Content uploaded by Marieke Liem
Author content
All content in this area was uploaded by Marieke Liem on May 23, 2014
Content may be subject to copyright.
http://hsx.sagepub.com/
Homicide Studies
http://hsx.sagepub.com/content/18/1/44
The online version of this article can be found at:
DOI: 10.1177/1088767913511460
2014 18: 44 originally published online 20 November 2013Homicide Studies Marieke Liem and Ashley Reichelmann
Patterns of Multiple Family Homicide
Published by:
http://www.sagepublications.com
On behalf of:
Homicide Research Working Group
can be found at:Homicide StudiesAdditional services and information for
http://hsx.sagepub.com/cgi/alertsEmail Alerts:
http://hsx.sagepub.com/subscriptionsSubscriptions:
http://www.sagepub.com/journalsReprints.navReprints:
http://www.sagepub.com/journalsPermissions.navPermissions:
http://hsx.sagepub.com/content/18/1/44.refs.htmlCitations:
at Universiteit Leiden \ LUMC on May 23, 2014hsx.sagepub.comDownloaded from at Universiteit Leiden \ LUMC on May 23, 2014hsx.sagepub.comDownloaded from
Homicide Studies
2014, Vol. 18(1) 44 –58
© 2013 SAGE Publications
Reprints and permissions:
sagepub.com/journalsPermissions.nav
DOI: 10.1177/1088767913511460
hsx.sagepub.com
Article
Patterns of Multiple Family
Homicide
Marieke Liem1 and Ashley Reichelmann2
Abstract
Previous research has treated multiple family homicide, or familicide, as a uniform
event. We sought to explore whether subtypes of familicide could be discerned,
making use of a decade of Supplementary Homicide Reports (SHR) coupled with
newspaper articles. The resulting 238 cases were analyzed through a two-step
cluster analysis, showing that the familicides can be subgrouped into four categories
based on the perpetrator’s age, relationship between perpetrator and victims, and
perpetrator’s suicide. The empirically grouped categories were labeled Despondent
Husbands, Spousal Revenge, Extended Parricide, and Diffuse Conflict. Familicide is thus
a heterogeneous phenomenon and must be viewed in unique terms to appropriately
determine prevention strategies.
Keywords
mass murder, subtypes, intimate partner, victim/offender relationship, child, parent/
parricide, siblicide
Introduction
Multiple family homicide is also known as familicide or the killing of multiple family
members. Despite the media coverage and public interest these events generate, famil-
icide is a relatively rare event. Even though familicide has an enduring presence
throughout history and across cultures, there is surprisingly limited research examin-
ing this most extreme form of family violence. It is important to note that some studies
lump multiple family homicides together with other forms of mass murder (see, for
example, Bowers, Holmes, & Rhom, 2010): the killing of two or more victims in one
event (Delisi & Scherer, 2006). Multiple family homicides, however, are believed to
originate and evolve from largely independent etiologies. Due to the intimate
1Harvard University, Cambridge, MA, USA
2Northeastern University, Boston, MA, USA
Corresponding Author:
Marieke Liem, Harvard Kennedy School, 79 John F. Kennedy Street, Cambridge, MA 02138, USA.
Email: Marieke_Liem@hks.harvard.edu
511460HSX18110.1177/1088767913511460Homicide StudiesLiem and Reichelmann
research-article2013
at Universiteit Leiden \ LUMC on May 23, 2014hsx.sagepub.comDownloaded from
Liem and Reichelmann 45
relationship between the perpetrator and victims, and given the motives underlying
these crimes, familicides stand apart from other forms of mass murder.
One of the main limitations in the work on multiple family homicides so far, how-
ever, is the lack of recognition of heterogeneity among familicides. The literature on
multiple family homicide typically focuses on just one prominent constellation: the
killing of spouse and children (Goldney, 1977; Leveillée, Lefebvre, & Marleau, 2009;
Liem & Koenraadt, 2008; Schlesinger, 2000; Websdale, 2010; Wilson, Daly, &
Daniele, 1995). Research has yet to determine if other constellations of familicide can
be discerned, and how these constellations can be characterized. The present study
uses the UCR Supplementary Homicide Reports (SHR), a national sample of homi-
cide data over the last decade, supplemented with newspaper articles, to examine the
characteristics associated with multiple family homicide in the United States. In con-
trast to earlier studies, the present research effort is designed to identify specific sub-
groups within multiple family homicides.
Prior Research
Spouse and Child(ren)
The most common type of familicide is one in which the spouse and one or more chil-
dren are victimized. The perpetrators are mostly men, typically in their 30s or 40s, who
commit the offense with a firearm. In the United States, familicide involving an inti-
mate partner and child(ren) occurs approximately 23 times per year (Liem, Levin,
Holland, & Fox, 2013). In their study of familicide in Canada, Wilson et al. (1995)
found 61 familicides in one 15-year period. In roughly the same period, 48 cases were
registered in England and Wales (Wilson et al., 1995). Because of its low incidence,
many studies have relied on describing a small number of cases, are based on single-
case descriptions (Scheinin, Rogers, & Sathyavagiswaran, 2011; Thaller, 2012), or
relied on clinical data (Leveillée et al., 2009; Liem & Koenraadt, 2008; Malmquist,
1980). In spite of these limitations, several general characteristics can be outlined.
A common dichotomy among these familicides is the one proposed by Frazier
(1975), who distinguished between the “murder-by-proxy” type and the “suicide-
by-proxy” type. The first applies to cases in which victims are chosen because they
are identified with a primary target—the spouse. These familicides are typically
preceded by a woman’s threat of withdrawal or estrangement (Wilson et al., 1995).
