Foreshadowing Targeted Violence: Assessing Leakage of Intent

Foreshadowing Targeted Violence: Assessing Leakage of Intent by Public Mass Murderers

Abstract

• The idea that identifiable behaviors presage violence is a core concept in threat assessment.
• "Leakage" is when offenders intentionally or unintentionally reveal insights into their thoughts or feelings that suggest impending targeted violence.
• This study describes leakage in a sample of 115 public mass murderers in the U.S. between 1990 and 2014.
• Leakage is disaggregated into three forms:
  • Written statements
  • Verbal statements to the public
  • Verbal statements to family/friends
• A significant predictor of leakage is the presence of a grievance against a person or entity.

1. Introduction

• Identifiable behaviors presage violence, termed 'pre-event indicators' or “warning behaviors”, “signaling the attack”, “high risk indicators” and “pre-attack signals”.
• Leakage is the communication to a third party of intent to harm another (Meloy & O'Toole, 2011; O'Toole, 2000).
• Leakage can encompass a range of factors, from objective details to subjective factors like motivation.
• O'Toole (2000) defined leakage as occurring when a prospective offender reveals clues to feelings, thoughts, fantasies, attitudes, or intentions that may signal an impending violent act.
• Clues could be subtle threats, boasts, innuendos, predictions, or ultimatums, spoken or conveyed in various forms.
• Studies have examined leakage among various populations, including mass murderers.
• Hempel, Meloy, and Richards (1999) assessed leakage in a sample of 30 adult mass murderers and divided a “threat variable” into four categories:
  • Specific threat
  • Generalized threat
  • Mixed threat
  • No threat
• They found evidence of some type of threat in 67% of their cases.
• Meloy and colleagues found that in 44% of cases the offender discussed the act of murder with at least one person prior to the event in a 2001 study of 34 adolescent mass murderers.
• In 58% of the cases where sufficient evidence was available, offenders made threatening statements, generally to third parties.
• Fein and Vossekuil (1999) examined the behaviors of 83 persons known to have attacked or approached to attack public officials and public figures in the U.S. since 1949, finding that 63% of the offenders had made an indirect, conditional, or direct threat about the target, although only 11% had made a direct threat to the target or about the target to law enforcement.
• Vossekuil et al. (2004) examined 37 school-based attacks and 41 offenders for evidence that they “signaled” the attack, finding that in 81% of cases, at least one person had information that the offender was thinking about or planning the attack, and in 59% of the cases more than one person had information about the attack.
• Drysdale, Modzeleski, and Simons (2010) studied 272 incidents of targeted violence at institutions of higher education occurring in the U.S. between 1900 and 2008 and determined that in 13% of the incidents (n = 35) the attacker made verbal and/or written threats about the target which were transmitted to the target or to a third party.
• Gill, Horgan, and Deckert (2014) examined 119 cases of those who engaged in or planned to engage in lone-actor terrorist attacks in the West, finding that in 64% of the cases the offender verbally told friends or family about the intent to engage in terrorist activity and that in 59% of cases, the offender made public statements prior to the planned attack.
• Meloy and O'Toole (2011) re-examined the literature on leakage, concluding that progress had been hindered by the lack of a standardized definition and re-oriented it over eight conceptually related warning behaviors:
  1. Pathway warning behavior
  2. Fixation warning behavior
  3. Identification warning behavior
  4. Novel aggression warning behavior
  5. Energy burst warning behavior
  6. Leakage warning behavior
  7. Directly communicated threat warning behavior
  8. Last resort warning behavior
• Leakage would not include threats made directly to a target or law enforcement; that behavior is considered “directly communicated threat warning behavior”.
• Leakage is also separate from “last resort warning behavior” which, as the name suggests, involves behaviors which likely occur close in time to the attack and suggest that the individual has made a conscious determination that there is no alternative to violence. In contrast, leakage is a more general behavior that may be intentional or unintentional and is not constrained by temporal limitations.
• Research on leakage is in its infancy, with empirical studies mainly presenting its prevalence.
• This paper provides a disaggregated view of leakage's different forms and investigates what differentiates those who leak from those who do not, testing intuitive predictors.
• The research findings are exploratory, providing operationally useful guidance on the nature of leakage in public mass murderers.

