Why we can’t have nice things
In their 2025 article “First Impressions Matter,” Simon Cole and Justin Sola trace the long, winding road of forensic science’s* failed statistical revolution. They focus on fingerprint analysis, where, despite decades of critical scrutiny, software development, and even well-funded institutional backing, a full shift to probabilistic reporting has yet to materialize. Most forensic fingerprint testimony still leans on categorical, often binary language (identification or exclusion) with little to no reflection of the underlying uncertainty.
Cole and Sola focus their attention not on the usual suspects (practitioner resistance and courtroom inertia) but rather ask why the technology needed to implement statistical reasoning still doesn’t exist in usable form. Their answer is mundane but accurate: There is no working “forensic device,” as they call it, that can operationalize the abstract promise of forensic statistics. A statistical model is only part of the picture; it needs a functioning software front-end, validated data pipelines, update schedules, legal usability, and sustained institutional buy-in. It needs infrastructure and that infrastructure doesn’t yet exist, not for any fingerprint software.
The “model gap” is real for nearly every kind of evidence other than DNA. This gap is not just technical, it’s ontological.1 What kind of thing is being measured? How stable is it? What defines its population? How are we measuring it? For DNA, those answers are unusually tractable2; for friction ridge impressions (and all other forensic evidence), less so. For some evidence types, like fibers, they might be beyond reach entirely, at least in any unified, generalizable way.
Fibers are really, really complicated
This becomes especially clear when we shift the focus to textile fibers, as Tiernan Coyle has done in his ongoing critique of forensic materials analysis. Coyle’s work centers on the evidence itself, not just how it’s handled in court or compared in the lab, but how it is made, circulated, and encountered in the real world. His position is simple, but disquieting: Most fiber evidence is too variable, too poorly documented, and too embedded in global supply chains to support probabilistic modeling of any meaningful kind.3 And yet, fiber evidence isn’t useless, far from it. Its probative value, when it exists, emerges only through disciplined framing. That framing starts with how we describe what the fiber is, and just as importantly, what it is not.
Beyond “Consistent With”
Much of the language traditionally used in fiber reporting has relied on the phrase “consistent with.”4 A fiber found on a jacket might be described as “consistent with” a fiber recovered from a car seat. It’s a soft, squishy-but-quite-liquid phrase, less than a match, more than a shrug—a nod, perhaps? But it’s also vague. What is being asserted? That the fibers are the same “type”? That they’re indistinguishable under the tests conducted (what about test that weren’t conducted)? That they likely share a source? Without further clarification, “consistent with” risks becoming a verbal placebo, a phrase that appears scientific without delivering substance.
Maybe there’s a better way to phrase it, and it begins by directly referencing the properties that forensic examiners actually measure. What about:
The questioned fiber has the same morphological, optical, and chemical properties as the known fiber. Based on these shared characteristics, both fibers are assessed to come from the same manufacturing class. No significant distinguishing features were identified using the analytical methods applied.
This kind of statement does three things well. First, it tells the audience what was actually measured: morphology (such as fiber cross-section and surface texture), optical properties (like refractive index and birefringence colors under polarized light), and chemical composition (often via FTIR, microspectrophotometry, or thin-layer chromatography). These are the bedrock domains of fiber comparison, and they are grounded in laboratory practice.
Second, it locates the shared characteristics within an industrial reality, the “manufacturing class.” This phrase doesn’t imply that the fibers came from the same object or the same factory. Instead, it acknowledges that they fall within a category of production, defined by shared material inputs and processing methods. A class might include millions of fibers used in garments sold across multiple continents, or it might be as narrow as a production run for a custom textile. The examiner isn’t asserting identity, they are asserting indistinguishability at a known level of resolution.
And third, it draws a firm boundary around what is being claimed. Saying that “no significant distinguishing features were identified” is not the same as saying the fibers are the same. It means that within the constraints of the methods used, nothing was found to separate them. That’s an empirical observation, not a conclusion about origin.
The Weight of Rarity
At this point, someone might ask: OK, Smart Guy, if two fibers are indistinguishable in these ways, doesn’t that suggest they’re likely from the same source? Isn’t it rare to find fibers that match on all three levels—morphological, optical, and chemical? The short answer is: yes, it’s not common, but, no, that’s not the same as saying it's necessarily meaningful.
