Development and validation of the Measure of Initial Attraction (MIA) / Elaboración y validación de la Medida de Atracción Inicial (MAI)

Abstract Unilateral Initial Attraction (UIA) is a first unilateral awareness towards an unknown person and can be the starting point of an interest in voluntarily initiating an interaction or relationship. In order to create a measure tapping this feeling, Study 1 asked individuals to indicate attributes characterizing UIA (Phase 1), and to rate their centrality (Phase 2). These were used to develop the Measure of Initial Attraction (MIA) comprising one component of arousal and another of unilateral interest. While the former is shared with the love construct, the latter differentiates from measures of passion. The MIA proved to be a valid and reliable instrument with the capacity to discriminate UIA across different relationships (Study 2) and targets (Study 3), with good convergent validity (Study 3). Results are discussed within the framework of personal relationships.

However, sometimes we do not have information about the target, reciprocity of interest, or prior interaction to feel attracted. Indeed, some theoretical models (Bredow, Cate, & Huston, 2008;Levinger, 1983;Murstein, 1970) explicitly refer to the linkage between a first unilateral awareness and the interest in voluntarily initiating an interaction/relationship. We term this Unilateral Initial Attraction (UIA) and assume that, even when no relationship has been effectively started (e.g., Berscheid & Regan, 2005), experiencing UIA is central to promote voluntary interest.
Albeit its relevance, UIA has not been thoroughly studied and measured. This is our focus, presenting the development of a UIA measure and providing empirical evidences regarding its validity and reliability.
The Unilateral Initial Attraction (UIA) phenomenon Attraction may be elicited immediately after a unilateral awareness/perception of a target, and promote positive affect, interest and willingness to engage in a relationship (cf., Afifi & Lucas, 2008;Bredow et al., 2008;Levinger, 1983;Murstein, 1970). We term this the UIA phenomenon (Rodrigues, 2010). As Levinger and Snoek (1972) suggest, 'the beginnings of a relationship appear when one person (P) becomes aware of another (O)', and that 'it is unimportant whether or not O in turn notices P. The only pertinent event is that P has information that forms a basis for his unilateral evaluation of O' (p. 6).
This phenomenon has not been given sufficient attention by researchers, as literature in relationship initiation tends to overlook the importance of UIA and focus on variables associated to romantic/sexual attraction (e.g., Buss & Barnes, 1986;Shackelford, Schmitt, & Buss, 2005). Indeed, some relationships are not initiated by sexual/romantic attraction (Moser, 1994), but rather by a general UIA (e.g., Bredow et al., 2008). In this sense, understanding the UIA construct and reliably measuring it is extremely important for analysing relationship initiation/ development.
Based on assumptions drawn from literature, the construction and development of the UIA measure is based on two premises: (1) UIA is not necessarily romantic and/or sexual; (2) nor it is necessarily associated with, or a first step in, the search for potential dating/romantic partners. Such conceptualization may confuse UIA with constructs such as liking (e.g., Lamm & Wiesmann, 1997) or desire/lust (e.g., Regan, 2004). In this sense, we assume it to be distinct from these two constructs in two important ways. First, liking is a positive general evaluation (Hendrick & Hendrick, 1992) that can take different meanings (e.g., respect, empathy, friendship, love; Rubin, 1970), and does not necessarily signal willingness to voluntarily approach another person with sexual/romantic intent. Hence, liking is not necessarily associated with willingness to interact, contrarily to UIA, which is assumed as the first stage for developing voluntary relationships.
Second, a first evaluation of a target is sometimes associated with physical attraction or desire/lust (e.g., Regan, 2004). Indeed, Regan (2000) suggests that desire/lust is mainly associated with passion, and not necessarily with liking or loving (e.g., companionate love). On the other hand, conceptualizations of immediate attraction such as being in love (Regan, Kocan, & Whitlock, 1998), or limerence (Tennov, 1999) assume desire/lust to be associated with the experience of physiological arousal when encountering, being in the presence, and/or thinking about another person. However, UIA is not necessarily associated with desire/lust, as it may motivate us to simply get to know the other and eventually develop a friendship (not characterized by desire/lust; Moser, 1994).
In short, UIA seems to be associated with willingness for a first approach. Even though this unilateral perception includes liking, it is not necessarily associated with desire/lust. This subtle yet important distinction allows us to argue that UIA underlies different relationships. Nevertheless, it is reasonable to expect different degrees of UIA to be associated with different relationships, as UIA should be complemented with other specific attributes (e.g., friendships vs. love; Moser, 1994). Hence, we expect a valid and reliable measure of UIA to prove not only sensitivity in measuring UIA across relationships, but also to assess this specific feeling, when compared to measures of proximal constructs (e.g., passion).

