Content uploaded by Ray W Cooksey
Author content
All content in this area was uploaded by Ray W Cooksey on Apr 03, 2014
Content may be subject to copyright.
10.1177/0306624X02238159
International Journal of Offender Therapy and Comparative Criminology
Psychological Profiling of Serial Arson
Criminal Psychological Profiling
of Serial Arson Crimes
Richard N. Kocsis
Ray W. Cooksey
Abstract: The practice of criminal psychological profiling is frequently cited as being appli
-
cable to serial arson crimes. Despite this claim, there does not appear to be any empirical
research that examines serial arson offence behaviors in the context of profiling. This study
seeks to develop an empirical model of serial arsonist behaviors that can be systematically
associated with probable offender characteristics. Analysis has produced a model of offence
behaviors that identify four discrete behavior patterns, all of which share a constellation of
common nondiscriminatory behaviors. The inherent behavioral themes of each of these pat
-
terns are explored with discussion of their broader implications for our understandingofserial
arson and directions for future research.
Criminal psychological profiling is the forensic technique of analyzing crime
behaviors to construct a descriptive template of probable offenders (Wilson, Lin-
coln, & Kocsis, 1997). The practice of and research into profiling has predomi-
nantly been focussed on crimes of sexual violence such as murder and rape.
Although comparatively little research has actually been developed, profiling is
nonetheless frequently cited as also being applicable to the investigation of arson
crimes (Holmes & Holmes, 1996; Rossmo, 1997; Vorpagel, 1982). Despite this
reputation and acceptance of profiling by the law enforcement community, there
exists a surprising dearth of rigorous empirical research on the topic of profiling
(Kocsis, in press; Kocsis, Hayes & Irwin, 2002; Kocsis, Irwin, Hayes, & Nunn,
2000; Oleson, 1996). The objective of this study is to develop an empirical model
for the criminal psychological profiling of serial arson offences.
The majority of current social science research on arson is dominated by psy
-
chiatric or psychological studies that examine issues of mental status and/or
offender etiology (e.g., Barnett & Spitzer, 1994; Geller, 1992; Harris & Rice,
1996; Lewis & Yarnell, 1952; Rix, 1994), or criminological studies that propose
either varying motive-based classification taxonomies (Bennett & Hess, 1984;
Prins, Tennent, & Trick, 1985) or anecdotal case studies (e.g., Jeffers, 1992; Orr,
1989a, 1989b). Despite this body of literature on the broad topic of arson, little
empirical material actually exists on the psychological profiling of arson crimes
NOTE: Correspondence concerning this article should be addressed to Dr. Richard N. Kocsis, 16
Lynden Ave., Carlingford, Sydney NSW 2118, Australia; e-mail: richard_kocsis@hotmail.com.
International Journal of Offender Therapy and Comparative Criminology, 46(6), 2002 631-656
DOI: 10.1177/0306624X02238159
2002 Sage Publications
631
for criminal investigations and little, if any, exists on the specific topic of profiling
serial arsonists.
Possibly the first and largest body of research to examine arson for the specific
purpose of criminal psychological profiling was undertaken by the Behavioral
Science Unit of the American FBI (Icove & Estepp, 1987; Rider, 1980a, 1980b).
The central theme of this research program was the development of various
motive categories correlated with crime behaviors and typical offender character
-
istics. These studies culminated in the proposal of six broad motive categories for
arson (Douglas, Burgess, Burgess, & Ressler, 1992).
The first category identified was labeled vandalism and is described as being
motivated by wanton destruction. Offenders are typically juveniles, and their
crimes demonstrate little sophistication. The second category of excitement deals
with motives of psychological stimulation and includes crimes committed for his
-
trionic or heroic desires, to satisfy sexual fetishes, or owing to psychotic delusion.
Offenders in this category tend to be older, and such crimes are typically charac
-
terized by perpetrators remaining at crime scenes, engaging in masturbation, and
igniting low-risk targets such as dumpsters and vegetation. The third category of
profit deals with arsons motivated by some material gain and frequently involves
fraudulent insurance claims by the offenders and/or victims. As the title implies,
the fourth category of revenge is committed “in retaliation for some injustice, real
or imagined, perceived by the offender” (Douglas et al., 1992, p. 173). A prior
relationship between victims and offenders typically characterizes these offences,
and there is often evidence of premeditation and planning combined with the use
of accelerants. Offenders are typically adult men in blue-collar jobs. The fifth cat-
egory of crime concealment identifies arson as a means of concealing the evi-
dence of another crime. The sixth and final category, termed extremist, refers to
fires that are set to further some social, political, or religious objectives.
Perhaps the first study to advance any theory-driven development of profiling
techniques for arson was undertaken by Kocsis, Irwin, and Hayes (1998). This
study examined the organized-disorganized behavior dichotomy developed by
the FBI Behavioral Science Unit for sexual murderers within the context of arson
-
ists. The underlying premise of this dichotomy was the interpretation of sexual
murder crime scenes by their level of behavioral sophistication and the matching
of offender characteristics. The organized category describes a methodical, pre
-
meditated crime with offender characteristics such as maturity, resourcefulness,
and typically, sexual perversion being exhibited. Conversely, the disorganized
category represents a haphazard, almost random crime with offender characteris
-
tics such as immaturity, opportunism, and a likelihood of some form of mental
disorder being demonstrated (Ressler, Burgess, & Douglas, 1988).
Although the organized-disorganized typology was developed from a study of
sexual murderers, its generalization to arson profiling is evident in the exposition
offered by Douglas et al. (1992). The absence of any empirically derived data to
support this generalization prompted the study by Kocsis et al. (1998). Within the
confined parameters of two forms of arson offences, this study was able to repli
-
632 International Journal of Offender Therapy and Comparative Criminology
cate the dichotomous distinction of the organized-disorganized typology. The
results of this study found that such a simple dichotomy could not validly be sup
-
ported when applied to the full gamut of possible behaviors and motivations
involved in arson offences.
A key factor of previous profiling studies, such as that of Douglas et al. (1992),
has been the combination of inferred motivations with identifiable behaviors for
the construction of proposed typologies. However, as Canter and Heritage (1989)
indicated, the basic tenet of profiling is that offenders differ in their actions and
that these differences in behavior relate to offenders’ characteristics. Thus, the
interpretation of crime actions requires the classification of offence behaviors as
distinct from any inferred motivations. Consequently, the majority of studies used
for profiling that combine the inference of motivations with observable behaviors
are arguably empirically flawed (Godwin, 2000; Kocsis, 1999; Kocsis, Cooksey,
& Irwin, 2002; Kocsis, Irwin, & Cooksey, 2002).
The only previous study thus far to address this methodological issue within
the context of arson profiling was that of Canter and Fritzon (1998). This study
identified four basic patterns to arsonist crime scene behavior. The first pattern,
labeled instrumental person, is described as being the result of some form of dis-
pute between offenders and victims and is reminiscent of a revenge-motivated
scenario. Characteristics of this pattern include a pattern of threats and arguments
between offenders and victims, premeditation in the commission of offences, and
specific targets selected for attack. The second pattern, labeled instrumental
object, is an opportunistic style of offence with no coherent purpose for the com-
mission of crime. Behaviors characterizing this pattern include theft of property
and fire occurring in an external, visible location, typically on a weekday. This
pattern was found to be strongly associated with multiple juvenile offenders.
The third pattern, labeled expressive person, is characterized by some form of
histrionic goal of offenders with fires being set to “alleviate distress by seeking
attention” (Canter & Fritzon, 1998, p. 82). Characteristic behaviors in this pattern
include the presence of suicide notes, with offenders frequently presenting as vic
-
tims. The final pattern of expressive object is distinguished by multiple offences
believed to be committed to achieve some form of emotional relief. Behaviors
inherent to this pattern include multiple offences being perpetrated on hospitals,
businesses, or public buildings, with offenders being triggered into offending by
nonspecific events and remaining at crime scenes to observe the fires.
Although Canter and Fritzon (1998) made a definitive contribution to the
development of arson profiling in general, a number of issues remain to fully
inform our understanding of arson and serial arson profiling in particular. The
specific topic of serial arson profiling has been the focus of little empirical
research despite the additional difficulties such offences present to investigators.
