University of Huddersfield Repository
Canter, David V.
A Multivariate Model of Sexual Offence Behaviour: Developments in 'Offender Profiling'
Canter, David V. (1996) A Multivariate Model of Sexual Offence Behaviour: Developments in
'Offender Profiling'. In: Psychology in Action. Dartmouth Benchmark Series . Dartmouth
Publishing Company, Hantshire, UK, pp. 189-216. ISBN 1855213656
This version is available at http://eprints.hud.ac.uk/9229/
The University Repository is a digital collection of the research output of the
University, available on Open Access. Copyright and Moral Rights for the items
on this site are retained by the individual author and/or other copyright owners.
Users may access full items free of charge; copies of full text items generally
can be reproduced, displayed or performed and given to third parties in any
format or medium for personal research or study, educational or not-for-profit
purposes without prior permission or charge, provided:
• The authors, title and full bibliographic details is credited in any copy;
• A hyperlink and/or URL is included for the original metadata page; and
• The content is not changed in any way.
For more information, including our policy and submission procedure, please
contact the Repository Team at: E.firstname.lastname@example.org.
A Multivariate Model of Sexual
Offence Behaviour: Developments
in 'Offender Profiling'
The extrapolation of characteristics of criminals from information about their
crimes, as an aid to police investigation, is the essence of 'profiling'. This paper
proposes that for such extrapolations to be more than educated guesses they must
be based upon knowledge of a) coherent consistencies in criminal behaviour and b)
the relationship those behavioural consistencies have to aspects of an offender
available to the police in an investigation. Hypotheses concerning behavioural
consistencies are drawn from the diverse literature on sexual offences and a study
is described of 66 sexual assaults committed by 27 offenders against strangers.
Multivariate statistical analyses of these assaults support a five component system
of rapist behaviour, reflecting modes of interaction with the victim as a sexual
object. The potential this provides for an eclectic theoretical basis to offender
profiling is discussed.
Over the last decade a variety of 'profiles' have been created by behavioural
scientists to assist police investigations. Although most have been produced by the
FBI Behavioural Science Unit (Hazelwood, 1983), there have been some notable
successes in the United Kingdom (Canter, 1988). Indeed Hazelwood (1983) claims
an accuracy rate in excess of 80%, and in a more detailed review Pinizzotto (1984)
proposes that suspects were identified with the help of profiling in 46 % of the 192
cases he examined.
Holmes and DeBurger (1988) describe a psychological profile as a report on a
violent crime, utilising information and approaches from various social and
behavioural sciences, intended to assist law enforcement personnel in their
investigations. Other authors are even more vague, writing of a combination of
brainstorming, intuition, and educated guesswork (Geberth, 1983) or a collection
of leads (Rossi, 1982) about an offender. Vorpogel (1982) simply describes a
profile as a biographical sketch of a criminal's behavioural patterns, trends, and
tendencies. The question of how such profiles are produced and what the
190 Psychology in Action
underlying psychological principles are that enable them to be created is given less
emphasis in the existing publications than descriptive accounts of what profiles may
No profile is all inclusive, nor is the same information provided from one profile
to another, being based on what was or was not left at the crime scene and on any
other information that might be available from victim or witness statements, since
the nature and amount of information varies, the profile may also vary. Ault and
Reese (1980), for example, provide the following list of what may be included in
(1) the perpetrator' s race
(3) age range
(4) marital status
(5) general employment
(6) reaction to questioning by police
(7) degree of sexual maturity
(8) whether the individual might strike again
(9) the possibility that he has committed a similar offence previously
(10) possible police record.
But this list omits aspects such as likely style of social interaction, general
personality characteristics, possibility of associated undetected crimes and the very
important matter of possible area of residential location, all of which are aspects
that have been successfully included in University of Surrey profiles (Canter 1988,
The derivation of an account of the perpetrator of a crime from knowledge of the
events associated with a crime is a process open to scientific development. As
pointed out by Canter (1989) it reflects the central psychological questions of how
characteristics of individuals are reflected in their behaviour. In this case the
characteristics are those of value to the police in identifying suspects and the
behaviour is that associated with the crime.
Present focus on sexual offences against strangers
Profiles as an aid to criminal investigation have been produced for a variety of
offences covering homicide, rape and arson (Pinizzotto, 1984). As a starting point
for empirical research, though, serious sexual offences are particularly suitable.
They are crimes in which there is information about a great many of the
perpetrator's actions. These actions in themselves, focusing as they do around
interpersonal sexual aggression, are likely to be revealing of the individual who
The investigation of sexual assaults committed by a person unknown to the victim
are also particularly difficult for criminal investigators to solve. Yet they do account
A Multivariate Model of Sexual Offence Behaviour 191
for a high proportion of sexual assaults. Mulvihill et al (1969) found 53% of
victims reported being raped by a stranger. There may well be important cultural
differences in these figures, because Kocis (1982) reports for Eastern Europe that
55% of the victims are known to the offender and a further 40% of the victims
come into contact with the offender before the crime is committed; only a small
minority of the victims appear to be absolute strangers. In Britain the figures
available are rather closer to those for the U.S.A. Lloyd and Walmsley (1989)
report that in 1973, 49% of rape victims were attacked by strangers. In 1985 the
figure was 40
The studies reported here therefore focus on sexual assaults in which the victim
had no prior knowledge of the offender. Some of these offenders also committed
murders, usually with sexually related aspects. Sexual homicide is therefore
included in the crimes examined.
The need for an eclectic model of offender behaviour
The central hypotheses of profiling, open to direct empirical test, relate to the idea
that offenders differ in their actions when committing a crime and that these
differences reflect (and therefore correlate with) overtly available features of the
offender. However, most published conceptualisations of variations in offender
behaviour have tended to combine accounts of actions in an offence with
explanations of the intentions, motivations and inferred offender characteristics. For
example, a commonly cited approach to rapist typology, Groth's (1979), is
premised on the assumption that rape is not an expression of sexual desire but the
use of sexuality to express power and anger. The typology that is derived from this
perspective, as a consequence, emphasises the various psychological functions that
rape has for the offender not what varieties of action rape actually consists of. A
further example is given by the work of Prentky et al (1985). Their attempts to
characterise and classify rapists makes little distinction between the overt behaviour
as it occurs in the sexual assault and the psychodynamic processes that are taken to
account for or produce that behaviour. There is little attempt to distinguish aspects
of the offender's motivations and life-style from his offending behaviour. Yet any
attempt to understand the actions that occur in the offence requires the classification
of offence behaviour as distinct from classifications of the person in either
psychological or social terms. So although each approach to classification is guided
by a particular explanatory framework any composite modelling of offence
behaviour for use in 'profiling' will have to draw upon all those approaches that are
supported by scientific evidence.
