Content uploaded by Clive R Hollin
Author content
All content in this area was uploaded by Clive R Hollin on Jul 30, 2019
Content may be subject to copyright.
Incorporating Context in Linking Crimes: An Exploratory
Study of Situational Similarity and If-Then Contingencies
JESSICA WOODHAMS1*, CLIVE HOLLIN2 and RAY BULL2
1University of Gloucestershire, Cheltenham, UK
2University of Leicester, Leicester, UK
Abstract
Personality psychologists have suggested that the context of a behaviour should be con-
sidered in studying behavioural consistency. They have operationalised this as studying
‘if (situation)–then (behaviour)’ contingencies and have demonstrated an association
between situational similarity and behavioural consistency. Previous research of behav-
ioural consistency in the forensic setting has tended to focus exclusively on the ‘then
(behaviour)’ part of the contingency—the offender’s behaviour. This paper considers
methodological approaches that might be used to investigate whether situational similar-
ity is associated with behavioural consistency, and to develop if–then contingencies.
Seventy-eight offences by stranger sex offenders were subjected to constant comparison
framework analysis to develop an offender behaviour checklist and a victim behaviour
checklist, and a combination of constant comparison framework analysis and hierarchical
cluster analysis were used to develop victim behavioural themes. Consistency in offender
behaviour and similarity in victim behavioural themes (representing situational similarity)
were measured using Jaccard’s coeffi cient for offence pairs within 13 solved series of
stranger sexual assaults. Correlational analyses were used to assess the relationship
between situational similarity and behavioural consistency. Contrary to expectations, no
relationship was found. The utility of linguistic computational programs in creating
if(victim behaviour)–then(offender behaviour) contingencies was tested with encouraging
results. However, little evidence of consistency in if(victim behaviour)–then(offender
behaviour) contingencies was found within the offence series. Explanations are proposed
for these novel fi ndings and avenues for future research are suggested. Copyright © 2008
John Wiley & Sons, Ltd.
INTRODUCTION
Empirical studies of attempts to link crimes have reported some excellent levels of predic-
tive accuracy, as indicated by areas under the ROC (Receiver Operating Characteristic)
Copyright © 2008 John Wiley & Sons, Ltd.
Journal of Investigative Psychology and Offender Profi ling
J. Investig. Psych. Offender Profi l. 5: 1–23 (2008)
Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/jip.75
*Correspondence to: Jessica Woodhams, Department of Natural and Social Sciences, University of
Gloucestershire, Francis Close Campus, Swindon Road, Cheltenham, GL50 4AZ, UK.
Email: jwoodhams@glos.ac.uk
2 J. Woodhams et al.
Copyright © 2008 John Wiley & Sons, Ltd. J. Investig. Psych. Offender Profi l. 5: 1–23 (2008)
DOI: 10.1002/jip
curve of 0.90–0.94 (Bennell & Jones, 2005; Woodhams & Toye, 2007). However, the
more traditional modus operandi behaviours have often been reported as less accurate
predictors of linkage than geographical distance between crime pairs (Bennell & Jones,
2005). The types of modus operandi behaviours that have been investigated in studies of
burglary, robbery, and sexual assault include behavioural domains of control, escape,
sex, target selection, planning, and entry. This has led some researchers to suggest that
modus operandi behaviours might be less effective for linking purposes (Goodwill &
Alison, 2006). However, the question is raised as to whether predictive accuracy suffers
when solely considering behaviours devoid of the context (or situation) in which they
occur. Whether linkage accuracy could be improved by considering the interaction
between context (the situation) and behaviour was a question posed by Woodhams,
Hollin, and Bull (2007). Goodwill and Alison (2007) similarly questioned whether the
accuracy of profi ling offender characteristics could be improved by considering the con-
text of the crime. Specifi cally, they investigated whether the ability to predict offender
age from victim age was moderated by offence planning and gratuitous aggression. This
paper aims to open up a discussion of whether the situational aspects of crime can be
incorporated into the linking task, whether this is desirable, and, if so, how this might be
achieved.
The interactionist approach to behavioural consistency
Personality psychologists have proposed a model of personality and behaviour called the
Cognitive Affective Personality System (CAPS) (Mischel & Shoda, 1995). People are
purported to have their own distinct if (situation)–then (behaviour) profi les, or behavioural
signatures (Mischel, 1999; Shoda, Mischel, & Wright, 1994; Wright & Mischel, 1987).
The model predicts that individuals will show consistency in the way they interact with
situations. Situations that are experienced as psychologically similar are hypothesised to
elicit consistent behaviours in an individual (Furr & Funder, 2004). It follows that there
will be a relationship between behavioural consistency and situational similarity, with
greater situational similarity resulting in greater behavioural consistency.
These models also predict that when faced with a situation of similar psychological
meaning, different people will react in different ways because of their individual and
distinctive CAPSs. Shoda, Mischel, and Wright (1993) explain that individuals encounter
and react to situational demands differently. In this way, and importantly, different indi-
viduals will react in different ways to the same situation. For example, Mischel and Shoda
(1995) explain that in a situation of receiving criticism from an intimate partner, indi-
viduals more sensitive to rejection would be more upset than those less sensitive. This
theory suggests that we can expect to observe, under certain conditions, both behavioural
consistency and behavioural distinctiveness.
Personality psychologists discuss two phenomena that are relevant to the linking of
crimes. These are temporal stability and cross-situational consistency. Temporal stability
is consistency in behaviour across the ‘same’ situation, whereas cross-situational consis-
tency refers to behavioural consistency across different situations (Shoda, Mischel, &
Wright, 1993). However, rather than consider them separate entities, Shoda et al., describe
temporal stability and cross-situational consistency as representing two extreme ends of a
continuum, which represents situational similarity. Shoda et al., state that ‘Cross-
situational consistency should increase as the situation similarity increases, approaching
what we expect for temporal stability’ (p. 1024).
Incorporating context in linking crimes 3
Copyright © 2008 John Wiley & Sons, Ltd. J. Investig. Psych. Offender Profi l. 5: 1–23 (2008)
DOI: 10.1002/jip
Whether such hypotheses are correct has been empirically tested by personality psy-
chologists, who broadly accept the phenomenon of temporal stability as empirically sup-
ported. However, there is less consensus for cross-situational consistency (Pervin, 2000;
Shoda, Mischel, & Wright, 1993). Their fi ndings, in relation to behavioural consistency
in general, are outlined elsewhere (Woodhams, Hollin, & Bull, 2007). The following sec-
tions focus instead on how personality psychologists have conceptualised situational
similarity and psychological salience, and how they have empirically tested whether
situational similarity is positively associated with behavioural consistency.
Situational similarity
Psychological salience of situations is described in terms of features that activate feelings
and thoughts in the personality system (Shoda & Mischel, 2000). Situational similarity
in relation to psychologically salient factors has been conceptualised in several different
ways (Furr & Funder, 2004; Mischel & Shoda, 1995; Shoda, Mischel, & Wright, 1993).
Before considering how forensic psychologists might study the effect of situational
similarity on criminal behavioural consistency, these different conceptualisations are
described.
Situational similarity as task type and task demands
Several researchers have discussed situational similarity in terms of the demands the
situation places on the individual. For example, Shoda, Mischel, and Wright (1993) refer
to social, cognitive, self-regulatory, physical, and motor control demands. Similarly, Furr
and Funder (2004) partly defi ned situational similarity as similarity in the task the par-
ticipant was to complete. They manipulated the task type by presenting participants with
three scenarios. These were an unstructured ‘getting acquainted’ task, a co-operative task,
and a competitive task. In both of these studies, situational similarity, as defi ned by task
demands, was found to be associated with greater behavioural consistency.
Situational similarity as valence
Mischel and Shoda (1995) identifi ed the valence of situation (which can be positive or
negative) as a psychologically salient feature important in determining similarity between
situations. Being teased, provoked, or threatened are given as examples of negative
valence, and receiving positive peer contact and receiving praise from an adult were con-
sidered to be situations possessing positive valence. Whilst Mischel and Shoda (1995) do
not provide a defi nition of valence, these examples seem to suggest that valence relates
to the face-threatening or face-saving quality of a situation. According to Thomas
(1995), the term ‘face’ refers to an individual’s self-image, with communications either
being face-threatening, threatening to the hearer’s self image, or face-saving and protect-
ing of the hearer’s self-image. Mischel and Shoda (1995) cite empirical evidence that
supports the premise that situations of similar valence result in consistent behaviour.
However, they add that valence cannot be the only factor that affects situational similarity.
