1 Please Process the Signal, but Don’t Praise It: How Compliments on Identity Signals Result in Embarrassment LISA A. CAVANAUGH JOSEPH C. NUNES YOUNG JEE HAN Lisa Cavanaugh is Assistant Professor of Marketing at the Marshall School of Business, University of Southern California, Los Angeles, CA 90089-0443, lisa.cavanaugh@marshall.usc.edu. Joseph C. Nunes is Professor of Marketing, Marshall School of Business, University of Southern California, Los Angeles, CA 90089-0443, joseph.nunes@marshall.usc.edu. Young Jee Han is Assistant Professor of Marketing at Sungkyunkwan University, Seoul, South Korea, youngjee.han@gmail.com. Questions should be directed to either lisa.cavanaugh@marshall.usc.edu or jnunes@marshall.usc.edu. 2 CONTRIBUTION STATEMENT This research examines identity signaling as a two-way process by looking at the interaction between signaler and a signal’s recipient. In doing so, we document how praising a consumer’s identity signal can result in an arguably unforeseen negative response. Specifically, we show that compliments related to an identity signal often result in embarrassment, a generally unwelcome self-conscious emotion. We investigate the moderating role played by signal strength in terms of the signal being loud versus quiet (i.e., more or less observable, recognizable, and interpretable, respectively). This research also documents how congruency between consumers’ self-beliefs and how signalers would like others to see them impacts their emotional response. Across four studies, we show compliments related to a loud identity signal often result in embarrassment and that people are most embarrassed when the identity signal and one’s beliefs about oneself are incongruent. Finally, we examine possible boundary conditions to the effect prevalent in social situations involving signaling by varying: 1) whether the compliment giver is a friend or stranger, 2) the type of compliment or remark made, and 3) whether a compliment giver exhibits a similar signal at the time of the event. ABSTRACT As social beings, consumers seek approval from others and therefore frequently make an effort to communicate aspects of their identity, actual or ideal, with the intention of impressing those around them. They often express these aspects (e.g., being smart, funny, athletic, stylish, or sexy) through the products they use and brands they exhibit. Conspicuous branding helps consumers insure their signal will be noticed and validated by others. One of the most common forms of validation is a compliment. However, we know surprisingly little about how compliments, particularly compliments on identity signals, impact consumers. This paper explores how consumers respond to compliments related to an identity signal. Four studies show such compliments often result in embarrassment, an arguably unforeseen and generally unwelcome selfconscious emotion. We find the amount of embarrassment experienced depends on the conspicuousness of the signal as well as the extent to which the signal and one’s beliefs about oneself are incongruent. The emotional response is explained further by public self-awareness. Consumers experience embarrassment regardless of whether the compliment comes from a friend or stranger and whether or not the identity signal itself is explicitly acknowledged. 3 “Jennifer Metcalfe: Don't call me sexy – it’s embarrassing” Headline in the Daily Mirror newspaper When actress and model Jennifer Metcalfe stripped down to a miniscule bikini on a Spanish beach for a photo shoot, locals walking by who recognized the former fitness instructor called out “Wow, sexy lady” loudly enough for Metcalfe to hear. Metcalfe responded by giggling, hiding her face in a towel and saying “Oh my God, I’m embarrassed” (mirror.co.uk 2010). For someone whose public persona encompasses being perceived as a sex symbol (she was listed as #46 in FHM magazine’s world’s top 100 sexiest women that year), her response exemplifies the phenomenon under investigation; people often become embarrassed after receiving a compliment on some aspect of their identity they actively promote. Metcalfe apparently does not see herself as quite as sexy as the sex symbol she portrays herself to be. “These photo shoots come with the job and I know I’ve got to do it and be this sexy person,” she is quoted as saying. “It’s not that I’m insecure as a person, it’s just completely normal to look at your body and see insecurities.” Whether it’s being seen as sexy, sophisticated, intelligent, or affluent, people care how others perceive them. Impression management is the process by which people attempt to influence others’ perceptions of them, and individuals regularly engage in identity signaling as a means of influencing others’ perceptions (Schlenker 1980). According to signaling theory, “one party, the sender, must choose whether and how to communicate (or signal) information, and the other party, the receiver, must choose how to interpret the signal” (Connelly et al. 2011, 39). In consumption, individuals regularly use possessions and brands to signal various aspects of their identities to others (e.g., Escalas 2004; Fournier 1998; Belk 1988; McCracken 1989). Consumers 4 expect observers to make certain inferences about their identities based on which possessions and brands they choose to display and how they display them (Belk, Bahn, and Mayer 1982; Burroughs, Drews, and Hallman 1991; Richins 1994a, 1994b). Whether it is standing on the beach in a skimpy bikini, sporting clothes with prominent brand names or logos, wearing horn rimmed glasses, or driving up to a crowded event in an iconic car, consumers anticipate the signal will be seen and understood by others. What consumers are less likely to anticipate is how they themselves might respond to feedback on the signal. This research is the first of which we are aware to study the dyadic process of identity signaling vis-à-vis consumption. Previous research has investigated either: 1) consumers’ intentions and behaviors with respect to signaling with products/brands (Berger and Heath 2007; Dubois, Rucker, and Galinsky 2012; Wang and Griskevicius 2014) or 2) how recipients interpret such signals (Bellezza, Gino, Keinan 2014; Shalev and Morwitz 2012; Sirgy 1982). Instead, we study identity signaling as a two-way, iterative process with consequences for the signaler. Signal recipients often provide feedback, such as acknowledging a signal with a compliment. (For example, consider the common retail practice of complimenting customers on brand choices or aspects of their appearance.) Yet, surprisingly little is known about how signalers respond to compliments on their identity signals. In this research, we focus on the signaler’s emotional response to a positive acknowledgment of an identity signal—that is, a compliment. By focusing on interactions between individuals and the emotional consequences of compliments in this context, this work links the literature on identity signaling in marketing with the literature on self-conscious emotions. We show that compliments on identity signals under certain circumstances cause consumers embarrassment owing to heightened public self-awareness. We also show that the 5 extent to which the signaler feels embarrassed by a compliment depends on characteristics of both the signal and the signaler. We observe a similar pattern of effects across both a field study (study 1) and two different types of controlled experiments examining the effects of the signal and the signaler’s self-beliefs using both deception (study 2) and self-identified gaps (studies 3 and 4), using both self-reported (studies 2-4) and observed measures of embarrassment (studies 1 and 2), as well as in studies in which consumers both self-selected (studies 1, 3, and 4) and were randomly assigned (study 2) their identity signal. The remainder of this paper proceeds as follows. First, we briefly review the relevant literature on identity signaling, compliments, and embarrassment. Next, we present our conceptual framework followed by four studies supporting it using a multi-method approach. A mixed methods approach generally is viewed as providing more effective investigative tools than an isolated qualitative or quantitative approach (Tashakkori and Teddlie 2003). In study 1, a field study documents how consumers who have self-selected a particularly conspicuous as opposed to inconspicuous identity signal are more likely to feel embarrassed after receiving a compliment. Study 2 replicates the effect as well as documents how individual self-beliefs with respect to the validity of the signal help determine the extent to which a consumer becomes embarrassed. Studies 3 and 4 reinforce the robustness of the compliment effect across different types of compliment-givers (friends and strangers) and in contrast to more generic remarks addressed to the signaler. These last two studies also offer additional process evidence for the role of public self-awareness as a principal mechanism. We conclude by discussing the limitations of this work and offering avenues for future research. 6 IDENTITY SIGNALING, COMPLIMENTS, AND EMBARRASSMENT Individuals attempt to manage others’ impressions of them in ways they believe are in their best interest through their appearance and behavior, (Goffman 1959; Jones and Pittman 1982). Thus, strategic self-presentation is anchored in social motives. Through the construction of a desirable social identity, a person attempts to bring his social self closer to an ideal he expects to positively influence the way others treat him (Rosenfeld, Giacalone, and Riordan 1995). Previous research in marketing has shown consumers use possessions to communicate various aspects of the self (Belk 1988), and these possessions often include favored brands (Han, Nunes, and Dréze 2010). Using a brand to convey an otherwise unobservable aspect of one’s identity serves an explicit identity signaling purpose (e.g., I ride a Harley to show others I am a rebel). In this way, brands help consumers manage their social identities (Kleine, Kleine, and Allen 1995; Berger and Heath 2007). Consumers who employ identity signals are looking to garner attention from observers, make a chosen impression, and hopefully receive the desired response (Schlenker and Leary 1982). Observers, however, may not provide the intended response. For example, they may make negative inferences when the identity signal (e.g., brand) is seen as being driven by extrinsic needs for social approval versus intrinsic factors (Ferraro, Kirmani, and Matherly 2013). Knowing that others will appraise their signals presents both opportunities and potential liabilities for the signaler. Signalers want to be perceived as possessing those aspects of their identity signaled; however, they may worry about being perceived as sending a phony signal or behaving in a manner purely for identity signaling purposes. How the signaler responds to an appraisal of the signal predictably will depend on qualities of both the signal and the signaler. First, an identity signal can be deemed as more or less purposeful depending on whether 7 it is more or less conspicuous, referred to as loud versus quiet in the context of brands (Han et al. 2010). Consistent with Han et al.’s notion of brand prominence, we define the more general “signal strength” construct as the ease with which any identity signal (not just a brand) is: (1) observable, (2) recognizable, and (3) interpretable. Observable refers to how obvious the signal is to observers. Recognizable refers to how readily identifiable the signal is to observers (e.g., Louis Vuitton’s checked Damier print, the Oakland Raiders’ iconic logo of a pirate wearing a football helmet, the distinctive scent of a woman’s perfume). Interpretable refers to how decipherable the signal is in terms of what the signaler is trying to communicate (e.g., I am affluent, I am an Oakland Raiders’ fan, I am sexy). For simplicity’s sake, we maintain the same nomenclature as Han et al. (2010); a “loud” signal as opposed to “quiet” signal is considered easier to observe, recognize, and interpret and is usually perceived to be a more purposeful promotion of certain aspects of one’s identity. Second, identity signals normally reflect how the signaler would ideally like to be portrayed. Despite a sometimes confusing array of “self” terms in the literature (e.g., selfconcept, self-esteem, self-efficacy), notions of self share a common emphasis on an individual’s beliefs about his or her attributes and abilities as a person (Hattie 1992). The extent to which the signaler actually possesses the underlying attribute associated with the signal has been referred to by a host of differing terms including signal honesty (see Durcikova and Gray 2009) and signal veracity (Busenitz, Fiet and Moesel 2005). Importantly, whether the signal is objectively true or not is largely irrelevant here (as our opening example with Jennifer Metcalfe illustrates). Yet, how one depicts oneself as compared to one’s own prevailing self-belief regarding possessing the attribute associated with the signal is critical and plays a large part in determining how someone responds to feedback on his or her public portrayal. 