The Same Print, A Different Answer

This is about the sharpest test of an examiner's objectivity anyone has ever run. You take pairs of prints that real latent examiners had personally identified — clear, definite conclusions in live casework. You wait five years. Then you put the very same prints back in front of the very same experts. Nothing about the ridges has changed. The only thing you add is a story: these are the prints the FBI wrongly identified after the Madrid bombings. Prints from two different people, in other words. And then you watch.

That's the whole design, and it's what makes the result so hard to shrug off. There's no novice to blame here. No ambiguous second opinion, no two examiners disagreeing with each other. One expert, one print, one set of eyes, measured against themselves. Forget whether examiners disagree with each other. The sharper question is whether you would still agree with you, once someone whispers the wrong context in your ear.

The study at the door is a fingerprint study, and prints carry the story for the next few pages. But the mechanism belongs to the examiner, not to the fingerprint, and the evidence below runs through DNA, firearms, and the post-mortem room.

The numbers are small, and the authors are upfront about that: five experts, between them over 85 years on the job. The effect wasn't small. Four of the five changed the decision they themselves had made years earlier. Three flatly reversed it, now judging that the prints came from different people. The fourth retreated to I can't tell. Only one expert held the line and identified the prints again.

Set that against what these prints were. Clear, definite identifications, re-verified with no context at all by two more experts with over twenty years apiece. The prints weren't the problem; the story was. Before you decide this is about a handful of unlucky people, weigh how ordinary they were, and how ordinary you are.

There's a fair objection to everything you just read. It's one small study. Five examiners. If that were the whole of the evidence, you'd be right to hold the result loosely, and the authors themselves were the first to flag the modest numbers. But it isn't the whole of the evidence. It was the opening shot. In the years since, the same effect has surfaced again and again, in discipline after discipline, with different materials and different teams. That repetition is what turns a striking anecdote into a finding you can build a practice on.

And it's not just fingerprints. In a real Georgia case, the analysts who knew the police narrative had testified that the defendant could not be excluded from a DNA mixture. Dror and Hampikian (2011, Science & Justice) sent the same data to seventeen analysts who knew nothing about the case: twelve excluded him, four found the data inconclusive, and only one agreed with the original incriminating call. In firearms, Mattijssen, Witteman, Berger and Stoel (2020, Science & Justice) studied peer review of bullet and cartridge-case casework at the Netherlands Forensic Institute: when the reviewer could see the first examiner's interpretation and proposed conclusion — the usual, non-blind arrangement — the review was biased toward confirming it, and when examiners disagreed, the higher-status examiner tended to carry the outcome. It reaches into the post-mortem room too. Dror, Melinek and colleagues (2021, Journal of Forensic Sciences) gave 133 forensic pathologists a child's death with identical medical facts and varied only irrelevant context about the child and the carer who brought them in; the manner-of-death rulings moved with the story. The title of the same group's earlier comment said it plainly: No One is Immune to Contextual Bias — Not Even Forensic Pathologists (Dror et al., 2018).

Step back and the pattern is hard to wave away. A systematic review gathered the whole literature in one place (Cooper & Meterko, 2019), and a meta-analysis went further and measured it, concluding that forensic experts are neither perfectly consistent with themselves nor immune to context (Dror & Rosenthal, 2008). Notice what that second finding couples together: reliability (agreeing with yourself) and biasability (being moved by what's around you). They're two faces of the same human limit, and we'll keep meeting them together.

The biasing context in the original study wasn't invented for the lab. It was a real catastrophe, and the researchers used it precisely because every examiner would recognise it. After the 2004 Madrid train bombings, the FBI identified a latent print as belonging to an Oregon lawyer, Brandon Mayfield, and additional FBI examiners confirmed the identification. He was arrested. The print wasn't his. It belonged to an Algerian man, and only an unusual chain of events ever exposed the error.

So when a colleague handed each expert in the study a pair of prints and said this is the one the FBI got wrong on Madrid, that single sentence did enormous work. It told the examiner what the right answer was supposed to be, an exclusion, and it wrapped that answer in fear, embarrassment, and the gravity of terrorism. The prints in front of them were not Mayfield's at all. But the story was strong enough to override what their own eyes had already concluded once before.

And hold onto this. The Mayfield disaster wasn't a story about lazy or crooked examiners. It was skilled people, working hard, swept along by an urgent investigation and by each other's confidence. The study deliberately recruited experts who had never seen Mayfield's actual print. The bias couldn't come from recognising the case detail. It could only come from the meaning of the story.

That is the uncomfortable bridge to your own bench. You will rarely be handed a print with a label that says find a non-match. But you are routinely handed context (he confessed, there's CCTV, the victim picked him out) that tells you what answer the system is hoping for. Madrid was just a louder version of an ordinary Tuesday.

The instinct, when you first read this work, is to take it personally: I would never do that. Resist it, because that reaction is exactly what the research is warning you about. The experts who flipped weren't negligent and weren't dishonest. The authors go out of their way to say their findings do not reflect carelessness, personal fault, or fraud, and that blaming individual error is usually a way to “deflect deeper scrutiny.” What changed the answers was cognition working normally.

Your visual and decision system is built to use context. It fills gaps and uses expectation to resolve what is ambiguous, and that is what makes you fast and skilled most of the time. But in a comparison task it means the story can shape what you see, especially when the mark is degraded. The study found that the experts who held firm were the ones who managed to focus on the data and ignore the context. So the capacity for objectivity is real. It just isn't automatic, and it isn't guaranteed by being good at your job. If anything, expertise can cut the other way: skill runs on expectation, and expectation is exactly the lever that context pulls. No one is simply above this. Not the novice, and not the thirty-year veteran.

