Causal role of trust in interpersonal attraction from attitude similarity

Abstract

The most popular method for investigating the mediation of the similarity-attraction link by trust involved first the manipulation of attitude similarity between the partner and the participant and then assessments of trust before attraction. Such correlational data precluded unambiguous inferences of causal flow of attitude similarity effects from trust to attraction. In the present study, we experimentally manipulated attitude similarity first and trust in the partner next, and measured trust in, and attraction toward, the partner after each manipulation. We found that similarity’s impact on trust remained stable over time but that on attraction declined substantially from the first to second assessment. Mediation analyses and structural equation modeling were consistent with a model in which causality flowed from attitude similarity to trust and then to attraction but not from attraction to trust. Findings also suggested that similarity can be expected to be secondary (or even redundant) when trust is already established.

Keywords

Mediation reverse causation similarity-attraction trust two-stage design

One of the central precursors of long- and short-term relationships is trust—a belief that the partners would consistently respond to and try to satisfy needs and goals of each other (Repel, Holmes, & Zanna, 1985). From this view, trust underpins companionate love (Holmes & Repel, 1989), intimacy in cross-sex and same-sex friendships (Monsour, 1992), dependency in intimate relationships (Wieselquist, Rusbult, Forster, & Agnew, 1999), marital satisfaction (Repel, Ross, & Holmes, 2001), and satisfaction between dating couples (Wieselquist, 2009). In the context of interdependent teams or groups, trustworthiness was also the single most sought after characteristic of an ideal partner (Cottrell, Neuberg, & Li, 2007). Van Lange (2015) further regarded trust as “the social glue in relationships, groups, and societies …” (p. 1).

Consistent with the central role of trust in general relationship formation and maintenance, the role of trust in two well-known interpersonal attraction phenomenon, reciprocated attraction (Backman & Secord, 1959; Gouldner, 1960) and similarity-attraction (Byrne, 1961), has also been examined. In the former, people are attracted to those people who are attracted to them (see, Gordon, 1996, for a meta-analysis; Montoya & Horton, 2012, for a review). Montoya and Insko (2008) demonstrated that such reciprocated liking is mediated by the belief in the benevolent intent of the partner. In the similarity-attraction effect, people are drawn to those who have similar attitudes, beliefs, characteristics, and/or values (see, e.g., Montoya & Horton, 2013; Montoya, Horton, & Kirchner, 2008, for meta-analyses). Singh et al. (2015) demonstrated that attitude similarity effects on short-term interpersonal attraction are also mediated by trust.

The mediational role of trust in the similarity-attraction link was demonstrated by measuring trust alone, alongside other identified mediators of positive affect (Byrne & Clore, 1970), inferred attraction of the partner toward the participant (Condon & Crano, 1988), and respect for the partner (Montoya & Horton, 2004), and with all three of them. In each case, similarity’s impact on attraction was weakened more when attraction was regressed on both the predictor and the measured mediating variable(s) than when attraction was regressed on only the predictor (see, e.g., Kenny, Kashy, & Bolger, 1998, for a discussion). Given such consistent evidence, Singh et al. (2015) regarded trust as a “new key mediator” of attitude similarity effects on attraction.

Whereas there can be no doubt about the causal role of trust in relationship formation, in general, and interpersonal attraction, in particular, the mediating role of trust in the similarity effect requires a more solid foundation than what has been reported to date (Singh et al., 2015). In the studies reported, both the mediator and outcome variables were measured responses to the manipulated attitude similarity. The correlational nature of such data precludes creating compelling arguments regarding the causal flow from the supposedly mediating to outcome variable (Bullock, Green, & Ha, 2010; Fiedler, Schott, & Meiser, 2011; Spencer, Zanna, & Fong, 2005). Further, whereas support for mediation of attitude similarity effects on attraction did come from the widely popular multi-equation regression framework (e.g., Kenny et al., 1998), rejection of the reverse effect of attraction on trust remained unaddressed, particularly because both the mediator and outcome variables were measured just once. In studies that reported trust as a mediator of attraction from different kinds of predictors, such as partner’s attraction (Montoya & Insko, 2008), personality traits (Singh et al., 2009), or attitude similarity (Singh et al., 2015), trust was not measured after attraction. Thus, the issue of reverse causation (Kenny et al., 1998; Singh, Ho, Tan, & Bell, 2007), that is, whether attraction also leads to trust remained heretofore untested.

In the case of similarity-attraction studies, the failure to rule out reverse causation has serious conceptual implications. First, the effect of similarity on respect was previously shown to be mediated by attraction as well (Singh, Ho et al., 2007; Singh, Lin, Tan, & Ho, 2008). Second, there is evidence of reverse causation for the similarity effect, as liking results in assumption and/or perception of high similarity with that person (Byrne, 1971; Morry, 2005, 2007). In this attraction-similarity effect, the higher the level of attraction (Marks & Miller, 1982) and the quality of relationship between two persons (Morry, Kito, & Ortiz, 2011), the greater is the assumed and/or perceived similarity between them. Such demonstrations of causality in both directions suggest that orders of measurement of the mediator and outcome variables should be always considered to eliminate the possibility of reverse causation. In line with the demonstrated importance of trust in interpersonal and group relationships (e.g., Cottrell et al., 2007; McAleer, Todorov, & Belin, 2014; Montoya & Insko, 2008; Murray & Holmes, 2009; Van Lange, 2015) and the centrality of trust in a general model of attraction (Montoya & Horton, 2014), we employed an experimental design that might not only demonstrate trust as an unambiguous mediator of attitude similarity effects on attraction but also investigate possible reverse causation from attraction to trust.