This confusion of action and person is less problematic in the clinical context, in
which earlier theoretical formulations were derived. After all, unlike a police
detective, Groth had actual patients present in interviews when carrying out his
research and his therapeutic mission requires him to enable the offending person to
deal with his actions. Such typologies undoubtedly contribute to the understanding
of the motivations of rapists and this can help to indicate why certain sorts of rapist
192 Psychology in Action
will perform certain types of offence. Yet there remains the primary question of
what variations in offence behaviour can be reliably identified without any
knowledge of the person who committed them. The exploration of how any
empirically validated variations relate to offender characteristics is an important
issue for subsequent examination.
The focus on the perpetrator's actions is not a purely pragmatic requirement
shaped by the limitations of criminal investigation, nor is it naively behaviourist,
assuming that it is only behaviour that is open to scientific investigation. Rather the
emphasis points to the social/interpersonal nature of criminal behaviour, especially
in crimes against the person. It is the variety of actions that happen in sexual attacks
that indicate the different modes of relationship that offenders have with their
victims. Any empirical model of offence behaviour must therefore encapsulate and
explicate these variations in mode of interaction with the victim.
The literature points to a number of aspects of the relationship that a rapist has
with his victim. The most obvious of these are sexuality, and aggressiveness, the
behaviour on which Groth (1979) bases his typology. But other writers, notably
Rada (1978) and Scully and Marolla (1983) point to the fact that many rapes of
strangers are carried out by men who carry out other criminal acts and for whom
rape is one such mode of criminal activity. This perspective indicates that, besides
the sexual acts and the violence, attention also needs to be paid to those aspects of
the offence that relate to its fundamentally criminal nature.
In contrast to the issues of sexuality, aggressiveness and criminality emphasis has
also been given recently to the argument that it is the desire for social contact, or
intimacy that is a primary motivation in rape (Marshall, 1989). Yet it is the
difficulty the offender has in achieving intimacy that leads to an assault. This
perspective may be contrasted with the others, from the viewpoint of profiling
research, in the attention it draws to behaviour that goes beyond physical contact
to attempt some sort of personal relationship with the victim.
The contrast of Marshall's (1989) emphasis on intimacy with Groth's (1979)
focus on power and aggression serves to show that, as logical as each one is, there
is some potential for inherent contradictions between them. At the very least they
raise questions about how a quest for intimacy and the desire for power or
aggression are combined in actual behaviour in actual offences, if at all. Such
possible contradictions, although not as strong, may also be seen in the difference
between the emphasis on the essentially psychopathological nature of sexual
assaults, as argued by those with a clinical perspective (e.g., Groth 1979, Prentky
et aI, 1985), compared with the views of sociologists such as Scully and Marolla
(1983) who see these offenders as essentially normal males operating within
A further contradiction of perspective can also be seen between those such as
Marshall and Groth who emphasise the fact that the attack is based upon
psychological contact with a person, whether for aggressive or intimacy reasons,
and those who argue that rape is fundamentally impersonal (notably Scully and
Marolla 1983). From the latter perspective the victim is an object used to satisfy a
physical craving with whom the offender wishes to have completely 'impersonal
A Multivariate Model of Sexual Offence Behaviour 193
sex'. A view expressed strongly by Symons (1979: 284) is that 'males tend to
desire no-cost, impersonal copulations' .
Broadly, then, at least five modes of interaction with the victim are suggested by
these different perspective. Each of these would be expected to have an observable
counterpart in the actions that happened during an offence. Some of these actions
would be hypothesised as unlikely to co-occur in the same offence, being
In summary they refer to the following elements of sexual offence behaviour:
(2) violence and aggression
(3) impersonal, sexual gratification
(4) criminality, and
(5) interpersonal intimacy.
A number of hypotheses can be derived from this fivefold framework about the
likely co-occurrences of specific behaviour in sexual assaults, given that all of these
types of behaviour may potentially occur in any sexual assault.
One hypothesis is that they all occur with each other in any combination across
a range of assaults. This would suggest that none of the explanations provides a
basis for distinguishing between offences. A completely random combination of any
behaviour with any other would also suggest that there is no consistently coherent
distinction to support empirically the concepts used by each author. This is, in
effect, a null hypothesis that no interpretable relationships will be found between
the actions that occur in offences.
A second hypothesis is that a sub-set of conceptually related actions, (e.g. ,
physical and verbal aggression) will consistently happen together. Any such
grouping would support the perspective related to that behaviour. If, for example,
different forms of aggressive behaviour co-occurred but various attempts at
intimacy were quite independent of each other, then there would be support for
aggressiveness as a coherent salient aspect of sexual assault, but not for intimacy.
In effect, such a result would reduce the number of explanations that are
A third hypothesis is that all of these aspects of offence behaviour can be
identified in details of actual events and that they therefore combine together to
provide a composite model of offence behaviour. Such an eclectic model would be
expected to have an interpretable structure to it. For instance, those types of
behaviour that are associated in the literature, such as the sexual nature of the
offence and its violence (cf. Groth, 1979) would be expected to have some
empirical relationship distinct, say, from the relationship between the criminal
actions and those dealing with the victim as an impersonal sex object (cf. Scully and
Empirical evidence for either the second or third hypothesis would contribute to
scientific support for the possibility of offender profiling because it would indicate
that there are indeed structured variations between offenders, revealed in what they
194 Psychology in Action
did when they committed a crime. Furthermore, such a structure, or system of
behaviours, could be used as the basis for specific hypotheses about the aspects of
behaviour that would be associated with differences between offenders.
This proposal, then, hypothesises that an examination of the behaviours as they
occur in actual sexual offences will reveal a structure that reflects the variety of
modes of interpersonal interactions that underlie those offences. The study reported
here describes an empirical test of that hypothesis.
Relationships between offence and offender characteristics
The study to be reported is part of a series of studies being carried out at the
University of Surrey. The central quest throughout this research is to identify
associations between aspects of the offender's characteristics and offence
behaviour. There are a number of ways in which such associations can be
established, but whatever methods are used they will be more powerful in their
application if they are part of a logical explanatory framework. The framework (or
theoretical stance) adopted in the present paper (first outlined in Canter, 1989) may
be characterised as a cognitive social one, in which the offender's interactions with
others, on a daily basis, is seen as the key to his criminal behaviour.
