RPA-Assig #2 - see attached

sagepub.com/journals-permissions DOI: 10.1177/0011128721999338 journals.sagepub.com/home/cad Article Pre-Attack Warning Behaviors of 20 Adolescent School Shooters: A Case Study Analysis Meagan N. Abel 1 , Steven Chermak 1, and Joshua D. Freilich 2 Abstract This study examines the pre-attack warning behaviors of adolescent schoo\ l shooters in the US. We conducted 20 case studies of adolescent school shooters in the United States that committed non-fatal or fatal shooting\ s on K-12 school grounds between 1999 and 2016. We investigate whether the school shooters displayed warning behaviors before the attack, who i\ n the perpetrator’s life was aware of these warning behaviors, and what\ if any actions were taken in response. Given the emergence of online forms of communication, we also investigate how adolescent school shooters may variably communicate warning behaviors in online and offline contexts.

Keywords school violence, juvenile delinquency, violence, prevention Since the 1999 targeted mass shooting at Columbine High school, many US parents and children have been gripped by fear about potential school sh\ oot- ings (Addington, 2003; Graf, 2018). School shootings are statistically\ rare 1Michigan State University, East Lansing, MI, USA2John Jay College, New York, NY, USA Corresponding Author:

Meagan N. Abel, School of Criminal Justice, Michigan State University, B\ aker Hall, Room 557, 655 Auditorium Road, East Lansing, MI 48824, USA.

Email: [email protected] 999338 CAD XX X 10.1177/0011128721999338Crime & DelinquencyAbel et al.

research-article 2021 Abel et al. 787 events—The American School Shooting Study (TASSS) finds than an average of 24 shootings resulting in at least one injury occur on K-12 school gr\ ounds each year in the United States (Freilich et al., 2020). Despite their relative infrequency, acts of school gun violence receive disproportionate media atten- tion compared to other violent crimes (Maguire et al, 2002). Thus, many Americans view school shootings as a significant social problem (Elsass et al., 2014; Schildkraut et al., 2015). Given this level of public concern, both aca- demics and practitioners have conducted research to identify adolescents\ plan- ning acts of violence before they commit their attack. One key finding from this research is that the vast majority of individu\ als planning to commit a school shooting broadcast their intentions to other\ s (Alathari et al., 2019; Pollack et al., 2008; Vossekuil et al., 2002, 2004).

Researchers have begun identifying the type of behaviors that adolescent\ s planning acts of violence may exhibit, when these warning signs occur, and the responses undertaken (or not) by the recipients of this informatio\ n (Alathari et al., 2019; Meloy et al., 2001; Pollack et al., 2008). However, it remains unclear whether these warning behaviors and the responses to the\ m vary by the type of attack, offender characteristics, or bystander attributes. This study seeks to fill this research gap by examining 20 case studies \ of adolescents in the United States that committed a shooting on school grounds that injured at least one person between 1999 and 2016. We explore if the shoot- ers displayed warning behaviors prior to the violent offense and what, if any, actions were taken in response to those warnings. We identify how different case circumstances may correlate with variations in the manifestation of part\ icular warning behaviors, including timing, modality (online vs. offline warnings), and bystander response. We conclude with a discussion about the policy implica- tions of our findings and provide suggestions for additional research.

Literature Review Defining “School Shootings” While the public, influenced by media reports, often associates school s\ hootings with incidents where an adolescent randomly kills many of their classmat\ es, this is only one type of school shooting. Unfortunately, there is no universally accepted definition of school violence. Currently, researchers and practitioners employ a variety of definitions. The broadest definitions count brandishing a gun or a single injury by a firearm as a case of school violence (Fridel, 2019; Kalesan et al., 2017). Stricter operationalizations require a particular number of\ fatalities for an event to constitute a school shooting (Anderson et al., 2001; Gerard et al., 2016; Holland, 2019). Given these varying definitions, it is 788 Crime & Delinquency 68(5) difficult to compare findings across these studies. In this study, we select the case studies from an open-source database that includes all publicly known shootings that resulted in at least one injury that occurred on K-12 sch\ ool grounds in the US between January 1, 1990 and December 31, 2016. A positive development for researchers is the creation of a number of typologies or classification systems for acts of school violence. These typolo- gies recognize that not all school shootings share the same motivations \ or outcomes. Following an analysis of 51 completed and 51 averted incidents\ of school violence, Langman and Straub (2019) identified three broad type\ s of school shooting incidents: (1) large scale, random attacks planned in advance, (2) small scale attacks against specific people whom the attacker holds a grievance, and (3) unplanned attacks in which the attacker brought a g\ un to school without the initial intention of using it. The first category, large ran- dom attacks, can be further conceptually divided into mass shootings—\ those in which multiple (typically four or more) individuals are killed (Du\ we, 2020)—and rampage shootings—incidents in which a current or forme\ r stu- dent targets their victims at random or for symbolic reasons in a public area of the school (Newman et al., 2004; Rocque, 2012). Mass-rampage attacks are less common than targeted or unplanned attacks, and recent research sug- gests that mass-rampage offenders are significantly different from other per - petrators of school violence (Baird et al., 2017; Langman, 2009).

Threat Assessment and Warning Behaviors In response to the 1999 Columbine shooting, schools adopted an array of violence prevention initiatives including no-tolerance discipline polici\ es, increased student surveillance, and the use of shooter “profiles” \ based on descriptive, demographic, or psychological characteristics (Borum et al., 2010). Concurrently, the Department of Education and the US Secret Service engaged in a research program (subsequently named the Safe School Initiative) to investigate the planning and pre-attack behaviors of school shooters (Vossekuil et al., 2004). This research group sought to identify infor - mation that is “knowable” before an act of targeted school violence, defined as an act of violence where a known or knowable attacker selects the tar\ get prior to the violent attack (Fein et al., 1995). Significantly, this study and subsequent investigations found that there is no consistent “profile”\ of ado- lescents who engages in targeted violence. Student-offenders range in age, socioeconomic status, academic achievement, and psychological and social\ backgrounds, etc. However, most attackers do share their ideas or plans with others or engage in a behavior “that caused others concerns or indica\ ted a need for help” (Vosskuil et al., 2004). Abel et al. 789 Given these findings, the Safe School Initiative recommended that schools use threat assessment models as their primary method to prevent future acts of targeted violence. Threat assessments identify concerning individuals, assess their risk for engaging in future violence, and then take steps to manag\ e that risk (Alathari et al., 2019). A critical component of this model is for bystanders to report to authorities concerning behaviors they witness. This reporting ensures that threat-assessment teams become aware of the potential threa\ t and consider it as an element in the risk-investigation process (Borum et al., 1999). As the Secret Service and Department of Education were developing the Safe Schools Initiative, a research team at the FBI also called for the imple- mentation of a threat assessment model as the primary method of preventi\ ng school shootings (O’Toole, 2000). In this report, O’Toole coined the term leakage: “when a student intentionally or unintentionally reveals clues to f\ eel- ings, thoughts, fantasies, attitudes, or intentions that may signal an i\ mpending violent act” (O’Toole, 2000). In the decade following, scholars recognized that this definition was broad and overlapped with other behaviors that \ may forecast an individual’s intent to engage in violence. Thus, Meloy and O’Toole (2011) revised leakage to be one of eight warning behaviors, defined as “any behavior that precedes an act of targeted violence, is related to it, and may, in certain cases, predict it.” These warning behaviors include:

1. Pathway warning behavior—research, planning, preparation, or implementation of an attack 2. Fixation—an increasingly pathological preoccupation with a person or cause 3. Identification—a psychological desire to be a “psuedocommando,”\ have a “warrior mentality,” closely associate with weapons or other military or law enforcement paraphernalia 4. Novel aggression—acts of violence unrelated to attack behavior that are committed for the first time.

