Beyond Active/Passive: Using the Interpersonal Circumplex to Re-Conceptualize Victim Behavior During Violent Crime

ABSTRACT

This study used the interpersonal circumplex (IPC) – a probabilistic model of human behavior during social interactions – to conceptualize victim behavior during physically assaultive crime (domestic violence, physical assault, and sexual assault). Using data from the Australian Database of Victimisation Experiences, 101 victim behaviors were identified across a sample of 150 victim narratives. Categorical Principal Components Analysis and Smallest Space Analysis found that victim behavior during physically assaultive crime aligned with the IPC’s four behavioral styles: dominance, submissive, hostility, and cooperation. These findings provide new opportunities to explore victim agency including how victims may influence the offender and situation.

KEYWORDS:

• Victim behavior
• violent crime
• offender-victim interaction
• interpersonal circumplex
• smallest space analysis
• categorical principal components analysis

Introduction

Researchers have long acknowledged the dynamic and evolving nature of violence – with multiple pathways into, and out of, a violent incident. Wikström’s (2004, 2014) Situational Action Theory, for example, argues that the motivation to commit crime and violence stems from the interaction between an individual’s internal characteristics and specific features of the setting in which it occurs – meaning it will be different in every case. This is echoed in Thomas et al.’s (2022) recent study of neighborhood characteristics, which showed how variations in disadvantage, disorder, and access to legitimate opportunities for success influenced offenders’ willing to engage in, and perceptions of the benefits to committing, crime. Similarly, during the crime itself, the interacting elements influence how the violence unfolds. Luckenbill (1977) famously characterized homicide as a “situated transaction” where fatal outcomes were shaped by both the offender and victim and their reactions to each other, while gender, boredom, binge drinking, aggression, overcrowding, irresponsible service of alcohol, and interpersonal tension have all been linked to both the likelihood and nature of violence in the night-time economy (see Cozens & Grieve, 2014; Finney, 2004; Homel & Clarke, 1995; Macintyre & Homel, 1997; Savard et al., 2019; Swan, 2021). As such, it is important to acknowledge that understanding how crime emerges and unfolds requires a broader focus that moves beyond discrete aspects of an incident – such as just the offender or features of the situation – and instead explores the interaction between these elements (Sullivan et al., 2012). However, this level of analysis requires an accurate way to capture and conceptualize key elements of crime to ensure the interaction can be appropriately explored.

One such element that is often overlooked is the role of the victim and their behavior (Fuller, 2015; Karmen, 2004). Victim behavior can vary dramatically across different incidents, crime types and between individuals. Where one victim may yell at or fight with an offender, another might freeze, comply, or run away. Similarly, a victim may engage in a range of different behaviors across a single incident – mixing or alternating resistive and non-resistive actions as the situation unfolds (Jacques & Rennison, 2012). Prior research has provided valuable insight into how victims behave in specific circumstances but has yet to examine if patterns in how victims respond to offenders can be described beyond active/passive, resistance/no resistance dichotomies (discussed below). Such a knowledge gap has implications for how we view victims of crime while also inhibiting a deeper understanding and acknowledgement of victim agency during crime. Victim agency refers to the extent to which victims are deliberately and intentionally attempting to influence the crime in reference to some underlying, self-motivated goal. This is in line with Bandura’s (1999) Social Cognitive Theory which characterizes humans with agency as “self-organizing, proactive, self-regulating and self-reflecting” individuals who operate within a broad network of socio-structural influences (Bandura, 1999, p. 21). If acting with agency, victims would play a key role in shaping and influencing the nature and outcome of a crime; however, accurately capturing the full range of behaviors across different incidents is challenging. It is important to note that this perspective is not suggesting that victims are responsible for the outcome or how the crime unfolds. The offender chooses to commit violence and is therefore responsible for its outcome, regardless of how it manifests (i.e., completed or attempted; victim injured or not). Instead, it is about recognizing and capturing victim behavior as a key element during violent interactions. Failing to recognize and examine how victims respond not only impedes our understanding of how violence unfolds but may also contribute to misinformation about “normal” victim behavior during crime and promote negative, harmful stereotypes and victim blaming (Kearon & Godfrey, 2007; Spalek, 2006; Walklate, 2012).

Previously, researchers have used active/passive dichotomies to classify victim behavior (see Bachman et al., 2002; Block, 1981; Bovin & Marx, 2011; Papendick & Bohner, 2017; Rizvi et al., 2008). Active behaviors refer to physical or verbal actions that resist the offender or protect the victim; for example, kicking, swearing, screaming, or fighting back. Alternatively, passive responses occur in the absence of these more overt resistive or protective behaviors, such as when victims cried, did not react in any specific way or “went along with” the offender (Bovin & Marx, 2011). “Freezing” and “tonic immobility” – characterized by an inability to move, reduced ability to feel cold or pain, and unresponsiveness to stimuli – are extreme examples of passive, automatic victim responses (Coxell & King, 2010; Hagenaars et al., 2014; Humphreys et al., 2010; Marx et al., 2008). Active/passive classifications have been used in studies of sexual assault or rape to enable researchers to examine how victim behavior, specifically resistance, impacts the outcome of crime via mutually exclusive categorizations.

Yet a key weakness of active/passive classifications is that victim behavior is categorized with little consideration of the motivation or goals that may underpin it – implying only “active” victims have agency, while “passive” victims do not. For example, active/passive classifications fail to accurately describe behavior in situations where the victims deliberately choose not to resist the offender in the interest of protecting and minimizing harm to themselves or others. Such responses have been documented in, for example, Meyer’s (2016) examination of victim responses to domestic violence and Smith’s (2017) study of taxi drivers’ responses to robbery. In situations where victims are consciously aware of their behavior, they may consider it more beneficial to submit to the offender rather than resist. Classifying such behavior as “passive” fails to acknowledge the deliberate decision, or agency, behind the action. This necessitates a more flexible and nuanced way to conceptualize victim behavior, one that goes beyond simple dichotomies and describes the full range of victim behaviors, whilst providing insight into the possible reasons underpinning victims’ actions.

One model that has been used to understand human behavior during interactions is the interpersonal circumplex. The interpersonal circumplex describes and models behavior during social interactions with reference to two intersecting dimensions: dominance to submission and hostility to cooperation. In addition to describing the range of behavior displayed during any social interaction, it also presents hypotheses about how behavior influences and shapes subsequent reactions. As a probabilistic model of human behavior, the model can be used to make inferences regarding individual behaviors, the likely response, and an actor’s possible intent when engaging in actions associated with a particular behavioral style. It has been previously used to conceptualize offender actions during crime (see Alison & Stein, 2001; Bennell et al., 2001; Porter & Alison, 2004, 2006) and could provide a comprehensive way to categorize victim behaviors. Its application to victims has so far been limited to the small numbers of victim responses included in studies of offender behavior during violent crime (see Alison & Stein, 2001; Bennell et al., 2001; Porter & Alison, 2004, 2006; Woodhams et al., 2020, discussed in detail below). The present study aims to expand on this research by identifying the range of victim behaviors present in victims’ accounts of physical assault, sexual assault, and domestic violence and examining the extent to which they align with the interpersonal circumplex. By doing so, this research seeks to identify a more nuanced way to conceptualize victim behavior. This is necessary to create new opportunities to enhance our understanding of how crime unfolds and contribute to the theoretical perspectives around victim agency.

