In the Mind’s Eye: Exploring the Relationship Between Visual Mental Imagery and Stereotyping

The Role of Imagery in Social Cognition

While some early evidence suggests that stereotypes evoke visual imagery, social cognitive theories that emerged in the 1970s focused almost exclusively on the role of stereotypical beliefs regarding groups’ traits and characteristics. A seminal definition of stereotypes from that era reflects this focus, describing a stereotype as “a cognitive structure that contains the perceiver’s knowledge, beliefs, and expectations about a human group” (Hamilton & Trolier, 1986, p. 133). This definition lacks specificity regarding the visual aspects of stereotypes; while it does not explicitly exclude mental imagery, neither does it explicitly incorporate it. Indeed, research from the social cognitive perspective devoted relatively little attention to visual appearance and mental imagery related to groups.

Early social cognitive research emphasized the semantic content contained in stereotypes, aiming to identify the traits and behaviors presumed to typify groups and their members (e.g., J. W. Sherman, 1996; Stangor & Lange, 1994; Wyer & Martin, 1986). These beliefs were often assessed through self-reported checklists of traits and characteristics (e.g., Fiske & Tablante, 2015; Krosnick et al., 2005), as well as indirect measures relying on response times to trait terms (e.g., Fazio & Olson, 2003; Gawronski et al., 2011). Reflecting its origins in information theory and cognitive psychology, early social cognitive research heavily relied on written stimuli. Stereotyping processes were typically investigated by presenting written descriptions of behaviors purportedly performed by a group member that either reinforced or contradicted established stereotypes. Standard measures included free recall and recognition memory for the group’s behavioral descriptions, alongside trait judgments to discern how the information was processed and influenced social judgment (see the works of Hamilton & Sherman, 1994; Leyens et al., 1994). Social cognitive research provided invaluable insights regarding the crucial role of stereotypes in processing social information, affecting attention, inference, and attribution (for reviews, see the studies by Hamilton, 1981; Hamilton et al., 1994). However, the initial decades of social cognitive research on stereotyping largely overlooked the visual aspects highlighted in Lippmann’s analysis.

The Role of Imagery in Cognitive and Neuropsychology

During the period when social cognition predominantly focused on the semantic content of stereotypes, research in cognitive psychology and neuropsychology actively investigated the role of VMI in various cognitive and behavioral phenomena. This body of research demonstrated the central role of VMI in simulating future scenarios and predicting outcomes in complex situations (Keller & Mrsic-Flogel, 2018; Kilteni et al., 2018; Kosslyn, 1980; Schacter et al., 2007). Research on human VMI has grown steadily (see reviews from Pearson, 2019; Spagna et al., 2021; Winlove et al., 2018) and now spans multiple domains (J. Liu & Bartolomeo, 2023). These include imagery of faces (Ishai et al., 2002), places (Boccia et al., 2015; Palermo et al., 2008), trajectories and rotations (Shepard & Metzler, 1971; Suchan et al., 2006), colors (Zago et al., 2010), shapes and objects (Kashihara & Nakahara, 2011; Mazard et al., 2005), and letters and words (Kosslyn et al., 1993; Stokes et al., 2009).

Much of this research has aimed to identify the cortical regions underlying VMI. Studies assessing the impact of lesions on the ability to visualize objects strongly implicate the ventral temporal cortex (see reviews from Bartolomeo, 2008; Bartolomeo et al., 2020; Spagna et al., 2023). The pivotal role of this region in supporting imagination was vividly demonstrated by an architect who lost the ability to imagine after suffering a lesion to a widespread area of the inferior temporal lobe, including the left fusiform gyrus (Thorudottir et al., 2020). Furthermore, J. Liu et al. (2023) identified the neural pathways of VMI and revealed how the absence of mental imagery, known as congenital aphantasia (Milton et al., 2021; Zeman et al., 2015), impacts these neural mechanisms. Findings support the notion that our subjective visual experiences, whether perceived or imagined, depend on coordinated activity among a domain-general area in the fusiform gyrus (the fusiform imagery node, Spagna et al., 2021), fronto-parietal networks for attention and working memory, and relevant stimulus-specific regions in the ventral temporal cortex (Spagna et al., 2023).

Evidence suggests that similar neural substrates are involved in visual imagery and perceptual processes (Kreiman et al., 2000; O’Craven & Kanwisher, 2000). Specific category-selective regions contribute to both perceiving and visualizing stimuli (J. Liu et al., 2023; Mahon & Caramazza, 2011), and visual perception and visual imagery rely on similar neural substrates (Dijkstra et al., 2019; Kosslyn et al., 2001; Winlove et al., 2018; Xie et al., 2020). However, several properties differentiate visual imagery and perception (Farah, 1984; Spagna et al., 2021). For example, Khuvis et al. (2021) recently demonstrated that human ventral temporal cortex neurons can encode different face stimuli and are endogenously activated when participants imagine those faces during free recall. Intriguingly, neural activity in response to face stimuli emerges early (100 ms), primarily in the fusiform face area of the inferior temporal cortex (Axelrod et al., 2019; Kanwisher et al., 1997).