5. Energy burst—an increase in the frequency or variety of any noted activities related to the target 6. Leakage—the communication to a third party of an intent to do harm to a target 7. Directly communicated threat—the communication of a direct threat to the target or law enforcement beforehand 8. Last resort—increasing desperation or distress through declaration in\ word or deed A growing body of literature on adolescent school shooters finds that mos\ t display warning behaviors prior to committing the attack (O’Toole, 2000; 790 Crime & Delinquency 68(5) Vossekuil et al., 2002). O’Toole’s (2000) original study of 18 shooters found that all engaged in some degree of leakage under her broad definition. The US Secret Service’s study of 37 school shootings found that in 81% of cases, “at least one person had information that the attacker was thinking a\ bout or planning the school attack” (Vossekuil et al., 2004). The National Threat Assessment Center’s subsequent study on targeted school violence found that all 41 adolescent offenders identified from January 2008 to December 2017 exhibited “concerning behaviors” (Alathari et al., 2019). These behaviors included planning acts, direct threats, leakage, preparation of final co\ mmuni- cations, and interest in weapons or violence, among others. Significantl\ y, the authors found that while a majority of the offenders exhibited a “constellation of lower-level behaviors,” displaying two or more warning signs, peer bystanders did not alert adults prior to the incident (Alathari et al., 2019). Only one known study has examined leakage from the perspective of the person who received the information. Pollack et al. (2008) interviewed 198 bystanders who received leaked intent from a school shooter. This study begins to shed light on why bystanders often fail to report concerning b\ ehav- iors to the authorities prior to school shooting events. Of the 198 inte\ rviewed individuals, 34% were friends with the attacker, and 29% were acquaintances (Pollack et al., 2008). A majority (59%) learned of the shooter’s intentions shortly before the attack, only days or weeks in advance, giving them a \ short period of time to consider their response (Pollack et al., 2008). Significantly bystanders reported that school climate, belief in the attacker’s threat, and judgment of the likelihood and immediacy of the attack all impacted thei\ r decision to come forward (Pollack et al., 2008). It is clear from this growing body of literature that adolescent school \ shoot- ers exhibit a variety of warning behaviors prior to an act of school vio\ lence, and that there are significant barriers keeping bystanders from reporting those behaviors. However, this research does not disaggregate shootings that only involved injuries versus fatal events. It is thus unclear if offenders in each cir - cumstance differ in warning behaviors. Given that the majority of this research took place in the early 2000s, it also remains unclear if shifts in adol\ escent use of technology—cell phones and the internet—may have changed the wa\ ys in which offenders exhibit warning behaviors and strategies for detection.

School Shootings, Social Control, and Prior Warning Behaviors This study investigates whether Sampson and Laub’s life course approach could account for variation of warning behaviors across different kinds of school shooters. While the Safe Schools Initiative and other previously discussed studies provide descriptive analyzes of school perpetrators’\ Abel et al. 791 backgrounds, it remains unclear if these circumstances manifest in parti\ cular warning behaviors. In looking at offender’s histories, we rely on Sampson and Laub’s (1993) developmental/life course social control perspective.

This perspective includes a series of constructs that capture developmen\ tal patterns over the course of individuals’ lives and assess the impact of precur - sor, enduring, and contemporaneous variables. Importantly, the influence of larger social events can indirectly impact individual behavior via their in\ flu- ence on significant others. Further, the life-course perspective’s emphasis on human agency recognizes that individuals are not passive, but instead ma\ ke deliberate decisions within the structure around them. The life-course per - spective provides the basis for a more nuanced investigation of the path\ ways into, maintenance, and exit from behavioral patterns. Life-course models include two key constructs, trajectories and transi- tions. Trajectories are long-term patterns of behavior, while transitions are particular life events that occur over shorter periods of time (Elder, 1995; Sampson & Laub, 1993). It is possible for certain transitions to become\ turn- ing points in the life course and to modify or reorient behavioral traje\ ctories.

Many empirical studies on juvenile delinquency and crime investigate whether key life events (e.g., marriage, employment, military service)\ sig- nificantly impact future involvement in crime and most find that these t\ urn- ing points play an important role. While many studies look at positive turning points that promote desistan\ ce from crime, turning points can also be negative in nature and foster con\ tinued or accelerated involvement in deviance (Bersani & Chapple, 2007; Chung et al., 2002; Fergusson et al., 2000; Wiesner & Capaldi, 2003). Sampson and Laub (1993) and Laub and Sampson (2001) explain that if certain turn\ ing points occur, individuals may become more connected to society and its insti- tutions, have higher levels of social control, and are thus less likely \ to be involved in violence (low-risk). On the other hand, socially isolated youths with lower levels of social \ control will do poorly in school and are more likely to commit deviance.\ If such youths do not experience positive turning points, or continue to ex\ peri- ence negative turning points, they will have more difficulty forming bonds/ attachment with their parents, classmates and teachers, and will become \ less involved in collective activities; that is, have more time on their hand\ s. These students will also have lower attachment/commitment/involvement- social \ control- than others, making them at higher-risk for engaging in violence. Given the differences in social control, we expect that high- and low-risk school shooting offenders will exhibit different types of warning behaviors prior to an act of school violence. More significantly, given the peer-groups these individuals are differentially surrounded with, we expect that there will be a variation in bystander response between these two groups. 792 Crime & Delinquency 68(5) Research Questions The literature on targeted school violence has demonstrated that most perpe- trators exhibit knowable warning behaviors prior to a shooting event.

However, researchers have yet to explore how differences in attack type and offender background may result in variation in the type of warning behaviors demonstrated, the timing of those behaviors relative to the attack, and who “receives” those warning behaviors. The current study seeks to fill these gaps and will answer three research questions:

1. Do fatal school shooting offenders differ in type, timing, or recipient of warning behaviors than non-fatal school shooters?

2. Do school shooting offenders who are low in self-control (high risk) differ from offenders with high self-control in type, timing, or recipi- ent of warning behaviors?

We use a case-study analysis to examine the pre-attack behaviors of 20 adolescent school shooters to answer these questions. Utilizing open-sou\ rce data, we examined each adolescent’s history and applied a life-course per - spective. Each case was subsequently coded for the type of shooting (fa\ tal/ nonfatal), whether the offender had significant social deficits and antisocial behaviors, and evidence of any of Meloy and O’Toole’s (2011) eight catego- ries of warning behaviors.

Data and Method Sampling We used a sample of 20 adolescent (10 fatal, 10 non-fatal) school shoot\ ings from The American School Shooting Study (TASSS) (Freilich et al., 2020).

