Abstract
According to the spatial agency bias model, in Western cultures agentic targets are envisaged as facing and acting rightward, in line with writing direction. In four studies of Italian participants, we examined the symbolic association between agency and the rightward direction (Study 1, N = 96), its spontaneous activation when attributing agency to female and male targets (Study 2, N = 80) or when judging the authenticity of photographs of men and women (Study 3, N = 57), and its possible relation to stereotype endorsement (Study 4, N = 80). In Study 4, we used a conditioning paradigm in which participants learned a counterstereotypical new association; we developed a novel measure to assess the association between gender and spatial direction, namely, the spatial association task. Participants envisaged and cognitively processed male and female targets in line with the spatial agency bias model and reported lower benevolent sexism after learning a new counterstereotypical spatial association. Our findings raise awareness about the biased use of space (and its consequences) in the representation of women and men, so that all people, and especially communicators and policy makers, can actively intervene to promote gender equality. Additional online materials for this article are available to PWQ subscribers on PWQ’s website at https://journals-sagepub-com.web.bisu.edu.cn/doi/suppl/10.1177/0361684316676045
Keywords
Sexism comes in different shapes and sizes, from blatant denial of basic rights (such as denial of access to education, voting, driving, or equal pay) up to very subtle, barely perceptible biases (such as the order in which women and men are mentioned, or how they are displayed in artwork). Although such subtle forms of discrimination appear negligible, they may actually be quite influential, because people are rarely aware of their existence (Sojo, Wood, & Genat, 2016). Here we focus on a specific bias, namely, the spatial layout with which women and men are portrayed in images.
Images provide a particularly rapid and immediate means to convey information, and spatial layouts are surprisingly revealing about gender stereotypes. For instance, graphs in science tend to be arranged so that data concerning male samples are positioned to the left of female samples (Hegarty & Buechel, 2006). Also, in Western portrait paintings, female targets are more often portrayed facing leftward than male targets (Suitner & McManus, 2011), presumably because men are stereotypically perceived as more agentic and agency is revealed by the left–right (LR) trajectory (Chatterjee, 2002). Similarly, human interactions are envisaged with stereotypically agentic groups (males) to the left of less agentic groups (e.g., females; Maass, Suitner, Favaretto, & Cignacchi, 2009).
According to the spatial agency bias (SAB) model (Maass & Russo, 2003; Suitner & Maass, 2011; see Suitner & Maass, 2016, for an overview), the right/left asymmetry lies at the intersection of three jointly operating processes. First, it is driven by a culture-specific embodiment process, in which the repeated actions of writing and reading, dictated by script direction, are internalized (Barsalou, 2008; Niedental, Barsalou, Winkelman, Krath-Gruber, & Ric, 2005). This habitual embodiment creates a mental schema for action, in which the writing direction becomes the ideal trajectory of actions. The same process will lead to opposite biases in cultures in which language is written from left to right (e.g., English, Italian, Swedish) and in those in which it is written from right to left (e.g., Hebrew, Arabic, or Urdu).
Second, in most languages (83%), the sentence subject precedes the object, whereas the reverse order is extremely rare (3%; Dryer, 2005). Thus, when communicating who-is-doing-what-to-whom in standard active sentences, the large majority of the world’s population would mention the agent before the patient. As a consequence, people in LR writing cultures tend to envisage the agent to the left of the recipient, with action flowing from LR, agent to recipient, from the observer’s point of view (henceforth the observer’s perspective will be maintained throughout the article). By the same token, action is envisaged with a right–left (RL) trajectory in languages like Arabic that are written from RL and in which the sentence subject precedes the object (Maass & Russo, 2003; for a discussion of script direction and word order, see Maass, Suitner, & Nadhmi, 2014).
Third, stereotypes related to agency map onto these spatial trajectories. According to Abele, Uchronski, Suitner, and Wojciszke (2008), “Agency refers to a person’s striving to be independent, to control one’s environment, and to assert, protect, and expand one’s self. […] A lack of ‘agency’ manifests itself in, for instance, inactivity and apathy” (p. 1204). Individuals or groups that are stereotypically associated with agency (e.g., males, young people) tend to be represented as facing or acting rightward in cultures in which language is written from LR, but as facing or acting leftward in cultures in which language is written from RL (Maass et al., 2009). The existence of a direct link between spatial bias and stereotyping is further sustained by the evidence that (a) spatial bias in drawing tasks is mainly shown by those who strongly endorse traditional gender stereotypes but reverses in those holding counterstereotypical beliefs (Maass et al., 2009); (b) spatial bias in portraits is stronger among male than among female artists (Suitner & Maass, 2007) who, presumably, have more egalitarian gender-role attitudes (Larsen & Long, 1988); and (c) very powerful, hence counterstereotypical, women (e.g., queens) are portrayed much like men (Chatterjee, 2002), facing rightward. Given this extant literature, we argue that spatial asymmetries are used systematically to represent and maintain gender stereotypes.
Overview of Aims of the Present Research
Although there is now growing evidence that people from Western countries envisage agentic targets to the left, acting rightward, its application to the representation of social targets and its relation to stereotype content are still limited to a small number of archival studies (Suitner & McManus, 2011) and even fewer experimental studies (Carnaghi, Piccoli, Brambilla, & Bianchi, 2014; Hegarty, Lemieux, & McQueen, 2010; Maass et al., 2009; Suitner & Maass, 2011). It is important to first establish whether there is a consistent horizontal bias in the representation of social groups differing in agency, specifically women and men (Abele, 2003). Moreover, essential questions regarding the processes and the possible consequences of spatial bias are currently unanswered. First, several scholars (including the current authors) have interpreted the fact that men are envisaged as facing and/or acting rightward, as reflecting the association between agency (a stereotypically male characteristic) and the LR spatial trajectory (Chatterjee, 2002; Suitner & Maass, 2011, 2016). Yet, to date there is no direct evidence that people in Western cultures associate agency distinctly with the rightward trajectory. The first aim of the current research was to test the assumption of SAB, namely, that both at a purely symbolic level (Study 1) and when evaluating specific target individuals (Study 2), people indeed associate agency with the rightward trajectory.
Second, all prior studies have employed visual productions such as artwork or experimental drawing tasks in which participants deliberately sketched agentic or nonagentic targets. No previous research (e.g., Chatterjee, 2002; Maass & Russo, 2003; Suitner & Maass, 2007) has addressed the important question of whether people also use spatial information spontaneously when categorizing women and men or when making decisions about them. We therefore investigated the spontaneous use of spatial information when participants attribute agency to rightward-facing (vs. leftward-facing) individuals represented on photos (Study 2), when they decided whether photographs of women and men appear authentic (Study 3), and when they categorize the gender of a target person (Study 4). We hypothesized that rightward-facing targets would be perceived as more agentic (Study 2) and associated more easily with males (Studies 3 and 4), compared to leftward-facing targets. If these hypotheses were confirmed in a randomized experiment, we may conclude that people spontaneously use spatial cues when processing social information.
The third and most important aim of the present work was to investigate whether a new, counterstereotypical representation of gender (rightward-facing females, leftward-facing males) can be learned in a conditioning paradigm, and whether this new association between gender and space may constitute a source of attitude change. People’s associations are known to be a source of attitude formation (Gawronski & Bodenhausen, 2006), and conditioning paradigms have successfully been used to induce attitude change (Hofmann, De Houwer, Perugini, Baeyens, & Crombez, 2010). Usually in such procedures, the creation of novel associations consists of pairing target stimuli (conditioned stimuli [CSs]) with other stimuli (unconditioned stimuli [USs]) of either positive or negative valence (De Houwer, Thomas, & Baeyens, 2001; Hofmann et al., 2010). Put simply, learning to associate a CS with a “liked” US will improve the attitude toward the CS, whereas an opposite attitude shift will be induced by associations with a “disliked” US. The key characteristic of the US has primarily been its positive or negative valence; however, nonevaluative properties also can be associatively transferred (Houwer, Baeyens, Randell, Eelen, & Meersmans, 2005). We used an US that is characterized by agency, rather than valence, and asked participants to learn a novel counterstereotypical association. According to the SAB model, the LR vector is generally associated with agency and with males, and the strength of this association is linked to people’s gender-role attitudes. We proposed that learning an opposite association between space and gender through counterstereotypic conditioning would affect subsequent expressions of attitudes toward social groups (Parish, Shirazi, & Lambert, 1976; Olson & Fazio, 2006). Specifically, we hypothesized that learning a new asymmetric representation of gender (associating females with the rightward trajectory) would influence people’s self-reported attitudes toward women.
