Gregor: A man has been found lying dead at the side of the road with oozing, elongated wounds to his back. To solve the murder, a local official calls all the farmers in the village and demands they all bring their tools to the main square. In the square, each man is set to lay down his sickle. And then something curious starts to happen. In the rising heat of the morning, flies begin to descend, all landing on a single sickle. Its owner buckles, falling to his knees and admits to murder. This is from a 1247 book, "The Washing Away of Wrongs" by Song Ci. And it's the first writing we have of an empirical approach to forensics.

Gregor: But okay, big deal. Anyone could have just swapped the sickles or used their neighbor's. But forensics today are much more accurate, right? Fingerprints support investigations in over 70% of murders and DNA evidence in more than 90%. But then you also read an article like "Hair sample that put men into prison for 28 years, turned out to be dog hair."

Speaker: That is not science. That is not justice.

Gregor: See, in 2009, the National Academy of Sciences published this 350 page report stating that, with the exception of nuclear DNA analysis, no forensic method has been rigorously shown to consistently demonstrate a connection between evidence and a specific individual. And they also said that some tests do not meet the fundamental requirements of science. So we ask people to rank five famous forensic techniques, all of which are still being used in court. Okay, five forensic techniques here. And I just need you to rank them in order of accuracy from the least accurate down here to the most accurate. So what do you think?

Respondent A: I don't know if I'm gonna nail these.

Respondent B: Whoa. I dunno.

Respondent C: I don't know what forensic to put first here.

Gregor: Okay, hair. What are you thinking?

Respondent A: I think hair is last, eh?

Respondent B: I would probably put the hair on the top two.

Respondent C: I feel like hair would probably be the easiest to, to fake? To fake, if you like.

Respondent D: I think hair at the bottom, because if it's not got the follicle intact, just hair in general.

Respondent E: Do you think?

Respondent F: Yeah.

Respondent G: Yeah.

Gregor: Hair has always been an object of interest for crime scene investigators. And before DNA, they would look for the microscopic similarities in the surface and the cross section of hairs to try and get a match. For example, analyzing how rough the fish scale like structure of the surface of the hair is, or how the pigment is distributed through the hair shaft.

Announcer: The examiner's trained eye can learn many things. Is it human hair? Of what race? Animal hair? What family? This vital information may help to establish guilt or innocence.

Gregor: Between the 1970s and 1999, the FBI used this against defendants in 268 cases. But when they later reexamined this using DNA as evidence, it completely flipped the field on its head, and it turned out that even the most experienced FBI examiners would routinely claim they had a match between hairs that were from two different people. And they sometimes couldn't even tell the difference between a human hair and a dog hair. Out of those 268 cases that used microscopic hair analysis, 96% were declared false. 33 people were already sentenced to death, and nine were already executed by the time the errors were found. Today, the FBI only uses hair as evidence if it can be supported by DNA testing, which is why microscopic hair analysis goes to the bottom of this list.

Gregor: What do you think of bite marks?

Respondent A: On a crime scene?

Gregor: On a crime scene? Yeah, yeah.

Respondent B: It's been... Dracula, is Dracula there?

Respondent C: Least accurate is gonna be pretty close to bite marks, I think.

Respondent D: Midway.

Respondent E: Midway?

Respondent F: Last. That's last. That's not realistic, innit?

Gregor: Okay. Okay. Mm. Bite marks at the top. The idea behind bite marks is simple. If you can use dental records to identify a deceased person, surely you can identify a perpetrator by the bite marks left on a victim. Since the 1950s, bite mark analysis has been used in thousands of cases. But then some experts, like forensic dentist Mary Bush, started looking into the method more deeply.

Mary Bush: There was no scientific exploration into bite marks before they were used in the court of law to say that this was even a feasible technique that you could do.