This newly created open-source database catalogs all incidents in the Un\ ited States in which a gun is fired on a K-12 campus and at least one person \ is injured from January 1, 1990 to December 31, 2016. The population of school- shooting events was identified by collating existing lists of school sho\ oting events (academic listings, official reports, the Gun Violence Archive, etc.), con- ducting targeted keyword searches on Google and LexisNexis, and tracking shootings mentioned in the news articles collected. For each identified shooting, a targeted search was conducted to collect all known information about the incident. Searchers utilized a standardized \ open-source search protocol that included over 60 web-engines, news archives, government websites and databases, and criminal justice source\ s to identify material about the event, perpetrator(s), victim(s), and sc\ hool. A Abel et al. 793 minimum of two individuals searched each incident, with the average case\ producing 60+ documents. TASSS identified 652 school shootings that resulted in at least one injur\ y in the US between 1990 and 2016. Almost 75% of the shootings in TASSS were voluntary/intentional acts (n = 473) and the rest were suicide or attempted suicide only shootings, or accidental shootings. We went through six steps to select our 20 cases. First, we queried TASSS to extract all intentional shootings (n = 473) and exclude all suicide only and accidental shootings. Second, we extracted all intentional shootings com\ mit- ted by known juveniles (n = 252) and excluded adult and unknown offender intentional shootings. These 252 intentional juvenile shootings encompassed 135 non-fatal and 117 fatal shootings. Third, since we were interested in pos- sible online warning behaviors we only included intentional juvenile sho\ ot- ings that occurred between 1999 and 2016 (n = 153), when we assume the Internet was more common. Fourth, we constructed a detailed template for\ constructing the case studies because we wanted to understand the life c\ ourse of the offender and the nature of the school environment. The template attempts to capture development social control and situational crime pre\ ven- tion indicators/issues. There was considerable variation in the number and type of documents available about each incident. We pretested the template and recognized that it was impossible to write a complete case study whe\ n the number of documents was small so we eliminated all incidents with under \ 30 documents (about 30% of the sample). Fifth, for the remaining cases, w\ e divided them by fatal versus non-fatal in a random order. Finally, we selected 10 cases from each category. Thus, our case studies focus on cases that received considerably more pu\ b- licity providing good information about the issues of concern in this pa\ per. We reviewed the incidents that were less covered and had fewer than 30 docu\ - ments and found they were mostly cases where victims received minor inju\ - ries, the shooting was gang-related, or occurred in a high-crime area. I\ n other words, our case studies under-represented these types of shootings. Media outlets provided the primary source of information about all 20 incidents. In several cases, original materials such as subject’s journal entries or websites had also been archived. The 20 cases under investigation in this paper generated an average of 219 total pages of open-source material. The total source material used to compose each case study is presented in Table 1.

Case Study Template Each case study was written according to a uniform template (see Supplemental Appendix). The case study template, a revised version of a 794 Crime & Delinquency 68(5) PIRUS template for a school-shooting context, is based on Sampson and Laub’s (1993) Theory of Informal Social Control. Given this theory, the first section of the case study calls for a life-course narrative detailing the social- structural controls in a child’s life and the presence (or lack thereof) of any antisocial behaviors leading up to the shooting. This narrative largely con- sists of summaries of the perpetrator’s early childhood, experiences in school, friends, family, and so on. The second section asks more detailed questions related to social controls and antisocial behaviors including emotional \ attach- ments to individual family members or teachers, turning points, and asso\ cia- tions with delinquent peers.

Characteristics of the Sample The majority of the offenders are male (17 out of 20) and the median age at the date of the shooting was 15 (mean = 15.4, SD = 1.5). The mean ages were Table 1. Source Material for Case Studies.

Identifying number PagesWords Court/ government News articles Scholarly/ databases Primary sources 1 12250,362 7722 1 2 3417,331 0371 0 3 263120,432 5154 30 4 16874,963 0114 10 5 5826,843 0670 0 6 8639472 4541 3 7 449193,970 3225 49 8 19283,463 5146 10 9 22790,724 1171 20 10 316141,058 2209 21 11 5727,248 0451 0 12 437233,171 4289 40 13 12856,727 5109 00 14 11255,726 2101 40 15 370177,886 18354 00 16 303145,373 11240 01 17 316143,869 6280 11 18 10648,930 4951 0 19 423179,526 10243 10 20 22393,139 2170 10 Mean 219.5100,010.7 4.5158.8 1.50.8 Standard deviation 134.263,296.6 4.591.0 1.32.1 Abel et al. 795 similar for fatal and nonfatal offenders (15.3/1.32; 15.7/1.93) and high and low risk offenders (14.86/1.78; 15.69/1.32). Thirteen offenders have significant deficits in social control and antisocial behavior (high-risk) meaning that they experienced family turmoil incl\ uding parental separation, abuse in the home, had few friends, and were bullie\ d by peers, among other factors. The remaining seven offenders experienced mini- mal hardship in comparison (low-risk). The first group of offenders were largely known as troublemakers or delinquent students to school officials, while the second were well-liked by peers and perceived as average if no\ t good students. Sixteen of the 20 school shooters engaged in warning behav- iors prior to their shooting events.

Findings We present the findings in several sections. First, we present characteri\ stics of the high risk shooters, the type of incident they were involved in (\ fatal/ non-fatal), and whether they warned others about their behavior. Second, we discuss these same results for low-risk offenders. Third, we focus on charac- teristics of the warning behaviors and the nature of the leakage that oc\ curred.

Fourth, we discussed the modality of their warning behavior, focusing on whether it was on-line, off-line, or both. Finally, we discuss the timing, response, and outcome of these warning behaviors.

High-Risk The first group of offenders—who were largely known to school officials before the shooting events—experienced a high number of negative life\ experi- ences and resultant antisocial behaviors, as seen in Table 2. These offenders were frequently raised by either a single-parent or extended family. They were often physically or sexually abused as children and had few ties to thei\ r peer groups—described as “loners” or targets of bullying. It is thus clear that this group of adolescents did not form the strong ties to friends or family that Sampson and Laub suggest are critical to developing pro-social behavior. Open source material provides less detail about these adolescent’s emotional ties to school or individual teachers. While information is available regarding educa- tional performance, which varies among these offenders from high to very low, little additional information about their school experiences is availabl\ e. Given these social control deficits, it is unsurprising that the individ\ uals in this sample also exhibit a large degree of antisocial behaviors. Seven high- risk adolescents were diagnosed with and received some degree of treatme\ nt for mental illness. In many cases, this treatment was deficient, with mo\ st adolescents going off their medications or ceasing counseling. Five were 796 Table 2. Elements of Thirteen High-Risk Shooters. Identifying number Fatal/NonVictims Social control deficits Antisocial behaviorWarning behavior 1 Fatal1 killed 1 injured Parents separated Gang membership Violence Criminal justice system involvementY 3 Fatal2 killed 13 injured Parents separated Sexual abuse Recent move Bullied Delinquent peers Drug/alcohol use Mental illness Y 4 Non-fatal5 injuredParents separated Physical abuse (father) Few friends Anger management issues Violence leading to probation Mental illnessY 5 Non-fatal2 injuredMother died Housing instability Mental illness Criminal justice system involvement Gang involvementN 6 Non-fatal2 injuredParents separated Physical abuse (father) Behavioral/mental health issues Y 7 Fatal7 killed 5 injured Parents separated Father killed by police Bullied Few friends Mental illness Right-wing affiliation Y 8 Fatal1 killed 2 injured Parents separated Parental substance abuse Housing instability Behavioral/mental health issues Drug use ViolenceN (continued) 797 Identifying numberFatal/NonVictims Social control deficits Antisocial behaviorWarning behavior 10 Fatal1 killedParents separated Physical abuse (father) Sexual abuse (stepbrother) Few friends Behavioral issues Y 12 Fatal1 killedParents separated Parental substance abuse Physical abuse (father) Sexual abuse (cousin) Right-wing affiliation Y 13 Fatal1 killedParents separated Parental mental illness (mother) Father imprisoned Sexual abuse (family friend) Behavioral issues Mental illness Y 15 Fatal3 killed 3 injured Parental substance abuse Parental intimate partner violence Sibling substance abuse Mental illness Violence Criminal justice system involvementN 16 Non-fatal1 injuredParents separated Parental intimate partner violence Parental criminal history (father and stepfather) Bullied Behavioral issues Mental illness Y 17 Non-fatal1 injuredParents separated Bullied Few friends Right-wing affiliation Y Table 2. (continued) 798 Crime & Delinquency 68(5) noted by the school or caregiver as having behavioral issues—acting out in class or at home, emotional disturbances, or other disruptive behavior.