In spite of the popularity of mediation analyses from correlational data in management and social sciences, researchers (Bullock et al., 2010; Fiedler et al., 2011) and journal editors (Smith, 2012; Trafimow, 2015) have been skeptical of causal conclusions drawn. Whereas Tate (2015) emphasized the importance of temporal gap between measurement of the mediator and outcome variables, Spencer et al. (2005) recommended manipulating the mediator along with the predictor. That is, the predictor-outcome link should be weakened or nullified when the mediator is crossed with the predictor. However, such a result is not achievable as easily as it might seem if there is more than one mediator (Bullock et al., 2010; Clore & Baldridge, 1970; Fiedler et al., 2011; Singh, Lin et al., 2008). Accordingly, evidence for mediation from a measurement-of-mediator design is required nowadays to be corroborated in other ways, including experimentation (e.g., Montoya & Insko, 2008, Experiment 2).

To demonstrate that similar versus dissimilar attitudes predicted attraction independent of the impact of the degree to which similarity was assumed, Chen and Kenrick (2002) employed a two-stage experimental design. At Time 1, participants learned that their partner belonged to their in-group or out-group, inferred the attitudes their partner may hold and then indicated their attraction. At Time 2, participants examined an attitude survey that had views surreptitiously made similar to or dissimilar from their own in a previously completed attitude survey and then again reported their attraction. Assumed similarity with the in-group versus out-group partner at Time 1 moderated the impact of attitude similarity on attraction and of attitude dissimilarity on repulsion. This finding revealed that attitude similarity effects on attraction might have been mediated by assumed similarity of attitudes with the partner at Time 1.

A similar two-stage design may provide a stronger evidence for mediation of attitude similarity effects on attraction by trust than previously reported. Specifically, (a) attitude similarity with (i.e., the predictor) and (b) trustworthiness of (i.e., the mediator) the partner should be manipulated at Time 1 and Time 2, respectively. Moreover, trust should be assessed after each such manipulation and before the assessment of attraction. If trust mediates similarity’s impact on attraction, similarity should influence Trust₁ (i.e., trust assessed at Time 1) and Attraction₁ (Singh et al., 2015). Because information about their partner’s trustworthiness is made available later as a substitute for the trust inferred from a similarity manipulation, two patterns of results can be expected. First, the partner’s trustworthiness should influence Trust₂ (i.e., trust assessed at Time 2) and Attraction₂ measured but not the preceding Trust₁ and Attraction₁. Second, attitude similarity should still contribute somewhat to Trust₂ but become less effective at predicting Attraction₂ because of the trust manipulation (Spencer et al., 2005). Such patterns of significant and nonsignificant effects of similarity and trustworthiness on trust and attraction may provide a stronger experimental evidence for mediation of similarity on attraction by trust than from previously employed multi-equation regression framework with the data from a measurement-of-mediator design.

Two additional virtues of a two-stage design deserve mention. First, at Time 1, the impact of similarity on Attraction₁ should be reliably mediated by Trust₁ (as demonstrated previously, e.g., Singh et al., 2015). Second, whatever similarity effect remains on Attraction₂ should also be mediated by Trust₁. More interestingly, given the explicit information about their partner’s trustworthiness at Time 2, measured Trust₂ may serve more as a check on the trustworthiness manipulation rather than as a mediator per se. As Fiedler et al. (2011) pointed out, the effect of a predictor on a mediator also reflects on a simple correlation between the mediator and outcome variables or a spurious relationship arising out of the link between the predictor and an unmeasured variable related to the mediator.

Another virtue of the two-stage design lies in its potential to investigate the possibility of reverse causation, that is, whether attraction causes trust. Simply reversing arrows between attraction and trust in a statistical analysis is not an adequate test of reverse causation or mediation (Thoemmes, 2015). In our experiment, however, we hypothesize that similarity alone predict Trust₁ and Attraction₁, but at Time 2, both similarity and trustworthiness predict Trust₂ and Attraction₂. Thus, it may be possible to trace how the specific effects of the time-ordered manipulated variables and measured variables created Attraction₂. In Singh, Chen, and Wegener (2014), for example, the similarity-attraction link was represented better when positive affect, respect, and inferred attraction were treated as sequential processes (i.e., each mediator also building on each other) rather than parallel processes (i.e., each mediator acting independent of other mediators). Likewise, positive affect seemingly transmitted attitude similarity effects to attraction via the succeeding mediating variables of inferred attraction and/or trust (Singh et al., 2015, Note 5) in sequential mediation analyses. Therefore, we specifically predicted that Trust₁ should predict Trust₂ and Attraction₂ and that Attraction₁ should predict Attraction₂ (i.e., similar processes and/or causes lead to subsequent effects) but not Trust₂ (i.e., no reverse causation). It deserves emphasis that Attraction₁ was measured before Trust₂ in this study. Thus, the absence of a significant path from Attraction₁ to Trust₂ would cast doubt on reverse causal effect from attraction to trust.

Method

Design and participants

The design was a 2 (Attitude Similarity: similar, dissimilar) × 2 (Partner’s Trustworthiness: trustworthy, untrustworthy) between-participants factorial. Notably, partner’s trustworthiness was absent in this 2 × 2 design at Time 1, but both similarity and trustworthiness manipulations were present at Time 2.