This is a generalisation of the hypotheses underlying the study by Silverman et
al (1988). They analysed the case records of 1,000 consecutive rape victims seen
at a crisis centre. They found that broad differences in the approach to the victim
were related to many other aspects of the offender and the offence. They showed
that crimes in which an offender used a sudden attack (blitz) and those in which he
used a confidence trick form of access (con) were distinct in a number of ways; the
victims' characteristics, the rape settings, the victims' activities before they were
raped, the assailants' characteristics, and the victims' immediate responses to the
assault. The present study is the first step in elaborating a more detailed
conceptualisation of sexual assaults and their perpetrators.
A study of the structure of offence behaviour
As discussed, the scientific basis to profiling requires an identification of what the
main variations in the actions of offenders in relation to a given offence are. There
are many possible aspects of an offence that may be considered as significant,
especially if there is a victim's account to consider. The present study was an initial
exploration of a range of crimes on which full information was available. It is
therefore of interest as an indication whether future research following this
approach is likely to be worthwhile.
A Multivariate Model of Sexual Offence Behaviour 195
Sample selection and features considered
A total of 33 offence variables were identified through data available such as victim
statements and other police reports, in order to provide a list of categorical
descriptions of the behaviour across all the offences. Behavioural variables with
very low frequencies across the sample were not included since little would be
gained from their inclusion at this feasibility phase of the analysis. Indeed the rare
characteristics may be important for linking offences to one individual, but are
likely to be unique to particular individuals and therefore of less value in
developing general principles. Care was taken to define variables so as to allow an
easy decision to be made as to the category of behaviour. All variables were treated
as dichotomous with no/yes values based on presence/absence of each behaviour
in anyone offence.
The full list of variables, with explanatory elaborations, used to describe offence
behaviour is given in Appendix I in relation to the five modes of interpersonal
interaction discussed above.
Data were collected across 66 offences, made available by a number of English
police forces in response to a request for details of sexual assaults against victims
unknown to the offender. These offences were committed by 27 offenders. The 33
dichotomous variables across the 66 offences provided the data matrix on which
subsequent analysis was conducted.
Smallest space analysis (SSA) of behaviour matrix
These data were subjected to an SSA-I (Lingoes, 1973). In essence, the null
hypothesis is that the variables have no comprehensible relationship to each other.
In other words, it is possible that those offenders who change their actions in
response to the reactions of the victim are not the same as those who talk to the
victim and encourage her to indicate her reactions to the attack. It may be a
common sense assumption that these two variables will relate to each other because
they both indicate a desire to indicate some relationship with the victim, but the
SSA allows a test of this assumption and all the other possibilities suggested by the
relationship every one of the 33 variables has to every other variable.
Although the literature, reviewed above, does suggest a fivefold way of
classifying the variables and this provides a set of hypotheses for the interpretation
of the SSA, the use of SSA also allows the generation of hypotheses both about the
components of the behaviour under study and about the relationships between those
components, the system of behaviour that exists. In other words, the analysis to be
presented may best be regarded as both hypothesis testing and also of heuristic
value in helping to indicate if there are any directions from the results that can be
used to focus future studies aimed at developing profiling.
Smallest Space Analysis (Lingoes, 1973) is a non-metric multidimensional scaling
procedure, based upon the assumption that the underlying structure, or system of
behaviour, will most readily be appreciated if the relationship between every
196 Psychology in Action
variable and every other variable is examined. However, an examination of the raw
mathematical relationships between all the variables would be difficult to interpret
so a geometric (visual) representation of the relationships is produced.
SSA, then, is one of a large number of procedures that represent the correlations
between variables as distances in a statistically derived geometric space. Although
it was first used a number of years ago (Guttman, 1954) only recently have
developments in computers made it readily available for general use. As described
by Guttman (1968), Smallest Space Analysis (SSA) was so called because, when
compared with other approaches to multidimensional scaling, it produces a solution
of smallest dimensionality. This is primarily because it operates on the rank order
of the original correlations rather than their absolute values.
The SSA program computes correlation coefficients between all variables, then
rank orders these correlations. In this case transforming an original rectangular data
matrix into a triangular matrix consisting of correlation coefficients for each
variable as correlated with all other variables. It is these correlation coefficients that
are used to form a spatial representation of items with points representing variables,
the rank order of the distances between points being inversely related to the rank
order of the correlations. Iterations are performed comparing the rank order
assigned to the correlations with the rank order of the distance while adjustments
are made to the geometric representation. The closer the two rank orders the better
the 'fit' between the geometric representation and the original correlation matrix,
or as it is called technically the lower the 'stress'. The iterations continue until the
minimal 'stress' possible is achieved, within the predesignated number of
dimensions. A measure of stress called the coefficient of alienation (see Borg and
Lingoes, 1987 for details) is used within the computing algorithm as the criterion
to use in bringing the iterative procedure to an end. It can therefore be used as a
general indication of the degree to which the variables' intercorrelations are
represented by their corresponding spatial distances. The smaller the coefficient of
alienation, the better is the fit, i.e., the fit of the plot to the original correlation
matrix. However, as Borg and Lingoes (1987) emphasise, there is no simple answer
to the question of how 'good' or 'bad' the representation is. This will depend upon
a complex combination of the number of variables, the amount of error in the data
and the logical strength of the interpretation framework.
In the present case the data is mainly derived from statements taken by the police
from victims. As such they were not collected for research purposes, nor was the
information recorded against a detailed protocol and careful training. Furthermore,
the content analysis of this material was an initial exploratory attempt of the
possibilities for drawing out clear, descriptive variables from this data. It would
therefore be expected that the data was not error free and would contain
considerable 'noise' that would reduce the possibility of interpreting the results. On
the other hand, the published literature is quite rich in suggestions about the
behaviour under study and, as presented above, a reasonably clear set of
distinguishing concepts can be derived. A reasonable fit to the conceptual system
presented would therefore be acceptable, as of heuristic value for future research,
even with a high 'stress' value in the SSA results.
A Multivariate Model of Sexual Offence Behaviour 197
In the SSA configuration, then, in broad terms, the more highly correlated two
variables are, the closer will be the points representing those variables in the SSA
space. Since the configuration is developed in respect to the relationships among
variables and not from their relationship to some given 'dimension', or axis, the
orientation in space of the axes of the resulting geometric representation are
arbitrary, even though the relationships between the points are replicably
determined. Therefore, the pattern of points (regions) can be examined directly
without the need to assume underlying orthogonal dimensions.
The testing of the evidence for ways of classifying variables by examination of
the regional structure of an SSA is part of an approach to research known as Facet
Theory (Canter, 1985). The 'facets' are the overall classification of the types of
variables. The spatial contiguity of the points representing them provides a test of
the major underlying differences amongst these variables as revealed through their
co-occurrence in actual incidents, and is therefore a test as to whether the 'facets'
are empirically supported. The SSA representation therefore offers a basis for
testing and developing hypotheses about the structure of relationships between
offence behaviours. Contiguous behavioural variables, forming an element of an
interpretable facet, provide a productive basis for future research to distinguish
The postulation of facets goes beyond the rather arbitrary proposals of
'grouping', by using the principle of contiguity (Foa, 1958; Guttman, 1965; and
Shye, 1978), which states that because elements in a facet will be functionally
related their existence will be reflected in a corresponding empirical structure. In
other words, variables that share the same facet elements would be more highly
correlated and thus should appear closer together in the multidimensional space than
variables not sharing the same element.