The interpersonal circumplex

The interpersonal circumplex (also known as the interpersonal circle; hereafter referred to as the IPC) models the relationship between two or more people’s actions and can be used to understand the structure and progression of an interaction (Plutchik & Conte, 1997; Wiggins, 1996). Behavior is classified according to two bisecting axes: one that represents power and the other, affiliation (Figure 1). An action’s position relative to both axes signifies the behavioral styles it aligns with. The vertical power axis captures behaviors used to control the other person and extends along a continuum from dominance to submission (Leary, 1957; Wiggins, 1996). Dominance is defined as the expression of control while submission is the yielding of it. Dominant behaviors include making decisions and taking the lead while submissive behaviors include waiting for others to act and acquiescing (Moskowitz, 1994). The horizontal affiliation axis represents an action’s “friendliness” and ranges from hostility to co-operation (Kiesler, 1983). Hostility is defined as opposing and/or antagonistic behaviors such as aggression, insults or making demands. Alternatively, co-operation is defined as behaviors that invite collaboration and/or support such as showing affection, being agreeable or negotiating/compromising (Moskowitz, 1994). By arranging behaviors along these two axes, the IPC can be used to describe the underlying themes present in a human interaction. The IPC’s circular structure of interpersonal behavior has been empirically tested using a variety of methods in different contexts including dispositional mindfulness (Deits-Lebehn et al., 2019), spousal relationships (Smith & Williams, 2016) and drug dependency (Minges et al., 2020).

Figure 1. The interpersonal circumplex.

The IPC also models the relationship between the four behavioral styles and how they influence and shape those of the other people in the interaction. The principle of complementarity proposes that specific actions expressed during an interaction are at an increased likelihood of being met with a complementary responding behavior (Kiesler, 1983; Markey et al., 2003). It manifests as reciprocity on the power axis (i.e., dominant behavior is likely to elicit a submissive response and vice versa) and correspondence on the affiliation axis (i.e., co-operative behavior is likely to elicit a co-operative response while hostility is likely to elicit a hostile response). As the IPC is probabilistic, not deterministic, the principle of complementarity reflects the likelihood of a particular response based on a specific type of behavior as opposed to accurately predicting it in every case (Kiesler, 1983). In the context of the current study, if the IPC is an appropriate model for understanding the styles of victim behavior, the principle of complementarity may provide insight into why certain actions are present, as victims will either be reacting to, or attempting to influence, the behavior of the offender.

The IPC and victim behavior

Researchers have previously used the IPC to model offender behavior and victim reactions during crime (see Alison & Stein, 2001; Bennell et al., 2001; Hauffe & Porter, 2009; Porter & Alison, 2004, 2006; Woodhams et al., 2020; Woods & Porter, 2008). These studies demonstrated the utility of using the IPC to understand the offender-victim dynamic; however, while some victim actions appeared to reflect dominance, submissive, hostile, and cooperative behavioral styles, the focus on offender behavior meant only a small number of victim behaviors were included for analysis. The following sections summarize what is currently known about how victim behaviors map to the IPC behavioral styles, from the studies cited above. It should be noted that the bisecting structure of the IPC means that a single behavior reflects elements of both axes as opposed to just one – for example, dominance and hostility or cooperation and submission. As such, different studies classified some similar behaviors in slightly different ways.

Dominance-submission

Prior research has examined how dominant and submissive behavioral styles, and their reciprocal relationship with each other, have manifested in offender-victim interactions during sexual assault, child sexual abuse and robbery. While the type and nature of behaviors included in each study varied, dominant offender behaviors that were consistent across three or more studies included binding, gagging, or blindfolding the victim (Alison & Stein, 2001; Porter & Alison, 2004, 2006; Woodhams et al., 2020; Woods & Porter, 2008). Additional dominant offender behaviors included kidnapping the victim (Woodhams et al., 2020; Woods & Porter, 2008), intoxicating the victim (Woods & Porter, 2008) and lying-in wait for the victim (Alison & Stein, 2001). Victim behaviors that co-occurred with dominant offender behavior included performing sex acts on the offender (Bennell et al., 2001) and freezing, disassociating, or otherwise not responding (Woodhams et al., 2020). Using the principle of complementarity, the researchers concluded that, due to their correlation with dominant offender behaviors, these behaviors likely reflected victim submission.

Fewer studies have examined the alternate relationship between victim dominance and offender submission during crime. In their respective studies of sexual assault and group robbery, Alison and Stein (2001) and Porter and Alison (2006) stated offenders were unlikely to behave submissively as both crimes required the offender to act in an overtly dominant manner. Both noted, however, that offenders may engage in “pseudo-submissive” behaviors, such as pretending to need help to trick the victim, gain their trust and/or give them the appearance of control; or, demeaning or insulting the victim – with the latter potentially signifying the offender’s loss of control. Bennell et al. (2001) similarly identified child sexual assault offenders who engaged in non-dominant behaviors such as spending time with the victim prior to engaging in the crime (also known as a confidence approach), fondling/kissing the child, showing affection, providing reassurance, or minimizing or downplaying the situation. They concluded that offenders used these behaviors to exploit the child’s trust by appearing to allow the victim a degree of autonomy or control during the abuse. Collectively, these studies showed offenders, either deliberately or otherwise, may behave submissively during crime, which could elicit dominant victim responses in return.

This is supported by studies that found evidence of victim dominance. Porter and Alison (2004) studied 223 cases of sexual violence committed by groups of offenders in the United Kingdom. They found submissive offender behavior such as demeaning the victim or using a confidence approach correlated with victims struggling, refusing to comply, or running away. Based on the context, the authors concluded that these offender behaviors represented offender submission because the victim behaviors reflected a more dominant behavioral style (Porter & Alison, 2004). More recently, Woodhams et al. (2020) coded victim/offender behaviors during multiple perpetrator – also known as gang or group – rapes to understand the prevalence of the IPC behavioral styles during sexual assaults. Dominant victim behaviors included struggling with the offender, giving orders, actively refusing the offender, seeking help, and running or walking away. Submissive offender behaviors aligned with those in Bennell et al.’s (2001) study and included letting the victim go, using a confidence approach, untying the victim or removing their gag, giving the victim information, and minimizing the situation. Of interest, results showed that dominance was the most common victim behavioral style and accounted for 57% of the 1,031 victim actions in the 71 incidents of multiple perpetrator rape (Woodhams et al., 2020).

