People Do Not Always Act as Positively as They Feel: Evidence of Affiliation Suppression

Goethe (1891) illustrates the emotional toll carried by a person who experiences attraction but does not act on those feelings. Though this quote was taken from a novel published in 1891, the phenomenon is timeless, and even in today’s social world, daters often refrain from acting as positively as they feel (Spielmann, MacDonald & Tackett, 2012). Whereas models and definitions of attraction tend to assume a high degree of consonance between the subjective experience of attraction for another person (affective attraction) and how one acts toward the person (affiliative behavior), there is only a small but significant relation between the two (Montoya, Kershaw & Prosser, 2018). In this research, we investigated processes that produce such a discrepancy, what we call affiliation suppression, in which a person’s behavioral response is restrained compared to his or her subjective experience. We begin by describing a model of interpersonal attraction that provides a framework for understanding when and why feelings and behavior diverge, and then we describe various psychological and situational processes that may account for a divergence. Discrepancy between Affective Attraction and Affiliative Behavior Attraction is a process that motivates physiological, cognitive, and behavioral responses that regulate interdependent relationships. The two-dimensional model of attraction (TDMA; Montoya & Horton, 2012, 2014; Montoya et al., 2018) proposes that interpersonal attraction assessments begin with the fundamental cognitive appraisals that guide the person perception process: willingness and ability. The willingness appraisal refers to the evaluation of the target person’s benevolent orientation toward the perceiver. The ability appraisal refers to the perception of the target person’s capacity to facilitate/hinder the perceiver’s interests. These assessments combine to produce the subjective experience of attraction (i.e., affective attraction). The degree to which affective attraction aligns with the behavioral response is determined by considerations inherent to social exchanges. All types of interpersonal interactions, from the trivial (interacting with a bank teller) to the benign (a meeting between a student and faculty member during office hours) to the romantic (initiating a conversation at a discotheque) can be considered from within the social exchange framework. TDMA posits that the behavioral expression of attraction communicates trust, which acts to increase the likelihood that the target person adheres to the social exchange (Montoya et al., 2018). From this perspective, people smile, make eye contact, mimic, and laugh to signal their conditional willingness to cooperate during a social exchange. However, natural selection works against people whose behavior in a social exchange is unrestrictedly cooperative Montoya, R. M., and Sloat, N. T. (2019). People Do Not Always Act as Positively as They Feel: Evidence of Affiliation Suppression. International Review of Social Psychology, 32(1): 9, 1–17. DOI: https://doi.org/10.5334/irsp.202

. Expressing a cooperative intent in all situations produces opportunities for being taken advantage or, perhaps, taking advantage of another person. In the following sections, we discuss three key considerations that affect whether trust is communicated, and thus, whether affiliative behavior follows from affective attraction. 1

Fear of exploitation
Expressing one's intent to cooperate heightens the chance that one will be exploited (Frank, 1988). Although people may feel quite positively toward someone, their affiliative response may be suppressed to reduce the likelihood of exploitation. For instance, female daters may not express too much liking on a first date to guard against being exploited for sex by their male dates. Most commonly, this is manifested in a romantic relationships or friendship as an equity concern, concerns that a person's feelings are more positive than those of the other person. The person may engage in affiliative suppression to avoid the perception that he or she likes the other person more than the other person likes him or her.

Self-threat considerations
Self-threat concerns are worries that one is not worthy of a social exchange with a specific person. Two processes relevant to self-threat have received attention in the literature, specifically, perceived acceptance and shyness.
Perceived acceptance. Concerns regarding acceptance and rejection contribute to a divergence between affective attraction and behavioral responses (Montoya & Horton, 2014;Montoya & Insko, 2008). Indeed, fears concerning rejection activate self-protective systems that shield the person from harm and reduce behavioral responses toward someone who is likely to "reject" him/her. The self-evaluation maintenance model (Tesser, 2000), for example, posits that one way to reduce threat to the self is to distance oneself from a close other who surpasses the self on a relevant dimension (see also Cialdini et al., 1976;Snyder, Lassegard & Ford, 1986).
Shyness. Shyness is a trait associated with suppression of one's behavior in the interpersonal context. Shyness is characterized by concerns about and discomfort in interactions with acquaintances and strangers (Cheek & Buss, 1981). Individuals who are shy lack self-confidence in social interactions (Clark & Arkowitz, 1975) and question their ability to make a good impression (Leary, 2001). People who are shy experience similar desires for closeness and intimacy compared to those who are not shy (Baker & Oswald, 2010), but self-confidence and social anxieties reduce engagement with other people. For example, users of an online dating service who scored high in shyness were less likely to contact other people and to initiate contact with possible dating partners despite experiencing a desire to find a relationship (Scharlott & Christ, 1995). Individuals who are shy may routinely engage in suppression, in that they experience affective attraction for a desirable person, but concerns regarding their ability to succeed in making a connection inhibit behavioral responses.

