Using Virtual Reality in Sexual Aggression Alcohol Administration Research: Opportunities and Challenges

Abstract

Objective:

This article explores methodological opportunities, challenges, and potential solutions for alcohol administration research that uses a virtual reality (VR) analogue for sexual aggression.

Method:

We briefly review relevant sexual aggression, alcohol administration, and VR literature. The strengths and limitations of different types of analogues are described.

Results:

Most existing sexual aggression analogues use sexual assault scenarios that end the same way for all participants, who are then asked how they would respond in that situation. Studies using these analogues have produced valuable findings regarding alcohol's role in sexual aggression; however, one important limitation is that participants can distance themselves from the situation before responding. VR provides highly immersive behavioral analogues that require participants to make quick decisions; thus, they are less likely to be impacted by social desirability. In VR, participants make multiple decisions based on the feedback they receive from their virtual dating partner and the ending depends on the choices they make. For example, VR analogues can allow participants to select dates who are intoxicated or to encourage their virtual dating partner to drink alcohol. Thus, researchers can model different patterns of responses and strategies used to attempt to obtain sex, which can then be used to develop interventions targeted for individuals with different types of risk profiles.

Conclusions:

There are strengths and limitations associated with all experimental proxies. The field would benefit from greater discussion of the essential elements of sexual aggression and whether these vary for different types of perpetrators and situations that involve alcohol.

Despite decades of research and public outcry, rates of sexual assault remain alarmingly high throughout the world (Basile et al., 2022; World Health Organization [WHO], 2021). Sexual assault is commonly defined as nonconsensual sexual contact (penetrative or nonpenetrative) obtained through force, threats of force, verbal coercion, or when the victim is unable to consent (WHO, 2021). Although alcohol is involved in approximately half of all sexual assaults in the United States (Abbey et al., 2014; Basile et al., 2021; Lorenz & Ullman, 2016), few alcohol administration experiments examine sexual assault perpetration, which is labeled “sexual aggression” in this article. Random assignment of participants to alcohol conditions allows researchers to draw causal conclusions about alcohol's effects on people with different beliefs and life experiences. This article describes the ongoing development of a three-dimensional (3D) virtual reality (VR) analogue to assess men's sexual aggression proclivity toward women in alcohol administration research. The methodological issues described in this article apply to sexual aggression with perpetrators and victims of other genders, and more broadly to researchers evaluating the costs and benefits of using different types of experimental analogues. Because VR allows participants to make multiple decisions in an absorbing, evolving situation, it is optimal for understanding the in-the-moment effects of alcohol and other situational variables on potential perpetrators' actions. In this article, we review the relevant sexual aggression, alcohol administration, and VR literature. One major theme of this article is that there are benefits and limitations associated with all experimental analogues and that a careful comparison of their effects is needed to guide the development of more precise etiologic theories and evidence-based interventions.

Alcohol's role in the etiology of sexual aggression

Sexual assault perpetrators have been described as having heterogeneous motives, backgrounds, and triggers (Seto & Barbaree, 1995; Tharp et al., 2013). Commonly identified risk factors include personality traits, attitudes, perceptions of peer norms, and past experiences (Spencer et al., 2023; Tharp et al., 2013; Waterman & Edwards, 2022). In addition, men's self-reports of sexual aggression are positively associated with their sexuality-related alcohol expectancies, stereo-types about drinking women, overall alcohol consumption, alcohol consumption in sexual situations, heavy episodic drinking, and alcohol-related problems (for reviews, see Abbey et al., 2022; Davis et al., 2023). Surprisingly, there is little theory or research that considers patterns of perpetration over time and whether some types of risk factors are more applicable to some types of perpetrators or types of perpetration. However, a few studies have found differences between men who only use verbal coercion tactics and those who also use physical force or the victim's impairment (Schewe et al., 2022; Swartout et al., 2022).

