Overcoming Gender Stereotypes

Across two experiments the present research examined the use of social-consensus feedback as a strategy for overcoming spontaneous gender stereotyping when certain social role nouns and professional terms are read. Participants were presented with word pairs comprising a role noun (e.g. surgeon) and a kinship term (e.g. mother), and asked to decide whether both terms could refer to the same person. In the absence of training, participants responded more slowly and less accurately to stereotype incongruent pairings (e.g. surgeon/mother) than stereotype congruent pairings (e.g. surgeon/father). When participants were provided with (fictitious) social consensus feedback, constructed so as to suggest that past participants did not succumb to stereotypes, performance to incongruent pairings improved significantly (Experiment 1). The mechanism(s) through which the social feedback operated were then investigated (Experiment 2), with results suggesting that success was owing to social compliance processes. Implications of findings for the field of discourse processing are discussed. OVERCOMING GENDER STEREOTYPES 3 Considerable evidence suggests that readers often make gender inferences in text comprehension when explicit gender information is lacking (e.g. Carreiras, Garnham, Oakhill, & Cain, 1996; Duffy & Keir, 2004; Garnham, Oakhill, & Reynolds, 2002; Irmen, 2007; Kreiner, Sturt, & Garrod, 2008). Such inferences typically follow the use of social or occupational role nouns that have a strong gender bias, but that are not grammatically marked for gender e.g. the term beautician is strongly female-biased while the term builder is strongly male-biased. While grammatical gender languages can largely avoid gender stereotypic inferences by employing gender specific personal nouns to convey maleness and femaleness (e.g. le musician ‘the [male] musician’/la musicienne ‘the [female] musician’ in French versus the musician in English), this is rarely possible in English. Instead, inferences based on stereotypical biases play an important role in building a cognitive representation of gender and, once established, are very difficult to overcome. This can result in processing difficulties when gender-related expectancies clash with explicitly stated gender information. Reynolds, Garnham and Oakhill (2006) gave participants a slightly adapted version of the now well known ‘surgeon riddle’ referenced by Sanford (1985). In this riddle, a father and son are involved in a car accident where the father dies but the son is taken to hospital for an operation. However, once there, the surgeon looks at the boy and exclaims “Oh my god, that is my son!” (Sanford, 1985: “I can’t do this operation. This boy is my son.”). When readers are asked how this can be, they typically infer that the surgeon is male and, despite knowing that the boy’s father is dead, fail to override this inference so as to successfully conclude that the surgeon is the boy’s mother. Indeed, Reynolds and colleagues report that 75% of readers who had not previously seen the text failed to resolve the inconsistency and update the gender of the surgeon in their mental representation. However, participants who received a different version of this riddle with the term surgeon replaced by nurse showed no difficulty in solving it (Experiment 1A). Moreover, when these findings were followed up using a self-paced reading task (Experiment 2), reading times for the final clause (which contained the gender-biased role noun) were found to be 1,000ms slower on OVERCOMING GENDER STEREOTYPES 4 the original surgeon riddle than the stereotype consistent version. This processing delay is much longer than is typically found in previous experiments that require minor accommodations to mental representations, i.e. approximately 200ms (e.g. Carreiras et al., 1996, Experiment 1; Haviland & Clark, 1974). The above evidence led Reynolds et al. to conclude that inferring gender from stereotyped role-names is at least in some part an automatic process, with the experiments highlighting how entrenched such gender inferences are and the substantial processing difficulties that they induce. Such incorrect inferences based on stereotypicality biases clearly have negative implications for discourse comprehension but are also a pervasive example of how language contributes to the maintenance and propagation of gender stereotypes in English, by artificially constraining the roles on offer to men and women. Therefore, the current article investigates how spontaneous gender stereotypes can be overcome through the use of (fictitious) social consensus feedback. This feedback involves presenting participants with social norm information relating to role-based gender stereotypes in an attempt to sway their attitudes