Overview and aims
To understand the UIA construct, its associated attributes were analysed using a prototypical approach (cf., Mervis & Rosch, 1981). In Study 1, individuals were asked to characterize the UIA feeling when first becoming aware of another person (Phase 1). A second set of participants was then asked to analyse each attribute and indicate its perceived centrality to UIA (Phase 2). In Study 2, we present the Measure of Initial Attraction (MIA) and analyse its construct validity and reliability, as well as its ability to discriminate UIA across different relationships. Study 3 focused on convergent validity, as well as on the scale's sensitivity to different targets.

Study 1
To operationalize UIA, we first relied on individuals' knowledge. By adopting a prototypical approach (cf., Mervis & Rosch, 1981), we identified the central and peripheral attributes that characterize the subjective experience of UIA. This study was divided into two phases. Phase 1 asked participants to characterize their experience of UIA when first becoming aware of an unknown target, allowing us to identify the more frequently nominated attributes (Buss & Craik, 1983). Phase 2 asked another sample of participants to analyse each attribute according to its importance to UIA, allowing us to identify central and peripheral attributes (e.g., Fehr, 1988).

Procedure
In both phases, participants were asked to freely take part in a study about interpersonal attraction. Participants were handed a booklet for completion. In Phase 1, the first page had two control questions assuring that participants knew UIA (1 = Don't know what it is, 9 = Know what it is) and had previously experienced it (1 = Never felt, 9 = Already felt). In the second page, half of the sample was asked to freely remember and write attributes associated with the experience of UIA, while the remaining participants were asked to write a personal episode where they felt UIA, focusing on their first sensations.
In Phase 2, the first page of the booklet asked participants to think about a situation where they felt UIA. Next, the attributes extracted from Phase 1 were presented and, reporting to such feeling, the participants' task was to indicate for each attribute: (a) how characteristic it was for UIA (1 = characteristic, 7 = extremely characteristic); and (b) if it was considered mandatory to experience it (yes/ no). The last page presented all the attributes, asking them to choose the 10 most important for UIA. In both phases, after completion participants were debriefed and thanked.
Participants' responses were content analysed following Fehr's (1988) methodology. Sentences with one attribute were coded directly (e.g., I felt cold sweats), while complex sentences were divided in attribute units and coded accordingly (e.g., We glanced at each other and then I felt butterflies in my stomach). Two independent judges coded the attributes into broader categories (comparison between codings yielded a 95% level of agreement; disagreements resolved through discussion). No differences according to participants' gender were found. Table 1 summarizes the frequency of nomination for the attributes used in the MIA 1 .
In Phase 2, for each attribute we computed: (a) the mean score reflecting how characteristic it was for UIA; (b) the percentage of participants deeming it as mandatory; and (c) the percentage of participants selecting it as important (Table 1). Following other approaches (e.g., Fehr, 1988;Regan et al., 1998), attributes with scores equal/above the median for at least one variable were We found a high level of participant agreement, showing consistency in the identification, representation and characterization of the subjective experience of UIA. Indeed, the attributes' frequency of nomination (Phase 1) was correlated to all measures from Phase 2, .36 > r > .52, all p < .001, suggesting that the most frequently listed attributes of UIA by a first group of participants were perceived as more central by a second independent group of participants. Likewise, more characteristic attributes were considered mandatory, r = .97, p < .001, and more important, r = .81, p < .001, and mandatory attributes were considered more important, r = .76, p < .001.
These results clearly suggest UIA as a specific phenomenon with a shared knowledge structure (Fletcher & Thomas, 1996;Hardin & Higgins, 1996), assuring the development of a Measure of Initial Attraction (MIA) tested for its construct validity and reliability in Study 2.

