You are on page 1of 14

Deviant Behavior

ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/udbh20

Understanding School Shootings with Crime Script


Analysis

David A. Keatley, Sian Mcgurk & Clare S. Allely

To cite this article: David A. Keatley, Sian Mcgurk & Clare S. Allely (2020) Understanding
School Shootings with Crime Script Analysis, Deviant Behavior, 41:9, 1084-1096, DOI:
10.1080/01639625.2019.1596543

To link to this article: https://doi.org/10.1080/01639625.2019.1596543

Published online: 29 Mar 2019.

Submit your article to this journal

Article views: 1349

View related articles

View Crossmark data

Citing articles: 3 View citing articles

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=udbh20
DEVIANT BEHAVIOR
2020, VOL. 41, NO. 9, 1084–1096
https://doi.org/10.1080/01639625.2019.1596543

Understanding School Shootings with Crime Script Analysis


David A. Keatleya, Sian Mcgurkb, and Clare S. Allelyc
a
Murdoch University, Perth, Australia; bUniversity of Lincoln, Lincoln, UK; cUniversity of Salford, Manchester, UK

ABSTRACT ARTICLE HISTORY


Mapping the antecedents of school shootings is an important step towards Received 28 March 2018
understanding and predicting when the tragedies occur. This study uses Accepted 8 August 2018
Crime Script Analysis (CSA) to map the variety of behaviors and precursors
in sequence. An in-depth analysis of 16 school shooters in the United States
of America was used to build a crime script. The results showed five key
scenes through which school shooters progressed in the lead-up to their
school shootings. The approach is the first to provide a temporal account of
school shooters planning of their crimes and forms the foundation for
future school shootings to be added to. This research also shows how
a CSA approach can be used to generate crime prevention strategies for
future school shootings.

It used to be the case that parents could send their children to school without fear for the children’s
lives. Publicity surrounding recent events in America have raised awareness and concerns about the
safety of children in high schools (Abouk and Adams 2013; Rocque 2012). While school shootings
are not a new phenomenon, there remains uncertainty about the antecedents of these tragedies
(Gerard et al. 2016; Meloy et al. 2001). Multiple risk factors have been identified in the literature,
such as exclusion and isolation, bullying, and revenge motivations (Langman 2013; Leary et al. 2003).
However, individual risk factors studied and presented alone do not provide a dynamic account of
how individuals prepare for and execute their plans.
There was a sudden growth in research on school shootings in the 1990s, following the events at
Columbine High School, Thurston High School, and Westside Middle School, as well as others (Baird,
Roellke, and Zeifman 2017). Despite an increased focus on research into school shootings there remains
a lack of clear understanding about what causes these events, and how to predict them. Some general
trends have been shown, relating to economic status, ethnicity, and interpersonal disputes (Wike and
Fraser 2009). General school violence is likely to be committed by low-income, ethnic minorities in
urban school environments (Wike and Fraser 2009); school shootings, however, are more likely to be
committed by white, middle-class students (Baird, Roellke, and Zeifman 2017). While general trends
may provide the basis of understanding social processes and indicators of future violence, they do not
offer specific accounts of the temporal dynamics preceding a school shooting.
Some of the most extensive and well-cited studies of school shootings include one which was
conducted by the Federal Bureau of Investigation (FBI; O’Toole 2000) and another conducted by the
United States Secret Service and the United States Department of Education (Vossekuil et al. 2002).
Borum et al. (2010) have highlighted that one of the things that makes these studies notable is that
they both rejected efforts to develop a profile of the individual who perpetrates a school shooting.
They also both discouraged the use of lists of warning signs or “checklists” as they are too simplistic.
Instead, the focus should be shifted towards “threat assessment as a prevention strategy” (Borum
et al. 2010). In the FBI study, a review was carried out of 14 cases of actual shootings and also four
cases of planned shootings that were thwarted (O’Toole 2000). O’Toole identified 47 descriptors that

CONTACT David A. Keatley drdavekeatley@gmail.com School of Law, Murdoch University, Perth 6150, Australia
© 2019 Taylor & Francis Group, LLC
DEVIANT BEHAVIOR 1085