2. Method

• Data was obtained as part of a broader study into the demographics and behavioral underpinnings of public mass murderers in the U.S. (Horgan, Gill, Bouhana, Silver, & Corner, 2016).
• Data was collected from open source research on all identifiable offenders from 1990 to 2014 who met the study definition of public mass murderers.
• Public mass murder is defined as four or more homicide victims (not including the offender) killed at one location (or multiple but geographically close locations) within a short period, excluding certain types of mass murder events.
• The study excludes mass murders that involve primarily intimate partner violence (“IPV”) and occur in the home, deaths arising from gang and/or organized crime activities, and mass murders with more than one offender.
• To identify the sample, researchers examined the academic literature on mass murderers, databases created by Mother Jones, USA Today, and Mayors Against Illegal Guns, and the FBI's Supplementary Homicide Reports.
• Ultimately, 115 offenders met the specified inclusion/exclusion criteria.
• Krouse and Richardson (2015) identified 66 mass public shootings (with four or more killed) over a 15-year period (1999–2013; 4.4 events per year).
• During that same 15-year period, researchers found 88 public mass murders (1999–2013; 5.9 events per year), and over the longer 25-year period of the study, 115 public mass murders (1990–2014; 4.6 events per year) (although nine mass murderers in the sample used a weapon other than a firearm).

2.1. Codebook

• A codebook was created for data collection after an extensive review of the literature on violent crimes and high-risk behaviors.
• The codebook contains > 180 variables covering four major areas: demographic, antecedent event behavior, event specific behavior, and post-event behavior.
• Three separate coders independently coded each observation, and results were reconciled by the project manager, who considered the trustworthiness of the sources.

2.2. Analysis

• Leakage was measured using three variables:
  1. Written Statements
  2. Verbal Statements Public
  3. Verbal Statements Family/Friends
• The three types of leakage were aggregated into a “global leakage” variable.
• The global leakage variable was analyzed through a series of crosstabs along variables of theoretic interest.
• A logistic regression of global leakage and each of the three sub-types of leakage was conducted along variables found to be significant in the bivariate analysis.
• Predictor variables were separated into categories based on theoretical considerations and assumed practical relevance.
• Each variable was treated as dichotomous (“yes” or “not enough information to suggest a yes”).

3. Results

3.1. Descriptive results

• Leakage is defined as communication to a third party of an intention to harm a target.
• The aim is to identify facts that would be of use to a threat assessment professional or law enforcement officer in assessing the potential for violence.
• Leakage need not be contemporaneously recognized as a sign of impending violence to be relevant.
• Leakage of some sort was found in more than half of the sample (n = 67, 58.3%).
• Leakage was often related to the eventual victims.
• The leaked intent could be derived from the words themselves, or from ancillary factors such as the timing and/or audience.

3.1.1. Written statements

• The least common means of leakage was written communication (n = 14).
• Examples include:
  • George Hennard writing a letter describing women as “vipers”.
  • Bryan Uyesugi leaving threatening notes for co-workers.
  • Jaylen Fryberg tweeting about his anger and texting ominous notes.
  • Shayne Riggleman expressing disappointment and thoughts about revenge on Facebook.
  • Jared Loughner making nonsensical Internet posts about government conspiracies.

3.1.2. Verbal statements public

• Verbal leakage to the public was far more common (n = 35).
• Examples include:
  • Herbert Chalmers saying he was going to kill his boss.
  • Thomas McIlvane joking about a mass murder and saying the same would happen if he wasn't reinstated.
  • Nathan Dunlop making public statements about killing people and robbing a restaurant.
  • James Holmes speaking about his desire to kill “people”.
  • Floyd Zane telling a prostitute he wanted to kill the next people he saw.
  • Charles Thornton disrupting city council meetings with angry outbursts.

