It’s a staple of cop shows, but the jury’s still out on claims that criminals’ identities can be deduced from their modus operandi
WHEN Derrick Todd Lee was arrested in 2003 for a spate of brutal murders in Baton Rouge, Louisiana, the public reaction was a mixture of relief and outrage: relief that he was off the streets, outrage that he had not been caught sooner, given his track record of sexual and violent crime. One reason the police gave for the delay in tracking him down was that they had been looking for a white man, based on an FBI profile. Lee was black. Just months earlier, criminal profilers made the same mistake in the case of the Beltway sniper, who was using a high-powered rifle to pick off drivers in the Washington DC area. Again, he was described as white. It was three weeks before police apprehended John Muhammad and his teenage accomplice Lee Boyd Malvo. Both were black.
As anyone who watches television detective programmes will know, criminal profiling claims to predict the characteristics of an offender from an analysis of a crime’s circumstances. Most police forces in the developed world use profiling, but it is not always as successful as the small screen would have us believe. Indeed, the approach has been mired in controversy since it was introduced half a century ago.
Advertisement
Last year, a searching by the US National Academy of Sciences challenged all of forensic science to demonstrate its scientific credentials. For profiling, rising to that challenge may take some doing, as even some of its strongest advocates consider it as much an art as a science. Psychologist Brent Snook of the Memorial University of Newfoundland in St John’s, Canada, goes as far as to say, “Profiling should be suspended until proper testing of the method provides evidence that it works.” There are even mutterings that this is a perfect opportunity to expose criminal profiling for the pseudoscience it really is. Is this the end for the staple of detective dramas, or can profiling prove its worth?
The idea that you can predict the characteristics of an offender from an analysis of his or her behaviour at the crime scene goes at least as far back as Jack the Ripper, a killer who haunted east London in 1888. But profiling in the modern sense probably began in 1956, when police investigating New York City’s Mad Bomber consulted psychiatrist James Brussel. He famously predicted that when the man who had terrorised the city for 16 years was apprehended, he would be wearing a buttoned, double-breasted suit. In fact, George Metesky was in his pyjamas when officers arrived at midnight to arrest him. He went off to get dressed and returned in a double-breasted suit, and it was buttoned.
The American model of profiling emerged from this auspicious start. Developed by the FBI, it relies on the judgement of experienced police investigators. First, all the available information about the crime is assimilated. Then the crime is classified as either organised or disorganised, with organised crimes characterised by planning, advanced social skills and control over victims, and disorganised ones tending to be more impulsive and opportunistic.
Next comes a reconstruction of the sequence of the crime. Then the profiler looks for any idiosyncratic behavioural signature that might betray the identity of the offender – a particular way of tying up a victim, for example, or covering a rape victim’s face with her shirt. Finally, a profile is produced which might describe any or all of the offender’s socioeconomic, family and physical characteristics, plus age, education, personality and psychological quirks.
A challenger to the FBI system appeared on the scene in the 1980s. Troubled by what he saw as a lack of scientific rigour, David Canter, a psychologist then at the University of Liverpool, UK, began by asking the most basic question: is crime scene behaviour actually linked to characteristics of the offender? He used statistical tools to look for patterns in police databases, aiming to find out how often different behaviours were coupled at the same crime scene – to try to identify “offence styles”. In the case of rape, for example, his group found that rapists could not be distinguished by their physical or sexual acts at the scene, but they could be distinguished by non-sexual acts: for example, whether they stole from or apologised to the victim. More controversially, Canter assisted in police investigations by taking the trends that he was identifying in large criminal populations and applying them to individual cases.
Canter’s scientific analyses marked an important break with the past. Since then, the FBI approach has incorporated more empirical methods, including a centralised database to help investigators identify links between crimes (see “How to spot a serial offender”). Meanwhile, police forces in many other countries have adopted and adapted the systems pioneered in the UK and US.
Fabricated offence
Profiling now looks more scientific, but is it? Canter’s approach has never been properly evaluated, says Snook, a former student of Canter’s. And Mark Hilts, who heads the second of three behavioural analysis units at the FBI’s National Center for the Analysis of Violent Crime in Quantico, Virginia, admits that the FBI has carried out no controlled studies of profiling’s efficacy since the centre was established in 1985. The FBI does not publish its profiles, so nobody else has done it for them either. “I’m not really sure how you’d do it, quite honestly,” says Hilts.
Snook has a suggestion: take real, solved cases where you know who the perpetrators are from other lines of evidence. Then give profilers the case materials and ask them to draw up profiles, rating them on how closely they match the perpetrators. Several studies have done something similar already, comparing the predictions of profilers and non-profilers in mock crime situations where the characteristics of the “real” perpetrator are known. Drawing together the results of four of these studies in a meta-analysis published in 2007, Snook’s team found that the profilers did only slightly better than students without any experience of profiling, and that the predictive abilities of both were very low ().
The main performance measure used by both the FBI and the National Policing Improvement Agency, which is responsible for profiling in the UK, is the level of customer satisfaction in surveys of the law enforcement agencies that use their services. “We get a lot of return customers,” says Hilts. But that doesn’t necessarily mean profiling is effective. Instead, the profiles might simply be couched in such ambiguous terms that they can be made to fit almost any possible offender. This is exactly what Laurence Alison, director of the Centre for Critical and Major Incident Psychology at the University of Liverpool, and colleagues found. In a study published in 2003, they presented two groups of police officers with a profile from a real case and either the actual offender or a very different, fabricated offender. More than half of both groups rated the profile as an accurate description of the offender in front of them ().