towards the perceived (gender-fair) attitude of their peers. A sizeable body of research has now been devoted to the influence of other people’s beliefs on an individual’s own beliefs, with evidence emerging that perceivers frequently modify their intergroup attitudes and behaviours in order to align with those modelled by members of groups that they value. For example, informing participants that stereotyping is not typical of their in-group has previously been found to reduce stereotyping against groups such as racial minorities (e.g. Stangor, Sechrist, & Jost, 2001; Wittenbrink & Henly, 1996) and people suffering from obesity (Puhl, Schwartz, & Brownell, 2005) while, conversely, research has documented that discrimination against racial minorities and women is more tolerated when a racist or sexist joke has just been heard (Ford & Ferguson, 2004; LaFrance & Woodzicka, 1998). Various theories have been offered to explain this peer influence on intergroup prejudice, now thought to be driven by the basic human goals of understanding, social connection, or OVERCOMING GENDER STEREOTYPES 5 alternatively, self-definition (Paluck, 2011). For instance, Social Reality Theory posits that striving for understanding and connection pushes people to validate their experiences with others and to display actions and beliefs that others value (Hardin & Conley, 2000). Similarly, Group Norms Theory suggests that people assume the perceived attitudes and behaviours of others who exemplify admirable in-group identities in an effort to socially connect with the group (Crandall, Eshleman, & O’Brien, 2002; Kelman, 1958; Sherif & Sherif, 1953). While both of these theories predict that individuals would adjust their behaviour and attitudes in line with those of the valued reference group, other theories predict a contrasting pattern of results. Specifically, Deviance Regulation Theory (Blanton & Christie, 2003) claims that people may reject the perceived attitudes and behaviours of their peers as a means of self-definition, while the Focus Theory of normative conduct predicts that peer values will only influence individuals when they are made salient (Kallgren, Reno, & Cialdini, 2000). Overall, given past success in using social norm information as a strategy for overcoming prejudice in other domains, we posit that participants are more likely to be influenced by the perceived attitudes of their peers than to reject them as a means of selfdefinition. In a similar vein, Prentice and Miller (1993) posit that individuals will experience discomfort if they perceive their attitudes to be different from the normative attitude of their peer group. However, this discrepancy can be resolved in three ways (1) by moving an individual’s personal attitudes towards that of the perceived norm, (2) bringing the norm closer to the individual’s attitude, or (3) complete rejection of the group. Prentice and Miller maintain that the most straightforward way for an individual to reduce a perceived discrepancy in attitude is to bring their private attitudes in line with those of the group norm. Furthermore, as regards stereotyping, Stangor et al. (2001) propose three reasons why individuals should be particularly likely to be swayed by the opinions of others on this issue (1) the accuracy of stereotypes is difficult to assess objectively (2) stereotyping is a socially sensitive topic and (3) people are likely to be highly motivated to learn about the traits of individuals from different social groups. OVERCOMING GENDER STEREOTYPES 6 However, while many studies suggest people reform their attitudes and beliefs so as to mirror those of their peer group, less research has explored how well people can initially identify these social norms. In fact, it is now clear that they can make significant errors in their estimation of opinions held by others (e.g. Prentice & Miller, 1993). The use of fictitious norm information as a stereotype reduction strategy in this, and other, research therefore benefits from the fact that people are often poor at estimating social norms, yet are strongly influenced by what they perceive these norms to be. With the above information in mind, Experiment 1 was designed to investigate the effect of fictitious social consensus feedback on levels of gender stereotype endorsement on a straightforward judgement task. This task was originally devised by Oakhill, Garnham and Reynolds (2005) who conducted a series of six experiments investigating gender biases associated with single role nouns and the extent to which such bias information could be overcome. Participants were required to judge quickly whether two terms presented on screen could refer to one person. The terms comprised a role noun (that was either definitionally gendered or stereotype biased e.g. princess, beautician