Content uploaded by Clive R Hollin
Author content
All content in this area was uploaded by Clive R Hollin on Dec 04, 2015
Content may be subject to copyright.
Part II
New Techniques
and Applications
Chapter 6
Case Linkage
Identifying Crimes Committed by the Same
Offender
Jessica Woodhams, Ray Bull, and Clive R. Hollin
Summary
This chapter begins by explaining the purposes of linking crimes committed by the same
offender and what case linkage can add to a police investigation and prosecution. The various
steps involved in the process of case linkage are explained. The assumptions of behavioral
consistency and inter-individual behavioral variation, which case linkage rests on, are outlined,
and the research that has begun to test these assumptions is reported. The effect of poor-quality
data on the case linkage process and on empirical research is examined. Current methods and
future developments for overcoming this difficulty are described. The obstacles to identifying
linked crimes across police boundaries are discussed. Case linkage research and practice are
compared with various criteria for expert evidence with promising results. The chapter closes by
considering future avenues for research and practice in case linkage.
INTRODUCTION
If a police officer is investigating a rape and the perpetrator of the rape has
committed other sexual crimes, there are several reasons why it would be advan-
tageous for all of the perpetrators’ crimes to be investigated together. First, this
would enable the police force(s) to use their limited resources more efficiently.
Investigative efforts can be combined rather than the crimes being investigated
in parallel, which would result in the duplication of work. Knowledge about
From: Criminal Profiling: International Theory, Research, and Practice
Edited by: R. N. Kocsis © Humana Press Inc., Totowa, NJ
117
118 J. Woodhams et al.
the perpetrator or evidence against the perpetrator gathered from each crime
scene can be combined (1), which can potentially result in the more rapid
apprehension of the suspect or can potentially strengthen the case against them.
Identifying crimes committed by the same perpetrator can be relatively
straightforward if the victims know the identity of their attacker. However, if the
suspect is a stranger to the victim, then identification can be less straightforward.
Physical evidence, such as DNA, can be used to link crimes together committed
by an unknown suspect, but in a number of crimes, there is no physical evidence
to identify the offender (2). It is in such situations that case linkage can be of use.
Case linkage is a process that aims to identify crimes that are likely to
have been committed by the same suspect because of the behavioral similarity
across the crimes. Crimes committed in a similar manner are “linked” to form
a “series.” In other words, the crimes are linked together because the offender
has behaved in the crimes in a very similar way, and therefore, it is probable
that the same offender has committed all of these crimes. Evidence that an
offender is likely to have committed a group, or series, of crimes is not only
useful for investigative purposes, as outlined above, but can also be presented
as similar fact evidence in legal proceedings (3,4).
Case linkage is typically conducted by crime analysts or police officers.
It is sometimes called “comparative case analysis” (5) or “linkage analysis” (3)
and has been described as a type of behavioral analysis (2). It has most often
been used with crimes such as stranger rape and murder. However, as will
become clear later in this chapter, it can be, and is, used with volume crime,
such as burglary and robbery.
Previous writers have considered the linking of crimes to be a type of
offender profiling, and indeed, expert profilers are asked to link crimes (6). The
two approaches do share some common features: for example, both are most
often used for crimes committed by unknown offenders. Criminal profiling
and case linkage also share the assumption that offenders are consistent in
the way that they behave across their crime series. This assumption has been
termed the offender consistency hypothesis (7). However, although the two
approaches share common features, it is important to recognize their differences.
Profiling makes predictions about a person’s (demographic) characteristics from
their crime scene behavior. Hence, it requires a relationship between behavior
and a person’s (demographic) characteristics. It follows that offenders who
share similar criminal behaviors should therefore share similar demographic
characteristics. This assumption has been termed the homology assumption
(8,9). Case linkage does not, however, make this assumption. The assumptions
underlying case linkage are outlined later in this chapter. The next section
outlines the process itself.
Case Linkage 119
THE PROCESS OF CASE LINKAGE
There are typically two different scenarios in which a crime analyst would
seek to link crimes. An analyst can proactively search for linked crimes among
a database of crimes. Alternatively, they can engage in a reactive search having
been presented with an index offense (for which the offender might already have
been identified) with a request to identify other crimes potentially committed
by the same offender. Depending on the purposes for which an analyst has been
consulted, the processes involved in the linking of crimes can vary slightly.
However, the same general steps will be followed, and these are illustrated in
Figure 1.
First, the crime analyst must collate all the relevant information about the
crime(s) in which they are initially interested. Typically, the victim’s account
of the crime is the primary source of information. Clearly, were the victim
killed during the commission of the crime, this will not be available.
