Behavior Sequencing Violent Episodes in Forensic Youth Populations

Introduction

Every year, youths who have broken the law are sent to Secure Children’s Homes (SCH), akin to adult prisons. While these homes are meant as a safe place for youths to be placed while working through the criminal justice system, the individuals within these homes may have suffered a number of traumatic life events and not have developed coping skills that overcome frustration and aggression (Eltink et al., 2018; Van den Tillaart et al., 2018). A large proportion of children within SCH settings have a history of violence, with many having directly acted violently toward SCH staff members (Browne et al., 1995). As such, violent outbursts and episodes are relatively common and place a further strain on an already overworked and understaffed system. Although the United Kingdom Department of Health has provided guidance and protocols on the use of physical restraint in SCH, the guidelines remain open to interpretation in terms of how and, perhaps more importantly, when to use physical restraint (Epps et al., 1999). There is also a lack of recent research directly focusing on this topic. Given the needs for SCH managers to balance the care of the children, their duty to protect their staff under the Safety at Work Act (1974) and to not be perceived as oppressive by the children or external parties, there is an obvious requirement for further research into the process of violence in SCH (Epps, 1997). Understanding the antecedent precursors to violent outbursts is something that would be helpful to workers in the systems, especially newer workers who may not have the work experience and implicit awareness of those with longer careers. The aim of the current study, therefore, was two-fold: first, to use an applied approach to understanding interpersonal dynamics in the escalation toward violence, and second to offer a meaningful outcome that can be used as a future training and awareness raising tool to those in the Forensic Youth system. The current study was based in the United Kingdom. Youths in these homes were placed there following arrest and imminent police questioning; remand by a criminal court; a sentence of detention following conviction for an offense; or as a result of a Section 25 Accommodation Order made by a family court judge.

A large corpus of research has shown that some forms of physical violence do not spontaneously erupt, without any precursors (Felson & Massoglia, 2012; Leclerc & Wortley, 2014), while other forms may be more implicit or automatic (Keatley et al., 2017a). Models such as the aggression-frustration hypothesis (Berkowitz, 1989; Dollard et al., 2013) suggest that when needs or goals are thwarted by real or perceived blocks, then individuals become frustrated. If the frustration cannot be reduced or vented in some way, then this may lead to aggression. In men, this is more likely to lead to physically violent outbursts, as either a means of attempting to attain the goal or as a way of demonstrating the feeling of frustration (Björkqvist, 2018). Women are typically more likely to display their aggression through non-physical, indirect means (Björkqvist, 2018). Research has typically attempted to understand the relationship between frustration, aggression, and violence through self-report measures, such as the aggression questionnaire (Buss & Perry, 1992). These scales are useful in ascertaining individuals’ self-perceptions of the causes of their violence; however, these measures are almost always post-hoc and are open to obvious biases—such as individuals not having reliable or valid self-perception and demand characteristics.

More recent research into aggression has focused dispositional aggression as a predictor across an individual’s lifespan (Chester & West, 2020). Chester and West (2020) highlight that while the Buss-Perry Aggression model (1992) is widely used in the research literature and applied settings, it focuses primarily on cognitive and affective processes, while aggression typically relates to behaviors. Indeed, in an applied setting, it is the behaviors that are in many ways more important to accurately detect, as they may serve as a clearer indicator of aggressive, violent attacks. Chester and West (2020) suggest that aggression is related to antagonism, the counter-point of “Openness” in classic personality constructs (Lynam & Miller, 2019). By conceptualizing aggression as being correlated with antagonism, explanation of behavioral traits observed in aggressive outbreaks can be gained. For example, the aggressors callous, immoral, or combative approach to social interactions. Other researchers have suggested that neuroticism is more closely correlated with trait aggression (Berkowitz, 1989). Within forensic youth homes, it may be that the type of youth entering into these settings are by nature more antagonistic and prone to aggression. It is important, however, to understand how these traits map on to observable behaviors, so that staff can be better prepared to de-escalate situations.

There is a lack of research looking at behavioral markers and predictors of future violence in forensic health settings. There is an abundance, however, of research on aggression and violence, from a more cognitive and social psychological perspective. Traditionally this research has focused on self-report measures of violence and aggression (Keatley et al., 2017a). Recently, researchers have attempted to overcome the potential self-report biases of violence by using implicit measures, such as the implicit association test (Keatley et al., 2014). Keatley et al. (2014), for example, investigated the role of self-control as a means of managing violent outbursts. They used an IAT to measure implicit self-control and found that those individuals with increased self-control showed reduced violence. Clearly, cognitions are an important part of violent outbursts, and measuring the cognitive scripts or rationale for violence can help caregivers to understand why people become violence, it is less useful in observational prediction. However, while these measures provide conceptual and empirical support for models of aggression and violence, they have extremely limited use in real-world settings. It is unlikely forensic youth homes would want to test everyone with an implicit measure, and it would not be wise to do so either, given the recent criticism of such measures (see Fiedler et al., 2006, for review).

