We Investigated How Accurate Forensic Methods Are. Here's What We Found.

[00:00] (dramatic music)

[00:03] - A man has been found lying dead

[00:05] at the side of the road with oozing,

[00:07] elongated wounds to his back.

[00:09] To solve the murder,

[00:10] a local official calls all

[00:13] and demands they all bring

[00:16] In the square, each man is

[00:19] And then something

[00:23] In the rising heat of the

[00:26] all landing on a single sickle.

[00:29] Its owner buckles,

[00:30] falling to his knees and admits to murder.

[00:36] This is from a 1247 book,

[00:38] "The Washing Away of Wrongs" by Song Ci.

[00:40] And it's the first writing we have

[00:42] of an empirical approach to forensics.

[00:44] But okay, big deal.

[00:46] Anyone could have just swapped the sickles

[00:47] or used their neighbor's.

[00:50] But forensics today are

[00:54] Fingerprints support investigations

[00:58] and DNA evidence in more than 90%.

[01:00] But then you also read an article like

[01:02] "Hair sample that put men

[01:06] turned out to be dog hair."

[01:07] - That is not science.

[01:09] That is not justice.

[01:10] - See, in 2009, the

[01:13] published this 350 page

[01:17] with the exception of

[01:19] no forensic method has

[01:22] to consistently demonstrate a connection

[01:24] between evidence and

[01:26] And they also said that some tests

[01:28] do not meet the fundamental

[01:31] So we ask people to rank five

[01:34] all of which are still

[01:38] Okay, five forensic techniques here.

[01:39] And I just need you to rank them in order

[01:41] of accuracy from the

[01:43] to the most accurate.

[01:44] So what do you think?

[01:45] - I don't know if I'm gonna nail these.

[01:47] - Whoa. I dunno.

[01:48] - I don't know what

[01:50] - Okay, hair. What are you thinking?

[01:52] - I think hair is last, eh?

[01:53] - I would probably put

[01:57] - I feel like hair would

[02:00] - To fake?

[02:01] - I think hair at the bottom,

[02:04] because if it's not got

[02:05] just hair in general.

[02:07] - Mm.

[02:08] - Yeah.

[02:09] (dramatic music)

[02:11] - Hair has always been

[02:13] for crime scene investigators.

[02:15] And before DNA, they would look

[02:17] for the microscopic

[02:19] and the cross section of

[02:22] For example, analyzing

[02:25] like structure of the

[02:27] or how the pigment is distributed

[02:30] - [Announcer] The examiner's trained eye

[02:31] can learn many things.

[02:33] Is it human hair? Of what race?

[02:35] Animal hair? What family?

[02:37] This vital information may help

[02:39] to establish guilt or innocence

[02:41] - Between the 1970s and 1999,

[02:44] the FBI used this against

[02:49] But when they later reexamined

[02:52] it completely flipped

[02:55] and it turned out that even

[02:58] examiners would routinely

[03:00] between hairs that were

[03:02] And they sometimes couldn't

[03:04] between a human hair and a dog hair.

[03:07] (dog barking)

[03:08] Out of those 268 cases that

[03:12] 96% were declared false.

[03:15] 33 people were already sentenced to death,

[03:17] and nine were already executed

[03:19] by the time the errors were found.

[03:21] Today, the FBI only uses hair as evidence

[03:24] if it can be supported by DNA testing,

[03:26] which is why microscopic hair analysis

[03:28] goes to the bottom of this list.

[03:30] What do you think of bite marks?

[03:31] - On a crime scene?

[03:32] - Yeah, yeah.

[03:33] Dracula, is Dracula there?

[03:34] - Least accurate is gonna be

[03:36] pretty close to bite marks, I think.

[03:38] - Midway.

[03:39] - Midway?

[03:40] That's last.

[03:41] That's not realistic, innit?

[03:43] Mm.

[03:44] Bite marks at the top.

[03:45] The idea behind bite marks is simple.

[03:47] If you can use dental records

[03:50] surely you can identify a perpetrator

[03:52] by the bite marks left on a victim.

[03:54] Since the 1950s,

[03:55] bite mark analysis has been

[03:59] But then some experts, like

[04:02] started looking into

[04:05] - There was no scientific exploration

[04:08] into bite marks before they

[04:11] to say that this was even a feasible

[04:13] technique that you could do.