Victim accounts can be in two forms. The first is a victim statement. This
is typically a chronological account of the crime that is written collaboratively
with a police officer (10). Alternatively, the crime analyst may have access to
the written transcript of the victim’s interview with the police (11). In such
circumstances, the victim is often asked to freely recall the event and is then
questioned further on this. A transcript of an interview is not therefore a chrono-
logical account of the event and is likely to contain quite a bit of repetition. As
well as consulting the victim’s account, the crime analyst may consult other
records, such as the suspect’s account (if apprehended) and medical examination
Request and read documentation
Construct list of behaviors for index offence
Search for similar cases
Construct lists of behaviors for other
potentially similar offences
Identify similar and dissimilar behaviors
Write a report for prosecutor/police
Weight similarities and differences
Fig. 1. The process of linking crimes.
120 J. Woodhams et al.
reports, or they may choose to visit the crime scene itself. With murder cases,
the analyst can consult materials such as crime scene photographs or sketches,
the autopsy reports, and toxicology reports (3).
Having collated all relevant information, the crime analyst must compose
a list of the behaviors exhibited by the perpetrator. Some behaviors might be
more spontaneous, whereas others may be produced as a reaction to the victim
or witnesses. In some jurisdictions, as well as creating a list of behaviors,
the analyst classifies suspect behaviors as “modus operandi” behaviors or
“ritualistic” behaviors (3). In these circumstances, the term “modus operandi”
is used to refer to behaviors that are necessary for the offender to commit
the crime, whereas ritualistic behaviors are not and are fantasy-based. Alison
et al. (12) explain that modus operandi behaviors are “functionally significant”
and dependent on context. Whereas, what Alison et al. (12) term “signature”
behaviors are “psychologically significant” and are not dependent on context.
These are behaviors that seem similar to Hazelwood and Warren’s (3) ritualistic
behaviors.
There is a difficulty in referring to ritualistic behaviors as fantasy-based,
as fantasy-based behaviors will be more relevant to some crimes, such as sexual
crime, and perhaps less relevant to property crimes on which case linkage is
still conducted. Terms such as “psychologically significant” and “functionally
significant” may be more helpful. However, there are some inherent problems
with categorizing behaviors in this way.
First, it seems unlikely that psychologically relevant behaviors are truly
context independent. The psychological meaning of a situation is influential
in determining the behavior that is displayed (13), and during a crime, the
psychological meaning of a situation could change for an offender depending
on a number of factors. For example, Davies (14) describes a serial rapist whose
behavior toward his victims appeared to vary depending on his perception of
their status. The offender was complimentary of a middle-class young female
victim, but he physically and verbally abused his other victims, most of whom
were older and appeared less affluent.
Second, as recognized by both Hazelwood and Warren (3) and Alison et al.
(12), categorizing a behavior as modus operandi or ritual/signature requires a
subjective decision on the part of the analyst as to the psychological meaning
of a behavior. Both sets of authors comment on the difficulty of determining
whether a behavior constitutes modus operandi or ritual. As an example, the age
and sex of the victim in a sexual crime are noted as a modus operandi behavior
by Hazelwood and Warren (3), and yet, these could quite clearly be related
to an offender’s sexual fantasy. Hazelwood and Warren (3) also comment
that a behavior could be both modus operandi and ritual. The categorization
Case Linkage 121
of offender behaviors into modus operandi or ritual/signature therefore has a
number of associated difficulties and is perhaps an unnecessary optional step
in the case linkage process.
Once a list of behaviors has been created for the crime(s) in question,
the analyst’s next task is to search for crimes where similar behaviors were
displayed. When potentially similar crimes are identified, the analyst must
again collate information about the perpetrator’s behavior and create a list of
behaviors engaged in for each offense. The analyst can then consider similar-
ities and differences in behavior between the offenses. The context in which a
behavior occurs is also considered because this can alter the apparent similarity or
difference in behavior between crimes. A behavior that initially appears different
might be explained by situational influences, such as victim behavior or third
party disturbance. For example, apparent differences in the use of physical
violence by a rapist could be explained by variations in the victims’ resistance.
The analyst must in this case consider the context in which violence occurred.
Having identified similarities and differences between crimes, an important
stage in this process is to consider the base rates for such behaviors. Two crimes
might share a similar behavior, but if the behavior commonly occurs within
the particular class of offense (e.g., rape and robbery), this would not strongly
suggest that the same offender committed the two crimes. In some jurisdic-
tions, databases of behaviors are available to enable such weighting of similar-
ities and differences, whereas in others, the analyst has to rely on their own
expertise with a crime type or on the combined expertise of the analyst team.
Hazelwood and Warren (3) refer to the analyst identifying the signature of an
offender or the unique combination of behaviors they have engaged in. Although
different terminology is being used in the literature, the same point is essen-
tially being made: that this process does not only involve the identification of
shared behaviors but that the rarity of such behaviors must also be considered.