Hostility-cooperation

Unlike the power axis, on the affiliation axis cooperative behaviors are most likely to elicit a similar, cooperative response while hostile behaviors are likely met with hostile responses. According to Woodhams et al. (2020), hostile behaviors accounted for 17% of offender actions during group sexual assault compared with only five percent of victim actions. Though the specific behaviors comprising each style have varied across the different studies, common hostile offender behaviors present in three or more studies included using weapons for harm and removing the victim’s clothing (see Hauffe & Porter, 2009; Porter & Alison, 2004, 2006). Other examples included beating the victim (Woods & Porter, 2008) and hitting, kicking, pushing, burning, or cutting the victim (Hauffe & Porter, 2009). Fighting back was the most common victim behavior that co-occurred with hostile offender actions (Hauffe & Porter, 2009; Porter & Alison, 2004; Woods & Porter, 2008). Additionally, Porter and Alison’s studies of group rape (2004) and group robbery (2006) found that, in line with the complementarity principle, hostile victim behaviors such as fighting, raising the alarm, or screaming were correlated with hostile offender behaviors such as violence, using weapons or removing/tearing victims’ clothes.

At the opposite end of the axis, cooperative behaviors aim to facilitate an accommodating or compliant response from the other party in the interaction. Co-operative offender behavior has also been termed “compliance-gaining” and may capture pseudo-intimate behaviors. For example, studies have identified offenders who compliment their victims, apologize, ask questions about the victim, spend time with the victim beyond the sexual assault, give gifts, masturbate, or perform oral sex on the victim (see Bennell et al., 2001; Porter & Alison, 2004, 2006; Woodhams et al., 2020). Porter and Alison (2004, 2006) found that not all co-operative offender behaviors were conciliatory. In their studies of group rape (2004) and group robbery (2006), offender behaviors including threatening the victim not to report or to comply, single acts of violence or using a weapon to control co-occurred with more cooperative behaviors. These behaviors were hypothesized to reflect dominance, yet the authors stated offenders were possibly aiming to elicit victims’ active compliance and cooperation through these swift and effective measures as opposed to trying to obtain complete passive submission (Porter & Alison, 2004, 2006). Porter and Alison (2004) noted that identifying cooperative victim actions was difficult due to the lack of sensitivity in their coding. Specifically, that offender compliance-gaining behavior, such as forcing victims to remove their clothing, was not always distinct enough from the corresponding victim behavior (i.e., victim removes clothing) to infer actual cooperation. They then hypothesized that behaviors, such as pleading and crying, were victims attempting to elicit compliance from the offender as opposed to simply doing as they were instructed. Woodhams et al. (2020) expanded this to include other victim behaviors such as negotiating, commenting, making requests, and justifying behaviors. They found that cooperative behaviors accounted for 23% of victim behaviors during multiple perpetrator rapes compared with 14% of offender behaviors.

The present study

As victim behavior during crime can be so diverse, any theoretical model designed to understand it needs to be both flexible and nuanced. Active/passive dichotomies lack this flexibility and nuance by failing to account for victim motivations and, as a result, potentially misrepresent some behaviors. The IPC may provide a useful theoretical model as it can describe victim behavior with reference to the individual’s goals regarding power and control (dominance-submission) and level of affiliation (hostility-cooperation). Prior research suggests the IPC could be applicable to victim behavior; however, the model has yet to be applied to the full range of victim actions across different crime types. For the IPC to be relevant, the range of victim behaviors and actions expressed during crime needs to align with the four behavioral styles of dominance, submission, hostility, and co-operation. The present study first aimed to catalog the range of victim behaviors present in three crime types: domestic violence, physical assault, and sexual assault (hereafter collectively referred to as physically assaultive crimes). Second, the study aimed to determine the extent to which victim behavior during these crime types aligns with the four IPC styles. Including these three crimes ensured the study could capture a range of victim responses to better understand the applicability of the IPC to victim experiences across different incidents and situations. Categorical Principal Components Analysis and Smallest Space Analysis were used to statistically explore the underlying thematic structure of victim behaviors and establish whether the IPC provided an appropriate theoretical model to begin to understand how victims behave during physically crime and provide possible underlying motivations.

Method

Sample

The sample comprised victim experiences of physical assault (PA), domestic violence (DV) or sexual assault (SA). These crimes involve a violent interaction between victims and offenders and therefore share similarities in terms of victim experiences. The sample comprised 150 victim narratives of physically assaultive crime. Sixty-five percent of cases (n = 97) involved PA, 22% (n = 33) involved DV and 13% (n = 20) SA. Exactly 50% of cases involved male victims and 50% female victims, although gender proportions varied by crime type; for example, approximately 70% of DV and SA victims were female compared with 38% of victims of PA (see Table 1).

Table 1. Sample characteristics.

	Domestic violence	Sexual assault	Physical assault	Total
Victim gender
Female	24 (73%)	14 (70%)	37 (38%)	75 (50%)
Male	9 (27%)	6 (30%)	60 (62%)	75 (50%)
Victim age
16–19	3 (9%)	4 (20%)	7 (7%)	14 (9%)
20–29	8 (24%)	2 (10%)	16 (16%)	26 (17%)
30–39	14 (42%)	6 (30%)	22 (23%)	42 (28%)
40–49	0 (0%)	2 (10%)	15 (15%)	17 (11%)
50–59	4 (12%)	1 (10%)	17 (17%)	22 (15%)
60 years or older	1 (3%)	1 (10%)	4 (4%)	6 (6%)
Not stated	3 (9%)	4 (20%)	16 (16%)	23 (15%)
Total	33	20	97	150

Percentage totals may not add to 100 due to rounding.

Data collection

This study utilized data from the Australian Institute of Criminology’s Database of Victimisation Experiences (DoVE). The DoVE is a qualitative database comprising 730 psychological evaluations of victims of crime who sought compensation from Victim Services, New South Wales (NSW) between 2005 and 2010. It includes four crime types: SA, PA, DV, and robbery (see Fuller (2015) for more information on the DoVE, its sample, and methodology).