Alcohol affects a wide range of behaviors through two well-established mechanisms: pharmacologically induced cognitive processing deficits and beliefs about alcohol's effects (for reviews, see Davis et al., 2023; Parrott & Eckhardt, 2018). These processes have been applied to sexual aggression in Abbey's (1991, 2002, 2017) two-stage theoretical model. The first stage focuses on expectations for the situation formed early in the interaction, and the second stage focuses on the point at which forced sex occurs. Given that most sexual assaults occur between people in a current or potential dating or sexual relationship, the expectation that some sexual activity may occur is reasonable (Abbey et al., 2001; Blayney & Read, 2018; Davis et al., 2015). Past research demonstrates that men interpret a wide range of friendly cues as signs of sexual interest, including drinking alcohol together (Abbey, 1987; Treat et al., 2015). Once individuals have formed a hypothesis, they tend to focus on confirming cues and ignore disconfirming cues (Klein & Kunda, 1992; Snyder & Stukas, 1999). Thus, misperception of sexual interest frequently occurs without alcohol consumption (Abbey, 1982; Lee et al., 2020; Rerick et al., 2020; Wegner & Abbey, 2016). Nonetheless, this focus on expectancy confirming evidence is intensified by the higher-order cognitive processing impairments associated with intoxication (Farris et al., 2010; Treat et al., 2016).

If someone has convinced themselves that their partner wants to have sex with them, then the second stage of Abbey's model posits that this belief makes them feel comfortable persisting after a sexual refusal. Persistence is normalized by traditional heteronormative sexual scripts that encourage men to presume that their partner's initial reluctance can be overcome (Davis et al., 2023; DeGue & DiLillo, 2004; Vannier & O'Sullivan, 2011). As in the first stage, this process can occur without alcohol. However, among men prone to sexual aggression, the myopic cognitive processing deficits associated with intoxication (Steele & Josephs, 1990) can increase the likelihood of a sexually aggressive response. Instigating cues such as sexual arousal, anger, and entitlement are expected to be more salient than inhibitory cues such as the victim's discomfort, one's sense of morality, or knowledge of active consent guidelines.

Alcohol as a tactic. Another line of research examines perpetrators' use of the victim's intoxication as a strategy to obtain sex (Abbey & Jacques-Tiura, 2011; Pegram et al., 2018; Tyler et al., 1998; Zinzow & Thompson, 2015). Some perpetrators target women who are already quite intoxicated; others encourage women with whom they are already interacting to drink faster, drink stronger drinks, or play drinking games (Zamboanga et al., 2017). In addition, perpetrators who use impairment tactics report consuming more alcohol during the incident than perpetrators who used other tactics and nonperpetrators (Abbey & Jacques-Tiura, 2011; Zinzow & Thompson, 2015). Thus, there is likely some synergy between perpetrators' alcohol expectancies for the victim and alcohol's pharmacologic and expectancy effects on themselves.

Strengths and limitations of different types of sexual aggression analogues

Experimentalists emphasize the importance of considering the essential elements of the behavior of interest when determining the adequacy of an experimental analogue (Berkowitz & Donnerstein, 1982). This raises the question, “What are the essential elements of sexual aggression that need to be captured by an experimental analogue?” Do these elements include anger about thwarted sexual goals? Exerting one's power and dominance over another person's bodily integrity and sexual autonomy? Feeling led on and entitled to sex? Are physical contact and the ability to touch the other person essential elements? Is seeing the other person's response vital to some or all perpetrators? Would a general aggression analogue such as noise bursts or hot sauce administration ever be sufficient (Watkins et al., 2023; Wegner, 2014)? We posit that essential elements include actual or imagined sexual contact with another person; however, this is ultimately an empirical question. We encourage readers to consider what they view as the essential elements of sexual aggression as they read this article. We return to this issue in the recommendations for future research section.

Table 1 summarizes the strengths and limitations of different types of analogues based on published studies. The last column of the table highlights elements that could be integrated into each type of analogue in future research.

Table 1.

