as warmth being met by coldness (Horowitz et al., 2006) or dominant behaviors being met with dominant responses (Shechtman and Horowitz, 2006). Within college roommates, individuals experiencing more complementarity of warmth and dominance report feeling more connected to their roommate (Ansell et al., 2008), and high quality, connected relationships can be measured through reports of relationship satisfaction and commitment to the relationship (Canevello and Crocker, 2010).

These processes were examined using a series of APIMs. Based on prior work assessing the BAP and relationships (e.g. Jobe and White, 2007; Losh et al., 2008; Wainer et al., 2013), we predicted that participants with higher levels of BAP traits would report lower relationship satisfaction and commitment to their roommate relationship. Similarly, we predicted that roommates of high BAP individuals would report less satisfaction with the relationship and less commitment to continue the relationship, especially when these roommates have lower levels of BAP traits themselves. We further hypothesized that the association between BAP traits and relationship quality would be mediated by expressions of warmth and dominance. Specifically, we predicted that participants with higher levels of BAP traits would be perceived as less warm by their roommates, that their roommates would behave less warmly toward participants with higher levels of BAP traits, and that the overall level of relationship satisfaction and commitment to the relationship would be lower for both individuals. Finally, given prior research suggesting that social but not nonsocial BAP traits are associated with reduced social cognition and social skill (Sasson et al., 2013c), we predicted that these effects would primarily occur for social BAP traits.

Method

Data analysis models

To evaluate our hypotheses regarding how BAP traits of unfamiliar roommates are related to their interpersonal behaviors and relationship quality, we specified a set of APIMs for indistinguishable members using Multilevel Modeling with Restricted Maximum Likelihood estimation in Statistical Package for Social Sciences (SPSS). Figure 2 presents the paths connected with each of our hypotheses. The prefix letters of “A,” “P,” and “D” represent actor effects, partner effects, and actor–partner interaction effects (i.e. interaction terms were created from both individuals’ scores of the same BAP trait; e.g. Roommate 1’s Aloofness × Roommate 2’s Aloofness), respectively. For the remainder of the manuscript, actor effects are synonymous with effects attributed to a “participant,” and partner effects are referenced as effects attributed to a “roommate.”

OPEN IN VIEWER

Figure 2.

Conceptual model of pathways predicting relationship outcomes through broad autism phenotype traits and interpersonal behaviors.

For our first set of hypotheses, we predicted that participants with higher levels of BAP traits would report lower relationship satisfaction and commitment to the relationship (i.e. the “A3” path in Figure 2); that participants with roommates possessing high BAP traits would report less satisfaction with the relationship and less commitment to continue the relationship (i.e. the “P3” path in Figure 2); and that an interaction would occur where a low BAP participant would be less satisfied and committed when paired with a high BAP roommate (i.e. the “D2” path in Figure 2).

Drawing upon interpersonal theory (Horowitz et al., 2006; Leary, 1957; Pincus, 2005), we further hypothesized that the association between BAP traits and relationship quality would be mediated by expressions of warmth and dominance. There are several potential mediational pathways to consider when actor and partner effects are estimated for both the predictors and the mediators. Each pathway has been designated by a specific name that corresponds to the relationship between variables it contains. The Personal Relationship Shaping pathway involves only actor effects; as such, it captures the process by which participants’ traits predict their own interpersonal behaviors and thereby their own relationship quality (the “A1-A2” pathway in Figure 2). The Interpersonal Relationship Shaping pathway involves both actor effects and partner effects and reflects how the associations between BAP traits and the interpersonal behaviors exchanged within the relationship shape the relationship qualities of the other person (i.e. the “A1-P2” and “P1-A2” pathways in Figure 2). The Evocative Relationship Shaping pathway involves only partner effects and captures how evoked reactions from roommates subsequently predict the participants’ relationship quality (i.e. the “P1-P2” pathway in Figure 2). Moreover, our inclusion of the dyad-level interaction term allows for additional mediational pathways to exist. Specifically, the Synergistic Relationship Shaping pathway explores how the interaction of the same BAP traits between partners could influence participants’ expressions of interpersonal behaviors, which then serve to shape either their own relationship quality (i.e. the “D1-A2” pathway in Figure 2) or the relationship quality of their roommate (i.e. the “D1-P2” pathway in Figure 2). To be comprehensive, we explore all of these mediational pathways. Finally, although we did not specifically hypothesize that any of our predicted effects would differ as a function of the gender of same-sex roommate pairs, we re-ran all analyses with gender included to explore whether the effects of BAP traits on relationship outcomes differed between male and female roommate dyads.

