Facial reactions reveal that slim is good but fat is not bad: Implicit and explicit measures of bodysize bias

Facial electromyography (EMG) was used to gauge emotional responding towards images of slim and overweight individuals, and findings were compared with data from a series of alternative measures including two implicit attitudinal procedures, the Implicit Relational Assessment Procedure (IRAP) and the Implicit Association Test (IAT), and explicit measures of anti-fat prejudice and discriminatory behavior. Images of slim individuals elicited EMG responses consistent with more positive affect. Data from both the IRAP and IAT indicated higher levels of bias than were revealed on the explicit measures, and the IRAP also corroborated the EMG pattern by indicating responses consistent with pro-slim rather than anti-fat bias. The IRAP was moderately correlated with both EMG and the IAT and was the only measure to predict behavioral intentions. Copyright © 2011 John Wiley & Sons, Ltd. Weight-related attitudes, including pro-slim attitudes on the one hand and anti-fat on the other, have been linked with significant problems such as eating disorders (Pepper & Ruiz, 2007; Thompson & Stice, 2001) and discrimination, respectively (Gapinski, Schwartz, & Brownell, 2006; Puhl & Brownell, 2001) and as such constitute important domains of psychosocial research. Accurate measurement of these attitudes is key to assessment and potential intervention. However, social desirability effects have become a concern in relation to the explicit self-report measures employed in traditional research (e.g., Dittmar, Halliwell, & Stirling, 2009; Mills, Polivy, Herman, & Tiggemann, 2002). Hence, a recent trend in measurement is the use of alternative procedures, which circumvent sources of bias characterizing self-reports. Perhaps the most prominent of such measures is the Implicit Association Test (IAT; Greenwald, McGhee, & Schwartz, 1998). The IAT has been employed as a measure of weight-related implicit attitudes in several previous studies (see Morrison, Roddy, & Ryan, 2009 for a review). In the critical phase of a typical IAT study of this phenomenon, participants are required to respond as rapidly as possible in accordance with two separate patterns: (i) categorize “Slim” with “Good” and “Fat” with “Bad” and (ii) categorize “Slim” with “Bad” and “Fat” with “Good”. Several studies have found that participants are faster on average to demonstrate the first pattern than the second, which is taken as evidence of “anti-fat” bias (e.g., Brochu & Morrison, 2007; Roddy, Stewart and Barnes-Holmes, 2010). Furthermore, participants tend to produce higher levels of bias on the IAT than with self-report suggesting that the IAT is revealing levels of prejudice that might not otherwise become apparent. *Correspondence to: Ian Stewart, School of Psychology, National University of I E-mail: [email protected] Copyright © 2011 John Wiley & Sons, Ltd. Despite its popularity as an implicit measure, several limitations of the IAT have been noted. Perhaps the most cited of these is that the associations found for any particular concept are always relative (De Houwer, 2002). For example, in the context of “anti-fat” bias, the standard IAT does not provide a measure of separate attitudes to “fat” and “slim”, and thus the meaning of an IAT effect with respect to attitudes towards the overweight is ambiguous; for example, it could indicate a neutral attitude towards slimandanegative attitude towards fat, or apositive attitude towards slim and a neutral attitude towards fat. One key advantage of the Implicit Relational Assessment Procedure (IRAP; Barnes-Holmes, Barnes-Holmes, Stewart & Boles, 2010), a more recent measure of implicit cognition, is that in contrast with the IAT, it permits independent evaluation of concepts. Thus, in the context of weight-related attitudes it provides information on separate attitudes towards “slim” and “fat”. To illustrate, on a typical IRAP trial the participant responds by categorizing the relationship between a label (e.g., “Good” or “Bad”) and a target stimulus (e.g., “Slim” or “Fat”) as quickly as possible by pressing one of two keys (e.g., “Same” and “Opposite”). Four different trial types can be created by presenting different combinations of label and target stimuli. For example, in the case of an IRAP examining anti-fat bias, the four combinations would be Good-Slim, Good-Fat, Bad-Slim, and Bad-Fat. For each of these combinations, the participant can be required to respond in either a “pro-slim/anti-fat”, or a “pro-fat/anti-slim” manner, and the difference in mean response latency between the aggregate of the former and the latter trial types provides a measure of implicit bias. Importantly, however, in addition to an overall IRAP effect, a separate effect can be obtained for each of the four trial types, thus providing information regarding reland, Galway, Galway City, Ireland. Received 26 April 2010, Accepted 15 June 2011 Facial electromyography and body-size bias 689 specific relationships (e.g., between “Slim” and “Good” and also “Fat” and “Good”). In a recent study (Roddy et al., 2010) comparing the IRAP and IAT as measures of body shape relevant prejudice, both measures revealed bias; however, the IRAP trial type analysis indicated that participants were showing strong pro-slim attitudes in the absence of negativity towards the overweight. Both the IAT and IRAP are measures of implicit cognition and as such can provide useful information about underlying weight-related attitudes that explicit measures may fail to reveal. However, attitude theory suggests that cognition is only one facet of these phenomena and that another at least equally important component is emotion (Zajonc, 1980). Hence, exploration of implicit weight-related affective responding would constitute an important extension of research in this domain. In addition, immediate emotional responding may provide a particularly informative additional measure of implicit reaction. For example, evidence suggests affect is superior to cognition in the prediction of behavior (Vanman, Saltz, Nathan & Warren, 2004); thus affective reactions might yield better relationships with behavioral indices. One possible methodology for measurement of implicit emotion is facial electromyography (EMG; Vanman, Paul, Ito & Miller, 1997; Vanman et al., 2004). Facial EMG records muscle activity involved in facial expressions and as such allows for the identification of muscle contractions based on speed and sensitivity (Stern, Ray & Quigley, 2001). It has been found to provide a sensitive and precise measure of muscle activity to weakly evocative emotional stimuli even in the absence of a detectable facial response by an observer (Cacioppo, Petty, Losch & Kim, 1986) and to yield reliable information regarding both the valence and the intensity of emotional reactions to stimuli more generally (Cacioppo et al., 1986). Several studies have demonstrated that attitude valence is reflected in facial EMG activity recorded at the zygomaticus major, and corrugator supercilii, muscles (e.g., see Tassinary & Cacioppo, 2000). The former are linked with smiling responses, whereas the latter are linked with frowning, although EMG activity at either location can vary as a function of both positive and negative valence (e.g., Cacioppo et al., 1986; Larsen, Norris & Cacioppo, 2003). In the context of racial attitudes, facial EMG has revealed patterns of responding to target groups that contradict explicitly revealed attitudes. For example, Vanman et al. (1997) showed that Caucasian participants’ facial EMG responses to images of White and Black faces revealed bias against Blacks even though explicit friendliness ratings suggested pro-Black bias. More recently, Vanman et al. (2004) demonstrated that EMG responses of Caucasian participants to photos of White and Black targets were related to a measure of discriminatory behavior, whereas responding on the IAT in the same study was not. Based on its success as a measure of implicit prejudice in other domains such as race, the present study sought to use facial EMG as a measure of weight-related attitudes and to compare patterns found using this physiological measure with those revealed through more conventional implicit (IAT and IRAP) and explicit (Crandall’s [1994] AFA and behavioral intentions) measures. In light of extant research documenting internalization of the thin ideal and of previous research from Copyright © 2011 John Wiley & Sons, Ltd. Roddy et al. (2010), it was hypothesized that a pattern of stronger pro-slim than anti-fat responding would be detected across a number of measures.