The present study focused on exploring victim behavior during physically assaultive crime. The authors chose to exclude the 193 cases of robbery to increase the consistency in the victim experiences included in the final sample. Specifically, Youngs et al. (2016) distinguish between instrumental and expressive or hostile aggression. The former occurs when offenders are motivated to obtain something possessed by another person – such as money, property, or status. Alternatively, expressive aggression occurs in response to an “anger inducing situation,” with the offender’s primary purpose being to harm or injure the victim (Youngs et al., 2016, p. 403). PA, DV and SA are more likely to involve expressive violence compared with robbery where the aggression may be more instrumental (Cohn & Rotten, 2003). This may lead to fundamental differences in victim experiences. Further, robbery can involve property stolen under threat of violence, and so therefore may not actually involve a physical assault. In comparison, while the other three crime types share similarities in the nature of the violence, they also differ in some specific ways. PA, for example, involves non-sexual violence against the victim while SA involves sexual violence; DV commonly involves physical violence but may also include sexual violence. Including these three crimes provided greater scope to explore a range of victim behaviors.

The remaining 537 cases describing victim experiences of PA, DV or SA were subsequently extracted from the DoVE. Cases were then evaluated for inclusion in the final sample according to three criteria. First, 229 cases were excluded because the victim was under 16 years of age at the start of the offence. Research has shown children’s capability to accurately recall their experiences can vary (Goodman et al., 2019). Specifically, being older at the time of trauma, as well as the severity of the experience, have been found to increase the accuracy of recall at later times (see Goodman et al., 2019 for a discussion). The decision to exclude victims under the age of 16 years therefore aimed to ensure the consistency in victim recollection.

Second, 131 cases were excluded because the reports did not describe or detail specific victim behaviors, as this is the focus of the present study. Finally, 37 cases were excluded because the report did not contain reference to a discrete incident of victimization. Six reports that contained multiple discrete incidents were split to ensure each case only contained one incident. This resulted in an additional ten cases being added to the sample resulting in a final sample of 150 cases included for analysis. This process is summarized in Figure 2.

Figure 2. Case selection and exclusion criteria.

The DoVE has two limitations that need to be acknowledged. First, it only captures the experiences of victims who sought compensation from Victims Services NSW. There may be specific differences between these individuals and their experiences and those of victims who do not report to, nor engage with, government services or seek compensation. Second, the data were derived from psychological evaluations compiled for the purpose of compensation rather than research. Thus, the psychologist’s primary goal when conducting the interview was to assess the incident’s impact on the victim’s physical, psychological and social functioning as opposed to obtaining comprehensive details of the crime. This impacted consistency in the way the information was collected and the focus of each report. It is therefore important to note that, while unlikely, the lack of information – for example, the absence of a victim behavior – did not necessarily indicate its absence from the victim’s experience, only that it was not discussed with the report writer. Despite these limitations, the DoVE was still a valuable data source for this study as the rich qualitative information provided unique perspectives and insights into how victims behave during crime.

Data coding

Initially, a content analysis of victim narratives was conducted to identify the full range of victim behaviors present in the data set. This process was conducted using NVIVO 11. Analysis aligned with Elo and Kyngas (2008) three stage procedure: preparation, organization, and reporting. During preparation, the first author selected a unit of analysis and “immersed” themself in the data to gain a thorough and accurate understanding of the context in which the information is presented (Elo & Kyngas, 2008). The unit of analysis was the incident, and the units of observation were the individual behaviors reported by the victims during their psychological evaluation. Each case was read in depth to gain a full understanding of the types of victim behaviors present in the sample and the context in which they occurred. During the organization phase, content was coded into discrete variables reflecting individual actions using an unconstrained matrix.

Initially, 101 different victim behaviors were identified across the 150 cases of physically assaultive crime. Twenty behaviors were excluded then because they were not interpersonal and did not occur between the offender and victim; for example, falling over, losing consciousness, or talking to police once the offender had left the scene. As the IPC describes behavior during interactions, these actions were unsuitable for inclusion. As 49 of the remaining 80 behaviors occurred in less than five cases, they were aggregated into broader variables based on their similarity. This process resulted in the final sample of 18 victim behaviors that were then coded as present or absent in each of the 150 cases (see Table 2 for descriptions and frequencies, as well as examples of behavior aggregation). A random sample of 15 cases (10% of the total sample) were coded by an independent coder to test inter-rater reliability. Cohen’s Kappa showed there was substantial agreement between the two coders (k = 0.713, 95%CI, 0.635–0.774, p < .001). Specific disparities in coding were resolved through discussion until agreement was reached and did not result in substantial changes to the list of behaviors generated.

Table 2. Victim behavior during physically assaultive crimes.

Hypothesized IPC style	Behavior	Frequency (% cases)	Description	Example behaviors
Dominance	Leave	49 (35)	Victim removed themselves from the situation	Walk away; run away; leave
	Defend	41 (27)	Victim defended themselves or other	Protect self or other; resist offender; direct other to protect themselves
	Evade	31 (21)	Victim avoids the offender	Retreat; hide; evasive action; lock the door; move away from offender (but not leave scene)
	Hard no	20 (13)	Victim said no with some force	Command offender; tell offender to go away; protest
	Help	16 (11)	Victim called for help	Call for help
Submission	Do nothing	24 (16)	Victim did not physically respond to offender	Not resist; not stop offender
	Sex	21 (14)	Victim experienced a sexual act involving the offender	Sexual intercourse; oral sex; watch offender masturbate; sex with offender not specified.
	Soft no	8 (5)	Victim said no, less forcefully	Ask offender to stop; said no
	Uncontrollable response	8 (5)	Victim expressed an uncontrollable response	“Freak out” [sic]; freeze; immobilized; “space out” [sic]
	Cower	8 (5)	Victim cowered from offender	Cry; curl into fetal position; not look at offender; pretend to sleep
Hostility	Fight	34 (23)	Victim physically fought with offender	Struggle; wrestle; tackle offender; fight; kick; attack with weapon; restrain offender/other; punch; bit offender
	Argue	22 (15)	Victim verbally argued with offender or bystander	Argue; swear; yell at offender; scream at offender
	Uncooperative	22 (15)	Victim refused to comply with offender	Not co-operate with offender; refuse to do as told; escape
	Confront	14 (9)	Victim confronts or threatens offender	Confront the offender; told offender off; threaten offender with weapon; break down a door
	Scream	11 (7)	Victim screamed	Scream or shout (not at offender)
Co-operation	Co-operate	27 (18)	Victim cooperates with or acquiesces to offender	Give help to offender or others; remove clothes; open door
	Talk	26 (17)	Victim engaged in conversation with or questioned offender	Talk to offender; question offender
	Negotiate	11 (7)	Victim attempted to find a way to compromise with offender	Negotiate with offender; offer offender money; calm offender; beg

Hypotheses were then generated as to which of the four IPC styles each of the 18 behaviors represented and, for ease of interpretation, behaviors were allocated to the one style they were thought to align with most strongly. Hypotheses were based on the definition of each style, conceptualizations from prior research using the IPC to explore offender and victim behavior, as well as the context in which the behaviors commonly occurred in the sample cases. Five behaviors were hypothesized to reflect victim dominance: leave, defend, evade, hard no, and help (see Table 2). These behaviors were grouped because they reflected how victims maintained, or attempted to maintain, control over the offender or outcome of the crime. For example, leaving or evading the offender are examples of how victims attempted to end the interaction with the offender. Similarly, defending themselves or others, calling for help, forcefully telling the offender “no” (i.e., hard no) or commanding them to stop, involve victims issuing commands – consistent with broader definitions of dominance during social interactions. These hypotheses also aligned with previous research. For example, Woodhams et al. (2020) classified victim behaviors such as running away, help-seeking, directly declining and walking away as dominant.