From this perspective, a man might kill his children in addition to killing his spouse
because he regards them as an extension of her. The children are thus perceived as
“her” children and, accordingly, equally responsible for her betrayal (Dietz, 1986;
Fox & Levin, 2011; Liem & Koenraadt, 2008). Based on a study of more than 200
familicidal men, Websdale (2010) termed the emotional styles of these perpetrators
as “livid coercive,” because of their uncontrolled anger and tyrannical behavior
toward their partners.
Suicide-by-proxy familicides, conversely, are thought to be triggered by the perpetra-
tor’s loss of a job, continuous unemployment, and/or his subsequent inability to support
at Universiteit Leiden \ LUMC on May 23, 2014hsx.sagepub.comDownloaded from
46 Homicide Studies 18(1)
his family (Fox & Levin, 2011; Palermo, 1997). When trapped by the evaporation of
economic dreams, the familicidal man does not see any other option but to “protect” his
family from the fate that would befall them without his support. In such cases, the perpe-
trator commands a relationship in which he perceives that only he can satisfy the needs of
his victims. Faced with such overwhelming threat to their role as provider, controller, and
central figure in the lives of their families, these men become desperate, homicidal, and
suicidal (Marzuk, Tardiff, & Hirsch, 1992). In some cases, familicides are altruistically
motivated, as the perpetrator aims to protect his loved ones from a catastrophic future.
Typically, the latter is followed by the suicide of the perpetrator (Cooper & Eaves, 1996;
Liem et al., 2013; Wilson et al., 1995). Websdale (2010) classified these perpetrators as
“civil reputable.” They do not have a history of violence, and the killings are carefully
premeditated over a considerable length of time. It has been suggested that as the econ-
omy softens and the unemployment rate rises, there may be more opportunities for cata-
strophic losses to precipitate a familicide (Scheinin et al., 2011). When job loss or
indebtedness is involved, the motivation may become a lethal solution for the offender to
“protect” his family (Fox & Levin, 2011). Results from recent analyses based on U.S.
data showed, however, that the prevalence of familicide motivated by financial problems
was unrelated to periods of financial downfall (Liem et al., 2013). What both types of
familicide consisting of spouse and child(ren) have in common is that the perpetrators are
motivated by a sense of loss of control over their spouse as well as over their family life
(Ewing, 1997; Liem & Koenraadt, 2008; Wilson et al., 1995).
Parent(s) and Sibling(s)
A less frequent type of familicide consists of a combination of parricide (the killing of
one or more parents) and siblicide (the killing of one or more siblings). To the extent
that this type of familicide has been studied, it is mostly limited to case studies—the
most famous perhaps being Pierre Rivière, as described in a classic treatise by Michel
Foucault (1975). On June 3, 1835, Rivière brutally murdered his mother, sister, and
brother with a pruning hook (Fuentes, 2013). He intended to save his father from his
mother, who tormented and oppressed him.
In more recent familicides of this constellation, the perpetrator is thought to primar-
ily aim his aggression at one or both parents—the siblings are considered as allies of a
dominant, hostile father. They are, in the eyes of the perpetrator, equally responsible,
albeit actively or passively, for the parent’s dominant and repressive conduct. These
perpetrators are typically motivated by a desire to free themselves from the parent’s
tyranny, and to regain an identity the parent(s) inhibit (Koenraadt, 1996). Prior inci-
dences of domestic violence are particularly uncommon among this group of offend-
ers. This suggests that these incidences may not move along a continuum of
child-initiated family violence (Walsh & Krienert, 2009), but rather constitute a sepa-
rate dynamic. Studies of parricide in the United States have found that in the majority
of cases, fathers and mothers are killed by adult sons (Bourget, Gagné, & Labelle,
2007; Heide, 2013; Heide & Frei, 2010; Marleau, Auclair, & Millaud, 2006). Similarly,
studies of siblicide indicate that male offenders are more likely to kill a sibling in
at Universiteit Leiden \ LUMC on May 23, 2014hsx.sagepub.comDownloaded from
Liem and Reichelmann 47
addition to another family member compared with female offenders (Peck & Heide,
2012). A major limitation in past studies is that they treat parricide and siblicide as
separate phenomena and hence do not distinguish this type of multiple family homi-
cide as a unique category, but rather refer to it as “the killing of parents along with
other family members” (see, for example, Heide & Frei, 2010).
In sum, previous research on multiple family homicide has merely focused on one
subtype of familicide: the killing of spouse and children. We do not know if, and to
what extent, we can discern other types of familicide. This research aims to fill this
longstanding vacuum by determining whether various constellations of familicide can
be distinguished, and how these constellations can be characterized. To meet this aim,
we use cluster analysis techniques, which are particularly useful, because the classifi-
cation scheme can take into account multiple characteristics of the killing. Thus, this
analysis allows for multiple family homicides to be categorized along numerous
dimensions (Kubrin, 2003). Cluster analysis has been proven a useful tool in previous
scholarly work to seek out subtypes of violent offending (Coid, Kahtan, Gault, &
Jarman, 2000; Saunders, 1992; Stefurak & Calhoun, 2007), including homicide
(Bijleveld & Smit, 2006; Blackburn, 1971; Kubrin, 2003; Yarvis, 1994).
Methodology
Data Sources
To meet our aims, we used two data sources to extend a previously generated data set
(Liem et al., 2013): the FBI’s Supplementary Homicide Reports (SHR) data (spanning
the years 2000-2007) and newspaper reports on familicide (spanning the years
2000-2009).