Results

Table 1 contains descriptive statistics and bivariate correlations between all primary study variables within and between roommates. Participants’ BAP traits were not significantly related to their roommates’ BAP traits, which is not unexpected given their unfamiliarity at the beginning of the study. In contrast, measures of warmth and relationship satisfaction completed after 9–10 weeks of cohabitation demonstrated evidence of interdependence between roommates (i.e. roommates’ ratings of satisfaction and commitment were similar within dyads), underscoring the importance of using dyadic models that appropriately account for this non-independence. Furthermore, as shown in Table 1, pragmatic language was the only BAP trait to show a negative association between a participants’ trait and their roommate’s relationship satisfaction and commitment.

OPEN IN VIEWER

Table 1.

Zero-order correlations within and between roommates for all study variables (top) and descriptive statistics for study variables across males and females (bottom).

Variable	1	2	3	4	5	6	7
1. Aloof	0.15 †	0.11	0.06	−0.15 †	−0.17*	−0.11	−0.06
2. Prag Lang	0.41**	0.07	−0.06	0.04	−0.07	−0.17*	−0.16*
3. Rigid	0.46**	0.17*	0.08	−0.08	−0.01	−0.06	−0.05
4. Warmth	−0.34**	−0.19*	−0.12	0.53**	−0.14 †	0.60**	0.46**
5. Dominance	−0.43**	−0.18*	−0.30**	−0.04	0.22**	−0.10	−0.08
6. Rel Sat	−0.07	−0.02	−0.04	0.47**	−0.10	0.68**	0.52**
7. Commitment	−0.08	−0.04	−0.01	0.35**	−0.08	0.79**	0.41**
Mean (SD)	2.84 (0.87)	2.93 (0.62)	3.09 (0.74)	−0.03 (0.91)	−0.04 (0.88)	4.14 (0.89)	6.52 (1.81)
Males	2.86 (0.85)	2.87 (0.62)	2.99 (0.75)	−0.10 (0.83)	−0.11 (0.77)	4.07 (0.89)	6.35 (1.83)
Females	2.81 (0.92)	3.02 (0.62)	3.24 (0.70)	0.07 (1.01)	0.07 (1.00)	4.24 (0.88)	6.73 (1.78)

SD: standard deviation.

Values in boldface along the diagonal reflect between-person correlations on the same variable. Correlations below the diagonal reflect associations within persons. Correlations above the diagonal reflect associations between persons. Prag Lang = Pragmatic Language; Rel Sat = Relationship Satisfaction; Warmth and Dominance are informant reports of interpersonal behavior provided by the roommate.

†

p < 0.10; *p < 0.05; **p < 0.01.

We first specified APIMs to evaluate relations between participants’ and roommates’ self-reported BAP characteristics (e.g. social aloofness) and the participants’ relationship outcomes (i.e. relationship satisfaction and commitment). Moreover, interaction terms for BAP traits were created and used as predictors to assess whether some combination of BAP levels within dyads (e.g. high BAP + low BAP) predicts relationship quality. For this and all other analyses, predictor variables were grand-mean centered to reduce their multicollinearity with interaction terms and unstandardized coefficients are reported.

As can be seen in Table 2, the actor–partner aloof interaction term was a significant predictor of relationship satisfaction. To break this down, we computed simple slopes for the effect of the roommates’ aloofness on the participants’ relationship satisfaction for both low aloof and high aloof participants (i.e. 1 SD below and above the mean, respectively; see Aiken and West, 1991). Simple slope analyses revealed that low aloof participants were less satisfied with a high aloof roommate and more satisfied with a low aloof roommate (b = −0.34, 95% confidence interval (CI): (−0.61, −0.06), standard error (SE) = 0.14, p = 0.016). Similarly, high aloof participants reported less satisfaction with a low aloof roommate and more satisfaction with a high aloof roommate (b = 0.30, 95% CI (0.00, 0.61), SE = 0.15, p = 0.049). Figure 3 further clarifies how the relation between participants’ reports of relationship satisfaction and their roommates’ levels of aloofness differ for participants (i.e. actors) who are low versus high on aloofness. As can be seen, less aloof participants were more satisfied when their roommate was also low on aloofness. In contrast, more aloof participants were more satisfied when their roommate was higher in aloofness. Taken together, these findings suggest that roommate pairs characterized by differing levels of aloofness tend to be less satisfied with their relationships after 9–10 weeks of relationship development.