They observed some variation in behaviour between the three situations of negative
valence, suggesting that whilst these three situations shared the feature of negative valence,
they continued to represent psychologically distinct situations for the individuals encoun-
tering them.
4 J. Woodhams et al.
Copyright © 2008 John Wiley & Sons, Ltd. J. Investig. Psych. Offender Profi l. 5: 1–23 (2008)
DOI: 10.1002/jip
Situational similarity as partner type
As well as defi ning situational similarity with regard to the type of task a participant was
to complete, Furr and Funder (2004) hypothesised that situations involving the same
partner would appear more similar to the participant than situations involving different
partners. Similarly, Magnusson, Gerzen, and Nyman (1968, as cited in Shoda, Mischel,
and Wright 1993) conceptualised situational similarity in terms of interacting with the
same or different group of participants. These researchers have therefore conceptualised
psychological salience in situations as relating to the characteristics of the person with
whom the participant was interacting. Furr and Funder (2004) give the example of a par-
ticipant interacting with an extroverted versus an introverted partner. They investigated
this hypothesis empirically by asking each participant to engage in a set of tasks with two
different partners. They found greater behavioural consistency when the participants were
working with the same partners.
Shoda, Mischel, and Wright (1993) also partially conceptualised psychological similar-
ity as being affected by the type of person with whom the participant was interacting. In
their study, they compared behavioural consistency between situations where a child was
interacting with a peer versus an adult. They found greater behavioural consistency in
situations where the child interacted with the same type of person.
In summary, personality psychologists have suggested that the similarity an individual
perceives between situations will be affected by the type of person with whom they are
interacting, the type of task (competitive, cooperative, or unstructured) and the competen-
cies it demands, and the valence of the situation.
It should, however, be noted that the effect of situational similarity on behavioural
consistency is thought to be mediated (Shoda, Mischel, & Wright, 1993). That is to say,
where a situation demands competencies that the individual has diffi culty providing, this
will be experienced as more stressful and will elicit negative affect. Such stressful situa-
tions are hypothesised to result in more automatic behaviours, which show greater consis-
tency across situations (Furr & Funder, 2004; Shoda, Mischel, & Wright, 1993). In contrast,
less stressful situations are thought to allow greater cognitive mediation of behaviour, with
subsequent behaviours being less automatic and impulsive. Behaviours that are cognitively
mediated are hypothesised to show less cross-situational consistency because the demands
made by the situation become less important (Shoda, Mischel, & Wright, 1993). If victim
resistance was considered a situational feature that might cause an offender negative affect
or stress, this might partly explain why greater consistency has been observed in control
behaviours in serial stranger sex offences, homicide, and robbery (Bateman & Salfati,
2007; Grubin, Kelly, & Brunsdon, 2001; Woodhams & Toye, 2007; Woodhams, Bull, &
Hollin, 2007). Evidence from personality psychology for the mediation effect of automatic
versus cognitively mediated behaviours is, however, mixed (Furr & Funder, 2004).
Situational similarity between crimes
The pervading practice in studying case linkage has been to focus on consistency in
behaviour without taking into account how behaviour is the result of interaction with the
situation. It could be argued that this research implicitly assumes that the same type of
crime represents the same type of situation. However, as Shoda, Mischel, and Wright
(1993, p. 1024) state, ‘the “same” situation observed on two different occasions . . . is never
exactly the same’.
Incorporating context in linking crimes 5
Copyright © 2008 John Wiley & Sons, Ltd. J. Investig. Psych. Offender Profi l. 5: 1–23 (2008)
DOI: 10.1002/jip
That the situation can affect the behaviour that is expressed by an offender at a crime
scene is apparent in writings about the practice of case linkage (Hazelwood & Warren,
2003; Woodhams, Hollin, & Bull, 2007). Such writings advise practitioners of case linkage
to consider the possibility that the situation may lead to ‘aspects of the crime being diluted,
modifi ed or interrupted’ (Hazelwood & Warren, 2003, p. 590) and to ‘be prepared to
discuss any dissimilarities noted across the series of crimes and why they do not negatively
impact the opinion’ (Hazelwood & Warren, 2003, p. 594). Yet, whilst it is commonplace
for the situation to be considered in practice, and whilst some researchers recognise that
the situation may affect consistency (Santtila et al., 2008), empirical research on case
linkage has not yet investigated whether incorporating the interaction between the situation
and the behaviour improves linkage accuracy. This is not surprising, since even in person-
ality psychology, there are few studies on this topic (Furr & Funder, 2004). This is in part
because of the diffi culty of identifying what characteristics of the situation are psycho-
logically salient (Shoda, Mischel, & Wright, 1993), and thus relevant in studying how the
situation and behaviour interact.
Before such empirical research could commence in forensic psychology, what consti-
tutes psychological similarity and what features of situations are salient in criminal inter-
actions would need to be understood. The need for similar research on behavioural
salience was highlighted with regard to furthering research on offender profi ling (Canter,
2000).
A series of crimes is defi ned as offences committed by one offender (or group of offend-
ers) against different victims. The offender is, therefore, interacting with a different partner
in each situation. Personality psychology would suggest, therefore, that individual crimes
in a series could represent psychologically different situations. However, it could also be
argued that the task type, as defi ned by Furr and Funder (2004), remains the same across
a crime series. For example, a series of rapes represents a series of competitive tasks,
where the offender’s and victims’ aims are diametrically opposed (Grant & Woodhams,
2007). The same could be said for robbery (Luckenbill, 1981) and homicide. This would
suggest that crimes in a series represent psychologically similar situations. However,
whilst not articulated in this way, there is other forensic psychology research that suggests
that individual crimes in a series might vary in their psychological meaning for an
offender.
Victim behaviour has been the principal focus of this research (Davies, 1991; Davies,
1992; Grubin, Kelly, & Brunsdon, 2001). Victim resistance has been reported to result in
increased verbal and physical aggression (Davies, 1991), a reduction in sexual behaviour
(but conversely an increase in sexual behaviour for sexual sadists), and a decrease in
‘ingratiating social behaviours’ (Davies, 1992, p. 191) by offenders in sexual offence
series. Studies of victim resistance in rape and robbery offences report different victims
using verbal protests, running away, and fi ghting offenders (Block & Skogan, 1986;
Greenfi eld, 1997; Wright & West, 1991). Burgess and Holmstrom (1976) report the use
of additional behaviours, such as stalling, fl attery, and bargaining, by some victims of
rape.
In differentiating situations on the basis of the demands they place on individuals,
Shoda, Mischel, and Wright (1993) refer to demands that could have some relevance to
interpersonal types of crime, such as rape and robbery. These were ‘requires the ability to
resolve disagreements with peers’ (a social demand), ‘requires the ability to resolve dis-
agreements with adults’ (a social demand), ‘requires the ability to tolerate frustration’ (a
self-regulatory demand), ‘requires the ability to delay gratifi cation’ (a self-regulatory
6 J. Woodhams et al.
Copyright © 2008 John Wiley & Sons, Ltd. J. Investig. Psych. Offender Profi l. 5: 1–23 (2008)
DOI: 10.1002/jip
demand), ‘requires physical strength’ (a physical demand), ‘requires resistance to injury’
(a physical demand), and ‘requires physical quickness’ (a physical demand). These differ-
ent types of situational demands suggest that different types of victim behaviour could
place different demands on an offender, in terms of social, self-regulatory, and physical
demands. In addition, victim behaviour might affect the degree to which a crime represents
a competitive task. For example, in their study of victim resistance in attempted homicides,
Fritzon and Ridgway (2001) report that two-thirds of the cases sampled did not involve
victim resistance. Thus, psychological meaning and situational similarity in interpersonal
crimes, such as rape, robbery, and attempted homicide, could be affected by victim
behaviour.
As well as potentially making varying demands of the offender and affecting the task
type, victim behaviour is also likely to affect the valence of the offence situation by threat-
ening or saving the face of the offender. A victim trying to stall for time is less face-
threatening than a victim who shouts obscenities at his/her offender.
Personality psychologists have been able to manipulate situational circumstances (e.g.
changing partner and task type), varying the ‘if’ part of the if–then contingency to inves-
tigate its effect on behaviour. In addition, they are able to ensure that each participant
experiences each ‘if’. Thus, a complete data set of if(situation)–then(behaviour) contingen-
cies can be derived for each participant (C. Bennell, personal communication, June 13,
2008). However, in forensic psychology, this type of manipulation is not possible. The
if(situation)s an offender is exposed to is beyond the control of the academic investigator
because forensic data is naturally occurring. However, as is suggested above, it seems
plausible that the circumstances of crimes might vary in similar ways to the qualities of
situations manipulated and investigated in personality psychology.