Given the overlapping neural mechanisms supporting visual perception and VMI, it is unsurprising that both systems can similarly affect subsequent cognitive processes. For example, both imagining and seeing an object can produce comparable modulation in awareness thresholds (Ishai & Sagi, 1995; Pearson et al., 2008, 2011), and imagery can enhance priming on perceptual implicit memory tests (McDermott & Roediger, 1994). Moreover, envisioning interactions with a stimulus can improve performance on a later task involving a similar target (Tartaglia et al., 2009), and classical conditioning can occur using visual images and direct interaction with perceptual stimuli (Lewis et al., 2013).

However, until recently, research on VMI has largely overlooked social stereotypes. Some research explores visual imagery in various social phenomena, including autobiographical episodic memory (Bone et al., 2020; Vannucci et al., 2016), eyewitness testimony (Dobson & Markham, 1993; Krans et al., 2011), moral reasoning (Amit et al., 2014; Amit & Greene, 2012), prospection (Christian et al., 2013; Conti & Irish, 2021), and false memory (Garry et al., 1996; Gonsalves et al., 2004). Nevertheless, the potential role of visual imagery in stereotyping has yet to be thoroughly explored.

Understanding How Imagery Influences Stereotypes in Memory

Memory for stereotype-relevant information has received extensive empirical attention in social cognition (e.g., Allen et al., 2009; Fyock & Stangor, 1994; J. W. Sherman et al., 1998). Attention and memory biases for stereotype-consistent and stereotype-inconsistent information are assumed to be critical in perpetuating intergroup bias. Moreover, memory accuracy is central to reality monitoring, the ability to distinguish internally generated information from information from the outside world (Dijkstra et al., 2022; Simons et al., 2017). Research has shown that mental imagery contributes to memory errors and deficits in reality monitoring (Gancedo et al., 2021; Garrison et al., 2017; Simons et al., 2017).

Slusher and Anderson (1987) conducted research highlighting the benefits of integrating reality monitoring and stereotyping phenomena. Two studies examined whether individuals incorporate stereotype content into the mental images formed of others, thus affecting memory. In the first experiment, participants imagined members of various occupational groups and described their mental images. Analysis of these descriptions revealed that participants incorporated stereotypical traits into their mental images, with specific characteristics reflecting common occupational associations. In the second experiment, participants read, imagined, or saw individuals performing stereotypical or non-stereotypical actions for their occupations. Subsequently, when judging the frequency of trait presentations for each occupation, participants overestimated their exposure to stereotypic occupation-trait combinations. This overestimation was particularly pronounced when actions were imagined rather than described or observed. Participants failed to distinguish self-generated images from actual stimuli, potentially reinforcing beliefs in stereotype accuracy by inflating estimates of the frequency of stereotypic behavior.

More recent studies have provided further insight into how stereotypes influence imagined events to affect memory. For example, Kleider et al. (2008) instructed participants to read descriptions and view pictures of a man’s and woman’s actions or simply imagine them after reading about them. Some actions were gender stereotypical (e.g., “the handyman tightening a pipe under the sink”), while others contradicted gender stereotypes (e.g., “the handyman pouring cake batter into a pan”). Memory tests conducted soon after exposure showed minor differences in memory between conditions. However, memory tests revealed more substantial stereotype effects in the imagined versus the viewed conditions after a delay. Participants who imagined actions exhibited better memory for stereotype-confirming information and higher memory error rates for actions contradicting stereotypes than participants who viewed pictures of described actions. Essentially, those who imagined the actions tended to remember stereotype-confirming actions better but had worse memory for stereotype-disconfirming behaviors.

Overall, these findings demonstrate that mental imagery can lead to potent stereotype-confirming memory distortions, simultaneously perpetuating bias and impairing reality monitoring.

Using Imagery as a Means for Reducing Bias

Although reality monitoring research suggests that imagery can increase stereotypic bias, imagery might also reduce stereotyping and prejudice. Over recent decades, various interventions have been developed to reduce stereotyping and associated prejudice, with some showing promising results (FitzGerald et al., 2019). Among the most effective interventions are those that involve increasing exposure to members of stereotyped groups who defy stereotypes (Allport, 1954; Pettigrew & Tropp, 2006, 2011; for recent reviews, see the works of Paluck et al., 2019; Vezzali & Stathi, 2017). Experiences involving intergroup contact have decreased intergroup bias by reducing anxiety, fostering empathy and perspective-taking, and enhancing knowledge about stereotyped groups (Pettigrew & Tropp, 2008).

Blair et al. (2001) provided initial evidence that merely imagining counterstereotypical group members could reduce intergroup bias. Participants were asked to “imagine what a strong woman is like, why she is considered strong, what she is capable of doing, and what kinds of hobbies and activities she enjoys.” Participants in a control condition were prompted to “imagine what a vacation in the Caribbean would be like.” Performance on a subsequent gender implicit association task (IAT; Greenwald et al., 1998) showed that the imagery group reported reduced stereotypical gender associations compared with the control condition.