[04:15] - One study from the university at Buffalo

[04:17] used a set of model teeth to create

[04:19] 89 bite marks in cadavers,

[04:21] alongside controlled

[04:24] And out of those 89

[04:27] none of them matched the measurements

[04:28] of the ones made in wax or

[04:31] In fact, when they compared

[04:34] to a wider collection of 411 model teeth,

[04:37] it wasn't even the original

[04:40] that was the closest match.

[04:42] After 12 studies, their conclusion,

[04:45] bite mark transfer to

[04:48] Skin isn't a good medium

[04:49] to leave an imprint of

[04:51] It's soft, it's squishy, and

[04:54] But despite all this,

[04:57] and bite mark evidence is still

[05:00] We asked Dr. Bush for a comment

[05:02] and she said that while

[05:04] is happy to discard an

[05:08] the justice system prefers consistency

[05:09] in a historical precedent.

[05:11] So even the new flurry

[05:13] to get bite marks out of court.

[05:15] In fact, bite marks have

[05:17] as recently as 2025,

[05:19] which is why it gets the

[05:23] Blood, if you don't have DNA

[05:24] and you're just using

[05:26] that it leaves along the ground.

[05:27] How they splat.

[05:28] - [Gregor] Like the picture,

[05:29] that blood on a crime

[05:31] - Okay.

[05:32] thinking of putting blood at the end,

[05:33] but like, I don't know.

[05:34] Put blood at the end.

[05:35] - I would go blood then

[05:36] - Second.

[05:37] - Yeah, yeah, yeah, yeah.

[05:38] - So bloodstain pattern analysis.

[05:40] When someone is injured,

[05:43] and drips into the environment

[05:45] And in 2020, "Wired"

[05:48] with a crime scene analyst

[05:49] on how these different

[05:52] You can use physics and

[05:55] and extrapolate where in space

[05:57] the blood must have come from.

[05:59] This all started in 1971 when chemist,

[06:01] Herbert Leon MacDonell, published

[06:03] "Flight Characteristics and

[06:06] In this book, MacDonell

[06:08] the width and length of a stain

[06:10] to calculate the angle of impact,

[06:12] and then using trigonometry,

[06:14] trace that back to the point

[06:18] The book quickly became the foundation

[06:20] of bloodstain pattern analysis,

[06:21] and everyone was using it.

[06:23] In fact, MacDonell even

[06:26] where thousands of US police

[06:29] The Supreme Court of Iowa

[06:32] as relatively uncomplicated.

[06:34] So they didn't require

[06:37] They were happy to go

[06:39] that he did alone in his basement.

[06:42] And soon enough,

[06:44] was used in courts all over the states.

[06:48] But here's the problem.

[06:49] Plotting the flight

[06:51] using trigonometry

[06:53] were straight lines and gives

[06:57] which suggests the victim was standing up.

[07:00] The issue is,

[07:01] MacDonell and then many forensic

[07:04] and even "Wired," failed to account

[07:05] for the effects of gravity and drag

[07:07] as obvious as this sounds.

[07:09] When you include these

[07:11] the common origin point lowers,

[07:13] and you can see that in this example.

[07:14] It's more likely that the victim

[07:18] The first study to measure

[07:21] of bloodstain pattern analysis

[07:25] That's over 50 years

[07:27] after it started being admitted in court.

[07:29] And a large scale study from

[07:32] come to different conclusions

[07:35] about 8% of the time.

[07:36] And that's because the same stain

[07:37] can come from many different mechanisms.

[07:39] And the fact that blood

[07:42] depending on how many red blood cells

[07:43] versus the plasma you have.

[07:45] Men actually tend to have

[07:48] of blood cells than women,

[07:50] which makes their blood more viscous.

[07:52] Nowadays, investigators have new software

[07:54] which uses fluid dynamics

[07:56] to help them map these complex 3D seams.

[07:58] So bloodstain pattern analysis

[08:02] which is why it gets this

[08:04] Okay.

[08:05] But surely fingerprints

[08:07] must have a lot of

[08:09] How unique do you think fingerprints are?

[08:11] - I think they're pretty accurate.

[08:13] - They're fingerprints.

[08:14] - That's probably one.

[08:15] - Yeah.

[08:16] I mean, yeah, you use

[08:18] you unlock it, you pay stuff, right?

[08:19] - 'Cause fingerprints are unique.

[08:21] Even when you have a passport,

[08:22] like they scan you when

[08:25] - On March 11th, 2004,

[08:27] a terrorist organization

[08:29] set off explosions on

[08:32] The attack killed 193 people

[08:37] Quickly after, the police

[08:39] on a detonator bag left behind,

[08:41] and they matched it to Brandon Mayfield,

[08:43] a lawyer in Oregon.