Determining the appropriate level of investigation
As well as considering how personality psychology research on situational similarity
relates to the investigation of criminal behaviour, there is also a question of what level of
investigation is appropriate.
The crime type level
The crime scenario in its entirety could represent a situation, and thus, a series of the same
type of crime could be considered as representing similar, if not the same, situations. If
this is the case, behavioural consistency across crime series would represent more a form
of temporal stability (Woodhams & Grant, 2006) than cross-situational consistency. As
previous studies of case linkage have demonstrated, there can be quite considerable behav-
ioural consistency across series of the same class of offence (Bennell & Canter, 2002;
Bennell & Jones, 2005; Woodhams & Toye, 2007; Woodhams, Bull, & Hollin, 2007). Such
consistency suggests that this might not be an inappropriate level at which to consider
situational similarity. In this scenario, the ‘if’ part of the if(situation)–then(behaviour)
contingency would represent the overall class of crime (e.g. if[rape]–then[offender . . .]).
The individual crime level
Alternatively, different crimes within a series could be deemed to represent different situ-
ations. As noted above, it seems likely that crimes within an offender’s series might vary
in task type, task demands, and valence. In such a scenario, the if-part of the if(situation)–
then(behaviour) contingency could be if(resistant victim)–then(offender . . .).
Incorporating context in linking crimes 7
Copyright © 2008 John Wiley & Sons, Ltd. J. Investig. Psych. Offender Profi l. 5: 1–23 (2008)
DOI: 10.1002/jip
To investigate whether the fi ndings from personality psychology—that situational sim-
ilarity is associated with behavioural consistency—generalise to criminal behaviour, the
situational similarity of each crime pair in a series could be quantifi ed. Whilst the same
method used in personality psychology could not be used for reasons outlined above, it
should be possible to investigate whether greater situational similarity is associated with
greater behavioural consistency if similarity in psychologically salient characteristics, such
as valence, task type, and task demands, can be measured.
The offender–victim interaction level
A further possibility is to view each crime in a series as representing a collection of numer-
ous if–then contingencies. In an interpersonal type of crime, such as robbery, attempted
homicide, or rape, each separate interaction with the victim could represent an if–then
contingency. For example, if victim screams–then offender hits the victim. Whilst previous
studies using multidimensional scaling techniques have considered which victim behav-
iours co-occur with offender behaviours (e.g. Porter & Alison, 2004), such analyses cannot
preserve the temporal ordering of victim and offender behaviours. For example, Porter
and Alison note that the offender behaviour of demeaning the victim is located within the
same region of the SSA (Smallest Space Analysis) plot as the victim refusing to comply.
However, from this analysis, it is unclear whether the victim refusing to comply results
in the offender demeaning the victim (e.g. if[victim refuses to comply]–then[offender
demeans victim]), or whether the offender demeaning the victim results in the victim
refusing to comply (e.g. if[offender demeans victim]–then[victim refuses to comply]). A
technique is therefore needed, which can maintain temporal ordering between victim and
offender behaviours.
The fi eld of linguistics suggests a method through which these if(victim behaviour)–
then(offender behaviour) contingencies could be identifi ed. Linguists study collocations,
which are ‘relationship[s] of habitual co-occurrence between words’ (Stubbs, 1995, p. 23).
In studying collocations, linguists specify a query word, the word they are investigating,
and determine which words co-occur with it. Linguists have devised specialist software
and statistical measures, such as mutual information values, to identify collocations and
the degree of co-occurrence of specifi c words across large texts. In the same way, psy-
chologists could specify a victim behaviour and determine which offender behaviours
follow it in victims’ accounts of crimes.
The question as to the most appropriate level at which to consider situational similarity
is one that does not seem to have been answered in the personality psychology literature.
The research conducted thus far by personality psychologists appears to investigate situ-
ational similarity at the second level mentioned above. It is the characteristics of the overall
situation that are being manipulated such as its valence, the goals it stimulates, and the
skills it requires. However, as the purpose of this paper is to consider various methodolo-
gies that could be further explored in the future, investigation was not limited to this
level.
Rationale
At the crime type level, a degree of behavioural consistency has been identifi ed, at least
in some behavioural domains for some crime types. What remains to be explored is
whether offenders show consistency in behaviour across different types of crime. Some
offenders are versatile in their offending (Farrington & Lambert, 1997; Klein, 1984), and
8 J. Woodhams et al.
Copyright © 2008 John Wiley & Sons, Ltd. J. Investig. Psych. Offender Profi l. 5: 1–23 (2008)
DOI: 10.1002/jip
thus, data exists that could be used to investigate such a proposition. However, this would
require behavioural domains that are common to more than one offence type. Behavioural
consistency in control behaviours has now been investigated for serial robbery, rape, and
homicide (e.g. Bateman & Salfati, 2007; Grubin, Kelly, & Brunsdon, 2001; Woodhams &
Toye, 2007), and thus control appears to be one domain that is shared by more than
one offence type. Such an analysis was beyond the scope of the current paper due to the
available data set, but this research question would benefi t from exploration in the
future.
Instead, this paper illustrates some of the methodological issues that arise when trying
to incorporate the situation into behavioural linking using situational similarity at the lower
two levels (the individual crime level and the victim–offender interaction level). At both
levels, situational similarity was conceptualised as similarity in victim behaviours. These
are recognised in forensic psychology literature to affect valence, task type, and task
demands, all psychologically salient features of situations that have been investigated in
personality psychology.
At the individual crime level, we investigated whether there was an association between
situational similarity and behavioural similarity (consistency) of crime pairs. Qualitative
and quantitative analytical techniques were used to explore whether victim behavioural
themes could be identifi ed that differed on psychologically salient characteristics. Situa-
tional similarity was conceptualised as similarity in victim behavioural themes between
crime pairs. Whilst the cognitive affective personality system proposed by Mischel and
Shoda (1995) suggests that the way each individual will construe a situation will be rela-
tively unique, the research in personality psychology suggests that it is possible for
researchers to identify the psychologically salient features of situations and objectively
group situations for psychological similarity (Furr & Funder, 2004).
The fi rst step to investigating behavioural consistency at the victim–offender interaction
level requires the generation of if(victim behaviour)–then(offender behaviour) contingen-
cies. The utility of linguistic collocation software to generate such contingencies was
explored.
METHOD
Data
The data used to investigate these issues were obtained from the Serious Crime Analysis
Section (SCAS) of the National Policing Improvement Agency. This unit holds the most
comprehensive collection of data on stranger sexual assaults in the UK. The databases at
SCAS were searched for suitable offences. Thirteen juvenile1 serial stranger sex offenders,
who had been convicted of their offences, were identifi ed and all of their (known) 39
crimes were sampled. To be characterized as a serial offender, the offender had to have
committed two or more sexual offences against different victims. The 39 serial offences
represented all of the offences committed by juvenile serial stranger sex offenders in the
database that had resulted in a conviction. A further 39 non-series stranger sex offences
that had resulted in a conviction were sampled, forming an overall data set of 78 crimes.
Information about each crime was taken from records of the victims’ accounts in the crime
1A juvenile was defi ned as an offender aged less than 18 years. For some offenders in the sample, they were
aged less than 18 years at the start of their series but were aged over 18 years for some of their later offences.
Incorporating context in linking crimes 9
Copyright © 2008 John Wiley & Sons, Ltd. J. Investig. Psych. Offender Profi l. 5: 1–23 (2008)
DOI: 10.1002/jip
fi les and represented victim statements, victim interview transcripts, interview notes, court
reports, and crime analysts’ reports.
Most offences were committed by a lone offender (92%, n = 72). Where multiple offend-
ers were involved, 4% (n = 3) of all offences were committed by two offenders, 3% (n =
2) by three offenders, and 1% (n = 1) by four offenders. All offenders were male. The
mean offender age2 was 16.26 years (median = 16.00, SD = 2.65, range = 11–26). All
victims were female. The mean victim age was 26.39 years (median = 23.00, SD = 13.72,
range = 10–81). The number of offences within each series is shown in Table 1.
Procedure
The case fi le for each offence was accessed and a narrative of each offence was written
from the victim’s account or report. The narrative contained the behaviours of offenders,
victims, and witnesses in the order reported by the victim. Identifi ers such as the victim’s
name, offender’s name and place names were not recorded in the narratives to ensure they
were anonymous.
Offender behaviour checklist
Through constant comparison framework analysis (Ritchie & Spencer, 1994) of the 78
stranger sex offences (including the 39 serial offences), an offender behaviour checklist
was created. Constant comparison framework analysis involves several stages: familiarisa-
tion with the data, development of the framework, open-coding, indexing, and charting
(Bradley, Curry, & Devers, in press; Henwood & Pidgeon, 2006; Ritchie & Spencer, 1994).