This idea of contiguity can be extended as a general, regional hypothesis. Items
that have facet elements in common will be found in the same region of space.
Likewise, variables which have very low intercorrelations will appear in different
regions of the plot, indicating dissimilarity, and no membership of the same facet
element. Contiguous regionality in a multidimensional space is a quite specific
identification of a facet element, provided a clear statement can be made of what
the variables in that region have in common. Of course, once the exploratory phase
of hypothesis generation has led to the establishment of facets, or when a literature
suggests facets, then the existence of contiguous regions can be used as a strong,
precise test of the hypothesised facets. The usual processes of scientific replication
can also be carried out.
Areas of the SSA plot which contain few or no points are also of interest. Cases
such as this, may indicate weak areas in the data or in fact missing facet elements.
Subsequent studies may then be carried out with new data sets to test for the
existence of these missing elements. In this way the interplay between the formal
theory, as specified in the facets, and the empirical structure, as revealed in the
regional contiguity, can lead to the identification of issues not within the original
set of data.
The approach taken to hypothesis test and generation, then, is to establish
198 Psychology in Action
whether the SSA plot, shown in
1, has any interpretable regional structure
to it. The general hypothesis (null) being tested here is that the variations amongst
offenders as discussed above are so diffuse that no coherent interpretation of the
SSA plot is possible.
Results of SSA
The SSA-I was carried out on an association matrix of Jaccard coefficients, these
being the most appropriate measures of association for this type of binary data. The
3-dimensional solution has a Guttman-Lingoes' coefficient of alienation
apologeltc anal penetratIOn
~ SEX~~,~~Y (
~ \ ~ ~ verbal violence
~ fellatio seQuence
~ fellatIO'" ~
im?"rsonal . .
Lf27jvag,nal penetratIon ImplO8$
. r,nlo 2
IOQUISltlVe ~ ~/
8steahng tom clothIng
r2J1 . ~
demands goods ~ do not tell
SSA of behaviour in
sexuaL assauLts with regionaL interpretations.
(Numbers refer to variabLes in Appendix
LabeLs are brief summaries of content
with 22 iterations, indicating a reasonable fit for this type of data. However, the
interpretation of this configuration turns out to be very close to the regional
structure of the 2-dimensional solution (which has a coefficient of alienation of 0.30
in II iterations) so for simplicity the 2-dimensional structure will be presented.
Figure I shows this projection of the resulting configuration.
For clarity it should be reiterated, each point is a variable describing offence
behaviour. The numbers refer to the variables as listed in Appendix I, although for
simplicity a brief title for the variable has also been placed on the plot. The closer
A Multivariate Model of Sexual Offence Behaviour 199
any two points the more likely are the actions they represent to co-occur in
offences, in comparison with points that are farther apart.
Focal aspects of rapes
A first stage in the interpretation of Figure 1, to test the hypotheses and explore the
structure of offence behaviour, in the present case, is to consider the frequency of
occurrence of each of the variables. Because they are all binary indications of the
occurrence of actions each variable has an associated frequency in the whole
sample. The SSA is derived from the associations between the variables and so has
no inevitable link to their overall frequency. However, with dichotomous variables
Figure 2 SSA of 66 assaults illustrating one offence as circles, one as arrows.
(Numbers refer to variables in Appendix I.)
it is possible for those variables that are frequent to have higher associations with
each other, unless there is one or more subsets of variables with lower frequencies
that co-occur with high probability. The relationship, then, between the frequencies
of the actions and the SSA structure is not artifactual. It is an empirical one open
to some substantive meaning. Figure 3 presents the frequencies of occurrence of
every offence action.
As can be seen it is possible to draw very clear contours on this diagram to cover
variables that occur in more than 65% of cases, 40% to 60%,25% to 35%, 20%
Psychology in Action
and less than 15
of cases (Figure 3). This polar sequence lends strong
support to the focal (polarising) nature of the high frequency variables. It also
indicates those actions that differentiate between offences, being at the edge of the
plot. As discussed later, the identification of a high frequency core and a polarising
sequence from it also opens up the possibility that the activities around this circular
structure coalesce along radii, creating wedges of modes of interaction, relating
directly to the decreasing frequency of those particular sets of variables.
Percentage frequency of assault behaviour indicated on SSA configuration
with equal frequency contours.
The frequencies serve, also, as a heuristic summary of offence behaviour,
showing that those behaviours further out from the core are the ones that are most
distinct, giving any particular offence its specific characteristics.
The hierarchy of frequencies indicate that there are certain activities that are
conceptually central to rape, in other words, at the core of sexual assault. The
activities at the outer rim reflect different aspects of the same overall phenomena,
differing in their reference to some common focus.
This focus and the referents that make it up can be given a clearer meaning by
considering those items at the centre of the plot. Such items share most with all the
others around them and so are both literally and metaphorically central to the issues
being examined. In Figures I, 2 and 3 the following variables are central:
A Multivariate Model of Sexual Offence Behaviour 201
vaginal intercourse (27)
no reaction to the victim (8)
impersonal language (11)
surprise attack (2)
victim's clothing disturbed (13)
This core is not the overtly aggressive set of actions that Groth's (1979) typology
would emphasise. It includes sexual intercourse, but not the variety of sexual
activity that might have been expected if sexual desire was a totally dominating
aspect of the offence. Nor are those variables central that indicate a desire to relate
to the victim as might be indicated from Marshall's (1989) considerations. The
discussion that best fits these variables is that of Scully and Marolla (1983),
indicating an impersonal, surprise attack in which the victim's response is irrelevant
to the offender. The five variables here could be regarded as the sine qua non of
a sexual assault; dealing with the victim impersonally with a surprise attack
disturbing her clothes and having vaginal intercourse. Their position at the centre
of the plot, therefore, does add credibility to the whole structure and shows that the
use of a woman as a sexual object is at the core of sexual assault.
