Linkage Analysis for Crime Research Paper

This sample Linkage Analysis for Crime Research Paper is published for educational and informational purposes only. If you need help writing your assignment, please use our research paper writing service and buy a paper on any topic at affordable price. Also check our tips on how to write a research paper, see the lists of criminal justice research paper topics, and browse research paper examples.

Police investigators must often determine whether multiple crimes have been committed by the same offender. In ideal situations, this decision is based on an analysis of physical evidence left at crime scenes, such as DNA, fabric fibers, and/or fingerprints. However, despite what is portrayed in the popular media, such evidence is not always available to be processed (Davies 1991). Given this, the police have had to establish alternative methods for linking serial crimes. One of the most commonly used approaches is behavioral linkage analysis.

When using this form of analysis, an attempt is made to link crimes based on the behaviors that offenders engage in while committing their offenses. Specifically, the goal is to identify patterns of behavior across an offender’s crimes that meet two criteria: behavioral stability and behavioral distinctiveness (Canter 1995). Behavioral stability exists when offenders behave in the same or similar way across their crime series (i.e., high levels of within-series similarity). Behavioral distinctiveness exists when the behaviors exhibited by one serial offender are different from those exhibited by other offenders committing similar types of crimes (i.e., low levels of between-series similarity). When offenders behave in a relatively stable and distinct fashion, it may be possible to link them to their crimes and to differentiate between crimes committed by different offenders (Bennell et al. 2009).

In the investigative setting, linkage analysis is most often carried out by crime analysts or police officers who have specialized training. Generally speaking, there are two ways that the linking task is approached (Woodhams et al. 2007). Proactive linking involves attempts to determine whether a new crime series can be identified by examining the similarities and differences that exist between unsolved crimes archived in a large database. In contrast, reactive linking typically involves attempts to determine whether unsolved crimes can be linked to a particular offense of interest to the police (it may also involve attempts to determine whether a specified set of crimes are the work of a common offender). In some cases, the perpetrator of an index crime may already be known to the police, and the task is to determine whether any unsolved crimes are the responsibility of that particular offender.

Although there is no one method for conducting behavioral linkage analysis, many approaches consist of the following steps: (1) searching for crimes that share similar behavioral features; (2) isolating all the similarities and differences between the crimes that are identified; (3) evaluating the importance of those similarities and differences by, for example, considering the base rates of the behaviors within larger samples of offenses (given that frequently occurring behaviors are unlikely to be useful for distinguishing between crimes committed by different offenders); (4) determining the likelihood of actual crime linkages; and (5) reporting the results of the analysis to investigators (Woodhams et al. 2007).

Historical Developments

The idea that an offender’s behavior can be useful for linking crimes is not new. In Gross’s (1906) classic book, Criminal Investigation, he highlighted the potential value in using an offender’s modus operandi, or MO, to link his crimes. Even before this, police agencies in England and Wales had begun developing sophisticated systems for categorizing MOs for the specific purpose of linking crimes (Fosdick 1915). For example, around the turn of the century, Chief Constable William Atcherley of the West Riding Yorkshire Constabulary developed a coding scheme for burglary MOs that increased the ease with which these crimes could be compared to one another. This system was further refined in North America to allow for more detailed comparisons (Vollmer 1919).

Such coding systems, which were used throughout the early 1900s, assumed that offenders exhibit behavior in a highly stable fashion across their crimes. However, thinking around this issue gradually changed. Investigators began to realize that an offender’s MO can vary across his crimes for a number of reasons, including learning, maturation, and situational factors (Douglas and Munn 1992). This led to a search for crime features that would remain more stable over time. Law enforcement professionals began to distinguish between three related but distinct constructs: MO, ritual, and signatures (Hazelwood and Warren 2003). MO refers to functional behaviors that are required to successfully commit a crime. Ritual refers to fantasy-based behaviors, which are symbolic in nature and reflect the psychological needs of an offender. Finally, signatures refer to combinations of behaviors (MO or ritual) that are assumed to be relatively stable and unique to each offender. Despite the lack of empirical research to support their use, by the 1990s, linking crimes based on behavioral signatures became a popular approach (e.g., Keppel 1995).

