Body Image 12 (2015) 98–107

Contents lists available at ScienceDirect

Body Image journal homepage: www.elsevier.com/locate/bodyimage

Self-affirming trait kindness regulates disgust toward one’s physical appearance夽 Philip A. Powell a,∗ , Jane Simpson b , Paul G. Overton a a b

Department of Psychology, University of Sheffield, Sheffield, UK Division of Health Research, Lancaster University, Lancaster, UK

a r t i c l e

a b s t r a c t

i n f o

Article history: Received 20 May 2014 Received in revised form 21 October 2014 Accepted 23 October 2014 Keywords: Appearance Disgust Emotion Kindness Self-affirmation Self-disgust

In two studies, self-affirming the behavioral trait of kindness was examined as a method of regulating state disgust toward one’s physical appearance. In Study 1, 56 participants (37 women, 19 men, Mage = 33.16 years) completed either a questionnaire designed to self-affirm kindness or a control equivalent and rated their disgust, anger, sadness, and happiness toward their appearance and behavior. In Study 2, 116 individuals (83 women, 33 men, Mage = 24.90 years) participated in the same experiment over the internet in an ecologically valid context. When controlling for trait self-disgust, the self-affirmed in Study 1 reported significantly less disgust toward their appearance (p 2 = .12, p = .011). This effect was replicated in Study 2, but driven by lower state disgust levels in those higher in trait self-disgust (f2 = .10, p = .001). Affirming valued traits, like kindness, may be a useful tool for regulating disgust toward body image. © 2014 Elsevier Ltd. All rights reserved.

Introduction Disgust is a universal basic emotion with a multiplicity of triggers. The “yuck” reaction is elicited by stimuli as diverse as dirt and deformity, parasites and pedophiles. While the origins of revulsion lie in the prevention of infectious disease (Curtis, Aunger, & Rabie, 2004), this pre-adapted aversion response is theorized to have acquired broader functionality through the documented process of evolutionary exaptation (Chapman & Anderson, 2012). Hence, contemporary repugnance can be seen as a normative reaction to diverse (socioculturally defined) stimuli, including unattractive bodily features, atypical sexual acts, and certain immoral behaviors, leading to its metaphorical description as the emotion of body and soul (Rozin, Haidt, & McCauley, 1999). Both biological and cultural processes have contributed to the expansion in stimuli that evoke disgust (Rozin & Haidt, 2013). While some researchers disagree over the theoretical classifications used (e.g., Tybur, Lieberman, Kurzban, & DeScioli, 2013), the most influential model of disgust to date is that of Rozin, Haidt, et al. (1999). Rozin and colleagues (1999) traced disgust’s evolution

夽 This research was supported by a postgraduate studentship grant from the Howard Morton Trust, Sheffield, UK. ∗ Corresponding author. Present address: InstEAD, Department of Economics, University of Sheffield, 9 Mappin Street, Sheffield S1 4DT, UK. E-mail address: p.a.powell@sheffield.ac.uk (P.A. Powell). http://dx.doi.org/10.1016/j.bodyim.2014.10.006 1740-1445/© 2014 Elsevier Ltd. All rights reserved.

through four qualitatively distinct stages, in response to emerging threats to the human body and mind. Disgust elicitors are proposed to have expanded from a “core” set with an oral incorporation focus (e.g., animals, rotten foods), to include reminders of our “animal nature” (e.g., body envelope violations, death), “interpersonal” threats (e.g., strangers, undesirable conspecifics), and “moral” elicitors that defile purity and sanctity (e.g., drug use, incest; Rozin, Lowery, Imada, & Haidt, 1999). Though the stimuli that induce disgust have increased, the characteristic features of the emotion (i.e., behavioral rejection, discrete facial expression, parasympathetic response, associated cognitions, and feelings of revulsion/nausea) have remained largely the same (e.g., Chapman, Kim, Susskind, & Anderson, 2009). Disgust is linked to magical thinking (Rozin & Fallon, 1987); innocuous objects that bear some similarity to, or have been in contact with, established disgust elicitors may elicit comparable responses (e.g., acne resembling infectious disease; Oaten, Stevenson, & Case, 2011). Furthermore, disgust is molded socioculturally and cultural variations exist at all levels (e.g., the Swedish delicacy of Surströmming, or fermented Baltic Sea herring, which is repulsive to most other cultures), but are greatest in the interpersonal and sociomoral disgusts (Rozin, Haidt et al., 1999). Disgust and the Body As an emotion that evolved to guard the body border (Rozin, Haidt et al., 1999), disgust has a particularly intimate relationship