Unexpectedly, three of the 20 exhibited some affiliation with or acceptance of right-wing/white supremacist ideologies.

Low Risk In contrast, the low-risk group experienced relatively few deficits in s\ ocial control and few antisocial behaviors as demonstrated in Table 3. For example, compare Offender 17 from the high-risk group, and Offender 2 from the low- risk group. Offender 17 was raised by a single mother, had few friends, and was significantly bullied at school. In one traumatic incident, a group of stu- dents pinned Offender 17 to the ground and a male peer placed his genitals on offender 17’s face, “tea-bagging” him (Harvey, 2014). Offender 17 was fasci- nated by Hitler and sent texts to an unknown recipient expressing far-right sentiment. He spent most of his free time after school on the internet. In con- trast, Offender 2 was raised in a two-parent household, engaged in extracur - ricular activities, and made friends with other students who were “so\ cial outcasts.” She was also bullied, and was reportedly taking antidepres\ sants as a result, but did not experience the severity or number of adverse exper\ iences that Offender 17 did. Low-risk offenders did experience adversity, but the number of negative life events was significantly fewer than those experi\ enced by the high-risk group.

Warning Behaviors Only four school shooters did not engage in warning behaviors prior to t\ heir shooting events. Two of them—offenders 2 and 8—appear to have brought a Table 3. Elements of Seven Low-Risk Adolescent Shooters.

Identifying number Fatal/Non Victims Social control deficits Antisocial behavior Warning Behavior 2 Non-fatal 1 injuredBullied Mental illness N 9 Non-fatal 1 injuredParents separated Y Bullied 11 Non-fatal 10 injuredParents separated Mental illness Y 14 Fatal 1 KilledParents separated Minor delinquency Y 1 injured Recent move 18 Non-fatal 3 injuredBullied Y 19 Fatal 4 KilledParental infidelity Y 1 injured 20 Non-fatal 4 injuredParents separated Substance abuse Y Abel et al. 799 firearm to the school without the intention of engaging in a shooting th\ at day, making warning behaviors, by definition, impossible. 1 While the high-risk group of offenders was two times as lethal as the low-risk group (high risk mean = 1.31, SD = 1.93; low risk mean = 0.71, SD = 1.49), they did not appear to be any more likely to exhibit warning signs than their low-risk count\ er parts. Similar to Silver, Hogan, et al.’s (2018) findings, there does not appear to be a “profile” of a shooter who engages in warning behaviors. H\ owever, the modality of these warnings (online vs. offline), the type of warning behavior, and the time between the warning and the offense vary consider - ably among offenders. Modality Nine of the individuals in this sample engaged in warning behaviors excl\ u- sively off-line. Six of these nine were high risk and six were involved in non- fatal incidents. These offenders all expressed warning behaviors, largely leakage or direct threats, to peers either during school or during the a\ fter- school period when they were unsupervised by adults. Four offenders dem- onstrated warning behaviors in both online and off-line spaces—three were high risk offenders and two were involved in fatal incidents. Three offenders exhibited their warning behaviors exclusively online or through technolo\ gi- cally-mediated communication. All these offenders were involved in a fatal incident and two were low risk offenders (Table 4). Offenders who displayed warning behaviors online or through technologi- cally mediated communication largely did so through peer-to-peer platforms.

Five of the individuals who warned online/through technology did so usin\ g text, IM, or emailing a friend. They also frequently posted a message the day of the shooting on social media—Instagram, Facebook, or Twitter. Only two offenders—#7 and #17—were known to have exhibited warning behaviors \ in other online communities.

Type of Warning Behavior Thirteen shooters in this sample engaged in some form of leakage prior t\ o their attack. Seven of these 13 were involved in a fatal incident and the major - ity (N = 9) were high risk offenders. Offenders with a peer network appeared to leak largely to their friends, with these statements often being highly explicit regarding the shooter’s intentions. For example, offender 11 showed the bullets he brought to school to his friends and told them he planned\ to shoot up the school (Register-Guard staff, 2007). Offender 14 posted a sui- cide note on Facebook and texted a close friend just before engaging in 800 Crime & Delinquency 68(5) Table 4. Modality of Warning Behaviors. Warning behavior HR/LRF/NFOnline OfflineBoth 1 YHRF x 2 NLRNF 3 YHRF x 4 YHRNF x 5 NHRNF 6 YHRNF x 7 YHRF x 8 NHRF 9 YLRNF x 10 YHRF x 11 YLRNF x 12 YHRF x 13 YHRFx 14 YLRFx 15 NHRF 16 YHRNF x 17 YHRNF x 18 YLRNF x 19 YLRFx 20 YLRNF x Total offenders 16 394 violence (Schukar, 2011). Offender 3 told a friend at a skate park 1 day before the shooting, “tomorrow I’m going to bring a bunch of guns and I’m going to shoot a bunch of people. . . and you’re going to watch” (Langman, 2015; Table 5). Three offenders—#3, #6, and #17—told friends to either not come to school or to leave the school when their attack was imminent. Offender 6, who attempted to commit a rampage shooting at his high school, texted th\ ree of his friends 15 minutes before the attack, writing, “IM IN SCHOOL W/ GUN. GET OUT” (Lyons, 2004). Offender 17 told several students the day prior to his attack not to come to school the next day because he had “bad feelings” about it (Kotowski, 2014). Individuals who did not appear to have a peer network leaked to their potential victims. For example, offender 10 said during one class, “I’m going to laugh when everyone in this school gets hurt,” and told his church\ youth Abel et al. 801 group that their Principal wouldn’t live through homecoming (Held, 2007).

Offender 4 put his hands in the shape of a gun, aimed them at two girls in his classroom and told them, “I wish I could do Columbine all over again”\ (AP staff, 2001). Other offenders posted threatening messages on social media.

Offender 19 posted a string of messages on Twitter prior to the shooting, including, “ Your [sic] not gonna like what happens next‼” and “Your gonna piss me off. . . And then some shits gonna go down and I don’t think you’ll like it. . . .” Offender 16 wrote on Facebook the morning he took a gun to school, “First day of school, last day of my life.” Fewer offenders engaged in warning behaviors beyond leakage.

Offender 18 demonstrated a Pathway warning behavior—he planned his attack and documented these plans in a journal (Albuquerque Journal staff, 2014). Offender 11 displayed Identification behavior, fixating on violence and weaponry a year prior to his attack and demonstrating a strong interest in the 1999 Columbine school shooting (Stine, 2007).