Seventy-two female Chinese students (age range = 18–20 years; mean (M) = 18.99, standard deviation (SD) = 0.85) from an urban public university in Southeast Asia participated to fulfill their requirements for an introductory psychology module. We randomly assigned 18 participants to the four cells of the factorial design.

To ensure that our sample size had enough power to detect the hypothesized effects, we conducted power analysis with the G*Power software (Faul, Erdfelder, Buchner, & Lang, 2007). In general, .80 is regarded as a minimum acceptable power of an empirical test. In Experiment 1 of Singh et al. (2015), effect sizes (f) of attitude similarity on trust and attraction were .91 and .45, respectively, which can be regarded as rather large (Cohen, 1988). Given those effect sizes in the previous experiment and the present sample size, the power to detect corresponding similarity effects on trust and attraction was 1 and .96, respectively.

Attitude similarity manipulation

Attitudes were operationalized as responses to a 12-item attitude survey. Participants indicated their support (i.e., slightly for, moderately for, or strongly for) or opposition (i.e., slightly against, moderately against, or strongly against) to each of those controversial issues by selecting one of the six choices, without the neutral point (Byrne, 1971). The issues used were environmental protection, integrated resorts, retirement age, use of social networking websites, abortion, number of seats for foreign students in the public universities, protests and demonstrations, women in politics, demolishment of old buildings, compulsory accommodation on campus, gay and lesbian rights, and cohabitation.

Based on the participant’s responses to the first attitude survey, we prepared a bogus survey. Similar attitudes were on the same side of the scale and exactly one level upward or downward from the participant’s own response to an issue. However, dissimilar attitudes were three levels away from the participant own responses and on the other side of the scale. This method of manipulating similarity and dissimilarity of attitudes of the participant and the partner is the standard in the similarity-attraction studies (Byrne, 1971; Montoya & Horton, 2004; Singh et al., 2015).

Materials

In a Partner’s Opinion Questionnaire , we included three trust items (i.e., My partner would act benevolently toward me; … make me feel secure; and … look out for my interests) before three attraction items (i.e., I would like to meet my partner, I look forward to meeting my partner, and … get to know this person better). Responses to each item was sought along a 7-point rating scale, anchored by 1 (strongly disagree) and 7 (strongly agree).

The trust and behavioral attraction items were selected from the corresponding measures of previous studies (Montoya & Insko, 2008; Singh et al., 2009, 2015). As can be seen, the trust items assessed participants’ belief in the benevolent intent of their partner. In contrast, the attraction items assessed the degree to which participants desired to affiliate with their partner. The coefficients of reliability of both of these measures at Time 1 and Time 2 are presented in Table 1. We averaged the three responses to each measure to form an index of trust and of behavioral attraction. Responses ranged from 1 (lowest) to 7 (highest).

Table 1.

Reliability of and correlation between responses at Time 1 and Time 2.

	Time 1		Time 2
Time and responses	Trust₁	Attraction₁	Trust₂	Attraction₂
Time 1
Trust₁	.81	.41**	.58**	.55**
Attraction₁		.76	.29*	.46**
Time 2
Trust₂			.82	.55**
Attraction₂				.93

Note. The corresponding Cronbach alpha (α) is listed along the diagonal. The subscripts 1 and 2 refer to measures taken at Time 1 and Time 2, respectively. df = 70.

*p ≤ .05; **p ≤ .01.

Procedure

We conducted the study in two sessions. In the first session, participants completed an attitude survey and signed up for the second session with a stranger. In the second session, we informed participants that they would be interacting with another female student on a project. Before meeting her, however, “some judgments of that unknown student have to be made from the information presented.”

At Time 1, participants examined the bogus attitude survey for 2 min, formed an opinion of the partner for 1 min, completed the trust and attraction assessments, and then returned the completed booklet to the experimenter.

To manipulate trust of the participant in the partner at Time 2, we provided the participant with alleged opinions of four other students about the benevolent actions of their prospective partner. Participants were told that each of the four students “… had previously worked with her prospective partner in another experiment requiring mutual cooperation.” Each opinion about the partner’s trustworthiness was recorded on the 7-point scale below the statement of “I found the partner to be acting benevolently with me” (1 = strongly disagree; 7 = strongly agree). Thus, along a 7-point rating scale of trustworthiness, the four ratings were 5, 6, 6, and 4 for a trustworthy partner (M = 5.25, SD = 0.96) but 3, 2, 2, and 4 for the untrustworthy one (M = 2.75, SD = 0.96). To enhance realism of the manipulated opinions, ink colors of the four check marks, presented on four separate sheets of paper, were different. For the sake of credibility, the three opinions were on the intended side, but one opinion was neutral.

After learning the four opinions about trustworthiness of the partner, participants again evaluated the prospective partner for 1 min and then completed the trust and attraction measures. After collecting the second completed questionnaire, the experimenter informed participants that no interaction session would be scheduled. There was an interval of 2 min between the completion of Time 1 requirements and the manipulation of the partner’s trustworthiness. Participants did not indicate any awareness of the hypotheses tested. Before terminating the session, the experimenter debriefed the participants fully and thanked them.

Results

Construct distinction

To distinguish Trust₁ from Attraction₁, we first performed a two-factor confirmatory factor analysis (CFA) of the six responses of Time 1 in AMOS with a correlation between the two factors. The two-factor measurement model provided a satisfactory fit to the data: χ²(8) = 10.44, p = .24, non-normed fit index/Tucker–Lewis index (NNFI/TLI) = .97, incremental fit index (IFI) = .99, root mean square error of approximation (RMSEA) = .07, and standardized root mean residual (SRMR) = .04. In another one-factor CFA in which we specified the responses to the six items to be just one factor of general evaluation, by contrast, the fit was not only unsatisfactory, χ ² (9) = 53.90, p < .001, but also statistically worse, χ²Δ(1) = 43.46, p < .001.