There is no doubt that cases of mistaken identity do occur. One notorious example is that of Colin Stagg, accused of the murder of Rachel Nickell on Wimbledon Common, London, in 1992. Despite a lack of any forensic evidence against him, Stagg fitted the offender profile drawn up by criminal psychologist Paul Britton and the police designed a covert operation to entrap him. Stagg was acquitted at trial but the debacle was a wake-up call to British police about the dangers of relying on profiles. The role of profilers in British police investigations has since diminished.
“They are a very small cog now, and rightly so, in a much bigger investigative machine,” says Alison. The kind of information they provide has changed too, as has the way that it is presented. “Ten, 15 years ago there was the belief that you could turn up at a crime scene, look at what had been left behind and say, here’s a guy who doesn’t like his mother and drives a red van,” he says. These days, by contrast, every inference is supported by academic research or statistical analysis.
Even so, the thinking is that profiling should be treated with extreme caution. “It’s not proof, it just points the police in a particular direction,” says GĂsli GuĂ°jĂłnsson of the Institute of Psychiatry, King’s College London, who has provided profiles to the British police. Knut Sturidsson at the Karolinska Institute in Stockholm, Sweden, goes further. He believes the way that profiling is practised now, based on something intangible called experience and a rather thin body of empirical research, is fundamentally flawed. If the profiling prediction is wrong, it may blind police to conflicting but genuinely informative details and cause them to waste valuable time.
“Profiling as practised now, based on a thin body of research and something intangible called experience, is flawed”
There are two things that need to be done to make profiling viable, Sturidsson says. First, there has to be a better system for presenting the information to the police. One approach would be to use the Bayesian framework employed in other forensic domains. So, for example, you might calculate the probabilities of obtaining a certain profile if suspect X is guilty and if he is not, and compare the two to give a likelihood ratio indicating which alternative the evidence favours and how strongly. Investigators would then have some idea of what level of uncertainty they were dealing with. The second, and greater challenge, he says, is to do the research to reveal how evidence at the scene of a crime reflects the characteristics of offenders – in other words, to continue what Canter started.
Solid foundations
Gabrielle Salfati of John Jay College of Criminal Justice in New York City has been doing just that. In 2004, when Canter, Alison and others analysed the behaviour of 100 serial killers to look for behaviours that occurred together, they found no evidence for the organised-disorganised dichotomy the FBI uses. Salfati has found evidence that homicides fall into two categories: expressive, where the act is an expression of the killer’s emotion, and instrumental, where the killing is a means to an end, such as financial gain ().
Similarly, Alison’s and other groups have found that rape can be classified loosely into three types: criminal, where rape is just one element of the offender’s crime repertoire; pseudo-intimate, where sex is the goal; and hostile, where the sex is secondary to violence. These classifications are useful, Alison says, partly because they help to link crimes. However, not all criminals fit into them. In one study of 200 single murders, Salfati found that just over half the perpetrators could be classified according to her expressive-instrumental scheme. “That shows that we’re missing a big part of the picture,” she says.
Will profilers ever have the whole picture? Some law enforcement agencies are beginning to show signs of disillusionment. “I don’t think profiling has much of a future,” says statistician Chris Devery, who has conducted a review of the literature for the New South Wales police force in Australia. “There’s just no evidence that it works.”
Others see a future for it, but only if practised very differently. Salfati admits that profiling is “very, very young” in scientific terms, though she is optimistic that it could be an effective investigational tool, if rebuilt on solid scientific foundations. After all, she says, “It’s just another name for understanding how people act. By saying that profiling doesn’t work, we’re essentially saying that psychology doesn’t work.”
Canter, now at the University of Huddersfield, UK, is even more upbeat. He sees profilers, or investigative psychologists as he prefers to call them, influencing police methods in more ways than just profiling – in interview techniques, for example. He compares their role to educational or occupational psychologists. “You do not expect these psychologists to teach or run companies but to contribute to the procedures and training of those who do,” he says.
How to spot a serial offender
Pooling national data on serious crimes seems like the obvious way to spot connections between them. In the US, this is done through a system called the Violent Criminal Apprehension Program (ViCAP), while Canada has the Violent Crime Linkage System (ViCLAS). Nine other countries use variants of these.
Information-gathering begins with police filling out a questionnaire. It contains more than 100 questions covering all aspects of the crime, asking about the criminal’s verbal or sexual behaviour as reported by the victim, for example. This information is sent electronically to a central bureau where analysts trained to look for patterns in behaviour can pore over it.
It sounds like common sense, but several problems have been identified with this approach. For one, we don’t actually know whether these links exist as there is no concrete evidence that serial criminals behave consistently across crimes. For another, it is likely that individuals will vary in how they interpret a crime scene. A study commissioned by ViCLAS in 2008 found that the likelihood of an investigating officer corroborating a behaviour identified by a colleague was 38 per cent in cases of homicide, and a mere 25 per cent in cases of sexual assault – no better than if the officers had entered responses at random.
Olivier Ribaux, an expert in forensic intelligence at the University of Lausanne in Switzerland, thinks linkage analysis needs a radical overhaul. He says analysts should start by looking at material evidence, such as DNA, left at the scene of the crime. From this they can build a picture of the offender’s movements and his method of selecting opportunities. A behavioural analysis and profile would come last. “It’s exactly the opposite of what is done now,” he says.
- Forensic Psychology: a very short introduction by David Canter (Oxford University Press, 2010)