P.A. Powell et al. / Body Image 12 (2015) 98–107

with body image. It has been linked to body dissatisfaction, criticism, and stigma (e.g., Griffiths & Page, 2008; Park, Schaller, & Crandall, 2007), and the violation of particular sociocultural bodily ideals (e.g., female menstruation; Fahs, 2014; Roberts, Goldenberg, Power, & Pyszczynski, 2002). Disgust is also related to physical and mental health conditions that involve distortions of the body, such as cancer (Smith, Loewenstein, Rozin, Sheriff, & Ubel, 2007), eating disorders (Troop & Baker, 2009), and body dysmorphic disorder (BDD; Onden-Lim, Wu, & Grisham, 2012). In a number of papers disgust has been linked to obesity stigma. Vartanian (2010) found disgust reactions to be a stronger predictor of weight bias than attributions of control or demographic characteristics (e.g., body mass index). In addition, obese people were rated as more disgusting than 11 other marginalized groups, including the homeless, homosexuals, and African Americas (but not politicians or drug addicts). O’Brien et al. (2013) explored the tripartite associations between appearance concerns, disgust, and anti-fat prejudice in a large Icelandic student sample. The researchers reported significant univariate and multivariate associations between these constructs. Disgust toward overweight body types may be driven by evolutionary fitness concerns (Park et al., 2007) and breaches of morality (O’Brien et al., 2013), given that obesity is moralized in the West (see Lieberman, Tybur, & Latner, 2012). The links between disgust and physicality are not confined to weight, but extend to other body features as well. Propensities for “pathogen” (i.e., core) disgust have been shown to predict (physical) mate preferences. Heterosexual women higher in pathogen disgust prefer increased masculinity (Jones, Feinberg, et al., 2013), while heterosexual men prefer more facial femininity (Jones, Fincher, et al., 2013). Park, van Leeuwen, and Stephen (2012) examined the link between pathogen disgust and attractiveness ratings of mixedsex faces in students. Park et al. (2012) found that sensitivity to pathogen disgust was negatively correlated with attractiveness ratings for unattractive faces, but not attractive faces (as determined by upper and lower quartiles in attractiveness ratings). Thus, it appears disgust is related to (un)attractiveness judgments in general, not solely those regarding weight. Physical atypicality is also a major trigger of disgust (e.g., Smith et al., 2007). Examining disgust ratings of photos showing varying degrees of physical disfigurement, Shanmugarajah, Gaind, Clarke, and Butler (2012) found that those higher in trait disgust sensitivity reported greater disgust toward all images. Moreover, this pattern held across “core,” “animal-reminder,” and “contamination” disgust sensitivities, suggesting multiple influences. Kleyn et al. (2009) recorded a diminished neural response to disgusted faces, compared to healthy controls, in a small group of men with psoriasis. This was interpreted as evidence of an acquired coping response to others’ revulsion. Disgust, then, is elicited by physical bodies that appear to violate a range of sociocultural and evolutionary fitness ideals. Self-Directed Disgust Disgust elicitors are often contextualized as external to the agent, yet the self (and its attributes) can function as its own disgust object (Overton, Markland, Taggart, Bagshaw, & Simpson, 2008; Powell, Simpson, & Overton, in press; Power & Dalgleish, 2008). Indeed, individuals may experience disgust toward their bodies, psychological characteristics, and behaviors (Powell, Overton, & Simpson, 2014). O’Brien et al. (2013) observed significant associations between disgust sensitivity and concerns with one’s physical appearance. Fahs (2014) explored women’s narratives about their vaginas, reporting a theme of disgust, or “dirty or gross.” Physical self-disgust has also been recorded in experiments, for example in women asked to try on swimwear (Fredrickson, Roberts, Noll,

99

Quinn, & Twenge, 1998). Further, heightened self-disgust features in mental health problems that involve issues with the body and/or bodily awareness, such as eating disorders (e.g., Espeset, Gulliksen, Nordbø, Skårderud, & Holte, 2012), BDD (e.g., Onden-Lim et al., 2012), sexual trauma (e.g., Jung & Steil, 2012), and certain cases of depression (e.g., Castle & Phillips, 2002). Powell, Simpson, et al. (in press) sketched a model of dysfunctional self-disgust as an “emotion schema” (Izard, 2007), or a lasting, disgust-based cognitive-affective orientation to (some enduring and important feature of) the self. While the affective part of the schema is theorized as qualitatively similar to the emotional experience of disgust, the cognitive and higher-order content involves a lasting appraisal of (some feature of) the self as repulsive (i.e., “my body is revolting”). In this model, state and trait constructs are theorized to interact reciprocally, with state disgust contributing to elaborated cognitive content, which facilitates a lasting appraisal of (some aspect of) the self as a disgust object. Trait self-disgust, through its top-down influence on information processing, in turn facilitates more frequent state self-disgust reactions (i.e., to a range of triggers perceived as fitting with the self as disgusting). While disgust toward one’s body is likely to be concomitant with other complex, self-directed affective phenomena such as shame (Power & Dalgleish, 2008), it can be theoretically distinguished by its particular cognitive-affective content (see Powell, Simpson, et al., in press). Unique features such as phenomenological revulsion, a discrete physiological and expressive profile, link to contamination and magical thinking, and specific cognitions (e.g., “yuck, that is revolting”) suggest an emotional experience of disgust rather than something like shame (which is related to hierarchical submission and diminished social rank; Gilbert, 2007). It is possible for individuals to be self-critical and/or shameful about their bodies (e.g., when of smaller stature) without finding it disgusting, yet these phenomena can be often linked (Powell et al., 2014). Given that heightened disgust toward one’s body is negatively associated with psychological well-being (Powell, Simpson, et al., in press), it is surprising that little research has looked at techniques that may attenuate self-disgust. This is a notable oversight, as recent longitudinal work suggests trait self-disgust (especially toward physical appearance) may act as a vulnerability factor for later depression (Powell, Simpson, & Overton, 2013). Thus, it is of empirical and clinical interest whether feelings of physical selfdisgust can be reduced. The limited work to date has been restricted to simple exposure techniques, such as mirror-exposure in BDD (Neziroglu, Hickey, & McKay, 2010). However, it is plausible that other psychological methods may be effective. One possibility is self-affirmation (Steele, 1988). Self-Affirmation The act of self-affirmation is hypothesized to protect against, or offset, specific threats to one’s global self-worth and self-integrity through the assertion of an alternative valued characteristic (Sherman & Cohen, 2006; Steele, 1988). Central to this theory is the idea of a multifaceted self-concept, with manifold divisible aspects of the self contributing to one’s overall self-worth and integrity. Affirming an unrelated, but important, self-aspect is theorized to compensate for, and thus attenuate, a threat to self-worth in another domain (Steele, 1988). For instance, threats to physical appearance may be offset, or buffered, by affirming one’s academic capabilities (e.g., Bergstrom, Neighbors, & Malheim, 2009). Self-affirmation manipulations have been shown to have a range of beneficial effects (McQueen & Klein, 2006; Sherman & Cohen, 2006). For instance, self-affirming consistently reduces defensive biases and increases openness to uncongenial information (e.g., Cohen, Aronson, & Steele, 2000; Reed & Aspinwall, 1998;