Indeed, the technique of profiling has often been found to be of far more utility in
the investigation of recidivistic offences, as nonrecidivistic crimes can typically
be solved via regular investigative procedures (Ainsworth, 2001; Geberth, 1996).
Psychological Profiling of Serial Arson 633
This issue in particular acts as a special impetus to warrant far more detailed scru
-
tiny of serial arson crimes.
Irrespective of the crime modality, the majority of profiling studies fail to rec
-
ognize and account for possible commonalties in criminal behavior. That is, pre
-
vious studies typically employ categorical typologies that do not actually allow
for the discrimination of a specific pattern of offence from behaviors that may
simply be typical of the crime. In addition, these categorical constructs typically
do not provide any impression of the overall relationship between the varying cat
-
egories or any possible comorbidity of motives between patterns. These are
important issues, as the actual practice of psychological profiling in criminal
investigations does not adopt such static approaches. The published literature on
which such practice is supposedly based does not concord with these rigid con
-
structs (Britton, 1997; Dietz, 1985; Douglas, Ressler, Burgess,& Hartman, 1986).
In conclusion, this study seeks to develop an empirical model for the criminal
psychological profiling of serial arson crimes by independently analyzing offence
behaviors and thus avoiding the methodological pitfalls of previous studies. In
addition, this analysis will also explore and attempt to identify the existence of
common behaviors as distinct from those that can truly discern a specific pattern
of behavior. Furthermore, the developed model will provide some understanding
of the potential relationships between patterns. Finally, the sample for this study
will specifically focus on serial or recidivistic arsonists that thus far have received
surprisingly little examination in psychological profiling research.
METHOD
DATABASE AND DATA-SCREENING PROCESS
The database employed in this study consisted of 148 incidents of arson. The
cases were obtained from the New South Wales and Victorian police jurisdictions,
which represent the two largest police agencies in Australia. All cases featured a
minimum of three arson offences and satisfied the criteria for serial violent
offences as described by Kocsis and Irwin (1998). The cases dated as far back as
1980 and as recently as 1998. All offenders in the sample had been convicted and
incarcerated for their offences. The variables for this study were extracted from
those originally developed in Kocsis et al. (1998). These variables were screened
and markedfor retention in subsequent analysis if theydemonstrated (a) sufficient
nonmissing entries and (b) sufficient variability across categories within each
variable. Frequency distributions were computed for all variables in the database.
Extremely small variances (indicating a near constant) were deleted prior to anal
-
ysis, as were variables havingmissing valuesin more than 50% of database cases.
634 International Journal of Offender Therapy and Comparative Criminology
CONDENSATION OF THE VARIABLES
To facilitate analysis and interpretation, conceptually similar categories for
each variable were collapsed with a view to producing dichotomous (0, 1) mea
-
sures having a reasonable number of Category 1 responses. Most variables were
recoded on a presence-absence basis, whereas others were recoded into less or
more categories. Some variables having multiple categories were dummy coded
into several dichotomous variables (e.g., variables for arson target, distance
offender traveled to target, point(s) of origin for the fire, and so forth). Variables
with very few or no Category 1 responses remaining after this coding process
were deleted from the database. Several variables were almost perfectly corre
-
lated with other related variables in the database, and to avoid problems with
extreme multicollinearity, these redundant variables were also deleted from the
analysis.
The data-screening and variable condensation process yielded a final set of 71
variables. For the major analyses, variables were broadly grouped into conceptual
sets: offender personal characteristics (personal set containing 12 variables); gen-
eral offender behaviour characteristics (general set containing 14 variables);
arson event-specific offenderbehaviour and choice characteristics (event-specific
set containing 16 variables, including 7 variables coding the time and season cho-
sen for the arson event), and crime scene characteristics (crime scene set contain-
ing 29 variables, which included variables coding the target of the arson event).
An appendix detailing these variables is attached.
ANALYTICAL PROCESS
The analysis proceeded in several discrete stages, commencing with a
nonmetric multidimensional scaling (MDS) analysis of the 29 dichotomous vari-
ables in the crime scene set. This analysis was accomplished using the MDS pro-
gram in SYSTAT 9.0. Guttman’s coefficient of alienation minimization criterion
wasemployed to control the scaling process, and Jaccard’s measure of binary sim
-
ilarity was employed as the similarity measure (a process similar to that recently
employed by Kocsis, Cooksey, & Irwin, 2002 and Kocsis, Irwin, & Cooksey,
2002). The two-dimensional MDS solution that emerged from this stage was
retained for further analysis and interpretation. However, for purposes of MDS,
29 objects would be considered a sufficient sample for scaling (see Hair, Ander
-
son, Tatham, & Black, 1998).
The second stage of analysis subjected the resulting MDS dimensional coordi
-
nates to cluster analysis to facilitate a regional interpretation of the dimensional
solution (a process recommended by Coxon, 1982). Dissimilarity was measured
using the squared Euclidean distance, and the clustering algorithm employed was
Ward’s minimum variance hierarchical method. The standardized dimensional
coordinates of the MDS solution were then plotted on a scatterplot using different
symbols to distinguish cluster groupings and facilitate identification of crime
Psychological Profiling of Serial Arson 635
scene attribute regions. The number of entities clustered in this analysis could be
considered a sufficient sample of serial arsonists for the purposes of cluster analy
-
sis (see Hair et al., 1998).
The third stage of analysis focused on fitting external property vectors using
variables from the personal, general, and event-specific variable sets to the MDS
coordinates for each of the 29 variables in the crime scene set. This was accom
-
plished by building up a newdatabase containing the standardized coordinates for
the two MDS dimensions and conditional probabilities for each dichotomous
external property variable not contained in the crime scene set (i.e., those vari-
ables contained in the personal, general, and event-specific sets). For a combina
-
tion of specific external property variable and crime scene characteristic, the
mean for that external property variable within the category coded 1 for that spe
-
cific crime scene characteristic was computed. This mean directly represented the
conditional probability that the external property variable of interest (e.g.,
OLANG) equaled 1 (offender was bilingual or multilingual) when specific crime
scene characteristics (e.g., TOCCUPY) also equaled 1 (target was occupied with
people at the time of the arson). These conditional probabilities defined the exter-
nal property vector variables statistically fitted to the MDS coordinates. It is
important to acknowledge that the sample size for the conditional probability
analysis conducted would be considered too small to cross-validate the resulting
equation parameter estimates. However, as the purpose of this exercise was pri-
marily descriptive and relational as opposed to establishing equations for predic-
tive purposes, this sample was considered sufficientfor the purpose of this article.
Property fitting was accomplished using an extension of the multiple regres-
sion procedure for fitting direction cosines described by Kruskal and Wish (1978)
and implemented in the SYSTAT 9.0 using the vector method in the perceptual
mapping procedure. (Essentially, the procedure was that each external property
vector served as a dependent variable in a regression analysis where the two MDS
coordinates served as predictors.) Screening for external property variables sig
-
nificantly predicted by the two-dimension MDS configuration of crime scene
characteristics occurred on a variable-by-variable basis. The significance deci
-
sion was based on the omnibus F test for the regression analysis of that variable;
only those external property vectors that were predictable based on a criterion of p
< .05 were retained for display and interpretation. The final significant external
property vectors were displayed in sets by superimposing the fitted vectors (of
standardized unit length) in a series of two-dimensional MDS scatterplots. The
direction that each vector points is defined by the direction and magnitude of the
standardized regression coefficient for each of the two MDS dimensions. The fit
-
ted vector thus indicates where the relevant external attributes will tend to reflect a
code of 1 when the crime scene variablesin that region of the MDS space also tend
to reflect a code of 1 (i.e., the vector points to where the conditional probabilities
are highest).
636 International Journal of Offender Therapy and Comparative Criminology
RESULTS
MDS
Each of the two- through five-dimensional MDS solutions was examined, and
the two-dimensional solution was chosen as most interpretable (coefficient of
alienation = .230; R
2
= .816). Higher dimensional solutions produced a marginally
better fit to the data but at the cost of much greater interpretive complexity and
increased noncomparability to previous research findings (e.g., K ocsis, Cooksey, &
Irwin, 2002; Kocsis, Irwin, & Cooksey, 2002). Figure 1 shows the plot of the stan
-
dardized coordinates for the two-dimensional MDS solution for the 29 variables
in the crime scene set.