In order to systematize the analysis of an offender’s crimes, including their MO and ritual behaviors, police professionals have historically used charts that allow them to compare common and distinctive features exhibited across a set of crimes. While such charts are still commonly used today (Burrell and Bull 2011), sophisticated computer databases are also sometimes relied upon to track and analyze the behaviors of offenders (Collins et al. 1998). One of the first systems designed for this purpose was the Violent Crime Apprehension Program (ViCAP), which was developed by the Federal Bureau of Investigation for the purpose of enabling cross-jurisdiction crime linkages (Howlett et al. 1986). With the cooperation of police investigators from across the country who provide ViCAP with detailed information about the crimes they are investigating, this system allows the ViCAP team to organize, search, and analyze offense-related data in an attempt to identify patterns across crimes that may indicate the presence of serial offenders. Since the development of ViCAP, other systems have also been developed. Most notable among these is the Violent Crime Linkage Analysis System (ViCLAS). Constructed by the Royal Canadian Mounted Police in the mid-1990s (Collins et al. 1998), ViCLAS is currently being used by police forces around the world in an attempt to link serial crimes and is generally considered the gold standard for linkage systems.

Current Research

With the advent of standardized data collection protocols came the possibility of conducting empirical research to examine the degree to which behavioral information can be used to reliably link crimes to the same offender. While this field of research is still relatively new, the number of studies in this area has grown rapidly over the last decade. Existing studies examine a wide range of issues.

Tests Of The Linking Assumptions

The majority of linking research conducted to date has attempted to determine whether the assumptions of behavioral stability and distinctiveness (and thus linking) are empirically supported. Many studies in this area use the same basic procedure: (1) the researcher categorizes crime scene behaviors into domains based on their particular function (e.g., wearing a mask is assigned to a “planning” domain in cases of robbery); (2) behavioral similarity scores are calculated for each domain for pairs of crimes that have either been committed by the same offender (linked pairs) or different offenders (unlinked pairs); (3) a statistical procedure, such as logistic regression analysis, is then used to determine the degree to which these similarity scores discriminate between the two types of crime pairs; and (4) linking accuracy is quantified using a measure known as the area under the curve (AUC), derived from receiver operating characteristic analysis, which typically varies from 0.50 (chance accuracy) to 1.00 (perfect accuracy). In general, research of this type provides support for the assumptions underlying linkage analysis in crimes ranging from serial burglary (e.g., Bennell and Canter 2002) to serial homicide (e.g., Melnyk et al. 2011), with AUC values in the range of .60–.90 being frequently reported. Evidence is even emerging that it is possible to use certain types of behavioral information (e.g., spatial and temporal information) to accurately link offenses that fall into different crime categories (e.g., a residential burglary and a sexual assault committed by the same offender; Tonkin et al. 2011).

Using Behavioral Themes To Link Crimes

Rather than assigning behaviors to domains based on their particular function, other researchers have examined the linking assumptions by using clustering procedures (e.g., multidimensional scaling) to group crime scene behaviors into psychological themes (e.g., hostility). These themes then form the basis for further analysis. For example, Santtila et al. (2008) used Mokken scaling, which is similar to factor analysis, to derive behavioral themes (or dimensions) from the crime scene behaviors of Italian murderers. Using scores derived from seven dimensions as independent variables, and the series an offense belonged to as the dependent variable, they were able to use discriminant function analysis to correctly assign 62.9 % of crimes to the correct series. While lower accuracy rates have been reported for other types of crimes (e.g., serial arson; Santtila et al. 2004a), even these studies suggest that linkage accuracy exceeds levels that would be expected by chance.

Identifying Factors That Influence Linkage Accuracy

Many studies have also focused on identifying factors that might influence the accuracy with which crime linkages can be identified. While a range of factors have been explored (e.g., the type of similarity coefficient used to measure across-crime similarity; Melnyk et al. 2011), most research has examined whether certain crime scene behaviors (or themes) are more useful for linking purposes. Recent findings suggest that this is the case. For example, crime scene behaviors which are “offender-driven” appear to be exhibited in a more stable and distinct fashion by offenders (and are thus better predictors of whether crimes are linked) than behaviors that are more “situation-driven.” In cases of serial burglary, for instance, the distance between crime site locations (which is obviously determined by where an offender decides to commit their crimes) can be a very effective linking cue, whereas the type of property stolen by an offender (which depends on what is available to be stolen) tends to lack high levels of predictive accuracy (Bennell and Canter 2002). Similar conclusions can be reached from studies of linkage analysis that have used the thematic approach. In Grubin et al. (2001) study of serial sexual assaults, for instance, behaviors related to a “control” theme (e.g., having a weapon, which would appear to be planned in advance of the crime) were exhibited more consistently than behaviors related to a “style” theme (e.g., taking money from the victim, which depends on situational factors).