Summary of strengths and limitations associated with different types of sexual aggression analogues

	How sexual aggression is typically assessed	Major strengths as typically used	Major limitations as typically used	Elements that could be altered or added
Scenarios
Video	Self-report: How likely to behave like the male character if they were in that situation.	Usually based on sexual assault narratives; thus, high in mundane realism. Provides full range of visual and auditory cues available in natural interactions. Reading not required.	Actors' perceived ethnicity and social class can allow participants to distance themselves if perceived to be different from them. Easily visually outdated as styles and norms change, which allows participants to distance themselves. Significant costs involved with hiring actors, filming, editing, and revising. Outcome usually assessed through self-report responses (1–7 scales) postscenario.	Could use a behavioral assessment similar to audio scenario (i.e., point in story that participant would stop). Could be stopped multiple times to assess in-the-moment thoughts and feelings. Could be stopped multiple times allowing participants to determine what happens next rather than providing predetermined ending.
Audio	Behavioral: Number of seconds until participants stop the action (which is linked to specific point in the narrative).	Usually based on sexual assault narratives; thus, high in mundane realism. Provides auditory cues. Reading not required. Fairly inexpensive to initially develop and to alter as styles and norms change or to address different hypotheses. Use of behavioral assessment.	Does not provide visual cues, including nonverbal cues, which are often used to convey lack of consent. Some participants may not be able to project themselves into the situation without visual stimuli.	Could use a self-report measure of likelihood of behaving like the male character. Could be stopped multiple times to assess in-the-moment thoughts and feelings. Could be stopped multiple times allowing participants to determine what happens next rather than providing predetermined ending.
Written	Self-report: How likely to behave like the male character if they were in that situation. Often stopped multiple times to assess in-the-moment thoughts and feelings.	Usually based on sexual assault narratives; thus, high in mundane realism. Allows participants to imagine a situation and characters that are relevant to them. Minimizes superfluous cues like hairstyles or furniture styles that can subtly affect judgments. Inexpensive to initially develop and to alter as styles and norms change or to address different hypotheses.	Participants must be able to read and comprehend the full scenario, when sober or intoxicated. Some participants may not be able to project themselves into the situation without visual or auditory stimuli. Outcome assessed through self-report responses (1–7 scales) post-scenario.	Could use a behavioral assessment similar to audio scenario (i.e., point in story that participant would stop). Could be stopped multiple times allowing participants to determine what happens next rather than providing a predetermined ending.
Behavior
Toward confederate	Behavioral: Directed at confederate.	Interaction with confederate can create an absorbing situation enhancing experimental realism. Confederate can provide full range of visual and auditory cues available in natural interactions. Reading not required. Confederate's clothing, comments, and behavior can be easily altered as styles and norms change or to address different hypotheses. Use of behavioral assessment.	Depending on the behavioral outcome, mundane realism may be low. Depending on the behavioral outcome; there may be ethical concerns associated with the confederate's mental health and physical safety. Requires a trained confederate at every session.	Can code participants' verbal and non-verbal responses during interaction with confederate. Can include self-report questions at end of study to assess thoughts and feelings during the interaction.
Toward fictitious other	Behavioral: Directed at fictitious other.	Use of fictitious other ensures standardization across participants. Reading not required. Fictitious other's responses or other elements of situation can be easily altered to address different hypotheses. Use of behavioral assessment.	Depending on the behavioral outcome, mundane realism may be low. Depending on the precise procedures used, limited interaction with and feedback from fictitious other is likely to limit generalizability.	Can systematically vary elements of the analogue to determine what factors affect likelihood of sexually aggressive response (cf. Angelone et al., 2022). Can include self-report questions at end of study to assess thoughts and feelings during the session.
Virtual reality (VR)	Behavioral: Number of times try sexual activity that virtual date refuses.	High immersion and presence. Participants make some initial choices (i.e., ethnicity and physical appearance of their date) to enhance engagement and realism. Each decision is based on feedback from prior choices. Whether or not sexual aggression is portrayed depends on participants' choices, which may decrease reactance. Reading not required. Real-time data gathering allows analysis of behavioral sequences and identification of different trajectories without relying on participants' recall. Use of behavioral assessment.	Expense: Advanced programming skills required to develop the analogue. Expense: Requires high-quality VR headset, gaming-style computer, and graphics. Large amounts of data need to be stored. Can be ethical concerns about reinforcing harmful behaviors depending on what behaviors are allowed and how graphically outcomes are depicted. Requires exclusion of individuals with moderate to high levels of motion sickness or traumatic brain injuries, reducing the generalizability of the findings.	Can code participants' verbalizations throughout the session on multiple dimensions, such as immersion, anger, entitlement, and their perception of the effects of alcohol on themselves and their virtual date. Can integrate ecological momentary assessments of thoughts and feelings into the VR in a relatively unobtrusive manner.