Crisp and colleagues (Crisp & Turner, 2009, 2012; Vezzali et al., 2020) have expanded on this initial finding by developing imagined intergroup contact interventions, where individuals visualize positive interactions with outgroup members. In one study (Turner et al., 2007), participants were instructed to

spend the next five minutes imagining that you are talking to a gay man that has sat next to you on the train. You spend about thirty minutes chatting until you reach your stop and depart the train. During the conversation, you learn some interesting and unexpected things about him.

Compared with participants in a control condition who imagined being on a hiking trip, participants who engaged in imagined contact reported lower anticipated anxiety in future interactions with gay men, more positive attitudes toward gay men, and reduced perceptions of homogeneity among gay men. These findings expand on initial evidence to demonstrate that imagining contact with a member of a stigmatized group can reduce several hallmarks of prejudice.

Similar benefits of imagined contact have been observed even in contexts of intense intergroup conflict (Bağcı, Piyale, & Ebcim, 2018; Husnu & Crisp, 2010b; Ma et al., 2019) and among individuals who are notably high in prejudice and right-wing authoritarianism (Asbrock et al., 2013; Borinca et al., 2022; West et al., 2015). Effects of imagined contact have also been found to persist over time (at least 1 month; Falvo et al., 2014) and to extend to attitudes regarding groups not initially visualized (Harwood et al., 2011). Recent research has identified several critical moderators and boundary conditions for imagined intergroup contact (Miles & Crisp, 2014; Shamoa-Nir & Razpurker-Apfeld, 2023; White et al., 2021). Imagined contact improves beliefs about typically stigmatized groups but leaves beliefs about admired groups unaffected (Brambilla et al., 2012). Imagining contact tends to be more effective in children than adults. In addition, interventions are more successful when the visualization emphasizes intergroup similarities and differences (Ioannou et al., 2017), including a potential for friendship (Bağcı, Piyale, Birçek, & Ebcim, 2018). Finally, imagined contact has been found to be most effective when it involves detailed imagery of intergroup interactions (Husnu & Crisp, 2010a; Turner et al., 2007) and reflects a positive interaction (Kuchenbrandt et al., 2013; Stathi & Crisp, 2008).

This literature suggests that mental imagery can mitigate intergroup bias. Imagery-based interventions offer particular advantages because they eliminate the need to create actual intergroup encounters with outgroup members who defy stereotypes. Imagery interventions are also advantageous for reducing intergroup bias because they can avoid feelings of anxiety and threat that often arise in face-to-face interactions with outgroups (Finchilescu, 2010; Stephan, 2014). Moreover, if such concerns are reduced through imagery, imagined contact can produce more positive attitudes toward outgroups and a greater willingness to engage in future intergroup interactions (Richeson & Trawalter, 2008; Turner et al., 2013; Wölfer et al., 2019). Finally, research indicates that individuals spontaneously engage in imagined contact with outgroup members outside of controlled laboratory settings (Stathi et al., 2020), thereby enhancing the external validity of imagery-based approaches. Various forms of mental imagery then offer promise as practical and effective methods for combating intergroup bias in real-world contexts.

Assessing Visual Imagery Through Reverse Correlation

Until recently, research methods did not easily allow for examination of the content of visual mental images. Reliance on self-reports or drawing-based techniques limited researchers’ ability to ascertain precisely what individuals envisioned when imagining typical group members, potentially introducing biases toward socially desirable responses (Allbutt et al., 2005, 2008). Fortunately, the adoption of an indirect method from psychophysics, known as reverse correlation, has revolutionized the exploration of visual features associated with social groups and their members.

Reverse correlation is a data-driven technique that, over numerous trials, presents participants with two response options derived from a single base image, typically a face or body photograph (Brinkman et al., 2017). These response options feature the same base image overlaid with a different random noise pattern on each trial. Participants are asked to select the image that better represents a specific group between each pair of response options (e.g., male, Chinese, friendly, criminal). The images chosen across hundreds of trials are averaged mathematically to create a composite classification image, revealing features most strongly associated with the group. An independent group of raters then assess the classification image to identify characteristics (e.g., race, gender, trustworthiness, attractiveness) that distinguish it from a composite of the non-chosen images. Stimuli from a recent article using this method to explore racial biases in diagnosing oppositional defiant disorder (Stroessner et al., 2023) are displayed in Figure 1.

OPEN IN VIEWER

Figure 1.

Base Image (l) and Two Response Options (r) (Top Panel) and Classification (l) and Non-Chosen Images (r) (Bottom Panel) From an Experiment Using the Reverse Correlation Procedure to Test for Racial Biases in Diagnosing Oppositional Defiant Disorder (Stroessner et al., 2023).