Situational processes
Situational processes refer to environmental, situational, and pragmatic influences that work to produce suppression of an affiliative response. The three most prominent situational processes are: social norms, anticipated effort, and target person availability.
Social norms. Norms have been linked to the divergence between one's attitude and behavior. Injunctive and descriptive norms for noninterdependent behaviors (e.g., exercising, eating healthily) facilitate the prediction of behavioral intentions and behavior (Rivis & Sheeran, 2003). Descriptive norms (the perception of how most people act in a particular situation; Cialdini, Reno & Kallgren, 1990) may be a contributing factor for the lack of interracial relationships in the South during the 20th century (Kalmijn, 1998). Alternatively, injunctive norms (the perception of what behavior should be done in a particular situation; Cialdini et al., 1990) also affects the degree to which suppression occurs. For instance, norms regarding 'taboo' or forbidden relationships (e.g., dating one's cousin, relationships between people with a dramatic age difference) inhibit relationship formation, and social disapproval impacts relationship development (Frost, 2011). Thus, whereas one might experience tremendous attraction for a friend's relationship partner, normative prohibitions against sabotaging a relationship may reduce the likelihood that one will act on such feelings. Anticipated effort. Predictions derived from the theory of planned behavior submit that behavioral intentions may not match enacted behavior when the enacted behavior is ' difficult' to accomplish (Ajzen, Brown & Carvajal, 2004). Specifically, the degree to which a behavior is ' difficult' to accomplish can reduce the relation between affective evaluations and behavioral responses. Campbell (1963) reasoned that whereas it is relatively ' easy' to, say, state that one will engage in discriminatory writings or actions, it is more ' difficult' to discriminate when standing faceto-face with a potential victim. In the case of suppression, stating that one will approach a highly attractive person is less ' difficult' than actually engaging in the behavior.
Target person availability. Whether the target person is available for a friendship or romantic relationship may generate suppression. For pragmatic reasons, suppression may result when the person of interest lives far away, is dating someone, does not have the time to invest in another relationship, or is unavailable for another reason (Gouldner & Strong, 1987). These considerations reduce the likelihood that an attempt to initiate a relationship would be successful and as a result, behavioral responses are more likely suppressed.

Purpose of These Studies
The purpose of this research was twofold: (a) to investigate the degree to which self-reported and observed affiliation suppression occurs and impacts behavioral responses, and (b) to assess the processes that covary with, and result from, affiliation suppression. Specifically, we investigated a number of processes that work to affect self-evaluative processing or situational constraints, and thus, contribute to the divergence of behavioral responses from affective attraction. To this end, we conducted three studies. Study 1 was a narrative study, in which we coded participants' descriptions of a time when they did not act as positively as they felt toward a person to whom they were strongly attracted. Study 2 was a laboratory study in which participants interacted with a confederate during a social task. Participants' affiliative behavior (e.g., chair placement, eye contact) was observed and compared to their self-reported attraction relative to various suppression processes (e.g., shyness, anticipated effort). In Study 3, participants completed an ideal-actual discrepancy task that assessed the degree to which their ideal behavior toward a desirable person matched how they thought they would actually act in the same situation. We investigated whether the various suppression processes covaried with the degree to which participants' ideal behavior matched their actual behavior.

Study 1
The purpose of Study 1 was to assess (a) the degree to which suppression occurred in an undergraduate population, (b) the self-reported reasons for suppression, and (c) whether the reasons for suppression differed by sex. In this study, we asked participants not currently in a romantic relationship to think of a person to whom they were most romantically attracted and to describe a situation (if possible) when their behavioral response did not match their positive emotional experience for that person. We coded the descriptions and their reported reasons for why their behavior did not match their affective attraction. Results were analyzed by sex because social norms dictate that men, compared to women, should be more inclined to "act on how they feel" and initiate romantic relationships (Crawford & Popp, 2003).

Participants
Three hundred and fifty-four undergraduate students (136 men, 217 women, and 1 did not specify sex; M age = 19.02, SD = 1.11) not involved currently in a romantic relationship at a midsized Midwestern university completed the study for course credit.

Procedure
Participants completed questions as part of a mass testing session at the beginning of the semester. Participants were first asked to imagine the person in their life to whom they were most romantically attracted. Participants were instructed to describe an 'instance in which your actions toward the person did not match how positively you felt. In your description, be sure to describe (a) how you feel about the person, (b) what you ideally wanted to do, and then (c) what you actually did.' Next, participants were asked to describe the reason(s) why their behavior did not match their feelings. Participants were also asked the degree to which they were emotionally attracted to the person (1 = very much; 7 = not at all) and how frequently their behavior does not match their feelings (1 = almost never; 7 = all of the time).