Notes: This table provides general information about each type of experimental analogue, both how it has been used in existing research and elements that could easily be added in future research. Some analogues (e.g., written scenarios) have been used by many research teams and there are many differences between them (e.g., relationship between perpetrator and victim). Other analogues have (to our knowledge) only been developed by one research team (e.g., audiotape scenario), although other researchers could adopt this methodology and depict different types of sexual assault situations.

Scenarios. The vast majority of alcohol administration studies that assess participants' sexual aggression propensity require them to read, listen to, or watch a video depicting a sexual assault scenario (see Abbey & Helmers, 2020; Abbey & Wegner, 2015; Davis et al., 2014, for reviews).¹ These studies have found significant main, moderation, and mediation effects of alcohol on men's likelihood of engaging in sexual aggression toward the woman in the story (for reviews, see Abbey et al., 2014, 2022; Crane et al., 2016). The scenarios are high in mundane realism because they are based on victims' and perpetrators' descriptions of sexual assaults (Basile et al., 2021; Davis et al., 2015). Experimental realism is enhanced by encouraging participants to imagine that a situation is happening to them and by using explicit language to increase participants' involvement and sexual arousal.

With one exception, these analogues assess sexual aggression propensity at the end of the story by asking participants how likely they would be to act like the male character. Answering questions about what they would do in that situation requires participants to pause and assign numerical values to their current thoughts and feelings, a reflective process that is unlikely to occur in the heat of the moment. Reflection allows participants' immediate visceral response to diminish and increases the likelihood that self-protection concerns will affect their responses. The exception is the audio scenario in which participants are instructed to press a button when they believe the man “should refrain from making further sexual advances” (Gross et al., 2001, p. 382). Additional strengths, limitations, and possible modifications are described in Table 1.

Behavior directed at a confederate or fictitious other. Classic experimental analogues create a compelling situation that requires an immediate response, such as deciding whether to call for help when smoke billows out from under a closed door even though no one else in the room seems to notice it (Latane & Darley, 1968; bystander intervention paradigm). The goal is to create a situation in which participants act as they would in real life because they do not have time to remind themselves that they are in a study or to think about the most socially appropriate response. Confederates are often used in these classic studies, both to augment the experimenter's cover story and to provide a target for participants' responses. In some studies, participants are told that they are interacting with another person who is in another location (fictitious other) and receive standardized information allegedly from that person.

Very few sexual aggression–alcohol administration studies have included a confederate or a fictitious other. In a study that did not include alcohol administration, Pryor (1987) conceptualized sexual harassment as male participants' use of touch when teaching a female confederate how to play golf or poker. Although the author's hypotheses were supported, we are not aware of this analogue being used in later studies. By current standards, allowing participants to touch another person is unethical because of concerns about the confederate's distress and physical safety. Confederates can be involved in research that evaluates the first phase of Abbey's (2002, 2017) theoretical model by evaluating how participants' alcohol consumption affects their perceptions of a confederate's signs of sexual interest (Abbey et al., 2005).

A few alcohol administration researchers have used a sexual imposition analogue to assess sexual aggression proclivity, finding main effects (Marcantonio et al., 2024) and interactions (Leone et al., 2022) with alcohol. This paradigm was first used as an analogue for participants' interest in sexual erotica (Briddell & Wilson, 1976; George & Marlatt, 1986) and later modified to assess men's willingness to show sexual materials to a confederate or fictitious other whom they are told dislikes this type of material (Dermen, 1990; Franz et al., 2018; Hall & Hirschman, 1994; Maass et al., 2003). An important strength of this analogue is that it requires a direct behavioral response. In most studies, participants are not in the same physical space as the actual or fictitious confederate, they do not communicate with her directly, nor do they see how she responds to the video. Sharing disliked sexual materials with someone seems conceptually closer to technology-facilitated sexual harassment than sexual assault (Abbey, 2023; Angelone et al., 2022; Davis et al., 2014).