The FBI’s SHR includes nationwide incident-level information on nearly all mur-
ders and nonnegligent manslaughters in the United States (Fox & Swatt, 2009). It
should be noted that these are arrest data, and that these offenders have not (yet) been
convicted (Peck & Heide, 2012). These case-based reports include specific identifying
details, such as the month and year of the case, the reporting agency, the primary vic-
tim–perpetrator relationship per case, the number of victims per case, and sociodemo-
graphic characteristics of the primary victim and the primary perpetrator (Fox, 2004).
The SHR data, however, do not provide the relationships between the perpetrator and
each subsequent victim involved in the case. Therefore, the SHR data only afforded
the ability to narrow down to cases that included (a) a family relationship between the
primary victim and the primary perpetrator and (b) involved two or more victims—
both characteristics that are necessary to identify cases as familicides.
A potential limitation of the SHR data is the accuracy and completeness of informa-
tion provided within their data set. To overcome these limitations, we collected addi-
tional data from newspaper articles found through the search engine LexisNexis.
Newspaper accounts have been proven useful to assess in-depth information on sev-
eral types of homicide (Aderibigbe, 1997; Danson & Soothill, 1996; Liem & Koenraadt,
2007; Malphurs & Cohen, 2002). Because of the violent nature of the act and the fact
at Universiteit Leiden \ LUMC on May 23, 2014hsx.sagepub.comDownloaded from
48 Homicide Studies 18(1)
that multiple victims are involved, familicides typically receive a large amount of
media coverage (Liem et al., 2013).
This approach provided a closer look at the relationship between the perpetrator(s)
and victims as well as the motives underlying the acts. By acquiring more details, we
were able to investigate the differing types of familicides. Cases were retrieved by
searching the reporting agency and “murder” after limiting the search to the geograph-
ical state and month/year of the occurrence. If the results numbered more than 25, the
search was further filtered by a keyword from the SHR data, starting with the listed
relationship between the primary perpetrator and the primary victim. If necessary, it
was further filtered by the age of the perpetrator. If an article could not be found within
the LexisNexis database, Google was used to locate a local newspaper covering the
case. Cases after 2007 were not listed in the SHR data. Following methodologies used
in prior research on familicide in the United States (Liem et al., 2013), these cases
were retrieved via LexisNexis. We used a keyword search of “family murder” after
restricting the dates from January 2000 to December 2009 and the location to the
United States.
Inclusion Criteria
The cases only qualified as familicides if there were multiple victims and if the victims
fit into at least two different relational categories, or victim types. For instance, a
mother killing her two children does not qualify as familicide, but rather as a multiple
child homicide, whereas a father killing his spouse and his children or a son killing his
parents and his siblings fits the appropriate criteria. The victims must thus constitute
different categories of familial relative status.
Cluster Sample, Variables, and Measures
By making use of additional newspaper searches, 326 out of the total 961 SHR mul-
tiple family homicides could be retrieved. Eighty-eight cases were excluded because
they did not match our inclusion criteria (27%). These excluded cases involved homi-
cides that “only” involved one family member victim (N = 5), or victims belonging
to one family category, such as “only” children (N = 5), “only” (grand) parents or
parents-in-law (N = 64), or were in other constellations part of just one relational
category, or victim type, such as an aunt and uncle, or two cousins (N = 14). Cases
that were included were subsequently coded to reflect event characteristics (motive,
relationship between each victim and perpetrator, location), perpetrator characteris-
tics (sociodemographics, criminal history), and victim characteristics (sociodemo-
graphics, cause of death). For each case, information about the perpetrator(s), victims,
and event was recorded and captured in more than 50 variables related to each
familicide.
To determine whether familicides with different characteristics can be grouped into
archetypes, we used cluster analyses. Cluster analysis is the generic name for a variety
of procedures that can be used to create a classification. These procedures empirically
at Universiteit Leiden \ LUMC on May 23, 2014hsx.sagepub.comDownloaded from
Liem and Reichelmann 49
form clusters, or groups of highly similar entities (Kubrin, 2003). From more than 50
variables that characterize each killing, the list was narrowed to three cluster variables:
Perpetrator age, relationship between victims and perpetrator (victims were spouse
and children; victims were parents and siblings; victims were a constellation of other
family members), and perpetrator’s suicide (attempt) (present or absent). The ratio-
nale for choosing these characteristics was conceptual and empirical (see also Kubrin,
2003). Conceptually, the variables include fundamental aspects of the homicide that
were known for the majority of cases: Characteristics of the perpetrator(s) and the
victims, the relationship between them, and details about the homicide event.
Empirically, the so-called swamping variables, such as gender and race, were excluded
as well as variables that were statistically rare, such as history of mental health treat-
ment or drug-related problems.
Given that the clustering variables involved continuous and categorical variables,
we used a two-step cluster analysis. This method is designed to reveal natural group-
ings, or clusters. The two-step cluster constitutes a combination of hierarchical and
non-hierarchical (K-means) techniques, instead of only employing one or the other.
Although the subtypes are derived using cluster analysis, the interpretation of the clus-
ter results was guided by additional analyses of background information provided in
the newspaper articles. The cluster sample was based on information in the SHR and
in newspaper articles, from a total of 238 multiple family homicides that occurred in
the United States in one decade, between January 2000 and December 2009.
Results
The majority of the familicides were committed by white men in their 30s or 40s, who
killed their family members with a firearm. In two thirds of the cases, the perpetrator
shared the household with all victims. Suicide was reported in 57% of all cases—of
which 6% included serious attempts—and killings showed signs of premeditation in
approximately 70%. In over half of the cases, victims of the familicide were the
(estranged) spouse and children (62%); this included children of spouse and perpetra-
tor (51%), children of spouse alone (8%), or children of the perpetrator alone (2%).