[08:45] There's only one problem,

[08:47] there are no records of

[08:50] He doesn't even own a passport.

[08:52] So how could he have pulled this off?

[08:55] - I honestly felt like I was being framed

[08:57] because I hadn't been outta

[09:00] - But despite that, the

[09:02] was considered so damning that

[09:06] But this wasn't a glitch.

[09:08] It was built into the

[09:10] (groovy music)

[09:12] In 1890s Kolkata,

[09:13] the city's rich career

[09:15] by paying people to

[09:18] It was a good tactic since, at the time,

[09:20] there was no practical

[09:23] So the criminals were

[09:25] Three officers, Edward Henry, Azizul Haque

[09:27] and Hem Chandra Bose

[09:28] started taking the prints

[09:29] of everyone who came through the station,

[09:32] but they quickly ran into a problem.

[09:34] When they got a suspect in,

[09:37] how could they crosscheck their prints

[09:39] against their own database

[09:40] when there were 10 prints per person

[09:42] and thousands of people coming through?

[09:44] They needed a classification system.

[09:47] Haque proposed that they should look at

[09:49] whether a fingerprint has a whirl.

[09:51] It's the spiral pattern

[09:52] that a person might have on their fingers.

[09:54] And because each of your 10 digits

[09:56] can either have a whirl or not,

[09:59] there are a total of

[10:01] or 1024 ways your whirls can be arranged.

[10:05] So the officers just built

[10:09] to organize these fingerprints.

[10:13] Okay.

[10:13] But what if like me,

[10:15] Well, two thirds of the

[10:18] So the system had to go further.

[10:20] Once you have a pigeonhole number,

[10:21] you go for additional

[10:24] Some of the additional categories

[10:25] that you could assign included

[10:28] which was denoted with the letter A,

[10:30] or whether you had a tented arch,

[10:32] which was denoted with a T.

[10:34] Similarly, you could look at which side

[10:36] of the finger this loop structure is on.

[10:38] And with a few more

[10:41] we get a much more

[10:44] one that would go on to be

[10:48] Famously, the world's

[10:51] which was established in London

[10:54] used this fingerprinting system.

[10:56] - I started in 1988

[10:59] in forensic services.

[11:01] - Could you paint a

[11:04] would've looked like with

[11:06] How big was this space?

[11:07] - So we're looking at a space of about

[11:10] 30 to 40 meters in length

[11:12] and about 10 to 20 meters in width.

[11:15] The collection at the time was about

[11:18] 3 to 4 million I think.

[11:19] - That is intense.

[11:20] Yeah, that's wow.

[11:21] - It's a lot of fingerprints.

[11:24] - Now, today's system takes into account

[11:27] these little details along the ridges.

[11:29] It's called friction ridge analysis.

[11:31] And well, it's no surprise that

[11:32] combing through pigeon holes

[11:35] but online databases today can

[11:37] (phone ringing)

[11:41] No way this phone works.

[11:44] Hello?

[11:46] Just trying to get a hold of you.

[11:47] - Why are you calling

[11:49] (caller speaks faintly)

[11:50] You have no data?

[11:51] Just get Saily.

[11:53] - You know, I travel a lot for filming.

[11:54] I actually just got to London.

[11:58] I always wanna be able to

[12:00] as soon as I land,

[12:01] and with the help of

[12:04] it couldn't be easier.

[12:06] First, you just scan the QR

[12:09] then you search your

[12:12] And next you just pick a data plan.

[12:14] And finally, during checkout,

[12:15] apply the coupon code Veritasium

[12:16] to get your exclusive 15% discount.

[12:19] No more dodgy cafe wifi or

[12:24] All you need to do is just

[12:25] switch on your mobile data when you land,

[12:27] and that's it.

[12:28] Saily works with all

[12:30] And once an eSIM is installed,

[12:33] you can keep adding plans in

[12:36] And in the rare case that your phone

[12:37] isn't compatible with Saily,

[12:39] you can get a full refund

[12:42] I've gone with the Saily Ultra Plan,

[12:43] which gets you unlimited data globally

[12:45] with access to airport

[12:48] and NordVPN, NordPass,

[12:52] So to stay connected on the

[12:54] you can scan this QR code, download Saily,

[12:57] and get 15% off your plan with

[13:00] So I wanna thank Saily for sponsoring

[13:02] this part of the video,

[13:03] and now back to our case.

[13:05] So if all the details

[13:08] into these databases,

[13:12] How is Brandon Mayfield falsely accused?