8 Certain situations may make signalers acutely aware of their own self-presentation; one such situation is when a signaler receives a compliment. A compliment is defined broadly as a “speech act which explicitly or implicitly attributes credit to someone other than the speaker, usually the person addressed, for some ‘good’ (possession, characteristic, skill, etc.) which is positively valued by the speaker and the hearer” (Holmes 1986, p. 485). Beyond their apparently transparent function as a positively affective speech act, compliments are particularly interesting to study because, as anthropological linguists have pointed out, giving and responding to compliments is part of a much more complex social dynamic between the speaker (the signal recipient) and the hearer (the signaler). A compliment makes the signaler aware that the compliment-giver has noticed something about the signaler’s “interests, needs and wants” that warrants a response (Brown and Levinson 1978, p.108). Generally speaking, deciding how to respond to any compliment creates psychological tension due to two competing maxims of speech behavior: 1) the agreement maxim and 2) the modesty maxim (Pomerantz 1978). On the one hand, the “agreement” maxim compels the hearer to agree with the speaker by accepting the compliment. On the other hand, the “modesty maxim,” compels the hearer to be modest and avoid self-praise by deflecting or evading the compliment. Thus, a compliment creates internal tension—agree with the compliment at the risk of being immodest versus disagree with the compliment at the risk of creating social strain by not agreeing. This internal tension and self-awareness about how one may appear to others is likely to manifest itself in an outward emotional response. Previous work suggests that being the center of attention can cause embarrassment (Sabini et al. 2000). Embarrassment is a negative self-conscious emotion resulting from presumed deficiencies in one’s presented self (Goffman 1955; Tangney et al. 1996; Tracy and 9 Robins 2004). As it involves attentional focus on the public self-representation, embarrassment is distinguishable from other negative self-conscious emotions (e.g., guilt or shame) that can be experienced even when someone is alone (Keltner and Buswell 1997; Tangney et al. 1996; Tracy and Robins 2004; but see also Krishna, Herd, and Aydinoglu 2015). Although multiple theories of embarrassment have been described in the literature (see Keltner and Buswell 1997 for a review), the two most influential models describe embarrassment as a response to: (1) an awkward social interaction and (2) a concern regarding negative social evaluation. The former, or awkward interaction model, suggests embarrassment results from uncertainty and loss of direction within a social interaction while the social evaluation model suggests embarrassment arises from one’s concern for what others think rather than a discontinuity in communication. Our conceptualization of how consumers respond to compliments on identity-signals builds on insights from both of these models. With respect to the awkward interaction model, a loud (vs. quiet) signal intensifies the conflict arising from not knowing how to respond (i.e., whether the recipient should agree or avoid self-praise). The recipient becomes less sure of what to say next and is less able to gracefully continue the interaction, resulting in embarrassment. With respect to the social evaluation model, when the signaler feels out-of-sync with a presented social identity (a signal that is incongruent with one’s self-beliefs), the possibility of unwanted negative social evaluation – imagined or real – results in embarrassment. CONCEPTUAL DEVELOPMENT: THEORY AND HYPOTHESES Building from previous work examining how people respond to compliments (Pomerantz 1978), we propose that the internal tension between the agreement and modesty maxims is exacerbated when a consumer receives a compliment on an identity signal. This occurs because 10 the signaler who is intentionally self-promoting an aspect of his or her identity: 1) recognizes that he or she has done something deliberate to draw attention to the self (i.e., signaled) and 2) experiences apprehension regarding how the self appears or is perceived by others (i.e., as truly having the identity vs. simply using an identity prop). A compliment focuses the signaler’s attention on his or her own self-presentation motives and raises awareness of having executed a calculated self-enhancement strategy. This realization may also increase concerns that the strategy could be or become obvious to others. Thus, a compliment on an identity signal generally heightens public self-awareness for the signaler. To be clear, public self-awareness concerns how the self appears to others, that is considering how I may appear to others (Govern and Marsch 2001). The public self consists of the overt, externally observable aspects of self, such as appearance. In contrast, the private self consists of those aspects of the self that cannot be observed by others. An identity signal purposefully makes aspects of the self that otherwise cannot easily be observed by others observable. Public self-awareness is typically accompanied by a certain discomfort and evaluation apprehension due to feelings of being the subject of others’ appraisals (Govern and Marsch 2001, p. 367). If a signaler perceives the self as outwardly overstating aspects of his or her own identity, public self-awareness can cause feelings of embarrassment. The extent to which the signaler feels embarrassed by a compliment on an identity signal will likely depend on characteristics of both the signal (i.e., one’s outward display) and the signaler (i.e., one’s inward self-beliefs). First, an identity signal can be deemed as more or less purposeful depending on the signal strength (i.e., whether it is loud or quiet. Loud signals are more likely to exacerbate selfpresentational concerns because they are more obvious and leave less room for ambiguity about 11 self-representation. In other words, a compliment on a loud signal (relative to a quiet signal) increases the salience of the signaler’s own self-presentation motives and heightens the potential for anxiety about how one appears to others. Thus, we expect loud signals to elevate both the signaler’s public self-awareness and concomitant feelings of embarrassment. Second, a signaler’s self-beliefs can also contribute to the emotional response. More often than not, an individual’s self-belief differs from an ideal identity (e.g., I am not as a successful as I would like to be). If a signaler believes s/he does not fully possess the attribute(s) signaled, a compliment may cause the signaler to assess the extent to which the compliment was “internally caused” (i.e., brought on by an aspect of the “self”) versus caused by the signal (i.e., brought on by an identity prop). We expect such reflection and the realization of any self-perceived identity incongruence to similarly result in heightened public self-awareness and concomitant feelings of embarrassment. Consistent with this notion, Buss (1980) posits that blushing (a display of embarrassment) can occur in situations where people are overpraised, that is, when others’ evaluations are perceived by the signaler as more positive than warranted. An identity signal that overstates a signaler’s self-belief regarding an identity may imply that a compliment is unwarranted. If the identity signal is accurate, the signaler will be less likely to be concerned about being overpraised. However, if the signaler believes the signal exaggerates characteristics of the self, he or she is more likely to experience discomfort over how he or she appears to others that is, unless the signal recipient is signaling in the same way (for the sake of brevity, the question of Signal Similarly is taken up in more detail the General Discussion). 12 FIGURE 1 Conceptual Framework Our complete conceptual framework is presented in Figure 1. To summarize, we expect a compliment on a relatively loud signal to result in greater public self-awareness and embarrassment. Additionally, when a signal is relatively quiet, a compliment causes embarrassment when the signal promotes an aspect of one’s identity that is incongruent with the signaler’s self-belief. Stated differently, only when a signal is relatively quiet and congruent with one’s self-beliefs is the signaler unlikely to become embarrassed after being complimented. Finally, we identify an additional moderator of the effect, documenting how a compliment coming from someone perceived to be sending a similar signal can mitigate these effects. STUDY 1: EMBARRASSING PRODUCTS, EXPECTED AND UNEXPECTED In study 1, we test the prediction that a signaler utilizing a loud (quiet) signal will be more (less) embarrassed after their signal is acknowledged with a compliment. We do so in a field study in which a researcher interviewed women carrying luxury handbags in public. We chose handbags as our target product category in this study as they are widely regarded as a selfexpressive symbol of wealth and status and differ significantly with respect to signal strength. 13 Further, it seems counterintuitive to expect a compliment to embarrass someone who self-selects to signal status in this way. The compliment was delivered during an interview conducted under the pretense of a study investigating general shopping behavior. We used Layered Voice Analysis to assess whether women interviewed while exhibiting loud signals were more embarrassed by the compliment than their counterparts exhibiting quiet signals. Layered Voice Analysis Pilot Study Layered Voice Analysis (henceforth LVA) is software developed by Nemesysco Ltd., a provider of voice analysis technologies to fields ranging from security (e.g., border control, police) to insurance and banking providers. The software utilizes a set of proprietary signal processing algorithms to extract vocal parameters from each voice segment and identify stress levels and various types of emotional responses (embarrassment, anger, happiness, and sadness). By measuring changes in vocal patterns, LVA can help researchers efficiently measure changes in emotional states in a natural setting. The software creates an emotional profile by measuring changes in high and low frequencies in the speech waveform and classifying them in terms of emotional states correlated with these changes. Because waveforms are person specific, the software requires a calibration period for each individual to measure inherent characteristics of the speaker’s voice. Accordingly, the software creates an instantaneous benchmark when a person begins speaking after which it measures deviations at constant intervals that can range in length from 4/10th of a second to two seconds. The program then assigns a numerical score between 0 and 30 to each deviation and produces an emotional profile for each vocal segment across time (see figure 2 for an example). 14 FIGURE 2 Study 1: Example of an Emotional Profile from LVA Prior studies have tested the validity of LVA in distinguishing between truthful and deceptive speech segments (Adler 2009; Damphousse et al. 2007). Further, research published in the Journal of Finance by Mayew and Venkatachalam (2011) utilized LVA effectively to show changes in speech patterns among executives provide important clues about the financial prospects of their firms. The authors found executives’ affective states during earnings calls as measured by LVA conveyed important information about their emotional state that effectively predicted their firms’ future profitability and returns. However, we were unaware of any published studies that employed LVA metrics to capture changes in a single emotion. To reassure the reader of the reliability of the LVA technology’s (software version QA5) ability to adequately capture changes in embarrassment per se, we conducted the following pilot study. Fifty undergraduate business students (35 male and 15 female) at a major business school participated in exchange for course credit. An interview with each respondent was conducted and recorded digitally. Each interview was conducted individually in two phases. The first phase 15 consisted of a calibration period during which respondents’ answers to a pre-specified question were expected to require little emotional processing. In this phase, the interviewer asked all respondents to describe how college students typically spend their days. The second phase was described as a separate study intended to assess college students’ familiarity with a variety of products. Respondents were presented either with condoms, an embarrassing product (Dahl, Mamchanda, and Argo 2001), or instant oatmeal, a non-embarrassing product. Following the interview, respondents completed a questionnaire regarding their feelings while describing condoms or instant oatmeal. We determined the efficacy of the software by examining how LVA’s metric for embarrassment changed across phases between the two products as well as the relationship between LVA’s embarrassment measure and respondents’ self-reported measures of embarrassment. First, using Adobe Audition software 3.0, we removed the researcher’s voice from each recording. The resulting audio files were used to calculate an average level of embarrassment for each individual for each phase of the interview. If the LVA software was successful in capturing embarrassment, we would expect a difference between the two phases for those who described the condoms but not for those who described the oatmeal. This is exactly what we observe. Analyses of the voice samples revealed equivalent levels of embarrassment at baseline before participants were asked to provide descriptions of condoms or oatmeal (Mpre-condom = 3.37; Mpre-oatmeal = 3.04; F<1, NS). However, after providing the descriptions, the voices of individuals describing condoms registered higher levels of embarrassment than those describing oatmeal (Mcondom = 5.34; Moatmeal = 3.03; F(1, 48) = 17.94, p < .0001). Using the change in embarrassment (time 2 – time 1), we found similar results; those who described condoms exhibited an increase 16 in embarrassment (Mcondom = 1.98; Moatmeal = -0.01; F(1, 48) = 11.40, p < .001). No significant differences or changes were found for any of the other emotions captured by LVA (see table 1). TABLE 1 Pilot Study (Condoms vs. Oatmeal): LVA Analyses Condoms Oatmeal Mean (SD) Mean (SD) Embarrassment T1 Embarrassment T2 Change in Embarrassment Range 3.37 (2.17) 5.34* (2.04) 1.98* (2.14) 3.04 (1.52) 3.03 (1.74) -0.01 (1.96) 0 to 7.64, M = 3.22 (1.90) 0 to 9.50, M = 4.33 (2.22) -4.00 to 7.39, M = 1.10 (2.27) 0.02 (.11) 0.20 (.73) 0.18 (.74) 0.24 (.89) 0.16 (.44) -0.08 (.55) 0 to 4.19, M = .12 (.60) 0 to 3.79, M = .18 (.62) -2.27 to 3.79, M = -.06 (.67) “Happiness/Excitement” T1 “Happiness/Excitement” T2 Change in “Happiness/Excitement” 15.72 (1.17) 16.08 (1.53) 0.36 (1.32) 15.98 (1.49) 16.31 (2.01) 0.33 (1.64) 13.08 to 20.50, M = 15.83 (1.31) 12.35 to 21.64, M = 16.18 (1.74) -2.73 to 3.66, M = 0.34 (1.46) “Sadness” T1 “Sadness” T2 Change in “Sadness” 4.19 (3.22) 5.20 (3.55) 1.00 (3.48) 3.09 (2.08) 4.53 (2.51) 1.45 (3.10) 0 to 14.42, M = 3.71 (2.80) 0 to 14.00, M = 4.90 (3.12) -3.68 to 8.47, M = 1.20 (3.29) 7.87+ (2.12) 6.45 (1.80) -1.42 (1.59) 6.83 (2.08) 6.10 (1.86) -0.74 (1.55) 2.75 to 12.25, M = 7.41 (2.14) 2.32 to 10.14, M = 6.30 (1.82) -5.09 to 2.17, M = 1.12 (1.59) Anger T1 Anger T2 Change in Anger Stress T1 Stress T2 Change in Stress Note: *p < .001, +p < .10; for all other emotions p > .17 Further, we find a significant correlation between the embarrassment scores measured by LVA during the second phase and self-reports of embarrassment collected from respondents at the conclusion of the task (r = .41, p < .01). Taken together, these results affirmed the ability of the LVA software to detect changes in embarrassment in voice recordings and allowed us to proceed confidently with our field study. Main Study 17 In the main study, we test whether consumers exhibiting loud signals will be more embarrassed after receiving a compliment than those exhibiting quiet signals. This is accomplished by recording the interactions outside of a lab setting and subsequently analyzing the voice recordings, including their emotional responses to a compliment, using LVA. Method and Procedure Interview Procedure. A researcher visited three different prestigious shopping malls in a major metropolitan area where she approached women carrying designer handbags to request an interview about shopping behavior in general. Owing to our expressed interest in understanding differences between loud and quiet signals, the interviewer was instructed to sample at the extremes (McClelland and Judd 1993) by identifying women carrying perceptibly “loud” and “quiet” handbags (i.e., signal strength). Ninety women were approached. Thirty-eight (42%) women agreed to participate. The