Other predominant constellations of victims included siblings and parents (13%),
spouse and other family members (7%), and (estranged) spouse and spouse’s parents
(5%). Less prevalent victim constellations consisted of parents and other family mem-
bers (6%), siblings and other family members (2%), and secondary relations, such as
uncles, cousins, nephews, and grandparents (4%). With the exception of three cases,
all familicides were perpetrated by one perpetrator. Cases ranged from involving two
through ten victims, and on average, cases included three victims (M = 3.0; SD =
1.16)—excluding the suicide of the perpetrator. Domestic problems were not uncom-
mon, ranging from intimate partner problems (80%)—at times resulting in a restrain-
ing order (29%)—to conflicts related to child custody (23%).
The two-step cluster analysis revealed four distinct familicide types: Despondent
Husbands, Spousal Revenge, Extended Parricide, and Diffuse Conflict. To determine
which characteristics differentiate the subtypes, descriptives of perpetrator, victim,
at Universiteit Leiden \ LUMC on May 23, 2014hsx.sagepub.comDownloaded from
50 Homicide Studies 18(1)
and event variables were produced for cases comprising each subtype, as well as for
the total familicide comparison (see Table 1). What follows is a summary narrative for
each cluster based on its profile features and background event characteristics.
The first subtype, or cluster 1 familicides, constituted the largest group with almost
one half of all cases (N = 109; 46%) and were characterized by the following: Cases
consist of the killing of spouse and children. Perpetrators were men with a mean age
of 42, who were predominantly white and commit suicide after the homicide event.
Perpetrators typically shared the household with all victims, and were the biological
father of the children. Previously reported intimate partner problems were frequent
and problems related to child custody were present in about 25% of cases. Insofar as
event-specific information was available, these homicides were premeditated in 92%
of cases.
Further analysis of newspaper articles showed the commonality of men who had
difficulty financially supporting their growing families. For instance, in January 2001,
in Angelina, TX, a 24-year-old Hispanic man killed his wife and his four young chil-
dren, all of whom were under the age of five. Due to the poor weather of the season,
he was unable to find work as a roofer. Ultimately, he could not pay his bills and had
to steal electricity from his neighbor. He killed his family members as they slept, and
Table 1. Characteristics of Total (N = 238), Despondent Husbands (n = 109), Spousal
Revenge (n = 41), Extended Parricide (n = 31), and Diffuse Conflict (n = 57) Familicides.
Variable Total (%)
Cluster 1:
Despondent
husbands (%)
Cluster 2:
Spousal
revenge (%)
Cluster 3:
Extended
parricide (%)
Cluster 4:
Diffuse
conflicts (%)
Perpetrator characteristics
Gender
Male 94 92 98 97 93
Female 6 8 2 3 7
Age M = 37.9 M = 41.8 M = 40.0 M = 27.1 M = 39.2
Race
White 72 78 68 71 67
Non-White 28 22 34 29 33
Event characteristics
Household
All shared 59 72 67 63 22
Partially shared 18 13 11 15 35
Non-shared 23 14 22 22 44
Premeditated 70 92 62 60 68
Suicide 57 100 0 19 37
Domestic problems
Intimate partner 80 83 76 50 85
Restraining order 29 25 25 20 46
Child custody 23 41 0 0 17
at Universiteit Leiden \ LUMC on May 23, 2014hsx.sagepub.comDownloaded from
Liem and Reichelmann 51
then took his own life, which had spiraled out of his control due to ongoing financial
problems coupled with recent marital problems. Another example took place in
September 2009 in Maryland, where a 38-year-old White male killed his family of
four. He was self-employed as a carpenter and cabinetmaker, and worked the night
shift as a janitor at the local school to generate enough income to support his family.
After the recession confounded the financial hardship, he killed his wife, his 14-year-
old son, and seven-year-old daughter as they slept. He also killed the family dog before
committing suicide with the same firearm. Taking their overall profile into account,
the cluster label of Despondent Husbands was generated to describe this group.
Cluster 2 familicides also consisted of cases in which spouse and children were
victimized—oftentimes, these children were the perpetrator’s stepchildren. What dis-
tinguished this cluster from the abovementioned cluster was that the perpetrator did
not commit suicide following the event. Perpetrators were predominantly male, with
an average age of 40 and were more likely than perpetrators in Cluster 1 to be of non-
White descent. A closer look at background characteristics of these cases showed that
intimate partner problems were indicated in ¾ of the cases and restraining orders in ¼
of them. The perpetrator shared the household with at least one of the victims in two
thirds of the cases.
Examples of such cases included a 28-year-old man who killed his long-term girl-
friend and her 5-year-old daughter in Dallas, TX, in July 2000. After his girlfriend
announced that she would be leaving him, he shot her in the back and killed her daugh-
ter with blunt force trauma. He was convicted and sentenced to life in prison. Another
example involved a former Navy veteran in April 2001 in Scottsdale, AR. The 40-year-
old White male and his wife had attended marital counseling sessions. After being
traumatized by his parent’s divorce as a child, he attempted to avoid the same fate, by
controlling every aspect of his family’s life. Eventually, his wife announced to family
members that she intended to divorce him. After a loud argument, he shot his wife and
cut the throats of their two sons, aged 10 and 13. He then unhitched a gas pipe, causing
the house to explode. Given that the primary object of aggression is the (estranged)
intimate partner, we applied the term Spousal Revenge to these familicides. The results
showed that 41 (17%) of all familicides could be grouped in this cluster.