100

P.A. Powell et al. / Body Image 12 (2015) 98–107

Sherman et al., 2009), including health messages in at-risk populations (Harris & Epton, 2009; Sherman, Nelson, & Steele, 2000). Studies of self-affirmation also suggest it may reduce social prejudice (Fein & Spencer, 1997), lower physiological markers of stress (Creswell et al., 2005), lessen ruminative thinking (Koole, Smeets, van Knippenberg, & Dijksterhuis, 1999), and improve task performance (Sherman et al., 2009). Moreover, even brief affirmations have been shown to have at least short-term positive effects on health behaviors (Epton & Harris, 2008; Harris & Epton, 2009; Sherman et al., 2000), with a recent study attributing enhanced weight loss (over 2 months) to a brisk experimental self-affirmation task (Logel & Cohen, 2012). A few studies have extended the self-affirmation paradigm to explore its effect(s) on certain self-focused variables, including body image (e.g., Armitage, 2012; Boyce, Martens, Schimel, & Kuijer, 2012; Bucchianeri & Corning, 2012). Such work is pertinent in the current context, as disgust is often associated with body dissatisfaction (Powell et al., 2014). In recent research, Armitage (2012) randomly allocated adolescents to either a selfaffirmation manipulation or control task, before having them rate their body shape, weight, body satisfaction, and self-esteem. The author found that affirmed girls (vs. controls) were less worried about rating their appearance, reported greater body satisfaction, and were more likely to derive their state self-esteem from domains other than their appearance (see also Bergstrom et al., 2009). Similarly, Bucchianeri and Corning (2012) reported reduced body dissatisfaction and criticism in affirmed college students. Thus, if self-affirmation reduces body dissatisfaction, it may also be useful in regulating disgust toward the physical self.

Study 1 The aim of Study 1 was to examine the effects of a verified method of affirming trait kindness (Reed & Aspinwall, 1998) on state disgust toward participants’ appearance in a controlled laboratory environment, by randomly assigning participants to either a self-affirmation or control task and then measuring their state emotions. As people differ in trait self-disgust (Overton et al., 2008), and thus are likely to experience commensurate levels of threat when rating their appearance, we sought to control for this heterogeneity. This was achieved in two ways. First, by measuring participants’ trait self-disgust in advance of the experiment and block-randomizing to conditions. Second, by including an index of trait self-disgust as a covariate in the analyses, thereby improving their power and accuracy (Miller & Chapman, 2001). Based on previous work (e.g., Armitage, 2012) we hypothesized that, when controlling for trait levels of self-disgust, participants in the selfaffirmed group would report lower levels of disgust toward their physical appearance than controls. To test the secondary questions (that self-affirming kindness may have differing effects on distinct emotions and in different self-domains) we included measures of three other universal, basic emotional states: anger, sadness, and happiness (Ekman, 1992), and also measured the effect of self-affirming on emotions toward behavior. Behavior was chosen as a broad self-domain unrelated to physical appearance, but explicitly related to the affirmed trait of kindness. As disgust is related to anger and sadness (Powell et al., 2014), we predicted that lower levels of state disgust in the selfaffirmed group would be accompanied by lower levels of sadness and anger. Finally, given that self-affirmation has positive effects in unrelated domains (Blanton, Cooper, Skurnik, & Aronson, 1997), it was hypothesized that self-affirming kindness would have no effects on emotional ratings toward behavior.