This produced a checklist of 148 offender behaviours, the codes for which are listed in
Table 2.
It is commonplace in studies of case linkage to remove from the statistical analyses
offender behaviours that are rare (Grubin, Kelly, & Brunsdon, 2001; Santtila, Junkkila, &
Sandnabba, 2005). This practice was followed in the current study with rare offender
behaviours removed from the analysis. Rarity was conceptualised as behaviours or themes
occurring in less than 10% of the offences. None of the offender behaviours occurred in
more than 90% of cases. This reduced the number of offender behaviours from 148 to 55.
Following this, each offence was binary-coded for the presence and absence of the 55
offender behaviours. Offender themes were not developed from the offender behaviours
because the dominant approach in case linkage research has been to use checklists of
behaviours rather than themes (Bennell & Canter, 2002; Bennell & Jones, 2005; Goodwill
& Alison, 2006; Woodhams & Toye, 2007; Woodhams, Bull, & Hollin, 2007).
Table 1. Number of offences per series
Series number Number of offences Series number Number of offences
13 82
22 92
34104
42112
52122
66136
72——
2Where two or more offenders were responsible for an offence, their mean age was calculated.
10 J. Woodhams et al.
Copyright © 2008 John Wiley & Sons, Ltd. J. Investig. Psych. Offender Profi l. 5: 1–23 (2008)
DOI: 10.1002/jip
Victim behaviour checklist
A victim behaviour checklist was also created through constant comparison framework
analysis following the same stages.
Inter-rater reliability
Both the offender behaviour and the victim behaviour checklists were shown to be reliable
through tests of inter-rater reliability. Ten per cent of the accounts (eight narratives) were
dual-coded. Cohen’s kappas were calculated that were of a satisfactory level (0.62 and
0.59) (Cicchetti, 1994; Landis & Koch, 1977) and signifi cantly greater than chance (p <
0.001). These and the percentage agreement values (94.25 and 91.22%) were very similar
to those reported in past research using checklists of behaviours (Langstrom, Grann, &
Lindblad, 2000; Porter & Alison, 2005; Santtila, Junkkila, & Sandnabba, 2005).
Table 2. Offender behavioural codes
On foot Direct sexual Ignores request Seen before
Outdoor Undresses self Location—con Allows action
Surprise Restrain body Blindfold Blindfold hand
Stalks Departs calmly Complies Takes weapon
Location—force Conditional threat Masturbates self-hand Provokes
Depart quick Kiss Direct don’t look Innocence
Breast Weapon Hiding Masturbates self-feet
Grab Discloses intent Mocked Lies—self-image
Con Indoor Verbal abuse Masturbates self-breasts
Positions Question—personal Disclose criminal Complies with accomplice
Undresses victim Instrumental violence Lying Clarifi cation
Touch vaginal Pursues Invitation Demonstrates
Physical threat Question—sexual Rips clothes Scripting
Ceases—put off Blocks escape Directs accomplice Unspecifi ed digital
penetration
Property Expressive violence Returns property
Requests help Erection Erectile dysfunction Prevents look
Disclose personal Re-dressed Forces entrance Car
Attracts attention Verbal threat Excuses Lie-upset
Ceases resistance Disclose intimate Removes blindfold Styled hair
Gag hand Don’t report Disguise Plan
Direct quiet Touch bottom Refuses answer Lie—protect identity
Direct position Direct undress Direct resist Personal question to
accomplice
Direct non-sexual Concern Question—security
Penile vaginal penetration Digital vaginal penetration Expresses shock Return home
Restrains arms Touch penis Sits Can’t help
Releases Contradicts Simulated intercourse Fulfi lled promise
Offers assistance Apologies Broke in Urinates
Fellatio Clean Bicycle Look out
Intrudes Compliment Pushes past Feet
Hug Negotiation Touch stomach Sexual noises
Ejaculation Arousal Direct re-dress Hold hand
Direct stay Penile anal penetration Gag Complain
Exposure Leaves weapon Fingerprints Accepts guilt
Cut wires Offers gift Drugs Removes gag
Minimises actions Alcohol Educates Observed
Touched leg Criticises Binding Spits
Cunnilingus Extends time Gloves
Reassures Refuses request Offenders argue
Incorporating context in linking crimes 11
Copyright © 2008 John Wiley & Sons, Ltd. J. Investig. Psych. Offender Profi l. 5: 1–23 (2008)
DOI: 10.1002/jip
Victim behavioural themes
The victim behaviours were subject to both qualitative and quantitative forms of analysis
to identify higher level themes that might differ in task type, task demand, and valence.
A combined approach was chosen to introduce methodological rigour. The analytical
techniques chosen were constant comparison framework analysis and agglomerative
hierarchical cluster analysis. All 124 behaviours were subject to constant comparison
framework analysis to identify higher level themes. The process involved is outlined
above. Nineteen higher-level functional themes emerged. Details of these 19 themes can
be provided by contacting the fi rst author.
To ensure a suffi cient ratio of cases-to-variables, it was not possible to include all 124
victim behaviours in the cluster analysis. As has been the practice in past studies (Davies,
Wittebrood, & Jackson, 1998; Grubin, Kelly, & Brunsdon, 2001; Salfati & Canter, 1999),
frequent and infrequent victim behaviours were removed from the data set. Victim behav-
iours that occurred in less than 10% of cases were removed. No victim behaviours occurred
in more than 90% of cases. Only 21 of the 124 victim behaviours occurred in more than
10% of cases, which reduced the ratio of cases-to-variables to an acceptable ratio of 78
cases to 21 variables.
Hierarchical cluster analysis has been used in past studies with similar cases-to-variables
ratios (for example in identifying styles of robbery behaviour and types of juvenile sex
offender) (Langstrom, Grann, & Lindblad, 2000; Woodhams & Toye, 2007) and does
not require prior knowledge of the number of clusters (Everitt, 1980). The stability of the
clustering solution was assessed by varying the clustering method used, as recommended
by Everitt, Landau, and Leese (2001). Clustering methods were only used if they were
suitable for similarity data (Everitt, Landau, & Leese, 2001). The four clustering methods
used were single linkage (nearest neighbour), complete linkage (furthest neighbour),
average within-groups clustering, and average between-groups clustering. The dendro-
grams were cut where the distances between fusion points seemed at their largest and
where relatively clear clusters appeared to emerge (Everitt, Landau, & Leese, 2001). The
single-linkage method resulted in chaining, which is not uncommon with this method
(Everitt, Landau, & Leese, 2001). However, dendrograms with seven similar clusters
emerged when using the within-groups clustering method, the average between-groups
clustering method, and the complete linkage (furthest neighbour) method. Details of the
seven clusters identifi ed can be obtained by contacting the fi rst author.
It is not surprising that the two analyses resulted in different numbers of clusters since
because of the statistical requirements of cluster analysis, not all 124 victim behaviours
could be included. The differences and similarities between the two analyses were con-
sidered and the outputs amalgamated to produce eight higher-level victim behavioural
themes. These are displayed in Table 3.
It is suggested that the behaviours within each grouping share common valence (e.g.
positive, negative) and demands (e.g. physical, social, self-regulatory), and share an
outcome for the offender, thus representing a task type (e.g. cooperative, unstructured, or
competitive). For example, active resistance by the victim is very face-threatening, and
therefore possesses negative valence for the offender. It also represents a group of behav-
iours that confl ict with the offender’s goals and which often require a physical response
by the offender. Thus, it presents the offender with a competitive task, rather than a co-
operative or unstructured task, which places physical demands on him. In contrast, victim
compliance presents a situation that is face-saving, thus possessing positive valence. It
12 J. Woodhams et al.
Copyright © 2008 John Wiley & Sons, Ltd. J. Investig. Psych. Offender Profi l. 5: 1–23 (2008)
DOI: 10.1002/jip
represents a group of behaviours that comply with the offender’s goals, representing a
cooperative task, placing few demands on him. It is, therefore, tentatively suggested that
each of the eight groupings might represent behaviours of particular psychological meaning
that is distinct from that of the other groupings. It would follow that similarity between
crimes in these victim behavioural themes would represent situations of greater psycho-
logical similarity.
Table 3. Victim behavioural themes identifi ed through constant comparison framework analysis
and the hierarchical cluster analysis
Theme Description Example behaviours
1. Active resistance The victim actively resists the
offender’s assault. She engages in
behaviours that directly communicate
her lack of consent to the suspect and
others, or engages in behaviours that
prepare her for this.