In contrast to dominance, five behaviors were hypothesized to represent a submissive victim behavioral style. These were: not reacting (referred to as doing nothing to stop the offender), sexual contact with the offender, saying no without force, cowering, and uncontrollable responses such as freezing. These behaviors were grouped as they reflected victims’ yielding to the control of the offender. For example, victims commonly emphasized that sexual contact with the offender occurred in the context of their unwillingness and capitulation while cowering. Similarly, these behaviors were similar to Woodhams et al.’s (2020) groupings where submissive victim behaviors included no resistance, freezing and dissociating, and avoiding looking at the offender.

Five victim behaviors were hypothesized to reflect hostility: fighting and arguing with the offender, being uncooperative, confronting the offender and screaming. These behaviors were characterized by high degrees of victim aggression. They also aligned with similar behaviors identified in prior research as representing victim hostility (see previous discussion of Porter & Alison, 2004, 2006; Woodhams et al., 2020). Finally, the three hypothesized cooperative behaviors were co-operating or negotiating with, and talking to, the offender. These behaviors were similar in that they reflected victims either complying with the offender or trying to get the offender to comply. For example, negotiating included trying to calm the offender or offering money in return for behaving in a particular way. Overall, three behaviors hypothesized to be dominant (leaving, defending, and evading) and one hypothesized hostile action (fighting) were each present in over 20% of cases. Alternatively, three hypothesized submissive responses (soft no, uncontrollable response, and cower) were the least frequent – being only present in five percent of cases.

Data analysis

Behaviors that share an underlying theme, or style, should be related and co-occur more frequently than behaviors from different themes. Thus, the utility of the IPC as a model of victim behavior during physically assaultive crime would be reflected in the degree to which the hypothesized categorizations, outlined above, are statistically supported. Further, the two-axis structure of the IPC dictates that dominant behaviors should be spatially opposite to submissive behaviors, and hostile behaviours be spatially opposite cooperative ones.

Analysis comprised a two-stage process, utilizing Categorical Principal Components Analysis (CATPCA) and Smallest Space Analysis (SSA). CATPCA was first used to statistically identify the relationships between the 18 victim behaviors and whether they could be grouped according to the hypotheses in Table 2. Second, SSA was used to complement the CAPTCA by providing a visual representation of each behavior in space. Both methods are detailed further below. The combination of CATPCA and SSA enabled the current research to leverage the strengths of each to explore the relationships between, and structure of, victim behaviors during physically assaultive crime. As noted previously, the bisecting nature of the IPC axes means individual behaviors will reflect elements of both the power and affiliation axes; for example, behavior could be dominant/hostile or submissive/cooperative. This therefore necessitated a degree of statistical flexibility to be able to assess each behavior with regard to both axes. CATPCA identifies underlying dimensions, or components, as opposed to discrete categories of variables, with the component scores providing some insight into how strongly each behavior relates to each component. SSA complements the use of CATPCA by providing a visual representation of the relational structure underpinning victim behavior during physically assaultive crime. Thus, using these two methods in combination captured somewhat different aspects of the IPC, while also allowing the reliability of the structure to be tested. It also facilitated a nuanced examination of the utility of the model and victim behavior across physically assaultive crime.

Categorical principal components analysis

CATPCA is a non-linear equivalent to Principal Components Analysis that is appropriate for use with datasets that include nominal variables. It was expected that, if victim behavior aligned with the IPC, then the CATPCA would yield four underlying components in the data, reflecting the four groupings hypothesized in Table 2.

CATPCA identifies the relational structures between variables by finding the smallest number of components that explain the greatest amount of variance within a dataset (Linting & van der Kooij, 2012). However, the low frequency of victim behaviors across the 150 cases impacted the robustness of CATPCA. The victims’ narratives contained in the DoVE often focused on the impact of the crime or describing the offender’s behavior as opposed to their own. While Pearson’s correlation was used to explore the relationship between variables, only 43% of cases (n = 65) contained references to more than three victim behaviors. This ultimately diminished the strength of the correlations between the 18 behavioral variables (see Table 3) and impacted how they related to each other in the CATPCA model.

Table 3. Correlations.

	Leave	Defend	Evade	Hardno	Help	Donot	Sex	Softno	Uncont	Cower	Fight	Argue	Uncoop	Cnfrnt	Scream	Coop	Talk	Negoff
Negoff	0.022	0	−0.017	−0.035	−0.097	−0.053	−0.02	−0.067	−0.067	−0.067	−0.03	−0.117	−0.117	0.086	0.019	0.134	0.074	1
Talk	0.094	−0.004	0.158	0.079	0.013	0.184*	0.059	0.048	−0.109	−0.03	−0.038	−0.09	0.208*	0.035	−0.061	0.198*	1	.
Coop	0.081	−0.171*	−0.068	−0.031	−0.106	0.032	0.161*	0.043	0.043	−0.034	−0.088	−0.096	0.002	−0.031	0.001	1	.	.
Scream	0.077	−0.173*	−0.08	−0.11	0.234**	−0.053	−0.101	0.047	−0.067	0.047	0.031	0.028	0.028	0.086	1	.	.	.
Cnfrnt	−0.028	0.009	0.063	0.076	−0.037	0.048	−0.115	−0.076	0.026	−0.076	0.045	−0.068	0.061	1	.	.	.	.
Uncoop	0.113	0.084	0.207*	0.059	−0.021	0.076	0.09	0.153	−0.015	0.069	0.091	−0.012	1	.	.	.	.	.
Argue	0.073	−0.043	0.068	0.115	0.04	−0.181*	−0.148	0.069	−0.015	−0.015	0.271**	1	.	.	.	.	.	.
Fight	−0.105	0.061	0.117	0.116	0.019	−0.193*	−0.143	0.084	0.155	0.013	1	.	.	.	.	.	.	.
Cower	0.024	−0.079	0.172*	0.081	0.014	0.058	0.383**	0.472**	0.076	1	.	.	.	.	.	.	.	.
Uncont	0.151	−0.012	−0.048	0.081	0.014	0.139	0.102	0.208*	1	.	.	.	.	.	.	.	.	.
Softno	0.024	−0.012	0.099	0.169*	0.014	−0.023	0.290**	1	.	.	.	.	.	.	.	.	.	.
Sex	0.02	−0.078	0.077	0.107	−0.124	0.131	1	.	.	.	.	.	.	.	.	.	.	.
Donot	−0.071	−0.104	−0.043	−0.011	−0.033	1	.	.	.	.	.	.	.	.	.	.	.	.
Help	0.128	−0.067	0.037	−0.008	1	.	.	.	.	.	.	.	.	.	.	.	.	.
Hardno	0.103	−0.021	0.139	1	.	.	.	.	.	.	.	.	.	.	.	.	.	.
Evade	0.066	0.093	1	.	.	.	.	.	.	.	.	.	.	.	.	.	.	.
Defend	0.019	1	.	.	.	.	.	.	.	.	.	.	.	.	.	.	.	.
Leave	1	.	.	.	.	.	.	.	.	.	.	.	.	.	.	.	.	.