In summary, we pursued three aims: First, we investigated whether agency would be represented with the rightward trajectory at a symbolic level (Study 1). Second, we tested whether people spontaneously used spatial information when attributing agency to people (Study 2), when evaluating the originality of a photo (Study 3), and when categorizing gender-related information (Study 4). Third, we tested whether people could learn new associations between space and gender, and whether these newly conditioned associations affect subsequent gender-role attitudes (Study 4). Our four studies were conducted in Italian and involved Italian native speakers who were unfamiliar with language written from right to left.
Study 1: Symbolic Representation of Agency
In our first study, we investigated how people represented agency versus communion in space when they can choose among four alternatives, namely, leftward, rightward, upward, and downward arrows. Participants indicated which of the four arrows best represent a target person described by either agentic or communal traits. According to various models, stereotypes are composed of two basic, orthogonal dimensions that roughly match (but are not confined to) the male (agentic) versus female (communal) portion of the universe (Bakan, 1966; Bem, 1974; Spence & Helmreich, 1978). In line with Chatterjee’s (2002) hypothesis, we expected participants to deploy the horizontal dimension to depict agency, and specifically to choose the LR trajectory as the preferred spatial direction to represent agentic targets.
It is less clear what to expect participants to do when exposed to descriptions of persons as communal; to date, scant research has investigated spontaneous spatial representation of communal traits. One exception is the study by Nicholls, Clode, Wood, and Wood (1999), where participants were more likely to expose their left (vs. right) cheek when asked to pose for a family portrait than for the Royal Society. The relation between agency and communion is somewhat controversial (Leonard, 1997; for an overview, see Abele & Wojciszke, 2014), with some authors reporting that agency and communion are negatively related due to compensation processes (Judd, James-Hawkins, Yzerbyt, & Kashima, 2005), others finding that they are orthogonal constructs (Abele, 2003; Abele & Wojciszke, 2007), and others that they are positively related (Judd et al., 2005; Rosenberg, Nelson, & Vivekananthan, 1968) because they involve a common evaluative process (Suitner & Maass, 2008). Therefore, it was difficult to hypothesize how participants would spatially represent a communal target. Moreover, given that the leftward direction is not sustained by a congruent motor experience, it is also possible that communion does not map onto space.
Although we were mainly interested in the horizontal dimension, one may also wonder whether and how vertical space is deployed to represent agency or communion. We are unaware of any prior research on this issue; however, we suspect that agency would be represented with an upward direction for various reasons. First, pictorially, “activity” and “energy” are often represented as “jumping,” a clear upward movement, whereas “depression” (characterized by a lack of energy) is often described by downward metaphors (“feeling down,” “falling into depression”; McMullen & Conway, 2002). Second, in the economic realm, a dynamic development of the markets is generally associated with upward graphical representations and linguistic metaphors (White, 2003). Thus, we expected agency to be represented rightward and upward and communion possibly with opposite (or with no specific) trajectories.
Method
Ninety-six Italian participants (56% females, M age = 24.8, SD = 6.62, age range = 19–60) took part in this study. Participation was voluntary and unpaid. Participants were recruited on a university campus by asking whether students would be willing to participate in a study on imaginative abilities. The study was part of the first author’s doctoral thesis (Suitner, 2009).
Participants were asked to fill in a pen-and-paper questionnaire presented as a “study on imagination.” Participants were presented with descriptions of eight target persons, each described by one adjective (e.g., “an active person”) and asked to select one of the four arrows (↑↓→←) that they considered most suitable to represent the person. Although only rightward and leftward directions were of interest for the present work, we included four possible directions in order not to force our participants’ responses; the front-back dimension was not included, as the task was limited to the two-dimensional space involved in writing activity (i.e., the assumed cause of the asymmetry); this choice is also consistent with previous methodologies investigating spatial asymmetries (Dehaene, Bossini, & Giraux, 1993; Meier & Robinson, 2004; Schubert, 2005). Based on extensive pretesting (Suitner & Maass, 2007), we selected four adjectives (aggressive, overbearing, dynamic, and active) rated as high in agency but low in communion and four adjectives (warm, sensitive, shy, and fragile) rated as low in agency but high in communion. The two groups of adjectives were very similar in terms of valence, with two positive and two negative items per group. Also, all adjectives were higher than 20 in use frequency (CoLFIS database; Laudanna, Thornton, Brown, Burani, & Marconi, 1995), none longer than 10 characters, and had been judged by pretest participants to primarily describe human beings (but not inanimate objects). Given eight missing values, the final sample consists of 376 observations for each dimension across all participants.
Results and Discussion of Study 1
Participants more frequently preferred vertical representations (n = 428) over horizontal representations (n = 324); this is in line with the idea that the vertical dimension is spontaneously selected in spatial representations, due to our constant exposure to gravity and hence prevails over other dimensions (Bryant, Tversky, & Franklin, 1992). Critical for our aims, we tested participants’ choices of specific directions to represent agentic versus communal targets. In order to test the relation between agency, valence, and arrow direction, we performed a log-linear analysis on the three-way frequency table. The model that fits our data best and therefore served for the following analyses was the two-way interaction model (likelihood ratio, χ2 = 3.74, df = 3, p = .29, power = .83).
The contingency table of valence and direction (Figure 1) shows that positive adjectives were more strongly associated with rightward and upward arrows than negative ones, whereas negative adjectives were more often associated with leftward and downward than positive arrows, χ2 = 208.86, df = 3, p < .0001. The expected interaction between agency and direction, χ2 = 162.70, df = 3, p < .0001 (Figure 2), shows that targets described with agentic adjectives were more strongly associated with rightward and upward arrows, whereas targets described with communal adjectives were more strongly associated with leftward and downward arrows.
Frequencies and parameter estimates of directions chosen for positive and negative adjectives. Asterisks indicate significant parameter estimates with α = .05.
Frequencies and parameter estimates of directions chosen for communal and agentic adjectives. Asterisks indicate significant parameter estimates with α = .05.
Considering the vertical dimension separately, participants mainly represented agency with an upward direction (78%) and communion with a downward direction (70%). Considering only the horizontal representations, participants mainly represented agency with a rightward direction (83%) and communion without a clear horizontal preference (leftward direction 55%), and the bias for agency remained significant after controlling for valence and alternative vertical biases. Inspecting the estimated standard parameters (Figure 2), which contrast agency and communion, the rightward and upward direction was more strongly associated with agency than with communion, whereas leftward and downward direction showed weaker associations with agency than with communion. In both cases, the association with the vertical dimension was more pronounced, corroborating previous evidence regarding the critical role of the vertical dimension to represent socially relevant information (Giessner, Ryan, Schubert, & van Quaquebeke, 2011; Giessner & Schubert, 2007; McMullen & Conway, 2002; Schubert, 2005). More important, the pattern of the results is in line with the (so far untested) SAB assumption that agency maps onto the horizontal dimension with an LR trajectory.