Having considered the similarities and differences in behavior between
crimes, and the base rates of behaviors, the crime analyst must finally decide
whether, in their opinion, it is probable that the same offender committed
the crimes analyzed. A crime analyst would not expect perfect consistency in
behavior for the reasons outlined above. This point raises the question of how
similar two crimes must be before the analyst should give their opinion that
it is probable that the same offender committed both crimes. As recognized
by Bennell and Canter (5), there are financial and human costs associated
with setting the criteria for linking crimes either too low or too high. These
issues will also vary depending on whether the analyst’s opinion is to be used
for investigative purposes or in legal proceedings. Little research has thus far
considered this issue; hence, Bennell and Jones (15) have called for researchers
122 J. Woodhams et al.
to begin investigating this topic. Recommendations for practice are therefore
not possible at this stage until more research is completed; however, interested
readers are referred to Bennell and Canter (5), Bennell and Jones (15), and
Alison et al. (12) for further information.
The final stage of the process involves the analyst producing a written
report for their client in which they can draw their client’s attention to similar
crimes were any identified. Their clients include police officers, crime analysts
from other jurisdictions, and prosecutors. As well as providing a written report,
analysts can also be asked to give formal verbal presentations of their findings
to the client.
Case linkage requires the crime analyst to process a substantial amount of
information, and this can put considerable cognitive load on the analyst (16).
Some policing organizations have recognized this difficulty and have developed
databases, which the analyst can search for similar offenses (17). Without such
resources, the analyst’s memory of similar crimes would have to be relied on,
which is clearly undesirable. As well as using such databases for searching,
some efforts have been made in automating part of the linking process by
computerizing the actual comparison of offense behaviors between crimes to
produce a measure of similarity for each pair of crimes in the database. The
crime analyst can prioritize pairs of crimes with a high similarity score for
further analysis. The first author has been working with a UK police force
in developing such a system for linking robberies. Such automation does not
currently have the capacity to consider the context in which a behavior occurred,
and for such reasons, it is unlikely that the process of case linkage could be
fully automated.
Such developments in the use of technology have increased the efficiency
and potentially the accuracy of the linking process. However, the creation and
maintenance of large databases requires considerable input. Information has to be
collated for each crime and entered onto the database. Quality assurance proce-
dures are also required to ensure the accuracy of the database. These processes
are time consuming, and therefore, the question has been raised as to whether
data collection and entry can be focused on a smaller number of perpetrator
behaviors. Research is outlined in the section entitled “Evaluating Case Linkage”,
which is beginning to suggest that this step may be possible in the future.
THE THEORY OF CASE LINKAGE
The use of case linkage in advising and directing police investigations
and its potential use as similar fact evidence in legal proceedings requires
that it has a sound theoretical basis. The process of linking crimes rests on
Case Linkage 123
two key assumptions. The first assumption is that criminals are consistent in
the way they behave across their crimes. In psychology, the assumption that
people show consistent behavior across different situations is termed cross-
situational consistency (18). However, case linkage focuses on the similarity of
an individual’s criminal behavior within a crime type (e.g., within robberies or
within sexual offenses). This is a special type of cross-situational consistency,
termed the offender consistency hypothesis (7).
Thesecondassumptionisthatthereisvariationinthewaydifferentcriminals
commit crimes. For it to be possible to link crimes committed by the same
offender, criminals must show consistent but distinctive behavior. If offenders
were consistent in the way they commit crimes but committed crimes in the
same way, then it would be impossible to differentiate the crimes of one offender
from those of another. Thus, for case linkage to work, criminals must behave
in a stable but distinctive manner. Whether these two assumptions are valid
has been the focus of research attention by forensic psychologists in Europe,
the United States, and Canada. This research is reviewed in the next section.
EVALUATING CASE LINKAGE
Much of the research interest in whether offenders are consistent in their
offending behavior has focused on the more serious types of crime, such as
sexual assault (1,19–21) and murder (22). However, studies have also been
conducted for arson (23), commercial and residential burglaries (5,15,24), and
commercial robbery (25). Although these studies have used different statistical
methods, they all have reported a degree of consistency in offenders’ behaviors.
Analysts typically consider similarity in individual behaviors across
crimes. However, some research has considered whether it is possible to link
crimes at the thematic level. For example, in Salfati and Bateman’s (22) study,
the themes instrumental and expressive are used to describe types of homicide.
Although a degree of consistency was demonstrated in themes, in the real
world, such a dichotomy is unlikely to be sufficiently discriminating for either
criminal intelligence or prosecution purposes.