Verbal help-seeking
Takes weapon
Kicks
Punches
2. Invoking social
convention
Behaviours that signal to the offender
that his behaviour is wrong and
unwanted and perhaps contain the
expectation that the offender will
respond by ceasing his behaviour
because of empathy for the victim or
embarrassment for breaking social
conventions.
Reports frightened
Draws parallels with others
Tells suspect it hurts
Indicates child
3. Information gathering Behaviours with the intention of
gathering information, sometimes
with the aim of using this as
evidence at a future date.
Queries intentions
Removes blindfold
4. Put off Behaviours with the intention of
dissuading the suspect from his
actions.
Disease—put off
Police—put off
5. Face-threatening non-
compliance
Behaviours through which the victim
either completely or partially does
not comply with the suspect’s wishes
or social conventions that could
threaten his self-image.
Physical non-compliance
Refuses to answer question
Spits semen
6. Face-saving non-
compliance
Behaviours that are designed to
disguise the victim’s true intent of
non-compliance from the suspect and
instead give the impression of
compliance or a willingness to
comply with the suspect, thereby
maintaining social conventions.
False disclosure
Vague disclosure
Negotiation
7. Compliance The victim complies with the
offender’s wishes or his expectations
of her.
Obeys wishes
Makes assurances to stay
8. Spontaneous Spontaneous behaviours that do not
appear intentional, and from viewing
the accounts, were unlikely to be
misinterpreted as a form of non-
compliance.
Drops bags/belongings
Falls asleep
Incorporating context in linking crimes 13
Copyright © 2008 John Wiley & Sons, Ltd. J. Investig. Psych. Offender Profi l. 5: 1–23 (2008)
DOI: 10.1002/jip
It is important to note that a victim behaviour could only be present in one theme. A second
round of inter-rater reliability assessments was not, therefore, conducted since allocation
to a victim behavioural theme was based on the presence of victim behaviours, which had
already been assessed for inter-rater reliability and shown to be reliable. Each offence was
binary-coded for the presence and absence of each victim behavioural theme.
Quantifying similarity and consistency
It has been a common practice in studies of case linkage to quantify similarity in behaviour
between offences using Jaccard’s coeffi cient (e.g. Bennell & Canter, 2002; Bennell &
Jones, 2005; Woodhams & Toye, 2007). Jaccard’s coeffi cient has also widely been used
in other investigative psychology studies (e.g. Canter & Heritage, 1990; Canter et al.,
2003; Porter & Alison, 2004; Santtila et al., 2003). This is because of its suitability for
crime data, where there can be uncertainty as to whether the absence of an offender behav-
iour in a victim’s account is due to it not occurring or because of poor memory, amongst
other factors (Bennell & Canter, 2002). Similarly, a victim may fail to recall a behaviour
of their own. Jaccard’s coeffi cient, therefore, seemed an appropriate measure of similarity
to also use in the current study. It was not only used to measure behavioural similarity
(consistency) between offences but also to measure situational similarity (similarity in
victim behavioural themes). Jaccard’s coeffi cients were calculated for behavioural similar-
ity and situational similarity for each pair of offences using SPSS.
The relationship between situational similarity and behavioural consistency: The
individual crime level
An aim of the study was to investigate whether behavioural consistency was greater in
situations of greater similarity, as has been found in personality psychology research. If
the hypothesis were correct, a positive association should be observed. This was investi-
gated in two ways. Using the subset of 39 serial offences, 53 linked crime pairs were
created by pairing together the offences by the same serial offender. One method was to
conduct an overall correlation between behavioural consistency (as measured between
each of the 53 pairs) and situational similarity (as measured between each of the 53 pairs)
for all series. The amalgamation of several series, and thus the offending of several offend-
ers, into one analysis may, however, mask relationships between behavioural consistency
and situational similarity that vary in strength between individual offenders. An alternative
method was to conduct a correlation per series. The small numbers of offence pairs within
each series in the data set precluded the use of inferential statistics to investigate correla-
tions between situational similarity and behavioural consistency for each series. Instead,
for series with a length greater than two offences, a scatter plot was drawn to tentatively
investigate whether this suggested that there was an association between situational simi-
larity and behavioural consistency. Of the 13 series in the data set, only fi ve series con-
stituted more than two offences.
The relationship between situational similarity and behavioural consistency:
The offender-victim interaction level
The WordSmith program (© Mike Scott, 2004–2007) was chosen for the study since it
contained the function to study collocations. Using WordSmith, for victim behaviours
(representing each ‘if’ part of the contingency), all possible offender reactions (represent-
14 J. Woodhams et al.
Copyright © 2008 John Wiley & Sons, Ltd. J. Investig. Psych. Offender Profi l. 5: 1–23 (2008)
DOI: 10.1002/jip
ing the ‘then’ parts of the contingency) were generated. As this represents a preliminary
investigation, only the three most frequent victim behaviours were selected: victim strug-
gles, victim gives the suspect an order, and victim verbally seeks help. These behaviours
occurred in 56, 36, and 31% of the offences that composed the 13 series, respectively. As
this represented the fi rst study to use collocation software for this purpose, the span was
maintained at 0:1. In other words, the program only considered the offender behaviour
immediately following the victim behaviour of interest. That the span can be manipulated
with ease within this program is advantageous and would allow for the study of more
complex contingencies in the future.
RESULTS
The individual crime level
Descriptive statistics
Table 4 contains the means and standard deviations for behavioural consistency and situ-
ational similarity for the 53 linked pairs generated from the 13 offence series as measured
by Jaccard’s coeffi cient.
The overall association between situational similarity and behavioural consistency
A correlation of behavioural similarity and situational similarity was computed on the
subset of 53 linked crime pairs. Personality psychologists have reported that greater time
between observations can reduce behavioural consistency (Pervin, 2000). It was therefore
important to determine if time (in days) between offences was correlated with behavioural
consistency. The distributions of time between offences, behavioural consistency, and
situational similarity were assessed for normality using Kolmogorov–Smirnov tests. Non-
parametric correlations were used where the distribution(s) were not normal. A Spearman’s
correlation was computed between ‘time between offence pair’ and behavioural consis-
tency. Time between offences was not found to be correlated with behavioural consistency
(r = −0.03, n = 53, p = 0.823). Therefore, it was not necessary to control for time in the
correlation between behavioural consistency and situational similarity.
A Pearson’s correlation was conducted between behavioural consistency (as measured
between each of the 53 offence pairs using Jaccard’s coeffi cient) and situational similarity
(as measured in the same manner). Contrary to expectation, no correlation was found
between these two variables (r = −0.06, n = 53, p = 0.669).
The association between situational similarity and behavioural consistency within
individual series
As can be seen from Figures 1–5, the degree and direction of correlation between situa-
tional similarity and behavioural consistency varies considerably by series. Only the scatter
Table 4. Means and standard deviations for behavioural consistency and situational similarity for
the 53 linked pairs created from the 13 series of juvenile stranger sex offences
Mean Standard deviation
Behavioural consistency 0.34 0.15
Situational similarity 0.39 0.27
Incorporating context in linking crimes 15
Copyright © 2008 John Wiley & Sons, Ltd. J. Investig. Psych. Offender Profi l. 5: 1–23 (2008)
DOI: 10.1002/jip
plot for series 2 approximates a positive correlation, as would be hypothesised were greater
behavioural consistency observed in situations of greater similarity within a series. The
expectation that situational similarity would be positively correlated with behavioural
consistency within a series was not supported in the majority of the fi ve series amenable
to preliminary investigation.
0.20 0.30 0.40 0.50 0.60 0.70 0.80
0.20 0.40 0.60
1.00
0.80
0.40
0.60
0.20
0.00
Situational Similarity
0.40
0.20
0.00
Situational Similarity
0.60
0.40
Situational Similarity
0.20 0.40
Behavioural Consistency
Behavioural Consistency
Behavioural Consistency
0.00 0.20 0.40
0.80
0.60
0.40
0.20
0.00
Situational Similarity
Behavioural Consistency
R Sq Linear = 0.01 R Sq Linear = 0.116
R Sq Linear = 2.775E-5
R Sq Linear = 0.429
0.80
0.75
0.70
0.65
0.60
0.55
0.50
0.32 0.34
Situational Similarity
0.36 0.38 0.40
Behavioural Consistency
R Sq Linear = 0.833
0.42
Figure 1. Scatter plots of behavioural consistency and situational similarity for the fi ve series of juvenile
stranger sexual assault.