Study 2
In this study we present and analyse MIA's underlying factor structure by using exploratory and confirmatory factor analyses. Apart from focusing on construct validity and reliability, this study also sought to analyse the measure's capacity to discriminate UIA across relationships.
By assuming UIA as a necessary basis to initiate voluntary relationships, we also assume it as a shared characteristic of such relationships. However, UIA should be experienced differently across relationships. Hence, we expect MIA to be more sensitive in assessing UIA for an unknown other, than for one's romantic partner, friend or work colleague.

Participants and design
Three-hundred and seventy-four undergraduates (62.60% females, M Age = 21.10, SD = 2.78) took part in this study. Two subsamples with approximately 50% of the cases were randomly extracted. The first subsample was composed of 217 participants (129 females, M Age = 21.04, SD = 2.63), and was the focus of an exploratory principal components analysis. The second subsample, composed of 183 participants (116 females, M Age = 21.08, SD = 2.89), was the focus of a confirmatory factor analysis. Participants were randomly assigned to one of four conditions, defined by the target: (a) initial attraction (n = 94); (b) love (n = 97); (c) friendship (n = 94); or (d) colleague (n = 89).

Measure
Thirty-one of the 36 central attributes were selected for the MIA. Four attributes were dropped for their direct reference to UIA (e.g., immediate attraction) or sexual desire (e.g., seduction). Each selected attribute was transformed into the sentence 'I felt [attribute] him/her' (e.g., I felt interested in him/her), and associated to a 7-point scale (1 = Not at all, to 7 = A lot).

Procedure
Upon consent to freely take part in a study about relationships, participants were randomly handed a booklet. On the first page they were asked to either think about 'an unknown person for whom you felt an immediate attraction' (UIA target), 'the person with whom you have/had a romantic relationship' (love target), 'a close friend with whom you would not have a romantic relationship' (friendship target), or 'a work colleague that you like but with whom you would not have a romantic relationship' (colleague target). While thinking about the target, participants were asked to complete the MIA. After completion, participants were debriefed and thanked.

Results and discussion
Principal components factor analysis (PCA) and reliability A PCA analysis with Promax rotation was conducted. The extraction method and rotation used in our analyses were chosen for two main reasons: (a) literature suggests PCA as the preferred method of extraction when reducing the number of items of a new instrument measure (Stoner, Perrewé, & Hofacker, 2011); and (b) since we suspect MIA's underlying factors to be correlated, Promax rotation method allows the factors to correlate while finding the best fit for an orthogonal solution (Hendrickson & White, 1964).
A first PCA resulted in two components sharing the attributes vivacious, fascination, and willingness to look. These ambiguous attributes were discarded. Based on the Kaiser rule (Kaiser, 1960), a second PCA with the 28 remaining items resulted in two correlated components, r = .74, p < .001. The final solution presented a highly acceptable index of sample adequacy (KMO = .95; inclusion of item on component with loading > .40), explaining 60.72% of total variance. Also, both components presented high Cronbach's alphas, with each item contributing to the respective component's reliability as shown by the corrected itemtotal correlations ( Table 2).
The first component, Arousal (16 items) is defined by the experience of arousal (e.g., desire) associated with an inexplicable feeling (e.g., something strange), physiological reactions (e.g., butterflies in my stomach), and intrusive thinking about the other (e.g., thinking about). The second component, Interest (12 items) is defined by positive feelings (e.g., joy), interest (e.g., curiosity), and willingness to voluntarily approach the other (e.g., willingness to know).