OPEN IN VIEWER

Table 2.

Actor and partner broad autism phenotype characteristics predicting relationship outcome variables.

	Outcome variables
	Relationship satisfaction		Commitment
	B	SE	B	SE
Actor aloofness	−0.02	0.10	−0.11	0.20
Partner aloofness	−0.02	0.10	0.11	0.20
Actor–partner aloofness interaction	0.37*	0.13	0.47 †	0.24
Actor pragmatic language	0.00	0.12	−0.05	0.26
Partner pragmatic language	−0.23 †	0.12	−0.50 †	0.26
Actor–partner pragmatic language interaction	−0.13	0.26	−0.37	0.50
Actor rigidity	0.03	0.11	0.07	0.23
Partner rigidity	0.02	0.11	−0.02	0.23
Actor–partner rigidity interaction	0.07	0.16	−0.12	0.32

SE: standard error of unstandardized regression coefficient.

†

p < 0.10; *p < 0.01.

OPEN IN VIEWER

Figure 3.

Predicting participants’ relationship satisfaction as a function of their own (i.e. the actors’) and their roommates’ (i.e. the partners’) levels of aloofness.

We next investigated whether interpersonal behaviors (i.e. warmth and dominance) mediated the effects of the BAP characteristics on each relationship outcome variable. To obtain the relevant parameters to address our mediational inquiries (i.e. A1, P1, D1, A2, and P2 in Figure 2), we performed two sets of analyses. In the first set of analyses presented in Table 3, we specified APIMs to evaluate the relations between participants’ and their roommates’ BAP traits on average expressions of interpersonal behavior. As hypothesized, participants’ aloofness was negatively associated with both warmth (b = −0.34, 95% CI (−0.53, −0.15), SE = 0.10, p = 0.001) and dominance (b = −0.35, 95% CI (−0.53, −0.18), SE = 0.09, p < 0.001), indicating that more aloof individuals behaved less warmly and less dominantly toward their roommate. However, contrary to hypotheses, pragmatic language was not significantly related to interpersonal behaviors. Rigidity was also unrelated to interpersonal behaviors.

OPEN IN VIEWER

Table 3.

Actor and partner broad autism phenotype characteristics predicting interpersonal behaviors.

	Dispositional expressions of interpersonal behavior
	Warmth		Dominance
	B	SE	B	SE
Actor aloofness	−0.34*	0.10	−0.35*	0.09
Partner aloofness	−0.14	0.10	−0.16 †	0.09
Actor–partner aloofness interaction	0.18	0.12	0.05	0.10
Actor pragmatic language	−0.08	0.12	0.03	0.11
Partner pragmatic language	0.21 †	0.12	0.04	0.11
Actor–partner pragmatic language	0.11	0.25	0.20	0.20
Actor rigidity	0.08	0.11	−0.17 †	0.10
Partner rigidity	−0.02	0.11	0.08	0.10
Actor–partner rigidity interaction	0.10	0.16	0.06	0.13

SE: standard error of unstandardized regression coefficient.

†

p < 0.10; *p < 0.01.

The second set of analyses used APIMs to predict participants’ relational outcomes as a function of their own and their roommates’ BAP characteristics and interpersonal behaviors (see Table 4). The actor–partner aloof interaction term still significantly predicted relationship satisfaction (b = 0.24, 95% CI (0.05, 0.43), SE = 0.10, p = 0.016). Additionally, although nonsignificant when mediators were excluded from the model, here roommates’ levels of pragmatic language abnormalities predicted participants’ relationship satisfaction (b = −0.24, 95% CI (−0.43, −0.05), SE = 0.10, p = 0.016) and commitment (b = −0.51, 95% CI (−0.96, −0.05), SE = 0.23, p = 0.031).

OPEN IN VIEWER

Table 4.

Relationship outcome variables predicted by actor and partner BAP characteristics and interpersonal behaviors.