We did two similar CFAs of the six responses of Time 2. Whereas the fit of the two-factor model was satisfactory, χ²(8) = 12.35, p = .14, NNFI/TLI = .97, IFI = .99, RMSEA = .09, SRMR = .05, that of a single-factor was unsatisfactory, χ²(9) = 54.75, p < .001; χ²Δ(1) = 42.40, p < .001. Consistent with these results, the correlations among the measures reported in Table 1 indicate that the two constructs were correlated but distinct.

Causal effects

We performed four separate 2 × 2 (Attitude Similarity × Partner’s Trustworthiness) analyses of variance on the trust and attraction responses of Time 1 and Time 2. None of the four responses resulted in a significant two-way interaction, Fs(1, 68) ≤ 1.60, ps ≥ .21, η_p ²s ≤ .02. In Table 2, we present Ms and SDs of the trust and attraction responses from Time 1 and Time 2.

Table 2.

Means (and SDs) of trust and attraction at Time 1 and Time 2.

	Attitudes		Trustworthiness
	Dissimilar	Similar	Untrustworthy	Trustworthy
Time 1
Trust₁	4.44^b (0.91)	5.24^a (0.57)	4.69 (0.99)	4.99 (0.66)
Attraction₁	4.04^b (0.86)	4.68^a (0.63)	4.37 (1.00)	4.34 (0.59)
Time 2
Trust₂	4.18^b (1.03)	4.89^a (0.91)	3.94^b (1.01)	5.13^a (0.61)
Attraction₂	4.16 (1.18)	4.50 (0.93)	4.04^b (1.36)	4.70^a (0.52)

Note. The value below the M is the corresponding SD. The two column means with different superscript letters differ significantly at

p ≤ .05. ns = 36 per cell.

At Time 1, similarity was the only predictor in the design. Thus, there were effects for only similarity on both Trust₁, F(1, 68) = 20.48, p < .001, η_p ² = .23, and Attraction₁, F(1, 68) = 12.66, p < .001, η_p ² = .16. No effect of the partner’s trustworthiness on Trust₁, F(1, 68) = 2.84, p = .10, η_p ² = .04, or Attraction₁, F(1, 68) = 0.02, p = .88, η_p ² = .00, displayed on the top right side of Table 2, was as expected because this predictor was absent at Time 1. Nevertheless, these null effects are methodologically important to demonstrate that the random assignment of participants to conditions was successful.

At Time 2, both attitude similarity and the partner’s trustworthiness should influence Trust₂ and Attraction₂. As expected, the partner’s trustworthiness had an effect on Trust₂, F(1, 68) = 43.81, p < .001, η_p ² = .39, and Attraction₂, F(1, 68) = 7.64, p = .007, η_p ² = .10. Furthermore, and as predicted, similarity had an effect on Trust₂, F(1, 68) = 15.61, p < .001, η_p ² = .19, but not on Attraction₂, F(1, 68) = 3.12, p = .08, η_p ² = .04. Notably, similarity’s impact on Trust₁ (η_p ² = .23) and Trust₂ (η_p ² = .19) were nearly equal over time, but the impact of similarity on Attraction₂ (η_p ² = .04) was considerably lower than that on Attraction₁ (η_p ² = .16).

Mediation analyses

We coded attitude similarity (1 for similarity and 0 for dissimilarity) and the partner’s trustworthiness (coded as 1 for trustworthy and 0 for untrustworthy) and conducted two separate mediation analyses by PROCESS Model 4 in SPSS (Hayes, 2013). The number of bootstrap resamples was set to 5,000.

We first tested the mediation of similarity on Attraction₁ by Trust₁. The indirect effect of Trust₁ on the relation between similarity and attraction was significant, indirect effect = 0.22, 95% confidence interval (CI) = 0.02, 0.59.

In the second mediational analysis, we tested mediation of similarity and trustworthiness effects on Attraction₂ via both Trust₁ and Trust₂. These analyses tested for mediation by Trust₁ and Trust₂ by holding the effect of one predictor constant while testing the impact of the other effect. As expected, the nonsignificant similarity effects on Attraction₂, B = 0.43, p = .08, were consistent with mediation by Trust₁, indirect effect = 0.40, 95% CI: 0.02, 0.94, but not by Trust₂, indirect effect = 0.25, 95% CI: −0.06, 0.68. As shown in Table 2, the effect of partner’s trustworthiness on Attraction₂ was significant, B = 0.67, t = 2.78, p = .007. Nonetheless, there was no evidence that partner’s trustworthiness was mediated by Trust₁, indirect effect = 0.15, 95% CI: 0.00, 0.49, or by Trust₂, indirect effect = 0.43, 95% CI: −0.14, 1.07. These findings are consistent with the conclusion that Trust₂ ceases to act as a mediator when partner’s trustworthiness was explicit.

Structural equation modeling

Although the foregoing mediation analyses demonstrated that mediation of the similarity effect by trust was consistent with our hypotheses, the issue of reverse causation (i.e., whether attraction also leads to trust) remained unexamined. Accordingly, we constructed a path model in AMOS in which the predictors of Attraction₂ were attitude similarity, the partner’s trustworthiness, Trust₁, Attraction₁, and Trust₂. We designated Trust₁, Attraction₁, and Trust₂ as Mediators 1, 2, and 3, respectively, because they each preceded Attraction₂. The fit of the model to the data was satisfactory, χ²(4) = 5.58, p = .23, NNFI/TLI = .96, IFI = .99, RMSEA = .08, SRMR = .05. We display the path coefficients in Figure 1. Four results require further elaboration.