Gregor: One study from the university at Buffalo used a set of model teeth to create 89 bite marks in cadavers, alongside controlled bite marks made in wax. And out of those 89 bite marks made in skin, none of them matched the measurements of the ones made in wax or the original model teeth. In fact, when they compared the bite marks in skin to a wider collection of 411 model teeth, it wasn't even the original model that made the bite mark that was the closest match. After 12 studies, their conclusion, bite mark transfer to skin is not reliable. Skin isn't a good medium to leave an imprint of someone's bite mark. It's soft, it's squishy, and it distorts under pressure.

Gregor: But despite all this, there's still been no ban and bite mark evidence is still allowed in courts worldwide. We asked Dr. Bush for a comment and she said that while the scientific method is happy to discard an old and disproven theory, the justice system prefers consistency in a historical precedent. So even the new flurry of evidence wasn't enough to get bite marks out of court. In fact, bite marks have been used as evidence as recently as 2025, which is why it gets the second worst spot on our list.

Gregor: Blood, if you don't have DNA and you're just using like kind of the patterns that it leaves along the ground. How they splat.

Respondent A: Right?

Respondent B: Like the picture, that blood on a crime scene could give you.

Gregor: Okay.

Respondent C: I'm looking, thinking of putting blood at the end, but like, I don't know.

Gregor: Okay. Put blood at the end. Okay.

Respondent D: I would go blood then

Respondent E: Second.

Respondent F: Second place.

Respondent G: Yeah, yeah, yeah, yeah.

Gregor: So bloodstain pattern analysis. When someone is injured, the way their blood pools and drips into the environment is in a distinct pattern. And in 2020, "Wired" actually did an episode with a crime scene analyst on how these different stains can be interpreted. You can use physics and biology to rewind the clock and extrapolate where in space the blood must have come from.

Gregor: This all started in 1971 when chemist, Herbert Leon MacDonell, published "Flight Characteristics and Stain Patterns of Human Blood." In this book, MacDonell described how you can use the width and length of a stain to calculate the angle of impact, and then using trigonometry, trace that back to the point of origin of the spatter. The book quickly became the foundation of bloodstain pattern analysis, and everyone was using it. In fact, MacDonell even established a Bloodstain Institute where thousands of US police officers would enroll.

Gregor: The Supreme Court of Iowa actually referred to the practice as relatively uncomplicated. So they didn't require proof of its reliability. They were happy to go ahead with MacDonell's tests that he did alone in his basement. And soon enough, bloodstain pattern analysis was used in courts all over the states. But here's the problem. Plotting the flight path of each bloodstain using trigonometry assumes the trajectories were straight lines and gives this common origin point, which suggests the victim was standing up.

Gregor: The issue is, MacDonell and then many forensic investigators after him and even "Wired," failed to account for the effects of gravity and drag as obvious as this sounds. When you include these in your calculations, the common origin point lowers, and you can see that in this example. It's more likely that the victim was actually sitting down.

Gregor: The first study to measure the baseline reliability of bloodstain pattern analysis was only done in 2014. That's over 50 years after it started being admitted in court. And a large scale study from 2021 found that analysts come to different conclusions about a way a stain was made, about 8% of the time. And that's because the same stain can come from many different mechanisms. And the fact that blood differs from person to person, depending on how many red blood cells versus the plasma you have. Men actually tend to have a 15% higher concentration of blood cells than women, which makes their blood more viscous. Nowadays, investigators have new software which uses fluid dynamics to help them map these complex 3D seams. So bloodstain pattern analysis is improving in accuracy, which is why it gets this middle spot in our list.

Gregor: Okay. But surely fingerprints must have a lot of value in identification. How unique do you think fingerprints are?

Respondent A: I think they're pretty accurate.

Respondent B: They're fingerprints.

Respondent C: Fingerprints.

Respondent D: That's probably one.

Respondent E: That's number one.

Respondent F: Yeah.

Gregor: Okay Okay. I mean, yeah, you use fingerprints on your phone, you unlock it, you pay stuff, right? 'Cause fingerprints are unique. Even when you have a passport, like they scan you when you enter a country.