The two-dimensional MDS coordinates were hierarchically clustered, and five
clusters (two of which had meaningful embedded subclusters) of crime scene
variables were identified. These clusters divided the two-dimensional space of
crime scene variables into five nonoverlapping regions and two subregions. The
five clusters of coordinates are marked by distinct plotting symbols in Figure 1,
and the cluster regions have also been sketched in; a dotted ellipse and region
boundary denotes the meaningful subregion structure. Figure 1 could be inter
-
Psychological Profiling of Serial Arson 637
-2 -1 0 1 2 3
Crime Scene Characteristics - Dimension 1
-3
-2
-1
0
1
2
3
Crime Scene Characteristics - Dimension 2
EVIDENCE
PLANNED
TRELATIO
POOINT
TSECURTY
THEFT
MINFIRE
TOCCUPY
SINGPOO
ENTARGET
ACCELERA
MATERBRO
MAJFIRE
RANDOM
BUSPROP
TUNRELAT
POOEXTER
MULTIPOO
TUNOCCUP
VEHPROP
ADAMAGE
TRAILERS
RESPROP
SPECBURN
EDUPROP
COMPROP
MINPROP
SEXACTIV
STATPROP
5
4
3
2
1
Cluster
COMMON
THRILL
ANGER
WANTON
SEXUAL
Figure 1 Multidimensional Scaling of Crime Scene Variables in Two Dimensions With
Regional Hierarchical Clustering of Scaling Coordinates Superimposed
preted in several ways, but a regional interpretation is the clearest way. Crime
scene variables appearing in the same region of the plot were inspected for com
-
mon themes to obtain an interpretation of what each regionmight be indicating.
The central Cluster 1 (surrounded by the solid-boundary ellipse) represented
crime scene variables that were not clearly differentiated by the two-dimensional
MDS structure—they were associated by virtue of having similar coordinate pat
-
terns centered on or near zero for each dimension. Consequently, this cluster was
labeled the common behavior pattern. However, it is clear that within this central
region, a substructure was identified by the cluster analysis, differentiated some-
what along MDS Dimension 1. The small dotted ellipse surrounds those crime
scene variables,separating them from the remaining central variables.The pattern
suggests a planned arson event (PLANNED) where the target was related to the
offender(TRELATIO) and where physical evidence was left by the offender at the
scene (EVIDENCE). This leads to a suggestion that Dimension 1 may be tapping
some element of closeness or meaningfulness to offender as well as an element of
premeditation, an interpretation supported by patterns in other regions more
removed from the center.
Premeditation, especially, seems defensibly suggested by the patterns of vari-
ables moving from right to left along Dimension 1. Crime scene variables toward
the extreme left reflect many aspects associated with a deliberate and planned
arson event, bringing materials (MATERBRO), using accelerants (ACCELER)
and trailers (TRAILERS), burning specific items (SPECBURN), and deliberate
damage to other target items (ADAMAGE). Variables captured in the region of
Cluster 3 seem to especially reflect a deliberate and directed rage, perhaps of a
more personal nature because residential properties and vehicles are the targets
included in this cluster. Variables toward the right end of Dimension 1 suggest a
more random and anonymous arson crime pattern, with a suggestion of distur-
bance or perverseness (THEFT and SEXACTIV, respectively).
The central region is marked by complete omission of presence of any targeted
property variables. Inspection of Figure 1 seems to indicate that a major thrust of
Dimension 2 is to differentiate target properties that were large and public institu
-
tions (i.e., educational [EDUPROP], state-owned [STATPROP], or commercial
[COMPROP] property) toward the top of the figure—positive coordinates) from
those that were small and noninstitutional (VEHPROP, or vehicle property;
MINPROP, or minor property; BUSPROP, or bush property; toward the bottom of
the figure—negative coordinates). (RESPROP, or residential property, is at the
zero point along Dimension 2.) The variables captured in the region of Cluster 4
seem especially related to specific targeted destruction of educational and com
-
mercial properties where specific items are used to start fires. The suggestion here
might relate to a pattern of strong grudge and directed anger against large public
institutions. The region defined by Cluster 2 has a defined substructure, differenti
-
ated along Dimension 1 where the four variables (MATERBRO, ACCELER,
TUNOCCUP or target unoccupied, and MAJFIRE or major fire) to the left of the
dotted line are more closely aligned with deliberate planned arson that results in
638 International Journal of Offender Therapy and Comparative Criminology
large fires although in unoccupied, smaller-scale properties. Those variables to
the right of the dotted line in Cluster 2 are more suggestive of unplanned random
arson events having multiple and exterior points of origin (MULTIPOO and
POEXTER, respectively). Targets here tend to be outdoors (BUSPROP and
MINPROP) and unrelated to offenders (TUNRELAT). In general, Cluster 2 is dif
-
ferentiated from Clusters 1, 4, and 5 (toward the opposing end of Dimension 2) on
the basis of size and institutional nature of property targets, involving a physical
man-made structure to destroy (toward the positive pole of the dimension), as well
as by the general unprotected nature of the targets where access is relatively easy
(toward the negative pole of the dimension). Note, for example, that crime vari
-
ables linked to having to enter the target (ENTARGET), targets where security
systems are present (TSECURTY), targets with a single point of entry
(SINGPOO), and educational, commercial, or state-owned property types sug
-
gest protected, man-made targets to which offenders have to gain access to start
fires—features not captured in the Cluster 2 region.
EXTERNAL PROPERTY VECTOR FITTING
Figures 2 through 4 summarize the property vector fitting analyses designed to
explicitly employ specific characteristics in the personal, general, and event-
specific variable sets to facilitate interpretation and understanding of the
dimensionality of crime scene characteristics. These analyses explicitly link
crime scene characteristics to conditional probability patterns that would be most
useful in profiling serial arsonists. It should be noted that certain crime scene vari-
ables may be missing from specific figures due to the presence of missing obser-
vations for that particular variable in the property fitting analysis. Where this
occurs, slight butinconsequential perturbations in the preference mapping scaling
solution represented will be present. Table 1 provides relevant numerical data
(e.g., standardized regression weights, multiple r values, omnibus F-test values,
and p values) used to facilitate the property-fitting exercise for the personal, gen
-
eral, and event-specific variable sets, respectively. Table 1 also lists those external
property variableswhose fitted vectors are not displayed in the figures by virtue of
not being significantly predicted by the two-dimensional MDS coordinates.
These variables are therefore largely irrelevant for achieving an understanding of
the association between crime scene variables and offender-related variables.
Interpretation of Figures 2 through 4, one plot for each of the three conceptual
sets of offender-related external variables,is relatively straightforward, especially
when interpreted in conjunction with the regions identified in Figure 1 and the
interpretations of the dimensions offered in the previous section. Each fitted vec
-
tor on a figure summarizes the relationship that exists between a specific external
offender-related variable and the two dimensions (i.e., the spatial pattern) of the
MDS solution. The strength of the relationship is measured by the multiple corre
-
lation between the two dimensional coordinates and the conditional probability
scores for the variables. The direction of the relationship directly reflects the com
-
Psychological Profiling of Serial Arson 639
bination of signs of the standardized regression weights for the two dimensions.
Therefore, movement toward the circle along a particular vector is interpreted as
reflecting an increasing tendency (i.e., the conditional probability that the
offender-related variable takes on a value of 1 when a crime scene variable takes
on a value of 1) for the offender-related variable to take on a coded value of 1 in
association with the crime scene variables in the region of the circle and vice-
versa when moving in the opposite direction.