*Correlation is significant at the 0.05 level (2-tailed). **Correlation is significant at the 0.01 level (2-tailed). Variable key: Leave = Victim removed themselves from the situation; Defend = Victim defended themselves or other; Evade = Victim avoids the offender; Hardno = Victim said no with some force; Help = Victim called for help; Donot (do nothing) = Victim did not physically respond to offender; Sex = Victim experienced a sexual act involving the offender; Softno = Victim said no, less forcefully; Uncont (uncontrollable response) = Victim expressed an uncontrollable response; Cower = Victim cowered from offender; Fight = Victim physically fought with offender; Argue = Victim verbally argued with offender or bystander; Uncoop (uncooperative) = Victim refused to comply with offender; Cnfrnt (confront) = Victim confronts or threatens offender; Scream = Victim screamed; Coop (Cooperate) = Victim cooperates with or acquiesces to offender; Talk = Victim engaged in conversation with or questioned offender; Negoff (negotiate) = Victim attempted to find a way to compromise with offender.

To determine the extent to which victim behavior reflected the IPC’s four behavioral styles, promax rotation was used to simplify the variable component loading and reduce cross-factor loading. The 18 behavioral categories were then evaluated for inclusion in the final model based on the amount of variance they explained individually and within the model. The most important indication of variable and component fit is variance accounted for (VAF) which reflects the amount of variance within the dataset explained by the number of components. Comrey’s (1973) “rule of thumb” states that a variable that accounts for 10% or less variance is considered a poor fit, 20% is fair, 30% is good, 40% is very good, and 50% is excellent. Variables with a total VAF greater than 0.25 (i.e., account for over 25% or more) are considered suitable for inclusion. Two variables, confront and hard no, were removed from the final model due to total VAF being 0.18 and 0.25 respectively. Finally, variables defend and do nothing were removed because they did not load positively onto any component. In total, 14 victim behaviors were retained in the final CATPCA model. Three, four, and five component models were evaluated using eigenvalues and Cronbach’s alpha – a measure of internal consistency. Ultimately, the model that explained the greatest level of variance and had the strongest relationship between variables was selected.

Smallest space analysis

Similar to CATPCA, SSA computes the correlations between variables to identify and represent the structure and relationship underlying a specific behavioral or social construct (Guttman & Greenbaum, 1998). However, SSA also uses correlations’ rank order to represent the association between variables as distances on a plot. This produces a visual representation of a group of variables’ correlation coefficients whereby the higher the correlations between variables, the closer they are represented on the plot. The degree of “fit” between the plot and the original correlation matrix is measured using the coefficient of alienation. A smaller coefficient indicates better fit and generally below 0.2 is considered acceptable. Given the spatial nature of SSA, the plot can be produced in one, two, or three dimensions. Increasing the number of dimensions lowers the coefficient of alienation, as it allows more “space” in which to better represent the structure. SSA provides an intuitive way to understand the relationship between victim behavior during physically assaultive crime and the IPC’s four behavioral styles.

Jaccard’s coefficient was chosen to calculate the correlations between variables (behaviors) as it measures the association between dichotomous variables while ignoring any joint non-occurrences. This is important in the context of the DoVE data as the absence of a behavior does not necessarily mean it did not occur; it may not have been discussed in the narrative.

SSA is not affected by VAF meaning all the original 18 behaviors could be included. This meant the hypotheses regarding which style confront, hard no, defend and do nothing aligned with could be tested in a way not possible with CATPCA. If the hypotheses were supported, these behaviors should be co-located on the SSA plot with the other behaviors reflective of that behavioral style. It was expected that, if results reflected the IPC, then the position of the behavioral groupings in the final SSA plot would align with the hypothesized groups and structure presented in Figure 1.

Results

CATPCA

CATPCA found the 14 included victim behaviors were best described by a four-component model which explained 44% of the total variance in the dataset. The low variance was likely the result of the low frequency with which victim behaviors were described in each case. As previously discussed, the low frequency of behaviors resulted in weak correlations across the data and is a limitation of this study. In line with the labeling of the IPC, and prior research of offender and victim behavior utilizing the IPC, the labels of dominance, submission, cooperation, and hostility were assigned to the components.

Table 4 summarizes the four-component model, including Cronbach’s α and eigenvalues, and outlines how each individual behavior loaded onto the four components within the model. Cronbach’s α is a coefficient of reliability and reflects the strength of relationship between items in a group. A Cronbach’s α of 0.7 or higher is considered acceptable and none of the components reached this threshold. This is likely the result of the poor correlations between variables caused by the lack of victim behaviors described in each case. However, the similarity between the CATPCA results and those of SSA (described below) provides additional support for these groupings.

Table 4. Summary of the four- component model of victim behavior.

	Component 1: Submission	Component 2: Dominance	Component 3: Hostility	Component 4: Cooperation
Soft no	.771
Cower	.746
Sex	.625		−.399
Uncontrollable response	.466
Uncooperative		.657	*
Talk		.652		.309*
Evade		.574
Leave		.392
Fight			.703
Argue			.628
Scream			.447
Help		*	.308
Cooperate				.727
Negotiate				.700
Cronbach’s α	0.531	0.415	0.300	0.102
Eigenvalue	1.974	1.627	1.386	1.111
% total variance	14.098	11.623	9.900	7.936

Loadings with values less than 0.3 were suppressed for ease of interpretation. Bold denotes variables that did not appear in the hypothesized component. * denotes the variable’s hypothesized component placing.

Eleven of the 14 behaviors loaded onto components as hypothesized. The model therefore reasonably aligns with the behavioral style groupings hypothesized in Table 1, with some notable exceptions. The four behaviors hypothesized to reflect submission – soft no, cower, sex and uncontrollable response – loaded onto component one indicating that these behaviors reflected a submissive behavioral style. Four behaviors loaded onto component two (dominance): uncooperative behavior, talk, evade, and leave. Evade, and leave were hypothesized to be dominant victim behaviors; however, talk and uncoop were expected to align with the co-operative and hostile behavioral styles, respectively. While talk did have a reasonable component loading (.309) on the cooperation component (component four), its loading was higher on component two.