We also computed for each participant two spatial bias indexes separately for agency and communion, defined as the percentage of rightward (vs. upward) representations over the total of horizontal selections and the percentage of upward (vs. downward) representations over the total of vertical selections. The rightward bias for agency was negatively correlated with the rightward bias for communion, r(62) = −.50, p < .001. Similarly, the upward bias for agency was negatively correlated with the upward bias for communion, r(80) = −.58, p < .001. These correlations are consistent with the negative relation between the agency and communion reported in the literature (Judd et al., 2005; Suitner & Maass, 2008). Moreover, the fact that the biases were weaker for communion than for agency suggests that the bias that emerges in the parameter estimates for communion may be driven by its negative correlation with agency, rather than by an independent spatial association between communion and the leftward vector. This is in line with the SAB-theoretical claim that the two constructs are not opposite poles of a single dimension, rather they represent two separate yet related dimensions grounded in different nonverbal cues (Abele & Wojciszke, 2014). Of note, there is no correlation between the horizontal and vertical bias for agency or communion, suggesting that the biases are independent.
In Study 1, participants associated, at a purely abstract level, different personality characteristics with specific spatial trajectories. Although these personality characteristics are stereotypically linked to males versus females, the task in itself was “gender free,” as it did not require participants to judge women and men. In Study 1, we investigated the spatial representation of the fundamental dimensions of stereotypes but not the effect of the spatial representation on the perception of the members of social categories described by those stereotypes.
Also, in Study 1, we focused on how people exploit spatial coordinates to symbolize socially relevant attributes and interpersonal interactions. However, it was still unclear whether people spontaneously make use of spatial information when making gender-relevant judgments. In Study 2, we investigated the role of spatial cues when attributing agency to male and female targets.
Study 2: Attributing Agency to Rightward Versus Leftward Profiles of Men and Women
Although agency (and its counterpart communion) can, in principle, be used to define the stereotype content of any social group, scholars have most frequently investigated the two dimensions in relation to gender. Men are stereotypically defined by agentic characteristics, such as active and dominant, and women by a lack of agency, such as passive and less assertive; the gender stereotype has been widely confirmed across different models, times, and cultures (Abele, 2003; Altermatt, DeWall, & Leskinen, 2003; Bem, 1974; Diekman, Goodfriend, & Goodwin, 2004; Heilman & Okimoto, 2007). The capability of both women and men to engage in power and agency (Hyde, 2005), and the specific modality with which men and women display power, is not questioned here (Sagrestano, 1992; Yoder & Kahn, 1992). Rather, we focused on the stereotypic representation of gender and its relation with spatial cues. Although both the vertical and the horizontal dimension can be used to spatially distinguish social characteristics (see Study 1), we focus here only on the horizontal dimension that is of prime relevance to our theoretical model. Different from the metaphor-rich vertical dimension representing power and status (e.g., climbing the social ladder, upper status, to stand up, top management; see Schubert, 2005), the horizontal dimension is not reflected in culturally shared metaphors, and hence represents a “pure” instance of cultural embodiment (Ijzerman & Koole, 2011). Here, the involvement of metaphors is not necessary for spatial bias to occur. This is particularly relevant for the theoretical debate concerning the nature of embodied effects, in which explanations based on metaphors (Lakoff & Johnson, 1999), and on semantic activation (Mahon, 2015), are often contrasted with simulation accounts (Barsalou, 1999) involving the motor system, even in the absence of metaphorical language.
Our primary aim in Study 2 was to investigate whether people would make use of spatial information when processing male and female facial profiles. Participants observed right- and left-oriented profiles of two women and two men and were asked to judge their personality. In line with the SAB, we predicted that (a) participants would judge right-oriented profiles as more agentic than left-oriented profiles and, in line with literature on gender stereotypes, we also predicted that (b) participants would judge male targets as more agentic than female targets. We expected these effects to be additive, leading to the higher attribution of agency to male rightward-facing targets.
Method
Eighty Italian participants took part in the study (49% females, 92% right-handers, M age = 25.58, SD age = 10.14, age range = 18–82). Participation was voluntary, unpaid, and participants were recruited on a university campus by asking whether they would be willing to participate in a study on person perception; most of them (88.8%) were students, the remaining participants were workers or unemployed. Participants filled in a questionnaire that introduced the research as a study on person perception. The questionnaire consisted of three parts corresponding to three different tasks, of which only the first will be presented here since the others (concerning spatial bias in political party assignment) are irrelevant to the present study.
Participants observed four profile photos similar to those in Figure 3 and evaluated the targets’ personality traits. Two photos were profiles of women, one turned toward the right, the other toward the left, and two photos were profiles of men, again one right oriented and the other left oriented. The photos were taken from the Karolinska Directed Emotional Faces (KDEF) collection (Lundqvist, Flykt, & Oehman, 1998) that contains close-up photos of the same people taken from different angles.
Example of experimental material of Study 3. In the example, the original profile is the rightward profile for the female target (Karolinska Directed Emotional Faces [KDEF] code: AF02NEHR) and the leftward profile (KDEF code: AM08NEHL) for the male target.
The photos, all with neutral emotional expression, had been pretested for agency. In order to avoid spatial effects in the pretest, we used fully facing photos (whereas the photos included in the main study were rightward- or leftward-oriented photos of the same targets). Twenty-four pretest participants (who were similar to the participants in the current study but did not take part in it) rated how agentic they considered several targets from KDEF collection. This allowed us to select eight targets that, in their fully facing version, were of comparable agency.
In order to control for effects of facial asymmetries (Nicholls, Wolfgang, Clode, & Lindell, 2002) in the final materials, originally left- or right-facing profiles were horizontally flipped to produce mirror images. Half of the participants received the original, half the mirror images. Also, we counterbalanced across participants right and left orientation of each target and order of presentation of the photos, resulting in a total of 16 questionnaire versions, with five participants assigned to each version. Neither counterbalance variable affected the result pattern.
Eighty participants (who did not take part in the pretest) evaluated each of the four photos on 6-point semantic differential scales including four bipolar adjective pairs related to agency (namely, active–passive, dynamic–not dynamic, dominant–submissive, strong–weak, α = .81). The adjectives included in the semantic differential are all words that are commonly used in everyday spoken language and, indeed, are all present in the Lexicon of Spoken Italian (De Mauro, Mancini, Vedovelli, & Voghera, 1993), which contains Italian words that people spontaneously use. We counterbalanced across participants item order and the end points on the bipolar scales (the same adjective occurring either on the right or on the left side of the scale) in order to prevent a pseudoneglect-like effect (Nicholls, Orr, Okubo & Loftus, 2006).
In addition to agency, which was the primary variable in our study, we assessed persuasiveness (5 items: persuasive, convincing, charismatic, credible, and competent, α = .74), communion (3 items: empathic, trustful, and reliable, α = .61), and a single bipolar valence item (positive–negative), none of which was shown to be related to spatial layout and was dropped for further analysis.
Results and Discussion of Study 2
We submitted trait attributions of agency to a 2 (Participant Gender) × 2 (Target Gender) × 2 (Direction: Rightward vs. Leftward) analysis of variance (ANOVA; power = .80) in which the last two variables were within-participants factors. Participant gender did not produce any effects. Replicating studies suggesting that agency is a key attribute in gender stereotyping (Abele, 2003; Eagly & Steffen, 1984), the main effect of target gender, F(1, 78) = 4.19, p = .044,
Study 2 confirmed our predictions: (a) men were perceived as more agentic than women and (b) agency was associated with the rightward trajectory. Of note, participants did not associate femaleness or leftward direction to lack of agency. If agency maps onto the rightward trajectory and if males are stereotypically perceived as more agentic than women, then observers may find profiles representing male targets as more authentic when they are rightward oriented. We addressed this question in the next study.