Personal offender characteristics. Figure 2 shows fitted property vectors for
six significant personal characteristics of offenders.The OLANG (offenders were
bilingual) and OACCENT (offenders spoke with accents) vectors showed higher
conditional probabilities, with crime scene characteristics falling toward the plan
-
ning or deliberation end of Dimension 1 and especially oriented toward the Clus
-
ter 3 region. Thus, there were higher probabilities that offenders were at least
bilingual and spoke with accents at crime scenes where the TRAILERS,
RESPROP, and VEHPROP variables took on a coded value of 1. The
OHAIRCOL and OEYECOL vectors generally suggested a higher probability of
640 International Journal of Offender Therapy and Comparative Criminology
-3 -1 1 3
Dimension 1
-3
-1
1
3
Dimension 2
VEHPROP
TRAILERS
RESPROP
SPECBURN
EVIDENCE
PLANNED
ACCELERA
TRELATIO
MATERBRO
EDUPROP
MAJFIRE
POOINT
TSECURTY
ENTARGET
MINPROP
SEXACTIV
STATPROP
THEFT
RANDOM
MINFIRE
BUSPROP
TUNRELAT
POOEXTER
TOCCUPY
COMPROP
MULTIPOO
TUNOCCUP
SINGPOO
OLANG
OACCENT
OHAIRCOL
OUTFEAT
OTEETH
OEYECOL
Figure 2 Fitted (Significant) Personal Offender Characteristic Pattern Vectors
offenders having darker hair and eyes, respectively, at crime scenes targeting
commercial and state-owned properties and where sexual activity was evident
(Cluster 5 orientation). The OTEETH and OUTFEAT vectors revealed higher
probabilities of offenders having noticeably imperfect teeth and an outstanding
feature, respectively, at the more random outdoor crime scenes that had multiple
points of entry and were unrelated to the offender (Cluster 2 orientation). Finally,
readers are reminded that the data originate from police records. Consequently,
some of the offenders’personal characteristic variables were included to provide
descriptive richness to the results. These results should not be interpreted or used
in any manner that would suggest racial stereotyping.
General offender behavior variables. Figure 3 shows the fitted property vec
-
tors for nine significant general offender behaviors. The OCRIMSTAT,
OVEHUSED, and OINTERNA vectors generally suggested higher probabilities
of offenders having a prior criminal status, using a vehicle to commit crimes, and
having been overseas within the past 10 years, respectively, at crime scenes espe
-
cially associated with Clusters 2 and 4 and the small subregion of Cluster 1 (using
Psychological Profiling of Serial Arson 641
-3 -1 1 3
Dimension 1
-3
-1
1
3
Dimension 2
VEHPROP
ADAMAGE
TRAILERS
RESPROP
SPECBURN
EVIDENCE
PLANNED
ACCELERA
TRELATIO
MATERBRO
EDUPROP
MAJFIRE
POOINT
STATPROP
THEFT
RANDOM
MINFIRE
BUSPROP
TUNRELAT
POOEXTER
TOCCUPY
COMPROP
MULTIPOO
TUNOCCUP
SINGPOO
ENTARGET
TSECURTY
OINTERNA
OVEHUSED
OCRIMSTA
OCONFESS
OINTERST
OLIVEWTH
OJOBTYPE
ODRGALCO
OVESTAT
Figure 3 Fitted (Significant) General Offender Behavior Pattern Vectors
trailers, targeting educational and residential properties, causing other damage to
targets, starting fires with specific items and where evidence tends to be left). The
OINTERST vector revealed a higher probability of offenders having traveled out
of state within the past 10 years at crime scenes oriented toward Clusters 4 and 5
and the outer fringes of Cluster 1 (especially targeting educational, commercial,
and state-owned properties where theft tends to also occur) the past 10 years, hav
-
ing facial hair, and having darker hair color and shade. The OVESTAT vector
showed a higher probability of offenders not owning the vehicle used in a crime at
crime scenes especially associated with Cluster 5 and the outer fringes of Cluster
1 (where state-owned properties are targeted, theft tends to occur, and minor fires
result). The ODRGALCO, OLIVEWTH, and OJOBTYPE vectors suggested
higher probabilities of offenders showing evidence of drug or alcohol use, living
with other people, and being employed, respectively, at crime scenes especially
associated with Cluster 2 variables (random acts at targets unrelated to offenders,
targeting bush properties, with points of origin exterior to the target). The
OCONFESS vector suggested a higher probability of offenders having confessed
to similar crimes at crime scenes especially associated with the more deliberative
642 International Journal of Offender Therapy and Comparative Criminology
-3 -1 1 3
Dimension 1
-3
-1
1
3
Dimension 2
ADAMAGE
TRAILERS
RESPROP
SPECBURN
EVIDENCE
PLANNED
ACCELERA
TRELATIO
MATERBRO
EDUPROP
MAJFIRE
POOINT
MINPROP
STATPROP
RANDOM
MINFIRE
BUSPROP
TUNRELAT
POOEXTER
TOCCUPY
TUNOCCUP
SINGPOO
ENTARGET
TSECURTY
NOTPRES
DISTMAJ
SUMSPRIG
HOLIDAY
WEEKEND
ACCOMPLI
VISIBLE
WEEK
WINAUTM
DISTMIN
PRESENT
Figure 4 Fitted (Significant)Event-Specific Offender Behavior and Choice PatternVectors
Psychological Profiling of Serial Arson 643
TABLE 1
FIT STATISTICS FOR EXTERNAL
OFFENDER-RELATED PROPERTY VECTORS
Beta Weight Beta Weight
for for Multiple Omnibus
Variable Dimension 1 Dimension 2 Correlation F Test pValue
Personal set
Significant variables
a
OLANG –.57 –.09 .58 6.54 .01
OHAIRCOL .31 .46 .57 6.04 .01
OEYECOL .54 .19 .58 6.41 .01
OTEETH .38 –.33 .49 3.92 .03
OUTFEAT .33 –.37 .49 3.92 .03
OACCENT –.66 .16 .67 10.44 < .01
Nonsignificant variables
OAGE, OBUILD,
OHAIRSHA,
OHAIRLEN,
OFACHAIR,
OODOUR
General set
Significant variables
OINTERST .22 .63 .66 9.43 < .01
OINTERNA –.73 .18 .75 15.49 < .01
OLIVEWTH .33 –.70 .78 18.75 < .01
OJOBTYPE .24 –.48 .54 4.91 .02
ODRGALCO .43 –.25 .52 4.53 .02
OCRIMSTA –.36 .40 .57 5.76 .01
OCONFESS –.33 –.35 .46 3.31 .05
OVEHUSED –.51 .43 .67 10.45 < .01
OVEHSTAT .67 .10 .68 10.95 < .01
Nonsignificant variables
OLIFESTY, OSEXHAB,
OMENPROB,
OPOSPROP,
OVEHTYPE
Event-specific set
Significant variables
DISTMAJ –.48 –.56 .74 15.20 < .01
DISTMIN .48 .56 .74 15.20 < .01
ACCOMPLI .16 –.65 .67 10.75 < .01
VISIBLE .19 –.62 .65 9.52 < .01
PRESENT .52 .07 .53 4.99 .01
(continued)
subregion of Cluster 2 (where accelerants are used, a major fire results, and mate
-
rials are brought to the target by offenders).