participants were primarily Caucasian (75.7% Caucasian, 5.4% African American, 13.5% Hispanic, 2.7% Asian/Pacific Islander, and 2.7% other) and ranged in age from 16 to 60 (M = 36). These women carried handbags produced by Gucci, Louis Vuitton, and Hermès, among other prominent luxury brands. Respondents who completed the interview, which lasted about five minutes and was recorded with a digital voice recorder, received a $5 gift card as a token of appreciation. The researcher’s portion of the interview was entirely scripted. As in the pilot study, each interview included two phases. In the first phase, respondents were asked generic questions about shopping such as which day of the week they prefer to shop, how often they go shopping, and whether they prefer to go shopping alone or with others. The conversations about respondents’ shopping behavior masked the true intent of the study and were used for calibration purposes. In the second phase of the conversation, the researcher paid each woman the identical compliment, saying “I really like your bag. It is very nice.” The woman was 18 given time to respond, after which the researcher asked further questions including when they had obtained the handbag, what they liked about it, and their feelings about the brand. At the end of every interview, the researcher took a picture of the respondent’s handbag that was rated later by three independent judges on the 7-point brand prominence scale (Han et al. 2010). The women approached by the experimenter were chosen randomly, given they were carrying either a “loud” or “quiet” handbags (see figure 3). FIGURE 3 Study 1: Distribution of Brand Prominence Number of Participants 14 12 10 8 6 4 2 0 1 Quiet 2 3 4 5 Avg. Brand Prominence Rating 6 7 Loud As in the pilot study, we compared measurements of interviewees’ affective states before and after the compliment to assess any changes due to the compliment. Individuals differ in their general susceptibility to embarrassment (Modigliani 1968); hence, it was possible some respondents may have felt embarrassed not because of the compliment but simply because they were part of an unusual experience. Note, however, that each woman’s voice prior to the compliment served as an individualized baseline for our analyses, which takes into consideration any embarrassment resulting from being interviewed (the general context). Emotional Profile. Among the 38 respondents, four reported carrying counterfeit designer 19 handbags (we address the special case of counterfeit goods in our general discussion) and one interview was interrupted unexpectedly. Hence, five recordings were removed from the data, leaving 33 usable audio files. Using each of the four specific emotions (embarrassment, anger, happiness, and sadness) and the stress level measured by the LVA software, we calculated a series of averages for each individual both the period before (first phase) and after the compliment (second phase). We then computed change scores (post-compliment average – precompliment average) for each of the five different voice profile measures. In our analysis, we use the change in emotions and stress level as detected by the LVA software, as our dependent measure. For embarrassment, the focal DV, this change ranged from -4.76 to 5.13 with M = 0.78. Results We estimated a series of linear regressions with the change in emotion scores as the dependent variable and signal strength rating (the judges’ rating of brand prominence) as the predictor variable. Several additional measures were included as planned covariates: the brand of handbag, the number of designer handbags the respondent owned, age, and employment status. The results are shown in table 2. TABLE 2 Study 1: Effects of Signal Strength on Embarrassment Brand Employment status (LV=1, nonLV=0) The number of designer handbags Age .49** (.17) -1.44 (.94) .04 (.05) -.09* (.04) 1.52 (.85) Happy/Excited .07 (.15) 1.21 (.84) .02 (.05) -.06 (.04) -.80 (.76) Sad -.06 (.23) -.29 (1.30) <.01 (.07) .07 (.06) -.29 (1.17) Angry .02 (.12) -.11 (.66) <.01 (.04) -.03 (.03) .57 (.60) Stressed -.22 (.12) -.33 (.70) <.01 (.04) .04 (.03) .21 (.63) Signal Strength Embarrassed (Yes=1, No=0) Note: *p < .05, ** p < .01; Entries in the table correspond to parameter estimates in OLS regression. Standard errors 20 are in parentheses. Consistent with our prediction, consumers carrying louder signals exhibited greater embarrassment in response to the compliment (β = .49, p < .01); no such differences were found for any of the other emotions (i.e., happiness, sadness, anger). These findings indicate that consumers displaying loud signals experienced significantly more embarrassment in response to a compliment on that identity signal than consumers displaying quieter signals. These results support our prediction and provide initial support for our proposed conceptual model. Given the interviewer sought out respondents based on self-selected signals, it was important to consider whether the interviewer’s voice changed based on signal strength as a change in the interviewer’s voice might have influenced respondents’ affective states. To rule this out, we used a similar procedure to analyze the interviewer’s voice across conditions. We examined the interviewer’s emotional states throughout the interview with particular attention to both the moment when the interviewer delivered the compliment and when the interviewer asked questions about the respondent’s handbag. As shown in table 3, none of the interviewer’s emotional measures varied with the signal strength of respondents’ handbags. TABLE 3 Study 1: Analyses of Interviewer’s Emotional States DV: Interviewer’s emotional states IV: Interviewee’s signal strength While delivering compliment Embarrassed -.02 Happy/ Excited 0 Sad -.37 Angry 0 Stressed .02 Second part of interview (.53) .01 (.26) (0) -.02 (.02) (.32) -.52 (.32) (0) .03 (.03) (.25) -.08 (.14) Note: Entries in the table correspond to parameter estimates in OLS regression. Standard errors are in parentheses. 21 Discussion The results from study 1 demonstrate that consumers who signal more loudly are more likely to become embarrassed when complimented on their signal. Recall our prediction was that people are expected to become embarrassed with a loud signal regardless of self-beliefs. In this study, we demonstrate this effect with actual consumers displaying self-selected signals in a natural environment. By measuring respondents’ emotional reaction using LVA technology, we were able to measure embarrassment in a field setting where explicit self-report measures were likely to have conflicted with our cover story. This approach helps insure the external validity of our results and the applicability of our conceptual model to the real world. In study 2, we replicate the effects for signalers displaying loud signals while providing additional evidence for the proposed mechanism. We also document how and when a compliment on a quiet signal is likely to produce embarrassment. More specifically, in study 2 we provide additional evidence for the proposed public self-awareness mechanism by showing that compliments on a quiet signal also result in social anxiety and greater embarrassment when individuals’ self-beliefs are incongruent with an aspect of their identity signaled (“sexiness”). STUDY 2: VICTORIA’S SECRET VERY SEXY PERFUME Study 2 was designed to test if and how respondents’ self-beliefs about an important characteristic of their self-concept—sexiness—would impact the level of embarrassment experienced in response to a compliment. Unlike study 1, in this study, we randomly assigned respondents to different levels of brand-related identity signal strength. The study thus employed an identity Self-belief (measured) x 2 Signal Strength (manipulated: Loud/ Quiet) design. The 22 cover story revolved around Victoria’s Secret, the largest American retailer of lingerie and beauty products. The brand and its marketing are strongly associated with being sexy. For a sample of Victoria’s Secret branded materials used to ensure the credibility of the study, please see the web appendix. Female students at a major west coast university were recruited under the pretext of a special event on campus orchestrated by Victoria’s Secret (henceforth VS) in which participants would receive a free individual consultation concerning first impressions and the art of social influence. In addition, participants were offered a $15 VS brand gift card and goodie bag in exchange for providing VS product feedback. Upon arriving in the lab, each participant was greeted by a confederate who asked whether they were checking in for the VS consultation or an unrelated “mobile technology study.” This provided a plausible explanation for encountering a male confederate (the compliment giver) later in the study. Respondents were run individually in an elaborate four-stage session designed to reinforce the credibility of VS’s role. Fifty-three participants completed the study, and each session lasted approximately 25 minutes. In stage 1 (room 1), participants completed a three-part computerized questionnaire consistent with the cover story of getting to know them better prior to their consultation. Because women vary considerably regarding the extent to which their appearance influences selfevaluations (Breines, Crocker, and Garcia 2008), part one of the questionnaire was designed to collect individual difference measures unobtrusively for appearance self-esteem and appearance contingent self-worth. These measures were planned covariates in our analyses. Participants completed assessments for three facets of self-esteem (appearance, social, and performance) using 21-items (Heatherton and Polivy 1991) each on a 7-point scale (1=not at all, 7=extremely). Participants also completed 20 items comprising four facets of the contingencies of self-worth 23 scale (Crocker et al. 2003): academic competence, appearance, approval from others, and competition, each on a 7-point scale (1= strongly disagree, 7= strongly agree). To lend legitimacy to the false feedback provided in the next stage of the study, part two of the questionnaire measured each participant’s level of extraversion. This was achieved using items from the Ten-Item Personality Inventory (TIPI; Gosling, Rentfrow, and Swann 2003) and the Revised Eysenck Personality Questionnaire (EPQR-A; Francis, Brown, and Philipchalk 1992). Each item was measured on a 7-point scale (1=strongly disagree, 7=strongly agree). Responses to these measures were converted into what was dubbed the “Demeanor/Social Aptitude” score (see figure 4) presented to participants along with the false feedback. In part three of the questionnaire, respondents answered a series of questions related to managing first impressions and social relations with potential dating partners. These questions were gathered from non-scientific quizzes found in popular women’s magazines and were linked to the impending consultation to reinforce the premise of the study. Participants also filled in a paper and pencil “Self-Description” survey asking them to complete four fill-in-the-blank items [e.g., “I would describe myself as ____ (adjective).”]. These adjectives were used to facilitate the individual consultation conducted in stage 2. After finishing the entire questionnaire, the participant was escorted into an adjoining room (room 2) for stage 2 of the study. In room 2, a male confederate wearing a black suit and VS credentials acted the part of the VS Specialist brought in to conduct the individual consultation. The confederate explained that he was engaged by VS to research how women’s demeanor and body language affects others’ perceptions in various social situations. As part of the script, he stated that… “…recent research suggests that up to 93% of all communication is nonverbal. Nonverbal behaviors are important because they are often automatic and influence how individuals are perceived by others. I will ask you to complete some everyday tasks that will allow me to better 24 observe your non-verbal physical behaviors and what they communicate about you.” The entire consultation was conducted under the pretense of assessing the respondent’s first impressions and various abilities involving social influence. The VS Specialist began by asking respondents to complete the following three nonverbal tasks, some involving items provided in the room: 1) “Please walk across the room, and then turn around and walk back,” 2) “Please make yourself comfortable in that chair,” and 3) “Please select a cookie or brownie and eat it.” Next, the VS Specialist conducted a structured interview based on the participant’s responses to the Self-Description survey completed in Room 1. Paraphrasing a participant’s responses, he asked her to elaborate on her answers. For example, he would say “You described yourself as … enthusiastic. Give me an example of a situation when you exhibited being enthusiastic.” The verbal task was used for calibration purposes, as in study 1, to measure baseline levels of emotion. Throughout the session, the specialist made a series of notes on his clipboard. After the final task, the specialist left the room, explaining that he would return momentarily with the results. After returning, the VS Specialist explained how demeanor and body language are two major dimensions of sex appeal and that the individual’s results were based on comparisons with a representative sample of 25,482 female college students at universities across the U.S. While reviewing the results, the respondent was told that the assessment was informed by data collected from thousands of male college-aged students about what they find “sexy.” Following common practice in false-feedback paradigms (e.g., Mead et al. 2011; Stillman et al. 2009), participants first received accurate, self-relevant feedback to increase the perceived credibility and legitimacy of the false feedback. To this end, each participant received a form depicting her actual level of extraversion drawn from the measures collected in Room 1. More 25 critically for our purposes, the respondent also received manufactured scores supposedly based on her body language for affectation, confidence, mannerisms, and comportment (see figure 4). FIGURE 4 Study 2: Sample Consultation Results Provided to Participants These body language scores did not differ across participants and were intended to influence (decrease) the extent to which they viewed themselves as sexy. Each participant was told she ranked in the “27th to 32nd percentile in Sex Appeal, which is significantly below average.” Due to the elaborate nature of the study and relatively small sample size, the false feedback helped insure heterogeneity with respect to the congruency between self-beliefs and an 26 ideal regarding sexiness exemplified by VS models (displayed throughout the study). Note that because we are interested in the more realistic situation where one’s ideal exceeds one’s actual identity, we did not include positive false feedback (i.e., by telling participants that they were significantly