Some of these studies of offender consistency have gone further and
assessed the two assumptions of case linkage simultaneously. They have inves-
tigated whether crimes committed by the same offender can be differentiated
from crimes by different offenders (5,15,25). Other studies have identified for
each crime in their sample the 10 most similar crimes. They have then assessed
whether any of the crimes in this selection were in fact committed by the
same offender (1,21,23). All such studies have confirmed that it is possible to
link crimes and have therefore supported the two assumptions of behavioral
124 J. Woodhams et al.
consistency and inter-perpetrator behavioral variation. However, the results also
indicate that this process is not perfect and that linkage accuracy appears to
vary with crime type. The variation in methods used by researchers can make it
difficult to draw comparisons between studies; however, Santtila et al. (21,23)
used the same methodology making a comparison between arson and sexual
offenses possible. Santtila et al. (23) found a “linked” arson to be in the 10
most similar offenses in approximately only 50% of arsons. Santtila et al. (21)
found greater linkage accuracy with sexual crimes for which approximately
60% of the time a crime from the same series was found within the 10 most
similar offenses.
Comparisons can also be drawn between the studies of Bennell and Canter
(5), Bennell and Jones (15), and Woodhams and Toye (25), which have all used
similar methodologies. Measures of predictive accuracy, called areas under the
curve (AUCs), were calculated in all three studies using receiver operating
characteristic (ROC) analysis. The AUC indicates how well linked and unlinked
pairs of crimes were identified. An AUC of 0.50 indicates chance level and an
AUC of 1.0 indicates perfect discrimination (26); thus, a larger AUC represents
higher predictive accuracy. In Bennell and Canter’s (5) and Bennell and Jones’
(15) studies of burglary, the AUCs ranged from 0.63 to 0.81 and 0.52 to 0.94,
respectively. In their study of commercial robbery, Woodhams and Toye (25)
report AUCs ranging from 0.70 to 0.95.
Variations in performance between studies could be due to method-
ological differences or could reflect how amenable certain crime types are to
case linkage. Consistency might actually be less for some crimes than others,
although this has not yet been investigated. The ranges for accuracy reported
within the studies of Bennell and colleagues (5,15) and Woodhams and Toye
(25) reflect the use of different behaviors as predictors of linkage. These studies
provide preliminary evidence that offenders show greater consistency in some
behaviors used in committing a burglary or robbery than in other behaviors.
Greater consistency has been observed in behaviors that are more inherent
to the offender and are less influenced by situational factors. For example,
the property stolen in a robbery or burglary can be highly dependent on the
situation, whereas the offender has greater control over which addresses he or
she chooses to target and how he or she will seek to control any witnesses.
If offenders show greater consistency and distinctiveness in some
behaviors compared with others, such findings have implications for the
collation and entry of information onto crime databases. It would be more time
efficient to focus on the collation and entry of behaviors that are more reliable
indicators of linkage. These findings also suggest that crime analysts should
focus their attention on such behaviors when considering similarities between
Case Linkage 125
offenses. The research is not yet at a stage where clear recommendations can be
made for practice; however, it does suggest this will be possible in the future.
Most of the studies of case linkage have focused on testing the underlying
assumptions of the process and assessing whether it is possible to link crimes on
the basis of behavioral similarity. Only one study (16) thus far appears to have
considered how case linkage is conducted in practice. The methods of linking
car crimes used by (i) experienced car-crime investigators, (ii) investigators of
other types of crime, (iii) inexperienced investigators, and (iv) laypersons were
compared using the same 10 solved series of car crimes. Participants’ accuracy
was assessed, but they were also asked to articulate their method for linking
during the process itself and afterwards. As would be expected, experience
increased linking accuracy. The accurate linking of car crimes was more often
associated with variables such as the type of vehicle chosen and the time and
place the crime occurred, whereas the property stolen from the car was a poor
predictor of linkage. It appears that the behaviors associated with accurate
linking are those that are more under the offender’s control, whereas poor
predictors, such as the property stolen, are more situation-dependent. These
findings mirror those reported in statistical analyses of case linkage and again
suggest that more accurate and efficient linking could be achieved if certain
predictors over others are focused on. The identification of accurate predictors
of linkage is an important research goal for the future.
OBSTACLES TO LINKING CRIMES
Data Limitations
One obstacle to the accurate linking of crimes is the type of data crime
analysts must presently work with. As explained in the section entitled “The
Process of Case Linkage”, it is the victim’s account of the crime that is most
often used in determining how the offense was committed. It is important to
remember that this account is a secondary record of the offender’s actions.
It is rare to possess a primary record of an offense, such as a closed circuit
television recording. However, even where such primary records exist, they do
not constitute a complete record of an event: for example, the offender’s verbal
behavior may be missing. It is therefore likely that crime analysts will always
have to rely mainly on the secondary account of the victim.
The accuracy and completeness of a victim’s account are likely to be
imperfect for a number of reasons. The victim may be traumatized by the event
or they may have a poor memory of the event or part of the event (e.g., in the
case of drug-assisted sexual assaults). When the victim has a good memory of
the event and is able to accurately articulate what occurred, errors or omissions
126 J. Woodhams et al.
can still occur at the interview or statement-writing stage. The statement may
be taken or the victim interview may occur some considerable time after the
offense occurred (12). The collaborative nature of statement writing and the
selective nature of investigators’ questioning can result in information currently
perceived by them as irrelevant for prosecution being ignored and/or omitted
(27). Distortion of what occurred is also possible, in that certain features may
be suppressed with others being exaggerated (28).