Figure 1.

The path coefficients from the structural equation modeling of Attraction₂ on the two predictors of similarity and the partner’s trustworthiness manipulated at Time 1 and Time 2, respectively, and the preceding measured variables of Trust₁, Attraction₁, and Trust₂. The subscripts 1 and 2 refer to the responses measured at Time 1 and Time 2, respectively. *p ≤ .01.

First, similarity of attitudes reliably predicted Trust₁, t = 4.48, p < .01, which, in turn, predicted Attraction₁, t = 2.48, p = .01. Moreover, partner’s trustworthiness reliably predicted Trust₂, t = 6.69, p < .01, which, in turn, predicted Attraction₂, t = 2.16, p = .01. Second, Trust₁, t = 2.69, p < .01, and Attraction₁, t = 3.14, p < .01, also reliably predicted Attraction₂. Third, Trust₁, t = 5.26, p < .01, but not Attraction₁, t = 1.36, p = .17, predicted Trust₂, which is inconsistent with the possibility of reverse causation between trust and attraction in the similarity-attraction link. Finally, Trust₁, Attraction₁, and Trust₂ rendered the direct effects of similarity, t = −1.73, p = .08, and the partner’s trustworthiness, t = 0.82, p = .41, on Attraction₂ nonsignificant. The direct effect of similarity on Attraction₂ was negative, suggesting that it might have been suppressed by the manipulated partner’s trustworthiness. Overall, then, these results indicate that the similarity-attraction link was mediated by trust and not vice versa.

Discussion

In a two-stage experiment in which we manipulated similarity and the partner’s trustworthiness, we found that belief in the benevolent intent of the partner mediates similarity, but not partner’s trustworthiness, impact on attraction. Findings confirm that belief in the partner’s benevolent intent acts as a key mediator of attitude similarity effects on attraction (Singh et al., 2015). Three findings are particularly noteworthy.

First, attitude similarity effects on attraction are reduced when both similarity and partner’s trustworthiness are experimentally manipulated. That is, when information about the partner’s trustworthiness was readily available, the well-known similarity-attraction link (Byrne, 1961, 1971; Montoya & Horton, 2013; Montoya et al., 2008) was not as robust. In fact, a 75% reduction of similarity’s impact on Attraction₂ following the manipulation of trust is an experimental evidence for mediation of the similarity-attraction link by trust (Singh et al., 2015).

Second, Trust₁ activated by similarity mediated not only similarity’s impact on Attraction₂ but also the residual nonsignificant similarity effect on Attraction₂. However, Trust₂ did not mediate the impact of partner’s trustworthiness on Attraction₂. The absence of mediation of the trustworthiness effect on Attraction₂ by trust indicates that Trust₂ may have acted as either a manipulation check or as a correlate of attraction (Fiedler et al., 2011).

Finally, Trust₁, Attraction₁, respectively, and Trust₂ predicted Attraction₂ but Attraction₁ did not predict Trust₂. These findings are consistent with two conclusions: (a) the causal flow from trust to attraction and (b) attraction does not cause trust. Accordingly, trust can well be regarded as a key causal component in the general model of attraction (Montoya & Horton, 2014).

Previous studies (Singh, Yeo, Lin, & Tan, 2007; Singh et al., 2014) raised the possibility of sequential dependency among the mediators of the similarity-attraction link. Specifically, positive affect and trust emerged as mediators distal and proximal to attraction, respectively, in sequential mediation tests (Singh et al., 2015). Given those findings, we included the preceding responses of Trust₁ and Attraction₁ as if they were additional routes to Attraction₂. By opening these paths, we not only eliminated the reverse causation possibility (i.e., Attraction₁ predicting Trust₂) but also uncovered the sequential effect of Trust₁ on Trust₂ and that of Attraction₁ on Attraction₂. Although interest in sequential effects of attitudes on attraction has been demonstrated before (e.g., Byrne, Lamberth, Palmer, & London, 1969; Chen & Kenrick, 2002; Singh, Lin et al., 2008; Singh, Yeo et al., 2007), effects of the preceding responses on the succeeding similar responses were never pursued in ways reported here. Given recent advances in research methods and data analyses (e.g., Hayes, 2013), we consider modeling the role of the preceding responses in the formation and maintenance of existing relationships (Morry, 2005, 2007) as important challenges to future relationship researchers.

One implication of the present experimental evidence for the mediation of attitude similarity effects on attraction by trust lies in explaining the finding of Cottrell et al. (2007). When they investigated the attributes that made one an ideal partner across different sorts of interdependent groups, trustworthiness was the single most important characteristic. Although similarity between the participant and the ideal partner (i.e., similar to me) was also included among the list of characteristics available to the participants, similarity did not make the list of key desirable characteristics. Consequently, Cottrell et al. pointed out that the manipulation of similarity alone might have created “… the impression that similarity is more important variable for assessing others than it really is” (p. 229). In our experiment, the manipulated trust altered attitude similarity effects from measured Attraction₁ to Attraction₂ but not across Trust₁ and Trust₂. Thus, similarity is still as important in relationship formation as it was when Lazarsfeld and Merton (1954) observed homophily in American societies or Byrne (1961) demonstrated similarity-attraction among Texas college students. We further demonstrate that attitude similarity can lose out to trust in the partner because trust is a key mediator of the similarity effect.