Offender 4 also exhibited signs of Identification behavior: according to a\ neighbor he “did love his guns” and had a strong desire to enlist in the military (Texiera et al., 2001). Allegedly, it was his rejection from the Table 5. Warning Behaviors. RiskFatal/NF Pathway FixationIdentification AggressionEnergy burst Leakage Direct threat Last resort 1 HR F x 2 LR NF 3 HR F x 4 HR NF xx 5 HR NF 6 HR NF x 7 HR F x x x xx 8 HR F 9 LR NF x 10 HR F x 11 LR NF xx 12 HR F x 13 HR F x 14 LR F x 15 HR F 16 HR NF xx 17 HR NF x x x 18 LR NF x x 19 LR F xx 20 LR NF x 31 5 0013 23 802 Crime & Delinquency 68(5) Navy that sparked his engagement in violence. Offenders 1 and 9 directly threatened their future victims. Two offenders from this sample, however, stand out for demonstrating a number of warning behaviors and engaging in these behaviors in an online space. The first exceptional case in this sample is Offender 17, a targeted, high- risk offender who in addition to the already-mentioned leakage engaged in Pathway and Identification behaviors. Offender 17 came to the attention of school authorities a year prior to his shooting when a peer reported tha\ t he had a hit list of people who had picked on him—an example of a Pathwa\ y behavior. He underwent a threat assessment and was determined to not be a risk to the community. However, school officials were unaware of this stu- dent’s anonymous activity online. He was active in furry 2 and vore 3 commu- nities online. One video that offender 17 allegedly watched depicted an animated pony being restrained, tortured, and baked into a cupcake (Kot\ owski, 2014). While participation in these online communities do not fit in neatly with one of the warning behavior categories, they do represent extreme f\ orms of sexual deviancy and thus may be conceptualized as a form of anti-social behavior under Sampson and Laub’s theory. Offender 17 displays Identification warning behaviors—a “warrior men\ tal- ity” and affiliation for the military and military paraphernalia—in his online postings. Six months prior to the shooting, he posted a short story titl\ ed “Physopath” (sic) on a fanfiction website. The story describes a bullied high school student who hunts down his tormentors as an adult and kills them \ in ways that mimic their harassment (Kotowski, 2014). The contents of this short story are incredibly graphic. In one scene, the main character kills his\ bully’s family, drains their blood, and then drowns the bully in it. The protagonist bears a resemblance to the offender: like him, the character has one older and one younger brother; number 17 wanted to join the Army and his main char - acter does so as an adult; and the character “blacks out” when he \ commits his attacks, which the offender claims he did after shooting at the school. Offender 17 appears to have a desire to be some sort of military vigilante\ , getting the justice he perceives he cannot through school. His fictional\ alter- ego (assuming this story is a case of fantasied self-representation) i\ s stealthy and skilled with weapons. Perhaps most importantly, he is praised for what he does, rather than reviled. At the end of the story, when the protagonist throws a knife into “the base of the skull and top of the neck” of an ind\ ividual, a crowd bursts into cheers. Offender 7, another high-risk offender who engaged in a mass-rampage shooting, demonstrated the greatest number of warning behaviors both onl\ ine and offline: Pathway, Fixation, Identification, Last Resort, and leakage.

Offender 7 began planning his attack as long as 2 years prior (Pathway),\ Abel et al. 803 drawing up maps of the school, researching bomb-making online and dis- cussing the plot with friends. Two years prior to the shooting, peers reported that this offender appeared to have a fascination with violence and death.

Offender 7 showed a classmate a comic book he’d drawn with images of people shooting each other, another classmate described Offender 7 drawing multiple pictures of skeletons and skulls, others reported that he was f\ asci- nated with school shooters (Roberts, 2005). Similar to Offender 17, Offender 7 posted writing online that demon- strated identification with a military/warrior mentality. Offender 7’s writing indicates an interest in graphic violence and glorification of the anti-\ hero. In addition, there is an attention to detail that suggests an affinity for weaponry, including firearms. Offender 7 makes a point of detailing the weapons used in the story—a Colt M1911 .45, an M4, a Beretta handgun—as closely as he details the gore. Offender 7 demonstrated Fixation behavior in his participation in forums on Nazi.org, beginning a year prior to the shooting. As a member of a Native American tribe, this engagement appeared to be driven by an anti-Semitic\ and anti-Black belief system, rather than White Supremacism. He explained in writing: “Ever since the Jewish post-war propaganda has been taught in our school systems (on reservations), a lot of us have been brainwashed in\ to think- ing purity is wrong. . .I can’t help but notice how many pure-blooded Natives there are left since.” Number 7’s posts indicated an increasing fixation with ideas of Native purity and disgust with inter-ethnic procreation. Six months prior to the shooting, he wrote, “Where I live less than 1% of all th\ e people on the Reservation can speak their own language, and among the youth wantin\ g to be black has run ramped (sic)” (Verdantam, 2005). Surprisingly, despite this racial fixation, Offender 7 did not appear to advocate for violence on these Nazi or White Supremacist forums. Offender 7, like the majority of other offenders who exhibited warning behaviors, also engaged in leakage. He reportedly told a female friend o\ ver IM that he was planning to attack the school on April 20th, the anniversary of Hitler’s death and the 1999 Columbine School Shooting (Kaufman, 2005).

Offender 7 posted two Flash videos he’d created to Newgrounds.com, the first titled “Target Practice” and the second titled “Clown.” Target Practice shows a masked man with an automatic-style weapon shoot four people, blow up a police car, and then kill himself. Clown depicts a clown strangle and eat a teenager. As with the zombie story, both animated videos demon- strate a fascination with graphic violence. The mass-shooting video acts as a particularly clear instance of leakage indicating a comfort with, if not\ an affinity for, mass violence. 804 Crime & Delinquency 68(5) Offender 7 made additional references to mass shootings on his social media pages. On his MSN profile, he listed his favorite thing as, “times when maddened psycho paths (sic) briefly open the gates of hell, and let chaos flood through” and his hobbies as, “planning, waiting, and hating.” Another website utilized an image from the movie Elephant, a fictional movie based on the 1999 Columbine High School shooting, as the profile photo (Marqu\ ez Estrada & Nixon, 2005). This behavior appears to contain both Identification and leakage warning behaviors—associating with other mass shooters an\ d demonstrating an intent to do harm himself. Offender 7 appeared to exhibit offline Identification with the Columbine Shooting perpetrators as well, forming a small group of students called the “Elephant people” or \ “Elephant group” who would regularly watch Elephant (Ragsdale, 2006). Finally, Offender 7 evidenced Last Resort warning behaviors in the 2 months prior to the shooting. His posts online became far more emotional, indicating strong suicidal thoughts. One example from social media reads\ , “I’m starting to regret sticking around, I should’ve taken the \ razor blade express last time around. . . . Well, whatever, man. Maybe they’ve got another shuttle comin’ around soon?” (Kaufman, 2005) Where the other adolescents reviewed in this study only displayed one or\ two warning behaviors shortly prior to their violent actions, Offender 7 engaged in a varied and extended display of warning behaviors that could\ have been detected given sufficient monitoring. The Internet in this case pro- vides a record of this individual’s change from an interest in fantasized vio- lence to far-right extremism, identification with mass shootings, a suicidal turn, and a final resolve to carry through with his thoughts and desires\ . It is unclear if this perpetrator is an extreme outlier or if he represents a \ subgroup of offenders who are very open in sharing the details of their progression towards violence online.

Timing, Response and Outcome In addition to variation in the type and amount of warning behaviors exh\ ib- ited by these 20 offenders, there was also differences in the time between the onset of warning behaviors and the shooting act itself as demonstrated i\ n Table 6. Half of the offenders who engaged in warning behaviors gave very little time for anyone to respond to their warning behaviors: only 1 to \ 2 days lapsed between leakage or direct threat and the shooting event. The offenders with the longest time between initial warning behavior and their attack— offenders 11 and 17 at 1 year and offender 7 at 3 years—are also those who exhibited the greatest number of warning behaviors and thus the greatest\ opportunity for detection and prevention. It is notable that four out of the five 805 Table 6. Warning Behaviors and Their Response. Timing Warning behaviors Bystander Bystander response 1 3 days prior to shooting event Direct threatTargetUnknown 3 1 month prior–day of event LeakagePeers; AdultPeers did not take threats seriously and often believed offender was jok\ ing One peer dared him to commit the shooting Adult did not inform offender’s parents about concerning comments 4 2 months Identification, leakageClassmates and adult neighborsA student told a teacher about the leakage and the offender was referred\ to a Vice Principal. The offender was forced to engage in anger-management sessions with the Vice\ Principal, who was later targeted in the shooting incident.