were common across many shooters. Some of these common descriptors including 28 personality
traits and behaviors, seven family dynamics, seven school dynamics, and five social dynamics. It was
highlighted by O’Toole (2000) that not all the shooters had each of these features, but the identified
dynamics were seen as constituting significant trends. Some of the common individual features
included narcissism (which is consistent with a number of studies, e.g., Bondü and Scheithauer
2015), bigotry, alienation, poor anger management, a fascination with violence, low levels of self-
esteem and a lack of empathy.
The FBI study (O’Toole 2000) was based on a five-day conference, which was attended by 160
invited experts and professionals in law enforcement, education, and mental health. The conference
also included school staff members who had experienced a school shooting first-hand. The 160
delegates collectively assessed 18 completed or thwarted school shooting cases. The FBI experts in
criminal profiling argued that profiling was an inappropriate method for the prevention of school
shootings. It was unequivocally agreed that there exists no single set of characteristics that could
define a would-be school shooter with sufficient levels of specificity to make it of practical value.
They argued that the development of a “checklist” of warning signs to identify a would-be school
shooter can be short-sighted and potentially dangerous. It is possible that some students who have
no intention of perpetrating such an attack (are non-violent individuals) would be labeled as
potentially dangerous (O’Toole 2000). The FBI report recommended that an assessment should
comprise four key domains including personality traits and behaviors, family dynamics, school
dynamics, and social dynamics (Borum et al. 2010). There is no single cause that propels or drives
an individual to commit a school shooting. School violence needs to be seen as the outcome of
a wide range of causal processes (Henry 2009).
In the study conducted by the United States Secret Service and United States Department of
Education (Vossekuil et al. 2002), 37 incidents of school violence involving 41 students between 1974
to 2000 were examined. Findings indicated a variety of common features among the school shooters.
The majority of the shooters were depressed, had feelings of persecution, had grievances against at
least one of their targets and had an interest in violent entertainment. Interestingly, a history of drug
abuse, prior violence or criminal behavior or animal cruelty were not found in the majority of the
school shooters.
There have also been some other well-cited studies investigating school shooters. Verlinden, Hersen,
and Thomas (2000) carried out a review of risk factors among 10 school shooters. Similar to the O’Toole
(2000) study, a number of areas were examined including individual, family, school/peers, and societal/
environmental factors were examined in the study. The factors which were identified as being most
common were: a history of aggression, uncontrolled anger, depression and suicidal ideation, discipline
problems and feeling rejected and bullied (“picked on”). Another study carried out by Meloy et al. (2001)
reviewed 37 adolescent mass murderers. Eight of these 37 adolescent mass murderers were classified as
“classroom avengers” and are therefore of specific interest here. Findings showed that the school shooters
often were bullied but did not bully others. They were also found to be preoccupied with weapons and
fantasy. Many also had a history of substance abuse. The majority were not depressed and had no history
of antisocial behaviors. In another study, Leary et al. (2003) reviewed 15 school shootings and found
a number of common features. The sample comprised of cases which were well-documented cases of
school violence which occurred in the United States between January 1995 to March 2001. They started
from 1995 because that was when school shootings started to receive national attention. Findings from
the study showed that acute or chronic rejection (in the form of ostracism, bullying, and/or romantic
rejection) was present in all but two of the incidents. It was also found that there was a tendency for the
shooters to be characterized by one or more of three other risk factors, namely, an interest in firearms or
bombs, a fascination with death or Satanism, or psychological problems involving depression, impulse
control, or sadistic tendencies (Leary et al. 2003). The findings by Leary and colleagues are consistent
with the hypothesis that social rejection is a contributory factor in the majority of school shooters.
Specifically, 12 of the cases involved an ongoing pattern of teasing, bullying, or ostracism. Also, at least six
of the school shooters had experienced a recent romantic rejection. No clear evidence of rejection was
1086 D. A. KEATLEY ET AL.

found in only two of the 15 school shooters (Leary et al. 2003). Such findings support earlier findings
(e.g., Vossekuil et al. 2002). Vossekuil et al. (2002) found evidence for bullying, ostracism, and social
rejection in over two-thirds of the school shooting cases they examined (Vossekuil et al. 2002).
There are a variety of reasons which makes the study of school shooters difficult to study
(Langman 2009). For instance, thankfully, the event is relatively rare so sample sizes are relatively
small across studies. Also, the perpetrators very often commit suicide or are killed by police at the
scene. This further limits researchers to a retrospective review of the perpetrators’ lives, and/or
interviews with people who knew the perpetrators. Lastly, across studies investigating school
shooters, there are varying definitions of what a school shooting or a rampage school shooting is.
This results in somewhat different, but nevertheless overlapping, populations being investigated.
Some samples studied by researchers include student-perpetrated firearms deaths which took place
at school, while other samples studied by researchers include perpetrators of larger-scale attacks
(Langman 2009). Newman (2004) and Newman and Fox (2009) define rampage school shootings as
involving students who attend (or had formerly attended) the school where the shooting occurred;
took place on a school-related “public stage” (i.e., in full view of other people); and involving
multiple victims, at least some of whom were shot randomly or as a symbol (such as a principal
who represents the school). Other victims may be targeted as a result of a grievance or perceived
wrong (Langman 2009). The definition of school shooters outlined by Newman (2004) is the
definition used in the present study.
Clearly, the focus of research is being directed towards crime prevention through understanding
the complex dynamics leading-up to school shootings. This is the key focus of the current study, too:
to elucidate the complex pattern of behaviors and antecedent events that precede a school shooting.
Understanding individual risk factors for school shooting provides an overview of key warning signs
to be vigilant towards; however, mapping the temporal relationships between risk factors can provide
an insight into the predictive pathways that lead towards crime (Keatley 2018).
Recently, research in criminology has begun to focus on temporal methods, such as Behaviour
Sequence Analysis (BSA, Keatley 2018; Keatley et al. 2018), Indicator Waves (Keatley and Clarke 2018),
and Crime Script Analysis (Leclerc and Wortley 2014), which allow researchers to clearly map the life
histories and antecedent events and behaviors that lead to a school shooting. The focus of temporal
research is not to undermine the important research into risk factors; but, to take a step further and ask
not only what the risk factors are; but, whether the order they occur in matter. From here, important
issues can be highlighted, such as key “hot spots” or “critical points” (Keatley 2018), wherein inter-
ventions are most effective. Clearly, school shootings are underpinned by a wide variety of behaviors or
“risk factors”, and no single behavior is reliably predictive of later events. However, temporal analyses
allow researchers to understand whether the progression of several behaviors is more likely to lead to
a school shooting than other sequences.
However, a limitation of BSA research is that it requires complete, detailed datasets, which are
often hard to gather comprehensively in real-world events. Information regarding every step and
detail of school shootings is either vague, missing, or classified. This poses a problem for traditional
BSA approaches, as misleading chains may occur (if events or behaviors are systematically missing
across samples). There are, however, other temporal methods that provide an overview of the general
trends and chains in behavior, while requiring less detailed data, such as Crime Script Analysis (CSA;
Cornish 1994; Ekblom and Gill 2016).
Crime script analysis (CSA; Cornish 1994; Ekblom and Gill 2016) was developed from general
schemata and scripts in Cognitive Psychology (Schank and Abelson 2013); though, it can also be
conceptualized in terms of computer scripts in programming. In the same way that actors in a play
follow a theatrical script, with acts and scenes, researchers have suggested that criminals follow
a similar script in the commission of their crime (Leclerc, Smallbone, and Wortley 2013; Leclerc and
Wortley 2014). In a similar psychological process as individuals entering a restaurant and knowing
the general progression of events, as criminals act out their crimes they follow scripts that direct or
guide their behavior. Nisbett and Ross (1980) outline scripts as being temporal sequences that extend
DEVIANT BEHAVIOR 1087