above average in their sexiness). At the end of stage 2, the specialist explained that, due to time constraints, he was unable to go into more detail. The participant proceeded to room 3 for the VS product evaluation, considered stage 3 of the study. In room 3, a box set of the VS signature line of “Very Sexy” lotion and perfume was prominently displayed, and testers were provided for product trial (see figure 5). FIGURE 5 Study 2: Victoria’s Secret Very Sexy Perfume Gift Set Used in Stimuli Stage 3 allowed us to administer a two-part survey. The first part was designed to measure self-beliefs about being sexy. Participants were asked to indicate their agreement with ten separate “I believe I am a ______ person” statements, each on a 7-point scale (1=completely disagree, 7=completely agree). The focal self-belief rating (“sexy”) was embedded in the list, consisting of several aspirational attributes common among college students (items: powerful, intelligent, charming, patient, devoted, sexy, giving, hardworking, honest, and athletic). The second part of the survey was titled “Product Evaluation—Victoria’s Secret VERY 27 SEXY Lotion & Fragrance.” The VS associate dispensed some lotion on the participant’s left hand and spritzed the perfume on her right wrist. Participants next indicated their agreement with statements addressing the attractiveness of the packaging, ease of application, extent to which the VS brand was conveyed, and whether the lotion and perfume complemented each other, each on a 7-point scale (1=strongly disagree, 7=strongly agree). Participants also indicated how they would describe the scent by providing ratings for a series of fragrance notes (citrus, earthy, floral, fresh, herbal, spice, vanilla, warm, woodsy) each on a 7-point scale (1=not at all, 7=very much), all intended to maintain the cover story. After completing the product evaluation, the associate handed the participant a bag containing the $15 gift card, a sample of “Very Sexy” perfume, and some candy (pink Orbit gum and Hershey’s kisses). The “goodie bag” provided a seamless way to manipulate signal strength in a random manner. Half of the participants received a plain, unmarked, glossy pink bag (quiet) while the other half received the same glossy pink bag embellished with a “VERY SEXY Victoria’s Secret” brand decal (see figure 6 for photos of the stimuli). At the end of stage 3, the participant was thanked for her input and instructed to bring her things (including the “goodie bag”) with her to the next room where she would be asked to provide feedback about the entire experience. At this point, the official study ostensibly was complete. All participants exited the evaluation room wearing the same fragrance. They differed only in terms of the type of bag carried (loud vs. quiet signal) and thus the extent to which the identity signal (scent they were wearing) was observable, recognizable, and interpretable in the next room. 28 FIGURE 6 Study 2: Very Sexy Product Gift Bags (Left: Loud, Right: Quiet) As each participant entered the final room (room 4) for stage 4 of the study, she encountered an attractive male student (another confederate) completing the fictitious “Mobile Technology Study” mentioned at check-in. Careful staging ensured both the female participant and male confederate were there for clearly different purposes yet still noticed by each other. The male student sat at a table on the opposite side of the room containing a variety of mobile devices (e.g., cell phones, an iPod, and cameras), ostensibly completing a different survey. When the female participant entered the room, the male student looked up, making eye contact and smiling warmly before returning his attention to his own survey. As the associate logged the female participant into a computer, she told the participant to place her things on a side-table (engineered to make the VS bag clearly visible) and that someone would be by to sign her out shortly. The associate exited the room, leaving the participant and confederate alone. Participants saw the following instructions on the computer: “In order for us to get to know you and improve our offerings to the college student market, we would like to ask you some additional questions about your background and experience today.” On the first screen participants completed the following three items regarding their consultation: 1) I found the 29 assessment to be (1=very inaccurate, 5=very accurate), 2) I found the information to be (1=very useless, 5=very useful), and 3) I enjoyed the experience (1=not at all, 5=completely). As the female participant began to move the mouse over to click her first response, the male student paid her the following compliment: “Wow, I really like your perfume. Is that what you’re wearing?” while alluding to the pink VS bag. The confederate was instructed to allow the participant to respond without encouraging further conversation. Regardless of whether the participant simply smiled or spoke, the confederate always brought closure by saying “It’s nice” and returning his attention to his survey while allowing her to complete hers. (See web appendix for additional details.) Following the carefully timed and choreographed compliment, participants then just so happened to complete a “Thoughts and Feelings” questionnaire about the session on the second screen that captured our focal dependent variable—feelings of embarrassment. Participants completed a two-item embarrassment measure: “Right now I am feeling…” 1) not at all embarrassed/very embarrassed and 2) not at all awkward/very awkward, each on a 7-point scale (Dahlet al. 2001; Modigliani 1968; Parrott and Smith 1991). They also completed a third measure assessing the extent of their discomfort: 3) not at all uncomfortable/very uncomfortable. Participants completed measures for a variety of other emotions: happy/lively, calm/content, loving/caring, proud/self-assured, grouchy/fed up, tired/drowsy, and gloomy/sad, each on a 7point scale (1=not at all, 7=very much). Participants also rated the extent to which they were experiencing anxiety as measured by feeling jittery and nervous (α = .77), each on a 7-point scale (1=not at all, 7=very much). The composite Anxiety measure is considered a reflective indicator for the latent variable Public Self-Awareness and any mediating effect between Signal Strength and Embarrassment would be 30 seen as supporting evidence for our conceptual framework. Further, participants completed a measure of momentary self-consciousness (i.e., right now, I feel self-conscious) on a 7-point scale and the SCS Revised scale (Scheier and Carver 1985; Fenigstein, Scheier, and Buss 1975), a 23-item measure of dispositional self-consciousness; each item was measured on a 5-point scale (1=not at all like me, 5=very much like me). Finally, participants completed basic demographic questions (age, ethnicity, and academic major) and a funneled debrief. We also administered an extended debrief along with having them sign a nondisclosure statement intended to prevent them from telling others about their experience (the nature of the experiment) before leaving the building. Results Initial analyses indicated evaluations of the Session experience did not differ by condition. Participants reported finding the consultation to be equally accurate (Mloud = 3.26, Mquiet = 3.35; β = -0.09), useful (Mloud = 3.56, Mquiet = 3.58; β = -0.07), and enjoyable (Mloud = 3.52, Mquiet = 3.62; β = -0.10), all t<1 across Signal Strength conditions. Similarly, we find no differences in the levels of emotions other than embarrassment by condition (see table 4). Next, we examined our focal dependent measures: Embarrassment and Anxiety. A model predicting Embarrassment (α = .89) was estimated using participants’ SelfBelief regarding being sexy (continuous measure), Signal Strength as a dummy variable (0=quiet, 1=loud), their interaction, and appearance self-esteem as a covariate. The model revealed a significant effect for Self-Belief (β = -.58, SE = .23, t(48) = -2.51, p < .02) in the expected direction and a marginally significant Self-Belief X Signal Strength interaction (β = .69, SE = .37; t(48) = 1.89, p < .06). Appearance self-esteem was highly significant (β = -.49, SE 31 = .19; t(48) = -2.62, p < .01), and results substituting appearance contingent self-worth in lieu of appearance self-esteem were substantively the same. Within the quiet condition, simple slope analyses revealed that as one’s Self-Belief regarding sexiness increased, Embarrassment in response to the compliment decreased (β = -.62, t(48) = -2.73, p < .01). In the loud condition, one’s Self-Belief regarding sexiness had no significant effect on Embarrassment (β = -.15, t<1) in response to the compliment, as expected. TABLE 4 Study 2: Levels of Other Specific Emotions by Signal Strength Condition Emotion measure Loud condition Quiet condition Test statistic Mean (SD) Mean (SD) t(48) Happy (α = .75) 4.46 (1.12) 4.85 (1.09) t = -1.26, p>.21 Content (α = .39)* 4.98 (1.02) 5.38 (0.94) t = -1.49, p>.14 Loving (α = .81) 4.52 (1.35) 4.58 (1.30) t = -0.16, p>.87 Proud (α = .57) 4.17 (1.12) 4.58 (1.23) t = -1.27, p>.21 Angry (α = .82) 2.20 (1.32) 1.94 (1.32) t = 0.72, p>.48 Tired (α = .87) 3.09 (1.77) 3.02 (1.73) t = 0.15, p>.88 Sad (α = .86) 2.17 (1.41) 1.88 (1.21) t = 0.78, p>.44 *We re-analyzed the component items for emotion measures for which reliabilities were low (i.e., content and proud). No significant differences were found for those individual items; however, a marginal difference was found for the item “calm” (t = -1.80, p < .08). A second model with Anxiety (nervous, jittery; α = .77) as the DV was estimated using the same set of predictors. The model revealed significant main effects for Self-Belief (β = -.70, SE = .20, t(48) = -3.52, p < .0009) and Signal Strength (β = -3.51, SE = 1.38; t(48) = -2.54, p < .01), qualified by a significant interaction (β = .99, SE = .31; t(48) = 3.18, p < .003). Simple 32 slope analyses revealed that increased Self-Belief in one’s sexiness had a negative effect on Anxiety within the quiet condition (β = -.71, t(48) = -4.37, p < .0002) but a non-significant effect in the loud condition (β = .22, t<1). Robustness Checks. Given that our model implies changes in self-consciousness, as a robustness check we estimated a model predicting momentary self-consciousness using the same set of predictors. The model revealed a significant effect for sexiness Self-Belief (β = -.61, SE = .21, t(48) = -2.93, p < .005) qualified by a marginally significant sexiness Self-Belief X Signal Strength interaction (β = .62, SE = .33; t(48) = 1.89, p < .06). Simple slope analyses revealed that sexiness had a negative effect on anxiety within the quiet condition (β = -.63, t(48) = -3.37, p < .003), but a non-significant effect in the loud condition (β = -.10, t<1). In sum, the means and pattern of results are consistent with the results for Embarrassment and Anxiety, corroborating the predicted negative consequences brought about by a compliment. As an additional robustness check, we examined respondents’ physical and verbal responses to the compliment as recorded on video during the study. Previous work has shown that negative self-conscious emotions such as embarrassment are associated with more speech disfluencies and disturbances (Edelmann and Hampson 1981; Leary 1983), less eye contact (Barrett 2005; Edelmann and Hampson 1981; Modigliani 1971), and fewer Duchenne smiles (Ekman, Davidson, and Friesen 1990; see also Edelmann 1985a and 1985b). A Duchenne smile raises the corners of the mouth and the cheeks and forms crow’s feet around the eyes. Using the video recordings, verbal and physical reactions to the compliment were coded by an independent judge blind to the hypotheses with respect to four behavioral correlates of embarrassment: (1) number of words spoken, (2) speech disfluencies such as “ums” and “uhs,” (3) the amount of eye contact, and (4) Duchenne smiles. In table 5, we provide basic descriptive statistics for the 33 sample both combined and by condition. TABLE 5 Study 2: Responses to Compliment by Signal Strength Observed Behavior Variables Range Overall Loud condition Quiet condition Mean (SD) Mean (SD) Mean (SD) Test statistic Number of words spoken Min=1, Max=43 12.10 (11.44) 12.65 (11.66) 11.55 (11.49) F<1, NS Disfluencies Min=0, max=5 0.60 (.98) 0.90 (1.21) 0.30 (0.57) F = 4.02, p <.05 Smiles (no/yes) Min=0, max=1 67.5% 45% 90% Chi-sq = 7.59, Eye contact (no/yes) Min=0, max=1 57.5% p < .006 20% 95% Chi-sq = 13.74, p < .0002 Analyses of the behavioral responses revealed participants in the loud condition displayed fewer smiles (β = -.45, t = -3.38, p < .002) and used more disfluencies (β = .60, t = 2.01, p < .05) than participants in the quiet condition. None of the interactions were significant. An additional regression model was estimated on the number of words spoken using the same set of predictors as in our earlier models. This revealed significant effects for signal strength (β = 32.97, t = 2.50, p < .02) and sexiness (β = 5.08, t = 2.01, p < .02) qualified by a significant interaction (β = -7.29, t = 2.47, p < .02). Simple slope analyses revealed that sexiness had a positive effect on word count within the quiet condition (β = 5.08, t = 2.76, p < .01) but a nonsignificant effect in the loud condition (β = -2.21, t<1). Similarly, analyses on the amount of eye contact made with the confederate revealed a marginally significant effect for the Self-Belief by Signal Strength interaction (β = -.16, t = -1.76, p < .09). Simple slope analyses revealed that Self- 34 Belief had a positive effect on eye contact within the quiet condition (β = .09, t = 2.24, p < .04), and again a non-significant effect in the loud condition (β = -.07, t<1). These results are all consistent with and thus support our conceptual model. In addition to the behavioral and self-report data, we also conducted supplementary analyses using the LVA software used in study 1. Using the video recordings obtained in rooms 2 and 4, we obtained baseline and post-manipulation voice samples for each participant. Compared to study 1, participants’ responses to the compliment (in room 4) were exceedingly brief, sometimes only one or two words such as “Yes” or “Thank you.” In such cases, the software cannot produce output owing to a lack of sufficient data. This is essentially why we consider these analyses supplementary. Using the audio files, we calculated the average level of embarrassment for each individual for the periods before (baseline in room 2; nloud = 17, nquiet = 15) and after the compliment (post compliment in room 4; nloud = 16, nquiet = 15). A model predicting embarrassment was estimated using participants’ self-belief regarding sexiness, signal strength, and their interaction. LVA revealed equivalent levels of embarrassment in the loud and quiet conditions before the signal strength manipulation, or at baseline (Mloud = 1.04; Mquiet = 1.33; F(1, 28) = 1.08, p < .31). Baseline embarrassment ranged from 0 to 2.58 with M =1.18 (0.71). However, after the compliment, the