	Outcome variables
	Relationship satisfaction		Commitment
	B	SE	B	SE
Actor aloofness	0.13	0.09	0.14	0.21
Partner aloofness	0.17 †	0.09	0.45*	0.21
Actor–partner aloofness interaction	0.24*	0.10	0.26	0.21
Actor pragmatic language	−0.08	0.10	−0.21	0.23
Partner pragmatic language	−0.24*	0.10	−0.51*	0.23
Actor–partner pragmatic language interaction	−0.21	0.20	−0.50	0.42
Actor rigidity	0.02	0.09	0.07	0.20
Partner rigidity	−0.02	0.09	−0.09	0.20
Actor–partner rigidity interaction	−0.00	0.12	−0.24	0.27
Actor warmth	0.23**	0.07	0.32 †	0.18
Partner warmth	0.47**	0.07	0.84**	0.18
Actor dominance	0.02	0.07	0.05	0.17
Partner dominance	−0.01	0.07	0.01	0.17

SE: standard error of unstandardized coefficient.

†

p < 0.10; *p < 0.05; **p < 0.01.

Indirect effects (ab) were calculated for all mediational pathways associated with Figure 2 and a full list of these effects can be found on OSF. Indirect effects are usually computed by multiplying the paths that together comprise a mediational pathway. They convey the extent to which the proposed mediator transmits the “effects” of the predictor to the outcome. To evaluate whether indirect effects significantly differed from zero in our study, we derived CIs for these indirect effects using the Monte Carlo Method for Assessing Mediation (MCMAM; with 20,000 repetitions; Selig and Preacher, 2008). We also present effect sizes for each indirect effect in the form of the index of mediation (ab_cs; i.e. the completely standardized indirect effect), where values of |0.01|, |0.09|, and |0.25| are considered small, medium, and large, respectively.

As hypothesized, participants’ aloofness predicted their own expressions of warmth, which subsequently predicted their satisfaction with their roommate relationships (ab = −0.08, 95% CI (−0.15, −0.02), ab_cs = −0.08). This effect was medium in size, suggesting that participants’ expressions of warmth constitute an important mechanism by which their aloof BAP traits predict their own relationship satisfaction. In addition, participants with greater levels of aloofness behaved less warmly toward their roommates, and this led their roommates to feel less satisfied (ab = −0.16, 95% CI (−0.26, −0.07), ab_cs = −0.16) and less committed (ab = −0.28, 95% CI (−0.50, −0.11), ab_cs = −0.14). These medium to large size effects suggest that perceived expressions of warmth from participants contribute importantly to how participants’ aloof BAP traits are associated with their roommate’s feelings of relationship satisfaction and commitment. There were no significant indirect effects involving dominance as a mediator, and neither pragmatic language nor rigidity had significant indirect effects predicting either relationship outcome with either mediator. Similarly, none of the actor–partner interaction terms had significant indirect effects for either outcome.

Finally, we re-conducted all analyses with gender included. Importantly, all previously reported effects maintained their direction and statistical significance. Out of all the possible main effects and interactions with gender, 7 interactions or 10.3% of all possible effects reached statistical significance. Briefly, males with more pragmatic language abnormalities were less satisfied and expressed less warmth with their roommate and also had less satisfied and more dominant roommates. Additionally, males with more pragmatic language abnormalities were more satisfied with roommates with similar levels of this BAP trait. Interestingly, unlike males, females behaved more warmly if their roommate expressed more pragmatic language abnormalities. Finally, low aloof females were more committed to low aloof roommates and less committed to high aloof roommates. In summary, these results indicate that pragmatic language may be more informative for the formation of relationships between male roommates compared to females. These results, however, should be interpreted with caution given their exploratory nature. All analyses and results of gender interactions can be found on the OSF. We also re-conducted all analyses controlling for relationship satisfaction or commitment at baseline to account for the possibility that stability in the relationship outcomes may provide an alternative explanation for our findings. Results were the same with a few exceptions (see tables on the OSF): (a) participants’ rigidity significantly predicted their dominance; and (b) roommates’ pragmatic language now significantly predicted participants’ expressions of warmth, as well as their levels of commitment at Time 5, but became a nonsignificant (albeit marginal) predictor of their relationship satisfaction.

Discussion

Although previous research has demonstrated that the BAP is associated with reduced relationship quality within established relationships (e.g. Jobe and White, 2007; Pollmann et al., 2010; Wainer et al., 2013), understanding the processes by which this association emerges requires assessment prior to relationship formation. This study leveraged a unique population, previously unfamiliar college roommates, to examine how self-reported BAP traits predict relationship quality and interpersonal behaviors later in the relationship. Whereas the nonsocial BAP trait of rigid personality showed limited associations with interpersonal behaviors between roommates and relationship outcomes, the social trait of aloof personality demonstrated robust relations.