Long-term and close relationships depend on perceived partner’s responsiveness that encompasses constructs such as trust in the partner, feeling secure with him or her, and experiencing partner’s regard, commitment, and care for the self (Clark & Lemay, 2010). Our demonstration of the centrality of trust in interpersonal attraction has real-life relevance too. A simple introduction as a benevolent person appears to be important in initiation of short-term relationships with him or her. Importantly, one’s benevolent intent can be conveyed by similar attitudes as illustrated in this experiment, physical similarity (DeBruine, 2002), other-profitable traits (Singh et al., 2009), or the way one says “Hello” to each other (McAleer et al., 2014). It is unsurprising, therefore, that participants “… allocate[d] large portions of limited resources to increase target trustworthiness” in one of the experiments by Cottrell et al. (2007, p. 225).

Four limitations of our experiment also deserve mention. One relates to our way of manipulating the partner’s trustworthiness. A person’s reputation is essentially what others say about him or her (Asch, 1946; Kelley, 1950; Singh, Onglatco, Sriram, & Tay, 1997). Given such consistent literature on impression formation, we manipulated the partner’s trustworthiness by supplying opinions of four peers who had allegedly worked in a cooperative task. Such manipulation resulted in predicted effects on Trust₂ and Attraction₂ but not on Trust₁ and Attraction₁ (see the values listed on the right side Table 2). On this basis, we concluded for mediation of attitude similarity effects on attraction by trust (Singh et al., 2015). However, reputation of one’s general trustworthiness may not necessarily imply his or her benevolent intent toward the participant. If so, no mediation of the partner’s trustworthiness effect on Attraction₂ by Trust₂ can also be interpreted in a way different from that of us. To resolve this issue, future work should directly manipulate the partner’s benevolent intent toward the participant as in Montoya and Insko (2008, Experiment 2).

Another limitation relates to the use of only women as participants. We used a sample of women first because of the practical constraint of more women than men in the participant population of undergraduate students in psychology modules across the globe (Chen & Kenrick, 2002; Montoya & Horton, 2004). More important, sex of the adult participants did not previously moderate attitude similarity effects on attraction (Singh, Ng, Ong, & Lin, 2008, Experiment 2). So, we expect the reported experimental evidence of mediation of attitude similarity effects on attraction by trust to generalize to men as well. Nevertheless, we recommend use of an equal number of men and women in future work to make the finding more robust.

Still another limitation pertains to the rather short duration of the study that did not use any distractor task between Time 1 and 2, unlike in Chen and Kenrick (2002). Asking for the same responses of trust in, and attraction toward, the partner twice might have created an obvious demand for changed responses. Given our principal goal of ascertaining effects of attitude similarity presented alone at Time 1 and those of attitude similarity followed by the partner’s benevolent intent at Time 2, we avoided any disruption in causal flow of attitude similarity effects from Time 1 to Time 2. Had the responses been contaminated by demand characteristics of the experiment, there could have been uniform changes in both trust and attraction responses across time. Contrary to this expectation, attitude similarity effects on the two trust responses were nearly equal. As we noted, changes over time were in the predicted effect sizes of attitude similarity on attraction alone. Furthermore, results from mediation and path analyses were at odds with the possibility of demand artifacts. Nonetheless, we recommend a similar online two-stage study of attraction in which the two predictors of attitude similarity and the partner’s benevolent intent are also embedded within different kinds of filler stimuli (Jia & Singh, 2009).

The final limitation pertains to the manipulation of only one of the multiple mediators of attitude similarity effects on attraction. Singh et al. (2015) showed sequential mediation of the similarity-attraction effect by positive affect, respect, inferred attraction, and trust. When there are multiple mediators, experimental manipulation of only one of them might not always weaken the predictor-outcome link (Fiedler et al., 2011). In the present case, we were interested in mediation of similarity-attraction effect by trust, a causal variable that was previously found to be the most important and proximal to attraction. It appears that an important mediator which is also proximal to the outcome variable can attenuate the original effects of the predictor. With experimental manipulation of other and distal mediators, however, evidence might not be as clear as reported in this article. What our findings suggest, nevertheless, is that the two-stage successive (Chen & Kenrick, 2002), relative to the simultaneous (Singh, Ho et al., 2007; Singh, Lin et al., 2008), manipulation of the predictor and the mediator holds a greater potential. Accordingly, we recommend increased use of the two-stage successive manipulations of the predictor and the mediator in future experimental tests of mediation.

Footnotes

Acknowledgments

The authors thank R Mathew Montoya and three anonymous reviewers for their extremely constructive comments on early versions of this manuscript.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported in part by the National University of Singapore (grant R-581-000-049-112) and the Indian Institute of Management Bangalore (grants 7384 and 7399E) to the first author.

References

Asch

S. E.

(1946). Forming impressions of personality. Journal of Abnormal and Social Psychology, 41, 258–290.

Backman

C. W.

Secord

P. F.

(1959). The effect of perceived liking on interpersonal attraction. Human Relations, 12, 379–384.

Bullock

J. G.

Green

D. P.

S. E.

(2010). Yes, but what’s the mechanism? (Don’t expect an easy answer). Journal of Personality and Social Psychology, 98, 550–558. doi:10.1037/a0018933

Byrne

(1961). Interpersonal attraction and attitude similarity. Journal of Abnormal and Social Psychology, 62, 713–715

Byrne

(1971). The attraction paradigm. New York, NY: Academic Press.

