Crime Script Analysis and Behaviour Sequence Analysis are two of the most widely used statistical frameworks in modern criminal investigation, designed to map offender behavior into predictable sequences that investigators can compare across cases. A 2026 study out of Murdoch University’s School of Law and Criminology puts both tools under scrutiny, and its conclusion is uncomfortable for anyone who treats statistical modeling as a shortcut to solving crime: these frameworks may be quietly distorting the very cases they’re meant to clarify.
What the Study Actually Tested
Researchers Jolie Henriksen and David A. Keatley analyzed 48 closed child abduction-homicide cases, gathering data through open-source intelligence rather than direct case-file access. Their focus wasn’t whether Crime Script Analysis or Behaviour Sequence Analysis could identify patterns, it was whether the patterns those tools identify are actually reliable once you compare enough real cases against each other.
The comparison turned up a split result. Offender profiles across the 48 cases showed minimal variation, meaning the people who commit these crimes tend to share similar characteristics. Crime sequences, however, showed significant differences. In plain terms: the who tends to look similar, but the how rarely repeats in a way that fits a clean statistical model.
Two Ways to Measure the Same Thing, Two Different Answers
Part of the study’s contribution is a comparison between two statistical representations used to score offender behavior: Standardised Residuals and Prevalence Scores. These aren’t just academic technicalities, since which one an investigator uses can shift how confident they are in a given behavioral pattern, and that confidence can shape which leads get prioritized. The researchers found that both approaches can distort investigative outcomes when applied uncritically, not because the math is wrong, but because the underlying crimes don’t behave the way the statistics assume they will.
The Core Argument: Real Crimes Resist Linear Models
The study’s central claim is that both tools, as commonly applied, undermined the chaotic, interrupted, and non-linear nature of real-world crimes. Child abduction-homicides in particular rarely unfold as a clean sequence of steps. An offender’s plan can be interrupted by the victim’s resistance, unexpected witnesses, or simple logistical failure, and what actually happens on the ground often diverges sharply from what a script-based model predicts in advance.
This matters beyond academic debate. Investigators who lean too heavily on a statistical script risk two failure modes: dismissing a real case because it doesn’t match the expected pattern, or forcing a case into a pattern it doesn’t actually fit, both of which can misdirect an active investigation.
What the Researchers Recommend Instead
Henriksen and Keatley don’t argue for abandoning statistical tools altogether. Their recommendation is more specific: treat each case as unique rather than extrapolating from generalized data, and pair analytical tools with experiential insight, meaning the judgment investigators build through direct casework, not just pattern-matching software. In practice, that means using Crime Script Analysis or Behaviour Sequence Analysis as one input among several, not as a template the case is expected to conform to.
Why This Study Matters
Statistical crime analysis has been gaining ground in investigative work for over a decade, often marketed as a way to bring objectivity to cases that used to rely purely on investigator intuition. This study is a useful check on that trend, not because the tools are useless, but because treating any single model as authoritative, in a domain as unpredictable as child abduction-homicide, risks replacing one kind of bias (investigator hunches) with another (false statistical confidence).