Understanding The Complexities Of Linking Decisions

Another body of research has investigated some of the complexities involved in linkage analysis by examining how decision-makers perform on laboratory-based linking tasks. For example, Santtila et al. (2004b) examined the performance of experienced car crime investigators, experienced general investigators, novice general investigators, and naive participants on a mock linking task using car crime data. They found that, while investigators were significantly more accurate than naive participants, there were no differences between the different types of investigators in terms of their linking accuracy, with each group identifying about half of all possible links (although experienced car crime investigators did rely on significantly less information than every other group to make their decisions). The reasons why participants in these types of studies experience problems with identifying crime linkages are still not well understood, but they likely have to do with difficulties in selecting appropriate linking cues and in adequately processing this information (Bennell et al. 2010). It may also be the case that participants who have been examined in these types of studies lacked the necessary training or experience in linkage analysis to perform well on the tasks or that the tasks themselves were inappropriately designed.

The Development Of Computerized Linking Algorithms

In contrast to the research just described, studies have also explored whether more “mechanical” approaches can be used to identify crime linkages, which is potentially important given the problems that people seem to encounter when faced with this task. In addition to various statistical approaches that have been examined (e.g., Bennell and Canter 2002), a variety of computer algorithms have also been tested (e.g., Yokota and Watanabe 2002). Studies indicate that these algorithms have the potential to automatically link crimes, sometimes with a high degree of accuracy, and that these procedures are associated with a number of advantages over other, more subjective, procedures. For example, some of the tested algorithms can weight behavioral similarity scores between crimes, such that information which is more prone to errors is given less weight (Brown and Hagen 2002). While it is still too early to make recommendations for how these computer algorithms (or similar tools) should be used in investigative settings, they may be able to effectively support the decisions currently being made by linkage analysts.

Examining Design Decisions From Previous Linking Studies

Finally, much of the research being conducted recently examines the impact of potentially problematic design decisions that have been made in previous studies. For example, most studies conducted on linkage analysis have relied on samples of solved serial crimes because, in order to determine linking accuracy, researchers need to know which crimes in their sample are actually linked. While this makes sense, designing studies in this way can bias results if the sampled crimes were originally linked and solved because they were characterized by highly similar or distinct MOs (Bennell and Canter 2002). If that were the case, levels of accuracy reported in these studies may overestimate what is actually possible when linkage analysis is applied to unsolved crimes. Fortunately, recent research that has examined linking accuracy for crimes first linked by MO versus DNA has indicated that similar levels of accuracy are observed, suggesting that there may be no need to worry about previous results (Woodhams and Labuschagne 2012). Similar attempts have been made to explore the impact of other design decisions, such as the common but questionable practice in most linking studies of excluding non-serial offenses from samples.

Remaining Challenges

Despite the growth in studies that has occurred over the last decade, challenges still remain, both for researchers working in this area and for practitioners who conduct linkage analysis.

Dealing With Problematic Data

Some of the most difficult challenges faced by both researchers and practitioners relate to issues around data quality (Burrell and Bull 2011; Woodhams et al. 2007). In both research and practice, linking decisions tend to be based on certain sources of data – typically victim statements or crime reports – that can be highly problematic. For example, with respect to victim statements, victims can forget what occurred during a crime, they can be reluctant to talk about certain events, and reports of what they say can be distorted by the police (Alison et al. 2001). Crime reports, on the other hand, are often plagued with missing data, and certain pieces of information contained within them can be highly unreliable (e.g., the time that a burglary occurred). Any research results that emerge from such data sources, or any linkage decisions that are based on them, must obviously be viewed with caution, though improvements to data coding and recording protocols can improve the situation.

This type of data can also be very limited with respect to what it reveals about offenders and the crimes they commit, which can prevent, or at least hinder, useful lines of research or inquiry. For example, in other research contexts where high-quality data can be collected, it has been discovered that individual differences in noncriminal behavior are more stable across “psychologically similar” situations (e.g., Shoda et al. 1994). This is clearly an important finding that might have relevance to the linking context. Unfortunately, despite the efforts by some (e.g., Woodhams et al. 2008), it is difficult to examine these issues within the investigative context. It is typically not possible to observe crimes taking place, and data sources like victim statements do not tend to include detailed information about situational factors. This prevents researchers from examining potentially interesting (and useful) issues. For the linkage analyst, such problems limit their ability to understand how situations influence offenders and to interpret the meaning or significance of specific behaviors, which is often necessary to accurately link crimes.