The Present Research Method Gaps remain in the knowledge of what factors drive efficacious self-affirmation (Harris & Epton, 2010; McQueen & Klein, 2006; Sherman & Cohen, 2006). The evidence for self-esteem acting as a mediatory candidate is inconsistent (McQueen & Klein, 2006; Sherman & Cohen, 2006). Some have argued (e.g., Tesser, 2000) that self-affirmation could work by increasing positive affect, but the recorded effects of self-affirmation on mood are variable (McQueen & Klein, 2006). However, most studies have employed rather general, global measures of affect (e.g., Armitage & Rowe, 2011; Bucchianeri & Corning, 2012), and, to our knowledge, no work has explored how self-affirmation affects discrete emotions (i.e., disgust, anger, happiness etc.), or emotions to specific, separable aspects of the self. This is a crucial omission, as depending on the context, it is plausible that self-affirmation may affect discrete, rather than globalized, affective states toward specific aspects of the self. The main objective of the present research was to test selfaffirmation of an unrelated behavioral trait, kindness, as a potential mechanism of regulating state disgust toward physical appearance, across individuals with varying levels of trait self-disgust. Two secondary interests were to assess whether any observed effects were specific to disgust (vs. other discrete emotions, such as sadness), and whether the effects were limited to physical appearance (vs. another domain of the self). Two experiments were conducted. The first was an initial test of whether self-affirming kindness led to less reported disgust toward physical appearance in the laboratory. In Study 2 we investigated whether the effects in Study 1 could be replicated in a larger sample with greater variance in trait self-disgust scores, and who self-affirmed outside of controlled laboratory conditions (i.e., in a personal environment).

Participants. From a larger sample of 311 participants taking part in a related study, 56 volunteers (37 women, 19 men) were recruited by e-mail at the University of Sheffield, UK (see Procedure). Participants were either staff (n = 24) or students (n = 32) at this institution. Participants’ ages ranged from 19 to 64, with a mean of 33.16 (SD = 11.27), and the majority described their ethnicity as White British (n = 41). The experimental subsample (n = 56) did not significantly differ on any observed variable from those in the larger pool who did not participate (n = 255). A £50 (∼$83) prize draw incentive was offered for participation. Materials. Self-affirmation manipulation. The task used to self-affirm kindness was the Personal Attributes Inventory developed by Reed and Aspinwall (1998), which has been shown to have similar efficacy to a range of other self-affirmation techniques (Armitage & Rowe, 2011). For each of 10 kindness behaviors, participants were asked to indicate (yes/no) whether they can recall performing the behavior. Illustrative items include: “Have you ever attended to the needs of another person?” and “Have you ever forgiven another person when they have hurt you?” If a participant responded “yes” to a question they were prompted to write down brief examples of their past acts of kindness. Control comparison. Participants in the control condition completed the Personal Opinion Survey, designed by Reed and Aspinwall (1998) to be similar to the Personal Attributes Inventory, but contain no self-relevant content and be unrelated to kindness. For each of 10 statements, participants indicated (yes/no) whether

P.A. Powell et al. / Body Image 12 (2015) 98–107

they endorsed that opinion. Illustrative items include: “I think that cooking is an important skill to possess” and “I think that the beach is a great place to go on holiday.” As in the experimental condition, control participants were asked to elaborate on their opinion when they responded “yes” to an item. Measures. Trait self-disgust. Participants’ trait self-disgust was measured using either the Self-Disgust Scale (SDS; Overton et al., 2008) or Self-Disgust Scale-Revised (SDS-R; Powell, Overton, & Simpson, in press). The 18-item SDS measures trait disgust for the self. For each statement (e.g., “I find myself repulsive”, “My behavior repels people”) participants rate their agreement on a 7-point Likert scale (1 = strongly agree, 7 = strongly disagree). A total score is obtained by omitting six filler items, reverse coding nine items, and then summing the resulting 12 values. Possible scores range from 12 to 84, with a higher score indicating higher trait self-disgust. The test–retest reliability of the SDS is excellent, with correlation coefficients of .94, .82, and .73, over 1-week, 6-, and 12-months respectively (Overton et al., 2008; Powell et al., 2013). Internal consistency of the SDS in this study was excellent (˛ = .91). The 22item SDS-R is a modified version of the SDS designed to increase its face validity (Powell et al., 2014), and also measures trait selfdisgust. Relative to the SDS, the SDS-R has four new items and five revised ones. Participants rate how much they agree with each statement (e.g., “I find the way I look nauseating”, “I am sickened by the way I behave) on a 7-point Likert scale (1 = strongly disagree, 7 = strongly agree). A total score is obtained by removing seven filler items, reverse coding four items, and then summing the resulting 15 values. Possible scores range from 15 to 105, with a higher score indicating higher self-disgust. Internal consistency of the SDS-R in this sample was also excellent (˛ = .93). Perceived threat. The five items from Armitage (2012) were used to assess participants’ perceived threat toward rating their appearance and behavior. Two identical 5-item scales were used to measure threat toward rating physical appearance and behavior separately. For each of five adjectives (fearful, anxious, nervous, tense, and uncomfortable), participants were asked to complete the stem “Rating aspects of my appearance/behavior makes me feel . . .” on a 7-point Likert scale (e.g., 1 = not at all fearful, 7 = very fearful). The Cronbach’s alphas for the two scales in the present study were excellent (both ˛ = .97). State emotion. Eight visual analog scales (VASs) were used to record how much of four basic emotions (disgust, anger, sadness, and happiness) participants currently felt toward their physical appearance and the way they behaved. For each emotion, participants responded to the stem: “Thinking about it now, my appearance/behavior makes me feel . . .” on a 100-point VAS (e.g., 0 = not at all disgusted, 100 = extremely disgusted). Procedure. Ethics approval was granted by the Department of Psychology Ethics Committee at the University of Sheffield, UK prior to data collection. As part of a related study examining the revised version of the SDS (i.e., the SDS-R; Powell, Overton, et al., in press), 311 staff and students at the University of Sheffield responded to an e-mail invitation to take part in an online study on their “emotional responses.” As part of this study, participants completed demographic questions and a measure of trait self-disgust (with approximately half the participants completing the SDS and half the SDS-R). Participants were informed that they may be invited to take part in a follow-up experiment on their “personal values and emotional responding” and that further information would be provided.