Although they are always likely to contain some omissions and errors,
victim interviews are arguably a more accurate record of the victim’s account
than victim statements because they are a real-time record of the event in the
victim’s own words. In England and Wales, (tape-recorded) victim interviews
are beginning to replace victim statements with the implementation of the
Youth Justice and Criminal Evidence Act 1999. This change means that the
potential for the introduction of error into the victim’s account will be less.
One way for researchers (and crime analysts using automated linkage
systems) to deal with potential errors in victims’ accounts is through choosing an
appropriate measure of similarity for linking. Jaccard’s coefficient is a measure
of similarity that does not include joint non-occurrences in its calculation. In
other words, if a particular behavior did not occur in two crimes, this would
not increase their similarity. This point is advantageous if we consider that
a behavior may have occurred but that, for the reasons outlined above, its
occurrence has not been recorded in the victim’s account. This advantage has
been noted by Bennell and Canter (5) although it has also been recognized that
Jaccard’s coefficient has some disadvantages (15).
Because police records are unlikely to be complete records of the event
for the reasons outlined above, Alison et al. (12) have warned against using
individual behaviors for linkage. Instead, they recommend that crime analysts
and profilers use geographical proximity to link crimes because this is more
likely to be accurately recorded (12,28). This recommendation appears to be
partly based on the findings of Bennell and Canter (5) and Bennell and Jones
(15) that, for residential and commercial burglaries, inter-crime distance was the
most accurate single-feature predictor of linkage. However, as outlined in the
section entitled “Evaluating Case Linkage”, some studies have demonstrated
the ability to link crimes using other behaviors, such as those behaviors used to
control the victim (25). It is perhaps too early in the research process to make
such recommendations.
Researchers of case linkage typically begin their data analysis by devel-
oping a behavioral checklist from a content analysis of the offenses in their
sample. This checklist would capture all of the behaviors in the sample. Each
offense is compared against the checklist and the absence or presence of each
Case Linkage 127
behavior has typically been recorded. Poor-quality data can cause difficulties
in this procedure. If a checklist contains the behavior “vaginal penetration,”
a difficulty arises should the victim’s account solely state that “penetration
occurred.” In such a scenario, the researcher is unsure as to the nature of
the penetration, and this information is lost from the analysis. Where police
forces are moving toward computerized databases of crimes, this issue can also
be problematic, and potential links between crimes can be missed. The first
author is currently working with colleagues to devise a method of categorizing
offending behaviors that could overcome this very real difficulty for researchers
and practitioners alike.
Geographical Obstacles
A further obstacle to the linking of crimes relates to the geography of an
individual’s offending. Crime analysts often work for a specific police force
that only covers its own geographical area. However, criminals do not keep
within police borders when offending, they will cross borders and offend in
other police jurisdictions. A force analyst searching for similar crimes within
their own force’s databases can therefore fail to identify linked crimes that
occur in neighboring jurisdictions. For this reason, units conducting analysis at a
national level have been established, such as the Serious Crime Analysis Section
in England. The sharing of intelligence and good communication between
different force analysis units can also help overcome this obstacle.
Obstacles in the Courtroom
Although case linkage can be used to guide police investigations, it
has experienced some obstacles in its acceptance in the courts (3). Although
expert evidence on the similarity between crimes has been admitted into legal
proceedings in the United States (3), some limits have been put on its admittance.
For example, in the case of State of New Jersey v. Fortin, Robert Hazelwood
was not allowed to present his expert opinion as to whether the two crimes were
committed by the same offender although he was allowed to testify as to the
similarity between the two crimes (3,29). The grounds for this decision were (i)
that Hazelwood’s linkage analysis did not have sufficient scientific reliability,
(ii) that few people other than Hazelwood’s close associates practiced linkage
analysis, and (iii) that it had not received peer scrutiny.
Criteria for the admissibility of expert evidence are not as clear in the
United Kingdom and other jurisdictions as they are in the United States, where
expert evidence about a novel technique must meet the Daubert criteria (30).
These criteria and associated guidelines produced by the Supreme Court of the
United States have been outlined (31).
128 J. Woodhams et al.
Vrij (31) explains that the first question to be considered when evaluating
whether a novel technique will meet the Daubert criteria is whether the evidence
has a scientific hypothesis that is testable. The hypotheses underlying case
linkage are that criminals are both consistent and distinctive in their behavior.
Scientific studies can be conducted using solved crimes to determine whether
the offenders known to be responsible for the offenses behaved in a consistent
but distinctive manner. It is therefore suggested that the answer to this first
question is affirmative.