One of the earliest applications of reverse correlation in social perception (Dotsch et al., 2008) involved Dutch participants selecting the most Moroccan-like face from presented pairs. (Moroccans are a stigmatized group in the Netherlands, often associated with criminality.) The resulting classification image was rated higher in criminality and lower in trustworthiness, particularly among individuals with strong prejudices against Moroccans (based on their performance on an earlier IAT; Dotsch et al., 2008). This study provided initial evidence of reverse correlation’s utility in investigating mental imagery underlying stereotyping and prejudice.

While reverse correlation sheds light on the visible features associated with group members, it is important to note that composite images represent a distortion of a base face. Therefore, a classification image cannot precisely mirror individuals’ mental visualizations of the investigated group. Nevertheless, reverse correlation offers an innovative avenue for studying subjective social perception and the role of mental imagery in stereotyping. Reverse correlation has been instrumental in assessing differences in mental representations of ingroups versus outgroups. Studies have shown that ingroup members are perceived as more trusting, caring, intelligent, and attractive than outgroup members (Ratner et al., 2014; see also Hong & Ratner, 2021). In addition, the facial features of ingroup members more strongly suggest trustworthiness.

Regarding specific identity biases, reverse correlation has demonstrated associations between visual images of emotion and gender. Faces in one experiment (Brooks et al., 2018) were selected for gender (male, female) or emotions (happy, angry). Results reflected gender-linked stereotypes regarding emotions, with male images perceived as angrier and female images as happier. Similarly, the angry classification image was judged more male, and the happy classification image was judged more female. Moreover, this bias was related to the strength of participants’ stereotypical associations linking men to anger and women to happiness. Finally, classification images of physics and language teachers were highly gendered in stereotypical ways (Degner et al., 2019), and classification images of leaders were judged to appear particularly powerful for female faces compared with non-leaders (Giacomin et al., 2021).

Beyond gender, reverse correlation has also shown that criminality and immorality (Avery et al., 2021), reckless risk-taking (Wages et al., 2022), and clinical diagnosis of oppositional defiant disorder (Stroessner et al., 2023) are associated with faces judged as Black. Interestingly, classification images of biracial individuals generated by White participants tend to appear more prototypically Black than those generated by Black participants. Furthermore, prototypically Black images were also judged less worthy of receiving money by participants high in prejudice in the same study (Oliver et al., 2024).

In addition to furthering the understanding of well-established biases, reverse correlation has examined other forms of bias that have yet to receive extensive attention. For example, one set of studies explored the visual mental representation of theists compared with atheists (Brown-Iannuzzi et al., 2018). Atheist classification images were judged as less trustworthy, moral, competent, warm, kind, human, and active than theist classification images. Participants also rated the atheist classification image as less likable, happy, and attractive than the theist classification image. Another study (Rudert et al., 2020) assessed visual mental representation associations with people participants expected to be ostracized by others. Consistent with predictions, judgments of the ostracized classification images showed them to be rated low in conscientiousness and agreeableness.

Johnson et al. (2012) used reverse correlation to study the visual mental representation of gendered body types. Images selected to be women had more extreme waist-hip ratios than the anthropometric average cisgender woman, whereas male classification images’ bodies were typical for cisgender men. Finally, Lick et al. (2016) used reverse correlation to study perceptions of sexual interest. Participants made classification judgments involving both state-based (sexually aroused/unaroused) and trait-based (sexually promiscuous/non-promiscuous) dimensions. The features of the classification and non-chosen images differed markedly in pupil dilation, with sexually interested faces containing larger and darker pupils. These differences were perceptually obvious to naïve observers, indicating a tendency to associate sexual arousal with a visual mental characteristic as specific as pupil dilation. Each of these examples illustrates how a novel technique for studying mental imagery has advanced the breadth of topics relating to social bias and stereotyping and presents a powerful tool for future research.

Social Category Representation in Mental Imagery

Recent studies on VMI have generally avoided studying visualizations of people and social categories, focusing instead on the neurocognitive mechanisms underlying mental imagery (Pearson, 2019; Spagna, 2022). These studies often involve simple stimuli (e.g., colors or shapes; J. Liu & Bartolomeo, 2023; McNorgan, 2012; Spagna et al., 2023). However, considering more complex social stimuli could yield valuable insights into the purposes and functions of visual imagery.

For instance, are stereotypical mental images easier to imagine than non-stereotypical ones? Based on research indicating the salience of stereotypical mental images on judgments (Kleider et al., 2008; Slusher & Anderson, 1987), we expect that stereotypical images are more readily produced than those that defy stereotypes. Simple reaction time paradigms might help measure the ease of generating mental images of groups varying in stereotypicality. Understanding the ease of conjuring stereotypical representations is both theoretically and practically important. Various consequential effects have been shown to rely on the ease of recall, including estimates of commonality (Tversky & Kahneman, 1974), risk (Pachur et al., 2012), and trait judgments (N. Schwarz et al., 1991; see Reber, 2017 for discussion of effects associated with ease of retrieval).