Specifically, participants who were mismatched on aloofness—the participant being low and the roommate being high—reported less relationship satisfaction after a few months of cohabitation. In contrast, high aloof participants that were paired with similarly high aloof roommates reported higher relationship satisfaction. Given that high aloof individuals remained satisfied in their relationships when paired with other high aloof roommates, satisfaction does not appear predicated upon the presence of aloofness generally but rather reflects a product of dissimilarity in aloof profiles between roommates. Because aloofness corresponds to reduced social inclinations and an increased preference of solitary activity, a mismatch between roommates on this BAP trait may indicate some discordance in their social needs and expectations. Likewise, shared aloof profiles between roommates may suggest overlap in these characteristics, with both high and low aloof roommate pairs finding their relative desire for sociability met by each other. Consistent with this interpretation, previous studies of long-term close friendships and romantic relationships report moderate correlations of BAP traits between individuals within the dyad (Constantino and Todd, 2005; Losh et al., 2008; Wainer et al., 2013), though it remains unclear whether this reflects a tendency for similarities to drive relationship formation or rather an increase of similarity over time. The current findings suggest that, at least within roommate dyads, complementary levels of aloofness may be particularly important for early relationship development. Examining whether this finding extends to the development of romantic relationships is an open question worthy of future research.

Whereas aloofness between roommate pairs was significantly associated with relationship satisfaction when mediating interpersonal behaviors were considered, pragmatic language abnormalities, the other social BAP trait measured in this study, was not. This was contrary to our prediction, and may indicate functional differences in relationship outcomes between the two social subscales of the BAPQ (Wainer et al., 2011). Unlike aloofness, pragmatic language abnormalities do not relate to a propensity to be social, but rather a competence in navigating the conventions of social discourse. This distinction between social motivation and social understanding may be especially relevant to early relationship dynamics, with a mismatch between roommates on aloofness uniquely predictive of later reductions in relationship quality. Aloofness but not pragmatic language abnormalities may manifest in a decreased effort to establish and maintain a relationship over time, an interpretation that is consistent with previous studies that report differences in how aloofness and pragmatic language abnormalities affect social interactions (Lamport and Turner, 2014; Wainer et al., 2011). Furthermore, these results expand upon previous findings indicating that social and nonsocial characteristics of the BAP relate to distinct behaviors and abilities (Baron-Cohen et al., 2001; Losh et al., 2009; Sasson et al., 2013c) by suggesting that even within the social domain, BAP traits can be productively subtyped further into motivational (aloof) and competence (pragmatic language) traits that may serve separate behavioral and social functions.

To explore how a mismatch between roommates on aloofness might result in reduced relationship satisfaction, this study examined the possible mediating effects of warmth and dominance (Horowitz et al., 2006). The results indicated that the association between individuals’ aloofness and their own relationship outcomes was mediated by the degree of warmth expressed between roommates. More aloof participants were less warm toward their roommates. As a result, warmth mediated the degree to which participants’ aloofness related to their own and their roommate’s relationship outcomes, with their warmer behaviors toward their roommate leading to greater relationship satisfaction for themselves and greater satisfaction and commitment for their roommates. Given that social aloofness is characterized by reduced social interest and a mismatch on aloofness between roommates is associated with poorer relationship outcomes, these results suggest that specific interpersonal behaviors may serve as a mechanism by which aloofness can affect relationship development. Specifically, relationships comprised of individuals mismatched on aloofness may engage in noncomplementary combinations of warm behaviors that over time may lead to distancing behaviors between social partners.

Furthermore, once the interpersonal behaviors of warmth and dominance were entered into the model, roommates of participants high on pragmatic language abnormalities reported lower relationship satisfaction and commitment at the end of the study. Interestingly, the magnitudes of these effects were remarkably similar across analyses (see Tables 2 and 4). This suggests that other mediators, unaccounted for by this study, contribute to the processes by which lower self-reported social competence is associated with reduced roommate relationship satisfaction and commitment. If future work is able to identify these mediators, such information could be combined with the results found for aloofness to present a more comprehensive account of the collective processes underlying how individuals with higher social BAP traits experience relationships that are lower in quantity and quality (Jobe and White, 2007; Losh et al., 2008; Piven et al., 1997; Wainer et al., 2013).