Study 3: Detecting the Original Profile of Men and Women
Our primary aim in Study 3 was to investigate whether people would make use of spatial information when processing images of males and females and, in particular, when making judgments about the authenticity of an image. In Study 1, we confirmed the predicted association between agency and the rightward spatial orientation, and in Study 2, we found that rightward-facing targets were perceived as more agentic. Given the stereotype associating maleness with agency, we hypothesized that rightward-facing male profiles would be perceived as more authentic and “true.” Previous evidence has linked judgments of truth to familiarity, such that more familiar objects are also considered more likely to be true (e.g., Begg, Anas, & Farinacci, 1992; Zaragoza & Mitchell, 1996). Participants observed pairs of profiles, one being the mirror version of the other. The two profiles were therefore identical, except for their direction, and the participants’ task was to identify the original profile (see Figure 3). In line with the SAB, we predicted that, in the case of male faces, the profile exposing the right cheek (for simplicity, we will refer to as the right-facing profile in contrast to the left-facing profile) should be selected more frequently as the “original” photo. In other words, we predicted that photos of rightward-oriented males would appear more authentic.
Our secondary goal was to exclude the possibility that participants distinguished between original and mirror faces. This is of particular relevance in a social context in which images are the primary vehicle of social communication. In Studies 1 and 2, we have shown that the rightward direction is used to communicate agency, which is considered a key dimension for social judgments (Abele & Wojciszke, 2014). It is extremely easy to modify profile direction by horizontally flipping the image with any image processing software, and authenticity judgments are critical in order to assess whether people are aware of the link between profile direction and social judgment and, most important, whether the effect of profile direction on social judgment lies in the eye of the observer or in the stimulus. The use of mirror-reversed images is not uncommon in the field (Benjafield & Segalowitz, 1993). For example, Zaidel and Fitzgerald (1994) found that the original rightward profiles of female portrait sitters were overall liked more than original leftward profiles, whereas the preferences were in the opposite direction for mirror images; the question remains whether the direction of the sitter’s face profile, or other stylistic features selected by the portrait painters, was responsible for the differences in preferences between rightward and leftward faces. It was therefore important to test whether participants were able to distinguish between original and mirror faces. Authors of previous studies have suggested that the left hemiface is more expressive than the right hemiface (Wolff, 1933). Thus, observers may be able to correctly guess which of the two identical photos is the “original” one, by choosing the left-facing photo in the case of more expressive and the right-facing photo in the case of less expressive faces.
We predicted that participants would be unable to distinguish original from mirror images, but that they would make use of spatial information to make their guesses, choosing the rightward-oriented profile as authentic but only for male targets, which are stereotypically associated with agency. Given the results of Study 2, we hypothesized that participants would more easily associate male (but not female) targets with the rightward trajectory. Finally, we wanted to test whether the explicit gender stereotype associating males with greater agency would be associated with greater spatial bias.
Method
Fifty-seven Italian participants volunteered for the study (73% females, 93% right-handed, M age = 24.77, SD age = 7.18, age range = 20–68). We replaced one participant because he also spoke Hebrew, a language written and read from RL, and we removed another participant since he did not complete the questionnaire, with a final sample size of 56 participants. Participants were recruited on a university campus by asking whether they would be willing to participate in a study on the ability to distinguish authentic versus nonauthentic images; most of them were students (two participants did not provide occupational information, one participant did not provide age information). Participation was voluntary and unpaid. The study was part of the first author’s doctoral thesis (Suitner, 2009).
Eight male and eight female KDEF targets (Lundqvist et al., 1998) matched for attractiveness on the basis of a pretest using fully facing photos (participants in the pretest were university students otherwise not affiliated with the current study; n = 12). The left- and right-oriented profiles of the pretested targets were used in the study, together with their respective mirror images. Therefore, for each target, we obtained two left-oriented (original left-facing and mirror image of right-facing photo) and two right-oriented photos (original right-facing and mirror image of left-facing photo). Participants were presented with pairs of images (original and mirror image) that were identical except for direction (see Figure 3). Participants indicated which profile of each pair was the original one. Values could range from 0 (i.e., no rightward photo was ever selected as the “original”) to 16 (i.e., all rightward photos were consistently selected as the “original”) for the entire stimulus set and from 0 (i.e., no rightward profile indicated as the original) to 8 (i.e., all rightward profiles indicated as originals) when only male or only female targets were considered. We counterbalanced across participants’ presentation order of male and female targets, rightward and leftward profiles as original and mirror images, and position of the two face profiles in each pair (inward vs. outward facing). We constructed the eight questionnaire versions, so that we presented each target photo only once to each participant and we randomly assigned participants to the eight questionnaire versions. To assure that males were indeed perceived as more agentic than females, participants rated how much a standard definition of agency (Abele, Uchronski, Suitner, & Wojciszke, 2008) described the personality of each target person from 1 (not at all) to 4 (completely) in a subsequent task.
Results and Discussion of Study 3
In line with Study 2 and with previous studies (Abele, 2003; Eagly & Steffen, 1984), participants perceived male targets, M = 2.37, SD = 0.50, which is higher than the scale midpoint, one-sample t(55) = 5.56, test value = 2, p < .001, 95% CI [0.24, 0.51], as more agentic than female targets, M = 2.13, SD = 0.51, which does not differ from the scale midpoint, one-sample t(55) = 1.67, test value = 2, p < .10, 95% CI [−0.02, 0.23], paired sample t(55) = 3.40, p = .001, 95% CI [0.06, 0.39], Cohen’s d = 0.87.
We examined whether people chose the rightward profile as the “original,” more frequently for male targets. We calculated how often each participant chose the rightward profile as the original. Overall, participants showed a small preference for rightward profiles, which they selected as the original in 8.64 (54%) cases, one-sample t(55) = 1.79, p = .08, Cohen’s d = 0.25, power = .96. Thus, participants showed a borderline gender-unspecific preference for the right-facing profile, but this was neither strong nor significant. More important, we had predicted that the perception of rightward profiles as more authentic would only be observed for male targets, assuming that males (but not females) are stereotypically associated with agency. Looking separately at male and female targets, participants indicated the rightward profile of male targets as the original (57%), M = 4.52, SD = 1.60, one-sample t(55) = 2.43, p = .018, Cohen’s d = 0.32, 95% CI [1.13, 11.81], power = .99, whereas the bias was absent when the target was female (52%), M = 4.13, SD = 1.58, one-sample t(55) < 1, 95% CI [−3.74, 6.87]. The hypothesized difference between male and female targets was marginally significant in a paired t-test, t(55) = 1.88, p = .088, Cohen’s d = 0.15, 95% CI [−0.75, 10.57], power = .56. The expected correlation between the explicit attribution of agency to males in general and the selection of the rightward profile as more authentic was not confirmed (within-subjects r = −.04). This suggests that the spatial bias and the explicit attribution of agency may be grounded in two separate processes, differing in the degree of awareness, and therefore differently affected by response biases and introspection (Gawronski & Bodenhausen, 2006). This conjecture is based on the observation that the spatial bias reflects a spontaneous and arguably unconscious attitude (which is confirmed by the finding that none of our participants guessed the scope of our studies), whereas the attribution of agency is explicit.
Turning to the question of whether participants were able to differentiate original and mirror face profiles, they correctly identified 8.04 photos as original, about 50%, suggesting that answers were random. We believe that the spatial bias found in our and previous research is unlikely to be driven by different degrees of expressivity of the left versus right hemiface. This interpretation is in line with Brady’s (2004) argument that the bias lies in the perceiver and not in the target.