6 3 days–day of LeakagePeersNone 7 2 years–2 months Fixation, last resort, leakage, pathwayNone Peers did not take leakage seriously and did not alert authorities; some\ encouraged offender to engage in violence 9 Day prior Direct threatTargetBoth the offender and the victim were scheduled to talk to administrator\ s on the day of the shooting 10 15 days–6 days LeakageTeachers; PeersTwo weeks prior to the shooting, the offender was interviewed by a Sheri\ ff’s Deputy and suspended for 3 days after referral by the teacher 11 1 year–day of Identification, LeakagePeersNone 12 1 day prior LeakagePeersNone 13 1 day–day of LeakagePeersNone 14 Day of LeakageOnline community; PeersNone 16 1 day–day of Identification, leakageOnline community; PeersNone 17 1 year–1 day Identification, leakage, pathwayAdult; Peers; Online communityA student told administrators about the leakage one year prior to the of\ fense. The offender was suspended and the district conducted a threat assessment. The offender w\ as not determined to be a risk and was allowed to come back to school 18 Unknown Pathway, leakageOnline communityNone 19 3 days–day of Last resort, leakagePeers; Online communityAn ex-girlfriend texted the offender’s father about his concerning be\ havior 2 minutes prior to the shooting 20 Unknown Last resortFamilyUnknown 806 Crime & Delinquency 68(5) individuals who exhibited their warning sign a month or more prior to th\ e shooting fall in the high-risk category of offenders. Despite opportunities for intervention, only five offenders in this sample were reported to authorities on at least one occasion following a warnin\ g behavior. In one instance, that of offender 19, the bystander alerted adults too late to intervene—only 2 minutes prior to the shooting event. In the \ other four cases where warning behaviors were reported, administrative action was ineffective and failed to prevent the shooting. For example Offender 17 came to the attention of school officials almost a year prior to the shooting event. A classmate told administrators that Offender 17 had a “hit list” of students who bullied him. Offender 17 was subject to a threat assessment, including an interview with a psychologist and an examination of Offender 17’s back- ground. He was not determined be a risk and was allowed to come back to \ school. His peers and teachers did not appear to agree with the results \ of this assessment. One month prior to the shooting, an unnamed student tweeted,\ “If [Offender 17] shoots up the school, I swear to God I have like four classes\ with him” (McCall, 2013). Two female teachers reportedly developed a plan to strangle offender 17 with their purse straps if he attacked them or their students (Kotowski, 2015). The remaining individuals in the sample did not have their warning behav\ - iors reported to parents, school officials, or law enforcements. In many cases, peers told authorities after the shootings that they thought the offender was joking. In two exceptional cases—that of offender 3 and 7—bystanders encouraged the offender to engage in violent behavior. One of Offender 3’s friends called him a “pussy” and dared him to carry out his threat\ (McCarthy, 2001). Some of Offender 7’s peers who were exposed to his warning behav- iors appeared to also be co-conspirators, helping in the planning of the\ shoot- ing event (Davey, 2005). Law enforcement would tell victim’s families that up to 40 individuals had heard rumors of Offender 7’s intention prior to the shooting attack and 4 to 5 students had direct knowledge of the plot, ye\ t none of them reported this to school officials or police (Robertson, 2006). In the majority of cases, bystanders appeared to have simply ignored the strange or cryptic comments their peers made in person or on social media.

Discussion and Conclusion This study conducted and analyzed 20 case studies of adolescent school shooters with a particular focus on their preparatory activities related to the shooting. We examined their warning and leakage behaviors, building on a growing body of research that has looked at such issues for rampage shoo\ ters, terrorists, and other mass shooters. We extended this important literature to Abel et al. 807 the area of school shootings, documenting the modality, the types of warning behaviors, and comparing across the high and low risk offenders in the sam- ple. In the following, we highlight a number of critical findings. First, our research findings are consistent with reports from the Secret\ Service and others that focus on the importance of multidisciplinary thr\ eat teams and the impact of critical events and stressors. Our findings, whe\ re we demonstrate differences between high and low risk offenders, show that the school shooters have a significant number of challenging deficiencies, i\ nclud- ing struggles at home, in school, in the criminal justice system, and wi\ th mental health issues. Prevention strategies should employ a comprehensiv\ e approach involving community organizations and significant investment within schools to enhance counseling and mentoring initiatives. Second, similar to the work by O’Toole (2000) and Meloy and O’Toole (2011), the results show that the authorities and school officials had clear opportunities to intervene before most of the school shootings occurred.\ Indeed, 16 of the 20 school shooters in this sample made their intention\ s clear to other people before they committed the attack. An oft-cited slogan in ter - rorism security studies, also used in other areas such as global health com- munication and trauma surgery, is “if you see something, say something!” The slogan drives home how important it is for individuals to take respo\ nsi- bility for the safety of themselves and their communities and take actio\ n. It appears difficult in the moment, however, to take such action. Parents, school officials, and law enforcement can only take action to prevent potential s\ hoot- ings if the students who are the frequent recipients of leakage are will\ ing to share that information with them. One effective strategy to prevent shootings, therefore, may be implementation of programs that teach students to dete\ ct such behaviors and develop the trust necessary for a student to feel com\ fort- able reporting these signs to administrators. Third, there were interesting differences across modality, type of warning behavior, and the time lapse between behavior and shooting. Most of the shooters, when sharing their intentions, did so off-line, primarily conveying their thoughts to their friends. It is important to remember that our sa\ mple includes cases starting in 1999. Adolescents today are much more comfort- able with social media accounts and it is possible there is a higher per\ centage of shooters providing online warnings in today’s environment. These online behaviors, particularly when expressed in anonymous web forums like offenders 7 and 17 exhibited, pose a particular challenge for detection an\ d suggest a need for increased guardianship by parental figures. One key challenge for law enforcement and school officials, as noted in Pollack et al.’s (2008) research is that the majority of warning behaviors noted in this study occurred close in time the shooting event. It is cru\ cial to 808 Crime & Delinquency 68(5) develop policies that recognize this small window of opportunity and tak\ es advantage of it to prevent the attack. Future research should examine th\ e time between when an individual first decides to commit the shooting, when th\ e shooting actually occurs, and where the warning behaviors fall within th\ at timeline. We suspect that, as in our sample, there will be some individuals who plan, evaluate, and share their ruminations prior to actually commit\ ting act. The longer a shooter’s planning period, the more warning behaviors appear and the more opportunities there are to recognize them and take a\ ction to prevent the shooting. However, other school shooters are much more spon- taneous in their decision-making, often reacting to a recent significant\ event like being suspended from school and almost immediately then committing \ the attack. Future research should explore strategies to better detect t\ he warn- ing behaviors in this limited time frame.

