Forensic Techniques Accuracy: Key Takeaways and Limitations

A man was found dead with unusual wounds on his back. To find the killer, a local official gathered all the village farmers and their tools. When each man laid down his sickle, flies descended on one specific sickle. Its owner, overwhelmed, confessed to the murder. This account, from Song Ci's 1247 book "The Washing Away of Wrongs," is considered the first written record of an empirical approach to forensics.

While modern forensics are often perceived as highly accurate, with fingerprints supporting over 70% of murder investigations and DNA evidence over 90%, there are significant concerns. A 2009 report by the National Academy of Sciences, spanning 350 pages, concluded that, with the exception of nuclear DNA analysis, no forensic method has been rigorously proven to consistently link evidence to a specific individual. The report also stated that some tests fail to meet fundamental scientific requirements.

To illustrate these issues, five commonly used forensic techniques, still admissible in court, were ranked by individuals based on perceived accuracy.

Microscopic Hair Analysis

Hair has long been a focus for crime scene investigators. Before DNA analysis, experts would compare microscopic similarities in hair's surface and cross-section, such as the roughness of the cuticle or pigment distribution, to find matches. This was used to determine if hair was human or animal, and even its race.

Between the 1970s and 1999, the FBI used microscopic hair analysis in 268 cases. However, a re-examination using DNA evidence revealed significant flaws. Even experienced FBI examiners routinely claimed matches between hairs from different individuals and sometimes couldn't distinguish between human and dog hair. Of the 268 cases, 96% were found to be false. Tragically, 33 people had been sentenced to death, and nine executed, before these errors were discovered. Today, the FBI only uses hair as evidence if supported by DNA testing, placing microscopic hair analysis at the bottom of the accuracy list.

Bite Mark Analysis

The premise of bite mark analysis is that if dental records can identify a deceased person, bite marks left on a victim should identify a perpetrator. Since the 1950s, this technique has been used in thousands of cases. However, forensic dentists like Mary Bush began to question its scientific basis.

Research from the University at Buffalo involved creating 89 bite marks on cadavers using model teeth, alongside controlled bite marks in wax. None of the bite marks on skin matched the measurements of those made in wax or the original model teeth. When compared to a wider collection of 411 model teeth, the original model that made the bite mark was often not even the closest match. After 12 studies, the conclusion was that bite mark transfer to skin is unreliable because skin is soft, squishy, and distorts under pressure.

Despite this scientific evidence, there has been no ban on bite mark evidence, and it is still allowed in courts worldwide. Dr. Bush noted that while science readily discards disproven theories, the justice system often prioritizes consistency and historical precedent. Bite marks have been used as recently as 2025, earning it the second-worst spot on the accuracy list.

Bloodstain Pattern Analysis

Bloodstain pattern analysis interprets the patterns formed by blood pooling and dripping at a crime scene to reconstruct events. This field gained prominence with Herbert Leon MacDonell's 1971 book, "Flight Characteristics and Stain Patterns of Human Blood." MacDonell described how to use the width and length of a stain to calculate the angle of impact and, using trigonometry, trace it back to the point of origin. His work became the foundation of the field, and he even established an institute to train police officers. The Supreme Court of Iowa deemed the practice "relatively uncomplicated" and did not require proof of its reliability.

However, a critical flaw in MacDonell's method, and subsequent analyses, is the failure to account for gravity and drag. Assuming straight trajectories leads to an inaccurate common origin point, often suggesting a victim was standing when they might have been sitting. The first study to measure the baseline reliability of bloodstain pattern analysis was not conducted until 2014, over 50 years after its court admission. A 2021 large-scale study found that analysts reached different conclusions about how a stain was made approximately 8% of the time. This is partly because the same stain can result from various mechanisms, and blood viscosity varies between individuals (e.g., men tend to have 15% higher red blood cell concentration than women).

New software using fluid dynamics is improving the accuracy of bloodstain pattern analysis, which places it in the middle of the accuracy list.