voices of individuals in the loud condition registered higher levels of embarrassment than those in the quiet condition (Mloud = 4.87; Mquiet = 3.45; F(1, 27) = 3.54, p < .07); post compliment embarrassment ranged from 0 to 17.50 with M = 4.19 (4.67). We also calculated the difference or change in emotion (time 2 – time 1) for each individual and found similar results; those in the loud condition exhibited a larger increase in embarrassment (Mloud = 3.82; Mquiet = 2.12; F(1, 27) = 2.76, p < .10), albeit this difference was marginally significant likely due to sample size constraints. This 35 change in embarrassment ranged from -2.58 to 17.50 with M = 3.00 (4.80) (see table 6). Again, these voice analyses provide supplementary non-self-report evidence of embarrassment consistent with the primary self-report emotion measures. TABLE 6 Study 2: LVA Analyses Loud Mean (SD) Quiet Mean (SD) Range Embarrassment T1 1.04 (.67) 1.33 (.75) 0 to 2.58, M = 1.77 (.71) Embarrassment T2 4.87 (4.76)* 3.45 (4.61) 0 to 17.50, M = 4.19 (4.67) Change in Embarrassment 3.82 (4.85)+ 2.12 (4.75) -2.58 to 17.50, M = 3.00 (4.80) Anger T1 0.03 (.10) 0.04 (.10) 0 to 0.40, M = .03 (.10) Anger T2 0 (0) 0 (0) 0 to 0, M = 0 (0) -0.02 (.10) -0.04 (.10) -0.40 to 0, M = -.033 (.10) “Happiness/Excitement” T1 14.57 (1.21) 14.62 (1.76) 11.50 to 17.09, M = 14.59 (1.46) “Happiness/Excitement” T2 15.64 (2.89) 14.44 (3.25) 9.67 to 23.33, M = 15.60 (3.02) Change in “Happiness/Excitement” 1.09 (3.36) 0.93 (3.81) -4.45 to 7.93, M = 1.01 (3.53) “Sadness” T1 5.64 (2.16) 5.08 (3.00) 1 to 10.48, M = 5.38 (2.56) “Sadness” T2 2.54 (3.05) 2.62 (3.48) 0 to 11.33, M = 2.58 (3.21) Change in “Sadness” -3.15 (3.71)+ -2.46 (3.71) -10.36 to 3.91, M = -2.82 (3.66) Stress T1 14.96 (4.02) 11.77 (4.55) 5.85 to 23.14, M = 13.47 (4.51) Stress T2 6.99 (4.62) 7.45 (4.69) 0.67 to 17.67, M = 7.21 (4.58) Change in Stress -8.12 (5.58) Note: *p < .07, +p < .10; All other F<1. -4.33 (4.63) -18.00 to 3.64, M = -6.29 (5.42) Change in Anger Moderated Mediation. We also find evidence that the Signal Strength x Self-Belief interaction on Embarrassment was mediated by Anxiety. We tested for moderated mediation using Hayes’ (2013) PROCESS macro with 5,000 bootstrapped samples. Model 8 was specified 36 with Y = embarrassment, X = loud signal, W = sexiness self-belief, and M = anxiety. The indirect effect of the highest-order interaction (signal strength x sexiness self-belief) was significant (β = .6726, SE = .2711, 95% CI = .1863, 1.2812), providing evidence of moderated mediation. This result implies that both the direct and indirect effects are conditional on selfbelief concerning sexiness. The conditional indirect effect of signal strength on embarrassment was significant at the mean (β = .5318, SE = .2723, 95% CI = .0548, 1.1094) and one SD above the mean level of sexiness (β = 1.3358, SE = .4359, 95% CI = .5975, 2.2981). Discussion Study 2 provides compelling evidence that receiving a compliment on a loud signal produces embarrassment regardless of the congruency between the signal and one’s self-beliefs. In addition, receiving a compliment on a quiet signal produces embarrassment but only for those who believe themselves to be low in sexiness, a self-belief that is incongruent with the identity signal exhibited by all participants (VS’s Very Sexy perfume). As participants’ self-belief in their own “sexiness” increased, embarrassment decreased. In other words, congruency between the signal and self-belief moderated the effect. The embarrassment that results from a compliment is mediated by anxiety, a typical manifestation of public self-awareness. Both studies 1 and 2 demonstrate that compliments on loud identity signals produce embarrassment, and study 2 offers evidence that public selfawareness (reflected here in reported anxiety) underlies this process. However, both of these studies have their limitations. More specifically, the range of identities signaled (status and sexiness) and the base of comparison (quiet signals) were constrained owing to the complex logistics inherent in the interaction constructed for both studies. Perhaps most importantly, both 37 the field (study 1) and lab (study 2) studies involved compliments received from relative strangers, making it unclear whether the resulting embarrassment was purely in response to a compliment on the identity signal (as hypothesized) or some other aspect of the interaction. One may argue interactions with strangers are generally more disconcerting and likely to produce embarrassment in response to a compliment and therefore question the generalizability of our results. However, it is important to point out that in both studies the effects obtained depended on signal strength condition, and all conditions involved strangers. Nevertheless, to address these potential limitations, studies 3 and 4 employ a new study paradigm created to examine a broader range of self-selected aspects of one’s identity displayed in a gender-neutral product category (clothing) commonly used in identity signaling. Most importantly, in studies 3 and 4 we hold the loud identity signal constant while systematically varying both the source of the compliment and the type of remark made (control conditions), providing further evidence of the extent to which a consumers’ embarrassment can vary in response to compliments per se. STUDY 3: RECEIVING COMPLIMENTS FROM FRIENDS VS. STRANGERS Study 3 was designed to address three specific objectives. First, we vary the source of the feedback to the signaler to determine whether the same compliment received from friends (i.e., not strangers) would produce a similar emotional response. Second, we include two different control conditions, allowing us to compare the emotional response to a compliment versus a less flattering comment or an unrelated question. Third, our design allows us to identify an individualized and idealized aspect of a respondent’s identity that is both important to them and where they see a significant discrepancy between how they would ideally like to be perceived 38 and how they believe they are actually perceived. In addition, unlike our first two studies, our sample includes men as well as women, further extending the generalizability of our findings. Finally, we also provide additional process evidence by including an improved measure intended to capture the nature and source of the resulting public self-awareness as well as embarrassment. Method Participants and Design. Two hundred and twenty-eight adults (44% male) ages 19 to 77 (M = 37.16, SD = 13.31) completed a 10-minute study using Amazon’s Mechanical Turk (mTurk) and were compensated $0.50 for their time. Online sampling offers a number of advantages, such as increased demographic diversity (Gosling et al. 2004), and recent evidence supports the reliability of data obtained from mTurk samples in particular (Buhrmester, Kwang, and Gosling, 2011). Study 3 consists of a 2 (Source: friend v. stranger) x 3 (Remark: compliment v. comment v. unrelated question) between subjects design. Participants were randomly assigned to one of the six conditions. For this study and the next, to better accommodate the diversity of identities and their differential value, we developed technology that allowed individuals to personally customize t-shirts (the signal) online. Procedure. Participants were told that they would be completing a study on “Self-Identity and Social Interaction” and would be asked some questions about their identity and then asked to envision a social interaction. After logging into the study website, participants were given a list of adjectives describing commonly valued aspects of one’s identity (athletic, smart, funny, rugged, sexy, stylish, healthy, helpful, successful, rebellious, powerful, charming, and creative). Participants were told the list contained aspects of one’s identity that may or may not be of 39 particular importance to them and were asked to rate how important each aspect was to them on a 7-point scale (0=not at all important to me, 3=important to me, 6=very important to me). Next, they were asked to choose an aspect of their identity from the list that “represents how you would IDEALLY like to be seen but that is not reflective of how you ACTUALLY perceive yourself at present. That is, please select an identity where there is a significant gap between your ideal self and your actual self.” Participants selected the adjective using a dropdown menu, and the chosen aspect of their identity was used in the identity signaling scenario that followed. Signaling and Compliment Scenario. On the next screen, participants were presented with an image of a navy blue t-shirt. Based on the participant’s idiosyncratic response, the t-shirt was customized to incorporate the chosen aspect of the participant’s identity where there was a significant gap between the ideal and the self-belief. This adjective was printed across the t-shirt in large, all-capitalized white lettering constituting a loud signal. For example, if a participant chose “smart,” the t-shirt had the word “SMART” printed across the front; if a participant had chosen athletic, the t-shirt had the word “ATHLETIC” printed across the front (see figure 7). FIGURE 7 Study 3: Identity Signaling T-Shirt Stimulus Example 40 Note all participants imagined wearing an equally loud identity signal; only the Source (friend or stranger) and the type of Remark varied by condition. Directly below the image, participants were presented with a visualization scenario in which both the source (friend v. stranger) and the type of remark (compliment v. comment v. unrelated question) were manipulated. The remark always included a question to insure the participant felt engaged in a dialogue. This was the case for the compliment, the (signal unrelated) comment on the product, and the unrelated question. Participants read the following scenario: “Imagine that you are wearing the high thread count pima cotton (considered a superior blend of cotton) t-shirt pictured above while out shopping on a summer evening. While you are in one of the stores you encounter a CLOSE FRIEND that knows you well [STRANGER that does not know you well] who looks at you, smiles, and says ... “I like your shirt. Is it new?” [“Is your shirt black or blue?” / “Do you know if it is still light outside?”] Feelings Inventory. Following the scenario, participants completed a “Your Thoughts and Feelings” questionnaire containing a randomized list of focal and filler emotion items. Participants completed multi-item measures for embarrassment (α = .90: embarrassed, awkward, uncomfortable), pride (α = .83: proud, self-assured, confident), anxiety (α = .85: jittery, selfconscious, nervous), anger (α = .87: angry, grouchy, fed up), happiness (α = .84: happy, excited, lively), sadness (α = .89: sad, gloomy), love (α = .89: loving, caring), contentment (α = .88: content, calm), and tiredness (α = .76: tired, drowsy). For all items, their emotional response was described in terms of “In that moment, I would feel…” on a 7-point scale (1=not at all, 7=very much). Self-Awareness Measures. Our measure of public self-awareness is reflected by the public facet of Situational Self-Awareness scale (SSAS; Govern and Marsch 2001) that gauges 41 attentiveness to features of one’s self that are presented to others. For completeness, however, we measured all three facets of situational self-awareness (public self-awareness, private selfawareness, and surroundings awareness). Using the prompt “At that moment, I would be…,” respondents completed the following items for: (1) public self-awareness (“concerned about the way I present myself,” “self-conscious about the way I look,” “concerned about what other people think of me”), (2) private self-awareness (“conscious of my inner feelings,” “reflective about my life,” “aware of my innermost thoughts”), and (3) surroundings awareness (“keenly aware of everything in my environment,” “conscious of what is going on around me,” and “conscious of all objects around me”) each on a 7-point scale (1=strongly disagree, 7=strongly agree). Manipulation Checks and Demographics. Participants were asked to recall what the person they had encountered said to them and rate their perceptions about the remark (i.e., what was said) in terms of friendliness (1=not at all friendly, 7=very friendly), genuineness (1=not at all genuine, 7=very genuine), and valence (1=very negative, 7=very positive). As an additional check on the source manipulation, participants completed a single item measure of relational closeness (Aron, Aron, and Smollan 1992). They also completed explicit measures asking “who made the remark to you?” and “specifically, what did that person say to you?” Finally, participants provided basic demographic information including age and gender. Results Preliminary analyses indicate participants did indeed choose a valued aspect of their identity as reflected by the average importance rating (M = 5.01, SD = 1.81). Recall the scale 42 went from 0-to-6 with 3=important to me and 6=very important to me. Initial analyses also indicated that our source and remark manipulations were successful. As expected, relational closeness for friends (M = 3.38) was significantly higher than for strangers (M = 1.62; F(1, 226) = 83.94, p < .0001). In addition, compliments were viewed as significantly more friendly (M = 5.72) than both comments (M = 4.96; F(1, 223) = 9.74, p < .002) and unrelated questions (M = 4.87; F(1, 223) = 11.99, p < .0006). Compliments were also viewed as more positive (M = 5.50) than comments (M = 4.46; F(1, 223) = 23.12, p < .0001) or unrelated questions (M = 4.34; F(1, 223) = 28.09, p < .0001). Compliments were perceived as equally genuine (M = 5.24) as comments (M = 5.00; F(1, 223) = 0.83, p < .36) and marginally more genuine than unrelated questions (M = 4.76; F(1, 223) = 3.25, p < .07). A model predicting Embarrassment was estimated using Source, Remark, and the Source by Remark interaction as predictor variables. This revealed only a main effect for Remark (F(2, 222) = 5.29, p < .006); yet neither the effect of Source (F(1, 222) = 2.55, p < .11) or the Sourceby-Remark interaction (F<1) were significant. Planned contrasts indicate compliments produced significantly more embarrassment (M = 3.42) than either comments (M = 2.54; F(1, 222) = 10.50, p < .001) or unrelated questions (M = 2.89; F(1, 222) = 3.71, p < .05); comments and unrelated questions did not differ from each other (F(1, 222) = 1.72, p < .19). This supports our conceptualization in that it is compliments per se that elicit these effects, whereas identical social interactions involving other feedback (comments and unrelated questions) do not. We also estimated the same model for anxiety, pride, and happiness. Analyses for Anxiety again revealed only a main effect for Remark (F(2, 222) = 3.99, p < .02); neither the effect for Source nor the Source-by-Remark interaction were significant (Fs<1). Compliments produced significantly more Anxiety (M = 3.32) than comments (M = 2.60; F(1, 222) = 7.29, p < .007) or 43 unrelated