Modes of interaction with the victim
The results allow a further development of the idea of sexual assault being
essentially an interaction with a woman as an object, by the identification of the
other related regions in the plot. Because there is an interpretable core to this plot
it is appropriate to consider the various emphases that can be given to this focus by
examining the variables around the plot. This allows a further test of the hypotheses
derived from the published literature, by exploring whether any of the emphases
in the literature have empirical support and, if they do, how they may relate to each
Attempted intimacy with the victim
As noted earlier, some discussions of rape suggest that it is the lack of ability to
form intimate relationships with women that is an important aspect of the
motivation to rape (Marshall, 1989). If this were ever dominant in rape then it
would be expected that those actions that would indicate an attempt at intimacy, or
at least some preparedness to relate to the victim as a person rather than an object,
would co-occur in some rapes. Five variables, particularly, indicate that the
offender is attempting to, or at least not deterred from, entering into some sort of
personal relationship with the victim:
the victim's reaction influences/deters the offender (variable 7)
the offender requires the victim to participate verbally during the assault (17)
the offender requires the victim to participate physically during the assault (18)
202 Psychology in Action
the approach is one of a confidence trick (1)
the offender is inquisitive about the victim (10)
offender compliments the victim (9)
the offender apologises to the victim (33)
These seven variables can be found in the lower left quadrant of the plot. There is
therefore some support for this interpersonal aspect of the offence being a coherent
and possibly significant feature of offence behaviour.
This distinct aspect of offence behaviour provides an initial heuristic for
generating hypotheses of associations between offender and offence. Those offences
in which most of these five actions happen, especially those actions which are less
frequent in the present sample, would be hypothesised to be correlated with
significant aspects of an offender's interpersonal background. For example,
following Marshall's (1989) arguments, it would be hypothesised that these
offenders would have had difficulty in formulating intimate relationships with
women, but that they may well have attempted this. For example, marriage to a
younger woman with a very short courtship and subsequent distancing in their
relationship would be predicted. Considerably more research, however, is needed
both to replicate this facet element and to establish the hypothesised correlations.
Although it is often underemphasised in clinical accounts of rape, there can be little
debate about the fact that sexual activity is a crucial component of the attack. There
is therefore an important question about whether the different types of sexual
behaviour are related and form distinct constituents of an attack or whether they are
diffuse, related more to other aspects of the offender's behaviour.
Six variables dealt specifically with the sexual behaviour:
vaginal intercourse (27)
fellatio initially (28)
fellatio as part of the attack sequence (29)
initial anal intercourse (31)
anal intercourse as part of the offence sequence (32)
The top left quadrant of the plot contains all these variables.
It is interesting to note that vaginal intercourse is central to the whole plot, so that
the upper left quadrant is defined by the other sexual activities.
The sexual variables, then, do form a region indicating that together they provide
an aspect of offence behaviour that needs to be considered further. This accords
with the arguments of Scully and Marolla (1983) that the desire for certain sorts of
sexual experience is a significant facet of rape, leading to hypotheses that when a
variety of sexual activities take place the offender may have be found to have either
considerable earlier sexual experience or a great interest in such experience as
A Multivariate Model of Sexual Offence Behaviour 203
revealed through a collection of pornographic material. In general, however, the
existence of a distinct region in the SSA configuration indicates that the actual
sexual aspects of a sexual assault should not be undervalued, as some authorities
have tended to do.
Overt violence and aggression
The clinical literature on sexual offences usually places most emphasis on the fact
that these are aggressive violent attacks. Violence therefore seems to be a salient
issue for consideration. Groth (1979), in particular, as discussed above, argues that
aggression is a primary motivation in sexual assault. The following four variables
dealt directly with overt violence and aggression:
violence used as means of controlling the victim (24)
violence used, but not as a means of control (25)
aggressive verbal behaviour (26)
insulting language (12)
All four variables are to be found in the top right quadrant of Figure 1. These
therefore appear to be coherent aspects of the offence behaviour. For some
offenders, then, this is a distinct aspect of their offending. However, it is a distinct
aspect which is not overtly apparent in many offences. Their relationship to prior
history of aggressiveness is worthy of exploration.
The adjacency of these aggression variables to the sexual variables is of interest,
showing that they are quite likely to be linked, as the clinical literature suggests. In
particular the closeness of the violence not used as control variable (25) and the
anal intercourse variables (31 and 32) does indicate that this particular form of
sexual act is likely to be associated with violence and may indeed be motivated by
similar psychological processes.
The antithesis of the actions that indicate the offender is trying to relate to a person
(as presented in (1) above) are the actions that treat the victim very much as an
object, dealing with her as an entity entirely for the criminal's use. Six variables
relate to these aspects of the offence:
'blitz' attack (3)
impersonal language (11)
no response to the victim's reactions (8)
surprise attack (2)
tearing of victim's clothing (14)
victim's clothing disturbed by offender (13)
These six variables are all to be found in the middle right segment of the plot. They
204 Psychology in Action
therefore provide a graphic combination of quite wide ranging behaviours
indicating the offender's callous disinterest in his victim. They do, however,
include a number of variables that were identified as having a high frequency and
at the core of sexual assaults. The existence of a distinct region indicates that for
some offenders this may be the dominant characteristic of their offending even
though overall offender behaviour is somewhat biased to this aspect of assaulting,
as reflected in the off-centre position of the 'core' actions.
The variable recording that the offender implies knowledge of the victim (20) is
also at the edge of this region. This is difficult to interpret at this stage but possibly
implies that the offender had prior knowledge of the victim, having identified her
as a desirable object.
The distant, impersonal contact with the victim, indicated by this sub-set of
variables is hypothesised to be a reflection of a general approach to women that
would be apparent in the offender's daily life. He would be predicted to be known
as someone who does not regard women as experiencing the world in the same way
he does, seeing them as vessels for men's desires. Clearly, such a perspective on
an offender has implications for approaches to therapy as well as for assisting
Criminal behaviour and intent
Rapists often also operate as criminals committing crimes not obviously sexually
motivated. There are also a number of aspects of their sexual crimes that have non-
sexual but still distinctly criminal components, such as the wearing of a mask to
hide the offender's identity, or the carrying of a weapon to the crime scene. The
question therefore arises as to whether these actions have some relationship and
There are seven variables that can be interpreted as reflecting criminality:
the use of bindings (5)
the use of gagging (6)
stealing from the victim (22)
the use of some form of disguise (19)
blindfolding the victim (4)
demanding goods (16)
controlling the victim with a weapon (15)
Some of these actions may be considered as more directly related to sexual sadism,
notably binding and gagging, with the implication that the offender obtained some
sexual gratification from these actions. That, of course, may be true in some
offences, but the position of these actions in the SSA configuration near to wearing
a disguise and carrying a weapon indicate that in the present sample the behaviour
of binding and gagging is more readily associated with criminality of the actions.
The position of these variables on the opposite side of the plot from the sexuality
and violence variables also supports the proposal that, within these offences at least,
A Multivariate Model of Sexual Offence Behaviour 205
the behaviours do not indicate sexual sadism.