across time and infer some causal relationships between scenes, in that earlier behaviors or events
precede or enable later events.
Crime Script Analysis is important in being able to understand the sequential flow of behaviors
and actions, allowing deeper understanding and insight into the process of crime. This provides
investigators with a more detailed account of how crimes unfold over time, rather than “snapshot”
approaches, which portray crime as a single event in space and time. However, a more important
output of CSA is in terms of crime prevention strategies. Recent research has focused on crimes such
as burglary (Homel, Macintyre, and Wortley 2014) and child sexual assault (Leclerc, Smallbone, and
Wortley 2013) through a CSA lens, not only to understand the process of the crimes; but, to
highlight key stages of the script in which intervention might have more effect. A similar approach
has been taken with terrorist acts to highlight how interventions such as transportation might have
been an effective deterrent (Meyer 2013). In terms of school shootings, purchasing or acquiring
a weapon necessarily precedes the school shooting event. The CSA approach outlines what other
behaviors and events are related to such key actions, and further analyses then allow researchers to
provide intervention strategies.

Current study
The aim of the current paper is to provide a novel method for mapping and describing the temporal
development of school shootings. Crime Script Analysis (CSA; Cornish 1994; Ekblom and Gill 2016) will
be outlined and shown to be an effective method, which has not been previously applied to the area of
school shootings. Crime Script Analysis (CSA) is a methodological approach that builds on the cognitive
psychology approach of script analysis (Schank and Abelson 2013), using schemata (Cornish 1994).
Script analysis essentially outlines the general trends of behavior, with the quintessential example being
the restaurant script, in which people can operate in different restaurants successfully by following
a generalized script of how to behave in such contexts. The script concept has been applied to crimes
(Beauregard et al. 2007; Ekblom and Gill 2016; Hutchings and Holt 2015), showing that criminals follow
a similar script and schemata process when planning and commissioning their crimes.
The aim of the current study is to investigate the behavioral sequences of school shootings, using
a crime script analysis approach. The general script will outline the preparation and planning stage
behaviors, and how these lead to the commission of the act itself, which leads to the “exit” or final
behaviors. A series of school shooting events will be used to define and clarify the scenes in the crime
script analysis, and provide examples of the typical behaviors exhibited in each scene. This is the first
paper, to the authors’ knowledge, to explicitly link crime script analysis to school shootings and
provides an initial framework to understand the dynamics of school shootings, and a starting point
for future cases to be added to.

Methods
Sample
The sample consisted of 16 individual cases of school shootings in the United States of America (see
Table 1). The majority of cases involved in a male shooter (n = 15); and the average age was 20.68 years
(SD = 10.09, range = 11–55). An average of 11.88 (SD = 12.41, range = 2–45) people were killed in the
school shootings. An average of 15.44 (SD = 14.14, range = 1–58) people were injured in the school
shootings. Guns were the most used weapon; however, bombs or dynamite were also used in several
cases. There were almost equal numbers of suicides (n = 6) as arrests (n = 7) in the sample.
When identifying individuals for the sample a variety of sources were consulted, including databases
for school shootings (e.g., schoolshooters.info; PACER.), news reports, court reports, police and
government reports, and general media accounts. In cases where data were not sourced from official
1088

Table 1. Descriptive information of school shooters.


Arrested/
Name Age Date Gender Location Type/name of school Number killed Number injured Weapons used Killed/ suicide
Andrew Kehoe 55 1st Feb 1872 M Michigan (U.S.) Bath Consolidated school 45 58 Bombs & dynamite Suicide
Charles Whitman 25 1st Aug 1966 M Austin (U.S.) University of Texas 14 31 Guns Killed by police
Brenda Spencer 16 29th Jan 1979 F California (U.S.) Grover Cleveland Elementary school 2 9 Gun Arrested
Eric Christopher Houston 20 1st May 1992 M California (U.S.) Lindhurst High school 4 9 Guns Arrested
Luke Woodham 16 1st Oct 1997 M Mississippi (U.S.) Pearl High school 3 7 Guns & knives Arrested
Michael Carneal 14 1st Dec 1997 M Kentucky (U.S.) Heath High school 3 5 Guns Arrested
Andrew Golden 11 24th March 1998 M Arkansas (U.S.) Westside Middle school 5 10 Guns Arrested (later released)
D. A. KEATLEY ET AL.