A further issue with many studies into aggression and violence is that the “violent outcome” is often construed as a single isolated behavior, such as “shouts” or “punches”. While this approach to researching violence is clearly important and provides an understanding of individuals’ psychological predispositions toward violence, the fact remains that operationalizing the outcome variable as an isolated behavior without any antecedent behaviors is overly simplistic and a large reason why statistical models of this kind are limited in their predictive validity (Epps et al., 1999). A method that maps the complex chains of behaviors that build toward a violent incident can clearly fill this gap in the literature and provide a much needed addition to our understanding of how and why violence escalates.

In relation to forensic settings, Papadopoulos et al. (2012) conducted a systematic review of 71 studies into the antecedents of violence and aggression in psychiatric in-patient settings. Staff interactions that were performed to limit in-patient freedoms were seen as a major predictor of future violent episodes. The review also showed that many staff record in-patients physical behaviors (rather than emotional states or mood). In-patient behaviors are more readily observable and easier to discriminate, so it is no surprise that they featured more in reports and research. Taking this approach further, a method that uses the observed behaviors and captures the dynamic processes underpinning antecedents of violent episodes would complement existing approaches and offer staff a new way to predict when a violent outburst might occur (Keatley, 2018).

Method

Results

Behavior Sequence Analysis

BSA was conducted to show the transitions of behaviors and events across the sample of violent episodes in a Forensic Youth Home. First, a transition frequency matrix was calculated, showing the transitions between pairs of behaviors. This table was then analyzed to show Standardized Residual scores, which could then be used to plot the state transition diagram (see Figure 1). Diagrams are created to provide a visual representation of the results. The results are lag-one BSA, which essentially measures A→B, B→C transitions; however, these can be placed into flow diagrams to show links between behaviors. The diagram is a representation of the results, with thicker lines connecting nodes indicating larger standardized residuals. The alternative is to present all results in a table; however, this would be a large table, and understanding or following the links between behaviors would be lost.

OPEN IN VIEWER

Figure 1.

State transition diagram of behavior chains leading to violent outbursts and restraints.

The diagram shows a cluster of transitions around the violent individual and staff, such that looking toward staff followed by talking to staff was observed multiple times (n = 10, SR = 6.16). However, a key transition in the chain was the individual looking toward their target and then looking away from them again (n = 7, SR = 8.12), and then looking back toward their target (n = 7, SR = 7.91). This cycle of looking at and then away from the target is important for staff to pick-up on, to raise awareness, as one of the next transitions from looking toward target was to walk toward target (n = 5, SR = 4.47), and from there, pushing target occurred more than would be expected by chance (n = 6, SR = 12.11).

Other nonverbal gestures, such as shaking head while speaking to target of aggression (n = 4, SR = 4.30) and pointing while speaking to the target of aggression (n = 5, SR = 4.81) were also observed, though not directly related to a subsequent physically aggressive act. Here again is a point in the sequence that staff could use to raise awareness of the following steps in the sequence. It should be noted, however, that behaviors such as crouching (possibly to become hidden) and breaking furniture (n = 8, SR = 23.39), to then brandish a weapon (n = 3, SR = 7.58) were not directly, clearly linked to a subsequent violent outburst. Clearly, staff should not ignore these behaviors; however, nor should staff hold-back until they see these behaviors, as the more inconspicuous behaviors are more closely connected to violence. It should also be noted that restraints typically occurred after staff were pushed past, or hit—indicating that staff were often the target of violence.

Discussion

The aim of the current study was to use Behavior Sequence Analysis to map episodes of violent outbursts in a Youth Forensic Home, in the UK. A series of 45 recordings of violent episodes resulting in the need for staff restraint of youths were coded and analyzed using the temporal method. The overarching outcome of the analysis is that these violent episodes are complex, involving numerous behaviors not directly related to the violent encounter, and therefore not a good predictor, in isolation. This is typical of research using BSA to map real-world episodes, which seldom happen along clear pathways (Clarke & Crossland, 1985; Keatley, 2018; Keatley & Clarke, 2019). While the results are complex, they do provide a more realistic account of violent episodes, and this research forms the basis of future research.

While previous research has focused on self-report (Buss & Perry, 1992) and implicit measures (Keatley et al., 2014) of violence, the current study used a behavior-focused approach to map the progression of actions and events leading-up to a violent outburst. Results indicated support for findings in Papadopoulos et al. (2012), showing that there were a number of antecedent behaviors to violence. There was commonly a staff-youth interaction before the violent outburst. Following on from this, staff could receive de-escalation training so that they are better trained to reduce the likelihood of violence ensuing. Given that staff were typically the final node before restraints being used (either having been pushed past, or hit), this indicates staff are often the target of violence. Staff may be seen as figures of authority and regarded as controlling or stopping the attainment of goals, therefore, frustrating the youths, which supports the frustration-aggression hypothesis (Berkowitz, 1989).

