“Inconclusive” Is Not the Safe Answer

Of all the words in your reporting vocabulary, inconclusive feels like the safe one. It’s humble. It admits the limits of the evidence. And here’s the appeal: it can never be proven wrong, because there’s no way to check it.

That last part is exactly the problem.

When you say identification or exclusion, you can be right or wrong, and one day someone might find out. When you say inconclusive, you step off the scoreboard. Nobody can ever say you erred. So it’s tempting to reach for it whenever a comparison gets hard.

This reading is about what that one word is doing. It shapes your conclusions, your discipline’s claimed error rate, and how you hold up on the stand. A scope note before we start: the studies and worked examples here are fingerprint studies, because that is where the measurement has been done. The same fight over how to count inconclusives is running right now in firearms identification, and the two ideas below carry to any comparison discipline — but the numbers are fingerprint numbers. There’s a bit of cognitive science underneath it (psychologists call it signal detection), but you won’t need a single equation. Just two ideas, kept in everyday terms. By the end, you’ll see why inconclusive is the most slippery word you use, and how to use it well.

Start with a fact that sounds obvious but explains a great deal of disagreement. When you compare two prints, two different things are going on at the same time, and they are easy to confuse.

The first is your eye, your ability to tell a true match from a true non-match. Call it skill. Some people are simply better at seeing whether two prints came from the same finger.

The second is your bar: how sure you need to be before you’ll commit. This isn’t about how good your eye is. It’s how high you set the threshold before you’ll say the word out loud.

The clearest way to feel the difference is the smoke alarm on your ceiling. Inside is a sensor. That’s its eye, how well it tells real house-fire smoke from harmless toast. On the side is a sensitivity dial. That’s its bar, how little smoke it takes before it screams. Turn the dial up and a wisp of toast sets it shrieking. Turn it down and it waits for a thick haze. The dial does nothing to the sensor. It only changes how much smoke the alarm demands before it will commit to fire.

Now put two identical alarms on two ceilings, same sensor, same factory, one dialled twitchy and one dialled calm. The same thread of toast smoke drifts up. One screams, the other stays silent. Opposite verdicts on the very same wisp, and neither alarm is better. They simply have the dial set in different places. That is your bar, laid bare, with skill held perfectly still. And note that even a flawless sensor still has a dial. Being a brilliant examiner doesn’t spare you the choice of where to set the bar. It only means your eye is sharp while you make it.

You already know a place where this plays out with everything on the line. Give two equally sharp juries the very same borderline case. One convicts. The other acquits. Not because one jury is wiser, or saw something the other missed, but because one holds beyond reasonable doubt at a hair’s breadth of lingering doubt and the other holds it at near-certainty. Same evidence, same eyes, different bar, opposite verdict.

That is the closest mirror there is to two examiners disagreeing on a borderline latent. You might say: but I’m not the jury; my comparison is meant to be objective. True, and beside the point. Every time you write identification or inconclusive, you reach a verdict at a bar you have chosen, and the bar is real whether or not you can feel yourself setting it. (Australians will know the umpire’s version. Two officials with equally good eyes still split on a 50/50 LBW, because one lives by when in doubt, not out and the other doesn’t. A higher bar doesn’t mean sharper eyesight; it’s a decision about which mistake you’d rather make.)

There is hard evidence the two really do come apart. When Thompson, Tangen and McCarthy (2013) — the study this reading is built on — tested examiners against novices, the experts did more than get answers right. They answered decisively. 92% of their judgements fell at the very ends of the scale (sure same or sure different), against only 32% for novices. The expert eye doesn’t dither in the middle. Which makes it all the more telling when an expert chooses the middle anyway.

Hold onto this: skill and willingness-to-commit are not the same thing. Almost everything confusing about inconclusive comes from mixing them up.

Now bring back inconclusive, and watch the trouble start.

In a proper test, we know the truth: these two prints really did come from the same finger, those two did not. That’s ground truth, and it is what lets us grade an identification or an exclusion as right or wrong.

But inconclusive isn’t a claim about whether the prints match. It’s a claim about whether the print is good enough to decide. It’s a claim about sufficiency. And the catch almost no one says out loud is this: there’s no ground truth for sufficiency. Nobody knows, as a fact, whether a smudged latent really does or doesn’t hold enough information for a human to make the call. So when you say inconclusive, there is nothing to check it against. It can’t be scored right or wrong. It simply stands unchecked.

Worse, inconclusive is two completely different statements wearing one label. It can mean “there isn’t enough detail here to decide.” Or it can mean “there might be enough, but I’m not willing to commit.” Those are opposites. One is about the evidence, the other is about you, and the single word hides which one you meant.