Component three was labeled “hostility” and comprised four behaviors: argue, fight, help and scream. Three of the four were hypothesized to reflect hostility. Help was originally hypothesized to reflect victim dominance; its inclusion on component three was likely the result of its relatively strong correlation with scream (Table 3). Component four was labeled co-operation and contained only two variables: coop and negoff, which captured instances where the victim behaved cooperatively or negotiated with the offender. Both were hypothesized to align with the cooperative behavioral style.

SSA

The structure of the 18 victim behaviors across the 150 cases was best represented in a three-dimensional SSA solution, with a coefficient of alienation of 0.13, confirming a good fit between the model and the correlations. Figure 3 shows two dimensions of the 3D SSA plot (the top face), selected because they showed the four behavior groupings most clearly. The SSA plot was partitioned to form groupings of variables based on the CATPCA groupings. The behaviors omitted from the CATPCA were explored in relation to their hypothesized IPC behavioral style and their spatial placing in the plot. The results provide additional support for the CATPCA results.

Figure 3. Dimension 1 × 3 of the 3D SSA plot of victim behaviour during physically assaultive crime.

The group at the top of the plot comprised evade, talk, leave and uncooperative behavior (supporting component 2 of the CATPCA) as well as defend (hypothesized to align with a dominant behavioral style) and were labeled “dominant” victim behaviors. The group toward the center of the plot comprised soft no, cower, sex and uncontrollable behavior (supporting component one of the CATPCA) as well as hard no and do nothing, and were labeled “submissive” victim behaviors. Doing nothing was hypothesized to align with the submissive behavioral style and its co-location with the other behaviors in this group supported this hypothesis. Hard no, however, was originally hypothesized to indicate a dominant victim behavioral style due to it being more decisive in nature. Its location in the SSA did not support this hypothesis. The group to the bottom left of the plot comprised help, argue, scream and fight (supporting component 3 of the CATPCA) as well as confronting the offender and were labeled “hostile” behaviors. Confront was excluded from CATPCA analysis but hypothesized to reflect hostility. Cooperation and negotiating with the offender comprised the final group, in line with CATPCA, and reflected the IPC’s cooperative behavioral style.

The SSA results provided some indication that the relative positioning of the four behavioral styles reflected the two axes of the IPC. Dominance and submission were located opposite to each other, as were hostility and co-operation. However, the structure deviated from that expected, and represented in Figure 1, in that the two axes did not bisect. This may be a result of the low frequency of behaviors in each case. This limitation is discussed in more detail below and the categorizations, and the frequency of each behavior, is summarized in Table 5.

Table 5. Victim behavior and the corresponding IPC behavioral styles.

IPC behavioral style (n)	Dominance (169)	Submission (89)	Cooperation (38)	Hostility (82)
Behaviors (n)	Defend (41) Evade (31) Leave (49) Talk (26) Uncooperative (22)	Cower (8) Do nothing (24) Hard no (20) Sex (21) Soft no (8) Uncontrollable response (8)	Cooperate (11) Negotiate (27)	Argue (22) Confront (14) Fight (34) Help (11) Scream (11)

Discussion

This study sought to determine the extent to which victim behavior during physically assaultive crime aligned with the IPC’s behavioral styles. By doing so, it aimed to identify new ways to conceptualize victim behavior in order to create opportunities to explore how crime unfolds and contribute to the theoretical understanding of victim agency. Overall, the results showed that, despite the range and diversity of actions across the three crime types, victim behaviors largely aligned with the IPC’s dominant, submissive, hostile, and cooperative behavioral styles. Though some differences existed, the findings were largely consistent with those from previous studies that utilized the IPC to model offender/victim dynamics.

Dominant, submissive, hostile, and cooperative victim behavior

Consistent with prior research, this study found fighting and other resistive actions reflected a hostile behavioral style while negotiating and cooperating aligned with a cooperative style. With regard to submission, CATPCA and SSA found that cowering, doing nothing/not resisting, saying no, and having an uncontrollable response such as freezing, aligned with this behavioral style. This finding was also in line with previous research. For example, Woodhams et al. (2020) classified not resisting, freezing, and disassociating as submissive victim behaviors. Similarly, Woods and Porter (2008) defined victim passivity and being too scared to scream – akin to not reacting/doing nothing in the current study – as submissive victim behavior.

The most interesting difference between the current and previous studies, and between hypothesized and realized categorizations of behaviors, was found in the theme of victim dominance. Indeed, since dominance during social interactions reflects the extent to which one individual exerts control, power, or influence over another it seems counter-intuitive that victims could behave dominantly while experiencing physically assaultive crime. Nonetheless, evading and leaving both appeared in this theme as hypothesized. This was in line with prior studies such as Woodhams et al. (2020) who described the victim running and walking away as dominant behaviors. Leaving the situation or evading the offender could be interpreted as the victim exerting influence by controlling the offender’s access to them. Additionally, the variable defend appeared in the SSA dominance theme as hypothesized and incorporated elements of resisting the offender as well as protecting others. Conversely, failing to cooperate and talking with the offender, were found unexpectedly in this theme. This suggests that uncooperative behavior reflects a dominant, as opposed to the hypothesized hostile, victim behavioral style. While not typically considered socially dominant behavior, uncooperative behavior is similar to dominant victim behaviors identified by other studies. For example, Woods and Porter (2008) classified refusing to cooperate with the offender as dominant. Similarly, Woodhams et al. (2020) included giving orders to the offender and directly declining an offender’s requests in their categorization of victim dominance.

The presence of “talk” within the dominance style was particularly unexpected. Woodhams et al. (2020) conducted the only previous study to capture talk-like victim behaviors such as “justifying offender behavior” and “comments on conversation”; however, these were considered cooperative and designed to gain offender compliance. In the current study, talking comprised engaging in conversation with the offender or asking questions, such as asking the offender for reasons behind their behavior. In line with Woodhams et al. (2020), it was originally hypothesized to be cooperative because it involved engaging with, as opposed to directly controlling, the offender. Its presence in victim dominance is likely explained by its correlation to other dominant behaviors and thus, it may reflect a victim’s attempts to influence or control elements of the crime – just in a less overt manner compared with leaving or being uncooperative.