Limitations The generalizability of these findings to the broader population of scho\ ol shooters is limited by several factors. While our sample was randomly selected from a known population of school shooting events, the small nu\ m- ber of cases we examined (N = 20) necessarily limits the diversity of behav- iors captured in this study. It is possible, and even likely, that there is some subset of school shooters that do engage in the mass shooting subculture or exhibit warning behaviors not demonstrated amongst this group. In addition, this study was necessarily limited by its use of open-sourc\ e data. First, it is well established in the literature that news outlets \ dispropor - tionately cover stories that are novel, sensational and match cultural s\ cripts.

Thus, there was far more information available regarding mass shootings \ or shootings with unique elements than the more common single-victim tar - geted shootings. Thus, this sample of offenders was over-representative of rarer types of shootings. In addition, in most cases the school shooter’s online posts had been taken down, forcing analysis to rely on government and jo\ ur - nalist’s selections of quotes from these documents. Finally, it is highly likely that some of the shooters in this sample engaged in online activity that was either not uncovered or not released to the public by investigators. In \ particu- lar, there was no information about the webpages shooters may have visited \ but did not directly interact with. These limitations of working with online data are not unique to studying\ school shooters but do pose a considerable challenge to understanding th\ e role of digital media in acts of violence. It would serve future researchers well to form partnerships with law enforcement and social media platform\ s to gain access to primary documentation of violent offender’s online activity. Abel et al. 809 Only with more robust data can the field develop a more complete empiric\ al picture of the digital activity of violent offenders, including school shooters, and an additional point of detection and intervention.

Declaration of Conflicting Interests The author(s) declared no potential conflicts of interest with respect\ to the research, authorship, and/or publication of this article.

Funding The author(s) disclosed receipt of the following financial support for\ the research, authorship and/or publication of this article: This project was supporte\ d by Award No.

2016-CK-BX-0013, awarded by the National Institute of Justice, Office of\ Justice Programs, U.S. Department of Justice. The opinions, findings, and conclu\ sions or recommendations expressed in this publication are those of the authors and do not necessarily reflect those of the Department of Justice.

ORCID iD Meagan N. Abel https://orcid.org/0000-0002-6586-0082 Supplemental Material Supplemental material for this article is available online.

Notes 1. Offender 8 brought the firearm to school to trade for drugs, only opening \ fire on administrators when he was caught. Offender 2 brought a firearm with the intent to commit suicide.

2. Furries are individuals who are interested in being, and are sometimes s\ exually attracted to, animal characters with human characteristics (Hsu & Baile\ y, 2019).

3. Vorarephilia, shortened to vore online, is a fetish in which individuals \ are aroused the idea of being eating, eating another person, or observing so\ meone being eaten (Lykins & Cantor, 2014).

References Addington, L. A. (2003). Students’ fear after columbine: Findings f\ rom a randomized experiment. Journal of Quantitative Criminology, 19(4), 367–387.

Alathari, L., Drysdale, D., Driscoll, S., Blair, A., Carlock, A., Cotkin\ , A., Johnston, B., Foley, C., Mauldin, D., McGarry, J., Nemet, J., Vineyard, N., & Bull\ winkel, J. (2019). Protecting America’s schools: A U.S. secret service analysis of tar- geted school violence. U.S. Secret Service National Threat Assessment Center.

Albuquerque Journal staff. (2014, January 17). Police look at Campbell’s Instagram posts. Albuquerque Journal. 810 Crime & Delinquency 68(5) Anderson, M., Kaufman, J., Simon, T. R., Barrios, L., Paulozzi, L., Ryan, G., Hammond, R., Modzeleski, W., Feucht, T., & Potter, L. (2001). School-associated violent deaths in the United States, 1994-1999. JAMA, 286(21), 2695–2702.

https://doi.org/10.1001/jama.286.21.2695 AP staff. (2001, March 23). Latest school shooting has parents seeking alternatives. News on 6.

Baird, A. A., Roellke, E. V., & Zeifman, D. M. (2017). Alone and adrif\ t: The associa- tion between mass school shootings, school size, and student support. The Social Science Journal, 54(3), 261–270.

Bersani, B. E., & Chappie, C. L. (2007). School failure as an adolesce\ nt turning point. Sociological Focus, 40(4), 370–391.

Borum, R., Cornell, D. G., Modzeleski, W., & Jimerson, S. R. (2010). W\ hat can be done about school shootings? A review of the evidence. Educational Researcher, 39(1), 27–37.

Borum, R., Fein, R., Vossekuil, B., & Berglund, J. (1999). Threat asse\ ssment: Defining an approach for evaluating risk of targeted violence. Behavioral Sciences & the Law, 17(3), 323–337.

Chung, I. J., Hill, K. G., Hawkins, J. D., Gilchrist, L. D., & Nagin, D.\ S. (2002). Childhood predictors of offense trajectories. Journal of research in crime and delinquency, 39(1), 60–90.

Davey, M. (2005, March 29). In a twist, a second youth is accused in s\ chool attack. The New York Times.

Duwe, G. (2020). Patterns and prevalence of lethal mass violence. Criminology & Public Policy, 19(1), 17–35. https://doi.org/10.1111/1745-9133.12478 Elder, G. H., Jr. (1995). The life course paradigm: Social change and \ individual devel- opment. In Moen, P., Elder, Jr, G. H., & Lüscher, K. (Eds.), Examining lives in context: Perspectives on the ecology of human development (p. 101–139).

American Psychological Association. https://doi.org/10.1037/10176-003 Elsass, H. J., Schildkraut, J., & Stafford, M. C. (2014). Breaking new\ s of social prob- lems: Examining media consumption and student beliefs about school shoot\ ings.

Criminology, Criminal Justice, Law and Society, 15(2), 31.

Fein, R. A., Vossekuil, B., & Holden, G. A. (1995). Threat assessment: An approach to prevent targeted violence. National Institute of Justice. https://doi.org/10.1037/ e517592006-001 Fergusson, D. M., Horwood, L. J., & Nagin, D. S. (2000). Offending tra\ jectories in a New Zealand birth cohort. Criminology, 38(2), 525–552.

Freilich, J. D., Chermak, S. M., Connell, N. M., Klein, B. R., & Greene-\ Colozzi, E. A. (2020). Final report: Understanding the causes of school violence using open source data (NIJ Grant #: 2016-CK-BX-0013). National Institute of Justice, OJP, Department of Justice.

Fridel, E. E. (2019). The contextual correlates of school shootings. Justice Quarterly, 1–30. https://doi.org/10.1080/07418825.2019.1666907 Gerard, F. J., Whitfield, K. C., Porter, L. E., & Browne, K. D. (2016)\ . Offender and offence characteristics of school shooting incidents. Journal of Investigative Abel et al. 811 Psychology and Offender Profiling, 13(1), 22–38. https://doi.org/10.1002/ jip.1439 Graf, N. (2018). A majority of U.S. teens fear a shooting could happen\ at their chool, and most parents share their concern. Pew Research Center. https://www.pewre- search.org/fact-tank/2018/04/18/a-majority-of-u-s-teens-fear-a-shooting-\ could- happen-at-their-school-and-most-parents-share-their-concern/ Harvey, K. (2014). Taft HS shooting trial Day 4. Bakersfield Now. https://bakers- fieldnow.com/news/local/taft-hs-shooting-trial-day-4-defendant-oliver-takes- the-stand Held, T. (2007, July 28). Hainstock angry, violent DA says. Milwaukee Journal Sentinel.