In addition, a lack of clarity exists regarding how prototypical mental images, procured regularly in daily life (Stathi et al., 2020), might shift over time. Imagined contact research has demonstrated the potential for VMI to modulate attitudes and behaviors related to stereotypes and bias, at least for a period of interventional aftermath (Crisp & Turner, 2012; Stathi et al., 2011). However, to our present knowledge, research has yet to investigate how the specific content of mental images is altered when stereotypes change.

It appears a reasonable, though untested, assumption that the prototypical mental image of an outgroup or one of its members might be modified following an imagined contact intervention involving counterstereotypical imagery. More broadly, does attitude change (e.g., reduction in prejudice or one’s belief in a stereotype) produce some modification in one’s imagery of that group? Using reverse correlation methods before and after various interventions designed to promote attitudinal change could allow assessments of shifts in prototype representations in mental imagery. One crucial direction of this research might examine whether these changes correspond with a weakening of the accessibility of stereotypical imagery, a strengthening of the accessibility of alternative imagery, or both. Future research can also explore the salience of these changes over time.

Another area for additional exploration involves mental imagery regarding concealable social identities (e.g., identities that are difficult to discern visually, such as sexual orientation, cognitive ability, and religion). Imagined contact interventions have improved attitudes toward concealable identity groups (Turner et al., 2007), and reverse correlation studies show that participants asked to select members of concealable categories produce classification images that can be visually distinguished from non-chosen composites (Brown-Iannuzzi et al., 2018; Deska et al., 2020). One intriguing question regarding mental images of concealable identities is what features people use to distinguish members from non-members since concealable categories, by definition, cannot be visually distinguished easily. One possibility is that holding negative attitudes and stereotypes toward a group might produce exaggerated valenced features in visual imagery. For example, in a study examining mental images of theists vs. atheists (a group for which many people hold negative attitudes; Gervais et al., 2011), the classification image for atheists exhibited facial characteristics implying low levels of trustworthiness, morality, competence, and kindness (Brown-Iannuzzi et al., 2018). A second possibility is that mental imagery for concealable identities might reflect beliefs regarding other social categories believed to be similar in some respect. For example, one study using reverse correlation techniques showed that feminist women are imagined to appear more masculine than traditional women, whereas feminist men appear more feminine than men with traditional gender attitudes (Gundersen & Kunst, 2019). Such findings represent compelling evidence that stereotypes about specific and related groups can color mental imagery.

This line of inquiry investigating whether social categorization is a function of mental imagery provides an ecological model for understanding how the mental images we create are beholden to attitudinal and cognitive functions. Expanding the exploration beyond stereotypical imagery might provide a greater understanding of the interplay between mental imagery and other cognitive processes. For example, creative thinking has been tied to (a lack of) automatic stereotype activation (Sassenberg & Moskowitz, 2005), as well as the ability to transform mental imagery (Palmiero et al., 2011, 2015). Cross-cultural contact has also been found to stimulate creativity, particularly through the process of cultural adaptation (Maddux & Galinsky, 2009). Future research might capitalize on stereotypical mental imagery change to further explore a reciprocal relationship between mental imagery and creative processes. If creativity is uncovered as the fuel for mental imagery changes and consequent stereotype alterations, how else does creativity in mental imagery drive attitude changes? This overall shift in research focus can further provide insights into how individual differences in these high-level functions and their interactions play a role in predicting problem-solving, decision-making, and behavior in real-world scenarios.

Vividness in Stereotyping

Despite extensive research in VMI, imagery vividness (at least as defined in cognitive psychology) remains largely unexplored in social cognitive approaches to stereotyping. Vividness has been defined in several ways over time. Marks (1973) argued that the vividness of a mental image relates to “a combination of clarity and liveliness: the more vivid an image, the closer it approximates an actual percept” (Marks, 1973). Furthermore, Marks (1999, 2019) claimed that clarity (brightness of colors and sharpness of outline and details) and liveliness (how dynamic, vigorous, and alive the image seems) are two independent but equally crucial aspects of VMI. According to Marks (2019), vividness pertains to the intensity of an image, independent from the level of imagery detail.

Vividness has also been defined in terms of realism. For example, Campos et al. (2008) argue that a vivid image “exhibits the greatest similarity to reality” (p. 338). Finally, Fazekas and colleagues (2020) have argued that vividness is defined by subjective intensity (how much an image stands out from its background), specificity (how much an image is precise, clear, and unambiguous), and stability (how durable an image is). Regardless of how it has been defined, imagery vividness has been shown to vary between individuals (Cui et al., 2007; Marks, 1999; Sheldon et al., 2017), and vividness declines across the lifespan (Gulyás et al., 2022). The vividness of mental images has also been found amenable to willful control (Cochrane et al., 2021).