questions (M = 2.74; F(1, 222) = 4.50, p < .03), which did not differ from each other (F<1). This is consistent with our conceptualization and the role of public self-awareness. Analyses for Pride again revealed only a main effect for Remark (F(2, 222) = 3.06, p < .05); neither the main effect for Source nor the Source-by-Remark interaction were significant (Fs<1). Compliments produced significantly more pride (M = 3.90) than either comments (M = 3.33; F(1, 222) = 5.17, p < .02) or unrelated questions (M = 3.39; F(1, 222) = 4.08, p < .04), which did not differ from each other (F<1). It is important to point out that similarly selfconscious emotions such as pride and embarrassment can and often do co-occur and that compliment recipients can feel both embarrassed and proud simultaneously. Analyses for Happiness revealed a main effect for Remark (F(2, 222) = 3.34, p < .04) such that happiness was higher in the compliment condition (M = 3.87) than the unrelated question condition (M = 3.25; F(1, 222) = 6.25, p < .01) but not higher than in the comment condition (M = 3.41; F<1). No such differences were found for any of the other emotions assessed in this study: anger, sadness, contentment, tiredness, or love (all Fs<1.12, p > .33). Self-Awareness Measures. Analyses for public self-awareness revealed only a main effect for Remark (F(2, 222) = 4.77, p < .009); neither the effect of Source nor the Remark-bySource interaction was significant (both F<1). Compliments produced significantly higher levels of public self-awareness (M = 4.76) than both comments (M = 4.10; F(1, 222) = 5.73, p < .02) and unrelated questions (M = 3.94; F(1, 222) = 8.48, p < .004), which did not differ from each other (F<1). Analyses for private self-awareness revealed an effect for source (F(1, 222) = 6.30, p < .01), such that remarks from friends produced higher levels of private self-awareness (M = 4.73) than strangers (M = 4.22); no other effects were significant (Fs<1). Additionally, no differences in surroundings awareness were found for source, remark, or their interaction (all 44 Fs<1.27). Mediation. We tested for mediation using Hayes’ (2013) PROCESS macro Model 4 with 5,000 bootstrapped samples. To test mediation involving a three-group independent variable (Hayes 2013, p. 196), we constructed two dummy variables, X1 and X2, representing the comment and unrelated question conditions respectively. Because there were three groups, there are two indirect effects: 1) the indirect effect of comment vs. compliment on embarrassment through public self-awareness and 2) the indirect effect of unrelated question vs. compliment on embarrassment through public self-awareness. As outlined by Hayes (2013), running PROCESS twice, once with X1 as the IV and X2 as a covariate and once with X2 as the IV and X1 as the covariate, allows one to recover each indirect effect. The indirect effect of comment vs. compliment was β = -.4039 (SE = .1658) with a 95% bias-corrected bootstrapped confidence interval that excluded 0 (95% CI [-.7284, -.0881]). The indirect effect of unrelated question vs. compliment was β = -.5003 (SE = .1748) with a 95% bias-corrected bootstrapped confidence interval that excluded 0 (95% CI [-.8592, -.1704]), supporting our hypothesis. The positive indirect effect of a compliment relative to both a comment and an unrelated question provides evidence that the mediational path predicting embarrassment from a compliment is explained by increasing one’s public self-awareness. Discussion Study 3 provides additional evidence supporting the assertion that compliments on loud identity signals result in embarrassment above that of merely interacting with someone unknown (i.e., a stranger) and above that of drawing attention to a non-identity relevant aspect of the product (i.e., being asked about a characteristic—in this case, color—of the shirt). Feelings of 45 embarrassment increased in response to a compliment regardless of whether it came from a stranger or a close friend. Notably, neither the main effect for Source nor the Remark-by-Source interaction was significant for any of the dependent measures. Study 3 also offers further evidence of public self-awareness as the underlying mechanism illustrating how it is critical to explaining the emotional response—embarrassment—that results in response to the compliment. Worth reiterating here is that participants who received compliments reported experiencing both greater embarrassment and pride (i.e., a mixture of negative and positive selfconscious emotions). This result is consistent with findings from study 2 whereby loud signalers who received a compliment also reported experiencing elevated levels of both embarrassment and pride. These results are also consistent with the notion that individuals’ emotional reactions to compliments are neither universally positive nor universally negative but rather are emotionally rich, complex, and perhaps mixed as in these studies. STUDY 4: COMPLIMENTS ON IDENTITY SIGNALS VS. OTHER ASPECTS OF SELF Study 3 demonstrated how a compliment on a loud identity signal causes more embarrassment than either a comment or unrelated question. Building on this finding, study 4 tests an important boundary condition. Specifically, study 4 was designed to determine: 1) whether receiving a more explicit identity-related compliment (i.e., explicitly calling out the identity signal—product or brand—per se) would cause greater embarrassment and 2) whether receiving any appearance-related compliment (i.e., about the person rather than the product) could result in similar levels of embarrassment. Method 46 Participants, Design, and Procedure. One hundred and fifty three adults (51.3% male) ages 18 to 65 (M = 35.22, SD = 11.79) completed a 10-minute study online. Study 4 was a single-factor, between-subjects design employing four different types of compliments (person/ product/ identity-explicit product/ control). The set-up and procedure were identical to that of study 3 with the exception of the wording of the compliment. Participants completed the same identity ratings and identity selection task but were then randomly assigned to one of four compliment conditions. In the person condition, the compliment was made about an aspect of the person (i.e., a new haircut). In the product condition, the compliment was made about the shirt (“your shirt”). In the identity-explicit condition, the compliment explicitly referred to the selfselected aspect of that person’s identity, collected the same way as in study 3 (e.g., “your smart shirt”). In the control condition, the remark made was about the light outside, just as in study 3. Identity Signal and Compliment Scenario. In conjunction with an image of the t-shirt displayed, participants read the following scenario: “Imagine that you recently had your hair cut and are wearing the high thread count pima cotton (considered a superior blend of cotton) t-shirt pictured above while out shopping on a summer evening. While you are in one of the stores you encounter a stranger that does not know you who looks at you, smiles, and says... “I really like your haircut. Is it new?” [“I really like your shirt. Is it new?” / “I really like your [identity piped] shirt. Is it new?”/ “Do you know if it is still light outside?”] Participants then completed the same feelings inventory, self-awareness measures, manipulation checks, and demographics measures as in study 3. Results Preliminary analyses indicated that participants again chose valued identities as reflected 47 by importance rating of the identity chosen (M = 5.43, SD = 1.62). Initial analyses indicated that the complimentary remark manipulation was successful. All remarks were viewed as friendly (all M > 5.21), genuine (all M > 5.03), and positive (all M > 4.36) such that all conditions were well above scale midpoints (see table 7 for details). TABLE 7 Study 4: Manipulation Checks on Compliment Received Measure Person Product Unrelated Question Mean Test statistic Mean Identified product Mean Mean Friendliness 6.34b 5.68a,b 5.90a,b 5.21a F=5.16, p < .002 Genuineness 6.03b 5.41a 5.23a 5.03a F=3.39, p < .02 Valence 6.32c 5.57b 5.59b 4.36a F=14.35, p < .0001 F(3,149) Note: Means within the same row with different superscripts (a, b, c) differ at p < .05. Notably, no significant differences were found between the product and identity-explicit product conditions for any manipulation checks, emotion measures, or self-awareness measures; thus, we collapsed across the product and identified product conditions for parsimony in reporting the results from the remaining analyses. We note that the results are substantively the same when each condition is compared separately. A model predicting Embarrassment (α = .92) was estimated using the remaining three compliment types as conditions. Independent contrasts indicate that product compliments produce marginally more embarrassment (M = 3.37) than person compliments (M = 2.82; F(1, 150) = 2.57, p < .10) and significantly more embarrassment than unrelated questions (M = 2.33; F(1, 150) = 9.11, p < .003). We estimated the same model for anxiety (α = .88), pride (α = .86) and happiness (α = .80). Contrasts indicate that product compliments also produced marginally 48 more Anxiety (M = 3.28) than person compliments (M = 2.71; F(1, 150) = 2.96, p < .08) and significantly more anxiety than unrelated questions (M = 2.51; F(1, 150) = 5.45, p < .02). Somewhat interestingly, product compliments produced significantly less Pride (M = 4.08) than person compliments (M = 5.17; F(1, 150) = 12.53, p < .01) but the same amount of pride as unrelated questions (M = 3.97; F<1). Finally, product compliments produced significantly less Happiness (M = 3.92) than person compliments (M = 4.83; F(1, 150) = 10.86, p < .001), but the same amount of happiness as unrelated questions (M = 3.88; F<1). Self-Awareness Measures. Analyses for public self-awareness revealed a marginally significant effect for the type of compliment (F(2, 150) = 2.83, p < .06). Both compliments on products (M = 4.70) and the person produced similar levels of public self-awareness (M = 4.32; F(1, 150) = 1.36, p < .24); however, product compliments produced significantly more public self-awareness than unrelated questions (M = 3.93; F(1, 150) = 5.50, p < .02). Analyses for private self-awareness revealed no significant differences (F(2, 150) = 1.16, p < .32). Finally, analyses for surroundings awareness revealed a significant effect for remark (F(2, 150) = 3.14, p < .05) whereby product compliments produced marginally less surroundings awareness (M = 4.48) than person compliments (M = 4.95; F(1, 150) = 3.17, p < .08) and unrelated questions (M = 5.07; F(1, 150) = 5.12, p < .03). Mediation. We tested for mediation using Hayes’ (2013) PROCESS macro Model 4 with 5,000 bootstrapped samples. To test mediation of a three-group independent variable (Hayes 2013, p. 196), we constructed two dummy variables, X1 and X2, representing the unrelated question and person compliment conditions respectively. Because we examined three groups, 49 there are two indirect effects: 1) the indirect effect of unrelated question vs. product compliment on embarrassment through public self-awareness and 2) the indirect effect of person compliment vs. product compliment on embarrassment through public self-awareness. The indirect effect of unrelated question vs. product compliment was β = -.2960 (SE = .1404) with a 95% biascorrected bootstrapped confidence interval that excluded 0 (95% CI [-.6360, -.0731]). The indirect effect of person compliment vs. product compliment was β = -.1483 (SE = .1375) with a 95% bias-corrected bootstrapped confidence interval that included 0 (95% CI [-.4562, .0917]). These findings provide some additional evidence that the mediational path predicting embarrassment from a product-related compliment is explained by public self-awareness. Discussion Study 4 provides converging evidence that receiving a compliment on an identitysignaling product (i.e., a product-related compliment) produces greater embarrassment. This embarrassment ensues even when the compliments are perceived as friendly, genuine, and positive. Study 4 also suggests that product-related compliments—for products linked to signaling aspects of one’s identity—produce a similar emotional response (i.e., roughly equivalent levels of embarrassment, anxiety, pride, and happiness). The resulting embarrassment and anxiety generated by product compliments tended to be higher (although not necessarily achieving traditional levels of statistical significance) than compliments related to one’s personal appearance (person compliments) such as a compliment on one’s haircut. Study 4 also provides some additional evidence supporting the notion that compliments produce embarrassment through heightened public self-awareness. 