Initial Evidence that Individuals Form New Relationships with Partners that More Closely Match their Ideal Preferences

different questions about the role ideal preferences and mate choice.Research largely using the first approach has provided some support for the notion that ideal preferences are positively associated with the qualities of romantic partners. For example, when women prefer more masculine men as romantic partners, they are indeed in relationships An important assumption in interpersonal attraction research asking participants about their ideal partner preferences is that these preferences play a role in actual mate choice and relationship formation. Existing research investigating the possible predictive validity of ideal partner preference, however, is limited by the fact that none of it has focused on the actual process of relationship formation. The current research recruited participants when single, assessed ideal partner preferences across 38 traits and attributes, tracked participants’ relationship status over 5 months, and successfully recruited the new partners of 38 original participants to assess their self-evaluations across the same 38 traits and attributes. Using multilevel modeling to assess the correspondence between ideal partner preferences and self-evaluations within couple, the results showed a positive within-couple association that was not accounted for by personality similarity or stereotype accuracy. We discuss these results with respect to the current literature on the predictive validity of ideal partner preferences in relationship formation.

Decades of published research have focused on interpersonal attraction, particularly the qualities and characteristics that people report being important to them in a romantic partner [13]. These ideal partner preferences are often detailed and encompass many domains (see [6]). Implicit in research on this topic is the assumption that ideal partner preferences are importantly involved in guiding mate search and mate choice. In other words, what people say they want in a future partner should be associated with the actual characteristics of their future partners. If this link did not exist, understanding the process of relationship formation would be scarcely informed by asking individuals about their ideal partner preferences. Indeed, in recent years an increasing amount of attention has been devoted to examining the role of ideal partner preferences in actual mate choice.
There are two approaches typically used in this growing body of research. The first approach assesses the degree of similarity between, for example, preferences for a given trait and how much a current romantic partner possesses that trait, and if this similarity is statistically different from zero. It is assumed that mate choice was guided in some way by preferences if the degree of similarity assessed is positive and statistically significant. The second approach assesses if the degree of similarity between preferences for a given trait and whether a potential partner possesses that trait predicts outcomes associated with mate choice. It is assumed that if higher vs. lower degrees of similarity predict, for example, romantic interest, then individuals may preferentially pursue potential partners because of a high degree of similarity between this person's qualities and the pursuer's preferences. Both approaches are similar in that they can be used to investigate the role of ideal partner preferences in mate choice. They also differ, however, in that the first approach investigates the degree of similarity between preferences and partner characteristics, whereas the second approach uses this degree of similarity as a predictor of outcomes associated with mate choice; each approach can thus be used to answer different questions about the role ideal preferences and mate choice.
Research largely using the first approach has provided some support for the notion that ideal preferences are positively associated with the qualities of romantic partners. For example, when women prefer more masculine men as romantic partners, they are indeed in relationships ORIGINAL RESEARCH REPORT Initial Evidence that Individuals Form New Relationships with Partners that More Closely Match their Ideal Preferences Lorne Campbell * , Kristi Chin * and Sarah C. E. Stanton * An important assumption in interpersonal attraction research asking participants about their ideal partner preferences is that these preferences play a role in actual mate choice and relationship formation. Existing research investigating the possible predictive validity of ideal partner preference, however, is limited by the fact that none of it has focused on the actual process of relationship formation. The current research recruited participants when single, assessed ideal partner preferences across 38 traits and attributes, tracked participants' relationship status over 5 months, and successfully recruited the new partners of 38 original participants to assess their self-evaluations across the same 38 traits and attributes. Using multilevel modeling to assess the correspondence between ideal partner preferences and self-evaluations within couple, the results showed a positive within-couple association that was not accounted for by personality similarity or stereotype accuracy. We discuss these results with respect to the current literature on the predictive validity of ideal partner preferences in relationship formation.