The third cluster constituted the smallest cluster with 31 cases (13%), and mostly
consisted of cases in which parents and siblings were killed. Perpetrators in this group
were White men with an average age of 27, who did not typically commit suicide after
the event. More than half of these men were in their teens or early twenties. The famili-
cide was less likely to be premeditated compared with the other three clusters, and
suicide only followed in one fifth of the cases. In the majority of cases, the perpetrator
shared the household with at least one victim. Typically, in such cases, siblings were
perceived as extensions of the parents, or potential witnesses to the event who need to
be removed. In other cases, the perpetrator’s problems were focused on an issue sepa-
rate from the actual victims, such as intimate partner problems, unemployment, or
alcohol/drug-related problems. In such cases, the origin of aggression included outside
circumstances, for which parents and siblings were regarded as in some way (partially)
responsible.
at Universiteit Leiden \ LUMC on May 23, 2014hsx.sagepub.comDownloaded from
52 Homicide Studies 18(1)
Two examples of the third cluster shared the commonality of antisocial behavior in
young males. The first concerned a case in Stevens, WA, in December 2000. A 16-year-
old boy confessed to killing his parents and two siblings with two separate rifles and
then burying them in their snowy backyard. Although the neighbors described the son
as nice because he often played with their young children, he had engaged in criminal
behavior, causing his parents to be strict on him. In the year of the murders, his parents
home-schooled him because of his behavior, while his 18-year-old sister and 11-year-
old brother attended the local public schools. Although no motive was officially identi-
fied, neighbors described the relationship between the parents and the 16-year-old as
quite conflicted, due to the number and extent of arguments that occurred between
them. Another example of this familicide cluster included a 14-year-old boy who
killed his father after years of abuse in New Mexico in July 2004. After his father’s
multiple marriages, the death of his mother, and a newly acquired stepsister, the young
White male had trouble transitioning into the new family structure. After an alleged
abusive episode by his father, he ran and hid in the woods. He returned hours later,
killing his father and stepmother. He then killed his younger stepsister, because she
witnessed the murders. In court, the defense argued that he was a withdrawn child who
was frequently abused by his father. The label Extended Parricide was generated to
capture the salient aspects of this group’s profile.
Cluster 4 contains 57 cases (24%) and consisted of men with an average age of 39,
who killed multiple family members with a firearm in a variety of constellations.
These included the killing of a spouse and in-laws, the killing of parents and other
family members, or the killing of secondary relations, such as combinations of grand-
parents, cousins, aunts, uncles, nephews. Perpetrators committed suicide following the
event in more than one third of the cases. In cases in which the spouse and in-laws
were killed, the spouse constituted the primary target; here, the perpetrator perceived
the in-laws as equally guilty of betrayal. This was reflected in the relatively high prev-
alence of intimate partner problems, including prior restraining orders. In other cases
in this cluster, the physical and geographic closeness of family members caused them
to be present and thus become a victim of the familicide. Information from newspaper
articles revealed that such events were frequently induced by alcohol, or preceded by
conflicts over money or other arguments. Typically, the victims and perpetrator did not
share a household. The event was premeditated in about two thirds of the cases.
Examples of cases in this cluster include the following: In Gary, IN, in September
2003, a 20-year-old man, who had been estranged from his family and was living on
the streets, knocked on his family door in the middle of the night. Before taking his
own life, he killed his two siblings, his mother, and his nephew. Only his 82-year-old
grandmother survived the attack. The family had previously become estranged from
the son due to mental illness and drug use. Cases such as these, clustered by offender
and offense characteristics, all have in common the perpetrator’s diffuse nature of
aggression, rather than aggression being directed at the one deemed responsible.
Another case took place in the small town of Brunswick, GA, in September 2009,
when a 22-year-old White man brutally beat eight of his family members to death. The
victims included his father, aunt, uncle, four cousins, and his cousin’s boyfriend. He
at Universiteit Leiden \ LUMC on May 23, 2014hsx.sagepub.comDownloaded from
Liem and Reichelmann 53
then called the police announcing, “my whole family’s dead!” (Deutsche Presse-
Agentur, 2009). On police arrival, he was arrested on drug charges, later to be charged
with murder. Another case in this cluster involved a 12-year-old boy in Washington,
OH, in November 2004, who occasionally shared a household with his grandmother.
He killed her and his aunt with a shotgun, before developing an elaborate story for the
police. He was seen as a troubled youth and claimed that his grandmother always put
him down. He killed his aunt because she happened to be there and tried to stop him.
Because of the diffuse nature of the conflicts underlying these cases, as well as the
diverse choice of victims, we applied the term Diffuse Conflict to describe this
cluster.
Discussion
This study represents the only effort to date to systematically analyze various constel-
lations of multiple family homicide. The results indicated that familicide is a hetero-
geneous phenomenon: based on a national sample of more than 238 multiple family
homicides that occurred in the course of a decade, we were able to discern four spe-
cific familicide clusters.