Event-specific offender-behaviors and choices variables. Figure 4 shows the
fitted property vectors for 11 significant event-specific offender behaviors and
choices. It should be noted that some of these vectors are concerned with offend
-
ers’ choice of timing for arson crimes, and the remainder concern offenders’
actions taken in relation to a specific arson event. The timing-related vectors will
644 International Journal of Offender Therapy and Comparative Criminology
NOTPRES –.52 –.07 .53 4.99 .01
WEEK .40 .72 .69 9.35 < .01
WEEKEND –.40 –.72 .69 9.35 < .01
HOLIDAY –.28 .37 .54 4.29 .03
SUMSPRIG –.48 –.54 .58 5.31 .01
WINAUTM .48 .54 .58 5.31 .01
Nonsignificant variables
THREAT, ACALLS,
AEXTIN, NIGHT, DAY
NOTE: OLANG = offenders were bilingual or multilingual; OHAIRCOL = offenders had darker hair;
OEYECOL = offenders had darker eyes; OTEETH = offenders had noticeably imperfect teeth;
OUTFEAT = offenders had outstanding features; OACCENT = offenders spoke with accents; OAGE =
offenders’age; OBUILD = offenders’ build; OHAIRSHA = offenders’ hair shade; OHAIRLEN = of
-
fenders’ hair length; OHAIRCOLoffenders’ hair colour; OFACHAIR = offenders’ facial hair;
OODOUR = noticeable offender odour; OINTERST = offenders traveled out of state within the past 10
years; OINTERNA = offenders had been overseas within the past 10 years; OLIVEWTH = offenders
lived with other people; OJOBTYPE = offenders were employed; ODRGALCO = offenders showed
evidence of drug or alcohol use; OCRIMSTA = offenders had prior criminal status; OCONFESS = of-
fenders confessed to similar crimes; OVEHUSED = offenders used a vehicle to commit crimes;
OVEHSTAT = offenders’vehicle status; OLIFESTY = offenders’ general lifestyle; OSEXHAB = of-
fenders’ sexual habits; OMENPROB = offenders displayed symptoms or had been treated for mental
problems; OPOSPROP = offenders possessed others’s property; OVEHTYPE = offenders’ vehicle
type; DISTMAJ = offenders traveled more than 1 km to the target; DISTMIN = offenders traveled less
than 1 km to the target; ACCOMPLI = offenders had accomplices in committing the arson; VISIBLE =
offenders lit fire in highly visible location with potential witnesses; PRESENT = offenders were pres
-
ent at the crime scenes watching the fires; NOTPRES = offenders were not present at the crime scenes
watching the fires; WEEK = offenders set fires on weekdays; WEEKEND = offenders set fires on
weekends; HOLIDAY = offender sets fires during some type of holiday period; SUMSPRIG = offend
-
ers set fires during the summer or spring seasons; WINAUTM = offenders set fires during the winter or
autumn seasons; THREAT = offenders made threats to someone about committing arson; ACALLS =
offenders reported fires they actually started; AEXTIN = offenders were involved in attempts to extin
-
guish fires they actually set; NIGHT = offenders set fires at night; DAY = offenders set fires during the
day.
a. Only significant offender-related variables will have the fitted property vectors displayed in Figures
2 through 4.
TABLE 1 (continued)
Beta Weight Beta Weight
for for Multiple Omnibus
Variable Dimension 1 Dimension 2 Correlation F Test pValue
be interpreted first. The HOLIDAY vector suggested a higher probability of
offenders committing arson during holiday periods at crime scenes particularly
associated with Cluster 4 and the small subregion of Cluster 1 (where educational
properties are targeted, specific items are used to start the fire, evidence is left
behind, the offenderis related to the target, and other damage is done at the target).
This relationship pattern is probably most closely linked to vandalism activity at
schools during holiday periods. The WEEK vectorwas oriented more toward both
Cluster 4 and 5 variables and the outer fringes of Cluster 1. Here, a higher proba
-
bility of offenders choosing weekdays for their crimes was associated with crime
scenes showing features such as educational or state-owned properties, where
security systems were present, theft was likely to have occurred, point of origin of
the fire was interior to the target, and there was evidence that offenders actually
entered the target. The WEEKEND vector showed the opposite directional trend
where offenders tended to choose a weekend day to start their fires, which was
associated with those crime scene variables located in the more deliberative sub
-
region of Cluster 2 (where materials were brought to the target, accelerants were
used, and a major fire resulted). The SUMSPRIG and WINAUTM vectors tended
to work almost in parallel (i.e., showingnearly identical association patterns) with
the WEEKEND and WEEK vectors, respectively.
With respect to event-specific offender behaviors, the PRESENT vector sug-
gested a higher probability of offenders actually being present at the crime scene
during the fire, at crime scenes associated with Cluster 5 (where the crime was
random rather than planned, state-owned property or minor property had been tar-
geted, small fires resulted, and where theft was likely to have occurred). The
NOTPRES vector worked in the opposite direction, suggesting that offenders
tended not to be present where crimes had been planned and accelerants and trail-
ers had been used. The ACCOMPLI and VISIBLE vectors were nearly parallel
and revealed higher probabilities of offenders working with accomplices and
starting fires in visible areas with possible witnesses at crime scenes associated
strongly with Cluster 2 features (where targets tended to be outdoors with multi
-
ple and exterior points of origin for fires, where targets were unoccupied, where
the crime tended to be random rather than planned, and where a major fire
ensued).
DISCUSSION
The results have produced an empirical model for serial arson crime scene
behaviors (depicted in Figure 1) that can be systematically associated with proba
-
ble offender characteristics (see Figures 2-4). The model depicted in Figure 1
shows that serial arson crime scene behaviors are composed of a centrally located
constellation of common behaviors surrounded by four outlying patterns. Each of
these four outlying patterns represents a distinct and coherent style to the commis
-
sion of a serial arson attack.
Psychological Profiling of Serial Arson 645
Focusing first on the central cluster, it can be seen that serial arson is composed
of a broad constellation of common behaviors, which also contains a further sub
-
set of behaviors (shown by the smaller dotted ellipse) that are more closely related
to crime scene behaviors to the left of Figure 1. These common behaviors provide
a core description of the characteristic behaviors common to all forms or patterns
of serial arson. The subset indicates that crime patterns toward the left of Figure 1
are quite likely to contain these three specific behaviors. Thus, for example, Clus
-
ter 3 will have a very high propensity for planning in the commission of the
offence, a relationship between the victim and target, and evidence left at the
crime scene.
A number of interesting theoretical implications emerge when considering the
significance of the behaviors located in the common behavior cluster. The com
-
mon presence of planning and evidence in the commission of most offences is at
odds with the basic tenet of the organized-disorganized behavior dichotomy. The
main premise of the dichotomy is the categorical distinction of behaviors by their
offence sophistication. The presence of planning suggests an organized offender,
and its absence is indicative of a disorganized offender. However, given that plan-
ning is located in the common behavior cluster, generally all serial arson offences
will typically involve planning, which therefore serves to question the validity of
such a categorical distinction. Similarly, the common presence of evidence at
most serial arson crime scenes is not congruent with the postulates of the
organized-disorganized dichotomy, which cites the detection of evidence as a key
indicator of a disorganized offender (Douglas et al., 1992).
Planning as an apparently central element to all serial arson offences is
matched by the observation that some form of relationship usually exists between
victims and offenders. This result is contrary to previous conceptions, which
describe these crimes as seemingly random and motiveless (e.g., Holmes &
Holmes, 1998) and suggests that at some psychological level, there is indeed
coherency or the proverbial method to their madness. Examples of this relation
-
ship can range from cognitive knowledge of their environment to some internal
fantasy that is then superimposed on presented targets. From an investigative per
-
spective, this point may proveespecially useful as it suggests that a careful consid
-
eration of the target is likely to provide some insight into offenders. Indeed, the
nature of this relationship may become especially more overt when considered in
conjunction with behaviors derived in one of the outlying patterns.
This theme of planning as a central element to all offences is congruent with
those previously observed by Kocsis, Cooksey, and Irwin (2002) and Kocsis,
Irwin, and Cooksey (2002) in sexual murder and rapes and to some extent by Can
-
ter and Fritzon (1998). Indeed, these findings match those previously found by
Kocsis et al. (1998) in highlighting the theoretical limitations of the organized-
disorganized dichotomy and suggest its reconsideration beyond a categorical
classificationand perhaps into a conceptual continuum. Indeed, a tentativeindica
-
tion of such a continuum model can be conceived when examining behaviors in
Figure 1 along Dimension 1 or from left to right.
646 International Journal of Offender Therapy and Comparative Criminology
Looking at the other actions encompassed by the common behaviors cluster, a
number of other features emerge in the inherent nature of serial arson crimes. Pro
-
vided targets are structures, arsonists commonly enter targets, steal items if avail
-
able, and then initiate small fires via single internal points of origin. This is partic
-
ularly interesting as offenders may engage in such behaviors although targets may
be equipped with various security and/or fire retardant devices or even occupied
by individuals who may detect and apprehend them. This constellation of behav
-
iors suggests that the inherent psychological nature of serial arson is a somewhat
brazen crime that does not concord with common behavior patterns observable in
other crime modalities such as murder or rape in which offenders are typically
deterred by a greater degree of risk in apprehension or a diminished capacity to
complete their crimes (Kocsis, Cooksey, & Irwin, 2002; Kocsis, Irwin, & Cook
-
sey, 2002).