Keywords: ideal partner preferences; close relationships; relationship formation; attraction; multilevel modeling; longitudinal with more masculine-looking men [2,8]; age preferences for marriage partners reflect the actual ages of individuals' marital partners [18]; and men with greater occupational status both prefer and tend to marry more physically attractive women ( [12], but see [26], for an alternative explanation of these data [27]). In none of this research, however, were ideal preferences assessed in individuals prior to forming a romantic relationship (i.e., when they were not romantically attached), meaning it was not possible to determine the association between ideal preferences of participants when single and the qualities of the new romantic partner.
Research largely using the second approach, however, offers mixed results. For example, research using a speed-dating paradigm, wherein groups of men and women interact for short periods of time with a number of actual potential partners, has not found evidence that the similarity between ideal preferences reported prior to the speed-dating event and the qualities of speed-dating partners predicted attraction to these interaction partners [9,28]. Additionally, Eastwick, Finkel, and Eagly [10] found, in a laboratory setting, that participants were more attracted to a study confederate when his/her written profile was created to more closely match their own ideal partner preferences, but this link disappeared following an actual interaction between participants and the confederate. Eastwick and Finkel [9] have therefore suggested that individuals lack introspective awareness of what influences their actual mate choices (but see [21] and [14] for data suggesting that similarity between individuals' preferences and the qualities of interaction partners do predict actual choices using similar "get acquainted" paradigms).
A recent meta-analysis by Eastwick, Luchies, Finkel, and Hunt [11] was conducted to better determine the nature of the predictive validity of ideal partner preferences in the existing literature. The analytic strategy used the second approach described above, testing if the degree of similarity between ideal preferences and qualities of the partner (hypothetical, live interaction partners, or current romantic partners) predicted study outcomes. The results of this meta-analysis suggest that individuals are more satisfied with current romantic partners who more closely match their ideal preferences, and single individuals are more attracted to hypothetical partners who more closely match their preferences. The preferences of single individuals, however, do not appear to predict how attracted those individuals were to actual potential partners following live interactions with them. Eastwick et al. [9] concluded that ". . . just because participants claim to value particular qualities in a mate does not mean that they will preferentially pursue partners who possess such qualities" (p. 647). These results seem to strongly suggest an absence of predictive validity of ideal partner preferences in relationship initiation or formation.
A close inspection of all of the published studies included in this meta-analysis, however, found that essentially none of it directly assessed mate choice in the actual formation of new relationships [7]. Instead, the bulk of this research focuses on the influence of ideal partner preferences in (a) initial interpersonal attraction (sometimes referred to as early relationship initiation), or (b) established relationship processes. Only three studies arguably attempted to assess individuals' transitions into new relationships, two using speed-dating paradigms [1,9] and one that paired men and women together for short "dates" and measured subsequent interpersonal attraction [25]. Only Asendorpf et al. [1] reported the probabilities for various kinds of future contact among study participants, with rates of actual relationship formation being quite low (i.e., 6.6% at 6 weeks after the speed-dating event, and 4.4% at 1 year after the speed-dating event), meaning it was not possible to determine if participants entered new relationships with other speed-dating participants that more or less matched their preferences.
Regardless of the approach used in the study ideal preferences and mate choice, therefore, the lack of research on the process of relationship formation means that we are not yet able to determine (a) if individuals enter new relationships with others that match their ideal preferences more closely, or (b) if the degree of similarity between one's ideal preferences and the qualities of potential partners predicts relationship formation and development. The present research was designed to address question (a), or to ascertain the degree to which individuals enter relationships with partners who embody their ideal preferences. Specifically, by measuring individuals' ideal partner preferences across many traits and attributes when they were single (i.e., not currently romantically attached), and the self-evaluations of their new romantic partners across the same traits and attributes (for individuals who began a relationship when involved in the study), we were able assess the correspondence (similarity) between ideal partner preferences stated when single and the selfevaluations of new partners. A positive degree of correspondence would suggest that individuals do tend enter relationships with others that have qualities more similar to their own preferences, providing initial empirical evidence for the notion that what people say they want in a future partner is indeed associated with the actual characteristics of their future partners.