Coding
The two authors independently generated a list of dimensions based on a subset of participants' responses. The authors then discussed the lists and created a final coding scheme comprised of overlapping codes. Three judges, blind to the purposes of the study, coded the responses. Any discrepancies between the coders were discussed and a consensus response was used. Using Krippendorff's alpha to estimate agreement between coders (Hayes & Krippendorff, 2007), we found that interrater reliability (before discussion) was moderate, α = 0.47.
The categories were shyness/anxiety ('I felt too shy around the person to act like I liked him/her.'), physiological ('I just froze up.'), perceived acceptance/fear of rejection ('I was afraid the person would turn me down.'), social exchange/equity consideration (i.e., ensuring no 'imbalance' in the expressed attraction between persons, 'I was not sure that the person felt as strongly for me as I felt for him/her, so I did not express my feelings.'), target person availability ('The person was already in a relationship, so I did not act on my feelings.'), situation norms (e.g., concerns regarding public displays of affection; 'I wanted to run up and hug the person, but I thought everyone would stop and stare at me.'), and a miscellaneous category (specific codes that appeared fewer than three times; e.g., pragmatic or idiosyncratic, 'She's on my intramural team,' and 'I didn't feel like it.') that was coded as ' other reasons.'

Results
Overall, participants described a target person about whom they experienced affective attraction (M = 2.31, SD = 1.52; 1 = very much, 7 = not at all). The self-reported frequency of suppression was near the midpoint (M = 3.44, SD = 1.33).
From the initial sample of 354, 242 (68%) generated a description of when their reported attraction did not match their behavior, and the remaining 32% of responses failed to provide a description of a situation. Of those who did provide a description, 171 participants were able to provide a codeable reason for the discrepancy. Percent and frequency of reasons for suppression are presented in Table 1. For both men and women, the most commonly reported reasons were shyness/anxiety, perceived acceptance/fear of rejection, and social exchange/equity considerations. Over one-third of participants reported shyness as a reason, and close to a quarter of participants reported concerns regarding acceptance/rejection. Chi-square tests that explored whether the reasons differed by participant's sex were not significant (χ 2 values ranged between 0.01 and 0.77).
An investigation into the relation between affective attraction and the inclusion of specific suppression reasons revealed that men were more inclined to cite social norms for their reason for their affiliation suppression with more affective attraction, B = 0.51, χ 2 (1) = 6.35, p = 0.01, but women were not, B = 0.16, χ 2 (1) = 0.82, p = 0.34. No other relations between affective attraction and suppression reasons reached significance.

Discussion
Consistent with predictions and with the proposition that suppression is a common experience, 68% of participants were able to generate an instance in which affiliation suppression occurred. Moreover, as expected, considerations focused on social exchanges (social exchange/equity considerations) and self-threat (perceived acceptance, shyness/anxiety) were most responsible for the regulation of affiliative expressions. All told, results were consistent with the contention that suppression is common and that participants can generate reasons for why they did not act as favorably as they felt.

Study 2
Due to the possibility that Study 1's findings resulted from biases associated with recalling past behavior, we conducted a laboratory investigation in which participants' behavior with a confederate was recorded and coded. Specifically, we asked participants to interact with a confederate as part of a social task in which we unobtrusively evaluated participants' affiliative behavior toward the confederate. We then investigated whether the difference between participants' self-reported affective attraction for the confederate and the participants' behavioral response could be explained by the various suppression processes. Behavioral responses included chair placement, eye contact, smiling, and the self-reported liking for the confederate.
Importantly, to fully assess the plausible covariates of suppression, we evaluated additional processes that may produce suppression. Whereas we had expectations that the processes identified in Study 1 would continue to exert influence in the laboratory context, we recognized the possibility that other processes may contribute to suppression. As a result, we included measures of avoidant attachment, anxious/ambivalent attachment, familiarity, and the participant's physical attractiveness (as assessed by independent coders). We had a priori expectations that these processes likely covary with affective attraction and affiliative behavior (e.g., avoidant attachment would produce both fewer affiliative behaviors and less affective attraction), but our expectations were neutral regarding whether such processes would produce suppression.

Method Participants
Participants were 61 men (M age = 19.38, SD = 1.15) recruited from an introductory psychology course not currently involved in dating relationships. Participants completed the study in exchange for course credit. Data were collected for one semester.