These findings should be interpreted in the context of several limitations. First, although the use of relatively unfamiliar roommates avoids the potential confounds of examining BAP traits within preformed relationships, the processes by which the BAP influences roommate dynamics may differ for other types of relationships (e.g. voluntary friendships and romances, family relationships). Second, college students cohabitating with relatively unfamiliar roommates may differ socially from those who are able to room with someone from an established relationship. Thus, it is possible that maximizing unfamiliarity in our sample came at some cost to representativeness. Relatedly, given that endorsement of BAP features can vary based upon cultural norms (Freeth et al., 2013), the reporting of BAP traits by the ethnically diverse sample in this study may differ relative to the predominantly Caucasian samples often found in previous BAPQ studies (Sasson et al., 2013a). We also did not collect data on socioeconomic status, which potentially could impact the quality of roommate relationships. Third, this study relied upon self-reports of the BAPQ, which may underreport BAP traits relative to informant report (Sasson et al., 2014). However, if this bias occurred here and was consistent across all participants, all rank-order relations would be preserved and thus not affect the regression analyses. In retrospect, collecting informant reports from an individual on his or her roommate at the end of the study may have been particularly informative for understanding how one’s perception of these personality characteristics relates to relationship outcomes and aligns with the participant’s own self-report. Future studies may want to include alternative assessment methods of the BAP, including informant report and objective clinical ratings. Fourth, roommates had limited contact but were not completely unfamiliar with each other prior to data collection, which would have required data collection immediately upon enrollment. It is possible that results may have differed if initial data collection occurred prior to any contact between roommates. Furthermore, the nature of prior contact (e.g. meetings, social media) was not assessed, which may relate to differences in the quality of initial familiarity and roommate formation. Fifth, future studies may choose to examine more deeply whether BAP traits impact relationship development differently for males and females. In our study, complementary profiles of pragmatic language between roommates were related to relationship outcomes for males only, with few other gender effects emerging. However, given prior research demonstrating gender differences in the BAP and in ASD, further exploration of the effect of BAP traits on relationships for males and females beyond same-sex roommate pairings is encouraged. Sixth, although we chose to investigate the interpersonal behaviors of warmth and dominance because of their perceived importance to relationship development, these behaviors do not encompass all of the potential variables that might mediate how BAP traits relate to relationships (e.g. Lamport and Turner, 2014; Pollmann et al., 2010). Future research may choose to explore alternate variables that may account for other effects. Finally, given the large number of tests performed, replication is encouraged to verify the robustness of our findings.

These limitations notwithstanding, this study demonstrates how the BAP in the general population is associated with the quality of newly formed relationships. By examining roommate dyads with limited familiarity previous to cohabitation, associations between self-reported BAP traits and later relationship outcomes were examined free of the confounding interdependence inherent in the study of preformed relationships. This design enabled the identification of aloofness as particularly relevant to relationship outcomes. Roommates matched on aloofness, either both high or both low, indicated higher relationship satisfaction after 10 weeks of cohabitation than roommates mismatched on aloofness. The preference for similarity on aloofness suggests that complementary profiles of social motivation may facilitate later relationship satisfaction. These effects did not extend to pragmatic language abnormalities, suggesting that similarity in social motivation but not social competence is associated with later relationship satisfaction between roommates. Further analyses indicated that warmth mediated the effects of aloofness on relationship outcomes, highlighting it as a specific interpersonal behavior that may serve as a mechanism by which aloofness influences relationship quality. Future studies that examine whether warmth and other interpersonal behaviors may underlie associations between BAP and relationship development in other contexts (e.g. friendships, romances, and family relationships) are warranted. Furthermore, because these findings collectively suggest that comparable levels of social motivation support positive relationship outcomes among previously unfamiliar individuals, colleges and other organizations may seek to increase relationship compatibility between assigned partnerships (e.g. roommates, work partners, and office mates) by matching dyads on this feature. Future research is also encouraged to determine whether the BAP results found here extend to the autism spectrum. If so, universities may be able to leverage such information to increase the likelihood of a positive transition to college for individuals with ASD by pairing them with compatible roommates, facilitating the development of social and support groups, and informing college disability services.