respectively) and a kinship term (that was definitionally gendered in all cases e.g. brother, sister). In order to successfully respond, participants were required take definitional gender into account (i.e. a brother is always male) but to dismiss stereotypical gender (e.g. that most beauticians are female). Presentation of stimuli and the details of the instructions were varied across the studies, but in all studies participants consistently rejected gender incongruent word pairings (e.g. beautician/brother) more frequently than gender congruent pairings (e.g. beautician/sister). This pattern was still evident (although to a lesser extent) when participants were explicitly provided with a strategy to aid performance with incongruent pairings; they were reminded that nowadays many jobs can are not clearly marked for gender (i.e. have ‘man’, ‘woman’ or ‘ess’ in the title), and that they should carefully consider whether the presented role could be occupied by a man, woman or both (Experiment 4). Oakhill et al. concluded that there is likely an automatic component to OVERCOMING GENDER STEREOTYPES 7 responding, as participants still struggled to suppress the gender stereotype information associated with the role nouns, despite it being counter-productive to task performance. Other strategies have also been proposed to overcome stereotype biases to social/occupational role nouns. These terms are frequently used in sentence comprehension studies using a match/mismatch paradigm in which a stereotyped term is followed by gender congruent or incongruent information. In the former condition processing is typically unproblematic (e.g. the builder went to work although he was not feeling well), whereas difficulty arises in the latter version (e.g. the builder went to work although she was not feeling well). This difficulty in integrating the unexpected gender information into the reader’s representation of the text is often conveyed through slower judgement or reading times relative to the gender matching condition (e.g. Carreiras et al., 1996; Duffy & Keir, 2004; Garnham, Gabriel, Sarrasin, Gygax, & Oakhill, 2012; Garnham et al., 2002; Irmen, 2007; Kennison & Trofe, 2003; Kreiner et al., 2008). However, while these gender stereotype biases can be successfully overcome by establishing the sex of a character before a role noun is encountered (e.g. Duffy & Keir, 2004; Kreiner, et al., 2008; Lassonde & O’Brien, 2013), such an approach is clearly not always practicable. But what about strategies to overcome stereotyping more broadly, outside of the domain of language processing? Social psychologists have typically sought to reduce stereotypes by changing attitudes and behaviour within people’s volitional control using a combination of awareness and effort (Dasgupta & Greenwald, 2001). For instance, this has been achieved by replacing automatic, culturally stereotypic responses with more considered responses that reflect personal beliefs (Devine, 1989; Monteith, 1993; Monteith, Devine, & Zuwerink, 1993) or by encouraging the suppression of negative stereotypes (Macrae, Bodenhausen, Milne, & Jetten, 1994). However, more recently cognitive psychologists have sought to address stereotype reduction by targeting automatic stereotypes using strategies such as stereotype negation/counter-stereotype affirmation training (e.g. Gawronski, Deutsch, Mbirkou, Seibt, & Strack, 2008; Kawakami, Dovidio, Moll, Hermsen, & Russin, 2000), mental imagery (Blair, Ma & Lenton, 2001), or through use of a counter-stereotype OVERCOMING GENDER STEREOTYPES 8 expectancy strategy (Blair & Banaji, 1996). Indeed, methods for overcoming both automatic and controlled stereotypes are of relevance to the current work as the judgement task of Oakhill et al. (2005) involves both. While stereotype activation takes place automatically and stems from increased cognitive accessibility of traits/features connected with a specific group, stereotype application is typically under the conscious control of a perceiver and involves the actual use of stereotypes in response to a group member (Kawakami, Dovidio, & van Kamp, 2005; see also Schneider & Shiffrin (1977) and Shiffrin & Schneider (1977) for a more detailed discussion of automatic and controlled processing). In the present context, participants must first overcome stereotype activation upon being presented with a gender-biased role noun to ultimately make the correct response. Therefore, in using this judgement paradigm of Oakhill and colleagues, the current article is ultimately concerned with achieving a reduction in stereotype application. If participants can learn to overcome spontaneous biases in a laboratory-based training, such interventions could have important implications for text processing and discourse comprehension more broadly.