Mitchell Scott Johnson 13 24th March 1998 M Arkansas (U.S.) Westside Middle school 5 10 Guns Arrested
Kipland Phillip Kinkel 15 30th Aug 1998 M Oregon (U.S.) Thurston High school 4 26 Guns Arrested
Dylan Klebold 17 20th April 1999 M Colorado (U.S.) Columbine High school 13 20 Guns & pipe bombs Suicide
Eric Harris 18 20th April 1999 M Colorado (U.S.) Columbine High school 13 20 Guns & pipe bombs Suicide
David Attias 20 23rd Feb 2001 M California (U.S.) University of California, Santa Barbara campus 4 1 Car Arrested
Seung Hui Cho 23 16th April 2007 M Virginia (U.S.) Virginia Tech University 32 18 Guns Suicide
Adam Lanza 20 14th Dec 2012 M Connecticut (U.S.) Sandy Hook Elementary school 27 2 Guns Suicide
Elliot Rodger 22 23rd May 2014 M California (U.S.) University of California, Santa Barbara campus 6 14 Knives & handgun Suicide
Christopher Harper Mercer 26 1st Oct 2015 M Oregon (U.S.) Umpqua Community college 10 7 Guns Suicide
Evidence of mental health issues/brain injury/ History of bullying Parents Expelled from Academic attainment Previous Access to weapons (owned by shooter/ Threats/
neurodevel-opmental disorders in childhood divorced/ school deterioration criminal family) & weapon proficiency verbalising violent thoughts to
separated in history others, in writing etc
childhood
UN UN UN UN UN N Y Y
Y UN N Discharge from Y N Y Y
military
Y UN Y N Y Y Y
Y UN UN N – dropped N N UN Y
out
Y Y Y N N N N UN
Y Y N N N N N Y
N N N N UN UN Y UN
UN N Y UN UN Y N Y
Y UN N N UN Y Y Y
N Y N N N Y N Y
Y Y N N N Y N Y
Y N N N UN UN Y UN
Y Y N N UN N N Y
Y N Y N Y N Y Y
Y Y Y N UN N N Y
Y Y Y Y – discharged UN N Y Y
from army
Note. Y = yes (present); N = no (none/absent); UN = unknown; school shooters are listed individually, even if they worked in pairs (e.g., Klebold & Harris; Golden & Johnson).
DEVIANT BEHAVIOR 1089

Law Enforcement sources, facts were cross-referenced between independent sources to ensure validity.
While this does not guarantee accuracy, it is an approved method in applied research (Keatley 2018).
Inclusion criteria were that the school shooting was documented in enough detail to allow clear
and comprehensive analyses. Inclusion criteria in the current study were that: the offense happened
in a “public location” – here operationalized as the school property; there were multiple victims
(both killed and injured); multiple victims targeted and untargeted; the offender had current or
former links with the school. A final inclusion criterion was method-specific and considered the
depth and quality of information available. For CSA to provide a systematic, step-by-step account of
developing behaviors, reports of the school shooting had to provide enough detail to satisfy this
criterion. This was important as the current research forms the foundation framework for future
research; therefore, complete accounts allowed for more informed crime script analysis. Future cases
with less data may be included as long as the base framework is well informed. This allows for
consistent comparative analyses between cases.

Script development
Crime scripts can be built from a variety of data, and there are no clear guidelines on the type of data
that can be used (Borrion 2013); indeed, the inclusion of multiple sources and types of data is
a strength of the crime script approach. There have been several revisions to the crime script
approach (Ekblom and Gill 2016); therefore, the current study will apply the most recent advance-
ments in crime script developments, to provide clear and concise scripts that break down long chains
of behaviors into five clear scenes or stages: preparation; pre-condition; instrumental actualization;
doing; and exit. The current study will follow Leclerc, Wortley, and Smallbone’s (2011) recommen-
dations, and include a visualization approach to outline the crime script and details. The current
study visualizes the data in the form of a Crime Script table (see Table 2).
The process of developing a crime script in the current study followed previously approved
protocols in the literature (Chiu 2015). First, cases were found with in-depth, sequential information
(rather than simply stating, “suspect had a gun”, the information needed to be grounded in some
form of timeline so that we could understand when they obtained the gun). Once multiple cases of
school shooter timelines had been sourced, a process of familiarization took place, wherein research-
ers read through each timeline and began mapping the sequential flow of scenes within each case,
using CSA scenes as a building block. For each case, behaviors were analyzed and discussed in terms
of which scene they belonged to. This process was iterated for each new case, and scenes were
developed to represent the data, rather than the data being forced to fit into scenes, which is why the
scene codes for the current study follow a similar structure as previous research; but, have different
scene names. Once researchers agreed that all behaviors were included in the crime script, and the
scenes within the crime script were sequential but distinct, the CSA was finalized.

Coding procedure
Cornish (1994) suggested that a standardized method for promoting richer, more comparable crime
scripts is to use a universal script. The universal script allows researchers to develop and organize
crime scripts in a systematic way that provides a cohesive framework and allows for future research
to easily build on. The universal script also outlines procedural aspects and requirements of the
planning and commissioning of a crime, at a more generalized, global level. More specific scenes can
then be developed for particular subtypes of related crimes (e.g., “school shootings” → Schools/
Colleges/Universities). As the present research is the first of its kind, a global, universal script was
developed. The data in the current study were initially coded through open coding and data-driven
approaches, using the universal script scenes proposed by Cornish (1994). All scenes and informa-
tion therein were agreed on by two researchers.
1090

Table 2. Crime script analysis of school-shooter behaviors.


School-shooter Universal script
script scene scene Behaviours Examples
1. Influences Preparation ● Attended therapy, parents separated. ● Lanza – mother = gun enthusiast
● Raised around firearms, previous crimes. ● Whitman – taught by parents to handle guns
● Said to have shown angry behaviors and may have previous ● Kinkel – father bought him the gun
charges for unrelated crimes.
D. A. KEATLEY ET AL.