101

All participants were invited to participate in the experiment by e-mail approximately two weeks after the questionnaire study. Participants were told that the experiment would take place in the Psychology Department, University of Sheffield, UK and involved a “short experimental task and filling out some questionnaires.” Individual appointments for those who responded to the e-mail were scheduled by a research assistant (RA). Consenting volunteers were block-randomized (within quartiles of trait self-disgust scores) to the self-affirmation (n = 28) or control condition (n = 28) using randomized allocation lists. Upon arrival participants were greeted by an RA, who introduced the study and then waited outside the room to facilitate anonymity. The study was run in the same location, on a local computer, on each occasion. All materials were administered electronically, beginning with informed consent. Participants were then presented with the Personal Attributes Inventory or Personal Opinion Survey (i.e., the manipulation) depending on their condition. Following this, participants were informed they were about to rate their appearance and behavior (i.e., the threat), and completed the perceived threat items from Armitage (2012). Finally, participants reported their state levels of emotion toward their appearance and behavior, and were prompted to fetch the researcher. Participants were fully debriefed. Data analysis. In the experimental sample, 26 participants had completed the SDS and 30 had completed the SDS-R. As the SDS and SDS-R both measure the same latent construct of trait self-disgust1 (but on different scales), the variables were first standardized (making scores across the measures comparable), and then combined into a composite measure of trait self-disgust across all 56 participants. A series of analyses of covariance (ANCOVAs) were conducted with experimental condition as the independent variable (IV), and perceived threat, state disgust, anger, sadness, or happiness toward appearance or behavior as the dependent variable (DV). Trait self-disgust was entered as a covariate in all analyses. In each analysis, the marginal effects of condition and trait self-disgust were modeled as well as their product, to check the homogeneity of regression slopes (Field, Miles, & Field, 2012). If the product term was non-significant, the model was re-specified to estimate main effects. Any DVs with significant within-group deviations from the normal distribution (skew and/or kurtosis z-scores >±1.96) and/or heterogeneous variances across groups (Levene’s test with p < .05) were optimally power transformed prior to ANCOVA using the BoxCox procedure (Osborne, 2010). In one instance (i.e., measure of perceived threat to rating appearance), where power transformation was unsuccessful, ANCOVA on rank-transformed scores was used as a robust, non-parametric alternative (Conover & Iman, 1982; Quade, 1967). Untransformed values (adjusted for the covariate) were used for plotting and interpreting effects. To correct for multiple analyses, a partial Bonferroni correction was applied to the chosen alpha value (˛ = .05), which takes into account the average shared variance between the DVs. For 10 outcome variables, and an average Pearson correlation of .51, a corrected alpha value of .016 was employed in the present study (see Uitenbroek, 1997). All analyses were carried out using SPSS v. 20 (IBM Corp., Armonk, NY, USA). Results Preliminary analyses. Experimental and control groups did not significantly differ on measures of gender, 2 (1, N = 56) = 0.72,

1 The SDS and SDS-R were found to be very highly correlated when administered together in a large independent sample, r(230) = .94, p < .001.

102

P.A. Powell et al. / Body Image 12 (2015) 98–107

Fig. 1. State ratings of perceived threat and emotion toward physical appearance in Study 1. Participants in the self-affirmation condition (n = 28) reported significantly lower levels of state negative emotion toward their physical appearance, than those in the control condition (n = 28). MPT = measure of perceived threat. Means adjusted for trait self-disgust scores. Error bars represent the standard error of the mean. Symbols denote significance at † p < .05, *p < .016, and **p < .01.

Fig. 2. State ratings of perceived threat and emotion toward behavior in Study 1. Levels of state emotion toward behavior were not significantly different between the self-affirmation (n = 28) and control (n = 28) conditions. MPT = measure of perceived threat. Means adjusted for trait self-disgust scores. Error bars represent the standard error of the mean.

Discussion p = .397,  = .11; ethnicity, 2 (1, N = 56) = 2.28, p = .131,  = .20; student status, 2 (1, N = 56) = 0.29, p = .589,  = .07; or age, U = 384, N1 = 28, N2 = 28, p = .896, r = .02. Moreover, there were no significant differences in trait self-disgust, t(54) = 0.42, p = .678, r = .06. These analyses suggest experimental randomization was successful and the critical assumption of independence between the covariate and IV in ANCOVA was tenable (Field et al., 2012; Miller & Chapman, 2001). For all DVs, the covariate*IV interaction was not significant, and was omitted in the models below (Field et al., 2012).