It is well established that people are able to distinguish between original and mirror versions of familiar faces and that this ability improves when asymmetric details (such as hair parting) are available (Tomita & Onodera, 1994); yet, the ability to recognize the original version of an unknown face is still underinvestigated. Although participants in our study were unable to distinguish original from mirror photos, they did tend to differentiate between rightward and leftward profiles, wrongly identifying the rightward profile as the original image more often than would be expected by chance. In line with our hypotheses, this was true only for male targets, whereas rightward or the leftward profiles of females were chosen at random. Although the dependent variable used here is, to our knowledge, different from any prior study, this result is conceptually in line with analyses of artwork, in which male sitters are more frequently portrayed with a rightward profile than female sitters (Suitner & McManus, 2011). However, in our data, there was no reverse bias for female targets; we therefore concluded that agency is more strongly associated with the rightward trajectory than communion is associated with the leftward trajectory. We suggest that differential degrees of embodiment may explain this stronger spatial representation of agency than communion. That is, while agency coincides with the (LR) activity of writing, there is no complementary (RL) enactment of movement corresponding to communion.
In Study 3, we confirmed that the additive attribution of agency to the rightward trajectory, and to male targets observed in Study 2, also affects participants’ judgments of authenticity, with rightward profiles being perceived as more authentic, but only when representing male targets. Thus, there appears to be an association between maleness and the rightward spatial direction. To further examine this association, in Study 4, we tested whether it is easier to process male profiles that are facing rightward (since these profiles imply agency, in line with the stereotype) and female profiles that are facing leftward (since these profiles imply low agency, again, in line with the stereotype).
Study 4: Gender Categorization and Stereotype Endorsement
In Study 4, we developed a task we called the spatial association task to measure the implicit association between gender and profile direction. Participants were asked to perform a label-matching task in which they had to indicate whether a picture (the profile of a female or male target) matched a previously presented label (i.e., “male” or “female,” in Italian “uomo” or “donna”). We expected a spatial bias, such that, when under time pressure, people would categorize male (but not female) targets more accurately when profiles were rightward oriented. We hypothesized that participants would use spatial information even when completely irrelevant to the task at hand.
Moreover, in Study 4, we tested whether people could learn a new association between gender and space and whether learning a new association would reduce sexism. Our hypothesis was based on the assumption that attitudes reflect patterns of activation that involve both the activation of previous associative paths (in our case, rightward direction, agency, and maleness) and input of contingent novel associations (in our case, rightward direction, agency, and femaleness) that can be experimentally induced with a conditioning paradigm (Gawronski & Bodenhausen, 2006).
There is ample evidence that the exposure to positive exemplars of a stigmatized social category reduces the automatic association between that social category and negativity, and eventually such associations also influence subsequent beliefs. For example, being exposed to female leaders enhances the associations between women and leadership qualities in an Implicit Association Test (IAT) measure (Dasgupta & Asgari, 2004). Similarly, being exposed to admired African American persons reduces the automatic associations of Blacks and negativity in an IAT measure (Dasgupta & Greenwald, 2001). Ramasubramanian (2011) briefly exposed participants to stereotypical versus counterstereotypical African American media characters and reported that the exposure to stereotypical out-group exemplars enhanced stereotype endorsement and prejudice against the entire out-group. Thomas Parish and his collaborators used conditioning paradigms in a series of studies in the 70s, confirming that both a positive (Parish & Fleetwood, 1975) and a neutral (Parish, Fleetwood, & Lentz, 1975) US can reduce prejudice against African American people. They also found that the effect of conditioning on subsequent social attitudes increased as the number of conditioning trials increased (Parish & Fetwood, 1975). More recently, Olson and Fazio (2006) reported that a conditioning paradigm effectively improved White participants’ implicit racial attitude toward Black people even after a delay of 2 days following the manipulation, but the effect was not present on explicit measures of racism. Previous literature has shown the importance of exposing people to positive exemplars and the importance of inducing positive associations in order to reduce negative attitudes toward a stigmatized group. We took this line of reasoning one step further in Study 4; we tested whether spatial information can work as US in counterstereotypic conditioning.
We developed a conditioning paradigm in which different associations were created, by modifying the exposure ratio to rightward (vs. leftward) male and female targets, with gender (CS) being conditioned by agentic versus nonagentic spatial direction (US). In the control condition, the number of rightward and leftward profiles was equal across male and female targets. In the counterstereotypical condition, the majority of male targets faced left and the majority of female targets faced right. Given that non-stereotypic associations are best learned in training procedures affirming a counterstereotypic association, rather than in unbiased associations as in our control condition (Gawronski, Deutsch, Mbirkou, Seibt, & Strack, 2008), we expected that only the counterstereotypic condition would promote a counterstereotypic association between femaleness and rightward orientation and that this novel association would be reflected in subsequent lower scores on a measure of sexism.
We were primarily interested in benevolent sexism, which is often mistaken as a nonsexist, protective attitude toward women; it is widely endorsed by both women and men around the world (Glick & Fiske, 2001). Benevolent sexism scores reflect the view that women are weaker than men; the view is often used to rationalize gender inequality in society (Jost & Kay, 2005). Benevolent sexism pertains to the paternalistic component of sexism that assigns women a passive and men an active role, describing a world in which women are protected and rescued by males (e.g., “In a disaster, women ought to be rescued before men”; “Women should be cherished and protected by men”). This definition of gender stereotypes maps onto the rationale of the SAB model, in which dynamic agents are portrayed rightward and passive patients leftward. This is quite different from hostile sexism in which women are envisaged as actively struggling against gender discrimination (e.g., “Women seek power by gaining control over men”; “Once a man commits, she puts him on a tight leash”) and hence as displaying a high degree of agency, typically ascribed to feminist women (DeWall, Altermatt, & Thompson, 2005). We expected a link between spatial associations and benevolent sexism such that the more people associate males with the rightward vector and females with the leftward vector, the more they will subsequently endorse benevolent sexist attitudes (that assign women a passive role) but not hostile attitudes (that assign women an active role). We also examined whether those who have successfully learned a new, counterstereotypical association between space and gender would show decreased benevolent (but not hostile) sexism.
The Spatial Association Task
To develop the spatial association task, faces of 10 female actors (F07NEH, F14NEH, F15NEH, F22NEH, F24NEH, F26NEH, F29NEH, F30NEH, F31NEH, and F32NEH) and 10 male actors (M08NEH, M10NEH, M12NEH, M21NEH, M26NEH, M28NEH, M29NEH, M30NEH, M31NEH, and M34NEH), used with permission from the KDEF collection (Lundqvist et al., 1998), were randomly selected and pretested (participants were university students otherwise not involved in the current study, n = 14), using full-facing photos that were equally attractive and dynamic. For each actor, four images were used, all with neutral facial expression: originally left facing, originally right facing, and their respective mirror images (similar to Figure 3).
The same participants were instructed to press a computer space bar (“go”) only if an image matched the cue word (“male” or “female”; in Italian “uomo” or “donna”) preceding the image but to wait (“no-go”) if the word did not match the image. The sequence of fixation point, cue word, and image (photo of male or female face) and the respective time windows are explained in Figure 4. A blank wait window remained available for 1,000 ms for participants to provide their response before the subsequent fixation point appeared. Each picture appeared twice (once as a “go” and once as a “no-go” trial). Presentation order was randomly determined, constrained by the following fixed criteria: Each image was maximally distant from its mirror image and from the picture of the other profile of the same actor. Each participant had a practice block of 8 trials and an experimental block of 80 trials. Given the nature of the task (go/no-go task), we evaluated accuracy using d′ scores (Nosek & Banaji, 2001; Stanislaw & Todorov, 1999), defined as the difference between the percentage of go answers when the correct response is go (hits) and the percentage of go answers when the correct response is no-go (false alarms). We adjusted extreme rates in the ratio of hits and false alarms following Banaji and Greenwald (1995; see also http://projectimplicit.net/nosek/gnat/). The higher was the d′ value, the more accurate were the participants’ responses.