Turning to the first of the outlying behavior patterns, the thrill pattern (Cluster
2) is the only outlying pattern that also contains a subset of behaviors differenti
-
ated along Dimension 1 (i.e., from left to right) of Figure 1. The thrill pattern
embodies a somewhat sporadic style of offence. However, this sporadic nature
should not be mistaken for incoherence such as with the classic disorganized
offender category but rather indicates that multiple targets may be attacked. This
nuance is more apparent when it is recognized that the thrill pattern can nonethe-
less demonstrate quite sophisticated and premeditated behaviors.
Indeed, the distinguishing element between the subsets of the thrill pattern is
evidence of such behavioral sophistication such as the use of various resources to
initiate a fire. Thus, the subset to the left is represented by behaviors such as
offender use of materials and accelerants that subsequently achieve a much larger
fire. The subset to the right is characterized more by behaviors concerned with
committing multiple attacks on comparatively unrelated targets. Greater insight
into the nature of any offendertarget relationship becomes apparent when it is rec
-
ognized that the predominant target in this pattern is some form of bush, forest, or
vegetation. Consequently, the scope of possible relations between such targets is
lessened in comparison to other targets such as residences against whose occu
-
pants offenders harbor prior grievances.
A curious set of offender characteristics is associated with this style of offence.
In physical appearances, these offenders tend to have poor dental work and some
type of outstanding physical feature (e.g., scarring). In addition, this pattern will
typically be composed of multiple offenders who are employed, live with others
(i.e., are not social loners), and ingest alcohol and/or drugs prior to the commis
-
sion of offences. As they commonly travel over 1 mile to commit offences, they
will typically be quite mobile. Looking at temporal features, these offenders usu
-
ally commit their offences in the summer or spring and predominantly on week
-
ends. Finally, striking characteristics also typically emerge from these offenders.
First, they will initiate a fire in a highly visible location where they may be poten
-
tially identified and apprehended. Second, upon apprehension, these offenders
are likely to confess to having committed similar other crimes.
Psychological Profiling of Serial Arson 647
What emerges from this pattern are older individuals who are socially compe
-
tent in their lives and cognitivelyaware of their actions yet engage in a high degree
of risk to commit arson (hence the thrill label for this pattern). Unlike the other
patterns discussed herein, no significant element of animosity appears to exist
with this pattern. Instead, these offenders seem to derive satisfaction from the
destruction of property. Although they are not explicitly sexual in their behaviors,
the fact that these offenders are typically older and physically unattractive may
suggest the sublimation of a possible sexual drive. Thus, this pattern holds some
similarities to the broad excitement category proposed by Douglas et al. (1992).
This pattern seems to capture the theme of a recreational fascination with fire.
Some anecdotal case examples of this psychological pattern include restless
firefighters seeking to create some action for themselves or bored teenagers deriv
-
ing excitement from the risk of igniting fires. These scenario-based examples are
quite different, but both share the underlying theme of the thrill pattern in creating
excitement or entertainment for themselves through setting fires.
Movingto the next outlying behavior pattern (Cluster 3), labeled the anger pat
-
tern, we find a style of offence in which some animosity or rage seems to find
expression in the commission of an arson attack. A significant feature of this
behavior pattern is that targets are predominantly residential properties or motor
vehicles. This distinction in target selection is important as it suggests the vio-
lence in the crime is associated with expressing or inflicting personalized harm
rather than general destruction inflicted on an intangible organization or entity.
Indeed, this intent to cause harm also findsexpression in the other two characteris-
tic behaviorsin this pattern. First, offenders employ trailers to ensure the thorough
spread of the fire. Of more significance, however, is that offenders’anger will typ-
ically find further expression in physically destroying household items in addition
to the subsequent damage caused by igniting fires. So, for example, the offender
will enter the residence and then manually destroy some of the household items
(e.g., television or stereo system) before lighting a fire.
The perpetrators of such crimes in this sample tend to be foreign nationals who
are bilingual and consequently tend to possess noticeable accents. They also have
a certain degree of financial stability, as they possess and use vehicles for trans
-
port. Finally, the burningof targets does not appear to hold any deeper psychologi
-
cal meaning other than to inflict harm, and consequently, offenders promptly
leave crime scenes once fires have been lit.
At a cursory level, there are similarities between the anger pattern and previ
-
ously developed conceptions such as the revenge-motivated arsonist by Douglas
et al. (1992) or the instrumental person proposed by Canter and Fritzon (1998).
However, a number of important differences do exist between these previous con
-
ceptions and the anger pattern. Foremost among these is the nature of the relation
-
ship between offenders and targets. Both the revenge motive and the instrumental
person pattern share the identical theme of retaliation and retribution as the under
-
lying motives of the offences. However, the concept of prior relations between tar
-
gets and offenders in this study is based more on cognitive knowledge or recogni
-
648 International Journal of Offender Therapy and Comparative Criminology
tion of targets. Consequently, offenders in the anger pattern may not actually
possess any previous animosity toward the target but may instead attack due to
some perceived familiarity or recognition, although not necessarily due to the
existence of a grievance with the specific target. This distinction is further high
-
lighted as the anger pattern describes the actions of serial arsonists, whereas the
revenge motive and instrumental person are both primarily conceived as
nonrecidivistic offences that find expression in very specific targets. Indeed, the
theme of the anger pattern is similar to that found by Kocsis, Cooksey, and Irwin
(2002) and Kocsis, Irwin, and Cooksey (2002) in studying sexual murderers and
rapists, both of which contained a behavior pattern demonstrating the expression
of an unfocused internal rage on targets.
Cluster 4 has been labeled the wanton pattern. The theme of this pattern
appears to be a generalized sentiment of animosity visited on a vague class of tar
-
gets. Thus, the wanton pattern predominantly features attacks on educational
facilities such as schools and universities or on commercial properties or business
establishments. In addition, offenders in this pattern will initially ignite specific
items within targets, suggesting some specific meaning in the initial destruction
of these items. Typical offender characteristics associated with this pattern
include a prior criminal history and a proclivity for offenses to be committed on
weekends.
There is some similarity between the theme of the wanton pattern and previous
categorizations such as vandalism-motivated arsons (Douglas et al., 1992) or the
instrumental object offence pattern described by Canter and Fritzon (1998). They
are similar in that the attacks are of an unfocussed nature. However, akin to the
observations in the anger pattern, these previous categorizations were not specifi-
cally developed in considering the actions of serial offenders. Focussing exclu-
sively on the targeting of educational facilities, some resemblance does exist
between the wanton pattern and a crime phenomena referred to as school fires
(Burns, 1991) in which juveniles specifically attack educational facilities. This
specific scenario also accommodates offences that occur predominately on week
-
ends. Nonetheless, these previous conceptions fail to adequately explain the
potential existence of an element of animosity as expressed by the initial burning
of specific items or the propensity for commercial properties to also be attacked in
similar circumstances. Clearly, this pattern will warrant further scrutiny in future
research.
The fifth and final cluster is labeled the sexual pattern and embodies an offence
style in which offenders associate the ignition of fires with sexual excitement and/
or gratification. The most distinguishing behavioral element to this pattern is evi
-
dence of sexual activity by offenders in or nearby the crime scene. The common
targetin this offencepattern is state-owned (public) premises that are easily acces
-
sible such as trash receptacles, post boxes, public toilets, or any other publicly
accessible facilities. These arson attacks are relatively minor in size and do not
typically escalate into major fires that may cause serious destruction. Indeed, this
Psychological Profiling of Serial Arson 649
pattern demonstrates the lowest amount of behavioral sophistication and as such,
within the regionalinterpretation of the data, is found in the far right of Figure 1.
Looking at typical offender characteristics associated with this pattern, offend
-
ers are likely to have dark colored hair and eyes and a history of domestic travel,
although not via their own vehicles. In the commission of their offences, these
offendersdo not generally travel far and typically light fires on weekdays and dur
-
ing winter and/or autumn months. However, most significant is that these offend
-
ers have a tendency to remain at the crime scene, typically to observe the fires and/
or their extinguishments.
The sexual pattern has a number of clear similarities with previous psychiatric
and/or psychological studies of arsonists as representations of forms of sexual
perversion or paraphilic compulsion (e.g., Barnett & Spitzer, 1994; Lewis &
Yarnell, 1952). However, within the context of the profiling literature, the sexual
pattern has a number of similarities to the excitement motive by Douglas et al.