Analyses of covariance.

Emotion ratings toward appearance. The rank ANCOVA on perceived threat to rating appearance indicated a significant positive main effect of trait self-disgust, F(1, 53) = 24.50, p < .001, p 2 = .32, and a trend for less threat in the self-affirmation condition, F(1, 53) = 5.48, p = .023, p 2 = .09. Parametric analyses revealed that trait self-disgust significantly, positively predicted state disgust, F(1, 53) = 37.05, p < .001, p 2 = .41; state anger, F(1, 53) = 26.44, p < .001, p 2 = .33; and state sadness, F(1, 53) = 31.70, p < .001, p 2 = .37, toward appearance; and negatively predicted state happiness, F(1, 53) = 11.92, p = .001, p 2 = .18. There were also significant effects of condition on state disgust, F(1, 53) = 6.90, p = .011, p 2 = .12; state anger, F(1, 53) = 7.54, p = .008, p 2 = .13; and state sadness, F(1, 53) = 9.20, p = .004, p 2 = .15, toward appearance, with affirmed participants reporting lower levels. However, there was no significant effect of condition on state happiness, F(1, 53) = 2.99, p = .089, p 2 = .05. Group means (adjusted for trait self-disgust) are presented in Fig. 1.

Emotional ratings toward behavior. For behavior, there were significant positive main effects of trait self-disgust on perceived threat, F(1, 53) = 33.90, p < .001, p 2 = .39; state disgust, F(1, 53) = 68.23, p < .001, p 2 = .56; anger, F(1, 53) = 53.47, p < .001, p 2 = .50; and sadness, F(1, 53) = 36.74, p < .001, p 2 = .41; and a negative effect on happiness, F(1, 53) = 31.50, p < .001, p 2 = .37. However, there were no significant effects of condition on perceived threat, F(1, 53) = 1.13, p = .292, p 2 = .02; state disgust, F(1, 53) = 0.75, p = .390, p 2 = .01; anger, F(1, 53) = 0.14, p = .710, p 2 < .01; sadness, F(1, 53) = 1.01, p = .321, p 2 = .02; or happiness, F(1, 53) = 2.70, p = .109, p 2 = .05, toward behavior. Group means for the behavioral variables (adjusted for trait self-disgust) are presented in Fig. 2.

Study 1 assessed the effect of self-affirming kindness on state disgust ratings toward participants’ physical appearance. Analyses showed that, when controlling for trait levels of self-disgust, affirmed participants reported significantly less state disgust toward their own appearance than controls. This effect was the same for the discrete emotions of anger and sadness, but not happiness. As predicted, the effects were also specific to emotions toward participants’ physical appearance, and not emotions toward behavior. The current findings complement work reporting reductions in body dissatisfaction as a consequence of affirming personality traits (e.g., Armitage, 2012; Bucchianeri & Corning, 2012). That self-affirming the trait of kindness affected emotion toward appearance, but not behavior, is consistent with the wider literature. Past work suggests that affirming a related domain may have adverse or null effects, potentially increasing defensiveness and dissonance (e.g., Blanton et al., 1997; Sherman & Cohen, 2006). This result may help explain the inconsistencies between studies assessing the effects of self-affirmation on emotion, which have used globalized affect measures (e.g., Armitage & Rowe, 2011; Sherman et al., 2000). Indeed, few studies have considered whether self-affirming alters self-directed feelings (Harris & Epton, 2010), and those that have used global indices with disparate results (e.g., Armitage & Rowe, 2011; Sherman et al., 2000). These findings suggest that self-affirming kindness may exclusively alter self-directed emotion in the unrelated, physical domain. In this study, the beneficial effects of self-affirming were driven by the regulation of several discrete, yet related, negative emotions (i.e., disgust, anger, sadness) toward appearance in the affirmed, with no significant effects on positive emotion (i.e., happiness). Although there was a trend for greater happiness toward appearance in self-affirmed versus control participants, the effects on negative emotion were much stronger. This finding has notable implications for studying bodily oriented emotions; using global indices (without at least differentiating positive and negative affective states) may lead to unreliable and/or incomplete conclusions. Indeed, positive and negative affect may be best conceptualized as orthogonal to one another, rather than opposites (Watson, Clark, & Tellegen, 1988). Study 2 Study 1 showed that self-affirming the behavioral trait of kindness led to lower reported disgust (and anger and sadness) toward one’s physical appearance. The second study presented here replicates and expands upon Study 1 by examining whether the effects could be replicated in a larger sample (with greater variance in