Confirmatory factor analysis (CFA)
We ran CFA analyses in the second subsample of participants. Three CFAs were performed, and fit indexes of a two-correlated factors model (our hypothesized model), a two-uncorrelated factors model, and a one-factor model were obtained.
Regarding each MIA component, we expected different patterns. Specifically, and although one may experience UIA (high arousal and interest) towards an unknown target (e.g., attraction at first sight) or one's romantic partner, the UIA for a friend or acquaintance should not be characterized by high arousal (Moser, 1994). Hence, we expected both UIA and love targets to share high arousal (vs. both friendship and acquaintance targets), and UIA target to reveal a higher interest score (vs. each of the remaining targets). As no differences according to the participants' gender were found previously, this factor was discarded in this analysis. A 4 (Target) × 2 (MIA components) repeated measures ANOVA revealed the expected main effect across targets, F(3, 370) = 43.29, p < .001, η 2 p = .26. Analysing more specifically the scores for the arousal component, results reveal that scores were higher for the UIA and love targets combined (M = 5.10, SD = 1.01) when compared to the combined scores for both friendship and colleague targets (M = 3.97, SD = 1.22), t(372) = 9.76, p < .001, d = 1.01.
In a nutshell, these results assure the validity and reliability of MIA and support our conceptualization of UIA. The UIA construct seems to be experienced as arousal and interest. However, and given the sensitivity results, both components do not seem to be necessary to experience UIA. Indeed, the arousal component seems to be shared with passionate relationships (e.g., Hatfield, Bensman, & Rapson, 2012;Moser, 1994). The interest component seems to be important for interpersonal approach and relationship initiation (given the scores for the UIA target), as well as for the development of different relationships (given its higher scores for all targets). In Study 3 we aim to further validate the MIA, focusing on convergent validity.

Study 3
In the previous study, no empirical evidences were presented regarding MIA's capacity to differentiate UIA from a passion/passionate love feeling (both feelings seem to share an arousal component, see Study 2). Hence, in this study participants were asked to report their feelings of UIA and passion/passionate love for a famous person with whom they never interacted. Half the participants thought of a target associated with UIA, while the other half thought of a neutral target. By doing so, participants reported their feelings solely based on unilateral personal knowledge about the other person and with no information regarding reciprocity of feelings or interest, thus converging with our conceptualization of UIA.
Based on previous evidences, we expect scores on all measures to be higher for the UIA (vs. neutral) target, with participants reporting higher MIA scores (vs. passion). We also expect MIA's interest scores to be higher (vs. all measures) for the UIA target, with no differences between MIA's arousal, and measures of passion. Similarly, for the neutral target we expect higher MIA scores (vs. passion), especially in the interest component. Again, no differences are expected between MIA's arousal, and measures of passion.

Method
Participants and design Two-hundred and five undergraduates (152 females, M Age = 22.80, SD = 5.82) participated in this study. Participants were randomly assigned to one of two target conditions: (a) UIA (n = 104); or (b) neutral (n = 101).

Procedure and measures
Upon consent to freely take part in a study about relationships, participants were randomly handed a booklet for completion. On the first page, they were asked to either think of 'an actor/actress with whom they had never interacted before, but on whom they have a crush' (UIA target) or 'an actor/actress with whom they had never interacted before and on whom they do not have a crush' (neutral target). While thinking about the target person, participants were asked to complete the MIA, the Passionate Love Scale (PLS; α = .91; Hatfield & Sprecher, 1986), and Eros sub-scale (α = .70; Hendrick & Hendrick, 1986). Responses were given in a 7-point scale (1 = Not at all, to 7 = A lot). Scales had no specific identification or instructions, and were presented in random order within conditions. After completion, participants were debriefed and thanked.
By assuming UIA to have an arousal and an interest component, it was reasonable to expect MIA scores to be correlated with PLS and Eros scores. Indeed, overall MIA scores were moderately correlated with PLS, r = .68, p < .001, and Eros scores, r = .53, p < .001. Furthermore, MIA's arousal component was moderately correlated with the PLS, r = .69, p < .001, and Eros, r = .54, p < .001, and similarly MIA's interested component was moderately correlated with both PLS, r = .64, p < .001, and Eros, r = .50, p < .001. These results suggest that even though these measures tap certain shared attributes of UIA and passion, the magnitude of correlations also suggests that MIA and its components tap into specific and non-shared attributes characterizing UIA.