Conducting (And Getting Access To) Valid Research

A related challenge for researchers in this area is to conduct studies which produce results that generalize to real investigations (i.e., externally valid studies). The challenge for practitioners is to make sound linking decisions in the absence of externally valid linking research. The primary issue for researchers is that attempts to improve the external validity of their studies can sometimes make the research more challenging to conduct. For example, in order to increase the validity of studies concerned with linking accuracy, researchers might want to include serial and non-serial crimes in their samples, rely on serial crimes that have been linked initially by DNA, and include every crime from each offender’s series. The problem is that it is difficult, though not impossible (e.g., Woodhams and Labuschagne 2012), to get access to such data. Unfortunately, the further studies stray away from these sorts of samples, the more the generalizability of results will become a serious concern.

Given these issues, practitioners often face challenges in applying existing research to linking tasks. While the sorts of sampling issues described above are certainly part of the problem, there are additional (potentially more serious) issues that need to be addressed. Indeed, some practitioners argue that much of what is studied by researchers in this area is largely irrelevant for decisions that analysts routinely have to make, leading to calls for research that is more pragmatic in nature (Alison and Rainbow 2011). For example, most research to date has examined proactive linking, where large samples are searched, crime linkages are established, and accuracy is determined. However, analysts often encounter reactive tasks where they are asked to determine whether a set of offenses, thought to be connected by an investigator, are likely to be linked (Rainbow in press). At best, research on proactive linking tasks is relevant here and researchers have simply not made this clear. At worst, a large body of research that has been conducted is of little use in assisting analysts on common tasks that they encounter.

Withstanding Legal Scrutiny

Increasingly, linkage analysis is finding its way into court when questions are raised about whether a defendant is responsible for multiple crimes (Labuschagne 2006). This presents challenges for both researchers and practitioners. While practitioners will likely be the ones presenting at trial, for their testimony to be admitted in many jurisdictions good linking studies will be required. Indeed, although standards of admissibility differ across jurisdictions, at least some rely on criteria that relate directly to research. For example, beyond having to establish expertise and demonstrate that testimony is relevant and useful (i.e., goes beyond the common understanding of the judge or juror), if one were to apply the Daubert criteria from the USA to any sort of linkage analysis, issues related to its reliability would also come up (Woodhams et al. 2007). Questions about the testability of the linking assumptions would likely be asked, in addition to other research-related questions pertinent to this issue (e.g., have the assumptions been tested, have the results of those tests been adequately published). It is doubtful that all the Daubert criteria would be met at the moment (Woodhams et al. 2007). Whether they are in the future will depend, to some extent, on the type of research that is conducted and published. In this way, researchers have a role to play in developing research that can allow linkage analysis to withstand legal scrutiny.

For practitioners, there are two issues they could potentially speak to at trial – behavioral similarities that exist across the crimes in question and/or how peculiar (distinct) the similar behaviors actually are (Ormerod 1999). As just discussed, the primary challenge for practitioners will be getting their evidence admitted. Usefulness might prove to be a controversial issue. Despite evidence to the contrary in some jurisdictions (e.g., Labuschagne 2006), it is debatable whether testimony related to similarities would be commonly heard by the courts given that jurors and judges can arguably perceive similarity for themselves (Ormerod 1999). However, evidence related to distinctiveness might be viewed as more useful. Issues of reliability could also be contentious. The practitioner will have to clearly demonstrate that distinctiveness exists and convince the court that their testimony is sound. Currently, there appears to be no standard (i.e., commonly accepted) protocol for measuring distinctiveness, and there is insufficient published research on this particular topic, especially as it relates to reactive linking tasks. While the increased use of crime linkage systems, such as ViCLAS, will assist in establishing base rates of behavior, the reliability of the data contained within these systems will have to be confirmed.