Materials
Affective attraction. Six questions evaluated the degree to which the participant was emotionally attracted to the target person (Herbst, Gaertner & Insko, 2003). Participants rated their attraction on 7-point scales between the following anchor pairs: unpleasant/pleasant, cold/warm, negative/positive, unfriendly/friendly, distant/close, and dislike/like. Subjective norms. Adapted from Ajzen (2008), one question measured descriptive norms by asking how people similar to the participant would respond ('Most people like me would act the way I acted.'), and one question measured injunctive norms by asking how participants thought people who were important to them would evaluate their actions ('Most people who are important to me would approve of how I acted.'). Participants responded on a scale from 1 (strongly disagree) to 7 (strongly agree).
Perceived acceptance. Four questions evaluated the degree to which participants believed the target person felt positively/negatively about him/her. Each question was evaluated using a 7-point scale, from 1 (not at all) to 7 (very much). Sample items included: 'I believe that the person likes me,' and 'The person doesn't really like me very much.' Target person availability. Three questions evaluated participants' perception of their own ability to date the target person. Questions were rated on a 7-point scale from 1 (not at all) to 7 (absolutely). An item was, 'I believe that my partner is romantically available to date me.' Anticipated effort. Four questions evaluated how much time and work participants felt they would have to exert to date the target person. The questions were rated on a 7-point scale ranging from 1 (not at all) to 7 (absolutely). A sample item is, 'The amount of time spent on school/job/family stops me from pursuing my partner.' Attachment style. A 36-item questionnaire, including anxiety and avoidance subscales, measured participants' attachment styles to romantic partners based on their behavior and perception of situations in a relationship (Fraley, Waller & Brennan, 2000). High scores on the anxiety subscale reflect greater concerns about one's relationship, whereas high scores on the avoidance subscale indicate less engagement and a propensity to distance oneself from relationships. Participants responded using a 0 (do not at all agree) to 8 (agree completely) scale. Shyness. Nine questions from Cheek and Buss's (1981) scale evaluated trait inhibition and anxiety in social situations. Sample items include, 'I am often uncomfortable at parties and other social functions,' and 'I feel nervous when speaking to someone in authority.' The items were evaluated on a scale ranging from 1 (very uncharacteristic or untrue) to 5 (very characteristic or true).
Self-esteem. Participants' feelings about themselves were assessed using Rosenberg's (1965) self-esteem scale. Participants indicated strongly disagree, disagree, agree, or strongly agree in response to each of the 10 questions.

Procedure
Each session began with a female experimenter greeting the participant in the hallway in front of the laboratory. While walking to the participant's experimental room, the participant was led past an open door, through which the participant saw the female confederate. At this time, the experimenter informed the participant that he would participate in a two-part study, first, responding to questionnaires and second, participating in a collaborative task with the person he saw in the other room (the confederate).
On reaching the experimental room, the participant completed the attachment style, shyness, and self-esteem questionnaires and then the participant was told that he would interact on an interdependent task with his partner (the confederate). He was given a four-item introductory questionnaire to complete for his partner. This questionnaire contained items such as, 'What is your favorite movie of all time? And why?' When the participant finished the introductory questionnaire, the experimenter left the room and ostensibly exchanged the questionnaires between partners. In reality, the participant received a standardized questionnaire that reflected responses of a typical female undergraduate student (created by Montoya and Insko, 2008).
Next, the participant was led into the adjoining room to complete the collaborative part of the study. While on route to the adjoining room, the participant was told that his partner was still finishing her questionnaires. On entering the adjoining room, the participant was asked to 'help' the experimenter set up for the next task by moving a chair relative to the extant chair in the room while the experimenter went to get his partner. The experimenter explained that the indicated chair should be moved relative to the extant chair 'so that the two participants could complete a social interaction task.' The placement of the participant's chair relative to the extant chair served as a behavioral measure of attraction (chair proximity; Byrne et al., 1971).
After the confederate entered the room, the participant and confederate were told that the social task involved asking and responding to a series of pre-determined questions. The experimenter stated that the participant was randomly selected to answer the questions first and handed the questions to the confederate. The confederate then asked five questions (e.g., 'If you could live as an animal, what animal would you be, and why?' 'What was the last "new thing" you tried? What happened?'). Each session lasted approximately five minutes. A concealed video camera recorded the participant's responses to the questions.
Before the confederate could respond to those same questions, the experimenter re-entered and asked the participant to return to his original room to ' complete a brief questionnaire.' Participants then completed the perceived acceptance, partner availability, anticipated effort, social norms, and the affective attraction questionnaires.
After participants completed the questionnaire packet, participants were handed a final questionnaire to complete. This questionnaire was a modified version of the affective attraction questionnaire that they had completed previously. Participants were told explicitly that this questionnaire would be immediately given to their partner. The degree to which they favorably responded constituted a second behavioral assessment of attraction (called expressed attraction).
Videotapes of the interaction were later coded by two independent raters for behaviors associated with the experience of affective attraction. Specifically, the coders rated (a) the total duration of eye contact and (b) the total duration of smiles (Grammer, et al., 2000;McCormick & Jones, 1989; for a review, see Montoya et al., 2018). These two behaviors constituted the third and fourth behavioral assessments of attraction. 2