● Mercer – showed interest in mass shootings & IRA

2. Operational Pre-condition ● Stole from parents or given to them by a parent. Stole weapon
● Knives and homemade bombs also used. ● Lanza (Mother’s); Corneal (own and friend’s house)
Given weapon
● Spencer (Father); Kinkel (Father)

3. Reconnaissance Instrumental ● Their own school/university, currently attending or previously. Own/Current School
actualization ● Johnson, Golden, Woodham, stole a parent’s car for transport. ● Lanza; Klebold & Harris
Previous School
● Kinkel

4. Activity Doing ● Some began at home before making their way to the school/ Harm family members
university. ● Lanza (Mother); Kehoe (Wife); Whitman (Wife, Mother); Woodham (Mother);
● Most seemed disorganized, shooting randomly. Kinkel (Father and Mother)
● Some showed clear motivation to kill certain students. Target individual
● Woodham (Girlfriend)

5. Withdrawal Exit ● Several committed suicides at the scene. Suicide


● All who were arrested showed some sort of resistance. ● Lanza, Kehoe
● Some given diagnosis after arrest. Arrested

● Houston, Kinkel, Woodham


DEVIANT BEHAVIOR 1091

Results
Cornish (1994) suggested a series of universal script scenes should be used in crime script analysis
(preparation; entry; pre-condition; instrumental actualization; doing; post-condition; exit). These
were used as a starting point in the current research. After analyzing and structuring the data in the
current study, five empirical scenes emerged (see Table 2). These five script scenes were renamed to
make them more specific and relevant for a school shooting crime script; however, their overlap with
Cornish’s universal script have been outlined in Table 2 for clarity. The five scenes in the current
crime script analysis were: Influences; Operational; Reconnaissance; Activity; and Withdrawal.

Crime script analysis


Influences
The influences scene, which relates to Cornish’s (1994) preparation scene, relates to the school
shooters upbringing and past events that may be seen to be precursors or influencers of the school
shooting. In this scene, factors like parents separating (n = 6) and the individual having mental
health issues (n = 11) were common influential precursors. Furthermore, the influencing effects of
being raised in a pro-gun culture (i.e., being raised in a family where guns are prevalent) is observed
in school shooters’ childhoods. For example, Lanza, Kinkel, and Whitman were raised around guns
and taught to handle them by their parents. Many school shooters are also known to have exhibited
anger issues and have a criminal history for unrelated crimes (n = 5). Being bullied was seen in
a number of cases (n = 6), as well as a marked deterioration in school academic performance (n = 2;
however, there were only data available for 7 of the cases).
It would be easy to suggest that the presence of guns in childhood is a contributing factor to later
school shootings; however, the influence may be more nuanced than this. It is also not as easy as
blaming the parents for having guns. Indeed, Kinkel’s father, for instance, tried to discourage gun
use and enrolled Kinkel on a gun safety course. Also, it is not simply guns alone that are a recurring
factor in this scene. Anger issues and criminal misdemeanors are also prevalent in the sample, for
example, 13 of the 16 cases had a history of verbalized or written threats and violent ideation.
Therefore, it is a number of factors, occurring in chains that are precursors to later school shootings,
in this scene.

Operational
The operational scene relates to Cornish’s (1994) pre-condition, scene and outlines the initial pre-
paratory behaviors that school shooters exhibited. The preparatory behaviors may be subtle and not
immediately overt, which is why the crime script approach is so important – allowing individuals to
become aware of progressive behaviors, however subtle. The development of ideation to written and
verbalized anger, for instance, could be an example of operationalizing ideas to actions. An important
factor in this scene is the acquisition of the weapon, which was later used in the school shooting. In
several cases (e.g., Spencer, Kinkel) the school shooter was given a weapon. However, when a weapon
has not been given to the individual, then they may resort to stealing a weapon from their parent (e.g.,
Lanza) or another source (e.g., Carneal). In this scene, school shooters may also begin researching and
preparing the other weapons used in school attacks (e.g., bombs).
It is important to note that there is no clear, systematic pathway that individuals use to obtain
guns. There is no clear single indicator that predicts later school shootings. More problematic for
prevention is the fact that in several cases weapons were obtained from parents as a gift or present (n
= 8). In these cases, there is no warning behavior for parents or friends to become aware of – it is not
the case that a parents gun goes missing and parents should become aware of the dangers. If the
school shooter is given a weapon, then they legitimately own it and can use it without warning
precursors.
1092 D. A. KEATLEY ET AL.

Reconnaissance
The reconnaissan, relates to the behaviours and movements that individuals begin once preparations
have been made. Obviously, in the case of school shootings, many students are or have recently
attended the school and have an existing working knowledge of the attack location. Indeed, the
majority of cases show school shooters attacking their own, current school (e.g., Lanza, Klebold,
Harris). Kinkel attacked a school he had recently been expelled from; therefore, he still had a good
geographical knowledge of the location.
The difficulty for crime prevention is that there may be no overt irregular behaviors leading up to
the school shooting. The reconnaissance scene may be subtle and otherwise undetectable to
bystanders. Typically around the time of this scene, the individual will begin preparing and running
reconnaissance regarding the use of vehicles, if they intend to drive on the day of the attack. Several
school shooters used a car (e.g., Johnson, Golden, Woodham); and it is likely they considered this
ahead of the attack. However, from a crime prevention perspective, it would be very difficult to
intervene at this scene, owing to the subtlety of behaviors.

Activity
The activity scene, similar to Cornish’s (1994) doing scene, relates to the behaviors performed on
the day of the attack. There are several sub-streams within this scene. Several school shooters began
their attack at home (n = 4), before traveling to school. In a number of cases, it is the parents that are
harmed or killed, before school shootings occur (e.g., Kinkel, Lanza, Woodham, Whitman).
While exact timings are not included in a crime script analysis, there is a very close temporal
proximity between the harming of parents and subsequent school shootings. This does not leave law
enforcement with a large window of opportunity to react to the initial crime commissioned at home,
and the subsequent attack at the school. However, knowing this link could raise alert levels quicker
for future attacks.
Finally, there does appear to be a difference in the focus of attacks. Many school shooters (n = 15)
appear to fire randomly and in a disorganized way (i.e., not focusing on particular groups or
individuals).1 However, there are cases (e.g., Woodham) in which a particular individual is the
primary focus and target. In the case of Woodham, it was his former girlfriend and could show
a motivation for his attack.