Byrne

Clore

G. L.

(1970). A reinforcement model of evaluative responses. Personality: An International Journal, 1, 103–128.

Byrne

Lamberth

Palmer

London

(1969). Sequential effects as a function of explicit and implicit interpolated attraction responses. Journal of Personality and Social Psychology, 13, 70–78.

Chen

F. F.

Kenrick

D. T.

(2002). Repulsion or attraction? Group membership and assumed attitude similarity. Journal of Personality and Social Psychology, 83, 111–125. doi:10.1037//0022-3514.83.1.111

Clark

M. S.

Lemay

E. P.

(2010). Close relationships. In Gilbert

D. T.

Fiske

S. T.

Lindzey

(Eds.), The handbook of social psychology (4th ed., Vol. 2, pp. 898–940). Boston, MA: McGraw-Hill.

10.

Clore

G. L.

Baldridge

(1970). The behavior of item weights in attitude-attraction research. Journal of Experimental Social Psychology, 6, 177–186.

11.

Cohen

(1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, N.J.: Erlbaum.

12.

Condon

J. W.

Crano

W. D.

(1988). Inferred evaluation and the relation between attitude similarity and interpersonal attraction. Journal of Personality and Social Psychology, 54, 789–797. doi:10.1037/0022-3514.54.5.789

13.

Cottrell

C. A.

Neuberg

S. L.

N. P.

(2007). What do people desire in others? A sociofunctional perspective on the importance of different valued characteristics. Journal of Personality and Social Psychology, 92, 208–231. doi:10.1037/0022-3514.92.2.208

14.

DeBruine

L. M.

(2002). Facial resemblance enhances trust. Proceedings: Biological Sciences, 269, 1307–1312.

15.

Faul

Erdfelder

Lang

A.-G.

Buchner

(2007). G*Power 3: A flexible statistical power analysis program for the social, behavioral, and biomedical sciences. Behavior Research Methods, 39, 175–191.

16.

Fiedler

Schott

Meiser

(2011). What mediation analysis can (not) do. Journal of Experimental Social Psychology, 47, 1231–1236. doi:10.1016/j.jesp.2011.05.007

17.

Gordon

R. A.

(1996). Impact of ingratiation on judgments and evaluations: A meta-analytic investigation. Journal of Personality and Social Psychology, 71, 54–70.

18.

Gouldner

A. W.

(1960). The norm of reciprocity: A preliminary statement. American Sociological Review, 25, 161–178.

19.

Hayes

A. F.

(2013). Introduction to mediation, moderation, and conditional process analysis. New York, NY: Guilford Press.

20.

Holmes

J. G.

Repel

J. K.

(1989). Trust in close relationships. In Hendrick

(Ed.), Review of personality and social psychology: Vol. 10. Close relationships (pp. 187–219). Newbury Park, CA: Sage.

21.

Jia

Singh

(2009). Asymmetrical attention allocation to dissimilar and similar attitudes. Journal of Experimental Social Psychology, 45, 1259–1265. doi:10.1016/j.jesp.2009.07.012

22.

Kelley

H. H.

(1950). The warm–cold variable in first impressions of persons. Journal of Personality, 18, 431–439.

23.

Kenny

D. A.

Kashy

D. A.

Bolger

(1998). Data analysis in social psychology. In Gilbert

D. T.

Fiske

S. T.

Lindzey

(Eds.), The handbook of social psychology (4th ed., Vol. 1, pp. 233–265). Boston, MA: McGraw-Hill.

24.

Lazarsfeld

P. F.

Merton

R. K.

(1954). Friendship as a social process: A substantive and methodological analysis. In Berger

Abel

Page

C. H.

(Eds.), Freedom and control in modern society (pp. 18–66). New York: Van Nostrand.

25.

Marks

Miller

(1982). Target attractiveness as a mediator of assumed similarity. Personality and Social Psychology Bulletin, 8, 728–735. doi:10.1177/0146167282084020

26.

McAleer

Todorov

Belin

(2014). How do you say ‘Hello’? Personality impressions from brief novel voices. PloS one, 9, e90779. doi:10.1371/journal.pone.0090779

27.

Monsour

(1992). Meanings of intimacy in cross-and same-sex friendships. Journal of Social and Personal Relationships, 9, 277–295. doi:10.1177/0265407592092007

28.

Montoya

R. M.

Horton

R. S.

(2004). On the importance of cognitive evaluation as a determinant of interpersonal attraction. Journal of Personality and Social Psychology, 86, 696–712. doi:10.1037/0022-3514.86.5.696

29.

Montoya

R. M.

Horton

R. S.

(2012). The reciprocity of liking effect. In Paludi

(Ed.), The psychology of love (pp. 39–57). Santa Barbara, CA: Praeger.

30.

Montoya

R. M.

Horton

R. S.

(2013). A meta-analytic investigation of the processes underlying the similarity-attraction effect. Journal of Social and Personal Relationships, 30, 64–94. doi:10.1177/0265407512452989

31.

Montoya

R. M.

Horton

R. S.

(2014). A two-dimensional model for the study of interpersonal attraction. Personality and Social Psychology Review, 18, 59–86. doi:10.1177/1088868313501887

32.

Montoya

R. M.

Horton

R. S.

Kirchner

(2008). Is actual similarity necessary for attraction? A meta-analysis of actual and perceived similarity. Journal of Social and Personal Relationships, 25, 889–922. doi:10.1177/0265407508096700

33.