The second question is whether this proposition of consistency and
distinction has been tested. The answer to this question is suggested to
be partially affirmative. The hypotheses that offenders are consistent and
distinctive have been tested with some crimes. Three studies have suggested that
offenders are consistent and distinctive enough for linked and unlinked crimes
to be accurately differentiated (5,15,25). Furthermore, five studies have demon-
strated that crimes can be associated with other crimes in their series, which
would require both offender consistency and distinctiveness (1,19,21,23,24).
Such research has ecological validity; therefore its findings should be relevant
to practice. However, it is important to recognize that the samples used in this
research consist of solved cases, whereas case linkage, in practice, is used with
unsolved cases. As noted by Bennell and Canter (5), such cases might actually
have been solved because of their greater behavioral similarity. This issue is
problematic and it would be difficult to overcome this limitation; however, one
solution would be to conduct research with samples of unsolved crimes that
had been linked through DNA testing. To strengthen further the affirmation
to this question, cross-validation studies would be required as well as studies
assessing other types of crime.
The third question is “Is there a known error rate?” It is suggested that
the answer to this question would also be partially affirmative. Studies such
as those of Bennell and colleagues (5,15) and Woodhams and Toye (25) have
used logistic regression and ROC analyses, which enable overall estimates of
error to be calculated on samples of solved cases. These studies have indicated
that predictive accuracy rates can be as high as 90% or represented by an AUC
in ROC analysis of 0.95 with an area of just 0.05 for error. Although some
crime analysts do rely on statistical analyses to aid them in making decisions as
to whether crimes are linked, this is not always the case. Crimes would also not
be linked purely on the outcome of a statistical analysis. A crime analyst would
be involved in making this final decision, potentially in light of additional
information. How the subsequent input of an analyst would affect accuracy and
how accurate analysts are at linking crimes without computational aid require
testing.
Case Linkage 129
There are no known field studies of the accuracy of case linkage. Thus
far, the closest to this would be Santtila et al.’s (16) study of the linking
of car crimes; therefore this is clearly an area for future research. However,
establishing real-world error rates will always be problematic. To determine
whether the decision to link a pair of crimes was correct or incorrect, the
perpetrator has to have been convicted for both crimes or DNA evidence of
linkage would be required. First, it should be noted that conviction cannot
be considered a perfect indicator of identity as miscarriages of justice do
occur. Second, were conviction considered a reliable indicator of identity, the
conviction rates for some crime types for which case linkage is most commonly
used are notoriously low, namely sexual crimes (32,33). Third, with regard to
DNA evidence, as noted above, often no such evidence exists (2). It is therefore
highly probable that in the real world, analysts will make predictions about
linkage, yet whether these decisions are correct or incorrect will be impossible
to establish.
The fourth question of “Has the hypothesis and/or technique been
subjected to peer review and publication?” would also have an affirmative
answer. However, it is suggested that the answer to the fifth question, “Is the
theory on which the hypothesis and/or technique is based generally accepted
in the appropriate scientific community?,” would be “not yet.” It is suggested
that case linkage has not yet received sufficient academic interest; therefore the
answer to this question lies in the future. In summary, case linkage evidence
does not yet appear to meet all criteria set out in the Daubert ruling.
In England and Wales, expert scientific evidence is not required to meet the
full Daubert criteria. Instead, in recent cases, it has been required that evidence
has general acceptance in the scientific community. In other cases, evidence
based on novel techniques has been admitted but accompanied by a warning
from the judge as to how such evidence should be considered by the jury (30).
Ormerod and Sturman (30) have specifically considered the likelihood
that case linkage evidence (or as they term it comparative crime scene analysis
evidence) would be accepted as expert evidence. They conclude that it is quite
possible that evidence of behavioral similarity and distinctiveness would be
accepted by the courts. They explain that the purpose of such evidence is to
prove similarity rather than proving a suspect’s guilt. Therefore, if the method
used was considered reliable, if the evidence was unlikely to distract the court
(i.e., the jury) unnecessarily, and if it was not rejected for being prejudicial,
such evidence could be considered legally relevant. However, they caution
that even if such evidence were considered relevant and admissible, it may
be rejected for other reasons. These reasons could include if the witness were
not considered suitably qualified to be granted expert status, if the evidence
130 J. Woodhams et al.
were considered within the experience and knowledge of a layperson, and if
the evidence was considered unreliable.
The second of these points, that is, whether the experience and knowledge
of a crime analyst about similarities and differences in criminal behavior is
outside that of the layperson, warrants discussion. It could be argued that very
clear similarities in criminal behavior across crimes would be obvious to the
layperson and hence the evidence of a crime analyst would be unnecessary.
However, it is important to consider whether the layperson will know which
behaviors are actually rare or common in a population of crimes. Crime analysts
can have read thousands of crime reports and hundreds of victim statements.
Therefore, they will arguably be better informed as to which behaviors are
common or rare than the layperson whose knowledge of a type of offense may
be based merely on media portrayal or on their limited personal experience.