16 J. Woodhams et al.
Copyright © 2008 John Wiley & Sons, Ltd. J. Investig. Psych. Offender Profi l. 5: 1–23 (2008)
DOI: 10.1002/jip
The offender–victim interaction level
Descriptive statistics
Tables 5–7 display the various offender behaviours that follow the three victim behaviours
and their respective frequencies, as identifi ed by WordSmith.
Table 5. Frequencies of if(victim struggles)–then(offender X) contingencies
If–then contingency Frequency
Victim struggles—offender positions the victim 3
Victim struggles—offender bodily restrains the victim 3
Victim struggles—offender forcibly moves location 2
Victim struggles—offender makes intimate disclosure 2
Victim struggles—offender grabs the victim 2
Victim struggles—offender undresses victim 2
Victim struggles—offender directs the victim into a position 1
Victim struggles—offender blindfolds the victim 1
Victim struggles—offender stalks the victim 1
Victim struggles—offender directs the victim in a sexual behaviour 1
Victim struggles—offender gags victim with hand 1
Victim struggles—offender uses instrumental violence 1
Victim struggles—offender kisses victim 1
Victim struggles—offender discloses his intent 1
Victim struggles—offender touches victim’s vaginal area 1
Victim struggles—offender touches victim’s breast 1
Victim struggles—offender makes conditional threat 1
Table 6. Frequencies of if(victim gives suspect order)–then(offender X) contingencies
If–then contingency Frequency
Victim gives order—offender kisses victim 2
Victim gives order—offender uses instrumental violence 2
Victim gives order—offender tells victim he has weapon 1
Victim gives order—offender reassures victim 1
Victim gives order—offender ceases assault 1
Victim gives order—offender asks victim a sexual question 1
Victim gives order—offender mocks victim 1
Victim gives order—offender makes a conditional threat 1
Table 7. Frequencies of if(victim verbally seeks help)–then(offender X) contingencies
If–then contingency Frequency
Victim verbally seeks help—offender uses instrumental violence 4
Victim verbally seeks help—offender gags victim with hand 4
Victim verbally seeks help—offender makes conditional threat 3
Victim verbally seeks help—offender ceases assault 3
Victim verbally seeks help—offender tells victim he has weapon 2
Victim verbally seeks help—offender orders victim to be quiet 2
Victim verbally seeks help—offender undresses victim 1
Victim verbally seeks help—offender forcibly moves location 1
Victim verbally seeks help—offender positions the victim 1
Victim verbally seeks help—offender re-dressed himself 1
Victim verbally seeks help—offender grabs the victim 1
Incorporating context in linking crimes 17
Copyright © 2008 John Wiley & Sons, Ltd. J. Investig. Psych. Offender Profi l. 5: 1–23 (2008)
DOI: 10.1002/jip
Tables 8–10 show the frequency of these contingencies for each serial offender.
The statistics are limited to descriptive statistics because of the low frequencies of
if(victim behaviour)–then(offender behaviour) contingencies within this data set. From
examining Tables 8–10, it becomes apparent that whilst some offenders did repeat the
same if(victim behaviour)–then(offender behaviour) contingency within their series, this
was only within the same offence.
DISCUSSION
This paper reports an exploratory study of the relationship between situational similarity
and behavioural consistency. This was investigated at the individual crime level and the
victim–offender interaction level.
The individual crime level
The mean Jaccard’s coeffi cients for behavioural consistency and situational similarity were
both quite small. The mean Jaccard’s coeffi cient for behavioural consistency was lower
Table 8. Distribution of if(victim struggles)–then(offender X) contingencies across series
If–then contingency Total S1 S2 S3 S4 S5 S6 S7 S8 S9 S10 S11 S12 S13
Struggles—Positions 3/25 — — — 1 — — — — 2* — — — —
Struggles—Restrains 3/25 — — 1 — — 1 — — — — 1 — —
Struggles—Move location 2/25 — — 1 — — — — — — — 1 — —
Struggles—Int. disclosure 2/25 — — — — — — — — — — 2* — —
Struggles—Grabs V 2/25 1 — — — — 1 — — — — — — —
Struggles—Undresses V 2/25 — — — — 1 — — — — — 1 — —
Struggles—Direct position 1/25 1 — — — — — — — — — — — —
Struggles—Blindfold 1/25 1 — — — — — — — — — — — —
Struggles—Stalks 1/25 — — — — — — — — — — — — 1
Struggles—Directs sexual 1/25 1 — — — — — — — — — — — —
Struggles—Gags hand 1/25 — — 1 — — — — — — — — — —
Struggles—Ins. violence 1/25 — — — — — 1 — — — — — — —
Struggles—Kisses 1/25 — — — — 1 — — — — — — — —
Struggles—Dis. intent 1/25 — — — — 1 — — — — — — — —
Struggles—Touch vaginal 1/25 — — — — — — — — — — 1 — —
Struggles—Touch breast 1/25 — — — — — — — — 1 — — — —
Struggles—C. threat 1/25 — — — — — — 1 — — — — — —
*Consistency only within the same offence in the series.
S, series; V, victim.
Table 9. Distribution of if(victim gives order)–then(offender X) contingencies across series
If–then contingency Total S1 S2 S3 S4 S5 S6 S7 S8 S9 S10 S11 S12 S13
Gives Order—Kisses 2/10 — — — — — — — 1 — — — — 1
Gives Order—Ins. viol 2/10 — — 1 — — — — — 1 — — — —
Gives Order—Weapon 1/10 — — — — — — — — 1 — — — —
Gives Order—Reassure 1/10 1 — — — — — — — — — — — —
Gives Order—Ceases 1/10 — — — — — — — 1 — — — — —
Gives Order—Sex Q 1/10 — — — — — — — — — — — — 1
Gives Order—Mocks 1/10 — — — — — — — — — 1 — — —
Gives Order—C. threat 1/10 — — — — — — 1 — — — — — —
S, series.
18 J. Woodhams et al.
Copyright © 2008 John Wiley & Sons, Ltd. J. Investig. Psych. Offender Profi l. 5: 1–23 (2008)
DOI: 10.1002/jip
than fi gures reported for linked pairs of commercial robberies (Woodhams & Toye,
2007).
No evidence was found for a positive association between behavioural consistency and
situational similarity, as would have been predicted from research in personality psychol-
ogy. This was the case for an overall correlation incorporating 53 linked crime pairs, and
largely the case for each of the fi ve series that contained more than two offences. Only
one series showed any indication of a positive association between situational similarity
and behavioural consistency.
There are several reasons that might explain why little evidence of an association was
observed. It is possible that the conceptualisation of situational similarity and psycho-
logical salience in this study was inappropriate. In which case, a positive association
between behavioural consistency and situational similarity might be found if more appro-
priate means of measuring situational similarity were developed. There may be factors
other than the victim’s behaviour that will affect the psychological meaning of a situation
for an offender. For example, Davies (1992) reports how the physical appearance of a
victim can affect offender behaviour. She gives the example of a serial rapist who was
typically both verbally and physically aggressive towards his victims. In contrast, with
one victim, who was young and middle class, he was complimentary and considerate. This
reported variation in behaviour does not seem to be related to task type, task demands, or
valence as they have been discussed in the personality psychology literature. Hazelwood
and Warren (2003) and Santtila et al. (2008) also report how an offender’s mood, mental
state, and the external circumstances of a crime can affect the consistency of offending
behaviour. It is possible that mood and mental state affect the perceived valence of a
situation and the external circumstances affect task demands.
Mischel and Shoda (1995) observe that participants may not agree with the researcher’s
interpretation of psychological similarity. This is an important point and one that suggests
a potentially valuable avenue for further research. Studies could investigate how offenders
construe the psychological similarity of situations by asking them to rate offences for
similarity. This would be a similar approach to that taken by Furr and Funder (2004),
where they asked participants to rate situations subjectively for similarity. One method of
achieving this would be to use a card-sorting task, asking offenders to group together
Table 10. Distribution of if(victim verbally seeks help)–then(offender X) contingencies across
series
If-Then Contingency Total S1 S2 S3 S4 S5 S6 S7 S8 S9 S10 S11 S12 S13
Seeks Help—Ins. viol 4/23 — — 3* — — 1 — — — — — — —
Seeks Help—Gags 4/23 — — — — 1 — — — — — 3* — —
Seeks Help—C. threat 3/23 — — 1 ————— 2*— — — —
Seeks Help—Ceases 3/23 — — — — — 1 — — — — 1 — 1
Seeks Help—Weapon 2/23 — — 2* ——————— — — —
Seeks Help—Quiet 2/23 — — 1 — — 1 — — — — — — —
Seeks Help—Undresses V 1/23 — — 1 ——————— — — —
Seeks Help—Move location 1/23 — — — — 1 — — — — — — — —
Seeks Help—Positions 1/23 — — 1 ——————— — — —
Seeks Help—Re-dressed 1/23 — — — 1 — — — — — — — — —
Seeks Help—Grabs V 1/23 — — — — — 1 — — — — — — —
*Consistency only within the same offence in the series.