[13:15] Well, say you have a mark from a scene

[13:17] and you wanna run it through

[13:20] you can't actually search the

[13:24] because you first need to identify

[13:25] a number of features on the print.

[13:28] These are called minutiae.

[13:30] It's where the ridge line split or end.

[13:33] - To even search the database,

[13:36] you need a minimum number

[13:40] It's very efficiently sorted,

[13:42] hundreds of million prints,

[13:43] but it cannot make an ID.

[13:45] So what it does, it pulls

[13:49] at least 10, 20 up to

[13:54] And then the human

[13:58] - [Gregor] This works in

[14:00] how these minutiae are

[14:04] - What is interesting in my research

[14:06] published scientific reviewed research,

[14:09] the same thing one examiner

[14:14] and one sees 10 minutiae.

[14:16] So 10 minutiae is enough

[14:19] However, if the other are

[14:23] they couldn't run it.

[14:24] So whether you find a criminal or not

[14:27] depends on the luck of the draw.

[14:29] Even if you give the same pair of prints

[14:31] to the same examiner

[14:35] they will reach a different decision.

[14:38] - And these disagreements

[14:41] where the examiners had no

[14:44] but in reality, they often do.

[14:46] As many as 42% of the requests

[14:48] that fingerprint experts process

[14:50] state whether a suspect

[14:53] - In a recent case that I was involved

[14:57] where a firearm was involved

[14:59] and a forensic firearm expert

[15:02] had to decide whether a

[15:06] on the firearm, on the suspect.

[15:08] On the former submission form

[15:11] that was given to the forensic scientist,

[15:14] it said that it's homicide,

[15:17] it said the age of the

[15:20] and it even said the race.

[15:22] It said the suspect is black,

[15:24] and the person who died

[15:28] - For over a century,

[15:29] experts have held so much

[15:32] that in the FBI case for Mayfield,

[15:34] the examiners claimed with 100% certainty

[15:37] that that was his mark.

[15:39] This likely happened because

[15:41] was brought in to verify the match,

[15:43] they probably already knew the verdict

[15:45] of the original analyst,

[15:47] and they probably also knew the stakes

[15:49] of what a match could mean for the case,

[15:51] which made it all the easier to agree

[15:53] with the initial verdict.

[15:55] Even the fingerprint expert

[15:57] agreed that the mark matched

[16:01] when in reality, it belonged to a man

[16:03] who had links to terrorist

[16:07] This is the impact of conformity bias.

[16:09] - So all of this pressure,

[16:10] all of this context and intervention

[16:13] in the forensic science,

[16:15] I say to the police and the prosecutor,

[16:17] leave the forensic examiners alone.

[16:21] Give them independence of mind

[16:23] to make decision based on the

[16:28] and do the work rather than

[16:33] - Well, if there were issues

[16:35] with all these other techniques,

[16:37] why don't we just use,

[16:38] - DNA?

[16:40] - Yeah.

[16:41] - DNA has got to be surely,

[16:44] - It has to be DNA because,

[16:46] - You already know everyone

[16:48] - But even DNA analysis can

[16:52] Back in the 80s,

[16:53] you had to do this with a sample the size

[16:56] of a dime of blood or saliva.

[16:58] But today you can do it with

[17:00] In fact, just this year

[17:03] was solved with just 0.4 nanograms of DNA.

[17:08] But the paradox is, the

[17:11] of DNA analysis could

[17:15] In 2012, paramedics in

[17:18] about a severely intoxicated homeless man.

[17:21] They treated him and

[17:24] but along the way, they

[17:26] of his DNA on their equipment and gloves.

[17:29] Then later that night,

[17:31] of a businessman and they

[17:34] the homeless man's DNA to

[17:38] Because of that, the

[17:40] was charged with murder and

[17:43] despite being hospitalized at

[17:46] He spent five months in jail

[17:48] before charges were finally dropped.

[17:51] - So this is trace DNA and touch DNA,

[17:54] where you can get this kind

[17:57] - I see how that could be abused easily.

[17:59] So my DNA could be in a

[18:01] - Yeah.

[18:02] - And if DNA is deposited

[18:05] say a dark dry place, it can

[18:10] - And if, for example,

[18:14] well, loads of people are gonna

[18:16] so you're definitely going to get

[18:18] more than one profile off of there.

[18:21] - [Gregor] These samples

[18:23] are called DNA mixtures,

[18:25] and they have been found to be

[18:26] the most common source of

[18:29] - Some people shed more cells than others,

[18:32] so the majority of the

[18:35] off of that door handle may not be

[18:36] from the last person who touched it.