Method
This study was registered on the Open Science Framework (OSF). All of the study materials and procedures are publically available and can be accessed via the following link: https://osf.io/9gf4q [4]. The present research was approved by the University of Western Ontario Non-Medical Research Ethics Board.

Participants
We first recruited 450 participants through posters displayed on the University of Western Ontario campus and in local London, Ontario grocery stores; an advertisement placed in the university newspaper and the popular online classified venue Kijiji (www.kijiji.com); as well as word of mouth. To be eligible for participation in the study, individuals had to be single (i.e., not in a casual or serious romantic relationship) at initial testing. We excluded 24 participants at Time 1: Fourteen participants reported their own height or preferred ideal mate height to be above seven feet tall, heights that are exceedingly rare in nature, and they were removed due to concern that their information was completed incorrectly; five participants were removed because their responses showed anomalies such that they came from the same IP address on the same day within a very narrow timeframe (e.g., less than five minutes apart); and five participants were removed because they did not fully complete the survey.
Of the 426 remaining participants, 167 became involved in a romantic relationship over the 5-month period of the study; of these, 85 provided us with the contact information of their new partners. Of the new partners that we contacted, 45 accepted the invitation to participate in the study, though seven of the new partners recruited to participate in the study did not complete the survey.

Materials and Procedure
Single individuals interested in participating in the study emailed the researchers and were sent a letter of information containing general information on the study. Individuals were informed that they would be sent an email containing a link to an online survey that consisted of questionnaires concerning their thoughts, feelings, and behaviours. After providing informed consent, individuals were added to the study and emailed online surveys each month over a six-month consecutive period (i.e., five monthly surveys after completion of the Time 1 survey). The Time 1 survey took approximately 30 minutes to complete. Ideal partner preferences and self-perceptions were assessed across 38 qualities (e.g., "understanding," "good lover," "ambitious,") used extensively in prior research. Specifically, the measure combined the 18-item short form of Ideal Standards Scale [15], and the 20-item Interpersonal Qualities Scale [20]. Participants rated the degree to which these qualities were important to them in describing their ideal partner in a close, romantic relationship on a 7-point scale (1 = very unimportant, 7 = extremely important). They also rated how characteristic each quality was of them on a 7-point scale (1 = not all characteristic, 7 = very characteristic). Participants who completed the Time 1 survey were emailed an Amazon gift card worth CAD-$10.00 as compensation for taking part in the initial portion of the study.
Each month for the next consecutive five months, participants were emailed links to the remaining online surveys, each of which took approximately 10 minutes to complete. Participants were emailed Amazon gift cards worth CAD-$5.00 per survey completed (i.e., up to CAD $25.00) as compensation for taking part in the over-time portion of the study. Each monthly survey first inquired if the participant had entered into a romantic relationship since the prior survey (i.e., "Have you become involved in a romantic relationship since the last survey?"). 1 If at any point over the five-month period after completion of the Time 1 survey participants become involved in a relationship, they were given the opportunity to provide their partner's email address, and, if they chose to share their new partner's contact information, the new partner was contacted by the researchers and invited to participate in the study. The survey completed by the new partners took approximately 30 minutes to complete and contained the same ideal partner preference and self-perception scales administered to the original participants. New partners who completed the study were emailed Amazon gift cards worth CAD-$10.00 as compensation. Lastly, all participants were emailed a debriefing form explaining the nature of the study 48 hours after the completion of their final monthly survey.

Results
For descriptive purposes, the range of responses, means, and standard deviations for ideal partner preferences and self-perceptions from both the original participants and their new partners are presented in Table 1. The SAS code used to estimate all of the effects reported below, and instructions on how to obtain access to the data used to test the models discussed below, are posted on the OSF and can be accessed via the following link: https://osf.io/ me7jp [3].

Data Analytic Strategy
Models were tested using multilevel modeling (MLM, also known as hierarchical linear modeling; [16,23], following the suggestions of Kenny, Kashy, and Cook ( [17]; see also [5]). In the present study, data have a nested structure, with original participants' and new partners' ratings of ideal preferences and self-perceptions across the 38 qualities (Level 1) nested within couple (Level 2). We ran two models: Model 1, the primary test of our hypothesis, estimated the association between the original participants' ideal preferences at Time 1 and the new partners' self-evaluations. Model 2, a secondary test of our hypothesis, estimated the association between the new partners' ideal preferences and the original participants' self-evaluations at Time 1. 2 All variables were grand-mean centered, and the predictor variable was modeled as a random effect. Table 2 displays the results from the original analyses for Models 1 and 2, as well as the results from the discriminant analyses discussed in the following section. The analysis testing Model 1 revealed a positive and significant association between original participants' ideal preferences and their new partners' self-perceptions. In other words, individuals' ratings of the qualities they preferred in their ideal partner when they were single positively predicted the self-perceptions of their new partners across the 38 qualities. The analysis testing Model 2 also revealed a positive and significant association between the new partners' ideal preferences and the original participants' self-perceptions; that is, the new partners' ratings of the qualities they preferred in their ideal partner were similarly related to the self-perceptions reported by original participants at Time 1.