Once you treat bias as a mechanism rather than a character flaw, a second thing comes into focus: the reason it's so hard to admit. We are far better at spotting it in other people than in ourselves. When examiners around the world were surveyed about it, a revealing gap opened up. Most accepted that bias is a real concern for the field, fewer thought it touched their own discipline, and fewer still believed it could reach them personally (Kukucka et al., 2017). That descending staircase even has a name: the bias blind spot. It is the most damaging of a handful of reassurances that keep good examiners exposed, each one comforting and widely repeated and flatly contradicted by the evidence.

This is why the framing matters so much for testimony. Treat bias as a character question (are you the sort of person who gets influenced?) and the only candid answer sounds like an admission of weakness, so every examiner is cornered into denial. Treat it instead as a mechanism of normal perception, and the conversation changes. Of course context can influence a conclusion. It influences everyone's, because that's how brains work. So the useful question is not whether you're immune, but what your process does to keep the mechanism from reaching the decision.

There's a detail in the procedure of Dror's 2006 study that is easy to skim past and shouldn't be. The experts were explicitly told to ignore the context. They were asked to set the background aside and focus only on the ridges, the very thing examiners always promise a jury they do. Most of them changed their answers anyway.

That detail matters. Bias of this kind doesn't feel like bias from the inside. It doesn't arrive as a thought you can notice and override. It arrives baked into what the print looks like to you. You can't try harder your way out of it, because the influence has already shaped the perception by the time you go looking for it. Worse, the trying can backfire. The harder you strain to suppress a piece of information, the more it presses on the decision.

Telling an examiner to disregard what they've been told is about as effective as telling someone not to think of a red door. The instruction names the thing, and the naming is the problem.

Once you grasp that exposure can't be reversed, the whole strategy for handling bias changes shape. There's no point in promising to be vigilant about information you've already absorbed. The damage, if there is any, is already done. The only move that works is upstream: control what reaches you, and when, so the biasing information never touches the analysis in the first place.

This is also why “I just put it out of my mind” is the worst possible answer on the stand. It claims a power over your own cognition that this research shows you don't have, and a sharp cross-examiner knows it. Far stronger to say it directly: I can't unknow what I was told, so my process is built to make sure I'm not told it until after I've decided.

When examiners get asked about bias, they tend to reach for a small set of reassurances. Most of them feel airtight and prove nothing. The trouble is that the comforting answers all share a shape. They locate the protection inside the examiner (my experience, my focus, my second look), and this whole literature is precisely about how the examiner is the part that gets moved.

A real safeguard has a different shape. It's structural. It lives outside your head, in the way information flows and the way work is checked, so it keeps working even on a day when your judgement is tilted. Run the list below. For each one, decide whether it protects the conclusion, or whether it just feels like it does.

The research isn't a counsel of despair. One expert in five held the line, looked past a strong, frightening context and called the print as the data demanded. The authors are clear that this proves objectivity is possible, and that some examiners simply aren't optimising for it. The difference between the one and the four is not moral fibre; it's whether the protection is left to the individual or built into the process.

So build it into the process. Control the information first. A case manager should screen out domain-irrelevant detail, such as confessions, criminal history, or what the detective believes, so it never reaches the analysis. Sequence your exposure. Examine the crime-scene mark and commit to what it contains before the comparison print is unmasked, so your conclusion forms from the evidence rather than working backwards from a suspect. That approach is known as linear sequential unmasking. Line up alternatives. Rather than carry one expected answer, hold several competing possibilities and make the data choose between them. Verify blind. The checking examiner should not know what you decided, or the verification just re-runs your bias with a second signature. And document as you go. Record what you knew, and when, so the order of events can be examined later.

Notice what every one of these has in common: none of them asks you to be smarter or more disciplined than the experts in the study. They assume you are exactly as human as the four who flipped, and they put the safeguard where your willpower can't be reached. That is the whole point. You don't earn objectivity by being above bias. You protect it by arranging your work so bias has nowhere to enter.

And there's a bonus. Every one of these safeguards is also an answer. When counsel asks how you guarded against influence, you don't have to protest your integrity. You can describe a system: what you were shielded from, in what order you looked, who checked you blind, what you wrote down. A process you can point to is far harder to cross-examine than a virtue you can only assert.

Sooner or later, a cross-examiner armed with this research will ask the question it was built to set up: given everything you'd been told about my client, isn't it possible that context influenced your conclusion? There is a wrong answer that feels like the right one, and this is the moment to get right.

The wrong answer is denial. No, I'm too experienced, I focus only on the evidence, context doesn't affect me. It feels strong and it's fatal, because this very body of work shows experts with decades of experience reversing their own calls, after being told to ignore the context, on prints they themselves had matched. The denial is also the textbook blind spot, the belief that bias is something that happens to other examiners, and counsel can quote you the surveys to prove it. The moment you claim immunity, you are now both wrong and the kind of witness who doesn't understand their own field.

The strong answer concedes the mechanism and then stands on the process. Yes, context can influence any examiner's conclusion, including mine, and it can work without my noticing. That's exactly why I don't rely on willpower. Here's what I was shielded from, here's the order in which I examined the material, here's who verified my work without knowing my answer, and here's what I documented. You've agreed with the science, refused to claim a superpower you don't have, and handed the jury a reason to trust you that doesn't depend on taking your word for it.

That is the shape of credibility this research points to. The evidence has already refuted the witness who insists they can't be moved. What holds up instead is the witness who understands precisely how they could be moved, and built their work so it doesn't happen.