Montoya

R. M.

Insko

C. A.

(2008). Toward a more complete understanding of the reciprocity of liking effect. European Journal of Social Psychology, 38, 477–498. doi:10.1002/ejsp.431

34.

Morry

M. M.

(2005). Relationship satisfaction as a predictor of similarity ratings: A test of the attraction-similarity hypothesis. Journal of Social and Personal Relationships, 22, 561–584. doi:10.1177/0265407505054524

35.

Morry

M. M.

(2007). The attraction-similarity hypothesis among cross-sex friends: Relationship satisfaction, perceived similarities, and self-serving perceptions. Journal of Social and Personal Relationships, 24, 117–138. doi:10.1177/0265407507072615

36.

Morry

M. M.

Kito

M. I. E.

Ortiz

(2011). The attraction-similarity model and dating couples: Projection, perceived similarity, and psychological benefits. Personal Relationships, 18, 125–143. doi:10.1111/j.1475-6811.2010.01293.x

37.

Murray

S. L.

Holmes

J. G.

(2009). The architecture of interdependent minds: A motivation-management theory of mutual responsiveness. Psychological Review, 116, 908–928. doi:10.1037/a0017015

38.

Repel

J. K.

Holmes

J. G.

Zanna

M. P.

(1985). Trust in close relationships. Journal of Personality and Social Psychology, 49, 95–112.

39.

Repel

J. K.

Ross

Holmes

J. G.

(2001). Trust and communicated attributions in close relationships. Journal of Personality and Social Psychology, 81, 57–64.

40.

Singh

Chen

Wegener

D. T.

(2014). The similarity-attraction link: Sequential versus parallel multiple-mediator models involving inferred attraction, respect, and positive affect. Basic and Applied Social Psychology, 36, 281–298. doi:10.1080/01973533.2014.912583

41.

Singh

L. J.

Tan

H. L.

Bell

P.A.

(2007). Attitudes, personal evaluations and interpersonal attraction: On the direct, indirect and reverse-causal effects. British Journal of Social Psychology, 46, 19–42. doi:10.1348/014466606X104417

42.

Singh

Lin

P. K. F

Tan

H. L.

L. J.

(2008). Evaluations, attitude similarity, and interpersonal attraction: Testing the hypothesis of weighting interference across responses. Basic and Applied Social Psychology, 30, 241–252. doi:10.1080/01973530802375052

43.

Singh

Ong

E. L.

Lin

P. K. F.

(2008). Different mediators for the age, sex, and attitude similarity effects in interpersonal attraction. Basic and Applied Social Psychology, 30, 1–17. doi:10.1080/01973530701665165

44.

Singh

Onglatco

M. L.

Sriram

Tay

A. G. B.

(1997). The warm–cold variable in impression formation: Evidence for the positive-negative asymmetry. British Journal of Social Psychology, 36, 457–477.

45.

Singh

Simons

J. J. P.

Young

D. P. C. Y.

Sim

B. S. X.

Chai

X. T.

Singh

Chiou

S. Y.

(2009). Trust and respect as mediators of the other- and self-profitable trait effects on interpersonal attraction. European Journal of Social Psychology, 39, 1021–1038. doi:10.1002/ejsp.605

46.

Singh

Wegener

D. T.

Sankaran

Singh

Lin

P. K.F.

Seow

M. X.

, … Shuli

(2015). On the importance of trust in interpersonal attraction from attitude similarity. Journal of Social and Personal Relationships, 32, 829–850. doi:10.1177/0265407515576993

47.

Singh

Yeo

S. E.

Lin

P. K. F

Tan

(2007). Multiple mediators of the attitude similarity-attraction relationship: Dominance of inferred attraction and subtlety of affect. Basic and Applied Social Psychology, 29, 61–74. doi:10.1080/01973530701331007

48.

Smith

E. R.

(2012). Editorial. Journal of Personality and Social Psychology, 102, 1–3. doi:10.1037/a0026676

49.

Spencer

S. J.

Zanna

M. P.

Fong

G. T.

(2005). Establishing a causal chain: Why experiments are often more effective than mediational analyses in examining psychological processes. Journal of Personality and Social Psychology, 89, 845–851. doi:10.1037/0022-3514.89.6.845

50.

Tate

C. U.

(2015). On the overuse and misuse of mediation analysis: It may be a matter of timing. Basic and Applied Social Psychology, 37, 235–246. doi:10.1080/01973533.2015.1062380

51.

Thoemmes

(2015). Reversing arrows in mediation models does not distinguish plausible models. Basic and Applied Social Psychology, 37, 226–234. doi:10.1080/01973533.2015.1049351

52.

Trafimow

(2015). Introduction to the special issue on mediation analyses: What if planetary scientists used mediation analysis to infer causation? Basic and Applied Social Psychology, 37, 226–234. doi:10.1080/01973533.2015.1064290

53.

Van Lange

P. A.

(2015). Generalized trust: Four lessons from genetics and culture. Current Directions in Psychological Science, 24, 71–76. doi:10.1177/0963721414552473

54.

Wieselquist

(2009). Interpersonal forgiveness, trust, and the investment model of commitment. Journal of Social and Personal Relationships, 26, 531–548. doi:10.1177/0265407509347931

55.

Wieselquist

Rusbult

C. E.

Foster

C. A.

Agnew

C. R.

(1999). Commitment, pro-relationship behavior, and trust in close relationships. Journal of Personality and Social Psychology, 77, 942–966.