P.A. Powell et al. / Body Image 12 (2015) 98–107

trait self-disgust) outside of the laboratory, in a less controlled (but more ecologically valid) setting. This has important implications, as psychological interventions are rarely administered in controlled conditions, and there are efforts to move toward accessible, and less resource burdening, psychological self-help methods, both in-print and online (e.g., Anderson et al., 2005; Riper et al., 2008). Study 2 was designed and disseminated entirely online, allowing participants to complete the self-affirmation task in a personal environment (e.g., at home). In all other respects Study 2 replicated Study 1. Participants’ trait self-disgust was measured online approximately two weeks before the experiment and included as a covariate in the analyses. Participants were randomized to conditions, and the same DVs were measured. Based on the results of Study 1, we predicted that the self-affirmed would report significantly less negative emotion (i.e., disgust, anger, sadness) toward their appearance, but not their behavior. Method Participants. From a larger pool of 308 participants, 116 consenting student volunteers (83 women, 33 men) were recruited from the e-mailing lists at Lancaster University, UK (see Procedure). Participants’ ages ranged from 18 to 68, with a mean of 24.90 (SD = 10.23), and the majority were White British (n = 79). The experimental subsample (n = 116) did not significantly differ demographically from those in the larger sample who did not participate (n = 192). However, those who took part had significantly higher levels of self-disgust, U = 9593.50, N1 = 192, N2 = 116, p = .042, r = .19, Thus, the generalizability of the present findings may be restricted in that regard. A £30 (∼$50) prize draw incentive was offered for participation. Materials and measures. Participants completed the SDS (Overton et al., 2008) as an index of trait self-disgust. All other materials were identical to Study 1. The internal reliabilities of the SDS (˛ = .95), measure of perceived threat toward appearance (˛ = .95), and behavior (˛ = .94) were high. Procedure. Ethics approval was granted by the Ethics Committee of the Department of Psychology, University of Sheffield, UK prior to data collection. The present study was an online replication of Study 1. To avoid contaminating the efficacy of the affirmation procedure (i.e., by disseminating measures assessing negative selfconstructs immediately before it), and to maintain consistency with Study 1, a larger participant cohort were invited to complete their demographic details and the SDS online in a study on “emotional responding,” approximately two weeks in advance of the experimental study. Participants were later invited by e-mail to take part in an online study looking at “personal values and emotional responding.” Participants were block-randomized (within quartiles of trait self-disgust scores) to the self-affirmation (n = 57) or control (n = 59) condition using randomized lists, and received a respective link to the experimental or control version of an online survey, designed on SurveyGizmo (www.surveygizmo.com). In the invitation to participate, and at the informed consent stage, participants were explicitly asked to complete the survey in one sitting, and a validation question was included at the end of the study to check that they had done this. Participants were debriefed online. All other details were the same as Study 1. Data analysis. Study 2 adopted the same analytic strategy as Study 1, involving ANCOVAs with trait self-disgust included as the covariate, and modeling of the covariate*IV interaction term to check the homogeneity of regression slopes (Field et al., 2012). Trait self-disgust and DVs with significant within-group non-normality (skew and/or kurtosis z-scores >±1.96) and/or heterogeneous

103

variances across groups (Levene’s test with p < .05), were optimally power transformed using Box-Cox (Osborne, 2010). In one instance (i.e., happiness toward behavior), where transformation was unsuccessful, rank ANCOVA was used as a robust alternative (e.g., Conover & Iman, 1982). As in Study 1, a partial Bonferroni correction was applied to mitigate an inflated experiment-wise type I error rate. With 10 outcome variables, and an average Pearson correlation of .54, a corrected alpha value of .018 was applied (Uitenbroek, 1997). All analyses were done on SPSS v. 20 (IBM Corp., Armonk, NY, USA). Results Preliminary analyses. Self-affirmed and control participants did not significantly differ on gender, 2 (1, N = 116) = 0.01, p = .929,  = .01; ethnicity, 2 (1, N = 116) = 1.26, p = .261,  = .10; or age, U = 1532, N1 = 59, N2 = 57, p = .403, r = .08. Moreover, there were no significant group differences in trait self-disgust, U = 1639, N1 = 59, N2 = 57, p = .814, r = .02. Accordingly, randomization to groups was successful, and ANCOVA was the appropriate analytical tool. All participants indicated they had completed the experimental survey in one sitting. For all outcome variables, except disgust toward appearance (see below), the covariate*IV interaction term was not significant and was omitted for the accurate interpretation of effects. Analyses of covariance. Emotional ratings toward appearance. There was a significant positive main effect of trait self-disgust on perceived threat, F(1, 113) = 47.36, p = .001, p 2 = .30; state anger, F(1, 113) = 39.18, p < .001, p 2 = .26; and state sadness, F(1, 113) = 88.09, p < .001, p 2 = .44; and a significant negative effect on happiness, F(1, 113) = 68.58, p < .001, p 2 = .38, toward appearance. Conversely, self-affirmation condition did not significantly predict threat, F(1, 113) < 0.01, p = .965, p 2 < .01; anger, F(1, 113) < 0.01, p = .967, p 2 < .01; sadness, F(1, 113) = 0.02, p = .899, p 2 < .01; or happiness, F(1, 113) = 1.28, p = .261, p 2 = .01. For state disgust toward appearance, there was a positive marginal effect of trait self-disgust, F(1, 112) = 90.57, p = .001, p 2 = .45, and condition, F(1, 112) = 7.98, p = .006, p 2 = .07, but also a significant trait self-disgust by condition interaction, F(1, 112) = 8.92, p = .003, p 2 = .07. Thus, trait self-disgust moderated the effect of self-affirmation on appearance disgust, violating the assumption of homogeneous regression slopes (Field et al., 2012). To account for this, a moderated regression was conducted and the findings reported below. Absolute group means (adjusted for the covariate) are presented in Fig. 3. Emotional ratings toward behavior. Trait self-disgust positively predicted behavioral threat, F(1, 113) = 24.49, p < .001, p 2 = .18; state disgust, F(1, 113) = 37.09, p < .001, p 2 = .25; anger, F(1, 113) = 52.80, p < .001, p 2 = .32; and sadness, F(1, 113) = 53.50, p < .001, p 2 = .32; and negatively predicted (ranked) happiness, F(1, 113) = 36.40, p < .001, p 2 = .24. There were no significant betweengroup differences on perceived threat, F(1, 113) = 0.50, p = .483, p 2 < .01; disgust, F(1, 113) = 0.24, p = .625, p 2 < .01; anger, F(1, 113) = 0.19, p = .668, p 2 < .01; sadness, F(1, 113) = 2.23, p = .138, p 2 = .02; or happiness, F(1, 113) = 0.27, p = .607, p 2 < .01, toward behavior. Group means are shown in Fig. 4. Moderated regression. A moderated multiple regression was conducted by z-standardizing trait self-disgust and condition, computing their product, and regressing state disgust toward appearance on these three predictors (Dawson, 2014). In order to calculate the additive effect of the interaction, the predictors were entered hierarchically, with trait self-disgust and condition in the