S, series; V, victim.
Incorporating context in linking crimes 19
Copyright © 2008 John Wiley & Sons, Ltd. J. Investig. Psych. Offender Profi l. 5: 1–23 (2008)
DOI: 10.1002/jip
offences that they consider to be similar and explaining what qualities these offences
possess that make them similar. Interviews with offenders may be productive in further
understanding what characteristics of an offence situation are salient. Just such an approach
was suggested by Grubin, Kelly, and Brunsdon (2001). Forensic psychologists working
with offenders would be well-placed to pursue research in this area.
Future studies could also take a similar approach to the current study and apply the
same or different analytical techniques to samples of victim behaviours to determine if
there are alternative ways of categorising such behaviour in terms of valence, and the
demands victim behaviour places on offenders. For example, whilst the hierarchical cluster
analysis suggested that the victim behaviour of giving orders should be located in the
theme ‘active resistance’, unlike other behaviours in this theme, this behaviour does not
place a physical demand on the offender.
It might also be the case that the conceptualisation of situational similarity was appro-
priate, but its measurement between situations was inaccurate because the information
about victim behaviour from the police fi les contained omissions or distortions. In the
current study, a victim account of the offence was not always available, and instead, a
report produced for the courts or by a crime analyst had to be used instead. The limitations
of using police data for psychological research have been discussed elsewhere (Alison,
Snook, & Stein, 2001; Canter, 2000).
In personality psychology research, situations have been rated on a 1–7 scale for the
degree of demand they placed on the individual, and situations have been correlated for
their similarities (Shoda, Mischel, & Wright, 1993). In the current study, situations were
coded for their qualities in a binary fashion and their similarity was assessed using the
similarity coeffi cient—Jaccard’s coeffi cient. Future studies may wish to investigate other
means of assessing situational demands and measuring situational similarity.
It is possible that there is greater opportunity for the situation, as defi ned by victim
behaviour, to impact on offenders’ behaviour in some crimes than others. Santtila et al.
(2008) recently observed in their study of murders that the immediate murder of the victim
meant that the victim’s behaviour had little effect on the offender’s behaviour. They explain
‘in some of the murders included in the present study the victim was shot immediately at
the beginning of the event, leaving the offender free to express his personality and psy-
chological needs with the body of the victim without any behavioural interference from
the victim’ (p. 19). In relation to sexual offences, surprise-approach style offences, where
the victim is physically dominated from the start, would be similar, with con-approach
style offences perhaps allowing greater opportunity for victim behavioural infl uence. If
offenders are consistent in their approach styles, some offenders might, therefore, experi-
ence greater situational variation than others.
The victim–offender interaction level
Linguists use statistical programs to study collocation, and it was proposed that the same
method could be applied to develop if(victim behaviour)–then(offender behaviour) con-
tingencies. The program performed well, and if–then contingencies were identifi ed for
three frequent victim behaviours. In principle, this therefore appears to be a relatively
straightforward task; however, the question arises of whether one should only consider the
victim behaviour/theme immediately preceding an offender behaviour, as was the case in
this study. It would seem likely that offenders do not just consider the immediately preced-
ing victim behaviour when deciding how to act. To account for this would require a more
20 J. Woodhams et al.
Copyright © 2008 John Wiley & Sons, Ltd. J. Investig. Psych. Offender Profi l. 5: 1–23 (2008)
DOI: 10.1002/jip
complex if–then contingency, e.g. ‘if victim struggles and verbally seeks help—then
offender gags the victim’. Such investigations were beyond the scope of the current study
but are clearly necessary.
The investigation of consistency in if–then contingencies within the series was based
on descriptive statistics, and was therefore a very crude measure. On inspection of the
frequency data, it appeared that whilst offenders, on occasion, responded to the same
victim behaviour in the same way, this was limited to a degree of consistency within the
same offence rather than across offences within the same series. The fact that the three
victim behaviours studied only occurred in between 31 and 56% of offences highlights
that not all offences within a series will necessarily contain the same victim behaviours.
This may partly explain why consistency within the same series was not found. To inves-
tigate this further would require studies sampling only offences in a series that contained
the same victim behaviour ‘if’. The frequencies in the current data set were too low to
pursue this. Larger samples may offer more scope to investigate this in a more sophisti-
cated manner.
The question also arises of whether police records of crimes contain suffi cient informa-
tion to determine victim ‘ifs’. Therefore, even if future research found the inclusion of
context in linking crimes to be benefi cial, a lack of detail might prevent this. There are
many reasons why an account held by the police might not represent a complete and
accurate record of a crime. Even if the interviewing of a victim is thorough, he/she may
struggle to recall the exact sequencing of behaviours. This limitation cannot be overcome,
and since it is the victim accounts and police fi les that are used by crime analysts in the
practice of case linkage, fi ndings produced using such data will be of direct relevance to
the work of the user. If victim accounts are not suffi ciently detailed to accurately code for
the if-part of the if–then contingency, research that improves the accuracy of linkage deci-
sions based solely on the then-part of the contingency will be important. Alternatively, a
hierarchical system of linking crimes might help overcome the problem of missing details
(Woodhams, Grant, & Price, 2007). A lower level in the hierarchy could represent the
if(victim behaviour)–then(offender behaviour) contingencies with the next level up
the hierarchy representing just the offender behaviours. If information were missing at the
if–then contingency level, similarity could still be measured at the offender behaviour
level. Clearly this proposition needs testing.
CONCLUSION
Unlike research in personality psychology that has found an association between situa-
tional similarity and behavioural consistency, no supportive evidence for such an associa-
tion was found at the individual crime level. Frequency data also suggested little evidence
of consistency in if–then contingencies within series. It is possible that case linkage
research has already accounted for situational similarity through its investigation of
behavioural consistency across specifi c crime types (e.g. rapes, burglaries, homicides, and
robberies). That serial offences appear to be suffi ciently similar and distinctive in some
behavioural domains for the accurate differentiation of linked and unlinked crime pairs
when focusing solely on the ‘then(behaviour)’ part of the if–then contingency suggests
this might be an appropriate interpretation. Such fi ndings suggest that the more cumber-
some task of coding if–then contingencies for case linkage databases might be unneces-
sary. However, it is certainly too early to conclude this. As illustrated by this paper, there
Incorporating context in linking crimes 21
Copyright © 2008 John Wiley & Sons, Ltd. J. Investig. Psych. Offender Profi l. 5: 1–23 (2008)
DOI: 10.1002/jip
are several avenues of research to pursue and obstacles to overcome in relation to inves-
tigating the utility of incorporating context in linking crimes and, if appropriate, realising
this goal.
ACKNOWLEDGEMENTS
The authors extend their thanks to the Serious Crime Analysis Section of the National
Policing Improvement Agency for allowing access to their database of serial sex offenders.
The authors also extend their thanks to Dr Tim Grant and Ruth Hatcher who read earlier
drafts of this manuscript. This work was completed during a period of study leave granted
by the University of Leicester. The authors would also like to thank the two anonymous
reviewers and the editor for their helpful feedback on the manuscript.
REFERENCES
Alison, L., Snook, B., & Stein, K. (2001). Unobtrusive measurement: Using police information for
forensic research. Qualitative Research, 1, 241–254.
Bateman, A. L., & Salfati, C. G. (2007). An examination of behavioral consistency using individual
behaviors or groups of behaviors in serial homicide. Behavioral Sciences and the Law, 25,
527–544.
Bennell, C., & Canter, D. V. (2002). Linking commercial burglaries by modus operandi: Tests using
regression and ROC analysis. Science and Justice, 42, 153–164.
Bennell, C., & Jones, N. J. (2005). Between a ROC and a hard place: A method for linking serial
burglaries using an offender’s modus operandi. Journal of Investigative Psychology, 2, 23–41.
Block, R., & Skogan, W. G. (1986). Resistance and nonfatal outcomes in stranger-to-stranger
predatory crime. Violence and Victims, 1, 241–353.
Bradley, E. H., Curry, L. A., & Devers, K. J. (2007). Qualitative data analysis for health services
research: Developing taxonomy, themes and theory. Health Services Research (in press).
Burgess, A. W., & Holmstrom, L. L. (1976). Coping behavior of the rape victim. American Journal
of Psychiatry, 133, 413–418.