Coyle points to an important but often overlooked strand of literature: Target fiber studies. These are controlled experiments designed to simulate the actual conditions under which fiber transfer and recovery occur. They account for wear, contamination, environmental loss, and the background noise of shared environments, among other stuff. What these studies show is instructive. In general, finding a fiber on a person or object that is indistinguishable from a known source is uncommon, especially in cases where no contact is expected. In other words, chance matches in blind conditions are rare.
But “rare” doesn’t mean unique. And it doesn’t mean source-identifying. One reason is volume. The textile industry produces over four billion pounds of fibers every year. Add to that the roughly 500 fiber cross-section types in use at any given time, the thousands of dyes available, and the near-infinite combinations of dye formulations5, treatments, and post-processing finishes, and the scale becomes untraceable. A few fibers that appear rare in a study may become several thousand pounds when scaled up to global consumption. And, yet, it is the context of the crime and those involved that further sets any meaningfulness the evidence may have.
A further complication the problem of what could be called forensic amortization. Fibers, once manufactured, do not behave 6according to a fixed or predictable lifespan; they are fungible.7 A coat might be discarded after a year or worn for twenty. Upholstery may be in use for decades. Clothing circulates through secondhand markets, donation centers, and storage bins. The total amount for any given fiber group, however, is always lessening and eventually asymptotic8 to zero. This variability obliterates any straightforward relationship between a fiber’s physical state and its date or source of origin.
So, even when a fiber appears to be rare or distinctive under laboratory conditions, there is no way, short of proprietary manufacturing records or controlled supply chains9, to estimate how often it might be encountered in the population. And even with a meaningful population frame, there is no denominator. Which means there can be no meaningful probability estimate, and no likelihood ratio.
Reporting Within Limits
This leads to a critical shift in how fiber evidence (all evidence, really) should be reported and discussed in court. Instead of framing findings as matches or exclusions or relying on vague hedges like “consistent with,” forensic analysts should describe their observations as clearly and neutrally as possible:
What properties were compared?
What methods were used?
Were any differences observed?
If no differences are found, the statement should anchor the evidence within a manufacturing class. This places them within an industrial and material logic, not an evidentiary fiction. A full statement might look like this:
The questioned fiber and the known fiber share the same morphology, including cross-sectional shape and surface characteristics, as well as indistinguishable optical and chemical properties as determined by [insert methods used]. Based on these shared characteristics, the fibers are assessed to come from the same manufacturing class. This classification does not imply a specific source or object, only that the fibers are not distinguishable under the conditions of analysis.
This phrasing avoids speculation, communicates the technical result, and opens the door for contextual interpretation, where the weight of the evidence belongs. It also guards against the temptation to let evidentiary language drift into certainty. As Cole and Sola argue, many forensic tools, like fingerprint algorithms and statistical software, have failed to gain traction not just because of philosophical resistance or legal inertia, but because the tools themselves couldn’t be easily embedded. They lacked usability, transparency, and infrastructure.
In the case of fibers, there is no such device even on the horizon. And that’s not a failure of imagination, it’s a reflection of the material conditions: The fibers themselves resist clean categorization, they multiply in uncontrolled ways, they enter the world without registration and move through it without audit.10 Any model that ignores that is not a model of the evidence, it’s a model of wishful thinking.
Empiricism Without Overreach
The best forensic reporting does not claim more than the evidence can support. It doesn’t mistake analytical indistinguishability for source attribution. It doesn’t suggest that a fiber "came from" a particular object when all that can be said is that it cannot be told apart from it in specific, testable ways. What Coyle reminds us is that this restraint is not a weakness: It’s a strength. It reflects scientific maturity, not caution. It communicates clearly to courts and juries what the examiner saw, what they measured, and what those findings mean within the framework of manufacturing, use, and the circumstances of the scene, not within a fantasy of evidentiary purity.
To say a fiber has the same morphological, optical, and chemical properties as another is to say something valuable. To claim more would be to pretend the world is simpler than it is. And forensic science*, if it is to be credible, cannot afford to do that anymore.