MIA's sensitivity analysis
To further test MIA's sensitivity in measuring UIA, we compared the scores of MIA and its components, PLS and Eros for the UIA target and for the neutral target. We expected higher scores in all measures for the UIA (vs. neutral) target. For the UIA target we also expected scores on the MIA's interest component to be higher ( In a more stringent test, we compared scores within each target. Planned contrasts on the UIA target reveal higher scores for MIA, when compared to the combined PLS and Eros scores, t(203) = 5.72, p < .001, d = .80. Also, planned contrasts show MIA's interest component scores to be higher than scores on both PLS and Eros combined, t(203) = 7.17, p < .001, d = 1.01. Contrary to our predictions, MIA's arousal component scores were also higher than PLS and Eros scores combined, t(203) = 4.25, p < .001, d = .60.
Considering the neutral target, planned contrasts show higher scores on the overall MIA, when compared to PLS and Eros scores combined, t(203) = 2.02, p = .044, d = .28. MIA's arousal component scores were not different from PLS and Eros scores combined, t(203) = .98, p = .327, d = .14. MIA's interest component scores were also higher than scores on both PLS and Eros combined, t(203) = 3.16, p = .002, d = .44.
These results show that the MIA has sensitivity in measuring UIA (vs. measures of passion). Indeed, MIA tapped into a feeling characterized by arousal and interest, not exclusively characterized by, and not overlapped with, more intimate and intense feeling of passion. This is especially evident when considering that, for the UIA target, scores on both MIA components were significantly higher (vs. PLS/Eros scores), while for the neutral target no differences were found between MIA's arousal component PLS and Eros. Importantly, MIA's interest component scores for the UIA target were above the mid-point of the 7-point response scale, t(103) = 2.56, p = .012, d = .50, while PLS and Eros scores were below the mid-point, t(103) = −3.34, p < .001, d = −.66 and t(103) = −4.22, p < .001, d = −.83 respectively.
Although the finding that MIA's arousal component was higher than PLS and Eros for the UIA target is contrary to our original hypothesis, this does not question our measure's validity. We believe it strengthens it. Note that in Study 2 we asked participants to think of either an UIA or a love target, and found that both shared MIA's arousal component. This was not the case in the present study. However, this result is not the same as saying that the feeling of UIA is characterized by a component of passion (as measured by PLS and Eros) for a public figure with whom one has never interacted before. Indeed, these results suggest that UIA is characterized by a component of arousal that is not necessarily passion, and is also characterized by a component of interpersonal interest that seems to be always present (hence the higher scores for this component in both targets). This evidence stays in line with the argument that what is tapped by the MIA's interest component is more general and may promote an interest in wanting to know more about the other person, independently of physiological reactions.

General discussion
This article is a first step in studying UIA, an innovative concept that fills a gap in literature, clarifying the UIA construct, providing a reliable measure, and opening new lines of research. UIA is experienced unilaterally towards an unknown person, being the base to initiate voluntary relationships (Bredow et al., 2008;Levinger, 1983;Murstein, 1970). Relying on individuals' knowledge and experience, we analysed the centrality of UIA's attributes (Study 1) and developed the MIA, a valid and reliable instrument (Study 2), with convergent validity (Study 3) and sensitivity to different relationships/targets (Studies 2 and 3).
Study 1 suggests UIA to have three main characteristics: (a) experience of positivity/affection; (b) personal interest/willingness to interact; and (c) arousal. Indeed, UIA is associated with positivity, a characteristic of attraction, thus not being surprising its overlap with liking (Lamm & Wiesmann, 1997), limerence (Tennov, 1999), falling/being in love (Regan et al., 1998), and love (Fehr, 1988). UIA is also associated with interest/voluntary willingness to interact, a necessary condition to initiate a relationship shared with liking. However, we question if this is necessary for liking, given that for UIA such interest is associated with empathy/fascination and for liking is associated with other's overvaluation. Finally, UIA is characterized by arousal/desire/lust shared with limerence, being in love and love. However, such experience for UIA has less intensity and emerges at a fantasy level, eliciting desire for reciprocation. Contrarily, limerence and being in love are associated to continued interactions, certainty of reciprocation, and intimacy.
In Study 2 we presented the MIA and attested this instrument's construct validity, reliability and sensitivity. Two components underlie the MIAarousal and interest. Also, MIA distinguished UIA across relationships, with the arousal component shared only with love (characterized by desire/lust), and the interest component differentiated across relationships (higher for the UIA target). This is in line with results from Study 3, where we found moderate correlations between MIA and measures of passion. Furthermore, we showed MIA's sensitivity by differentiating scores across targets. Higher MIA scores were obtained for the UIA target (vs. PLS/Eros), while differences in the neutral target were due to MIA's interest component (vs. MIA's arousal component or PLS/Eros). Hence, our measure tapped a specific feeling not assessed by passion measures.
More empirical data are needed to further validate the UIA construct and MIA. Indeed, we did not address divergent validity, and future studies should compare the MIA with other measures to further support for the differentiation of UIA. Furthermore, we did not address criterion-related validity. Assuming UIA to be the starting point of voluntary interpersonal relationships, it is important to analyse if the MIA can predict such initiation and development. Futures studies should ask individuals to recall the initiation of a close relationship (e.g., friendship, love), list the factors that lead them to first approach another person and then report their UIA. To have a comparison basis, individuals could do the same while considering a person with whom they did not develop a close relationship. The differentiation in UIA scores for each target would argue for the predictive value of the MIA. Finally, future studies should also consider developing a shorter version of the MIA in order to reduce eventual redundancy between the items and facilitate its applicability in experimental settings.