The first two clusters of Despondent Fathers and Spousal Revenge have been pre-
viously described in the literature—In line with previous (international) studies
(Byard, Knight, James, & Gilbert, 1999; Ewing, 1997; Friedman, Hrouda, Holden,
Noffsinger, & Resnick, 2005; Liem et al., 2013; Scheinin et al., 2011; Somander &
Rammer, 1991), these husbands and fathers were White men who kill their (estranged)
spouses and children. The difference between these two clusters lied in the underlying
dynamics and in the family composition: While the Despondent Father killed his fam-
ily members out of pseudo-altruistic reasons, frequently instigated by financial trou-
bles and followed by suicide, perpetrators in the Spousal Revenge cluster acted out of
anger over abandonment. When his spouse threatens to leave and/or threatens to
exclude him from contact with the children, the anger, jealousy, and rage against the
spouse is extended to the children (Alder & Polk, 2001; Leveillée et al., 2009; Liem &
Koenraadt, 2008; Wilczynski, 1997). In this category, the killing of stepchildren was
not exceptional—a finding reported in previous studies (Wilson et al., 1995). When
marital problems occur, children can be viewed as a source of hostility and anger,
particularly when coupled with doubts of paternity. The vast majority of the perpetra-
tors in these two clusters were male. Traditionally, the loss of a job and subsequent
inability to support the family are arguably more detrimental to the masculine identity
as a provider (Palermo, 1997), than the feminine identity—hence leading to a higher
prevalence of men among Despondent Parents than women. The high prevalence of
men among the Spousal Revenge cluster can be attributed to psycho-evolutionary
goals: Men have a stronger need for control over their spouses’ reproductive capacity
(Daly & Wilson, 1988; Wilson et al., 1995) compared with women. In an attempt to
regain control following a threat of withdrawal or estrangement, men may respond
with lethal violence, in which the children are perceived of as extensions of their inti-
mate partner.
at Universiteit Leiden \ LUMC on May 23, 2014hsx.sagepub.comDownloaded from
54 Homicide Studies 18(1)
We are the first to describe the existence of two additional familicide clusters that
were thus far neglected in the literature: Extended Parricide and Diffuse Conflict. The
lack of attention for these familicide clusters appears unjustified, particularly given the
sizable proportion of these categories. Extended Parricides are committed by rela-
tively young, troubled men, whose primary anger is directed toward one or both par-
ents. The fact that suicide rarely occurred in this cluster of familicides is in line with
earlier work on parricides alone (Marleau et al., 2006; Millaud, Auclair, & Meunier,
1996; Mouzos & Rushforth, 2003). In the case of adolescents tormented by abuse,
parricide resulted from a desire either to commit suicide or remove the tormentor.
Meloy (1992, p. 58) summarized this dynamic as “either he or I must die, something
has to give.” He argued that a polarization takes place between the “all bad” parent and
the “all good” self; once the adolescent has killed his tormentor and the associated
culprits (his siblings), the need for taking his own life is removed. Insofar as we could
retrieve information from newspaper articles on these cases, siblings thus appeared to
be victimized because of their physical presence as witnesses, or because they were
considered extensions of the parent. Future empirical research should uncover the
extent to which the dynamics of this type of familicide overlap with (double) parri-
cide-only events and siblicide-only events.
Diffuse Conflict familicides represented the second-largest familicide cluster and
could be distinguished from the abovementioned clusters by the diverse constellation
of victims—ranging from in-laws, uncles, aunts, cousins, and secondary family rela-
tions. Based on the available information, it could be hypothesized that the more dis-
tant the family relation between victim and perpetrator (such as involving cousins,
nephews, uncles, and grandparents), the more likely the homicide resembles nonfam-
ily homicides. This was reflected in the lack of suicidal behavior following the event,
the perpetrator’s relatively frequent use of alcohol, the prevalence of conflicts over
money, and other conflicts of nonfamilial nature—characteristics seen in nonfamily
homicides nationwide (Fox & Zawitz, 1999). Future research should further assess
potential overlap between Diffuse Conflict familicides and multiple nonfamily
homicides.
In contrast to earlier research merely focusing on just one type of multiple family
homicides, we revealed the existence of four separate groups. What all four familicide
categories had in common, other than the defining fact that the victims consist of mul-
tiple family members, was the presence of primary and secondary victims. Secondary
victims were either killed because they are seen as extensions of the primary victim,
or because of their physical closeness to the primary victims. Arguably, this even
accounts for the Despondent Father, who perceives his victims not as extensions of
each other, but rather as extensions of himself. Such men considered themselves to be
the central figure in the lives of their family members (Cooper & Eaves, 1996), and as
such, aimed to protect his loved ones by taking them “along” in his suicide.
In sum, the results from this study showed that multiple family homicides constitute
a heterogeneous phenomenon, which due to the intimate bond between perpetrators and
victims cannot be equated with other types of multiple murderers. What flows from the
numerous differences between the four clusters is that there is no “one-size-fits-all”
at Universiteit Leiden \ LUMC on May 23, 2014hsx.sagepub.comDownloaded from
Liem and Reichelmann 55
prevention strategy. However, the results from this study and previous US-based stud-
ies (Websdale, 2010) revealed that virtually all familicides are committed with a fire-
arm. The availability of guns in a conflict-ridden home environment can quickly change
a heated argument into a deadly confrontation. States vary greatly in their ability to
conduct background checks in a timely manner and thereby prevent firearms from
being acquired by persons who are subject to a domestic abuse restraining order or who,
because of a (violent) criminal record, are prohibited from acquiring a firearm (Barber
et al., 2008). The results suggest that improved enforcement of restrictions on firearm
purchase, as well as gun removal from the houses of those with previous violent behav-
ior, may be a useful tool to reduce the prevalence and the lethality of these events.
A number of limitations apply to this study. First, the SHR data depend on the qual-
ity and completeness of coding by law enforcement, which is not always accurate
(Peck & Heide, 2012). Second, making use of supplemental newspaper articles is not
without problems, particularly when it concerns the underreporting of certain impor-
tant characteristics of the murders; for example, the fact that no mention was made of
a restraining order does not necessarily mean that a restraining order was not in place
(Liem et al., 2013). Furthermore, although we relied on a search method validated by
previous studies, familicide cases in the period 2008-2009 that were not reported in
newswires or in newspapers were not included in our study, possibly leading to an
underrepresentation of the actual prevalence of familicides in this time period.