(1992) or the expressive object offence pattern observed by Canter and Fritzon
(1998). Indeed, the expressive object pattern is described as the only arson behav
-
ior pattern that is recidivistic in nature in terms of selecting public buildings as
common targets and is believed to be committed to achieve some form of emo-
tional relief. Despite these similarities, a number of significant inconsistencies are
also apparent. Whereas the expressive object pattern is said to be instigated to
achieve some emotional relief, the nature of this relief is not typically explicit and
consequently may not equate with sexual perversion. In addition, the selected tar-
gets described in the expressive object are large structures (e.g., hospitals, busi-
nesses), whereas the typical victim class in the sexual pattern can be conceived
more in the nature of minor nuisance fires on small, public, and easily accessed
targets.
In conclusion, these results do not represent an exhaustive appraisal of serial
arson behavior patterns but rather offer a previously unavailable empirical model
on which such behaviors can systematically be assessed. The key feature of this
model is a holistic depiction of all potential behavioral patterns, both common and
discriminatory, which was not initially formulated on the inference of potential
motives. It is hoped that this model will have a number of effects in both the
research and the practice of criminal psychological profiling. With respect to the
actual practice of profiling, this model now provides an empirically robust basis
on which future serial arson crime series may be assessed and may provide insight
to both offence style and offender characteristics. It is hoped that this model will
also provide impetus for further research into the crime of serial arson. Examples
could include an examination of psychological factors that instigate offences in
the various patterns, temporal trends between offences, or measured effectiveness
of varying investigative tactics dependent on the style of the offences committed.
Indeed, replication of this study and comparison of the underlying patterns to
other crime modalities such as murder and rape may offer some higher theoretical
insight into the nature of serial violent crimes. It is hoped that the options and
answers to such research and development lie in the not-too-distant future.
650 International Journal of Offender Therapy and Comparative Criminology
APPENDIX
Variable Sets, Names, and Definitions for the Serial Arsonist Data Set
Variable
Variable Set Label Definition (Category Labels and Coding)
Personal offender
characteristics
OAGE Offender’s age (20 years old or younger = 0;
21 years or older = 1)
OLANG Offender’s language background (monolingual = 0;
bilingual = 1)
OBUILD Offender’s build (small = 0; medium or large = 1)
OHAIRSHA Offender’s hair shade (lighter = 0; darker = 1)
OHAIRLEN Offender’s hair length (short or none (1-3) = 0;
medium or long (4-6) = 1)
OHAIRCOL Offender’s hair colour (red, grey, or white = 0;
brown or black = 1)
OEYECOL Offender’s eye colour (light eyes = 0; dark eyes = 1)
OTEETH Offender’s teeth (not noticed = 0; noticeably
imperfect = 1)
OFACHAIR Offender’s facial hair (no = 0; yes = 1)
OOUTFEAT Outstanding offender physical features (no = 0;
yes=1)
OACCENT Offender’s accent (no = 0; yes = 1)
OODOUR Noticeable offender odour (no = 0; yes = 1)
General offender
behavior variables
ODRUGALC Offender showed evidence of drug/alcohol use
(no=0;yes=1)
OINTERST Offender visited interstate in past 10 years (no = 0;
yes=1)
OINTERNA Offender lived or visited internationally during past
10 years (no = 0; yes = 1)
OLIVEWTH Offender living alone (8) = 0; with others (1-7) = 1
OJOBTYPE Offender’s job type (unemployed = 0; employed = 1)
OLIFESTY Offender’s general lifestyle: noncriminal
(1,2,4,8,11-13) = 0; criminal: 3,5-7,9-10 = 1
OCRIMST Offender’s criminal status (non-offender = 0;
statutory release = 1)
OSEXHAB Offender’s sexual habits (heterosexual = 0;
homosexual or bisexual = 1)
OMENPROB Offender displayed symptoms or had been treated
for mental problems (no = 0; yes = 1)
Psychological Profiling of Serial Arson 651
(continued)
OPOSPROP Offender possessed other’s property (no = 0; yes = 1)
OCONFESS Offender admitted to other similar crimes of violence
(no=0;yes=1)
OVEHUSED Offender used a vehicle in this incident (no = 0;
yes = 1)
OVEHSTAT Offender’s vehicle status (owned = 0; not owned = 1)
OVEHTYPE Offender’s vehicle type (car = 0; van, jeep, or
truck = 1)
Event-specific
offender behavior
and choice
variables
THREAT Offender makes a threat to someone about
committing the arson (no = 0; yes = 1)
DISTMAJ Offender travels more than 1 km to the target (no = 0;
yes=1)
DISTMIN Offender travels less than 1 km to the target (no = 0;
yes=1)
ACCOMPLI Offender had accomplices in committing the arson
(no = 0; yes = 1)
VISIBLE Offender lit fire in highly visible location with
potential witnesses (no = 0; yes = 1)
PRESENT Offender was present at the crime scene watching
the fire (no = 0; yes = 1)
NOTPRES Offender was not present at the crime scene watching
the fire (no = 0; yes = 1)
ACALLS Offender reports the fire he actually started himself
(no=0;yes=1)
AEXTIN Offender is involved in attempts to extinguish the fire
he actually set (no = 0; yes = 1)
NIGHT Offender set the fire at night (no = 0; yes = 1)
DAY Offender set the fire during the day (no = 0; yes = 1)
WEEK Offender set the fire on a weekday (no = 0; yes = 1)
WEEKEND Offender set the fire on a weekend day (no = 0;
yes=1)
HOLIDAY Offender set the fire during some type of holiday
period (no = 0; yes = 1)
SUMSPRIG Offender set the fire during the summer or spring
seasons (no = 0; yes = 1)
WINAUTM Offender set the fire during the winter or autumn
seasons (no = 0; yes = 1)
652 International Journal of Offender Therapy and Comparative Criminology
APPENDIX (continued)
Variable
Variable Set Label Definition (Category Labels and Coding)
Crime scene
variables
SINGPOO Fire was lit from a single point of origin or location
(no=0;yes=1)
MULTIPOO Fire was lit from multiple points of origin or
locations (no = 0; yes = 1)
POOEXTER Point of origin of fire was a location exterior to the
target (no = 0; yes = 1)
POOINT Point of origin of fire was a location interior to the
target (no = 0; yes = 1)
MATERBRO Offender consciously brought materials to start the
fire to the target (no = 0; yes = 1)
ACCELERA An accelerant was employed by the offender to light
the fire (no = 0; yes = 1)
TRAILERS There was evidence of a trailer (detectable burn line
of liquid accelerant) used at the fire (no = 0;
yes = 1)
PLANNED There was evidence the arson was planned with a
specific intended target (no = 0; yes = 1)
RANDOM There was evidence that the arson was unplanned
or random (no = 0; yes = 1)
ENTARGET Offender actually entered the target to light the fire
(no=0;yes=1)
MAJFIRE The resulting fire caused major damage (no = 0;
yes=1)
MINFIRE The resulting fire caused minor damage (no = 0;
yes = 1)
SPECBURN Specific items were initially burned by the offender
to start the fire (no = 0; yes = 1)
ADAMAGE Additional damage, other than fire damage, was
caused by the offender, for example, vandalism
(no=0;yes=1)
THEFT Offender stole something from the target (no = 0;
yes=1)
EVIDENCE Physical evidence was left by the offender at the
crime scene (no = 0; yes = 1)
SEXACTIV There was evidence that the offender engaged in
some sexual activity at the crime scene (no = 0;
yes=1)
RESPROP The target was a residential property such as a house
or apartment (no = 0; yes = 1)
Psychological Profiling of Serial Arson 653
APPENDIX (continued)
Variable
Variable Set Label Definition (Category Labels and Coding)
(continued)
COMPROP The target was a commercial property such as a
business, used for work, not living (no = 0; yes = 1)
EDUPROP The target was an educational facility such as a
school (no = 0; yes = 1)
STATPROP The target was a state-owned property such as a
government building or police station (no = 0;
yes = 1)
VEHPROP The target was a motor vehicle such as a car,
motorcycle, or truck (no = 0; yes = 1)
MINPROP The target was a minor item such as a rubbish bin,
letterbox, or abandoned property (no = 0; yes = 1)
BUSPROP The target was a bushland or forest, possibly
including property, fences, and hedges (no = 0;
yes=1)
TOCCUPY The target was occupied by people at the time of the
fire (no = 0; yes = 1)
TUNOCCUP The target was not occupied by people at the time of
the fire (no = 0; yes = 1)
TRELATIO The offender had some relationship with the target
such as their school or workplace (no = 0; yes = 1)
TUNRELAT The offender had no relationship with the target
TSECURTY The target had some form of security system, fire
alarm, sprinkler systems, and so forth (no = 0;
yes = 1)
ACKNOWLEDGEMENTS
We acknowledge with gratitude Commissioner Brett Hume and the Fire Inves
-
tigation and Research Unit NSW Fire Brigade in providing funds and support to
make this research possible. In addition, the first author wishes to personally
thank Inspector Terry Campbell and Superintendent Peter Dein, NSW Police Ser
-
vice, and all contributing officers within the respective police and fire brigade
agencies throughout Australia who contributed to this research.