Virtual reality analogue strengths and limitations

VR sexual aggression analogues can combine the strengths of scenario and behavioral analogues because they embed participants in a sexual situation that requires them to interact with a fictitious other. They can experience themselves touching another “person” by touching their date in the VR environment, who can be preprogrammed to respond positively or negatively (see later section on ethical issues and Table 1). Participants wear headsets that envelop them in the 3D virtual environment and block out competing external visual and auditory cues (Bengtsson & Van Couvering, 2023). Within that world, participants make choices that have consequences for what happens next, thereby increasing their perceived control (Banakou & Slater, 2014; Nierula et al., 2021; Slater et al., 2010). Social presence is the feeling of being connected to others in the space, whereas physical presence occurs when participants believe they are physically located in that space, not in a sterile laboratory room wearing a headset (IJsselsteijn et al., 2000).

Current VR strategy based on pilot research. Although true VR uses headsets so that participants are completely immersed in a 3D world, our pilot research used a two-dimensional dating simulation in which participants interacted with their virtual date through a high-definition computer monitor and keyboard (Abbey et al., 2018). Participants saw themselves through a first-person perspective (only their hands and legs were visible) sitting on a couch with the woman in her apartment. Participants were encouraged to act as they would on an actual date and could select nonsexual (e.g., talk about various topics, watch TV, you drink water or alcohol, she drinks water or alcohol) and sexual activities. In the pilot studies, some participants never chose to do anything sexual; some only chose acts like kissing and making out, which their virtual date enjoyed; and some chose penetrative sexual acts, which she refused. Among those participants who received a refusal, some never tried that level of sexual activity again and some tried multiple times. Thus, this 2D sexual aggression analogue provided participants with the opportunity to learn from their choices (e.g., she does like this; she does not like that—at least not now) and make their own decisions about how to respond. This is similar to the process that occurs in a real-life situation in which individuals have to negotiate sexual decisions and do not know in advance what the outcome will be. In contrast, current scenario-based analogues provide all participants with the same ending, rather than allowing them to make iterative decisions based on their choices during the study.

Sexual aggression was operationalized as the total number of times participants tried a sexual activity that their virtual date refused (Abbey et al., 2018). This behavioral assessment of persistence despite refusal was significantly positively correlated with past sexual aggression and numerous risk factors associated with sexual aggression in past research including heavy drinking in dating and sexual situations (Abbey et al., 2018).

Strengths of VR and this operationalization of sexual aggression. In our planned 3D VR analogue, as well as the pilot 2D dating simulation analogue, many participants had no reason to think they were in a sexual aggression study because they never made a sexually aggressive response. In vignette analogues, it is usually obvious that the study is focused on sexual aggression because participants are typically required to rate their likelihood of acting like someone who used verbal coercion and physical force to obtain sex. By focusing on sexual decision-making rather than force, this analogue should reduce reactance. This type of analogue also provides information about coping strategies used by men who want sex but do not persist, which could be included in prevention programs.

Most sexual assaults occur during an extended interaction, often beginning in a public location and then moving to someplace more private. Traditional dating and sexual scripts are often followed, with the man initiating sexual activities and the woman expected to make her limits clear (Bogren et al., 2023; Hirsch et al., 2019; Jozkowski et al., 2017; Muehlenhard et al., 2016). Most individuals report that they rely on indirect cues to determine their partner's willingness (Bedera, 2021; Jozkowski et al., 2017; Shumlich & Fisher, 2018). There is substantial literature that describes common refusal strategies and common persistence strategies used in the face of refusal (Edwards et al., 2014; Marcantonio & Jozkowski, 2023; Richards et al., 2022); however, we are not aware of research that describes the sequencing of these events. Thus, another strength of a VR analogue is that researchers have a detailed record of every choice made by participants and the order in which they were made without having to rely on their ability to accurately and honestly report what they did. Participants can be encouraged to talk to their virtual date as they would in real life, providing information about affective states and cognitions that are hypothesized to justify persistence after refusal. In an alcohol administration study using the 2D dating simulation, an alarmingly large number of participants spontaneously insulted and threatened their virtual date (Woerner et al., 2018).