Results
We first created an index of affiliation suppression by inspecting the relation between affective attraction and enacted behavior. The use of discrepancy scores to investigate the relation between a behavioral enactment and affective attraction raises a number of concerns. Specifically, difference scores (i.e., suppression = affective attraction minus affiliative behavior) are associated with reliability problems (Peters, Churchill, & Brown, 1993), unlikely assumptions regarding the relation between variables (Cohen, Cohen, West, & Aiken, 2003), and interpretative issues, as a difference score may not differentiate between the various reasons for a discrepancy (e.g., similarity, superiority, or discrepancy; Griffin, Murray, & Gonzalez, 1999). Therefore, we employed response surface modeling (Box & Draper, 1987) to address these issues and to comprehensively model the relation between affiliative behavior, affective attraction, and suppression processes. First, the results from a polynominal regression using the predictors (i.e., an affiliative behavior, affective attraction) and outcome (e.g., shyness) were used to estimate four surface test values (a 1 , a 2 , a 3 , and a 4 ) for a three-dimensional rendering of the data. An interpretation of the surface values allowed for a description of the relation between the predictors and outcome variables.
a 1 evaluates the linear agreement between the two predictors (behavior and affective attraction) relative to the outcome (an affiliation suppression process). A significant positive a 1 value would indicate that both behavior and affective attraction linearly and positively covary with, say, shyness. Alternatively, a negative a 1 value would indicate a negative and linear relation between the two predictors and the outcome. a 2 refers to the non-linear agreement between the predictors relative to the outcome. A significant positive a 2 value would indicate that both behavior and affective attraction have a convex (upward curvature) relation with the shyness score, whereas a negative a 2 score would indicate a concave curvature. In other words, a negative a 2 score would indicate that high or low shyness scores would rise or fall more dramatically as both behavior and affective attraction change together.
a 3 and a 4 focus on the discrepancy between the predictors relative to the outcome. a 3 tests the perfect negative relation between the predictors relative to the outcome. A positive a 3 value indicates, for instance, that shyness is higher when the discrepancy is such that affective attraction is greater than the behavior (than vice versa). Alternatively, a negative a 3 value would indicate that shyness is greater when the difference between the predictors is such that behavior exceeds affective attraction. a 4 is conceptually related to a 2 , such that a significant positive a 4 value indicates a convex surface and a negative value indicates a concave surface as a function of the discrepancy between the predictors relative to the outcome.
Given that our interest is the discrepancy between predictors as it relates to a specific outcome variable, we are most interested in a 3 (and a 4 ), but for completeness and for descriptive purposes, we report all four indices.

Descriptive Analyses
Means, standard deviations, and correlations for the indices of attraction and affiliation suppression are presented in Table 2. In general, the reported level of affective attraction revealed that participants liked their partner (M = 5.51, SD = 1.05; 1 = not at all, 7 = very much). The correlations between self-reported affective attraction and affiliative behavior (i.e., smile duration, chair distance, expressed attraction, and eye contact) ranged between 0.21 and 0.63.

Suppression Analysis
Chair proximity. Chair proximity was calculated by averaging the distance between the front right and front left chair legs. To aid in the interpretation of this variable, chair distance scores were reversed so that higher values were associated with greater proximity. A description of the analysis for chair proximity, affective attraction, and suppression processes is presented in Table 3. As identified by the significant a 1 value, and as expected, participants reported the highest levels of perceived acceptance when both affective attraction and chair proximity were highest. Specific to our interests, perceived acceptance was also associated with a significant a 3 value, which suggests that perceived acceptance was greatest when affective attraction was high and was discrepant from chair proximity.
Eye contact and smiling. Perceived acceptance was associated with a significant a 1 value, indicating that perceived acceptance was greatest when eye contact and affective attraction were highest. Avoidant attachment was associated with a significant a 1 value for both smiling and eye contact, revealing that avoidant attachment was highest when affective attraction and the behaviors were highest, and that avoidant attachment was lowest when affective attraction and the behaviors were lowest. The a 2 value for avoidant attachment and eye contact indicates that this effect was more prominent for those participants who were particularly high and low on avoidant attachment, such that it was those participants with the highest levels of avoidant attachment who made the most eye contact and liked the most. 3 Expressed attraction. Perceived acceptance was associated with a significant a 1 value, indicating that perceived acceptance was highest when both affective attraction and expressed attraction were high. With respect to affiliation suppression, avoidant attachment was associated with a marginal a 3 value, an effect consistent with the finding that participants were high in avoidant attachment when affective attraction exceeded expressed attraction.