Gregor: On March 11th, 2004, a terrorist organization operating in Madrid set off explosions on four commuter trains. The attack killed 193 people and injured thousands more. Quickly after, the police found a finger mark on a detonator bag left behind, and they matched it to Brandon Mayfield, a lawyer in Oregon. There's only one problem, there are no records of Mayfield even leaving the US. He doesn't even own a passport. So how could he have pulled this off?

Brandon Mayfield: I honestly felt like I was being framed because I hadn't been outta the country for over 10 years.

Gregor: But despite that, the fingerprint evidence against him was considered so damning that Mayfield was incarcerated. But this wasn't a glitch. It was built into the method from the start. In 1890s Kolkata, the city's rich career criminals were avoiding jail by paying people to serve sentences for them. It was a good tactic since, at the time, there was no practical way to identify a person. So the criminals were getting away with it.

Gregor: Three officers, Edward Henry, Azizul Haque and Hem Chandra Bose started taking the prints of everyone who came through the station, but they quickly ran into a problem. When they got a suspect in, how could they crosscheck their prints against their own database when there were 10 prints per person and thousands of people coming through? They needed a classification system.

Gregor: Haque proposed that they should look at whether a fingerprint has a whirl. It's the spiral pattern that a person might have on their fingers. And because each of your 10 digits can either have a whirl or not, there are a total of two to the power of 10 or 1024 ways your whirls can be arranged. So the officers just built 1024 wooden pigeonholes to organize these fingerprints. Okay. But what if like me, you don't have a whirl? Well, two thirds of the population actually don't. So the system had to go further. Once you have a pigeonhole number, you go for additional layers of identification. Some of the additional categories that you could assign included whether you had a plain arch, which was denoted with the letter A, or whether you had a tented arch, which was denoted with a T. Similarly, you could look at which side of the finger this loop structure is on. And with a few more identifiers like these, we get a much more in-depth classification, one that would go on to be used for the next 100 years. Famously, the world's first fingerprint bureau, which was established in London in the New Scotland Yard, used this fingerprinting system.

Expert: I started in 1988 with Metropolitan Police in forensic services.

Gregor: Could you paint a picture of what the room would've looked like with all these pigeonholes? How big was this space?

Expert: So we're looking at a space of about 30 to 40 meters in length and about 10 to 20 meters in width. The collection at the time was about 3 to 4 million I think.

Gregor: That is intense. Yeah, that's wow.

Expert: It's a lot of fingerprints.

Gregor: Now, today's system takes into account these little details along the ridges. It's called friction ridge analysis. And well, it's no surprise that combing through pigeon holes for these would've taken weeks, but online databases today can basically do it in minutes.

Gregor: No way this phone works. Hello?

Collaborator: Howdy, Gregor. Just trying to get a hold of you.

Gregor: Why are you calling me through this thing? You have no data? Just get Saily.

Collaborator: You know, I travel a lot for filming. I actually just got to London. I always wanna be able to connect to the internet as soon as I land, and with the help of today's sponsor, Saily, it couldn't be easier. First, you just scan the QR code and download the Saily app, then you search your destination, the UK, for example. And next you just pick a data plan. And finally, during checkout, apply the coupon code Veritasium to get your exclusive 15% discount. No more dodgy cafe wifi or unexpected roaming charges. All you need to do is just switch on your mobile data when you land, and that's it. Saily works with all iPhone and Android devices. And once an eSIM is installed, you can keep adding plans in the app for each new trip. And in the rare case that your phone isn't compatible with Saily, you can get a full refund at no questions asked. I've gone with the Saily Ultra Plan, which gets you unlimited data globally with access to airport lounges, fast track services, and NordVPN, NordPass, NordLocker and Incogni on top. So to stay connected on the go for an affordable price, you can scan this QR code, download Saily, and get 15% off your plan with the coupon code Veritasium. So I wanna thank Saily for sponsoring this part of the video, and now back to our case.

