Humans privilege the visual, says Wired writer Brooke Jarvis. All primates, including humans (we are animals, specifically primates; get over it), are visually oriented due to evolutionary pressures. Our eyes, which are an external expression of our brain, rely heavily on visual cues for fundamental activities like finding food, mates, and shelter. More complex behaviors, like caring for children and operating within society, also require the processing of visual cues. About half of the human brain is devoted or related to processing visual information. So it makes sense that when we try to make sense of the natural world, we use our eyes.
Morphology
Take, for instance, taxonomy. Taxonomy is the science of categorization or classification, according to Wikipedia; it is the hierarchical arrangement of things based on defined traits. Taxonomy is usually associated with science and biology but anything can be classified with a taxonomic set of rules. It might sound odd but there are multiple ways to arrange living (and non-living) things: There is no single set of rules to arrange the units of taxonomy, that is, taxa (singular, taxon). In biological taxonomy, there are taxonomic systems based on ancestry (phylogenetics), overall similarity (phenetics), and morphological traits (Linnean, named after the originator of scientific taxonomy, Carl Linnaeus). Without going down the Leporidae hole of taxonomy (which is very deep and complex, trust me), the Linnean system is a rank-based method, moving from the largest groups (Kingdoms) to the smallest (species). Linnaeus’ system used binomial nomenclature, or a “two-name system.” Each animal or plant falls into a genus (the group above species) and a species. For example, humans are Homo sapiens (genus: homo, species: sapiens), eastern cottontail rabbits are Sylvilagus floridanus, giant pandas are Ailuropoda melanoleuca, and so on. Whether an organism falls into a particular species (or genus or so on up the taxonomic ladder) depends on the traits it does and does not have. The decision about which traits to use comes from taxonomic keys, a set of questions (typically yes/no or by elimination) that lead you to what taxon the organism you’re looking at belongs to (at least arguably).
Still Waiting on the Forensic Stuff…
We’re getting there and, wow, will you be surprised. Here’s the thing about species: We don’t really know how many there are. It’s estimated that we’ve only found between one-fifth and one-thousandth of the current species (ignoring how many we destroy each day). Finding a new species is not uncommon--happens all the time. Another fun fact about species: We don’t really know what makes a species. Wait, what? That’s really confusing. Yes. Yes, it is. The traditional definition of a species is organisms that are or are potentially capable of interbreeding. But, despite the fact that we have named grizzly bears (Ursus arctos horribilis) and polar bears (Ursus maritimus) as separate species hasn’t stopped them from getting busy (finding each other due to shrinking environmental territories due to global warming) and making the too-cute-sounding-for-such-a-massive-animal-that-can-kill-you pizzly bears. This begs the question of what a species is and how it is defined.
In biology, this has come to a heated argument amongst entomologists who study braconid wasps. Braconids may have tens of thousands of species but most are largely unknown. Braconids are parasitic wasps with morally questionable reproductive habits (think Alien but worse; no, really) which means there’s a huge diversity based on all the other organisms they use to reproduce. It gets very, very specific. Jarvis, who was quoted at the beginning of this post, wrote an interesting article about the struggle to identify braconid species. Most entomologists have used morphology, that is, visual cues, listed in taxonomic keys to “key out” species. What’s the big deal?
Longevity (of Methods)
With the advent of cheap and fast DNA sequencing, identification of genetically distinct organisms became easy. Called DNA barcoding, the method uses the information of one or more gene regions to identify a species. This method has upset the apple cart of morphological taxonomy. One entomologist decided to revisit some of his old taxonomic work on braconids and
The results shocked him. The work hadn’t just been slow; much of it seemed to be wrong. According to the genetics, some of the animals he’d diagnosed as one species were best understood as four or five; others, which he’d named as multiple species, were only one. It seemed that as much as half of his work was, at best, misleading. “The morphological work I was doing was just garbage,” [the researcher] said. “I thought, my God! I’ve wasted 20 years of my life, or at least my professional life.”
This is not uncommon now amongst biologists. What was thought to be one giraffe species is actually four, orcas are at least three, and one well-studied butterfly species was actually ten different ones, genetically speaking. The taxonomic keys, the keys to the taxonomic kingdom, if you will (OK, sorry), were a set of methods biased by our visual orientation and us thinking that was all there was.
IN HIS OFFICE in Colorado, [the researcher] showed me old monographs and morphological keys meant to guide people in identifying various parasitoid wasps. He lamented how “useless” they were. Some of the written descriptions seemed like they wouldn’t be much easier to follow than a genetic code; many specimens didn’t key out to a species, or keyed out to the wrong one, because the keys included only the small subset of species that had been discovered at the time and no information on the much wider world that really existed.