[18:38] - The problem in interpreting a mixture

[18:40] comes down to how a sample gets analyzed.

[18:43] The most common way to do it

[18:44] uses short tandem repeats or STRs.

[18:48] DNA is made up of four nucleotides,

[18:50] G, A, T and C.

[18:52] And the SDR method looks

[18:55] usually three to five nucleotides long

[18:57] repeat in your genome.

[18:59] For example, G-A-T-A, G-A-T-A and so on.

[19:02] For one person, they might

[19:06] whereas another individual

[19:10] A standard SDR test will look at around

[19:13] 20 locations on your genome

[19:14] for where these repeats can occur,

[19:16] and it will count how many repeats

[19:18] you have in each location.

[19:20] - So if you are looking

[19:23] and you've got one from

[19:25] you're actually getting 40 different

[19:27] genetic markers coming back.

[19:29] So the chances of somebody

[19:34] of genetic markers as you

[19:38] - Here's what a crime

[19:39] just one contributor

[19:43] You can see how it's easy to compare this

[19:44] to a sample from a suspect.

[19:47] But if the crime scene sample contains DNA

[19:49] from multiple individuals,

[19:51] then their profiles will begin to overlap,

[19:53] all with varying signal strengths.

[19:55] And the more individuals in the mixture,

[19:57] the more difficult these

[20:00] With four or more individuals,

[20:01] it becomes increasingly hard to compare

[20:03] a clean DNA profile from

[20:07] And now it's hard to

[20:09] was actually in the sample.

[20:10] You just can't tell which

[20:13] It's kind of like having

[20:15] at the same time.

[20:17] You can hear all the noise,

[20:18] but it's really hard to single one out.

[20:20] Now, some labs claim they

[20:24] of up to five individuals.

[20:26] And to test this,

[20:27] the National Institute of

[20:30] ran a controlled study in 2013.

[20:33] They've sent out a DNA mixture

[20:35] to labs all across the US.

[20:37] Their aim to see how different facilities

[20:40] interpret the same mixture of

[20:44] 69% of the labs got the analysis wrong.

[20:47] And despite the sample

[20:50] only 21% of the labs

[20:54] and not possible to give a comparison on.

[20:56] After NIST published the study,

[20:58] new checks have been imposed

[21:01] but there's still no lower limit

[21:03] on the quality or the

[21:05] that labs are permitted to analyze.

[21:08] And labs themselves still decide

[21:10] if something is too mixed or too partial.

[21:13] Now, you might think that

[21:14] using the entire sequence

[21:17] but that can actually introduce

[21:20] Now, your analyst has

[21:22] eye color, ethnicity,

[21:24] and that could introduce discrimination

[21:25] in the way the sample is being analyzed.

[21:28] - This is where the guidelines

[21:29] and the guardrails have to come in here.

[21:32] What are the genetic markers

[21:36] This is where the really interesting

[21:37] ethical questions come into it.

[21:39] People think that DNA is

[21:41] that will answer everything.

[21:44] And it is true.

[21:45] DNA evidence is incredibly powerful,

[21:48] it's amazing for identifying individuals,

[21:52] but DNA can never be taken out of context.

[21:57] - Now, the point of this

[22:00] I think it's still better

[22:01] where forensics exist rather

[22:05] But if we want to keep

[22:08] we need to continue the work of the people

[22:10] who actually made this video possible,

[22:12] who've dedicated their lives

[22:13] to already reassessing the field

[22:15] and making it as accurate as it is today.

[22:17] Do you think it's fair calling

[22:19] - I still think there's science in it

[22:21] because you're trying to do the best

[22:22] with the information you have,

[22:24] which is often times what

[22:27] with the information you have,

[22:28] this is the conclusions you make

[22:30] until you know something better.

[22:32] (gentle music)

[22:44] - Hey, one last thing.

[22:46] Last year we launched the

[22:49] Elements of Truth.

[22:50] It's a trivia game with 800 questions

[22:52] covering science and technology,

[22:54] and we on the team get quite competitive.

[22:56] So every time we play

[22:59] See, there's this mechanic

[23:00] where you can't only guess an answer,

[23:02] you have to put down a

[23:04] to gauge how confident you

[23:07] So it gets very fun very quickly.

[23:09] The game is coming out later this year,

[23:11] so if you're interested in pre-ordering,

[23:12] you can click the link in the description,

[23:14] which will take you to the

[23:17] Thank you so much for your

[23:19] And as always, thanks for watching.