Discussion
The results of Study 2 provided support for the importance of affiliation suppression processes to understand the difference between affective attraction and expressed behavior. Our results largely align with the findings of Study 1, such that the self-generated reasons for affiliation suppression were similar to processes active in a controlled laboratory study (e.g., perceived acceptance).
We did, however, identify unexpected effects for those who scored high on the avoidant attachment measure. Avoidant participants were associated with more eye contact and smiling. However, people who scored high on avoidant attachment only behaved more favorably when they also reported liking the person; in other cases, avoidant participants demonstrated less favorable behavioral responses (less eye contact, smiling). This finding indicates that there was a desire to initiate and develop rapport when they liked the confederate, but this was expressed behaviorally via 'indirect' behaviors (e.g., eye contact) versus relatively more ' direct' behaviors (e.g., smiling and explicitly stating interest). The magnitude of effect sizes in Study 2 align with the magnitude of effect size for nonverbal behaviors identified in a recent metaanalysis (Montoya et al., 2018).
One question left unanswered by this analytic approach was identifying which affiliation suppression process was primarily responsible for the difference between the behavioral response and affective attraction. Whereas such an analysis would involve controlling for the various affiliation suppression processes (as covariates while investigating a particular process), current response surface modeling methods do not allow for such a test. However, close inspection of the data reported in Table 2 reduces such concerns. Specifically, (a) the low bivariate correlation between the various suppression processes and (b) the lack of consistent significant bivariate correlations between the suppression processes with the affiliative behavior indicates that the inclusion of a covariate would neither be warranted nor impactful (Little, 2013).
An important issue for this study is whether the effects identified for male participants would generalize to    female participants. As evident in Study 1 and a recent meta-analysis (Montoya et al., 2018), affiliation suppression operates in both men and women, and affiliative behaviors, relative to attraction, are expressed similarly in both romantic and nonromantic contexts. As a result, we are confident in the external validity of Study 2's findings. Regardless, in Study 3, we investigated affiliation suppression in both men and women.

Study 3
Study 3 was designed to replicate and extend Study 2. Whereas Study 2 provided initial evidence, we conducted another study to investigate suppression using a different method. In Study 3, we explored suppression using an actual-ideal discrepancy task. The actual-ideal task presented participants with a series of scenarios and asked how they might actually and ideally behave toward the person to whom they are most romantically attracted. We expected suppression processes to predict when ideal behavior (e.g., 'Ideally, I would talk to someone I "like".') would exceed actual behavior (e.g., ' Actually, I would not talk to the person.').
We also explored the emotional consequences that may result from affiliation suppression. Several research traditions propose that people experience negative emotional consequences when affective experiences do not match their behavioral reactions (Gross, 2015). At a basic level, dissonance theory (Festinger, 1957) proposes that individuals experience an aversive state when their affect does not match behavior. Furthermore, with respect to discrete emotions, research on moral emotions (guilt, shame, embarrassment;Tangney, 2003) postulates that such specific emotional responses result from self-reflection and evaluation. In the case of affiliation suppression, guilt may result when past behavior is evaluated negatively relative to a higher standard (e.g., 'I let myself down by not talking to him/her.'). Alternatively, shame results when one negatively evaluates one's global self (Tracy & Robins, 2004). Additionally, self-discrepancy theory (Higgins, 1987) posits that negative affective states result when one's actual self differs from one's ideal or ought self. Negative emotional reactions result when one's actual behavior (actual self) differs from one's ideal behavior (ideal self).
As with Study 2, we assessed several processes that may affect participants' various attraction responses (e.g., attachment style), but that we did not have any specific predictions regarding suppression. We also included an assessment of familiarity as it has been hypothesized to result in more affective and behavioral attraction (Moreland & Beach, 1992). Familiarity may reduce affiliation suppression by facilitating affective attraction by ' allowing' one's behavior to be expressed to the same degree as experienced affective attraction.

Method Participants
Two hundred and ninety-three participants (150 women, 141 men, and 2 did not specify sex; M age = 19.22, SD = 1.24) were recruited at a midsized, Midwestern university to complete the study for course credit in an introductory psychology course.
The sample size was determined by comparing the present study to previous research using the actual-ideal discrepancy task (e.g., Zuckerman, Gagné, Nafshi, Knee, & Kieffer, 2002). Only participants who were not currently in a romantic relationship completed the study.