Gregor: So if all the details are painstakingly placed into these databases, then where's the problem? How is Brandon Mayfield falsely accused? Well, say you have a mark from a scene and you wanna run it through the US fingerprint database, you can't actually search the database using the whole mark because you first need to identify a number of features on the print. These are called minutiae. It's where the ridge line split or end.

Expert: To even search the database, you need a minimum number of minutiae points. It's very efficiently sorted, hundreds of million prints, but it cannot make an ID. So what it does, it pulls out the most similar ones, at least 10, 20 up to 40 in some jurisdiction. And then the human examiner has to compare.

Gregor: This works in theory, but the problem is how these minutiae are identified in the first place.

Expert: What is interesting in my research published scientific reviewed research, the same thing one examiner see only three minutiae and one sees 10 minutiae. So 10 minutiae is enough to run the database. However, if the other are only the three minutiae, they couldn't run it. So whether you find a criminal or not depends on the luck of the draw. Even if you give the same pair of prints to the same examiner twice, 10% of the time, they will reach a different decision.

Gregor: And these disagreements are from controlled studies where the examiners had no context about the crime, but in reality, they often do. As many as 42% of the requests that fingerprint experts process state whether a suspect has a criminal record.

Expert: In a recent case that I was involved where a firearm was involved and a forensic firearm expert had to decide whether a bullet was actually fired on the firearm, on the suspect. On the former submission form that was given to the forensic scientist, it said that it's homicide, it said the age of the suspect and the victim, and it even said the race. It said the suspect is black, and the person who died the victim was white.

Gregor: For over a century, experts have held so much trust in fingerprints that in the FBI case for Mayfield, the examiners claimed with 100% certainty that that was his mark. This likely happened because when another fingerprint expert was brought in to verify the match, they probably already knew the verdict of the original analyst, and they probably also knew the stakes of what a match could mean for the case, which made it all the easier to agree with the initial verdict. Even the fingerprint expert working for the defense agreed that the mark matched Mayfield's fingerprint when in reality, it belonged to a man who had links to terrorist organizations in Spain. This is the impact of conformity bias.

Expert: So all of this pressure, all of this context and intervention in the forensic science, I say to the police and the prosecutor, leave the forensic examiners alone. Give them independence of mind to make decision based on the relevant scientific evidence and do the work rather than telling them about the case.

Gregor: Well, if there were issues with all these other techniques, why don't we just use DNA? DNA first?

Respondent A: Yeah.

Respondent B: DNA has got to be surely,

Respondent C: DNA should be top.

Respondent D: It has to be DNA because,

Gregor: Okay, okay.

Respondent E: You already know everyone have their own DNA.

Gregor: But even DNA analysis can fall prey to the same problems. Back in the 80s, you had to do this with a sample the size of a dime of blood or saliva. But today you can do it with less than a pin hair's worth. In fact, just this year a 61-year-old cold case was solved with just 0.4 nanograms of DNA. But the paradox is, the increasing sensitivity of DNA analysis could also be its downfall.

Gregor: In 2012, paramedics in California responded to a call about a severely intoxicated homeless man. They treated him and took him to the hospital, but along the way, they also picked up traces of his DNA on their equipment and gloves. Then later that night, they responded to a murder of a businessman and they accidentally transferred the homeless man's DNA to the victim's fingernails. Because of that, the homeless man, Lucas Anderson, was charged with murder and he faced the death penalty despite being hospitalized at the time the crime occurred. He spent five months in jail before charges were finally dropped.

Expert: So this is trace DNA and touch DNA, where you can get this kind of DNA transfer that happens.

Gregor: I see how that could be abused easily. So my DNA could be in a place I've never been before.

Expert: Yeah. And if DNA is deposited under the right conditions, say a dark dry place, it can persist for hundreds of years.