All of this information allows researchers to examine behavioral patterns associated with sexual aggression and to potentially develop different trajectories of response to sexual rejection. For example, are there some men who apologize and seem to really mean it (i.e., turn to other activities and do not try again)? Are there some men who apologize but seem to use it as a manipulative strategy (i.e., try sex again soon?) And what if the apology doesn't work and they are rejected again? Are there common behaviors selected by men who persist, such as encouraging the woman to drink more alcohol? Researchers can assess (preferably in a separate session) participants' past sexual aggression, as well as risk and protective factors to determine if they are differentially associated with different patterns of behavior in the VR analogue. This type of information should be extremely valuable for developing more effective prevention programs.

Limitations. There are important ethical issues that need to be carefully considered when developing a VR environment. One crucial concern is how far to let the sexual activities progress. This was also a concern in our 2D simulation; however, 3D technology raises additional concerns because it allows researchers to show participants their own faces on the male avatar and to see themselves using physical force to obtain sexual intercourse or engaging in intercourse with a woman passed out and unable to respond. We do not allow participants to see their own faces or engage in forced sexual activities out of concern that these experiences could be rewarding and reinforce their sense that this is acceptable behavior. As this technology continues to develop, new ethical concerns are likely to arise.

The costs of VR are also a limitation. Although there are some “drag and drop” products on the market, creating a VR environment requires programming expertise; thus, this research requires a multidisciplinary team with the appropriate skill sets and ability to communicate across their disciplinary boundaries. Our team includes faculty and students in psychology, social work, communication design, and computer programming. Data management expertise is required to handle the large data files produced by VR. Like video scenarios, some VR elements can become outdated, thereby reducing immersion. All of our study's 3D models and source code are being created in a modular fashion and are fully documented so that later updates are possible.² See Table 1 for more information about VR requirements.

Another potential limitation of VR paradigms is the degree of responsiveness that can be provided. There are two ways that participants can interact with another standardized person in VR. One approach relies on a confederate who responds at the appropriate points. In Jouriles et al.'s (2014) sexual assault victimization analogue, female participants sit next to the male confederate but cannot see him because they are wearing a headset that projects them into a car or bedroom with him. This approach allows the confederate to tailor their responses (within the constraints of the script) to reflect the participant's previous response. The disadvantages are the same that exist for any study that uses confederates, which include the intensive training required to ensure that the procedures are properly executed, the challenges with running sessions that must fit participants' and confederates' schedules, and the difficulty of disseminating the paradigm widely. Thus, the most common approach is to include an avatar with preprogrammed responses (commonly known as an agent). There can be elaborate contingencies built into the programming to ensure that the agent's responses are appropriate, but free-flowing conversation is not possible nor are responses that address specific comments made by participants.

Opportunities and challenges associated with embedding alcohol cues in a virtual reality environment

VR analogues provide opportunities for new insights into the psychological and pharmacological effects of alcohol on sexually aggressive behavior. Sex and aggression alcohol expectancies are associated with an increased likelihood of acting like the male character in vignette studies, with some studies producing main effects (regardless of alcohol condition) and others finding these effects only for intoxicated participants (Davis, 2010; Davis et al., 2020; Norris et al., 2002). With VR, researchers can determine if participants with strong alcohol expectancies act more sexually and aggressively toward their virtual date from the start of their interaction or if the expectancies seem to be triggered by sexual refusals. In addition, participants' decisions about how much alcohol (if any) they want to consume and how they act after consuming alcohol in the VR environment can be compared for participants who are randomly assigned to environments that do or do not include alcohol cues such as liquor bottles or characters in the background who are drinking quickly or slowly (Bordnick et al., 2008; Mirijello et al., 2023; Schneider et al., 2022).