Imagery vividness has significant consequences for other processes and judgments. It correlates with performance on certain memory tasks (Baddeley & Andrade, 2000), arousal levels (Barrowcliff et al., 2004; Bywaters et al., 2004), emotional valence (Alter & Balcetis, 2011), and visual cortex activity (Cattaneo et al., 2011, 2012; Farah & Peronnet, 1989; Sparing et al., 2002). Vivid images are easier to remember (Cohen & MacNeilage, 1974), produce more accurate recall (Marks, 1973; Swann & Miller, 1982), elicit incidental recall (D’Angiulli et al., 2013), increase realism (Stephan-Otto et al., 2017), and contribute to false memory (Gonsalves et al., 2004). Furthermore, higher vividness produces faster, more accurate, and more sensitive detection of subsequently presented threatening stimuli (Imbriano et al., 2020).

In research on imagined intergroup contact, vividness has been implicated as a potential moderator of intervention efficacy. Studies have shown that greater vividness in imagining intergroup contact leads to more positive intentions. However, vividness in this context has, at times, been operationalized differently from how it is defined in VMI research (i.e., the level of detail in mental images). In the first of two studies investigating vividness, Husnu and Crisp (2011) prompted some participants to consider when and where an imagined contact occurred. Participants who visualized this elaborated imagined contact reported greater intergroup contact intentions, an effect mediated by subjective ratings of imagery vividness (e.g., clarity, brightness, liveliness). In a second experiment, researchers had participants imagine contact with their eyes open or closed (a vividness manipulation used in several VMI studies, e.g., Campos et al., 1999). Heightened intentions to engage in future events were found for participants whose eyes were closed.

Based on these preliminary investigations of vividness in social cognition (Blair et al., 2001; Swann & Miller, 1982), we speculate that just as vivid images increase arousal, recall, and visual cortex activity, vivid stereotypical images might lead to more extreme stereotypic judgments. This would be expected particularly if visual images confirm stereotypes (typically negative ones) rather than defy them (given the increased salience of stereotypical mental images; Kleider et al., 2008). Alternatively, vividness might reflect control over mental imagery, such that greater control might be associated with decreased stereotype content (c.f., Gordon, 1949; see also Moskowitz, 2010; Payne, 2005). Future research could directly manipulate vividness within subjects (e.g., by highlighting clarity and liveliness in induction scripts) or focus on variations in vividness as a predictor of stereotyping outcomes to address these competing theories.

Conversely, future studies exploring the relationship between vividness and stereotyping strength can yield insight into modulators of mental imagery vividness more broadly. As the strength or frequency of stereotyping can be effectively reduced through intervention (FitzGerald et al., 2019; S. Liu et al., 2021), the vividness of associated mental imagery may be impacted in tandem. If this is the case, research might investigate whether this impact increases or decreases vividness across various types of imagery, the durability of changes in vividness, and how vividness of non–stereotype-relevant (or even non-social) mental imagery changes.

Visual Perspectives in Stereotypical Imagery

Another opportunity for integration concerns the role of visual perspectives within mental imagery. Imagery has been found to reflect either first-person or third-person perspectives (Nigro & Neisser, 1983). First-person perspectives immerse the perceiver in the imagined scenario as if events were happening to them (e.g., imagining a group staring intently at you while giving a speech). In contrast, third-person perspectives involve viewing the scenario as an observer, often from another person’s viewpoint (e.g., imagining watching yourself delivering a speech). Numerous studies have explored differences between first-person and third-person mental imagery regarding past events. First-person perspectives are typically associated with recollections of greater emotional intensity (McIsaac & Eich, 2002, 2004), greater vividness (as defined classically; Sutin & Robins, 2008), and a stronger sense of reliving a remembered event (Berntsen & Rubin, 2006). However, third-person perspectives often contain more detailed elaboration (McIsaac & Eich, 2002, 2004).

Differing perspectives have also been studied for events that have yet to occur (Atance & O’Neill, 2001; Schacter et al., 2017). In imagining an anticipated event, first-person perspectives emphasize sensory-affective information, whereas observer perspectives highlight more abstract and contextual information. The perspective used to imagine a future event can result in dramatically different interpretations with differing behavioral consequences (Libby & Eibach, 2011; Sutin & Robins, 2008). For example, research has shown that imagining future events from a third-person perspective increases motivations for subsequent behaviors involving actions such as voting (Libby et al., 2007) and retirement saving (Macrae et al., 2017), often by shifting attitudes toward desired behavior.

First-person and third-person perspectives might also moderate stereotyping, although research findings are inconsistent regarding the efficacy of first- versus third-person perspectives on bias reduction. Several experiments have used manipulations to encourage first-person perspectives over default third-person views of marginalized groups. For instance, research participants have been asked to “imagine the thoughts, feelings, and actions” of members of a stereotyped group (Lai et al., 2014) or to “imagine how the person being interviewed feels about the experiences he or she describes and how it has affected his or her life” (Vescio et al., 2003). Other studies manipulate perspective by prompting participants to write essays about a stereotyped target from a first-person perspective (e.g., using “I” while describing a person’s imagined experiences) or a third-person perspective (e.g., using “He” to describe a person’s experiences; Marx & Stapel, 2006; Wheeler et al., 2001). In these studies, encouraging a first-person perspective regularly produces more favorable attitudes toward marginalized groups than a third-person perspective (Vescio et al., 2003). However, it is important to note that such interventions may not necessarily induce VMI, and few studies have tested whether such prompts instigate VMI.