Withdrawal
The withdrawal scene, or exit scene in Cornish’s (1994) universal script, is the final scene of the
crime script. The crime script ends with one of two main outcomes: school shooter committing
suicide or school shooter being arrested by police (in one case – Whitman- the school shooter was
killed by police). The data show that the type of withdrawal from the school shooting is largely a part
of an individual’s plan. Evidently, in several cases, the school shooter planned to end the attack
through suicide (n = 7). It is also apparent that in some cases, arrest occurs before suicide or owing
to an obstruction of suicide. For instance, Kinkel appears to have wanted to die as a part of the
school shooting; but, was overwhelmed by students.
There were not any cases, in the current sample, in which the shooter escaped and continued
their attacks beyond the school shooting. While crime script analysis does allow for continued scripts
beyond withdrawal or exit scenes (Cornish 1994), this extension is not shown in the current data. It
is not always entirely clear, from the data, when the suicide occurs, in terms of antecedent events.
Clearly the suicide marks the end of the school shooting; however, the choice point at which a school
shooter decides to end their life is not a systematic or clear moment.

1
This may be a caveat of the data collection approach. Those cases in which students shoot multiple students are more likely to be
widely documented and detailed – allowing for inclusion in the analyses. Instances of single-victim shootings may receive less
mainstream attention and therefore are simply omitted from the current analyses owing to a lack of information.
DEVIANT BEHAVIOR 1093

Discussion
The aim of the current study was to apply Crime Script Analysis to school shooting events, to provide an
account of the dynamic behaviors and antecedents that school shooters progress through leading up to
their attacks. A sample of 16 school shootings was investigated in-depth to provide detailed accounts of
the precursors of each school shooting. The depth of data acquisition allowed for more informed scripts
to be developed, and a stronger framework to be built for future research to be added to. A strength of the
crime script methodology is that diverse cases can be collected and analyzed parsimoniously, to provide
a generalized, or universal script. This also means that a variety of further cases can be added to the
current dataset, to build a general overview of the school shooting crime script.
The current research did not make a priori hypotheses about the outcomes; however, it was not
atheoretical. Based on previous research, a number of factors were expected to occur. For instance, in the
majority of cases in the current study included behaviors in the influences scene that have previously
been shown in the literature. For instance, several school shooters showed signs of anger management
issues, which supports previous findings in the literature (Langman 2013; Meloy et al. 2001). Also, many
school shooters had previous criminal charges for unrelated crimes. There are, therefore, precursors of
school shootings; however, in terms of a crime prevention strategy approach, it is not simply a case of
labeling every child who shows anger and criminal tendencies as a potential future school shooter.
Similarly, in this scene, there is clearly an influence of guns in the home environment, with many cases
reporting parents buying guns for the child. However, it would be simplistic to suggest this is a clear
causal relationship, given that parents also encouraged gun safety.
Crime Script Analysis is not simply a descriptive tool; but, if used effectively, should also be a crime
prevention method (Leclerc and Wortley 2014). Understanding the sequential process of a crime commis-
sion allows investigators to pinpoint crime prevention points. One of the repeated preventative measures
that is commonly suggested is gun control legislation. This would certainly lower the prevalence of guns
and would lower the opportunity for individuals to attain guns. However, in several cases, school shooters
stole the guns; therefore, gun control would need to ban guns on a national level and impede all access to
guns. While guns may provide an easily accessible weapon, removing guns might not stop school attacks.
Several individuals also used explosive devices in their attacks, which means if guns were made more
difficult to attain, it is possible that school attacks would take the form of a different method.
The reconnaissance scene poses the most difficulty for crime prevention strategies. Most school
shooters have intricate knowledge of the location of their attack, which is gained as part of their daily
activity in attending the school. Attempts to ascertain potential school shooters at this scene is
extremely difficult, and this is problematic given that is the final scene before the attack occurs. Any
attempt to intervene at this stage would most likely need to be based on observing changes in
students’ behavior or detecting behavioral markers. This information was not available for the
current analyses, and it is difficult to suggest how it might be validly obtained. Interviews with
school friends and teachers might indicate behavioral patterns; however, this would need to be
systematically conducted and validation would be difficult.
The current sample showed some differences in the activity scene. Several school shooters began their
attack at home, injuring or killing members of their family. There was typically very little time between
the attacks at home and the subsequent school shooting. It would not always be possible for police to
respond to or even become aware of the attacks at home, prior to the school shooting. However, the link
between killing of family members and school shootings should be investigated further to allow police
investigators quicker insight into possible follow-up attacks if they are called to a homicide case.
It remains unclear why some cases in the sample ended in suicide, while others resulted in arrest
during the withdrawal scene. Some of the school shooters who were apprehended claimed to want to
die as part of their school shooting crime. A strength of the crime script approach is that the
withdrawal scene can be elaborated and developed with further investigation. It may be that there are
behavioral predictors of whether a school shooter is likely to commit suicide or be arrested. At
present, the data are not clear enough on this outcome, especially given the fact that intervention
1094 D. A. KEATLEY ET AL.