Canter, D. (2000). Offender profi ling and criminal differentiation. Legal and Criminological Psy-
chology, 5, 23–46.
Canter, D., & Heritage, R. (1990). A multivariate model of sexual offence behaviour: Developments
in ‘offender profi ling’. I. Journal of Forensic Psychiatry, 1, 185–212.
Canter, D., Bennell, C., Alison, L. J., & Reddy S. (2003). Differentiating sex offences: A behav-
iourally-based thematic classifi cation of stranger rapes. Behavioral Sciences and Law, 21, 157–
174.
Cicchetti, D. V. (1994). Guidelines, criteria, and rules of thumb for evaluating standardized assess-
ment instruments in psychology. Psychological Assessment, 6, 284–290.
Davies, A. (1991). The use of DNA profi ling and behavioural science in the investigation of sexual
offences. Medicine, Science and the Law, 31, 95–101.
Davies, A. (1992). Rapists’ behaviour: A three aspect model as a basis for analysis and the identifi -
cation of serial crime. Forensic Science International, 55, 173–194.
Davies, A., Wittebrood, K., & Jackson, J. L. (1998). Predicting the criminal record of a stranger
rapist. Special Interest Series Paper 12. London: Home Offi ce.
Everitt, B. (1980). Cluster analysis. (2nd ed.). Chichester, UK: John Wiley & Sons.
Everitt, B., Landau, S., & Leese, M. (2001). Cluster analysis. (4th ed.). London: Hodder Arnold.
Farrington, D. P., & Lambert, S. (1997). Predicting offender profi les from victim and witness descrip-
tions. In J. L. Jackson, & D. A. Bekerian (Eds.), Offender profi ling: Theory, research and practice
(pp. 133–158). Chichester, UK: John Wiley & Sons.
Fritzon, K., & Ridgway, J. (2001). Near-death experience: The role of victim reaction in attempted
homicide. Journal of Interpersonal Violence, 16, 679–696.
22 J. Woodhams et al.
Copyright © 2008 John Wiley & Sons, Ltd. J. Investig. Psych. Offender Profi l. 5: 1–23 (2008)
DOI: 10.1002/jip
Furr, R. M., & Funder, D. C. (2004). Situational similarity and behavioural consistency: Subjective,
objective, variable-centered, and person-centered approaches. Journal of Research in Personality,
38, 421–447.
Goodwill, A. M., & Alison L. J. (2007). When is profi ling possible? Offense planning and aggression
as moderators in predicting offender age from victim age in stranger rape. Behavioral Sciences
and the Law, 25, 823–840.
Goodwill, A., & Alison, L. (2006). The development of a fi lter model for prioritizing suspects in
burglary offences. Psychology, Crime and Law, 12, 395–416.
Grant, T., & Woodhams, J. (2007). Rape as social activity: An application of investigative linguistics.
In J. Cotterill (Ed.), The language of sexual crime (pp. 1–15). Basingstoke, UK: Palgrave.
Greenfi eld, L. (1997). Sex offenses and offenders: An analysis of data on rape and sexual assault.
Washington, DC: Bureau of Justice Statistics, Offi ce of Justice Programs, US Department of
Justice.
Grubin, D., Kelly, P., & Brunsdon, C. (2001). Linking serious sexual assaults through behaviour.
Home Offi ce Research Study 215. London: Home Offi ce.
Hazelwood, R. R., & Warren, J. (2003). Linkage analysis: modus operandi, ritual, and signature in
serial sexual crime. Aggression and Violent Behaviour, 8, 587–598.
Henwood, K. L., & Pidgeon, N. F. (2006). Grounded theory. In G. Breakwell, C. Fife-Schaw, S.
Hammond, & J. Smith (Eds.), Research methods in psychology (3rd ed.). (pp. 342–364). London:
Sage.
Klein, M. W. (1984). Offence specialization and versatility among juveniles. British Journal of
Criminology, 24, 112–127.
Landis, J. R., & Koch, G. C. (1977). The measurement of observer agreement for categorical data.
Biometrics, 33, 159–174.
Langstrom, N., Grann, M., & Lindblad, F. (2000). A preliminary typology of young sex offenders.
Journal of Adolescence, 23, 319–329.
Luckenbill, D. F. (1981). Generating compliance: The case of robbery. Journal of Contemporary
Ethnography, 10, 25–46.
Mischel, W. (1999). Personality coherence and dispositions in a cognitive–affective personality
system (CAPS). In D. Cervone, & Y. Shoda (Eds.), The coherence of personality: Social cognitive
bases of consistency, variability and organisation (pp. 37–60). London: Guilford Press.
Mischel, W., & Shoda, Y. (1995). A cognitive–affective system theory of personality: Reconceptual-
izing situations, dispositions, dynamics and invariance in personality structure. Psychological
Review, 102, 246–268.
Pervin, L. (2000). Current controversies and issues in personality. New York: John Wiley &
Sons.
Porter, L., & Alison, L. (2004). Behavioural coherence in violent group activity: An interpersonal
model of sexually violent gang behaviour. Aggressive Behavior, 30, 449–468.
Porter, L., & Alison, L. (2005). The primacy of decision-action as an infl uence strategy of violent
gang leaders. Small Group Research, 36, 188–207.
Ritchie, J., & Spencer, L. (1994). Qualitative data analysis for applied policy research. In A. Bryman,
& R. G. Burgess (Eds.), Analysing qualitative data (pp. 173–194). London: Routledge.
Salfati, C. G., & Canter, D. V. (1999). Differentiating stranger murders: Profi ling offender charac-
teristics from behavioral styles. Behavioral Sciences and the Law, 17, 391–406.
Santtila, P., Häkkänen, H., Canter, D., & Elfgren, T. (2003). Classifying homicide offenders and
predicting their characteristics from crime scene behaviour. Scandinavian Journal of Psychology,
44, 107–118.
Santtila, P., Junkkila, J., & Sandnabba, K. (2005). Behavioural linking of stranger rapes. Journal of
Investigative Psychology and Offender Profi ling, 2, 87–103.
Santtila, P., Pakkanen, T., Zappala, A., Bosco, D., Valkama, M., & Mokros, A. (2008). Behavioural
crime linking in serial homicide. Psychology, Crime and Law, 00, 000–000.
Shoda, Y., & Mischel, W. (2000). Reconciling contextualism with the core assumptions of personal-
ity psychology. European Journal of Personality, 14, 407–428.
Shoda, Y., Mischel, W., & Wright, J. C. (1993). The role of situational demands and cognitive
competencies in behaviour. Journal of Personality and Social Psychology, 65, 1023–1035.
Shoda, Y., Mischel, W., & Wright, J. C. (1994). Intraindividual stability in the organization and
patterning of behavior: Incorporating psychological situations into the idiographic analysis of
personality. Journal of Personality and Social Psychology, 67, 674–687.
Incorporating context in linking crimes 23
Copyright © 2008 John Wiley & Sons, Ltd. J. Investig. Psych. Offender Profi l. 5: 1–23 (2008)
DOI: 10.1002/jip
Stubbs, M. (1995). Collocations and semantic profi les: On the cause of the trouble with quantitative
studies. Functions of Language, 2, 23–55.
Thomas, J. (1995). Meaning in interaction. London: Longman.
Woodhams, J., & Grant, T. (2006). Developing a categorization system for rapists’ speech. Psychol-
ogy, Crime and Law, 12, 245–260.
Woodhams, J., & Toye, K. (2007). An empirical test of the assumptions of case linkage and offender
profi ling with serial commercial robberies. Psychology, Public Policy & Law, 13, 59–85.
Woodhams, J., Bull, R., & Hollin, C. H. (2007). The behavioural consistency of serial juvenile sex
offenders and the implications for linking crimes. Paper presented at the Third International
Psychology and Law Conference, Adelaide, July 2007.
Woodhams, J., Grant, T., & Price, A. (2007). From marine ecology to crime analysis: Improving the
detection of serial sexual offences using a taxonomic similarity measure. Journal of Investigative
Psychology and Offender Profi ling, 4, 17–27.
Woodhams, J., Hollin, C. H., & Bull, R. (2007). The psychology of linking crimes: A review of the
evidence. Legal and Criminological Psychology, 12, 233–249.
Wright, J. C., & Mischel, W. (1987). A conditional approach to dispositional constructs: The local
predictability of social behavior. Journal of Personality and Social Psychology, 53, 1159–1177.
Wright, R., & West, D. J. (1991). Rape—a comparison of group offences and lone assaults. Medicine,
Science and the Law, 21, 25–30.