Another promising line of recent research uses virtual reality (VR) to alter visual perspectives in intergroup contexts (O’Donnell et al., 2021; Tassinari et al., 2022). One notable study (Chen & Ibasco, 2023) involved participants who assumed the perspective of either an immigrant or an ingroup character experiencing a microaggression at work. Those who adopted the immigrant’s perspective reported more favorable attitudes and reduced stereotypes toward immigrants than participants who assumed an ingroup member’s perspective. This improvement was attributed to an increase in empathy toward immigrants. Thus, VR can enhance intergroup relations by enabling individuals to “see the world through another’s eyes.” Nonetheless, it is crucial to recognize that the outcomes of VR interventions are not uniformly positive, as some studies have reported a rise in prejudice (Tassinari et al., 2022).

On the other hand, other studies have shown that third-person perspectives more effectively reduce intergroup bias. Several experiments in which participants visualize intergroup contact using first- or third-person perspectives show that improved attitudes are more likely when third-person views are encouraged. In these studies, participants who imagine an intergroup encounter involving a marginalized group from a third-person perspective tend to make more positive judgments of the group and report stronger intentions to engage with the group (Crisp & Husnu, 2011; Libby et al., 2011, 2014). Two related mechanisms have been invoked to explain these effects. Crisp and Husnu (2011) have argued that bias is reduced when using an objective perspective because behaviors are seen as more reflective of one’s character when imagined from a third-person perspective. In other words, “seeing” oneself engage in an intergroup interaction is likely to produce the inference that one must have positive intergroup attitudes to engage in a successful intergroup interaction. Niese et al. (2022) have suggested that third-person perspectives enhance the tendency to interpret actions through one’s beliefs and attitudes. In one study showing the mediating effect of values (Libby et al., 2014), White participants were prompted to visualize themselves interacting with a Black partner using a first-person or third-person perspective. The reported anxiety that participants expected to experience during the interaction was predicted more strongly by their personal motivation to avoid prejudice in the third-person than in the first-person condition.

Recent reviews of perspective-taking manipulations have produced inconsistent results (FitzGerald et al., 2019). However, some variability in research outcomes may reflect differences in specific manipulations used to induce perspective-taking. As our examples illustrate, some studies help participants focus on the experiences or emotions of interaction partners, while other studies emphasize visual shifts in imagining intergroup encounters more heavily. We suspect that shifting visual perspectives alone might have different effects compared with manipulations invoking emotional, value, or empathic considerations. It will be crucial to explore which combinations of these factors are necessary and sufficient for manipulations involving visual perspectives to ameliorate intergroup bias.

Conscious and Unconscious Processes in Imagery

Distinguishing implicit social cognition—automatic cognitive processes of which participants are unaware—from explicit social cognition—cognitive processes requiring initiation and effort—has been a research goal for the last several decades (for reviews, see Gawronski et al., 2020; Gawronski & Payne, 2011; Greenwald & Lai, 2020; Winkielman & Schooler, 2011). Much of this research has focused on the nature and role of implicit attitudes and stereotypes in accounting for intergroup behavior. Implicit stereotypes appear to be related but distinct from their explicit counterparts, and each can independently predict distinct attitudes and behaviors (Bar-Anan & Vianello, 2018; Kurdi & Banaji, 2022; Kurdi et al., 2019; Ratliff & Nosek, 2010). For example, implicit beliefs have been found to uniquely impact aspects of interpersonal behavior (Asendorpf et al., 2002), judgments (Banaji & Greenwald, 1995), and intergroup interactions (Pardal et al., 2020).

Whether mental imagery occurs unconsciously has yet to be extensively explored. However, some have speculated (albeit controversially; see the study by Block & Phillips, 2017) that imagery can occur both consciously and unconsciously (Brogaard & Gatzia, 2017; Cabbai et al., 2023; Mammarella, 2011; Nanay, 2021). Evidence of some aspects of implicit imagery already exists. For example, visual imagery can occur both voluntarily (e.g., imagining a pleasant intergroup exchange) and involuntarily (e.g., experiencing intrusive memories and flashbacks in post-traumatic stress disorder; Pearson et al., 2015). Thus, some imagery does not require intentional activation and cannot be controlled, two defining characteristics of relatively automatic processes (Bargh, 1994). In addition, preparing to encounter a stimulus can trigger the activation of a mental image of its appearance prior to conscious awareness, suggesting not only involuntary but also unconscious imagery (Koenig-Robert & Pearson, 2019; Kok et al., 2014).