(from police, security, teachers, or students) means that cases which would have otherwise been
“suicide” are instead coded as “arrested”.
While the current research takes a more nomothetic approach to grouping data, this is not to
undermine the importance of idiographic approaches. Langman (2009) highlights that it is impor-
tant to understand the differences between school shooters, as well as similarities between them.
Variations in demographics and childhood backgrounds in school shooters highlight the importance
of understanding individual differences (Langman 2009; Verlinden, Hersen, and Thomas 2000).
Future research could seek to integrate crime script tracks to account for different individuals with
distinct typologies; however, the literature in both crime script research and typology verification
would both need to be developed to make integrated analyses more effective.
The researchers made every effort to ensure all data in the current study were verified and valid.
This means that the dataset was limited to only those cases where enough information was available,
and facts were supported by cross-referencing. A strength, therefore, of the current dataset is the
depth and quality of the data included. This was important to build a strong initial framework
foundation for future cases to be applied to. Unlike other temporal methods, such as behavior
sequence analysis, crime script analysis can be conducted with incomplete, limited, or idiosyncratic
datasets (Keatley 2018). The current findings should be expanded to include further cases, especially
those that also have detailed information. When a large and strong enough dataset is developed,
subthemes and sub-scenes may be developed to account for different types of school shooting (e.g.,
those involving only guns; those motivated by revenge; those involving other weapons). A further
area of research would be to investigate the differences between high school shootings and college/
university shootings (see Newman and Fox 2009); this would allow for sub-scripts to be formed and
more specific understanding of different shootings gained.

Conclusion
The current paper uses crime script analysis to understand school shooters’ preparation and attack
behaviors. Five scenes emerged, which provided a concise overview of the temporal chain of
precursors leading-up to school shootings. The five sequential scenes were: 1) influences; 2) opera-
tional; 3) reconnaissance; 4) activity; and 5) withdrawal. School shootings are complex events with
multiple precursors; crime script analysis allows for a complete and systematic overview of these
precursors to be investigated. The outcomes of the current study can be used to understand other
school shootings, as well as provide the foundation for future research to extend.

Notes on contributors
David A. Keatley is a Senior Lecturer in Criminology at Murdoch University, Australia, and the Director of
Researchers in Behaviour Sequence Analysis (ReBSA). Dr Keatley collaborates on a number of international research
projects into violent crime resulting in multiple publications and case reports, and has written a book on investigating
crime. Dr Keatley is also a researcher and consultant in criminal investigations, cold cases, and behavioural analysis,
and regularly provides assistance to Law Enforcement worldwide.

Sian Mcgurk is an undergraduate student at the University of Lincoln. She is currently a member of Researchers in
Behaviour Sequence Analysis (ReBSA) and has assisted with multiple papers. Sian will be continuing her work on
applying Crime Script Analysis to Forensic Psychology, and will be carrying on with her education, studying for
a Masters beginning in September 2019.

Clare S. Allely is a Reader in Forensic Psychology at the University of Salford in Manchester, England and is an
affiliate member of the Gillberg Neuropsychiatry Centre at Gothenburg University, Sweden. Clare holds a PhD in
Psychology from the University of Manchester and has previously graduated with an MA (hons.) in Psychology from
the University of Glasgow, an MRes in Psychological Research Methods from the University of Strathclyde and an MSc
in Forensic Psychology from Glasgow Caledonian University. Clare is also an Honorary Research Fellow in the College
of Medical, Veterinary and Life Sciences affiliated to the Institute of Health and Wellbeing at the University of
Glasgow.
DEVIANT BEHAVIOR 1095