Importantly, the results showed that the typical chains to violent outbursts did not have any key, single, indicators that were overtly violent in nature. Most of the behaviors, if taken individually, were relatively passive (such as “looking at target” and “looking away from target”). This adds support to the theme of the current research, that it is the sequence rather than individual behaviors that is important in predicting violent outcomes. This is similar, also, to previous BSA research in violent episodes. Taylor et al. (2018), for example, showed that violence occurred after a complex interaction between multiple variables, none of which alone was violent or aggressive. These results, however, do lend support to the trait aggression approaches (Chester & West, 2020). While the focus of the current research was on behavioral analysis, given the large corpus of research focusing on personality factors related to aggression and implicit cognitive processes, it would be possible to suggest links between behaviors and cognitions. It may be the case that youths who are in these forensic centers are more likely to have individual difference factors (e.g., antagonism, neuroticism, callous disregard). These personality factors predispose the youths to perceive or interpret stimuli (e.g., other youths’ behaviors) as being threatening, disrespectful, or combative, resulting in a propensity to escalate aggression and violence. Owing to the possibility of implicit cognitions, it may be that youths’ decision to escalate a fight is largely non-conscious and a matter of trait “habit,” rather than reasoned action. Future research could investigate to what extent youths are able to verbalize or justify their decision for each behavior in the sequence. It may be that certain segments of the sequence are not available for cognitive reflection. Training methods to overcome these impulsive sub-sequences may be more effective than global-level approaches that focus on the episode as a whole.

Clearly, the current research is not without limitations. First, and most notably, the episodes in the current research all ended with staff having to intervene to restrain the youth. This was a purposefully done within the current research for practical reasons. Though a “control” condition of non-restrain episodes sounds like a good idea, it is difficult to impose. Clearly, staff need to intervene and restrain at a certain point, once the escalation reaches physical violence. It would be difficult to judge what a non-violent episode looks like—in one sense, any interaction or period of time wherein a violent event does not occur could represent a non-violent, non-restraint episode, but it would not make for a compelling or clear contrast condition. It would be important, however, to find episodes in which de-escalation occurred without outside interference (such as restraint) so that staff can better judge whether an episode is escalating or not. Comparison and contrast of escalating and de-escalating scenes, through sequence matching would show points in the sequence that may predict more clearly when an episode may turn into violent outbursts. Also, as researchers we may be interested to see the difference between violent and non-violent encounters; however, staff are typically overworked and need to prioritize time. It is beneficial that they understand and learn the maps of sequences that have previously required restraint, rather than know those that do not. Given the complexity of the current results, the “non-violent” condition would likely be extremely complex and require more time to understand and learn than most staff have. That said, future research could see to investigate non-violent incidents as a comparison and contrast condition. Perhaps operationalizing “non-violent” as any encounter that has aggressive signs or behaviors, but does not ultimately end in restraint. This might also help researchers and practitioners to understand de-escalation pathways.

A further limitation is the fact that only young men were used in the current study, meaning results cannot be generalized to young women in similar contexts. As research has previously indicated that men and women typically exhibit different forms of physical and verbal violence (Björkqvist, 2018), the results in the current study should not be generalized to young women. Therefore, further research should be conducted on young women’s Secure Chidlren’s Homes to measure and map the types of behaviors exhibited there. A final limitation was in the coding of the data, owing to the fact that only one researcher was allowed to see the source material. Ideally, multiple coders should be used and inter-rater reliability taken; however, researchers agreed on coding schemes and the coder reported no difficulties with coding material. This is typically the case for BSA research in which behaviors are quite clear to observe and code.

Future research should also include more developmental factors and individual difference predictors on violent outbursts. A growing wealth of research shows the cognitive, social, and developmental predictors of violence, future research could stratify the sample into individuals with different demographics or measurement scores. Taking a typology approach to BSA would be more akin to recent research (e.g., Keatley et al., 2019), in which different typologies of serial killers showed different sequences, or Marono et al. (2018) research into different personality clusters and deception. Therefore, measuring individual’s personality and developmental backgrounds might offer clearer sequential predictions.

Conclusion

Young people in Secure Children’s Homes are likely to engage in acts of aggression or violence. While self-report and implicit measures of aggression and violent tendencies exist, these are not clearly visible or known to staff who may have to deal with young people on a daily basis. Therefore, understanding the observable behaviors may provide a more applied approach to understanding violence. The current research provides a novel way for practitioners to understand the progression toward violence that may occur, and therefore better prepare and perhaps thwart the physically violent outburst. The findings suggest that relatively non-confrontational behaviors occur in the build-up chain toward violent outbursts that require restraining. Though there were several notable limitations in the current research, it stands as a foundational framework for future research to build upon and add toward.