Gregor: And if, for example, somebody touches a door handle, well, loads of people are gonna be touching a door handle, so you're definitely going to get more than one profile off of there.

Gregor: These samples with multiple DNA profiles are called DNA mixtures, and they have been found to be the most common source of error in DNA interpretation.

Expert: Some people shed more cells than others, so the majority of the DNA that you're getting off of that door handle may not be from the last person who touched it.

Gregor: The problem in interpreting a mixture comes down to how a sample gets analyzed. The most common way to do it uses short tandem repeats or STRs. DNA is made up of four nucleotides, G, A, T and C. And the SDR method looks at how chunks of DNA, usually three to five nucleotides long repeat in your genome. For example, G-A-T-A, G-A-T-A and so on. For one person, they might have G-A-T-A repeat six times, whereas another individual might have nine repetitions. A standard SDR test will look at around 20 locations on your genome for where these repeats can occur, and it will count how many repeats you have in each location.

Expert: So if you are looking at 20 of these markers and you've got one from mom and one from dad, you're actually getting 40 different genetic markers coming back. So the chances of somebody having the same combination of genetic markers as you is around one in a billion.

Gregor: Here's what a crime scene sample containing just one contributor looks like on an STR test. You can see how it's easy to compare this to a sample from a suspect. But if the crime scene sample contains DNA from multiple individuals, then their profiles will begin to overlap, all with varying signal strengths. And the more individuals in the mixture, the more difficult these results are to interpret. With four or more individuals, it becomes increasingly hard to compare a clean DNA profile from a suspect to the mixture. And now it's hard to say whether the suspect was actually in the sample. You just can't tell which peak belongs to who. It's kind of like having five people talk to you at the same time. You can hear all the noise, but it's really hard to single one out. Now, some labs claim they can reliably separate samples of up to five individuals.

Gregor: And to test this, the National Institute of Standards and Technology ran a controlled study in 2013. They've sent out a DNA mixture from a fictional crime scene to labs all across the US. Their aim to see how different facilities interpret the same mixture of four people's DNA profiles. 69% of the labs got the analysis wrong. And despite the sample being deliberately complex, only 21% of the labs deemed the mix inconclusive and not possible to give a comparison on. After NIST published the study, new checks have been imposed to address some of the issues, but there's still no lower limit on the quality or the quantity of a DNA sample that labs are permitted to analyze. And labs themselves still decide if something is too mixed or too partial. Now, you might think that using the entire sequence genome would be better, but that can actually introduce different kinds of problems. Now, your analyst has access to hair color, eye color, ethnicity, and that could introduce discrimination in the way the sample is being analyzed.

Expert: This is where the guidelines and the guardrails have to come in here. What are the genetic markers that are being used for this? This is where the really interesting ethical questions come into it. People think that DNA is like the silver bullet that will answer everything. And it is true. DNA evidence is incredibly powerful, it's amazing for identifying individuals, but DNA can never be taken out of context.

Gregor: Now, the point of this video isn't to bash forensics. I think it's still better for us to live in a world where forensics exist rather than one where they don't. But if we want to keep calling forensics a science, we need to continue the work of the people who actually made this video possible, who've dedicated their lives to already reassessing the field and making it as accurate as it is today. Do you think it's fair calling this a forensic science?

Expert: I still think there's science in it because you're trying to do the best with the information you have, which is often times what science is trying to do is, with the information you have, this is the conclusions you make until you know something better.

Gregor: Hey, one last thing. Last year we launched the official Veritasium game, Elements of Truth. It's a trivia game with 800 questions covering science and technology, and we on the team get quite competitive. So every time we play things get very saucy. See, there's this mechanic where you can't only guess an answer, you have to put down a number between one and 10 to gauge how confident you are that your answer is right. So it gets very fun very quickly. The game is coming out later this year, so if you're interested in pre-ordering, you can click the link in the description, which will take you to the Elements of Truth website. Thank you so much for your support in this project. And as always, thanks for watching.