Bibliography:

Alison LJ, Rainbow L (eds) (2011) Professionalizing offender profiling: forensic and investigative psychology in practice. Routledge, London
Alison LJ, Snook B, Stein KL (2001) Unobtrusive measures: using police information for forensic research. Qual Res 1:241–254
Bennell C, Bloomfield S, Snook B, Taylor PJ, Barnes C (2010) Linkage analysis in cases of serial burglary: comparing the performance of university students, police professionals, and a logistic regression model. Psychol Crime Law 18:507–524
Bennell C, Canter DV (2002) Linking commercial burglaries by modus operandi: tests using regression and ROC analysis. Sci Justice 42:153–164
Bennell C, Jones NJ, Melnyk T (2009) Addressing problems with traditional crime linking methods using receiver operating characteristic analysis. Leg Criminol Psychol 14:293–310
Brown DE, Hagen S (2002) Data association methods with application to law enforcement. Decis Support Syst 34:369–378
Burrell A, Bull R (2011) A preliminary examination of crime analysts’ views and experiences of comparative case analysis. J Gov Inf 13:2–15
Canter DV (1995) Psychology of offender profiling. In: Bull R, Carson D (eds) Handbook of psychology in legal contexts. Wiley, Chichester, pp 345–355
Collins PI, Johnson GF, Choy A, Davidson K, MacKay RE (1998) Advances in violent crime analysis and law enforcement: Canadian violent crime linkage analysis system. J Gov Information 25:277–284
Davies A (1991) The use of DNA profiling and behavioral science in the investigation of sexual offences. Medic Sci Law 31:95–101
Douglas JE, Munn C (1992) Violent crime scene analysis: modus operandi, signature, and staging. FBI Law Enforce Bull 61:1–10
Fosdick RB (1915) The modus operandi system in the detection of criminals. J Am Inst Crim Law Criminol 6:560–570
Gross H (1906) Criminal investigation. G. Ramasawmy Chetty, Masdras
Grubin D, Kelly P, Brunsdon C (2001) Linking serious sexual assaults through behaviour. Home Office, London
Hazelwood RR, Warren JI (2003) Linkage analysis: modus operandi, ritual, and signature in serial sexual crime. Aggress Violent Behav 8:587–598
Howlett JB, Hanfland KA, Ressler RK (1986) The violent criminal apprehension program: a progress report. FBI Law Enforce Bull 55:14–22
Keppel RD (1995) Signature murders: a report of several related cases. J Forensic Sci 40:670–674
Labuschagne G (2006) The use of linkage analysis as evidence in the conviction of the Newcastle serial murderer, South Africa. J Investig Psychol Offender Profiling 3:183–191
Melnyk T, Bennell C, Gauthier D, Gauthier D (2011) Another look at across-crime similarity coefficients for use in behavioural linkage analysis: an attempt to replicate Woodhams, Grant, and Price (2007). Psychol Crime Law 17:359–380
Ormerod D (1999) Criminal profiling: trial by judge and jury, not criminal psychologist. In: Canter DV, Alison LJ (eds) Profiling in policy and practice. Ashgate, Aldershot, pp 207–261
Rainbow L (in press) A practitioner’s perspective: theory, practice and research. In: Woodhams J, Bennell C (eds) Crime linkage: theory research and practice. CRC Press, Boca Raton
Santtila P, Fritzon K, Tamelander AL (2004a) Linking arson incidents on the basis of crime scene behavior. J Police Crim Psychol 19:1–16
Santtila P, Korpela S, H€akk€anen H (2004b) Expertise and decision-making in the linking of car crime series. Psychol Crime Law 10:97–112
Santtila P, Pakkanen T, Zappala´ A, Bosco D, Valkama M, Mokros A (2008) Behavioural crime linking in serial homicide. Psychology, Crime Law 14:245–265
Shoda Y, Mischel W, Wright JC (1994) Intra-individual stability in the organization and patterning of behavior: incorporating psychological situations into the idiographic analysis of personality. J Pers Soc Psychol 67:674–687
Tonkin M, Woodhams J, Bull R, Bond JW, Palmer EJ (2011) Linking different types of crime using geographical and temporal proximity. Crim Justice Behav 38:1069–1088
Vollmer A (1919) Revision of the Atcherley modus operandi system. J Am Inst Crim Law Criminol 10:229–274
Woodhams J, Bull R, Hollin CR (2007) Case linkage: identifying crimes committed by the same offender. In: Kocsis RN (ed) Criminal profiling: international theory, research, and practice. The Humana Press, Totowa, pp 117–133
Woodhams J, Hollin CR, Bull R (2008) Incorporating context in linking crimes: an exploratory study of situational similarity and if-then contingencies. J Invest Psychol Offender Profiling 5:1–23
Woodhams J, Labuschagne G (2012) A test of case linkage principles with solved and unsolved serial rapes. J Police Crim Psychol 27:85–98
Yokota K, Watanabe S (2002) Computer-based retrieval of suspects using similarity of modus operandi. Int J Police Sci Manag 4:5–15