50 GENERAL DISCUSSION This research examines what occurs when a signaler receives feedback on his or her identity signal, exploring the dynamic process of signaling with real brands and real behavior within dyadic interactions. We document a distinct and arguably unforeseen consequence of identity signaling—consumer embarrassment. We investigate the role of signal strength and show how consumers who display loud signals, implying they are seeking attention for their signal, become more embarrassed when others pay them a compliment on the signal. We also show how the signaler’s self-belief, particularly any incongruence between an ideal and the signaler’s self-belief, influences his or her emotional response to a compliment. The results consistently show that signalers are more likely to feel embarrassed when: (1) their signal is loud (versus quiet) and (2) when a signal is quiet but the signaler’s self-belief is incongruent with the signal. While previous research has focused on consumer behavior involving either the signal being sent or the impressions that signal made on recipients, this research differs from past work by viewing identity signaling as a two-way interactive process and investigating how the nature of feedback from the signal recipient impacts the signaler’s emotional response. More specifically, this work focuses on how the signaler responds when a recipient notices the signal and positively acknowledges it with a compliment. We show that the signaler’s emotional reaction is explained by heightened public self-awareness (i.e., concern about how he or she might appear to the compliment giver and how to reply). Such concerns about others’ evaluation depend on the signaler’s signal choice regarding the strength of the signal as well as the individual’s perceptions regarding the congruency of the signal with his or her self-beliefs. Furthermore, by documenting embarrassment as a consequence of impression management efforts—and more specifically identity signaling—we link the growing body of 51 work on identity signaling with the literature on self-conscious emotions. While previous research on identity signaling has focused on cognitive aspects of the phenomenon, our research begins to explore an important emotional consequence of identity signaling. People often expect and desire others’ attention for their signals, particularly when the signals express their desired identities. Previous research has focused on undesired social attention as the causes of blushing and embarrassment (Leary et al. 1992). However, the current research demonstrates how people also can feel embarrassed when they receive the attention they seemingly desire. These findings are thought-provoking in light of the prevalence of conspicuously branded goods in the marketplace that are clearly intended to facilitate consumer signaling. Evidence of larger brand marks on products ranging from consumer packaged goods (e.g., Coke) to handbags and automobiles (e.g., Gucci and Mercedes) are indicative of a more general trend toward louder signal offerings. Over time, both the size of these logos and emblems as well as the amount of brand ornamentation on products appears to have increased markedly. Further, blatantly conspicuous consumption (usage of loud signals) does not appear to diminish even during a recession, when others (potential signal recipients) are suffering financially and might be expected to become resentful (Nunes, Drèze, and Han 2011). Perhaps somewhat ironically, increased usage of conspicuously branded products makes it more likely that these loud signals will be noticed and thus more likely that consumers will become embarrassed. One might ask why people purchase and display conspicuously branded products in the first place if many are going to feel embarrassed. Previous research has shown that people use products and brands for symbolic self-presentation (Escalas and Bettman 2003, 2005; Kleine, Kleine, and Kernan 1993; Belk 1988) and to associate or dissociate themselves from others (Han et al. 2010). First, people may expect others to see the signal but not mention it. Second, it is 52 possible that people could expect to feel embarrassed when others acknowledge the signal, but even so, they may still choose loud signals over quiet signals when they think the benefits of sporting conspicuously branded goods outweigh the psychosocial costs of experiencing embarrassment. This is not unheard of; Lau-Gesk and Drolet (2008) show that consumers may be willing to suffer the embarrassment that comes from purchasing certain products when the products can prevent other negative experiences. In this research, we offer four studies that provide compelling evidence of the effect of signal strength on consumer embarrassment. Study 1 demonstrates the effect of loud vs. quiet signals in a natural field study involving self-chosen signals. Study 2 identifies public selfawareness (as reflected by anxiety) as a key mechanism explaining the relationship between identity signals and embarrassment. Study 3 provides additional evidence for the public selfawareness mechanism as well as further distinguishes the effect of compliments from other types of feedback (i.e., product related remark, unrelated question). Study 4 further demonstrates the robustness of the effect by comparing more vs. less explicit identity signal-related compliments and distinguishes the effect of identity signal-related compliments from other types of compliments unrelated to an identity signal (e.g., one’s haircut). The main effect of signal strength found in study 1 could be due to the fact that consumers with discrepancies between their ideal and actual self tend to prefer loud signals, perhaps in an effort to compensate for their insecurities about those identities. As a consequence, the distribution of consumers would be skewed toward consumers whose signal is not consistent with their self-beliefs and who use loud signals on one hand versus consumers with consistent self-beliefs and quiet signals on the other hand. The results from an additional pilot study (not reported here) indicate people who have larger differences between their actual and ideal selves 53 are more likely to favor louder products in terms of brand prominence. Further, the proposed explanation for why we observe such a strong main effect in study 1 is consistent with work by Braun and Wicklund (1989) who find that people who feel inadequate in a self-identified domain tend to acquire materialistic displays of their identity. For example, beginners in tennis were more likely to wear branded clothing than were expert players who were presumably more selfconfident. It is likely that many consumers who do not believe they possess a particular characteristic prefer to use loudly-branded goods to bolster themselves or perhaps simply pose as something they are not. To more fully understand how self-beliefs and signal strength interact, we investigated these factors as orthogonal factors in study 2. Limitations and Future Research This research has limitations that highlight opportunities for future research. As mentioned earlier, as a follow-up to study 1 we ran an additional version of the field study in which we relied on the same interview procedure except for one key difference. The researcher carried either a very loud or very quiet Louis Vuitton purse in terms of signal strength, or a “no name” brand handbag, constituting Signal Similarity (see figure 1). We observed that a compliment from someone who signals in the same way (i.e., loudly) leads to less embarrassment as compared to a compliment from someone who signals quietly or not at all. This result is consistent with prior work suggesting people expect more favorable evaluations from similar others (Byrne and Clore 1967). Given that embarrassment is caused by an acute concern for self-presentation, the intimacy or relational closeness between the signaler and the compliment giver would be expected to affect the signalers’ emotional response (Keltner and Buswell 1996; MacDonald and Davies 1983). And yet we find no such difference in reported 54 embarrassment in response to compliments received from close friends (presumably more similar) versus strangers (study 3). Future work examining in more detail which attributes (e.g., familiarity, social status) of a signal recipient who provides feedback impact the signaler’s concerns about others’ evaluations is warranted. Our research documents the mediating role of public self-awareness in inducing embarrassment. Additional work could help us understand what exactly loud signalers are concerned about with regard to others’ evaluations. Signalers may be concerned recipients are perceiving them, due to the strength of their signal, as signaling something about their identity that is not accurate or presenting themselves as something that they are not. But they may also be concerned about being perceived as trying too hard to impress others through consumption. At present, we are agnostic about what exact inferences are being made by the signaler, but we believe this latter explanation for embarrassment warrants further study. The symbolic use of brands is robust across cultures (Aaker 1997). However, the nature of the symbolic use and responses to others’ acknowledgment might differ across cultures. Aaker and Schmit (2001) show individuals with an independent self tend to self-express by demonstrating their points of differentiation while those with an interdependent self tend to selfexpress by demonstrating points of similarity. Further, Singelis and Sharkey (1995) find that strength of independent self-construal is negatively correlated with embarrassability. The interaction between the differences in motives of self-expressive behavior and embarrassability across cultures offers additional opportunities for future research. In this research, we focused on compliments and, as points of comparison, generally positive or natural verbal acknowledgments. However, there are other forms of feedback, such as an extended gaze or overt criticism, that we do not study. In the present research, we identified 55 the conditions whereby a compliment leads to concerns about social evaluation. A gaze is less likely to lead to further interaction with the signal recipient, and concerns about others’ evaluation might be less intense with others simply staring at their signal. This may very well be what loud signalers desire. Future work could examine how signalers respond to different forms of acknowledgement that are less interactive and perhaps less emotionally intense. In study 1, we eliminated respondents who were carrying a counterfeit handbag. However, researchers could also examine how signalers with counterfeit goods would respond to others’ acknowledgment. Wilcox, Kim, and Sen (2009) show that consumers’ moral beliefs about counterfeit consumption affect counterfeit brand preferences, particularly when their luxury brand attitudes serve a value-expressive function. Additionally, Gino, Norton, and Ariely (2010) show that wearing counterfeit goods causes people to be something that they are not and behave unethically. Therefore, consumers may experience more negative cognitive and emotional consequences of such behavior when using counterfeit brand signals. However, further research is needed to fully understand the emotional consequences of using counterfeit goods as an identity signal and how these signalers would respond to acknowledgments. Managerial Implications What consumers experience while using branded products as a signal is not only of interest from a theoretical perspective but also from the practitioner standpoint. As consumers, we constantly use products and brands hoping to achieve the image we desire through signaling. Managers should remember that people can be very sensitive to others’ reactions when they use conspicuously branded products to express their desired self-image. This can apply to a retail store context that involves interactions between consumers and salespeople. Consumers who use 56 loud items to express their desired identities may be more comfortable when salespeople are wearing similarly loud and conspicuously branded items. This may be particularly relevant within the context of luxury goods stores, especially when one considers some consumers feel intimidated in luxury stores due to the notion that such stores and salespeople have a cold and judgmental disposition (Okonkwo 2007; Ward and Dahl 2015). If salespeople at the store are also wearing loud items, this could suggest the signal is entirely appropriate and may reduce consumers’ concerns about how they are being perceived with their choice of branded product. If consumers feel embarrassed when using brand signals, it might damage their attachment to a brand and make them wonder whether the purchase was a good choice. Consumers might be able to overcome the embarrassment themselves, but marketers may be able to help, too. For example, they could run ad campaigns to help reduce concerns about others’ judgment (e.g., advising them to expect compliments and directing them in how to respond). Brand prominence, or how prominently a brand is displayed on a product, has been left to the designers as opposed to the marketers who are actually managing the brand. However, the findings of this research suggest that brand prominence as a form of signal strength should be used strategically and actively managed by marketers and should be considered when making advertising and other communication decisions. In doing so, marketers are likely to be more effective in managing consumers’ emotional responses related to their brands. 57 REFERENCES Aaker, Jennifer L. (1997), “Dimensions of Brand Personality,” Journal of Marketing Research, 34 (3), 347-56. Aaker, Jennifer L. and Bernd Schmitt (2001), “Culture-dependent Assimilation and Differentiation of the Self: Preferences for Consumption Symbols in the United States and China,” Journal of Cross-Cultural Psychology, 32 (5), 561-76. Adler, J. Michael (2009), “Detecting Deceptive Responses in Sex Offenders: A Comparison of Layered Voice Analysis (LVA) and the Polygraph,” working paper, Counseling and Consultation Services, Inc. Limestone, TN. Aron, Arthur., Elaine N. Aron, and Danny Smollan (1992), “Inclusion of Other in the Self Scale and the Structure of Interpersonal Closeness,” Journal of Personality and Social Psychology, 63, 596-612. Barrett, Karen Caplovitz (2005), “The Origins of Social Emotions and Self-regulation in Toddlerhood: New Evidence,” Cognition and Emotion, 19(7), 953-979. Belk, Russell (1988) “Possessions and the Extended Self,” Journal of Consumer Research, 15 (2), 139-68. Belk, Russell, Kenneth D. Bahn, and Robert N. Mayer (1982), “Developmental Recognition of Consumption Symbolism,” Journal of Consumer Research, 9 (June), 4-17. Bellezza, Silvia, Francesca Gino, and Anat Keinan (2014), “The Red Sneakers Effect: Inferring Status and Competence from Signals of Nonconformity,” Journal of Consumer Research, 41(1), June, 35–54. Berger, Jonah (2008), “Who Drives Divergence? Identity-Signaling, Outgroup Dissimilarity, and the Abandonment of Cultural Tastes,” Journal of Personality and Social Psychology, 95(3), 593-607. Braun, Ottmar L. and Robert A. Wicklund (1989), “Psychological Antecedents of Conspicuous Consumption,” Journal of Economic Psychology, 10 (2), 161-87. Breines, Juliana G., Jennifer Crocker and Julie A. Garcia, (2008), Self-objectification and Wellbeing in Women’s Daily Lives,” Personality and Social Psychology Bulletin, 34, 583598. Brown, Penelope and Stephen C. Levinson (1978), Universals in Language Usage: Politeness Phenomena. Question and Politeness, ed. by Esther N. Goody. Cambridge: Cambridge University Press. Buhrmester, Michael D., Tracy Kwang, and Samuel D. Gosling (2011) “Amazon’s Mechanical Turk: A New Source of Inexpensive, Yet High-Quality, Data?” Perspectives on 58 Psychological Science, 6(1), 3-5. Burroughs, W. Jeffrey, David R. Drews, and William K. Hallman (1991), “Predicting Personality from Personal Possessions: A Self-Presentational Analysis,” Journal of Social Behavior and Personality, 6 (6), 147-63. Busenitz, Lowell W., James O. Fiet, and Douglas D. Moesel, (2005), “Signaling in Venture Capitalist – New Venture Team Funding Decisions: Does It Indicate Long-Term Venture Outcomes? Entrepreneurship Theory and Practice, 29(1), 249-265. Buss, Arnold H. (1980), Self-consciousness and Social Anxiety, San Francisco: Freeman. Byrne, Donn and Gerald L. Clore (1967), “Effectance Arousal and Attraction,” Journal of Personality and Social Psychology, 6 (4), 1-18. Connelly, Brian L., S. Trevis Certo, R. Duane Ireland, and Christopher R. Reutzel (2011), “Signaling Theory: A Review and Assessment,” Journal of Management, 37(1), 39-67. Crocker, Jennifer, Ria K. Luhtanen, M. Lynn Cooper and Alexandra Bouvrette, A. (2003), “Contingencies of Self-worth in College Students: Theory and Measurement,” Journal of Personality and Social Psychology, 85, 894-908. Cupach, William R. (1992), “The Effects of Type Of Predicament and Embarrassability on Remedial Responses to Embarrassing Situations,” Communication Quarterly, 40, 149161. ——— (1994). Facework. Thousand Oaks, CA: Sage. Dahl, Darren W., Rajesh V. Manchanda, and Jennifer J. Argo (2001), “Embarrassment in Consumer Purchase: The Roles of Social Presence and Purchase Familiarity,” Journal of Consumer Research, 28 (3), 473–81. Damphousse, Kelly R., Laura Pointon, Deidra Upchurch, and Rebecca K. Moore (2007), “Assessing the Validity of Voice Stress Analysis Tools in a Jail Setting,” Final Report to the United States Department of Justice. Dubois, David, Derek D. Rucker, and Adam D. Galinsky (2012), “Super Size Me: Product Size as a Signal of Status,” Journal of Consumer Research, 38(6), 1047-1062. Durcikova, Alexandra, and Peter Gray (2009), “How Knowledge Validation Affects Knowledge Contribution,” Journal of Management Information Systems, 25 (4), 81-107. Edelmann, Robert J. (1985a), “Individual Differences in Embarrassment: Self-consciousness, Self-monitoring and Embarrassability,” Personality and Individual Differences, 6 (2), 223-30. 