That is the uncomfortable consequence. Suppose the information was there, the print was good enough, and you said inconclusive anyway. You didn’t avoid an error. You committed one: a missed call. And because sufficiency can’t be measured, that miss is invisible. Inconclusive becomes the place errors go to hide.

This isn’t a fringe worry. When the FBI ran its landmark accuracy study, disagreement about sufficiency was right there in the findings: “examiners frequently differed on whether fingerprints were suitable for reaching a conclusion.” The thing we can’t score is also the thing examiners can’t agree on.

This is where the abstract problem turns into a courtroom problem.

When your discipline reports an error rate, the famous “about one in a thousand” kind of figure, someone had to decide what to do with all the inconclusive answers in the study. That one decision sets the number.

Leave the inconclusives out, treat them as “no answer, move on,” and the rate shrinks, because many of the hardest comparisons, the ones most likely to trip an examiner, never enter the tally. Count the inconclusives that should have been decided as errors, and the number climbs, sometimes steeply.

So a single bookkeeping choice, made by researchers and not by you, can move the headline error rate from reassuring to alarming. That argument is live in the literature right now, unresolved, with serious people on both sides. Itiel Dror and Nicholas Scurich (2020) argue that setting inconclusives aside flatters the figures and buries real mistakes. Others argue the reverse, that an inconclusive is a legitimate “I can’t tell,” not an error, and that counting it as one is unfair to the examiner.

You don’t have to settle that fight. But you do have to know it exists, because a prepared cross-examiner does.

None of this means inconclusive is a dodge. Often it is exactly right. Some latents really are a hopeless smear of three ridges, and calling them inconclusive is just telling the truth.

The difficulty is that the same word covers a truthful “there isn’t enough here” and a much more evasive “I’d rather not stick my neck out.” Occasionally it covers something worse: a conclusion that moved because of the case story rather than the ridges — an examiner heading for a definite call until word arrived that the suspect had confessed, who then wrote inconclusive instead. That is contextual bias wearing the safest label, and it has a reading of its own (The Same Print, A Different Answer). From the outside, and often from the inside, these kinds of inconclusive are hard to tell apart.

Researchers have even asked whether examiners, as a group, have drifted too cautious: so wary of the career-ending false identification that they retreat to inconclusive on calls they could, and arguably should, make. There’s no tidy way to measure it, for the reason you now know. But you can at least be straight with yourself about which kind of inconclusive you are writing. Try it below.

There’s a tidy way out of all this, and it’s worth understanding, along with why it isn’t the whole answer.

The tidy way is the one this reading has been building toward: stop arguing about the third category, and measure the two real things separately. Measure the eye, how well examiners tell matches from non-matches. Measure the bar, how willing they are to commit. Keep those apart, and much of the confusion about inconclusive dissolves. A cautious examiner isn’t a worse examiner, just one with a higher bar. A buried inconclusive is a threshold set out of view. This is the signal-detection view, and on its own terms it’s right.

But there’s a catch, and the reason the argument is still running. That tidy framework treats an inconclusive as, in effect, a very cautious no. In a laboratory, fine. In a courtroom, an inconclusive and a wrong identification are not the same thing at all. They fall on a defendant’s life with completely different weight. A framework that scores them on one scale captures the maths and can miss the stakes. Sinha and Gutiérrez made exactly that point in 2023, and they’re not wrong either.

So this is one of those rare places where the science hasn’t resolved itself. Careful people, looking at the same evidence, still disagree about what an inconclusive even is. You don’t need to pick a winner. You need to know the ground is contested, so that when you’re asked, you can show you understand the debate instead of pretending it doesn’t exist. On the stand, that’s the difference between an expert and a technician.

Put the pieces together, and a small, clear playbook falls out.

Know that “inconclusive” is doing hidden work. It’s the one answer that can’t be scored, the place errors hide, and the hinge your discipline’s error rate swings on. Treating it as the bland, safe option is the one thing you can’t afford.

Don’t use it as a hedge. If the information is there, make the call. An inconclusive you can justify only on your nerves, not on the evidence, is a miss with better public relations.

Separate your eye from your bar, out loud. Asked why you and a colleague differed, you can say what most witnesses can’t. You may have seen the print the same way and simply set the threshold for committing differently. Then explain where you set yours, and why.

And be straight about the number. When counsel asks for your error rate, the answer that survives is the one that begins: it depends on how the studies counted the cases where examiners didn’t answer. Then say what that means for what you can claim today.

All of this makes you more authoritative, not less. The examiner who understands what “inconclusive” really is, and says so directly, is far harder to rattle than the one who reaches for it because it feels safe.