104

P.A. Powell et al. / Body Image 12 (2015) 98–107

Fig. 3. State ratings of perceived threat and emotion toward physical appearance in Study 2. Participants in the self-affirmation (online) condition (n = 57) reported similar levels of perceived threat, anger, sadness, and happiness to those in the control condition (n = 59). There was a significant interaction between experimental condition and trait self-disgust in predicting state disgust toward physical appearance. MPT = measure of perceived threat. Means adjusted for trait self-disgust scores. Error bars represent the standard error of the mean. Symbols denote significance at *p < .01.

Fig. 5. State ratings of disgust toward appearance across three levels of trait selfdisgust. Affirmed participants (n = 57) reported significantly lower levels of state disgust toward appearance than controls (n = 59) at a “high” level of trait self-disgust (p = .001), but not at “low” or “moderate” levels of the trait.

Discussion

Fig. 4. State ratings of perceived threat and emotion toward behavior in Study 2. Levels of state emotion toward behavior were not significantly different across the affirmed (online) group (n = 57) or control group (n = 59). MPT = measure of perceived threat. Means adjusted for trait self-disgust scores. Error bars represent the standard error of the mean.

first block, and then the interaction term second. The results are presented in Table 1. The effects of trait self-disgust and condition were qualified by a significant interaction. To interpret the interaction, simple slopes were examined, estimating the effect of condition on appearance disgust at three levels of the moderator: the mean (moderate self-disgust), one SD below (low self-disgust), and one SD above the mean (high self-disgust). This analysis indicated that condition did not significantly predict appearance at low, ˇ = .08, SE = 0.10, t(112) = 0.87, p = .388, f2 = .01, or moderate trait self-disgust, ˇ = −.12, SE = 0.07, t(112) = −1.78, p = .078, f2 = .03, but did so when trait self-disgust was high, ˇ = −.32, SE = 0.10, t(112) = −3.37, p = .001, f2 = .10, with the affirmed group reporting significantly less state disgust toward their appearance. Group means across the three levels of the covariate are presented in Fig. 5.

Study 2 was an online replication of Study 1 in a larger, independent sample. The results of Study 1 were partially replicated. In particular, self-affirming with the Personal Attributes Inventory had no significant effects on any of the behavioral DVs. Furthermore, the manipulation had no significant effects on participants’ perceived threat or happiness toward their appearance. However, in contrast to Study 1, rather than predicting all negative emotions, self-affirmation predicted state disgust toward appearance only. Moreover, this marginal effect was driven by significantly lower disgust ratings in the self-affirmed (vs. control) group in those that were high (vs. low) in baseline levels of trait self-disgust. The inconsistent findings between Study 1 and 2 may be explained, to some degree, by differences in the samples’ characteristics. First, Study 2 used an exclusively student sample (vs. staff and students in Study 1), and thus participants were significantly younger. Second, there was greater variance in participants’ trait self-disgust scores in Study 2 than Study 1, and trait selfdisgust was significantly higher in participants who volunteered than the broader sample they were recruited from. This may have contributed to the different effects observed, including – to the extent that trait self-disgust is specific to the disgust emotion – an exclusive effect of self-affirmation on disgust, rather than anger and/or sadness. More than likely, the different modes of dissemination may have also influenced the findings. While Study 1 was conducted in a controlled environment, Study 2 was carried out over the internet, letting individuals participate in more ecologically valid personal settings, but at the expense of decreased control. Thus, one may have expected the effects of an online manipulation to be weaker. Further, there is likely to have been (measured and unmeasured) differences in the ways participants engaged with the affirmation process. For instance, while the mean number of “yes”

Table 1 Hierarchical moderated regression analysis predicting appearance disgust (N = 116).

Step 1 Constant Trait self-disgust Condition Step 2 Constant Trait self-disgust Condition Trait self-disgust*Condition

B [95% CI]

SE B

ˇ

p

138.11 [122.36, 153.85] 75.70 [59.88, 91.52] −13.67 [−29.49, 2.14]

7.95 7.98 7.98

.66 −.12