Materials
Emotion measure. The Differential Emotion Scale (Izard, 1971) evaluated participants' emotional reactions to discrepancies between their ideal and actual responses. On a scale from 1 (does not apply at all) to 7 (applies very much), participants rated affective words categorized within the following subscales: joy, distress, anger, disgust, fear, guilt, shame, and embarrassment. Participants were instructed to choose the degree to which the adjective describes their current feeling.
Familiarity. Two questions assessed how well the participant knew the target person. Items included: 'How well do you know the target person?' and 'How frequently do you speak to the target person?' The questions were evaluated on scales ranging from 1 (not at all/never) to 7 (very well/very frequently).
Satisfaction with life. The Satisfaction with Life Scale (SWLS) evaluated how presently satisfied participants were with their lives (Diener, Emmons, Larsen, & Griffin, 1985). Five questions were rated on a 7-point scale ranging from 1 (strongly disagree) to 7 (strongly agree).

Procedure
Participants completed the questionnaires online using SurveyMonkey. Participants first completed the same questionnaires included in Study 2 (i.e., attachment style, shyness, and the self-esteem questionnaire). Next, to identify the discrepancy between actual and ideal responses, we adapted a task originally used to measure prejudice/discrimination (by Devine, Monteith, Zuweink, & Elliot, 1991) to measure attraction responses. In our procedure, participants responded to six hypothetical scenarios (created by the authors) regarding how they would ideally and actually act toward the person to whom they were most attracted. One scenario read, 'While at a party, you notice that the person to whom you are most attracted is standing alone. [Ideally,] You walk over and talk with the person.' All the scenarios are presented in the online supplemental materials. Each question began with the prompt, 'Based on how you wish would act or how you would ideally act [Based on how you believe you would actually act], indicate the number between 1 (strongly disagree) and 7 (strongly agree), that best reflects your response to the situation.' Next, participants completed the emotion measure, the questionnaires described in Study 2 (affective attraction, perceived acceptance, social norms, and anticipated effort), and the familiarity and satisfaction with life questionnaires.

Results
We began by computing aggregate actual behavior and ideal behavior scores by summing participant responses to the six scenarios (actual behavior, α = 0.78; ideal behavior, α = 0.72). Discrepancy scores were initially estimated by subtracting actual behavior scores from ideal behavior scores. Of the 293 responses, 50 produced negative suppression indices, indicating greater actual than ideal behaviors. Because of our theoretical interest in instances in which participants did not express positive behavior to match their positive emotional experience, these individuals were excluded from all analyses (Devine et al., 1991).
Descriptive analyses. Reported levels of affective attraction (M = 5.43, SD = 1.04) indicated that individuals were able to think of a person to whom they experience attraction, with men reporting comparable levels of attraction ( Correlations between the affiliation suppression processes and indices of suppression are presented in Table 4. As may be expected, affective attraction was correlated positively with familiarity, target availability, and negatively correlated with anticipated effort. Table 5. First, as identified by the significant a 1 values, most processes, as expected, covaried with the agreement of actual/ideal behavior, such that higher actual/ideal responses covaried with more affective attraction, salience of social norms, and greater familiarity, but lower levels covaried with avoidant attachment and more shyness. The negative a 1 value for avoidant attachment indicates that higher avoidant attachment scores were associated with lower actual and ideal behavior scores, but the nonsignificant a 3 score indicates that avoidant scores were not associated with affiliation suppression.

Suppression processes and actual/ideal behavior. A description of the relation between actual/ideal responses and suppression processes is presented in
Second, several processes were associated with suppression. Specifically, shyness, injunctive norms, and familiarity were associated with significant a 3 values. For shyness, as illustrated in Figure 1, the a 3 value revealed that for persons who scored high on shyness, ideal responses exceeded actual responses; with the a 4 value indicating that this pattern was particularly evident for those highest in shyness. The a 3 value for familiarity revealed that low familiarity was associated with higher ideal behavior relative to actual behavior, a pattern consistent with affiliation suppression.
We also assessed the responses separately for men and women. With respect to the processes that were significant for men or women, shyness and target person availability produced a significant a 3 effect for women but not for men, whereas avoidant attachment style, shyness, and descriptive norms were associated with curvilinear effect for affiliation suppression for men.
Suppression processes and emotional consequences. As presented in the bottom half of Table 5, greater actual and ideal behavior was associated with lower levels of aversive emotions, including disgust, fear, shame, and embarrassment, with such distress, anger, disgust, and shame significant for men, and fear, guilt, and embarrassment significant for women. With respect to affiliation suppression, three effects were identified. For the overall sample, as illustrated in Figure 1 (bottom), the a 3 score for guilt indicates that guilt was highest when ideal behavior exceeded actual behavior. Sex differences reveal that these effects were significant for women but not men.