As quoted by Jarvis in the article, Carol Kaesuk Yoon, in her book Naming Nature, says, “There is nothing harder to see than one’s own frame of reference.”
Incept Dates: Forensic Hair Comparisons
Now we get to forensic science*. Microscopical hair comparisons (MHC) have a checkered past as a forensic method. The whole hair thing could be its own book (I know, I know). The point to be made here is that MHC used morphological taxonomic methods to compare unknown hairs to known samples. In doing so, MCH used taxonomic keys (explicitly or implicitly) to identify animal hairs and also used “taxonomic-light” methods to categorize hairs by race, body area, and potential source (that is, from whom could the hair have originated), among other things.
The MHC frame of reference, to borrow Yoon’s phrase, was that microscopic morphology could yield reliable answers about these questions. “Accidents apart, it is difficult to see how data could be searched for without a theory of some sort.” So, those morphological traits were used to make inferences about questioned hairs and known hair samples. And that was the MHC orthodoxy until--again--DNA analysis appeared on the scene, in this case, mitochondrial DNA (mtDNA). A paper that I co-authored was the first to compare MHC results with mtDNA results. It showed that MHC was strong in exclusions (most forensic examinations are; a possible future post to explain that) and relatively good at inclusions:
Overall, for this limited sample (170 hairs total) from FBI casework, morphology-based MHC had a precision of 0.88 and an accuracy of 0.91 backed by mtDNA analysis. Sound good? Well, here’s the rub.
Questions
Translating what one sees under the microscope to reality is very difficult. Unlike the entomologists discussed earlier, MHC doesn’t purport to identify an entire species; it goes way, way beyond (or below) that and talks about the possibility that the questioned hairs found at a crime scene came from a specific individual. This process runs somewhat counter to how much, but not all, of science works. Most science uses individual observations to make more general statements about the world. “Thus by means of the similarity of properties of diverse individuals a bridge is made to the generality of classes.” Thus, the more a scientific theory can explain, the better.
Another factor in what we’re heading towards is that in forensic science*, we never know the real answer. Even (or especially) if a person of interest confesses, that is no guarantee of what happened. Forensic science* is blind to the actual set of circumstances leading to the samples on the lab bench for examination; the details can seem irreducible but, honestly, the truth is not known. My teacher, mentor, and friend Jay Siegel once told me, “Justice is to law as health is to medicine.” Subject states rendered by imprecise methods.
The morphological identification of a species and its refutation by genetics is one hot topic but both are backed by objective methods. MHC are subjective evaluations of objective criteria. What’s a “few” cortical fusi? What is “sparse” or “dense” pigmentation? How “discontinuous” is “discontinuous medullation”? More to the point, what does it mean? How do you convert the subjective evaluation of a sample of hairs to a population that is unknown and unknowable? How many individuals could have hairs that display a similar enough range of morphological traits when compared to the range in the known sample? Who knows? “Every individual is of course unique to some extent, but it is one of the scientific axioms that there are no absolutely unique individuals, no actual material objects which are unlike all others in every respect. If there are individuals, then there are classes, and once given classes they can be related systematically. Thus the scientist can build his data into systems easier than he could completely describe any datum.” MHC seems to have worked from individuals into classes and then forgot about the build the data into systems part.
How many people could be the source of a questioned hair? No one knows, that’s the point. Using words like “rare,” “common,” “few,” “many,” all imply some metric or statistic that does not exist. We lack the precision of statistics to make measured statements of relatedness or similarity, so we use colloquial terms, everyday words, to describe and evaluate the morphology of evidence. These words all have different meanings, depending on how they are used and who is hearing them.
And even if we use technical terms to run through taxonomic keys to place a hair in a class (Fred’s possible head hair, for example), morphology can be trumped by genetics. Although the area of human perception of probabilistic words continues to be an area of somewhat active research and articles have been written about how chemists should word reports or how the strength of statements is perceived, most forensic scientists* and their agencies haven’t really addressed the issue. If “could have come from” or “is consistent with” or “unable to exclude” all end up being vacuous statements with no statistical underpinning, then many forensic sciences* are screwed. “Could have come from” implies the counterfactual “Could not have come from,” so what’s being said, really?
Forensic science* needs to rethink how it approaches evidence interpretation (and I do not mean Bayesean or likelihood ratios: We don’t have a base rate and starting at 50% is facile). Chemometrics offers some guidance, along with network analysis and (I can’t believe I’m saying this) maybe AI. A better understanding of the production (biological or manufactured, along with as-used) of what becomes evidence is required to have the vision of producing accurate methods of interpretation. “There are no sacrosanct systems within science; any formulation may have to be abandoned.” However, we’ve never been very good at getting rid of our dead weight: We do it only when forced and then in a ham-fisted way that doesn’t solve the problem.
Does that sound like a culture of science?