Experimental procedure of Study 4. In Study 4, participants were asked to press the key bar if, and only if, the gender of the target in the figure matched the previously presented word. The word could be either UOMO (i.e., man) or DONNA (i.e., woman). The image includes (from top to bottom) the Karolinska Directed Emotional Faces stimuli AF30NEHR, AM12NEHR (mirror reversed), and AM08NEHR as examples.
In order to obtain an overall index of stereotype-congruent spatial association (ScSA), we subtracted the accuracy of leftward-oriented male and of rightward-oriented female targets from that of rightward-oriented male and of leftward-oriented female targets. Higher scores in ScSA indicate that males were associated with rightward orientations and females with leftward orientations. The ScSA index simultaneously contrasts men versus women and rightward versus leftward orientation; hence, in line with the SAB model, it describes, in a single index, the relative spatial positioning of men and women.
Method of Pilot Study
Since this was the first time that the SAB was assessed with a measure of implicit association and the first attempt to experimentally influence sexism by varying the exposure to rightward (vs. leftward) male and female profiles, we first conducted a pilot study to verify the effectiveness of the spatial association task and its sensitivity to the manipulations. Three conditions were created for this pilot study: a stereotypical condition, in which the majority of male profiles were rightward oriented and the majority of female profiles leftward oriented; a control condition, in which an equal number of rightward and leftward profiles were included for male and female targets; and a counterstereotypic condition, in which the majority of male profiles were leftward oriented and the majority of female profiles rightward oriented.
Fifty-four Italian university students (87% females, 96% right-handed, M age = 20, SD age = 1.86, age range = 18–28) volunteered in exchange for course credit. They were all undergraduate students attending an introductory social psychology course. None of the students knew any language written from LR. Because of an error in recording, the data of one participant were not available. The study was part of the first author’s doctoral thesis (Suitner, 2009).
Before (t1) and after (t2) administration of the spatial association task, participants completed the Italian versions of the Ambivalent Sexism Inventory (ASI) and the Ambivalence Toward Men Inventory (AMI; Glick & Fiske, 1996 , 1999; Manganelli Rattazzi, Volpato, & Canova, 2008) and attributed the 16 adjectives (8 agentic, 8 communal) used in Study 1 to women and men in general on two separate 1–7 Likert-type scales. Given that AMI and ASI tap a consistent attitude toward traditional gender roles (Glick et al., 2004), we created a benevolent (from now on referred to as benevolent sexism) and hostile (from now on referred to as hostile sexism) index by averaging all the items of the Italian versions of the ASI and AMI measuring benevolence (αt1 = .85, αt2 = .82) and all items measuring hostility (αt1 = .77, αt2 = .76). Participants were then randomly assigned to three versions of the spatial association task: the control condition (N = 18) in which male and female profiles were half leftward oriented and half rightward oriented, the stereotypic condition (N = 18) in which 80% of male profiles were rightward oriented and 80% of female profiles leftward oriented, and the counterstereotypic condition (N = 17) in which 80% of male profiles were leftward oriented and 80% of female profiles rightward oriented.
Results and Discussion of Pilot Study
We first examined whether participants learned a counterstereotypical association when repeatedly exposed to leftward-oriented males and rightward-oriented females. Two outlier participants, with a performance of more than two standard deviations below the mean, were excluded from the subsequent analyses, leaving 51 participants who participated in the pilot study.
To test whether the conditioning paradigm was effective, we conducted an ANOVA (power = .34) with the ScSA index as dependent variable. As expected, the ScSA index varied as a function of experimental manipulation, F(2, 48) = 8.64, p < .001,
The four components of the ScSA index were positively correlated with each other (average r = .50) and reflect the overall performance on the task. The highest correlation emerged between rightward-facing male and leftward-facing female targets (r = .74, p < .001), confirming the comparative nature of the spatial bias.
To test whether this (already established or newly learned) spatial association affects subsequent attitudes, we performed an ANCOVA with benevolent sexism at t2 as a dependent variable, experimental condition as predictor, and benevolent sexism at t1 as a covariate (power = .34). Although, at the descriptive level, sexism was lower in the counterstereotypic condition (M = 3.32, SD = 0.74) than in the control (M = 3.54, SD = 0.84) and stereotypic conditions (M = 3.55, SD = 0.40), this difference was small and statistically nonsignificant. However, the correlation between the ScSA index and the t2 sexism scores (controlling for t1 sexism) confirms our idea that spatial associations and attitudes are related. As expected, only benevolent, but not hostile, sexism was related to spatial association, r = .32, p = .023. Despite the high correlation between benevolent sexism at t1 and t2, r = .74, p < .001, only the t2 measure was related to ScSA, with no correlation between t1 benevolent sexism and ScSA (r = .08). Although means of benevolent sexism were in the expected direction (M stereotypic = 3.63, SD = 0.16; M control = 3.45, SD = 0.17; M counterstereotypical = 3.32, SD = 0.17), the experimental condition did not have a significant effect on benevolent sexism (Fs < 1).
A path analysis testing for the indirect effect of experimental condition on benevolent sexism is reported in the Online Supplemental Material together with additional information about the study methodology. The same analyses were performed using hostile sexism, attribution of agency, and attribution of communion as dependent variables; all were unrelated to the experimental condition and to the ScSA index.
The pilot study provided tentative support for the hypothesis that learning a specific spatial representation of social targets affects the subsequent attitude toward the targets; this hypothesis was tested in greater detail and on a larger sample in the main study reported below.
Method of Study 4
In the main study, we slightly modified the methodology of the pilot study. First, given the small effect sizes of the pilot study (and low power observed), we doubled the sample size, and we evened the distribution of male and female participants across conditions. Given that there was no difference between control and stereotypic condition in the pilot study, we administered in the main experiment only the control and counterstereotypic conditions. We also assessed, in an exploratory way, paternalistic chivalry (Viki, Abrams, & Hutchison, 2003) in order to further verify whether the relation between spatial bias and sexism was specific to benevolence. Similar to benevolent sexism, paternalistic chivalry implies a seemingly positive and protective attitude toward women but confines them to a traditionally passive role (e.g., “It is up to the man to make sure a woman enjoys herself during a date”; “It is inappropriate for a woman to make sexual advances toward a man”). The Paternalistic Chivalry Scale is more limited in scope than the Benevolent Sexism Scale, as it focuses specifically on courtship behavior. As a consequence, it only has a moderate correlation with benevolent sexism (Viki et al., 2003).
Eighty Italian participants (50% females, 90% right-handed, M age = 22.78, SD age = 2.87, age range = 19–35), who had not participated in the pilot study, volunteered for the study. Participants were recruited on a university campus by asking whether they would like to participate in a study on face perception; the majority of them (96%) were students and did not receive remuneration for their participation in the study. None of them knew any language written from LR.
As in the pilot, all measures were administered in Italian. Immediately before (t1) and after (t2) the computer task, participants completed the ASI and AMI. Given the short time span between t1 and t2, we varied the order of the questions in the two administrations. In the pilot, we measured the key dimension (benevolent sexism) with a pool of questions taken from two inventories (ASI and AMI). Given that ASI and AMI contain redundant questions, it was reasonable to divide the items between t1 and t2 by matching the meaning of the questions (e.g., ASI 5 at t1—“Every woman needs a male partner who will cherish her” and AMI 16 at t2—“Women are incomplete without men”). We applied the same logic to all the dimensions. As a result, the items of pre- and posttest were very similar in meaning but were never repeated in the exact same form. The final questionnaire at t1 included 11 benevolent sexism items (Items 12, 8, 6, 22, 17 of ASI and Items 1, 3, 7, 10, 5, 11, 13 of AMI; α = .78), 10 hostile sexism items (Items 10, 4, 2, 7, 14 of ASI and Items 2, 4, 8, 11, 14 of AMI; α = .81), and 8 paternalistic chivalry items (Items 1, 16, 5, 2, 4, 8, 11, 3; α = .74). The questionnaire at t2 contained 10 benevolent sexism items (Items 1, 3, 9, 13, 19, 20 of ASI and Items 12, 16, 18, 20 of AMI; α = .87), 10 hostile sexism items (Items 15, 16, 18, 21 of ASI and Items 6, 9, 15, 17, 19 of AMI; α = .82), and 8 paternalistic chivalry items (Items 6, 7, 9, 10, 12, 13, 14, 15; α = .82). All scales were consistent between t1 and t2 (all rs > .74). Given that both paternalistic chivalry and hostile sexism were statistically unrelated to the performance on the spatial association task and were not affected by the experimental conditions, we will not discuss them in the Results section.