Exploration of the instrumental use of alcohol to achieve sexual goals requires an experimental paradigm that allows participants to target an intoxicated woman or to encourage a woman's heavy drinking. VR agents can be programmed to show common signs of intoxication (e.g., laughing, unsteady gait, slurred speech) and for these to become more extreme with increased consumption. These types of cues can be embedded in the description of the incident in scenario-based analogues, and comparisons can be made between the responses of men randomly assigned to interact with a clearly impaired woman or a nonimpaired woman (although these cues are seldom included or manipulated; see Abbey & Helmers, 2020). VR, however, allows participants to control aspects of the situation as they might try to do in an actual sexual assault. There are many possibilities depending on the researcher's focus. For research questions regarding perpetrators who target drinking women, participants could be allowed to interact with different virtual women at a bar who display different levels of impairment and then asked which one they wanted to try to pick up (cf. Tuliao et al., 2018). Or participants could explore a house party in which some partygoers are playing drinking games and their decisions could be assessed regarding whether they joined those games, encouraged others to drink heavily, and/or tried to engage in sexual activities with an impaired virtual woman. For research questions about perpetrators who encourage dates to become intoxicated, the VR environment could allow them to buy their date drinks or encourage her to do shots with them. For research questions about sexual aggression in more established relationships, participants could be watching a movie at home with their partner and have the opportunity to bring them an unrequested alcoholic drink or top off a glass of wine repeatedly. Other substances could also be made available, thereby allowing researchers to examine participants' preference for their use instead of or in combination with alcohol.

Comments from focus group participants have encouraged us to consider how the female agent responds when her date gives her a drink she did not request. Is it reasonable for her to just smile and take the drink or should she sometimes refuse or show reluctance, particularly if participants give her a large number of drinks in quick succession? We have not yet resolved this issue because of the competing goals of wanting the woman to respond realistically and allowing participants to do what they really want to do. Although we want the female agent to act realistically, findings from the pilot 2D dating simulation suggest that complete realism may not be necessary (Abbey et al., 2018). Participants could choose “She drinks some beer.” When selected, the woman took a sip from a bottle of beer on the coffee table; she never refused or complained although some participants made this choice many times in succession. The number of drinks participants gave her was significantly, positively correlated with narcissism, sexual dominance, lifetime number of sexual partners, sexual alcohol expectancies, and frequency of drinking in dating and sexual situations (assessed a month earlier); with the number of lower level consensual sexual activities selected in the VR and how often they attempted higher level sexual activities that she refused (operationalization of sexual aggression); and with their report of how much she was playing hard to get and how mad she made them (after the session ended). It is also noteworthy that although the agent was not programmed to act differently after drinking, many participants reported that she began to relax and act differently after they gave her alcohol to drink, demonstrating the power of expectancies. Future research could code verbalizations that map onto sex and aggression alcohol expectancies to explore their role in perceptions of the woman and as justifications for their actions.

Recommendations for future research

Validity data support the use of the analogues discussed in this article, although the magnitude of the associations between responses to the analogues and self-reports of sexual aggression are typically not large and leave a great deal of variance unexplained (Abbey & Wegner, 2015; Angelone et al., 2022; Davis et al., 2014; Franz et al., 2018). Some risk factors have been identified across a wide range of perpetrators, which suggests that a core group of perpetrators might be identified with any well-validated analogue (Waterman & Edwards, 2022). Researchers have also emphasized that past perpetrators do not engage in sexual assault at every possible opportunity, which suggests that different analogues may be required to identify different types of perpetrators and the situational factors that embolden them to act (Abbey, 2023; Swartout et al., 2022). There is a pressing need for more nuanced etiologic research to guide the development of prevention and treatment programs. We strongly encourage more dialogue between researchers regarding definitions of key concepts, identification of essential elements, and use of multiple analogues. We are not aware of experiments that randomly assign participants to different sexual aggression analogues to determine if the patterns of associations with risk factors are comparable. If such studies find comparable results across multiple types of analogues, that greatly increases confidence in the generalizability of the findings. If such studies find sizable differences between analogues, then it is crucial to identify which analogues best predict which forms of sexual aggression, for which types of individuals, and in which type of situations.

Concluding note

VR is being used in sexual violence bystander prevention programs (Porta et al., 2024), risk recognition secondary prevention programs (Jouriles et al., 2014), and domestic violence treatment programs (Seinfeld et al., 2018). VR provides the opportunity to practice identifying sexual partners' nonverbal signs of discomfort and to develop sexual negotiation skills privately without fear of embarrassment. Alcohol administration studies with adults provide a controlled environment for determining when and how intoxication undermines these skills. For adolescents, choices they make about how much alcohol they want to consume and how much they want their dating partner to consume on a VR date could provide them with insights about alcohol's effects without having to actually experience them (cf. Hadley et al., 2014). In sum, there are many opportunities for researchers to use VR to develop evidence-based interventions.