Yet talking was also significantly correlated with cooperation and the submissive victim response do nothing. It may be that talking to the offender reflects an information gathering exercise which then motivates the victim to maintain or change their behavioral style. A key assumption of the analysis in the present paper is that behavioral styles will remain relatively consistent within an incident – for example, a victim who exhibits dominant behavior will be more likely to exhibit further dominant behaviors throughout the incident. This may not be the case for victims of violent crime, however, who may change their response as they evaluate the risk and danger to themselves. While information-gathering would still likely reflect victim dominance, further research examining where in the victim-offender interaction such behavior occurs, more detail about the nature or tone of the conversation, as well as a victim’s rationale behind talking to the offender, would provide more insight into its relationship to dominance.

Additionally, these results may shed light on the more subtle ways an individual can attempt to exert power or control over another, even in the context of a violent crime. For example, these behaviors afford the victim some element of control over the situation or the offender. While, due to the IPC’s bisecting structure, these actions will likely contain aspects of hostility or cooperation, it is the element of control that distinguishes these behaviors as dominant. The current study is limited in its ability to explore the connection between these behaviors and a victim’s conscious attempts to influence the offender. However, in the context of victim behavior, these results suggest that dominance may reflect the individual’s attempts to influence or control elements of crime, just in a less overt, more subtle way than in other, non-violent, social situations. Overall, the finding that leaving, evading, talking, defending, and being uncooperative reflect “dominant” victim behaviors within the IPC highlights gaps in how we fundamentally conceptualize victim behavior during physically assaultive crimes. It also emphasizes the importance of understanding victim motivation during violence and the extent to which they are acting with agency.

Other behaviors that did not align with their hypothesized styles were calling for help – which appeared in the theme of hostility rather than dominance as hypothesized – and “hard no” which aligned with submission as opposed to dominance. Calling for help was significantly correlated with screaming, a hostile act, and may therefore represent a more reactive, unfriendly response than a deliberate attempt by victims to control the situation by seeking help. Similarly, it seems counter-intuitive that saying no, with or without force, would be a submissive act; however, this study did not capture where this response occurred within the broader offender-victim interaction. Thus, its presence here may reflect the lack of success saying no has in achieving offender submission. This is supported by the correlation between soft no and engaging in sex acts with the offender and their close positions on the SSA plot. Alternatively, differences between the hypothesized and realized styles may again indicate victims are changing their style in response to the offender or situation. While the variables included in the current study lack the detail needed to explore this possibility, it opens an interesting avenue for future research.

A strength of the current study is its use of two complementary statistical techniques to explore and confirm the thematic structure underlying the correlations between the behaviors – particularly in light of the fact that the lack of victim behaviors described in DoVE reports ultimately impacted the relationship between variables. The placement of behaviors within the four behavioral styles largely aligned between CATPCA and SSA, with the latter providing an additional test of the spatial relationship between the four themes. As noted above, however, some deviations from the hypothesized structure were found using these methods. In addition to the behaviors that differed between hypothesized and realized behavioral styles, the dominance-submission and hostility-cooperation axes seemed to be present but did not bisect. One explanation for this discrepancy may be the low frequency of victim behaviors described in the 150 cases. Only 43% of cases described four or more victim behaviors, which impacted the strength of the correlations between behaviors. It may also increase the statistical effect of idiosyncrasies in the data. This study was limited by its reliance on secondary analysis of DoVE narratives that were originally collected for the purposes of compensation claims and thus focused on the impact of the crime rather the offender-victim interaction. This limitation could be addressed in future research by interviewing victims and collecting detailed descriptions of their interactions with the offender. While the current results aligned reasonably well with prior research, stronger correlations would produce a clearer picture of how victim behaviors map to the four IPC styles.

Future research

Ultimately, this study found that the IPC can be used to model victim behavior during physically assaultive crime. Using the IPC opens new avenues for researchers to obtain a broader understanding of the dynamic, evolving nature of crime while also exploring ideas of victim agency. Specifically, removing the limitations of an active/passive binary classification allows for a more nuanced approach to understanding victim behavior during crime – both as it interacts with the offender and situation but also potentially as it shows how victims express their power and agency in shaping offender behavior. For example, like Woodhams et al.’s (2020) findings regarding multiple perpetrator rapes, victim dominance was the most common behavioral style displayed by victims during the present sample of physically assaultive crimes – accounting for 45% of victim behaviors identified. Submissive and hostile behavioral styles were the next most common – which differed from Woodhams et al.’s (2020) study where cooperative actions accounted for 23% of victim behaviors. This was almost certainly due to this study’s inclusion of a broader range of crime types but highlights how the IPC could provide a more nuanced way to explore a victim’s behavioral contribution to the incident compared with active/passive dichotomies.

The findings from this research also provide insights into victim agency and the way victims may be attempting to shape the offender and situation. As a probabilistic model of social interaction, the IPC can be used to understand how a victim’s behavior may both influence and be influenced by the behavior of the offender. Specifically, the principle of complementarity provides a framework to understand how a victim’s behavioral style – for example, submission – increases the likelihood of a specific response – for example, dominance – from the offender. Applying the IPC to explore sequences and outcomes of victim-offender interaction would provide a greater level of nuance needed to understand victim agency during crime and how it contributes to and shapes a violent incident. Researchers have used active/passive dichotomies to classify victim behavior and explore how it influences the outcome of violent crime (for example, Bachman et al., 2002; Block, 1981). While outside the scope of the current study, future research could utilize the IPC to understand the link between victim behavioral style and outcome, including whether certain behavioral styles are more likely when crimes are completed.

One particularly interesting avenue for exploration is the role of situational dynamics in shaping the nature and type of victim behavior present during physically assaultive crime. While the IPC suggests victim behavior will largely be influenced by offender behavior (and vice-versa) in an interaction, using the IPC to conceptualize victim behavior also creates opportunities to better understand the relationship between variations in victim styles – for example, the presence of cooperation as opposed to hostility – and underlying characteristics of the situation and/or crime type. While this study combined three crime types to explore a range of victim experiences, it is acknowledged that differences in physical assault, domestic violence and sexual assault may contribute to variations in victim behavior across incidents and crime types. While outside of the present study’s scope, exploring the interaction between situation, context and victim behavioral style would provide insight into victim decision making and agency during violence.

Finally, the insights and inferences the IPC provides regarding the way victims may be attempting to shape the offender and situation provides greater opportunity to understand their role in how violent crime progresses. For example, it creates opportunities for researchers to explore circumstances under which victims may engage in particular behavioral styles and determine the extent to which victims are deliberately attempting to shape the outcome of the violence as opposed to simply reacting to the offender. This does not suggest that victims should be blamed or held to account for what occurs, but it does shift theoretical focus away from the offender as the sole actor during crime to a more interactionist perspective that acknowledges a victim’s agency during crime.

Acknowledgments

The data used in this publication were made available through the Australian Institute of Criminology (AIC). The AIC does not bear any responsibility for the analysis or interpretations presented herein.

Disclosure statement

No potential conflict of interest was reported by the author(s).