We randomly assigned participants to two versions of the spatial association task: the control condition (n = 40) in which 50% of male and 50% of female profiles were rightward oriented and the counterstereotypic condition (n = 40) in which 80% of male profiles were leftward oriented and 80% of female profiles rightward oriented.
Results and Discussion of Study 4
We computed the ScSA index as in the pilot study. We excluded two participants from the subsequent analyses who had a performance (measured as number of correct answers) more than two standard deviations below the mean. We conducted an ANOVA (power = .60) with the ScSA index as dependent variable and experimental condition and participant gender as factors varying between subjects. As expected, the ScSA index varied as a function of experimental manipulation, F(1, 76) = 6.14, p = .015,
Again the four components of the ScSA index were positively correlated (average r = .69), and the highest correlation emerged between rightward-facing male and leftward-facing female targets (r=.74, p < .001), supporting the comparative nature of the spatial bias. We also computed one-sample t-tests, comparing the group means to zero (absence of bias), to assess the presence or absence of bias in each condition. Participants’ performance showed a reliable ScSA in the control condition, t(39) = 2.76, p = .009, 95% CI [1.92, 12.43], power = .85, suggesting that men were associated with the rightward direction. In the counterstereotypical condition, the spatial association was attenuated but not significantly reversed, t(39) = −1.03, p = .31, 95% CI [−10.45, 3.39]. In order to investigate whether the bias was driven mainly by male targets, we also computed separate ScSA scores for male and female targets. In the control condition, the bias was mainly driven by male, M = 5.49, SD = 10.36, t(39) = 3.36, p = .002, 95% CI [2.18, 8.81], d = 1.08, power = .99, rather than by female targets, M = 1.50, SD = 13.87, t(39) = 0.68 p = .5, 95% CI [−2.94, 5.94], and this was in line with the results of Study 2. In the counterstereotypical condition, the manipulation induced a counterstereotypic bias for female targets, M = −5.00, SD = 15.52, t(39) = −2.39, p = .048, 95% CI [−9.96, −0.04], d = 0.76, power = .99, whereas the bias for male targets was removed but not reversed, M = 0.62, SD = 13.63, t(39) = 0.29, p = .77, 95% CI [−3.73, 4.98]. Thus, the categorization of both male and female targets was sensitive to the experimental manipulation. The interaction between targets’ gender and experimental condition was not reliable (F = .14, p = .70), hence the experimental condition modified the spatial association for male and female targets in a similar fashion. Note, however, we manipulated within participants both target gender and profile direction; this methodological procedure, though highly ecological, as it mirrors natural social environments, may have strengthened gender differences in spatial associations compared to between participants designs.
Our results corroborate the preliminary data of the pilot study, namely, that the spatial association task could be used to (a) detect spatial associations and (b) modify associations between gender and profile directions. In the balanced situation, participants responded more accurately when categorizing rightward-oriented male and leftward-oriented female profiles, providing evidence for a spontaneous use of the SAB. However, a counterstereotypical layout modified the spontaneous spatial associations, by inducing a rightward bias toward female targets (which in the control condition was absent) and eliminating the rightward bias for male targets. Thus, the experimentally induced spatial associations modified existing associations (namely, the rightward bias for male targets and the absence of bias for female targets that emerged in the control condition). Thus, the SAB, which researchers have linked to sensory motor experiences that are repeatedly practiced over time and that are extremely consistent, may—at least temporarily—be overridden by contingent alternative associations (for similar conditioning effects concerning the association between power of groups and font size, see Schubert, Waldzus, & Seibt, 2009).
To test whether the (already established or newly learned) spatial association affects subsequent expressions of attitudes, we performed an ANCOVA with benevolent sexism at t2 as the dependent variable, experimental condition and participant gender as predictors, and benevolent sexism at t1 as covariate (power = .60). A main effect of experimental condition, F(1, 73) = 4.75, p = .032,
Different from the pilot study, the experimental condition in the main study exerted a direct, albeit small, effect on the expression of benevolent sexism. Thus, participants’ responses were consistent with the hypothesized spatial conditioning effect, namely, that the exposure to counterstereotypical spatial layouts reduced postexperimental self-reported sexism (after controlling for preexperimental sexism). The outcome is similar to the pilot study, but the effect is indirect and driven by ScSA, such that shifts in sexism are conditioned by the degree to which participants learn the new spatial association. Thus, both studies provide initial evidence that the SAB is sensitive to the exposure to spatially biased images of female and male targets and that the expression of benevolent sexism is affected by this exposure. Whether (or when) the influence of biased spatial representations of gender on sexism is direct (Study 4) or indirect (Pilot Study) should be further investigated in future studies.
General Discussion
Four main conclusions can be drawn from the above studies examining the relation between spatial layout and gender stereotypes. First, the findings corroborate and extend previous evidence according to which people appear to use space in a consistent way to represent agency (Maass et al., 2009). Through Study 1, we have provided the first evidence that people symbolically represent agency with a rightward trajectory, supporting the idea that abstract concepts are linked to physical modalities even in the absence of linguistic metaphors that would promote such associations (Barsalou, 2008; Casasanto & Boroditsky, 2008). This does by no means exclude the possibility that the same trajectory also serves to represent related concepts such as importance, relevance, priority, or status. Given the great number of abstract concepts that populate the human mind, they necessarily need to share the few spatial coordinates accessible to our senses, as in the case of the vertical dimension that accommodates diverse bipolar concepts such as good–evil, God–devil, powerful–powerless, conscious–unconscious, wealthy–poor, more–less, high–low status, and happy–depressed (Lakoff & Johnson, 1980; Meier & Robinson, 2004; Schubert, 2005). Similarly, the horizontal dimension was used in our first study to represent both agency (vs. communion) and positive (vs. negative) valence, despite the fact that agency and valence were not confounded in any way.
Across studies, horizontal spatial mapping was specific to agency and much lower (Study 1) or even absent (Studies 2 and 3) for communion. This finding is in line with the SAB model, according to which, agency is expected to mirror the motor activity of writing (Maass, Pagani, & Berta, 2007; Maass & Russo, 2003; Maass et al., 2009; Maass, Suitner, & Nadhmi, 2014), whereas communion is not anchored in any specific horizontal motor action but simply derives from the fact that agency and communion are often seen as opposites. This interpretation is sustained by the negative correlations between agency and communion in Study 1, in which participants who envisaged agency with a rightward (or upward) trajectory also tended to envisage communion with a leftward (or upward) trajectory. In a sense, then, communality appears to be an epiphenomenon of the SAB rather than a distinct and complementary bias.
A second conclusion that may be drawn from our results is that spatial information is used spontaneously, which is in line with previous studies (e.g., Maass & Russo, 2003). The novel aspect is that this automatic use of space emerges also when processing social information. Participants associated the rightward direction with greater agency (Studies 1 and 2), perceived males as more agentic than women (Study 2), and—in a complementary fashion—perceived rightward-oriented depictions of men as more authentic (Study 3). Together, these findings are consistent with the idea that people automatically use spatial information in processing social category membership even when such information is irrelevant to the task and when more diagnostic information (e.g., facial features, hairstyle) is available. Future research is needed to investigate whether these findings extend to social groups other than gender, defined for example by status, power, or age differences.