All seven of these criminality variables, then, are in the same regional segment
of the plot at the bottom right. Interestingly two other variables are clearly within
the same region; the threat to the victim not to tell anyone about the offence (21),
and the informing of the victim that she is known by the offender (23). It is not
difficult to conceptualise these actions as part of the criminal repertoire.
Taken together these nine variables indicate an emphasis to a criminal's
behaviour that, if present to any degree, would be hypothesised to correlate with
aspects of his previous criminal offences. In particular, it is hypothesised that
offenders who commit many of these actions are likely to have extensive history of
Summary and conclusions
In order to establish whether it is possible to derive the characteristics of a criminal
from his actions when offending ('offender profiling') it was proposed that it was
first necessary to demonstrate that the behaviour of offenders during a crime had
some comprehensible coherence to them. Focusing on sexual assaults against
strangers, theoretical accounts of variations between offenders were reviewed in
order to establish the range of offence behaviours that should be examined. This
review revealed that there were a variety of proposals as to the differences between
sexual assaulters, some of these being contradictory.
From these considerations five aspects of a sexual assault were identified. These
aspects provided a set of hypotheses about the behaviours that would co-occur
during an assault. To test these hypotheses 66 offences, committed by 27 offenders
were content analysed into 33 behavioural categories. The occurrence of these
categories of behaviour across all offences was examined using SSA-I.
The results of the SSA indicate that sexual assault can be understood as various
ways of carrying out sexual acts in an impersonal fashion, treating the unwilling
victim as an object. The results, further, lend support to all five different aspects
identified from the published literature. They show that the various explanations
given may be construed as different emphases of an assault, different ways of
engaging in rape, any offence drawing on one or more aspect.
This set of aspects of elements that make up a rape attack form a circular order
in the SSA space. This implies that those regions which are closer together are
more likely to contain actions that occur in the same offence. The sequence around
the plot is therefore of some substantive, theoretical interest.
Broadly, the actions on the top half of the SSA deal with actions of interest, and
often noted by those with a psychopathological perspective on rape such as Groth
(1979). They cover the variety of sexual activities and the aggressive acts. As such
they cover actions that may be akin to the expressive aggression to which Prentky
et al (1985) draw attention. Their concept of 'instrumental aggression' may be
more closely aligned with the behaviours in the lower half of the plot. These actions
are also more in accord with those perspectives that take sociological or social
206 Psychology in Action
psychological perspectives (e.g., Scully and Marolla, 1983), covering actions that
are very impersonal, criminally oriented. Interestingly the social psychological view
of Marshall (1989) falls between these two regions.
The closeness of the sexual activity to the violence does lend credence to the view
that many sexual offences have a strongly violent aspect. The adjacency of this
violence to the impersonal behaviour does reflect rather well the ways in which the
one can merge into the other. The criminal behaviour is adjacent to the impersonal
behaviour, reflecting the likelihood that criminality is indeed antisocial in the strong
sense that it relates to an unpreparedness, or inability, to relate to other people.
However, there are offences in which this inability is shown through the
inappropriate attempt to form a relationship with the victim, as revealed by the
adjacency of the 'intimacy' region. That such actions should be next to the sexual
behaviour is also logical in that, on occasion, that implies overlapping motivations.
The sequence of types of activity shown in Figure I, from sexual activity through
criminal behaviour back to sexual activity again is a circular sequence. No simple
linear dimension running from one obvious extreme to the other can be identified.
This leads to the hypothesis that all five types of activity represent different
emphases of the same overall phenomena, rather than providing positions along a
continuum. A further test of the validity of this hypothesis of a circular order of
activities is provided by considering the frequencies of the actions, as indicated in
Figure 3. As has been noted, although there is no statistical necessity for the higher
or lower frequencies to be found in a specific region of the plot, it none the less
turns out to be the case that th~ lower the frequencies the further are the actions
from the centre of the configuration.
Yet quite independently of the frequencies, it is the nature of an SSA
configuration that those actions at the centre of the configuration are the ones that
empirically have most in common with each other. Those at the periphery are the
most functionally discrete. Therefore, in the present results, as action frequencies
become lower so those actions became more distinctly part of the most functionally
specific regions of the plot. This means that there are some actions that have a lot
in common with each other, providing a core to these sexual assaults, whereas
other actions reflect more specific emphases for those assaults. In other words, the
distribution of the frequencies support the hypothesis of a coherent system of
behaviour that has different emphases to it, adding weight to the validity of the
The overall combination of the frequencies and the radial elements (a 'radex',
Guttman. 1954) all therefore reflects different aspects of the same overall
phenomena, differing in their reference around the focus of the victim being treated
as a sexual object by the offender, yet in a variety of different ways.
This radex model of sexual offending has a number of heuristic values. For
example, examination of it indicates gaps in the empirical space where no variables
are found. It can be hypothesised that the current sample does not include actions
which would have gone into those locations. Future data can be used to test this
hypothesis. One such instance is the occurrence of overtly sadistic aggressive
behaviour in which the victim is bound as a form of humiliating and is arousing to
A Multivariate Model of Sexual Offence Behaviour 207
the offender. There is no indication that any such extreme form of sexual sadism
was present in the sample. If it were it is hypothesised that an SSA would place it
at the extreme edge of the plot in the top region of the aggressive behaviour, but
close to the boundary with the sexual behaviour. Other researchers will be able to
derive many other similar hypotheses.
A further heuristic value derives from the fact that all five aspects of sexual
assault contribute to all offences, but there is likely to be different combinations of
constituents for different individuals. The differences in an offender's repertoire of
offence behaviour can therefore be clearly established. Establishing what these
combinations of major classes of behaviour are is an important objective for future
research. It could be used for considering how an offender develops over a series
of offences or more pragmatically for establishing whether two or more offences
were committed by the same person.
Each of these five aspects also has the potential of correlating with different sets
of characteristics of offenders. Study of these correlations will be a major step in
the furtherance of a scientific base to offender profiling.
208 Psychology in Action
Appendix I: Variables used to describe offender's behaviour during an offence
as derived from content analysis of victim statements
Coding of variables
Thirty three offence variables were created from a content analysis of available
police records and victim statements in order to provide a list of elements common
to offences. Variables with a very low frequency were not included. Care was taken
to describe the definition of variables so as to eliminate discrepancies in category
assignment. All variables are dichotomous with values based on the
presence/absence of each category of behaviour. A description of the categorisation
scheme is given below.