In addition, factors that are important to the study of multiple family homicides
may not have been uncovered by these two sources; these variables include family
violence history, perpetrator’s mental health status, substance abuse and criminal his-
tory, and variables related to family dynamics. Future research should attempt to avoid
these caveats by moving beyond SHR and newspaper data. This can be done by incor-
porating interviews with perpetrators and those closest to them, as well as hospital
reports and criminal records (Liem et al., 2013). Based on this information, a more
accurate assessment can be made of the perpetrator’s mental and physical health and
personality (Cavanagh, Carson, Sharpe, & Lawrie, 2003). Including such detailed data
would lead to a more thorough understanding of the complex nature of these acts.
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 research was supported by a Marie Curie Outgoing
Fellowship for Career Development in the project 299875.
References
Aderibigbe, Y. A. (1997). Violence in America: A survey of suicide linked to homicides.
Journal of Forensic Sciences, 42, 662-665.
at Universiteit Leiden \ LUMC on May 23, 2014hsx.sagepub.comDownloaded from
56 Homicide Studies 18(1)
Alder, C. M., & Polk, K. (2001). Child victims of homicide. Cambridge, UK: Cambridge
University Press.
Barber, C. W., Azrael, D., Hemenway, D., Olson, L. M., Nie, C., Schaechter, J., & Walsh, S.
(2008). Suicides and suicide attempts following homicide. Homicide Studies, 12, 285-297.
Bijleveld, C., & Smit, P. (2006). Homicide in the Netherlands: On the structuring of homicide
typologies. Homicide Studies, 10, 195-219.
Blackburn, R. (1971). Personality types among abnormal homicides. British Journal of
Criminology, 11, 14-31.
Bourget, D., Gagné, P., & Labelle, M.-E. (2007). Parricide: A comparative study of matricide
versus patricide. Journal of the American Academy of Psychiatry and the Law Online, 35,
306-312.
Bowers, T. G., Holmes, E. S., & Rhom, A. (2010). The nature of mass murder and autogenic
massacre. Journal of Police and Criminal Psychology, 25, 59-66.
Byard, R., Knight, D., James, R. A., & Gilbert, J. (1999). Murder-suicides involving children:
A 29-year study. The American Journal of Forensic Medicine and Pathology, 20, 232-327.
Cavanagh, J. T. O., Carson, A. J., Sharpe, M., & Lawrie, S. M. (2003). Psychological autopsy
studies of suicide: A systematic review. Psychological Medicine, 33, 395-405.
Coid, J., Kahtan, N., Gault, S., & Jarman, B. (2000). Women admitted to secure forensic psychi-
atry services: II. Identification of categories using cluster analysis. The Journal of Forensic
Psychiatry, 11, 296-315.
Cooper, M., & Eaves, D. (1996). Suicide following homicide in the family. Violence and
Victims, 11, 99-112.
Daly, M., & Wilson, M. (1988). Homicide. New York, NY: Aldine de Gruyter.
Danson, L., & Soothill, K. (1996). Murder followed by suicide: A study of the reporting of mur-
der followed by suicide in The Times (1887-1990). The Journal of Forensic Psychiatry, 7,
310-322.
DeLisi, M., & Scherer, A. M. (2006). Multiple homicide offenders. Criminal Justice and
Behavior, 33, 367-391.
Deutsche Presse-Agentur. (2009, September 5). US man arrested in murders of 8 people,
including 7 relatives.
Dietz, P. E. (1986). Mass, serial and sensational homicides. Bulletin of the New York Academy
of Medicine, 62, 477.
Ewing, C. P. (1997). Fatal families. London, England: Sage.
Foucault, M. (1975). I, Pierre Rivière, having slaughtered my mother, my sister, and my brother:
A case of parricide in the 19th century. Linclon: University of Nebraska Press.
Fox, J. A. (2004). Missing data problems in the SHR: Imputing offender and relationship char-
acteristics. Homicide Studies, 8, 214-254.
Fox, J. A., & Levin, J. (2011). Extreme killing: Understanding serial and mass murder.
Thousand Oaks, CA: Sage.
Fox, J. A., & Swatt, M. L. (2009). Multiple imputation of the supplementary homicide reports,
1976-2005. Journal of Quantitative Criminology, 25, 51-77.
Fox, J. A., & Zawitz, M. W. (1999). Homicide trends in the United States. Washington, DC:
U.S. Department of Justice, Office of Justice Programs, Bureau of Justice Statistics.
Frazier, S. H. (1975). Violence and social impact. In J. C. G. Schoolar (Ed.), Research and the
psychiatric patient (pp. 191-200). New York, NY: Brunner & Mazel.
Friedman, S., Hrouda, D. R., Holden, C. E., Noffsinger, S. G., & Resnick, P. J. (2005). Filicide-
suicide: Common factors in parents who kill their children and themselves. Journal of the
American Academy of Psychiatry and the Law Online, 33, 496-504.
at Universiteit Leiden \ LUMC on May 23, 2014hsx.sagepub.comDownloaded from
Liem and Reichelmann 57
Fuentes, P. L. (2013). The Pierre Rivierre: A case of parricide as a Foucaldian model of power/
knowledge relations. Luz y Saber, 4, 79-91. Retrieved from: http://ejournals.ph/index.php?
journal=LS&page=article&op=view&path[]=5805 .
Goldney, R. D. (1977). Family murder followed by suicide. Forensic Science International, 3,
219-228.