Fingerprint Analysis

Fingerprints are widely believed to be highly accurate for identification due to their uniqueness. However, a notable case highlights potential issues. On March 11, 2004, explosions in Madrid killed 193 people. Police found a finger mark on a detonator bag and matched it to Brandon Mayfield, an Oregon lawyer who had no record of leaving the US and didn't own a passport. Despite this, the fingerprint evidence was considered so damning that Mayfield was incarcerated.

The origins of fingerprint classification date back to 1890s Kolkata, where criminals were avoiding jail by paying others to serve sentences. Edward Henry, Azizul Haque, and Hem Chandra Bose developed a system to classify fingerprints. Haque proposed using "whirls" (spiral patterns) on fingers. With 10 fingers, each either having a whirl or not, there were 2^10 (1024) possible arrangements, leading to 1024 "pigeonholes" for organization. For individuals without whirls (two-thirds of the population), additional categories like plain arches (A) or tented arches (T), and the side of loop structures, were used for more detailed classification. This system was used for over 100 years, notably by the world's first fingerprint bureau at New Scotland Yard.

Today's system uses "friction ridge analysis," focusing on minute details along the ridges, called minutiae (where ridge lines split or end). While online databases can now process these in minutes, the problem lies in the initial identification of minutiae. To search a database, a minimum number of minutiae points are required. However, different examiners can identify vastly different numbers of minutiae on the same print. This means whether a criminal is found can depend on the "luck of the draw." Studies show that even when given the same pair of prints twice, examiners reach a different decision 10% of the time.

These disagreements occur in controlled studies without context. In reality, fingerprint experts often have information about the crime, which can introduce bias. For example, 42% of requests to fingerprint experts state whether a suspect has a criminal record. In the Mayfield case, experts claimed a 100% certain match, likely influenced by conformity bias, where subsequent examiners and even the defense expert agreed with the initial verdict, despite the print actually belonging to someone else. This highlights the importance of forensic examiners working independently, free from case context.

DNA Analysis

DNA is often considered the "silver bullet" of forensics due to its power in identification. While early DNA analysis required a dime-sized sample, today it can be done with less than a pinhead's worth. This increased sensitivity, however, can also be a downfall.

In 2012, paramedics in California inadvertently transferred a homeless man's DNA to a murder victim's fingernails after treating him and then responding to a murder scene. This led to the homeless man, Lucas Anderson, being charged with murder and facing the death penalty, despite being hospitalized at the time of the crime. He spent five months in jail before charges were dropped. This illustrates the issue of "trace DNA" and "touch DNA" transfer, where DNA can be found in places an individual has never been.

Crime scenes often yield "DNA mixtures" with multiple profiles, which are the most common source of error in DNA interpretation. The amount of DNA shed varies between individuals, meaning the majority of DNA on a surface might not be from the last person who touched it.

The most common method for DNA analysis uses Short Tandem Repeats (STRs), which examine how short DNA sequences repeat in the genome. A standard STR test looks at about 20 locations, with each person having two copies (one from each parent), resulting in 40 genetic markers. The chance of two unrelated individuals having the same combination is about one in a billion.

While a single-contributor sample is easy to interpret, DNA mixtures cause profiles to overlap with varying signal strengths, making interpretation difficult. With four or more individuals, it becomes challenging to compare a suspect's clean DNA profile to the mixture.

In 2013, the National Institute of Standards and Technology (NIST) conducted a controlled study, sending a fictional crime scene DNA mixture (from four people) to labs across the US. 69% of the labs got the analysis wrong, and only 21% deemed the mix inconclusive. While new checks have been implemented, there's still no lower limit on the quality or quantity of DNA samples labs can analyze, and labs themselves decide if a sample is too mixed or partial.

Using whole-genome sequencing could introduce different problems, as analysts would have access to information like hair color, eye color, and ethnicity, potentially leading to discrimination in analysis. While DNA evidence is incredibly powerful, it must always be interpreted within context.

Conclusion

While forensics are essential, it's crucial to acknowledge their limitations and continuously reassess their scientific validity. The goal is not to dismiss forensics but to ensure they meet rigorous scientific standards. The ongoing work of dedicated professionals in reassessing and improving forensic techniques is vital for maintaining their scientific integrity and accuracy.