In addition, the crime analyst may have used more objective measures of base
rates in determining commonality or rarity through consulting crime databases,
which hold information on thousands of crimes, to which the layperson would
not have access.
In the sister field of criminal profiling, factions have arisen over the most
appropriate way to conduct profiling. This has been termed the clinical versus
the statistical debate. Ormerod and Sturman (30) explain that because profiling
varies with regard to what it actually is and how it is conducted, it may not
be considered admissible in legal proceedings. For case linkage evidence to be
accepted by courts, it is crucially important that it receives scientific research
attention and has a sound theoretical underpinning. Its acceptance as expert
evidence would also be aided by standardization in the way in which it is
conducted. As outlined above, although small differences seem to exist in the
process, from what has been published, the basic underlying steps appear the
same. However, there does appear to be some variation in its practice, as, for
example, in the use of statistical methods and computerization in linking.
Thus, in summary, there are obstacles to case linkage in relation to (i)
the data that the practitioners and researchers must rely on, (ii) overcoming
police boundaries, and (iii) the acceptance of case linkage evidence in courts.
There are a number of ways in which these issues can be addressed, as outlined
above, and these will no doubt be the focus of future empirical research.
THE FUTURE OF CASE LINKAGE
Theoretically and empirically, case linkage has started with a good
grounding. Research seems to be supporting its underlying assumptions. The
future for research in this area will include studies investigating the validity
Case Linkage 131
of these assumptions with previously untested crime types. In addition, cross-
validation studies are required before existing findings regarding criminal
consistency and distinctiveness can be fully accepted.
Empirical research is suggesting ways in which the process of linking
crimes could be improved. For example, it appears that some offender behaviors
are more reliable indicators of linkage than others. However, further research
on this is required before any firm recommendations can be made to practi-
tioners. As research continues into case linkage, it is likely that reliable statis-
tical methods for linking crimes will be developed. These methods have the
potential to reduce the cognitive load placed on the analyst and will be more
reliable and scientific. They may also encourage standardization in the way
case linkage is conducted. They will, however, require the development and
maintenance of large-scale databases of crimes. The high standards demon-
strated at some crime analysis units such as the Serious Crime Analysis
Section in England could be considered best practice for the future estab-
lishment of such databases. The development of such databases also holds
the potential for calculating base rates of behavior, providing the analyst
with a more reliable method of weighting behavioral similarities between
offenses.
A number of these potential future developments also have implications
for the acceptance of case linkage evidence in legal proceedings, because they
will encourage standardization and reliability. However, to remove the analyst
from the process of case linkage in the pursuit of standardization and reliability
would be an undesirable development. The findings of the linking process will
always need to be considered in light of other information uncovered during
the investigative process (25), thus it is likely that the professional expertise of
the crime analyst will always be required.
The published empirical research has clear implications for the conducting
of case linkage. However, the results of such research must be balanced
with the practical application of the findings in the real world so that evalu-
ative research will be needed to ensure that recommendations are workable
in practice. In the absence of a large pool of research on case linkage,
practitioners have been cautious in its application and have been mindful in
allowing researchers access to their data for their practice to be indepen-
dently researched and scrutinized. The cooperative and reciprocal relationship
that has thus far existed between academics and practitioners will no doubt
continue through joint enterprises. The interested researcher will certainly not
struggle to find a topic to pursue: they will be contributing to a novel field
with important practical implications for both the policing and prosecution of
crime.
132 J. Woodhams et al.
REFERENCES
1. Grubin, D., Kelly, P. & Brunsdon, C. (2001). Linking Serious Sexual Assaults
Through Behavior (No. HORS 215). London: Home Office Research Development
and Statistics Directorate.
2. Davies, A. (1991). The use of DNA profiling and behavioral science in the inves-
tigation of sexual offences. Med Sci Law, 31, 95–101.
3. Hazelwood, R.R. & Warren, J. (2003). Linkage analysis: modus operandi, ritual,
and signature in serial sexual crime. Aggress Violent Behav, 8, 587–598.
4. Ormerod, D. (1999). Criminal profiling: trial by judge and jury, not criminal
psychologist. In D. Canter & L.J. Alison (Eds.), Profiling in Policy and Practice
(pp. 207–261). Aldershot, UK: Ashgate Publishing.
5. Bennell, C. & Canter, D. (2002). Linking commercial burglaries by modus
operandi: tests using regression and ROC analysis. Sci Justice, 42, 1–12.
6. Copson, G. (1995). Coals to Newcastle? Part 1: A Study of Offender Profiling.
London: Home Office.
7. Canter, D. (1995). Psychology of offender profiling. In R. Bull & D. Carson (Eds.),
Handbook of Psychology in Legal Contexts. (pp. 343–355). Chichester, UK: John
Wiley and Sons.