Estudio 3
En el estudio anterior no se ofrecieron evidencias empíricas de la capacidad de la MAI para diferenciar la AIU de un sentimiento de pasión/amor apasionado (ambos sentimientos parecen compartir un componente de activación, véase Estudio 2). Es por ello que en este estudio se solicitó a los participantes que describieran sus sentimientos de AIU y pasión/amor apasionado por una persona famosa con la que nunca habían interactuado. La mitad de los participantes pensó en un destinatario asociado con la AIU, en tanto que la otra mitad tuvo en mente a un destinatario neutral. Al hacer esto, los participantes describieron sus sentimientos basándose únicamente en un conocimiento unilateral de la otra persona y sin información sobre la reciprocidad de los sentimientos o del interés de la otra persona, es decir, convergiendo con nuestra conceptualización de la AIU.
Análisis de sensibilidad de la MAI Con el fin de comprobar una vez más la sensibilidad de la MAI para medir la AIU, comparamos las puntuaciones de la MAI y sus componentes, la PLS y el Eros para el destinatario de la AIU y para el destinatario neutral. Esperamos encontrar puntuaciones más altas en todas las mediciones para el destinatario de la AIU que para el destinatario neutral. Para el destinatario de la AIU también esperamos que el componente interés de la MAI fuera más alto que para todas las demás mediciones, aunque sin diferencias en el componente activación de la MAI, la PLS y el Eros. Para el destinatario neutral esperamos puntuaciones más altas tanto para la MAI en general como para el componente interés de la MAI que para el resto de mediciones.
A pesar de haber encontrado que, en contra de nuestra hipótesis original, el componente de activación de la MAI resultó más elevado que el de la PLS y Eros, esto no pone en duda la validez de nuestra medición. Creemos que, por el contrario, la fortalece. Nótese que en el Estudio 2 solicitamos a los participantes que pensaran en un destinatario de la AIU o de amor y que encontramos que ambos compartían el componente de activación. Este no fue el caso en el presente estudio. Sin embargo, este resultado no equivale a decir que el sentimiento de AIU se caracteriza por un componente de pasión (tal y como lo miden la PLS y Eros) hacia una figura pública con la que uno nunca ha interactuado. De hecho, los resultados sugieren que la AIU se caracteriza por un componente de activación que no es necesariamente pasión y, al mismo tiempo, se caracteriza por un componente de interés que parece estar siempre presente (de ahí las puntuaciones más elevadas para este componente para ambos destinatarios). Este hallazgo converge con la idea de que lo que mide el componente del interés de la MAI es más general y puede fomentar un interés en querer conocer más acerca de la otra persona, independientemente de las reacciones fisiológicas.