REFERENCES
Ainsworth, P. B. (2001). Offender profiling and crime analysis. Devon, UK: Willan.
654 International Journal of Offender Therapy and Comparative Criminology
APPENDIX (continued)
Variable
Variable Set Label Definition (Category Labels and Coding)
Barnett, W., & Spitzer, M. (1994). Pathological fire-setting1951-1991: A review. Medicine, Science &
Law, 34, 4-20.
Bennett, W.M.,&Hess,K.M.(1984).Thearsonist:A closer look. Springfield, IL: Charles C Thomas.
Britton, P. (1997). The jigsaw man. London: Bantam.
Burns, L. (1991). School arson in New South Wales. Sydney: Government Printer.
Canter, D., & Fritzon, K. (1998). Differentiating arsonists: A model of firesetting actions and charac
-
teristics. Legal and Criminological Psychology, 3, 73-96.
Canter, D., & Heritage, R. (1989). A multivariatemodelof sexualoffence behaviour: Developments in
offender profiling: Vol. 1. Journal of Forensic Psychiatry, 1, 185-212.
Coxon, A. P. M. (1982). The user’s guide to multidimensional scaling. London: Heinemann Educa
-
tional Books.
Dietz, P. E. (1985). Sex offender profiling by the FBI: A preliminary conceptual model. In M. H. Ben-
Aron, S. J. Hucher, & C. D. Webster (Eds.), Clinical criminology (pp. 207-219). Toronto: M & M
Graphics.
Douglas, J. E., Burgess, A. W., Burgess, A. G., & Ressler, R. K. (1992). Crime classification manual.
New York: Simon & Schuster.
Douglas, J. E., Ressler, R. K., Burgess, A. W., & Hartman, C. R. (1986). Criminal profilingfrom crime
scene analysis. Behavioral Sciences & the Law, 4, 401-421.
Geberth, V. J. (1996). Practical homicide investigation (2nd ed.). New York: CRC Press.
Geller, J. L. (1992). Pathological firesettingin adults. International Journal of Law and Psychiatry, 15,
283-302.
Godwin, M. G. (2000). Hunting serial predators. New York: CRC Press.
Hair, J. F., Jr., Anderson, R. E., Tatham, R. L., & Black, W. G. (1998). Multivariate data analysis (5th
ed.). Englewood Cliffs, NJ: Prentice Hall.
Harris, G. T., & Rice, M. E. (1996). A typology of mentally disordered firesetters. Journal of Interper-
sonal Violence, 11, 351-363.
Holmes, R. M., & Holmes, S. T. (1996). Profiling violent crimes: An investigative tool (2nd ed.). Thou-
sand Oaks, CA: Sage.
Holmes, R. M., & Holmes, S. T. (1998). Contemporary perspectivesonserialmurder. Thousand Oaks,
CA: Sage.
Icove, D. J., & Estepp, M. H. (1987). Motive-based offender profiles of arson and fire-related crimes.
FBI Law Enforcement Bulletin, 56(4), 17-23.
Jeffers, H. P. (1992). Profiles in evil. London: Warner Bros.
Kocsis, R. N. (1999). Criminal profiling of crime scene behaviors in Australian sexual murders. Aus
-
tralian Police Journal, 53, 113-116.
Kocsis, R. N. (in press). Criminal psychological profiling: Validities and Abilities? International
Journal of Offender Therapy & Comparative Criminology.
Kocsis, R. N., Cooksey, R. W., & Irwin, H. J. (2002). Psychological profiling of sexual murders: An
empirical model. International Journal of Offender Therapy and Comparative Criminology, 46,
532-553.
Kocsis, R. N., Irwin, H. J., & Cooksey, R. W. (2002). Psychological profiling of offender characteris
-
tics from crime behaviors in serial rapists. International Journal of Offender Therapy and Com
-
parative Criminology, 46, 144-169.
Kocsis, R. N., Hayes, A. F., & Irwin, H. J. (2002). Investigativeexperienceand accuracy in psycholog
-
ical profiling of a violent crime. Journal of Interpersonal Violence, 17, 811-823.
Kocsis, R. N., & Irwin, H. J. (1998). The psychological profile of serial offenders and a redefinition of
the misnomer of serial crime. Psychiatry, Psychology, and Law, 5, 197-213.
Kocsis, R. N., Irwin, H. J., & Hayes, A. F. (1998). Organised and disorganised behavior syndromes in
arsonists: A validation study of a psychological profiling concept. Psychiatry, Psychology and
Law, 5, 117-130.
Kocsis, R. N., Irwin, H. J., Hayes, A. F,. & Nunn, R. (2000). Expertise in psychological profiling: A
comparative assessment. Journal of Interpersonal Violence, 15, 311-331.
Psychological Profiling of Serial Arson 655
Kruskal, J. B., & Wish, M. (1978). Multidimensional scaling. Beverly Hills, CA: Sage.
Lewis, N. D. C., & Yarnell, H. (1952). Pathological firesetting (pyromania) (Nervous and Mental Dis
-
ease Monographs No. 82). New York: Coolidge Foundation.
Lewis, N. D. C., & Yarnell, H. (1952). Pathological fire-setting (pyromania) (Nervous and Mental
Disease Monograph No. 82). New York: Coolidge Foundation.
Oleson, J. C. (1996). Psychological profiling: Does it actually work? Forensic Update, 46, 11-14.
Orr, J. (1989a, January). Profilesinarson:Theserialfiresetter. American Fire Journal, 24-27, 32-33.
Orr, J. (1989b, May). Profilesinarson:Therevenge firesetter. American Fire Journal,30-32,37-38.
Prins, H., Tennent, G., & Trick, K. (1985). Motives for arson (fire raising). Medicine, Science & Law,
25(4), 275-278.
Ressler, R. K., Burgess, A. W., & Douglas, J. E. (1988). Sexual homicide: Patterns and motives.
Lexington, MA: Lexington Books.
Rider, A. O. (1980a). The firesetter: A psychological profile. Part 1. FBI Law Enforcement Bulletin,
49(6), 6-13.
Rider, A. O. (1980b). The firesetter: A psychological profile. Part 2. FBI Law Enforcement Bulletin,
49(7), 12-17.
Rix, K. J. B. (1994). A psychiatric study of adult arsonists. Medicine, Science & the Law, 34, 21-34.
Rossmo, D. K. (1997). Geographic profiling. In J. L. Jackson & D. A. Berkerian (Eds.), Offender pro
-
filing: Theory, research and practice (pp. 159-175). New York: John Wiley.
Vorpagel, R. E. (1982). Painting psychological profiles: Charlatanism, coincidence, charisma or new
science. Police Chief, 3(8), 156-159.
Wilson, P. R., Lincoln, R., & Kocsis, R. N. (1997). Validity, utility and ethics of profiling for serial vio-
lent and sexual offenders. Psychiatry, Psychology and Law, 4, 1-12.
Richard N. Kocsis, Ph.D.
Forensic Psychologist, Private Practice
16 Lynden Ave.
Carlingford, Sydney NSW 2118
Australia
Ray W. Cooksey, Ph.D.
School of Marketing and Management
University of New England
Armidale NSW 2351
Australia
656 International Journal of Offender Therapy and Comparative Criminology