3.1.3. Verbal statements family/friends

• The most common form of leakage was verbal statements to family or friends (n = 36).
• Examples include:
  • Charles Brownlow announcing his plan to kill family members.
  • Herbert Chalmers telling his ex-girlfriend that he was going to kill everyone who had made him suffer.
  • Eric Houston talking with his best friend about wanting to kill people at his old high school.
  • Kip Kinkel making troubling statements about killing someone, torturing animals, and explosives.
  • John Miller complaining about a child-support order and threatening to kill himself.

3.2. Bivariate analysis

3.2.1. Age, birthplace, education, employment, military service, criminal history

• These variables were analyzed for their relationship with leakage.
• None of these variables were significant at any level.
• Table 1 shows leakage by age, education, employment, military, and criminal history.

3.2.2. Social connection and mental health

• Social isolation, living alone, mental illness, and substance abuse were analyzed.
• Neither social isolation nor living alone had a statistically significant relationship with leakage.
• Substance abuse did not have a statistically significant relationship with leakage, while mental illness was significant at the 0.1 level.
• Table 2 shows leakage by social connection and mental health.

3.2.3. Stressors and grievance

• Stressors and grievances were analyzed for their relationship with leakage.
• Grievance (p = 0.003; Phi = 0.274) and grievance person/entity (p = 0.002; Phi = 0.289) were significant.
• Table 3 shows leakage by stressors and grievance.

3.2.4. Event

• Event characteristics were analyzed.
• The only event variable with a significant relationship with leakage was arson (p = 0.031; Phi = 0.199).
• Table 4 shows leakage by event.

3.3. Logistic regression

• Logistic regressions were performed using the three variables with statistically significant relationships to leakage: grievance person/entity, mental illness, and arson.
• The only significant relationship was global leakage with grievance person/entity and arson.
• Offenders who had a grievance against a person or entity were 3.8 times as likely to engage in some form of leakage.
• The findings regarding arson should be interpreted with caution, as there were only 10 cases of arson in the sample.
• Tables 5-8 show the results of the logistic regressions.

4. Discussion

• Leakage is apparent across all targeted violence offender groups.
• Leakage was present in 58.3% of the 115 public mass murderers in the present sample.
• Several factors that intuitively might seem to be related to leakage turn out not to be significant, such as social isolation, living alone, mental illness, and demographic factors.
• A grievance against a person or entity is strongly correlated with leakage.
• Grievance category and grievance idea do not correlate with an increase in leakage, suggesting that the phenomenon is a function of concreteness.

5. Limitations

• Limitations are inherent in the open source research method.
• Researchers may have missed some events.
• Media reporting does not address all the myriad biographical details relevant to researchers.
• The amount of data available may vary in a systematic way.

6. Conclusion

• There are likely multiple motivations for leakage, ranging from seeking attention, the desire to intimidate, a need for excitement or the simple inability to contain anxiety related to the impending violent act (Meloy & O'Toole, 2011).
• Those tasked with assessing potential violent offenders should be aware that leakage in its various forms is common among public mass murderers, although not universal, and consider the presence or absence of leakage with the understanding that it is most likely to occur when the offender has a grievance and that grievance is against a specific person or entity.
• The pathway to violence model does not suggest that all people with a grievance will move toward violent action.
• Many research opportunities concerning leakage remain.
• The present sample did not contain mass murder events that were primarily IPV, gang/organized crime related, or with more than one offender, limiting the generalizability of the findings, and indicating areas for similar study.
• Also, the approach taken by this study could also be applied to other related groups of targeted violence offenders, including the groups previously assessed for the prevalence of leakage.
• Ideally, future efforts will also address the issue of false positives – namely, comparing cases of targeted violence and leakage with instances where there was leakage but no targeted violence.