59 ——— (1985b), “Social Embarrassment: An Analysis of the Process,” Journal of Social and Personal Relationships, 2(2), 195-213. Edelmann, Robert J., and Sarah E. Hampson (1981), “Embarrassment in Dyadic Interaction,” Social Behavior and Personality, 9(2), 171-177. Ekman, Paul, R.J. Davidson, and W.V. Friesen (1990), “The Duchenne Smile: Emotional Expression and Brain Physiology II, Journal of Personality and Social Psychology, 58(2), 342-353. Escalas, Jennifer E. and James R. Bettman (2003), “You Are What They Eat: The Influence of Reference Groups on Consumers’ Connections to Brands,” Journal of Consumer Psychology, 13(3), 339-48. ——— (2004), “Narrative Processing: Building Consumer Connections to Brands,” Journal of Consumer Psychology, 14(1 & 2), 168-179. ——— (2005), Self-construal, Reference Groups, and Brand Meaning. Journal of Consumer Research 32, (3), 378-389. Fenigstein, Allan, Michael F. Scheier and Arnold H. Buss (1975), “Public and Private SelfConsciousness: Assessment and Theory,” Journal of Consulting and Clinical Psychology, 43(4), 522-527. Ferraro, Rosellina, Amna Kirmani and Ted Matherly (2013) “Look at Me! Look at Me! Conspicuous Brand Usage, Self-Brand Connection, and Dilution,” Journal of Marketing Research, 50 (4), 477-488. Fisher, Robert J. (1993), “Social Desirability Bias and the Validity of Indirect Questioning,” Journal of Consumer Research, 20 (2), 303-15. Fournier, Susan (1998), “Consumers and Their Brands: Developing Relationship Theory in Consumer Research,” Journal of Consumer Research, 24, 343-373. Francis, Leslie J., Laurence B. Brown and Ronald P. Philipchalk (1992), “The Development of an Abbreviated Form of the Revised Eysenck Personality Questionnaire (EPQR-A): Its Use Among Students in England, Canada, the USA and Australia,” Personality and Individual Differences, 13, 443-449 Gino, Francesca, Michael I. Norton, and Dan Ariely (2010), “The Counterfeit Self: The Deceptive Costs of Faking It,” Psychological Science, 21(5), 712–20. Goffman, Erving (1955), “On face-work: An analysis of ritual elements in social interaction,” Psychiatry: Journal for the Study of Interpersonal Processes 18, : 213-231 ——— (1959), The Presentation of Self in Everyday Life, New York: Doubleday. 60 Gosling, Samuel D., Peter J. Rentfrow, and William B. Swann Jr. (2003), “A Very Brief Measure of the Big-Five personality Domains,” Journal of Research in Personality, 37, 504–528. ———, Simine Vazire, Sanjay Srivasta and Oliver P. John (2004), “Should We Trust WebBased Studies? A Comparative Analysis of Six Preconceptions About Internet Questionnaires,” American Psychologist, 59(2), 93–104. Govern, John M. and Lisa A. Marsch (2001), “Development and Validation of the Situational Self-Awareness Scale,” Consciousness and Cognition, 10, 366-378. Han, Young Jee, Joseph C. Nunes, and Xavier Drèze (2010), “Signaling Status with Luxury Goods: The Role of Brand Prominence,” Journal of Marketing, 74(4), 15-30. Hattie, John (1992). Self-concept. Hillsdale, NJ: Lawrence Erlbaum Associates, Inc. Hayes, Andrew F. (2013), Introduction to Mediation, Moderation, and Conditional Process Analysis: A Regression-Based Approach. New York, NY: The Guilford Press Heatherton, Todd F. and Janet Polivy (1991), “Development and Validation of a Scale for Measuring State Self-esteem,” Journal of Personality and Social Psychology, 60, 895– 910. Holmes, Janet (1986), “Compliments and Compliment Responses in New Zealand English,” Anthropological Linguistics, 28 (4), 485-508. Jones, Edward E. and Thane S. Pittman (1982), “Toward a General Theory of Strategic Selfpresentation,” in J. Suls (ed.), Psychological Perspectives on the Self, 231-261, Hillsdale, NJ: Lawrence Erlbaum. Keltner, Dacher and Brenda N. Buswell (1996), “Evidence for the Distinctness of Embarrassment, Shame, and Guilt: A Study of Recalled Antecedents and Facial Expressions of Emotion,” Cognition and Emotion, 10(2), 155-71. ——— (1997), “Embarrassment: Its Distinct Form and Appeasement Functions,” Psychological Bulletin, 122, 250-270.k Kleine, Susan Schultz, Robert E. Kleine III, and Chris T. Allen. 1995. “How is a possession ‘me’ or ‘not me’? Characterizing types and an antecedent of material possession Attachment,” Journal of Consumer Research 22, (3): 327-343, Kleine, Robert E., III, Susan Schultz Kleine, and Jerome B. Kernan (1993), “Mundane Consumption and the Self: A Social-Identity Perspective,” Journal of Consumer Psychology, 2 (3), 209-35. 61 Krishna, Aradhna, Kelly B. Herd, and Nilufer Z. Aydinoglu (2015), “Wetting the Bed at Twentyone: Embarrassment as a Private Emotion,” forthcoming, Journal of Consumer Psychology. Lau-Gesk, Loraine and Aimee Drolet (2008), “The Publicly Self-Consciousness Consumer: Prepared to Be Embarrassed,” Journal of Consumer Psychology, 18 (2), 127-136. Leary, Mark R. (1983), “Social Anxiousness: The Construct and Its Measurement,” Journal of Personality Assessment, 47(1), 66-75. Leary, Mark R., Thomas W. Britt, William D. Cutlip II, and Janice L. Templeton (1992), “Social Blushing, “ Psychological Bulletin, 112 (3), 446-60. MacDonald, Linda M. and Martin. F. Davies (1983). “Effects of Being Observed by a Friend or Stranger on Felt Embarrassment and Attributions of Embarrassment,” Journal of Psychology, 113 (2), 171–74. Mayew, William J. and Mohan Venkatachalam (2011), “The Power of Voice: Managerial Affective States and Future Firm Performance,” Journal of Finance, forthcoming. McClelland, Gary and Charles M. Judd (1993), “Statistical Difficulties of Detecting Interactions and Moderator Effects,” Psychological Bulletin, 114, 376–90. McCracken, Grant (1989), “Who Is the Celebrity Endorser? Cultural Foundations of the Endorsement Process,” Journal of Consumer Research, 16 (3), 310-321. Mead, Nicole L., Roy F. Baumeister, Tyler F. Stillman, Catherine D. Rawn, and Kathleen D. Vohs (2011), “Social Exclusion Causes People to Spend and Consume Strategically in the Service of Affiliation,” Journal of Consumer Research, 37 (February), 902–19. Miller, Rowland S. (1995), “On the Nature of Embarrassability: Shyness, Social-evaluation, and Social Skill,” Journal of Personality, 63 (2), 315–39. Mirror.co.uk (2010) “Jennifer Metcalfe: ‘Don't call me sexy – it’s embarrassing,’” Celebrity News, August 1, accessed on August 13, 2014 at http://www.mirror.co.uk/3am/celebritynews/jennifer-metcalfe-dont-call-me-sexy-238057 Modigliani, Andre (1968), “Embarrassment and Embarrassability,” Sociometry, 31 (3), 313-26. ——— (1971), “Embarrassment, Facework, and Eye Contact: Testing a Theory of Embarrassment,” Journal of Personality and Social Psychology, 17 (1), 15-24. Nunes, Joseph C., Xavier Drèze, and Yan Jee Han (2011), “Conspicuous Consumption in a Recession: Toning it Down or Turning it Up,” Journal of Consumer Psychology, 21(2), 199-206. 62 Okonkwo, Uché (2007), Luxury Fashion Branding, Hampshire: Palgrave Macmillan. Parrott, W. Gerrod and Stefanie F. Smith (1991), “Embarrassment: Actual vs. Typical Cases, Classical vs. Prototypical Representations,” Cognition and Emotion, 5 (5&6), 467-88. Pomerantz, Anita. (1978), “Compliment Responses. Notes on the Cooperation of Multiple Constraints. In Jim Schenkein (Ed.), Studies in the Organisation of Conversational Interaction (pp. 79-112). New York, San Francisco, London: Academic Press. Richins, Marsha L. (1994a), “Valuing Things: The Public and Private Meanings of Possessions,” Journal of Consumer Research, 21(3), 504-21. ——— (1994b), “Special Possessions and the Expression of Material Values,” Journal of Consumer Research, 21(3), 522-33. Rosenfeld, Paul, Robert A. Giacalone, and Catherine A. Riordan (1995), Impression Management in Organizations: Theory, Measurement, and Practice, New York: Routledge. Sabini, John, Michael Siepmann, Julia Stein and M Meyerowitz, M. (2000), “Who is Embarrassed by What?” Cognition and Emotion, 14, 213–240. Scheier, Michael F. and Charles S. Carver (1985), “Optimism, Coping, and Health: Assessment and Implications of Generalized Outcome Expectancies,” Health Psychology, 4(3), 219247. Schlenker, Barry R. (1980), Impression Management: The Self-concept, Social Identity and Inter-personal Relations, Monterey, CA: Brooks/Cole. Schlenker, Barry R. and Mark R. Leary (1982), “Social Anxiety and Self-Presentation: A Conceptualization and Model,” Psychological Bulletin, 92 (3), 641-69. Shalev, Edith and Vicki G. Morwitz (2012), "Influence via Comparison-Driven Self Evaluation and Restoration: The Case of the Low-Status Influencer," Journal of Consumer Research, 38 (5), 964-980. Singelis, Theodore M., and William F. Sharkey (1995), “Culture, Self-construal, and Embarrassability,” Journal of Cross-Cultural Psychology, 26, 622-644. Sirgy, M.J. (1982), “Self-concept in Consumer Behaviour: A Critical Review,” Journal of Consumer Research, 9(3), 287-300. Stillman, Tyler F., Roy Baumeister, Nathaniel Lambert, Will Crescioni, Nathan DeWall, and Frank Fincham (2009), “Alone and Without Purpose: Life Loses Meaning Following Social Exclusion,” Journal of Experimental Social Psychology, 45 (4), 686–94 Tangney, June P., Rowland S. Miller, Laura Flicker, and Deborah H. Barlow (1996), “Are 63 Shame, Guilt and Embarrassment Distinct Emotions?” Journal of Personality and Social Psychology, 70 (6), 1256-269. Tashakkori, Abbas and Charles Teddlie (2003). Handbook of mixed methods in social and behavioral research (eds.) Thousand Oaks, CA: Sage. Tracy, Jessica L., and Richard W. Robins (2004), “Putting the Self into Self-conscious Emotions: A Theoretical Model,” Psychological Inquiry, 15, 103-25. Wang, Yajin and Vladas Griskevicius (2014), “Conspicuous Consumption, Relationships, and Rivals: Women's Luxury Products as Signals to Other Women,” Journal of Consumer Research, 40(5), 834-854. Ward, Morgan K., and Darren W. Dahl (2015), “Should the Devil Sell Prada? Retail Rejection Increases Aspiring Consumers’ Desire for the Brand,” Journal of Consumer Research. Wilcox, Keith, Hyeong Min Kim, and Sankar Sen (2009), “Why Do Consumers Buy Counterfeit Luxury Brands?” Journal of Marketing Research, 46(2), 247-59. 64 WEB APPENDIX Pilot Study: Compliment Validation Because the participants in the main study were unaware that the male student who complimented them was a confederate, we could not ask them about their perceptions of him and his statement. However, in an effort to ensure the consistency of how the compliment was delivered across the two hired male confederates, we recorded sessions each day. We later used these videos in a separate pilot study to further assess the perceived attractiveness of the male student confederates and the perceived genuineness of the compliments they delivered. Participants and design. One hundred and eighty female participants (M age = 32.6, SD = 11.1) were asked to view these recorded sample sessions in an online pilot study. Each female was randomly assigned to view one of eight video clips in a 2 signal condition (loud/ quiet) x 2 actor (actor 1/ actor 2) x 2 session sample (sample 1/ sample 2) with the expectation that we would collapse across the session sample factor. Participants were told that they would watch a short video clip of a social interaction and answer a series of questions about it. They read the following instruction: “We are interested in your reactions to what you see and hear, particularly what you hear the male student say. Please answer the questions to the best of your ability.” Each video clip lasted approximately 90 seconds, and it consisted of the female VS associate showing the participant to her computer and then leaving the room followed by the male confederate complimenting the female student and her response. Measures. After viewing the video, participants were asked to evaluate the statement made by the male student. Specifically they were asked to indicate the extent to which the statement made was positive, negative, genuine, and a compliment on a 5-point scale (1=not at all, 5=very much). On the next screen, participants were asked for their evaluations of the male student. They were asked to indicate the extent to which the male student was attractive, sincere, handsome, genuine, and friendly on a 5-point scale (1=not at all, 5=very much). Results. We first collapsed across session samples for each actor within a given signal condition (i.e., two samples of the same actor delivering his compliment in the same (e.g., loud) condition). A multivariate analysis of all the measures revealed no significant differences for signal condition (loud vs. quiet) or actor (actor 1 vs. actor 2). The statements delivered in the loud and quiet conditions were viewed as equally positive (F<1), negative (F<1), genuine (F(1, 176) = 1.38, NS), and complimentary (F<1). Moreover, the male student was perceived as equally attractive (F<1), handsome (F<1), genuine (F<1), sincere (F<1), and friendly (F<1) in both the loud and quiet signal conditions. Similarly, no differences were found between the two actors. The statement was viewed as equally positive, negative, genuine, and complimentary (all F<1), and the male student was perceived as equally attractive, handsome, genuine, sincere, and friendly (all F<1) regardless of which of the two male confederates had delivered it. 65 Pilot Study: Compliment Evaluation Means and Standard Errors By Condition Signal Strength Condition Confederate Dependent Variable Quiet Mean (SE) Loud Mean (SE) Actor 1 Mean (SE) Actor 2 Mean (SE) The statement made was positive. 4.49 (.07) 4.51 (.07) 4.47 (.07) 4.54 (.07) The statement made was negative. 1.27 (.07) 1.34 (.07) 1.32 (.07) 1.29 (.07) The statement was genuine. 3.86 (.10) 4.03 (.10) 3.90 (.10) 4.00 (.10) The statement was a compliment. 4.57 (.08) 4.52 (.08) 4.50 (.08) 4.59 (.08) The male student was attractive. 3.99 (.09) 3.94 (.09) 4.01 (.09) 3.92 (.09) The male student was handsome. 3.98 (.09) 3.92 (.09) 3.99 (.10) 3.91 (.09) The male student was genuine. 3.80 (.10) 3.90 (.10) 3.81 (.10) 3.89 (.10) The male student was sincere. 3.80 (.10) 3.85 (.10) 3.76 (.10) 3.89 (.10) The male student was friendly. 4.36 (.08) 4.33 (.08) 4.37 (.08) 4.33 (.08) Note: No significant differences were found between the two signal conditions or between the two actors. 66 Study 2: Recruitment Poster Study 2: Sample Victoria Secret Materials (e.g., Name Tag) Study 2: Screensaver