Discussion
Study 3 provided support for the importance of several affiliation suppression processes. Specifically, suppression was associated with more shyness, descriptive and injunctive norms, familiarity, and greater target person availability (in women). Interestingly and importantly, affective attraction was not associated with a significant a 3 effect. The lack of an effect is consistent with the conclusion that more attraction is not necessarily indicative of a greater association between actual and ideal behavior.
For the emotional consequences, affiliation suppression was paired with several negative emotional consequences, including distress and guilt (e.g., Butler et al., 2003). These findings are consistent with self-discrepancy theory and models of moral emotions, such that these models predict the presence of guilt when individuals negatively evaluate their own (in)actions. In this case, women whose actual behavior did not match their ideal behavior experienced more guilt and distress.

General Discussion
The goal of this research was to explore the processes that reduce affiliative behavior given the presence of affective attraction. In Study 1 (narrative study), the most common reasons for affiliation suppression were shyness/anxiety and concerns regarding acceptance. Study 2 (laboratory study) identified perceived acceptance as related to affiliation suppression. Study 3 (discrepancy task) provided further support for the conclusions of Study 1 by noting that shyness, norms, and familiarity were associated with affiliation suppression. Study 3 further extended the findings of Study 2 by revealing that there are emotional consequences (guilt, distress, and less joy) associated with suppression, particularly for women.
By considering the affiliation suppression processes that differentiate affiliative expressions from affective attraction, it is possible to predict when and why affiliative behavior will or will not align strongly with affective responses. The findings from Studies 1-3 provided clear evidence that several processes affect the expression of affiliative behavior relative to affective attraction. Specifically, perceived acceptance, shyness, familiarity, and social norms-each predicted by TDMA-played a role in affiliation suppression. For instance, the importance of self-threat considerations was highlighted by finding that familiarity and shyness both differentially speak more to affiliative expressions than to affective attraction. The results of Study 3 indicate that greater familiarity reduced the difference between actual and ideal behavior, or in Table 4: Means, standard deviations, reliabilities, and correlations between affiliation suppression processes, Study 3.  other words, familiarity facilitated individuals acting on how they feel. The correlation is consistent with the assertion that familiarity generates an environment in which individuals can act on their experienced attraction. Furthermore, the finding that familiarity was associated with greater ideal behavior for women, but not men, indicates that familiarity facilitates relationship development via increases in women's desire for a relationship. Notably, we did not find evidence in Study 3 that affective attraction predicted affiliation suppression. This finding further underlines the point that affective attraction (e.g., 'I like Person X') should neither be used in measurements of interpersonal attraction with measures of behavioral attraction (e.g., 'I would like to hang out with Person X') nor with indices of affiliative behavior (e.g., number of smiles, sitting proximity). These studies provided evidence for why affective and behavioral expressions may differ. Indeed, these findings are consistent with research that has identified differences between affective and behavioral expression of attraction, and proposed that such processes should be considered separate, but often highly correlated, processes (e.g., Montoya & Insko, 2008).
In addition, our findings align with and expand on extensive research on emotion regulation in dating couples (e.g., Levenson et al., 2014). Whereas the extant literature has investigated a number of negative consequences (e.g., lower relationship satisfaction) associated with emotion regulation and on the inhibition of a momentary emotion or the expression of a negative emotion (fear, anger; Levenson et al., 2014), the present research investigated the suppression of a 'positive' emotional experience and its negative consequences (in Study 3). Future research on the suppression of the expression of love/liking/ affection in extant romantic relationships may provide further insight into methods for enhancing well-being and relationship satisfaction.
Finally, we acknowledge that our data were collected using retrospective methods (as in Study 1) and using hypothetical scenarios (as in Studies 3), methods that are subject to the potential shortcomings, including dissonance (Study 1) and social desirability (Study 3). However, despite possible shortcomings, we found consistent evidence with a controlled laboratory study and a narrative study. Such consistency works to counter concerns regarding potential confounds associated with survey or narrative methods.

Notes
1 A fourth consideration, instrumental concerns, can also produce a discrepancy. People may instrumentally communicate more liking than they feel (Montoya et al., 2018). People will act as if they like someone to heighten the likelihood that other person will act benevolently during a future social exchange (i.e., flirting, brownnosing, kissing up). Because instrumental considerations involve exhibiting more affiliative behavior than affective attraction, they are not considered here. 2 We also manipulated the order of assessment. Specifically, participants in the 'before' condition completed the affiliation suppression questionnaires (e.g., perceived acceptance, partner availability, anticipated effort, and the affective attraction/affiliative behavior) before meeting the confederate. Participants in the 'after' condition completed those same questionnaires after meeting the confederate. An initial investigation into the assessment order manipulation produced no meaningful differences in affective attraction, affiliation behavior, levels of affiliation suppression predictors, or the interrelations between these variables. As a result, we do not consider the variable further. 3 We also coded smile frequency and eye contact frequency. Results for these two measurement techniques did not meaningfully differ from the reported duration indices.