Holland, K. M. (2019). Characteristics of school-associated youth homi\ cides—United States, 1994–2018. MMWR. Morbidity and Mortality Weekly Report, 68(3), 53– 60. https://doi.org/10.15585/mmwr.mm6803a1 Hsu, K. J., & Bailey, J. M. (2019). The “Furry” phenomenon: Char\ acterizing sexual orientation, sexual motivation, and erotic target identity inversions in male fur- ries. Archives of Sexual Behavior, 48(5), 1349–1369.

Kalesan, B., Lagast, K., Villarreal, M., Pino, E., Fagan, J., & Galea, S. (2017). School shootings during 2013–2015 in the USA. Injury Prevention, 23(5), 321–327.

Kaufman, G. (2005, March 23). Minnesota shooter’s friend says he wa\ rned her of attack. MTV.

Kotowski, J. (2014, November 13). Trial of school shooter opens with testimony from victim. Bakersfield Californian.

Kotowski, J. (2015, January 23). Court filing: Taft teachers, aware of\ bullying, planned to strangle Oliver if needed. Bakersfield Californian.

Langman, P. (2009). Rampage school shooters: A typology. Aggression and Violent Behavior, 14(1), 79–86. https://doi.org/10.1016/j.avb.2008.10.003 Langman, P. (2015, April 1). Charles Andrew Williams: Sorting out the \ contradic- tions. SchoolShooters.info.

Langman, P., & Straub, F. (2019). A Comparison of Averted and Completed School Attacks from the Police Foundation Averted School Violence Database. Office of Community Oriented Policing Services.

Laub, J. H., & Sampson, R. J. (2001). Understanding desistance from cr\ ime. Crime and Justice, 28, 1–69.

Lykins, A. D., & Cantor, J. M. (2014). Vorarephilia: A case study in m\ asochism and erotic consumption. Archives of Sexual Behavior, 43(1), 181–186.

Lyons, B. (2004, February 14). Cellphone records, home yield clues. The Times Union.

Maguire, B., Weatherby, G. A., & Mathers, R. A. (2002). Network news c\ overage of school shootings. The Social Science Journal, 39(3), 465–470. https://doi.

org/10.1016/S0362-3319(02)00201-X Marquez Estrada, H. & Nixon, R. (2005, March 24). An internet trail of\ a boy’s death wish. Minneapolis Star-Tribune. http://www.sitnews.us/0305news/032405/032405_ shns_schshoot.html 812 Crime & Delinquency 68(5) McCall, D. (2013, January 15). Witnesses describe shooting scene insid\ e the class- room. Midway Driller.

McCarthy, T. (2001, March 11). Warning: Andy Williams here. Time Magazine.

Meloy, J. R., Hempel, A. G., Mohandie, K., Shiva, A. A., & Gray, B. T. (2001). Offender and offense characteristics of a nonrandom sample of adolescent\ mass murderers. Journal of the American Academy of Child & Adolescent Psychiatry, 40(6), 719–728.

Meloy, J. R., & O’Toole, M. E. (2011). The concept of leakage in th\ reat assessment. Behavioral Sciences and the Law, 29, 513–527.

Newman, K., Fox, C., Roth, W., Mehta, J., & Harding, D. (2004). Rampage: The social roots of school shooters. Perseus Books Group.

O’Toole, M. E. (2000). The school shooter: A threat assessment perspective. Critical Incident Response Group, National Center for the Analysis of Violent Cri\ me.

Pollack, W., Modzeleski, W., & Rooney, G. (2008). Prior knowledge of potential school-based violence: Information students learn may prevent a targeted attack.

United States Secret Service and United States Department of Education.

Ragsdale, J. (2006, January 31). 39 Knew of Red Lake killer’s plan.\ Twin Cities Pioneer Press.

Register-Guard staff. (2007, April 13). A painful reminder: School shooting rein- forces need to communicate. The Register-Guard.

Roberts, J. (2005, March 22). Troubled life of Minnesota shooter. CBS News.

Robertson, T. (2006, January 30). Prosecutor: Dozens knew school shoot\ er’s plans. Minnesota Public Radio.

Rocque, M. (2012). Exploring school rampage shootings: Research, theor\ y, and pol- icy. The Social Science Journal, 49(3), 304–313.

Sampson, R. J., & Laub, J. H. (1993). Crime in the making: Pathways and turning points through life. Harvard University Press.

Schildkraut, J., Elsass, H. J., & Stafford, M. C. (2015). Could it hap\ pen here? Moral panics, school shootings, and fear of crime among college students. Crime, Law and Social Change, 63(1), 91–110.

Schukar, A. (2011, January 6). Suspension ignited fury in Millard Sout\ h shooting. Omaha World Herald.

Silver, J., Hogan, J., & Gill, P. (2018). Foreshadowing targeted viole\ nce: Assessing leakage of intent by public mass murderers. Aggression and Violent Behaviors, 38, 94–100.

Stine, M. (2007, April 13). Shooter was crying out for help, friends s\ ay. Pamplin Media Group.

Texiera, E., Krikorian, G., & Martelle, S. (2001, March 23). 5 Hurt in gunfire at high school near San Diego; Student is held. Los Angeles Times.

Verdantam, S. (2005, March 23). Gunman found haven on Nazi web site. The Washington Post. http://www.washingtonpost.com/wp-dyn/articles/A58397- 2005Mar22.html Vossekuil, B., Fein, R. A., Reddy, M., Borum, R., & Modzeleski, W. (200\ 2). The final report and findings of the Safe School Initiative. US Secret Service and Department of Education. Abel et al. 813 Vossekuil, B., Fein, R. A., Reddy, M., Borum, R., & Modzeleski, W. (200\ 4). The final report and findings of the safe school initiative: Implications for the \ prevention of school attacks in the United States. United States Secret Service and United States Department of Education.

Wiesner, M., & Capaldi, D. M. (2003). Relations of childhood and adole\ scent fac- tors to offending trajectories of young men. Journal of Research in Crime and Delinquency, 40(3), 231–262.

Author Biographies Meagan N. Abel is a doctoral student in the School of Criminal Justice at Michigan State University and recipient of a University Distinguished Fellowship. Her research primarily focuses on public perceptions of and reactions to extremist an\ d mass-casu- alty violence in the United States.

Steven Chermak is a Professor of Criminal Justice at Michigan State University and a lead investigator for the Natinal Counterterrorism, Innovation, Teachi\ ng and Education Center (NCITE), a Center of Excellence of the U.S. Departmen\ t of Homeland Security. Dr. Chermak’s research includes activities in the \ following areas:

far-right extremism, the effectiveness of strategies used to prevent ter\ rorism and crime, and the media’s role in relation to crime and terrorism issues\ . He and Professor Joshua Freilich have collaborated to create the Extremist Crime Database\ —the first of its kind National Database on criminal activities involving U.S. far rig\ ht, far left, and Jihadist extremists. Other research has focused on terrorism and media coverage of terrorist activities, including depictions of the militia movement and t\ he September 11th attacks. Current research includes research on the risk and protect\ ive factors related to extremist behavior, school violence, and predictors of violen\ ce.

Joshua D. Freilich is a professor at John Jay College, and a Creator and co-Director of three open source database studies: U.S. Extremist Crime Database (E\ CDB), U.S School Shooting Database (SSDB), and the U.S. Extremist Cyber Crime Database (ECCD). He is a member of two DHS Centers of Excellence (COE), NCITE\ and CAOE and a past member of START, an emeritus COE. Freilich’s research\ has been funded by DHS and NIJ and focuses on the causes of and responses to bias\ crimes, terrorism, extremist crime, cyber-terrorism, and targeted mass violence; open source research methods and measurement issues; and criminology theory, especia\ lly situa- tional crime prevention.