Heide, K. M. (2013). Understanding parricide: When sons and daughters kill parents. New
York, NY: Oxford University Press.
Heide, K. M., & Frei, A. (2010). Matricide: A critique of the literature. Trauma, Violence, &
Abuse, 11, 3-17.
Koenraadt, F. (1996). Parricide: An ultimate offense. Deventer, Netherlands: Gouda Quint.
Kubrin, C. E. (2003). Structural covariates of homicide rates: Does type of homicide matter?
Journal of Research in Crime & Delinquency, 40, 139-170.
Leveillée, S., Lefebvre, J., & Marleau, J. D. (2009). Profil psychosocial des familicides com-
mis au Quebec - 1986 a 2000 [Psychological profile of familicides committed in Quebec].
Annales Médico-psychologiques, revue psychiatrique, 167, 591-596.
Liem, M., & Koenraadt, F. (2007). Homicide-suicide in the Netherlands: A study of newspaper
reports, 1992-2005. The Journal of Forensic Psychiatry & Psychology, 18, 482-493.
Liem, M., & Koenraadt, F. (2008). Familicide: A comparison with spousal and child homicide
by mentally disordered perpetrators. Criminal Behaviour and Mental Health, 18, 306-318.
Liem, M., Levin, J., Holland, C., & Fox, J. A. (2013). The nature and prevalence of familicide
in the United States, 2000-2009. Journal of Family Violence, 28, 351-358.
Malmquist, C. P. (1980). Psychiatric aspects of familicide. Bulletin of the American Academy of
Psychiatry and the Law, 8, 298-304.
Malphurs, J. E., & Cohen, D. (2002). A newspaper surveillance study of homicide-suicide in the
United States. The American Journal of Forensic Medicine and Pathology, 23, 142-148.
Marleau, J. D., Auclair, N., & Millaud, F. (2006). Comparison of factors associated with par-
ricide in adults and adolescents. Journal of Family Violence, 21, 321-325.
Marzuk, P. M., Tardiff, K., & Hirsch, C. S. (1992). The epidemiology of murder-suicide.
Journal of the American Medical Association, 267, 3179-3183.
Meloy, J. R. (1992). Violent attachments. Northvale, NJ: Aronson.
Millaud, F., Auclair, N., & Meunier, D. (1996). Parricide and mental illness: A study of 12
cases. International Journal of Law and Psychiatry, 19, 173-182.
Mouzos, J., & Rushforth, C. (2003). Family homicide in Australia (Trends and Issues in Crime
and Criminal Justice, No. 255). Canberra, Australia: Australian Institute of Criminology.
Palermo, G. B. (1997). The berserk syndrome: A review of mass murder. Aggression and
Violent Behavior, 2, 1-8.
Peck, J. H., & Heide, K. M. (2012). Juvenile involvement in fratricide and sororicide: An empir-
ical analysis of 32 years of US arrest data. Journal of Family Violence, 27, 749-760.
Saunders, D. G. (1992). A typology of men who batter: Three types derived from cluster analy-
sis. American Journal of Orthopsychiatry, 62, 264-275.
Scheinin, L., Rogers, C. B., & Sathyavagiswaran, L. (2011). Familicide-suicide: A cluster of 3
cases in Los Angeles County. The American Journal of Forensic Medicine and Pathology,
32, 327-330.
Schlesinger, L. B. (2000). Familicide, depression and catathymic process. Journal of Forensic
Sciences, 45, 200-203.
Somander, L. K. H., & Rammer, L. M. (1991). Intra- and extrafamilial child homicide in
Sweden 1971-1980. Child Abuse & Neglect, 15, 45-55.
at Universiteit Leiden \ LUMC on May 23, 2014hsx.sagepub.comDownloaded from
58 Homicide Studies 18(1)
Stefurak, T., & Calhoun, G. B. (2007). Subtypes of female juvenile offenders: A cluster analysis
of the Millon Adolescent Clinical Inventory. International Journal of Law and Psychiatry,
30, 95-111.
Thaller, J. (2012). The children are listening an autoethnographic account of the years leading
up to familicide in a quiet, suburban neighborhood. Qualitative Inquiry, 18, 266-272.
Walsh, J. A., & Krienert, J. L. (2009). A decade of child-initiated family violence: Comparative
analysis of child–parent violence and parricide examining offender, victim, and event char-
acteristics in a national sample of reported incidents, 1995-2005. Journal of Interpersonal
Violence, 24, 1450-1477.
Websdale, N. (2010). Familicidal hearts: The emotional styles of 211 killers (Vol. 5). New
York, NY: Oxford University Press.
Wilczynski, A. (1997). Child homicide. London, England: Greenwich Medical Media.
Wilson, M., Daly, M., & Daniele, A. (1995). Familicide: The killing of spouse and children.
Aggressive Behavior, 21, 275-291.
Yarvis, R. M. (1994). Patterns of substance abuse and intoxication among murderers. Journal of
the American Academy of Psychiatry and the Law Online, 22, 133-144.
Author Biographies
Marieke Liem is a Post-Doctoral Fellow with the Program in Criminal Justice Policy and
Management (PCJ) at Harvard Kennedy School (HKS). Her research interests involve domestic
homicide, homicide by the mentally ill, homicide followed by suicide, criminal careers of homi-
cide offenders, and international comparative research in lethal violence.
Ashley Reichelmann is a Ph.D. candidate in the Sociology Department at Northeastern
University, focusing on conflict and violence as well as social psychology.
at Universiteit Leiden \ LUMC on May 23, 2014hsx.sagepub.comDownloaded from