Nanay (2021) offers three types of evidence for the claim that VMI may be conscious or unconscious. First, much research has demonstrated that images presented subliminally, below the threshold of perceptual awareness, can nonetheless affect judgments and actions (e.g., Bar & Biederman, 1998; Elgendi et al., 2018; Kunst-Wilson & Zajonc, 1980). Second, many aspects of visual imagery are not accessible to conscious awareness or available for self-report, other hallmarks of automatic processes (Bargh, 1994). For example, although the orientation of an imagined stimulus affects performance on mental rotation tasks (Shepard & Metzler, 1971), participants are generally unable to report the impact of orientation on rotation efficiency or how stimuli are mentally rotated. Third, several phenomena can be better explained by positing the existence of unconscious mental imagery than by accounts relying on conscious imagery alone, including unilateral neglect (Bartolomeo & Chokron, 2002; Bourlon et al., 2008; Grossi et al., 1993) and unconscious priming (Kwok et al., 2019; Nanay, 2021).

The existence of implicit mental imagery provides opportunities for integration with research on implicit processes in social cognition. As has been established in several decades of research, implicit and explicit processes are somewhat independent and differentially associated with distinct judgments and actions. Whereas implicit social cognitive processes have been more strongly linked to spontaneous behavior (i.e., behavior that is difficult to control or of which people are typically inattentive), explicit processes are more closely associated with deliberative behavior (i.e., behavior that is thoughtfully determined or seemingly can be controlled; e.g., Amodio & Devine, 2006; Dovidio et al., 2002; McConnell & Leibold, 2001). Therefore, it will be essential to determine the degree to which implicit and explicit imagery processes are distinct and whether they differentially affect social cognition. Furthermore, understanding the relationship between differing aspects of implicit cognition—for example, implicit mental images and implicit stereotypes—will be crucial in future research. For instance, can implicit imagery affect implicit stereotypes and vice versa? Can these distinct implicit processes be manipulated to causally produce changes in the other form of representation? Finally, what unique aspects of judgment and behavior rely on each representation, and do they function additively or interactively? These are promising areas for research, given clear evidence from social cognition demonstrating the interplay between multiple implicit processes and also between implicit and explicit phenomena.

Correspondence and Non-Correspondence in Imagery and Social Perception

Another promising area for cross-fertilization relates to the degree to which there exists a correspondence between VMI and encountered social stimuli. Numerous studies in social cognition have shown that the degree to which people behave consistently with their beliefs and attitudes depends on whether there exists a substantial overlap in the content of their mental representations of an imagined stimulus and the features of an encountered stimulus. The idea that “attitudes will predict behavior better when assumed and perceived target characteristics match rather than mismatch” (Lord & Lepper, 1999, p. 302) is central to attitude representation theory (Lord & Lepper, 1999; Sia et al., 1999). One classic study found that prevalent negative attitudes about “members of the Chinese race” held by hotel and restaurant proprietors in the 1930s did not affect how a well-dressed, educated Chinese couple was treated in their establishments (LaPiere, 1934). Presumably, the low overlap between common (negative) mental representations of Chinese people at that time and the couple’s appearance meant that any mental imagery involved had little bearing on how they were received. Several other studies have shown that when beliefs about groups are discordant with characteristics of encountered group members, people do not act consistently with their beliefs (e.g., Lord et al., 1984, 1991; Paulson et al., 2012). However, the theory is broad regarding the nature of mental representations underlying actions, ranging from beliefs about typical behaviors and traits to visual images of a group’s members.

To our knowledge, relatively little research has explored the implications of matches between VMI and encountered stereotyped individuals (Farah, 1989; Freeman & Johnson, 2016). Several potentially fruitful avenues of inquiry are apparent. For example, are imagined and encountered group members associated with more closely overlapping neural activity when they are highly similar compared with when they are not? What neural changes would be expected if an imagined group member bore little resemblance to an encountered group member? If these changes are pronounced, would effects associated with VMI vividness not affect responses to the encountered individual? Does correspondence have a differential impact on mental images of concealable identities compared to visible ones? And finally, how do visual mental images change over time as a function of (mis)matches with perceived group members? These are among the most exciting areas for future research that might provide integral insights regarding the intersections between mental imagery and stereotyping.

Imagery and Emotion

A final possible avenue for development involves leveraging insights from research on mental imagery and psychopathology to better understand imagery and stereotyping. Mental imagery has been implicated in a variety of emotional disorders (e.g., post-traumatic stress disorder, social anxiety disorder, and bipolar disorder; for reviews, see Holmes & Mathews, 2010; S. Schwarz et al., 2020) as both a symptom (e.g., intrusive images) and an amplifier of emotions. The relationship between mental imagery and emotional experiences has been evidenced in childhood psychopathology, although increases in mental imagery vividness across development may exacerbate the imagery-emotion link (S. Schwarz et al., 2020). Imagery modification interventions have also shown efficacy in reducing emotional disorder symptoms (Morina et al., 2017). With evidence that imagined contact interventions seem to exert changes in stereotypical attitudes through decreased anxiety (Vezzali et al., 2013), an appealing pathway for future study might further examine the mediation of emotion in stereotypic imagery by considering VMI. This research might explore whether emotional processes mediate or moderate the impact of mental imagery on stereotypic attitudes, as well as how stereotypic imagery influences emotional regulation processes in intergroup settings.