Ontology is the study of what exists and how we can group the things that exist into basic types. It looks for the features that different kinds of things share and asks how they can be classified. One key idea is the difference between particulars and universals. Particulars are one-of-a-kind things, like me (thank goodness). Universals are things that can show up in many places at once, like the color red or the idea of "fuzzy." Another important distinction is between things that exist in space and time, like a car, and things that don't, like the number 5; “5” doesn’t take up space or change over time, but we still treat it as real in some sense. Philosophers build systems of categories, like substance, property, relation, event, and state of affairs, to try to map out everything that exists. These systems aim to create a full picture of reality by organizing all the different kinds of things into a shared framework. At least, they do in theory.
Largely because DNA, as a molecule, is pretty straightforward and, until you start multiplying what are treated as independent alleles, it’s still just class evidence.
It’s not “Space is big. Really big. You just won't believe how vastly, hugely, mind-bogglingly big it is,” big, but, yeah, pretty big.
I’ve used it, you’ve used it, we’ve all used it. I don’t like it but, guess what? We don’t have numbers, so we have to use flabby words (see the discussion at the end about hairs).
But there are such things as “dead fibers,” those that are no longer in production, a kind of manufacturing death. A good example of this was a case I had in 1999, when police in San Diego responded to a refrigerator dumped behind a hardware store. The refrigerator had been opened and a body was inside. The door to the refrigerator had been secured with a rope to keep it closed. The police looked over the refrigerator and guess what they found? A piece of cardboard from the box the fridge was shipped in with the delivery address on it. I can hear the officers now. “Hey, Marty. Why don’t we start with this guy?” They get to his apartment and start questioning him. He says his roommate left a few days ago. One of the officers noticed four round impressions in the living room carpet that outlined a square; he took photos and measurements. During the interview, a woman entered the apartment. When questioned, she said she was the new roommate coming to drop off her first month’s rent to the man being interviewed. One officer asks to speak with her privately. She tells him she found out about the room a day or so ago. Did she see anything odd or suspicious on that first visit? Well, she said, there was a refrigerator in the living room on her first visit but on her second visit it was gone. When she asked the guy where it went, he told her ‘his old roommate was using it.’ (I know, right?) It turns out the guy killed his roommate because she annoyed him. He murdered her and stuffed her in a refrigerator that was tied shut with rope and the rope became a key piece of evidence. Some rope manufacturers put tracers, colored fibers or other identifiers worked into the ropes, to identify the product as theirs, largely for avoiding counterfeits and use in lawsuits. The rope from the refrigerator had a specific structure as well as a tracer identifying the manufacturer; the tracer was a thin ribbon of clear polyester with the company’s name on it. When I contacted the company, they indicated that the tracer had not been used since 1988; this case was in 1999, remember. The amount of that specific rope generally available was reduced each year, providing a deadline for the creation of that product. You couldn’t go down to the hardware store and buy this rope. Rope of the same manufacture with the same tracer was found on a furniture dolly in the possession of the victim’s roommate. For more like this, see Houck, M.M. ed., 2001. Mute witnesses: trace evidence analysis. Academic Press. and Houck, M.M., 2003. Trace evidence analysis: More cases in forensic microscopy and mute witnesses. Elsevier.
(Of a product or commodity) replaceable by another identical item; mutually interchangeable. One of my favorite words, right up there with defenestration.
Another favorite word. Go look it up. Hey! Make it fun. Use a physical dictionary. I’ll wait.
To be fair, I’ve done my fair share of this legwork. It is not easy, however, as anyone who’s tried it knows. Especially with a global economy, it becomes very difficult to track things. For example, here’s a quote I’ve used before that is an excellent example of the whack-doodle nature of the world textile economy:
Say we get an order from a European retailer to produce 10,000 garments. It's not a simple matter of our Korean office sourcing Korean products or our Indonesian office sourcing Indonesian products. This customer, we might decide to buy yarn from a Korean producer but have it woven and dyed in Taiwan. So we picked the yarn and ship it to Taiwan. The Japanese have the best zippers and buttons, but they manufacture them mostly in China. Okay, so we go to YKK, a big Japanese zipper manufacturer, and we order the right zippers from their Chinese plants. Then we determine that, because of quotas and labor conditions, the best place to make the garments is Thailand. So we ship everything there. And because the customer needs quick delivery, we may divide the order across five factories in Thailand... Five weeks after we have received the order, 10,000 garments arrive on the shelves in Europe…
Not a bad life, come to think of it.