References
Abouk, Rahi and Scott Adams. 2013. “School Shootings and Private School Enrollment.” Economics Letters 118
(2):297–99. doi: 10.1016/j.econlet.2012.11.009.
Baird, Abigail A., Emma V. Roellke, and Debra M. Zeifman. 2017. “Alone and Adrift: The Association between Mass
School Shootings, School Size, and Student Support.” Social Science Journal 54(3):261–70. doi: 10.1016/j.
soscij.2017.01.009.
Beauregard, Eric, Jean Proulx, Kim Rossmo, Benoît Leclerc, and Jean-François Allaire. 2007. “Script Analysis of the
Hunting Process of Serial Sex Offenders.” Criminal Justice and Behavior 34(8):1069–84. doi: 10.1177/
0093854807300851.
Bondü, Rebecca and Herbert Scheithauer. 2015. “Narcissistic Symptoms in German School Shooters.” International
Journal of Offender Therapy and Comparative Criminology 59(14):1520–35. doi: 10.1177/0306624X14544155.
Borrion, Herve. 2013. “Quality Assurance in Crime Scripting.” Crime Science 2(1):1–12. doi: 10.1186/2193-7680-2-6.
Borum, Randy, Dewey Cornell, William Modzeleski, and Shane. Jimerson. 2010. “What Can Be Done about School
Shootings? A Review of the Evidence.” Educational Researcher 39(1):27–37. doi: 10.3102/0013189X09357620.
Chiu. 2015. “Patterns in Unsolved Sexual Offences against Women by Strangers.” PhD Thesis, Submitted to: Griffith
University.
Cornish, Derek. 1994. “The Procedural Analysis of Offending and Its Relevance for Situational Prevention.” Pp.
151–96 in Crime Prevention Studies, Vol. 3. , Monsey, NY: Criminal Justice Press.
Ekblom, Paul and Martin Gill. 2016. “Rewriting the Script: Cross-Disciplinary Exploration and Conceptual
Consolidation of the Procedural Analysis of Crime.” European Journal on Criminal Policy and Research 22
(2):319–39. doi: 10.1007/s10610-015-9291-9.
Gerard, Jeane, Kate Whitfield, Louise Porter, and Browne Kevin. 2016. “Offender and Offence Characteristics of School
Shooting Incidents.” Journal of Investigative Psychology and Offender Profiling 13(1):22–38. doi: 10.1002/jip.1439.
Henry, S. 2009. “School Violence beyond Columbine.” A Complex Problem in Need of an Interdisciplinary Analysis’.
American Behavioral Scientist 52 (9):1246–65.
Homel, R, S. Macintyre, and R. Wortley. 2014. “How House Burglars Decide on Targets.” Pp. 26–47 in Cognition and
Crime: Offender Decision Making and Script Analyses, edited by B. Leclerc and R. Wortley. Oxford: Routledge,
Hutchings, Alice and Thomas J. Holt. 2015. “A Crime Script Analysis of the Online Stolen Data Market.” British
Journal of Criminology 55(3):596–614. doi: 10.1093/bjc/azu106.
Keatley, David. A. 2018. Pathways in Crime: An Introduction to Behaviour Sequence Analysis. London: Palgrave
Macmillan.
Keatley, David. A. and David. D. Clarke. 2018. “Indicator Waves: A New Temporal Method for Measuring Multiple
Behaviours as Indicators of Future Events.” Paper Presented at the 11th International Conference on Methods and
Techniques in Behavioral Research: Measuring Behavior Conference. Pp 92–95, Manchester, UK.
Keatley, David. A., Hayley Golightly, Rebecaa Shephard, Enzo Yaksic, and Sasha Reid. 2018. “Using Behaviour
Sequence Analysis to Map Serial Killers’ Life Histories.” Journal of Interpersonal Violence, Advance online publica-
tion. doi: 10.1177/0886260518759655.
Langman, Peter. 2009. “Rampage School Shooters: A Typology.” Aggression and Violent Behavior 14(1):79–86. doi:
10.1016/j.avb.2008.10.003.
Langman, Peter. 2013. “Thirty-Five Rampage School Shooters: Trends, Patterns, and Typology.” Pp. 131–58 in School
Shootings: International Research, Case Studies, and Concepts for Prevention, edited by N. Böckler, T. Seeger,
P. Sitzer, and W. Heitmeyer. London: Springer,
Leary, Mark R., Robin M. Kowalski, Laura Smith, and Stephen Phillips. 2003. “Teasing, Rejection, and Violence: Case
Studies of the School Shootings.” Aggressive Behavior 29(3):202–14. doi: 10.1002/ab.10061.
Leclerc, B., R. Wortley, and S. Smallbone. 2011. “Getting into the Script of Adult Child Sex Offenders and Mapping out
Situational Prevention Measures.” Journal of Research in Crime and Delinquency 48 (2):209–237. doi:10.1177/
0022427810391540.
Leclerc, Benoit, Stephen Smallbone, and Richard. Wortley. 2013. “Interpersonal Scripts and Victim Reaction in Child
Sexual Abuse: A Quantitative Analysis of the Offender-Victim Interchange.” Pp. 101–119 in Cognition and Crime:
Offender Decision Making and Script Analyses, edited by B. Leclerc and R. Wortley. Oxford: Routledge.
Leclerc, Benoit and Richard. Wortley. 2014. Cognition and Crime: Offender Decision Making and Script Analyses.
Oxford: Routledge.
Meloy, J. Reid, Anthony G. Hempel, Kris Mohandie, Andrew A. Shiva, and B. Thomas Gray. 2001. “Offender and
Offense Characteristics of a Nonrandom Sample of Adolescent Mass Murderers.” Journal of the American Academy
of Child and Adolescent Psychiatry 40(6):719–28. doi: 10.1097/00004583-200106000-00018.
Meyer, Sunniva. 2013. “Impeding Lone-Wolf Attacks: Lessons Derived from the 2011 Norway Attacks.” Crime Science
2(7):1:13. doi: 10.1186/2193-7680-2-7.
Newman, K. 2004. Rampage: The Social Roots of School Shootings. New York: Basic Books.
Newman, K., and C. Fox. 2009. “Repeat Tragedy: Rampage shootings in American high school and college settings,
2002-2008.” American Behavioral Scientist 52 (9):1286–1308. doi:10.1177/0002764209332546.
1096 D. A. KEATLEY ET AL.

Nisbett, Richard, E and Lee Ross. 1980. Human Inference: Strategies and Shortcomings of Social Judgment. New Jersey:
Patience Hall.
O’Toole, Mary, E. 2000. The School Shooter: A Threat Assessment Perspective. Quantico, VA: Critical Incident
Response Group. National Centre for the Analysis of Violent Crime, FBI Academy.
Rocque, Michael. 2012. “Exploring School Rampage Shootings: Research, Theory, and Policy.” Social Science Journal
49(3):304–13. doi: 10.1016/j.soscij.2011.11.001.
Schank, R. C. and R. P. Abelson. 2013. “Scripts, Plans, Goals, and Understanding.” Pp. 190–223 in Readings in
Cognitive Science: A Perspective from Psychology and Artificial Intelligence, edited by A. Collins and E. E. Smith.
Massachusetts: Morgan Kaufmann,
Verlinden, Stephanie, Michel Hersen, and Jay Thomas. 2000. “Risk Factors in School Shootings.” Clinical Psychology
Review 20 (1):3–56.
Vossekuil, B., R. Fein, M. Reddy, R. Borum, and W Modzeleski. 2002. The Final Report and Findings of the Safe School
Initiative: Implications for the Prevention of School Attacks in the United States. Washington, DC: U.S. Department
of Education, Office of Elementary and Secondary Education, Safe and Drug-Free Schools Program and U.S. Secret
Service, National Threat Assessment Center.
Wike, Traci L. and Mark W. Fraser. 2009. “School Shootings: Making Sense of the Senseless.” Aggression and Violent
Behavior 14(3):162–69. doi: 10.1016/j.avb.2009.01.005.

You might also like