The style of approach used by the offender in which any ploy or subterfuge is used
in order to make contact with the victim prior to the commencement of the assault:
this would include any verbal contact - questions asked, false introductions, story
Variable 2. Surprise attack
The immediate attack on the victim, whether preceded by a confidence approach
or not, where force is used to obtain control of the victim: force in respect of this
variable includes threat with or without a weapon. Violence is for the physical
control of the victim, i.e., exercised against the victim in order to render her
available to the offender, but not the actions covered in variable three.
Variable 3. Blitz attack
The sudden and immediate use of violence, whether preceded by a confidence
approach or not, which incapacitates the victim: typically this is the sudden blow
which leaves the victim unable to respond or react to the attack. This variable
focuses on the extreme violence of the initial assault which leaves the victim
incapable of reaction.
Variable 4. Blindfold
The use at any time during the attack of any physical interference with the victim's
A MultivariateModel of Sexual Offence Behaviour 209
ability to see: this only includes the use of articles and not verbal threat or the
temporary use of the offender's hands.
As above in respect of the use of articles to disable the victim: the categorisation
does not include the possible situational effect of partial stripping of the victim, nor
the temporary use of manual control of the victim.
As above in respect of the prevention of noise: this does not include the manual
gagging of victims commonly associated with the attack variables.
One of two reaction variables, to examine how the offender copes with, or reacts
to, active victim resistance: the resistance of the victim can be verbal or physical
but does not include the act of crying alone. The categorisation addresses the
offender and not the victim. This variable assigns a 2 to the offender who is
deterred or, who changes or negotiates his intended actions upon victim reaction.
The category emphasises the change or negotiation of any act as a result of victim
As above, but this variable categorises those offenders whose action and/or
intentions are not changed by victim resistance; this offender will continue the
assault against an actively resisting victim. An offence in which the victim offers
no resistance will be found in the categories of I on both variables 7 and 8.
The first of five variables concerned with the complexities of what is said by the
offender to the victim. This is not necessarily the result of verbal interchange but
is focused on the style of speech used by the offender, in the non-violent context.
This variable assigns a category of 2 to those offences in which the offender
compliments the victim, usually on some aspect of her appearance.
210 Psychology in Action
Variable 10. Language (2) Inquisitive
The second language variable categorises the offender's speech in being inquisitive
of the victim. This includes any questions asked about the victim's life-style,
associates etc. There are other variables which deal with the identifying of the
victim and the requirement, for example, of the victim to participate in the acts
committed against her. This therefore focuses on the questions asked of the victim
which are those of a non sexual nature.
This language variable categorises those aspects of the offender's
impersonal/instructive dealings with the victim. The focus is the impersonal style
of the offender rather than the categorised differences between personal/impersonal.
The personal style of speech will be shown in one or more of the other language
Variable 12. Language (4) Demeaning/insulting
A non-violence language variable which categorises offender's speech with or
towards the victim that is demeaning and/or insulting: this would include profanities
directed against the victim herself or women in general.
The focus of this variable is the insult and not sexually orientated comment.
Variable 13. Victim clothing disturbed
One of two clothing variables: this categorises the offender's removal of the
victim's clothing himself. The alternative category, i.e., category I, includes the
act of disrobing carried out by the victim. This act is always at the instruction of
the offender and therefore the same category is used in the circumstances of a
naked or semi-naked victim. The focus of this variable is on the actions of the
offender and can be seen in comparison with the activities of the offender in the
second clothing variable (14). It categorises any act of removal by the offender as
2, regardless of whether the victim assisted or not.
Variable 14. Victim clothing cut/torn
This variable addresses the offender's removal of clothing by particular methods.
Although there are obvious differences in the tearing or cutting of clothing, this
category deals with the offender who is prepared to use an apparently more violent
A Multivariate Model of Sexual Offence Behaviour 211
style in his treatment of the victim. Category 1 covers the disturbance of clothing
as well as the undressed victim. The focus is on the removal of clothing and not
what the offender does with it after removal.
Variable 15. Control weapon
1= Threat 2 = Weapon
The categories differentiate those offenders who are prepared to display a weapon
in order to control the victim, from those who do not. Threat of the possession of
a weapon and threat of physical presence are coded as '1'.
Variable 16. Demand goods
This variable categorises the offender's approach to the victim that includes a
demand for goods or money. Importantly the demand categorised in this context is
that which is made in the initial stages of the attack. A later variable deals generally
with stealing from the victim (V22).
Variable 17. Victim participation verbal
There are two variables dealing with the requirement of the victim to participate in
the offence. Both have been found to occur at the instruction of the offender. Those
instructions may appear in many forms, therefore this categorisation deals with the
offender's requirement that the victim say words or phrases to him at his insistence.
The category does not cover the occasions where an offender directs a question to
the victim which does not appear to require her to answer.
Variable 18. Victim participation acts
As above, but this is categorised to cover the offender's requirement that the victim
physically participate. The acts demanded of the victim are those which may be in
association with specific sexual demands made of her but are in addition to those
sexual acts. Therefore an example may be the requirement made of the victim to
kiss the offender, or to place her arms around him.
In other words, it focuses on the requirements that the victim participate in any
act committed against her; in this context the expectation is to differentiate between
those offenders who will commit, say, fellatio against the victim and those who
commit the same act but accompanied by instructions to do specific acts associated
with oral sex.
212 Psychology in Action
Variable 19. Disguise
Various disguises can and are worn by offenders, categorically the definition of
them all would result in an unwieldy variable. The category of disguise in this
variable, therefore, deals with those offenders who wear any form of disguise.
Variable 20. Implied knowledge
Instances occur within the attacks, at various times, in which the offender implies
knowing the victim. This categorisation records the implication that the offender
knew or knew of the victim before the sexual assault.
Variable 21. Threat ... No Report
This is specific categorisation of the verbalised threat made to a victim that she
should not report the incident to the police or any other person. This may take
many forms however the specific threat against the victim in this context is plain
Variable 22. Stealing
The general category of stealing differentiate those offenders who do steal from
those who do not.
Variable 23. Identifies victim
This categorisation covers offences in which offenders take steps to obtain or
attempt to obtain from the victim the details which would identify her. This may
take many forms including verbal approaches, the examination of personal
belongings before or after the actual sexual assault, or indeed the stealing of
personal identifying documents following the assault. The act is complete if the
offender acts in any way that allows him to infer to the victim that he has, or can,
Variable 24. Violence (1) Control
This categorisation of 'violence to control' identifies the use of force which is more
than the physical control of the victim and which, situationally, is not the initial
attack to obtain control of the victim.
A Multivariate Model of Sexual Offence Behaviour
The category in this variable describes the punching, kicking etc of the victim in
order to reinforce the control the offender is seeking to exercise on the victim.
This categorisation here deals with the offender who is prepared to use excessive
violence in retaliation to perceived resistance or, in some cases, the use of violence
apparently for its own sake.