8. Alison, L., Bennell, C., Mokros, A. & Ormerod, D. (2002). The personality paradox
in offender profiling: a theoretical review of the processes involved in deriving
background characteristics from crime scene actions. Psychol Pub Pol’y & Law,8,
115–135.
9. Mokros, A. & Alison, L.J. (2002). Is offender profiling possible? Testing the
predicted homology of crime scene actions and background characteristics in a
sample of rapists. Legal Criminol Psychol, 7, 25–43.
10. Rock, F. (2001). The genesis of a witness statement. Forensic Linguist, 8, 44–72.
11. Milne, R. & Bull, R. (1999). Investigative Interviewing: Psychology and Practice.
Chichester, UK: Wiley.
12. Alison, L.J., Goodwill, A. & Alison, E. (2005). Guidelines for profilers. In
L. Alison (Ed.), The Forensic Psychologists’ Casebook: Psychological Profiling
and Criminal Investigation (pp. 235–277). Devon, UK: Willan.
13. Shoda, Y., Mischel, W. & Wright, J. (1994). Intraindividual stability in the organ-
isation and patterning of behavior: incorporating psychological situations into the
idiographic analysis of personality. J Pers Soc Psychol, 67, 674–687.
14. Davies, A. (1992). Rapists’ behavior: a three aspect model as a basis for analysis
and the identification of serial crime. Forensic Sci Int, 55, 173–194.
15. Bennell, C. & Jones, N.J. (2005). Between a ROC and a hard place: a method for
linking serial burglaries by modus operandi.J Invest Psychol Off Prof, 2, 23–41.
16. Santtila, P., Korpela, S. & Hakkanen, H. (2004). Expertise and decision-making
in the linking of car crime series. Psychol, Crime Law, 10, 97–112.
17. Yokota, K. & Watanabe, S. (2002). Computer-based retrieval of suspects using
similarity of modus operandi.Int J Police Sci Manag, 4, 5–15.
18. Pervin, L.A. (2002). Current Controversies and Issues in Personality (3rd ed.).
New York: John Wiley and Sons.
Case Linkage 133
19. Canter, D., Heritage, R., Wilson, M., Davies, A., Kirby, S., Holden, R.,
McGinley, J., Hughes, H., Larkin, P., Martin, L., Tsang, E., Vaughan, G., and
Donald, I. (1991). A Facet Approach to Offender Profiling (Volume 1). Unpub-
lished manuscript.
20. Knight, R.A., Warren, J.I., Reboussin, R. & Soley, B.J. (1998). Predicting rapist
type from crime-scene variables. Crimin Justice Behav, 25, 46–80.
21. Santtila, P., Junkilla, J. & Sandnabba, N.K. (2005). Behavioral linking of stranger
rapes. J Invest Psychol Offender Profil, 2, 87–103.
22. Salfati, C.G. & Bateman, A.L. (2005). Serial homicide: an investigation
of behavioral consistency. Journal of Investigative Psychology and Offender
Profiling, 2, 121–144.
23. Santtila, P., Fritzon, K. & Tamelander, A.L. (2005). Linking arson incidents on
the basis of crime scene behavior. J Police Crim Psychol, 19, 1–16.
24. Green, E.J., Booth, C.E. & Biderman, M.D. (1976). Cluster analysis of burglary
MO’s. J Police Sci Admin, 4, 382–388.
25. Woodhams, J. & Toye, K. (2007). An empirical test of the assumptions of case
linkage and offender profiling with serial business robberies. Psychology, Public
Policy and Law, 13(1).
26. Swets, J.A. (1988). Measuring the accuracy of diagnostic systems. Science, 240,
1285–1293.
27. Canter, D. & Alison, L.J. (2003). Converting evidence into data: use of law
enforcement archives as unobtrusive measurement. Qual Rep, 8, 151–176.
28. Alison, L.J., Snook, B. & Stein, K. (2001). Unobtrusive measurement: using police
information for forensic research. Qual Res, 1, 241–254.
29. State vs. Fortin (2004). Available at http://www. dui1.com/ DuiCas-
eLawDetail3492/Page1.htm, accessed on November 21, 2005.
30. Ormerod, D. & Sturman, J. (2005). Working with the courts: advice for expert
witnesses. In L. Alison (Ed.), The Forensic Psychologists’ Casebook: Psycho-
logical Profiling and Criminal Investigation (pp. 170–193). Devon, UK: Willan.
31. Vrij, A. (2005). Criteria-based content analysis: a qualitative review of the first 37
studies. Psychol Pub Pol’y & Law, 11, 3–41.
32. Home Office. (2000). Statistics on Women and the Criminal Justice System.
London: Home Office.
33. Lea, S.J., Lanvers, U. & Shaw, S. (2003). Attrition in rape cases: developing a
profile and identifying relevant factors. Br J Criminol, 43, 583–599.