Discriminant Analyses
To determine the robustness of the effects from our original analyses and to rule out alternative explanations, we conducted several discriminant analyses. First, because people tend to be attracted to others that are similar to themselves [19], we ran Models 1 and 2 with the addition of the original participants' self-evaluations across the 38 qualities (Model 1) and the new partners'   Table 2: Multilevel Models Testing the Effects of Original Participants' Ideal Preferences Predicting New Partners' Self-Perceptions (Model 1) and New Partners' Ideal Preferences Predicting Original Participants' Self-Perceptions (Model 2). Note. Unstandardized regression coefficients are reported. Approximate effect sizes were computed using the formula r = √(t 2 /(t 2 + df )) (see [24]). Degrees of freedom ranged between 1333-1395 for models estimated with all couples, and were 1101 for the models with couples with no dating history. ***p < .001. self-evaluations (Model 2) as predictor variables. Selfevaluations contributed meaningfully to both Model 1 and Model 2, b = .18, SE = .05, p < .001 and b = .14, SE = .04, p = .001, respectively; however, when statistically controlling for these self-evaluations, the effects of ideal partner preferences predicting self-perceptions remained significant in both models. Second, we noted that, of the 38 couples included in our original analyses, only 30 couples indicated that they had never dated their partner before; five reported having a dating history with their partner and three did not answer this question. When rerunning Models 1 and 2 including only the 30 couples who indicated they had no previous dating history with their partner, the effects of ideal partner preferences predicting self-perceptions remained significant in both models.
Finally, and perhaps most importantly, the association between stated ideal preferences of one partner and the self-evaluations of the other partner may reflect stereotype accuracy, meaning that the original participants may share a similar view of an ideal partner (e.g., rating "honesty" as important but "moody" as less important), and the new partners may perceive themselves in a similar manner (e.g., rating themselves relatively high on "honesty" and relatively low on "moody") [29]. If so, the unique preferences of one partner would not predict the unique self-evaluations of the other partner. We therefore followed the suggestions of Kenny et al. [17] to control for stereotype accuracy, allowing us to determine if the association between the ideal preferences of one partner and the self-evaluations of the other partner remains robust when essentially taking into account the magnitude of the association between randomly paired dyads. Specifically, we subtracted the mean ideal preference value for each trait from the ideal preference reported by each participant, and we subtracted the mean self-evaluation value for each trait from the selfevaluation reported by each participant. The resulting values have less variability in proportion to the degree that responses were heavily influenced by stereotype accuracy. When controlling for stereotype accuracy, the effects of ideal partner preferences predicting self-perceptions remained significant in both Model 1 and Model 2.

Discussion
This is the first study to measure ideal partner preferences in a sample of individuals not currently involved in a romantic relationship, track their relationship status over time (i.e., five months), and then measure the selfevaluations of their new partners to determine if individuals tend to enter new relationships with others that possess traits and qualities that more (versus less) correspond to their preferences. Across 38 qualities that have been used in many studies on interpersonal attraction and relationship processes, there was a positive association within couples between stated ideal partner preferences when single and the self-reported attributes of the new romantic partner. A similar positive association emerged between the ideal partner preferences of the new partners and self-evaluations of the original participants. These positive associations were not accounted for by the similarity in self-evaluations between partners, previous dating history, nor by stereotype accuracy. The data presented therefore suggests that people do appear to enter relationships with others that self-report possessing qualities that correspond more closely to their stated preferences when single.
It is important to note, however, that our data do not allow us to directly test if individual's pursued their eventual new romantic partners because these partners possessed qualities that corresponded more closely to their own ideal preferences (i.e., question [b] discussed above). We assessed the degree of correspondence existing between ideal preferences of the original participant and the self-evaluations of the new romantic partner, not if variation in such correspondence predicted the desire to form new relationships. Now that we have initial evidence establishing a positive degree of correspondence between stated preferences when single and the self-evaluations of new partners across the same set of qualities, future research needs to test of the degree to which varying levels of such correspondence predict the likelihood of beginning, and maintaining, new relationships.
It is also possible that the link between ideal partner preferences and forming relationships with others that more closely approximate these preferences is indirect rather than direct, implying the presence of mediating variables in the relationship formation process. For example, people with particular ideal partner preferences may be more likely to seek out social contexts where potential mates with qualities similar to their preferences are also likely to frequent. Indeed, research by Motyl, Iyer, Oishi, Trawalter and Nosek [22] has demonstrated that individuals select communities with ideologies similar to their own in which to live. Individuals that ideally prefer someone who likes being physically active may therefore be particularly likely to meet a future partner while mutually engaging in these types of activities (e.g. while hiking on a trail, shopping for camping gear, while exercising at the gym), whereas someone else that ideally prefers someone that is more introverted may be particularly likely to meet a future partner in a different setting (e.g., at the café of a book store, after being introduced by friends, while strolling through the park). Future research should test this possibility.
Before concluding, we discuss a few limitations of the current research. Although efforts were made to recruit a larger sample of the new partners as relationships were formed, our final sample consisted of 38 couples, with 38 repeated measures (i.e., ideal partner, and self, ratings) per couple, totalling 1444 data points. In order to properly test between couple differences (e.g., Are couples happier when ideal preferences and self-evaluations match more closely?), data from many additional couples would be required. Additionally, this research was designed to determine if individuals enter new relationships with partners that more versus less closely match their own ideal preferences, and therefore is not able to determine if the degree of such correspondence that exists between partners is predictive of the development of these new relationships in the near-or long-term. Building on the initial empirical evidence presented here, it is also important for future research to replicate these findings to acquire a more precise estimate of the true effect size.