This variable is distinct from those dealing with the speech types directed at the
victim which can be categorised as impersonal (VII), or demeaning (VI2). The
categorisation in this variable is to address the use of intimidating language in the
form of threats to maim or kill which are not necessarily associated with control or
resistance. Focus is therefore on verbal violence which is not associated with
control or resistance.
Variable 27. Vaginal penetration
This variable covers whether vaginal penetration was achieved or attempted.
This is one of two variables dealing with the forced oral penetration of the victim.
The categories of this variable deal only with whether oral penetration was carried
out or attempted.
The second variable of fellatio categorises offenders' requirements that their victims
submit to oral penetration and are those whose performance of the act is part of a
sequence of sexual acts. The offender who does not engage in the oral penetration
of the victim will be identified as being categorised as 'I' in both variables 28 and
29. Similarly the offence in which only oral sexual activity occurs will be
differentiated by being categorised as '2' in variable 28, and' I' in the other sexual
214 Psychology in Action
Variable 30. Cunnilingus
This variable deals with the performance of a particular sexual act committed
against the victim's genitalia by the offender's use of his mouth. In the present
sample there is no sequential variable in this context as to date no cases have been
seen where this act is performed alone. There is always some other sexual activity
accompanying the act of cunnilingus.
Variable 31. Anal penetration
This is one of two variables dealing with penetration per anus committed against
a victim. This categorisation deals only with those cases where the act was carried
out. In the present sample categorised cases the penetration is by male organ only.
It includes attempts where there is clear indication of intent.
Variable 32. Anal penetration in sequence
The second variable dealing with anal assault: the category addresses anal assault
in sequence with other sexual acts. The offence in which anal penetration occurs
or is attempted will be categorised' I' on both variables. Similarly the attacks with
an anal assault only will be categorised as '2' in variable 31, and' I' in respect of
the other sexual act variables.
Variable 33. Apologetic
This is a further language variable to deal with the specific apologetic speech used
by an offender, most typically at the end of a sexual assault.
We are grateful to Milena Kovacik and Katherine King-Johannessen for their
contribution to this paper. This research benefited from Home Office Contract No.
and support from Surrey Police Constabulary.
Reese, J.T. (1980), 'A Psychological Assessment of Crime
Profiling', FBI Law Enforcement Bulletin, Vol. 49(3), pp. 22-5.
Borg, I. & Lingoes, J. (1987), Multidimensional Similarity Analysis, Springer
A Multivariate Model of Sexual Offence Behaviour 215
Verlag, New York.
Canter, D. (1985), Facet Theory: Approaches to Social Research, Springer-Verlag
Canter, D. (1988), 'To Catch a Rapist', New Society, 4 March.
Canter, D. (1989), 'Offender Profiling', The Psychologist, Vol. 2, pp. 12-16.
Davies, G .M. (1988), 'Use of Video in Child Abuse Trials', The Psychologist,
Vol. 1, pp. 20-2.
Foa, U.G. (1958), 'The Contiguity Principle in the Structure of Interpersonal
Relation', Human Relations, II, pp. 229-38.
Geberth, V.J. (1983), Practical Homicide Investigation, Elsevier, New York.
Groth, N.A. (1979), Men Who Rape: The Psychology of the Offender, Plenum,
Weihofen, H. (1952), Psychiatry and Law, Norton, New
Guttman, L. (1954), 'A New Approach to Factor Analysis: The Radex'. in
Lazarsfield, P.F. (ed.), Mathematical Thinking in the Social Science, Free
Guttman, L. (1965), 'A faceted Definition of Intelligence', Studies in Psychology,
Scripta Hierosolymitana, Vol. 14, pp. 166-8, Hebrew University,
Guttman. L. (1968), 'A General Nonmetric Technique for Finding the Smallest
Coordinate Space for a Configuration Point', Psychometrika, Vol. 33 (3),
Hazelwood, RR. (1983), 'The Behaviour-oriented Interview of Rape Victims: The
Key to Profiling', FBI Law Enforcement Bulletin, pp. 1-8.
DeBurger, J. (1988), Serial Murder, Sage, London.
Kocis, L. (1982), 'Posudzovanie sexualnych delikventov', in Forenzne Aspekty
Sexuality, Sexualita zeny, Zbornik referatov: Kosice, September, pp. 186-
Lingoes, J. (1973), 'The Guttman-Lingoes Non-metric Program Series', MA
thesis, University of Michigan.
Walmsley, R (1989), Changes in Rape Offences and Sentencing
(Home Office research study No. 105), HMSO, London.
Marshall, W.L. (1989), 'Intimacy, Loneliness and Sexual Offenders', Behavioural
Research in Therapy, Vol. 27(5), pp. 491-503.
Mulvihill, D., Tumin, M.
Curtis, L. (1969), Crimes of Violence. A Staff Report
Submitted to the National Commission on the Causes and Prevention of
Violence, Vols. 11-13, US Government Printing Office, Washington DC.
Pinizzotto, AJ. (1984), 'Forensic Psychology: Criminal Personality Profiling',
Journal of Police Science and Administration, Vol. 12(1), pp. 32-40.
Prentky, RA., Cohen, M.L.
Seghorn, T. (1985), 'Development of a Rational
Taxonomy for the Classification of Sexual Offenders: Rapists', Bulletin of
the American Academy of Psychiatry and the Law, Vol. 13, pp. 39-70.
Rada, R.T. (1978), 'Psychological Factors in Rapist Behaviour', in Rada, R.T.
(ed.), Clinical Aspects of the Rapist, Grune and Stratton, New York.
216 Psychology in Action
Rossi, D. (1982), 'Crime Scene Behavioural Analysis: Another Tool for the Law
Enforcement Investigator', Official proceedings of the 88th Annual IACP
Conference, The Police Chief, January, pp. 156-59.
Marolla, J. (1983), 'Incarcerated Rapists: Exploring a Sociological
Model', Final Report for Dept. of Health and Human Services, Rockville,
Shye, S. (1978), Theory Construction and Data Analysis in the Behavioural
Sciences, Jossey Bass, San Fransisco.
Silverman, D.C., Kalich, S.M., Bowie, S.1.
Edbiel, S.E. (1988), 'Blitz Rape and
Confidence Rape: A Typology Applied to 1,000 Consecutive Cases',
American Journal of Psychiatry, Vo1.145(1l), pp. 1438-41.
Symons, D. (1979), The Evolution of Human Sexuality,Oxford University Press,
Vorpogel, R.E. (1982), 'Painting Psychological Profiles, Charlatism, Charisma, or
a New Science?', The Police Chief, January, pp. 156-9.