Third, this is the first research to support the claim that people may acquire new space-gender associations when repeatedly exposed to images in which females are rightward oriented. Whereas previous research had mainly emphasized the distal causes of spatial bias, namely, script direction (Vaid, 2011), we have provided the first evidence that contextual information also plays a role in creating or modifying spatial schemata, demonstrating the potential malleability of such mental representations. This finding is conceptually in line with Casasanto and Bottini’s (2014) observation that people who engage in novel embodied processes (e.g., mirror reading) are likely to learn new associations between space and abstract concepts (in the case of Casasanto and Bottini’s study, time rather than agency).
Fourth, and most important, we found that the conditioned acquisition of gender-space associations affects people’s attitudes; participants expressed reduced or sexism after learning to associate women (rather than men) with the rightward direction. According to previous studies, counterstereotypical conditioning is a strategy to modify both recently acquired associations (Kerkhof, Vansteenwegen, Baeyens, & Hermans, 2015) and preexisting attitudes such as racial prejudice (Kawakami, Dovidio, Moll, Hermsen, & Russin, 2000; Oslon & Fazio, 2006). This is the first time that counterstereotypical conditioning was implemented by varying the exposure to different spatial layouts of photographs of male and female targets.
Although spatial and social cognition have traditionally been conceived of as unrelated abilities, a steadily increasing body of literature sheds doubts on this assumption. At a cultural level, the two areas of cognition are connected through metaphors that frame abstract interpersonal relations, such as interpersonal closeness (Williams & Bargh, 2008) or power differentials (Schubert, 2005), in terms of concrete spatial relations. At the cortical level, the two areas of cognition have been found to recruit overlapping neural substrates (Yamakawa, Kanai, Matsumura, & Naito, 2009), possibly reflecting an exaptation process by which evolutionarily older (spatial) circuitries are deployed for novel (social) purposes (Parkinson & Wheatley, 2013). The two approaches concur in showing that spatial and social cognition are closely and asymmetrically interconnected. Here, we take these arguments one step further by advancing the hypothesis of a mutual influence of spatial and social cognition. On the one side, we propose that abstract social cognitive concepts, such as gender stereotypes related to agency versus communion, map onto spatial coordinates and that they do so even in the absence of linguistic metaphors. On the other side, we argue that these spatial coordinates can be used to modify gender stereotypes.
Our findings show that spatially biased portrayals of men and women in the media, in picture books, or in the visual arts may in a subtle way contribute not only to the maintenance of gender stereotypes but potentially also to their change. A great effort has been made to prove that women are portrayed in a discriminatory way; however, scholars have mainly focused on the sexualized visual representation of women (e.g., Lanis & Covell, 1995), whereas little is known about more hidden ways to promote stereotypic gender roles. One exception is the literature on faceism (Archer, Iritani, Kimes, & Barrios, 1983) and head canting (Costa, Menzani, & Ricci Bitti, 2001); researchers report that men are more likely portrayed with a focus on their full face (Archer et al., 1983) and in straight positions (Kendon & Ferber, 1973), whereas women are more likely portrayed showing their body and with an inclined head position (see also Kang, 1997; Lindner, 2004). These biases are critical since the full face communicates higher social status and competence (Schwarz & Kurz, 1989), and an inclined position reduces their attributed power (Egolf & Corder, 1991; Judge & Cable, 2004). We propose a different and possibly even subtler bias is portrayed in the visual representation of women and men. We propose that spatial layout is a hidden tool to corroborate or disconfirm the status quo of gender differences in society. In acknowledging that spatial features may be currently used with important consequences, we contribute to the creation of a literacy that can enrich the semiotic understanding of visual communication of gender roles. At the same time, we offer a tool for changing gender attitudes without arousing resistance or counterargumentation.
Limitations
Besides the limitations that we have already identified, there is one critical limitation of this work, namely, that it does not include evidence about the process behind the spatial bias (e.g., Is it attentional in nature?) nor its origin (e.g., Is it related to writing habits?). There are also methodological limitations, related to the reliability of the measures as well as the ecological validity of the experimental designs. Do such spatial biases emerge also in contexts rich of additional information besides face profile direction? An additional important limitation is that it involves participants who are relatively homogenous in terms of age range and type of population (mainly students; all Italian). Future studies are needed to overcome these limits and identify the boundary conditions of the reported findings.
Practice Implications
We believe that the present findings have practice implications for different domains. First, professionals in the publishing industry have become increasingly sensitized to the detrimental effects of gender-stereotypical portrayals of women and men, yet they generally pay little attention to the spatial arrangement of images of males and females. Space is not only telling about people’s attitudes (Schoel, Zimmer, & Stahlberg, 2015) but, as we show here, it also contributes to shaping people’s attitudes. Avoiding stereotypical representations of rightward-oriented males and leftward-oriented females may be particularly important in the case of children’s books, given that gender stereotypes are learned and internalized very early in life (e.g., Coyne, Linder, Rasmussen, Nelson, & Birkbeck, 2016). Based on the last study reported above, we propose that authors and publishers use counterstereotypical spatial arrangements to modify existing gender stereotypes. Associating women systematically with the more agentic rightward orientation may help reduce traditional gender stereotypes. This strategy challenges the passive role subtly assigned to women in Western society, without provoking motivated reactions against the message. Thus, compared to more explicit messages, presenting women in a spatially agentic way may, over time, lead to attitude change without arousing resistance, discounting, or counterargumentation. Future research should address this issue. Whether this strategy may be useful as well to reduce prejudice toward other social groups remains an additional open question at the moment.
Second, nondiscrimination is part of the mission of many companies and organizations, including universities, and this is often communicated through corporate websites. For instance, many companies seek to display gender- and race-balanced photographic materials to communicate a nondiscriminatory attitude and an inclusive organizational climate. Given that members of minority groups are highly sensitive to such cues of inclusion versus exclusion (Emerson & Murphy, 2014), we suggest that spatial layouts of photographs representing majority and minority members ought to become part of web design strategy. Thus, web designers should place women in the agentic position, on top and oriented toward the right, at least with the same frequency with which males are portrayed in this position. Third, a similar argument can be made for recruitment websites and mobile recruiting apps. An organization communicates whether it is committed to gender and racial diversity by using either explicit statements or subtle cues (Walker, Feild, Berneth, & Becton, 2012). One such cue may be the spatial orientation of males and females in photographic or video materials.
Fourth, “fair” spatial arrangements are critical in both political advertisements and in political debates. Not by chance, presidential election debates in the United States are regulated by a “memorandum of understanding,” which generally includes balanced seating or standing arrangements. Thus, right- versus left-positioning of the candidates is counterbalanced across TV debates. Similar principles should guide all political debates to prevent women candidates from being systematically placed in the position seen as less agentic, namely, to the right. Similarly, professionals designing political advertisements should be aware that the rightward orientation of politicians may communicate greater agency than leftward orientation.
Conclusions
Although further applied research on spatial layouts is needed, we believe that fair spatial displays may contribute to gender fairness in different professional contexts. Further, we would like to emphasize that counterstereotypical spatial conditioning may serve to modify existing agency-related gender stereotypes in a subtle way without arousing motivated resistance. All people and especially communicators and policy makers might become aware of the biased use of space (and its consequences) when representing women and men and actively intervene to promote gender equality.
Footnotes
Authors’ Note
Data and research materials for all studies are available from the first author upon request via e-mail.
Acknowledgments
We are grateful to Angelika Teutsch for collecting the data for Study 2 and to Silvia Pellegrin for collecting the data for Study 4.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was funded by a grant to the second author from the Italian Ministry of Education, University and Research, PRIN 2008, on Embodiment and Social Cognition: The Role of Sensorimotor Processes in the Perception of Individuals and Groups.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
