Patterns, Not Categories

From below, a kettle of vultures over a Costa Rican valley reads as a single event: a dozen or more large, dark birds turning slowly on a rising column of warm air, wings held in a shallow V, heads bare and small against the body, none of them beating a wing for minutes at a time. Look longer and the kettle resolves into at least two species, turkey vultures and black vultures, often with a scatter of individuals riding higher on the same thermal that could, on a good day, include a king vulture drifting past on its far larger wings. From the ground, at that distance, they are doing the same thing in the same way for the same evident reason, and it would take a decent pair of binoculars and some practice to say with confidence which black shape belonged to which species, let alone to notice that the turkey vulture tends to rock and tip on the thermal in a way the flatter-winged black vulture does not, or that it is usually the turkey vulture that finds the carcass first and the black vultures that arrive afterward, drawn less by their own senses than by the sight of another bird already descending. The sky was offering, without any apparent effort, a single coherent category: vulture.

It is worth taking that category at face value for a moment, because doing so quickly generates a specific, testable expectation. If turkey vultures and black vultures look and behave this similarly because they are close relatives, sharing a scavenging body plan the way siblings share a family nose, then a family tree built from their DNA should place them near each other and, ideally, closer to each other than either is to anything that merely resembles a vulture without being one. This is a perfectly reasonable expectation, and it is also, as it happens, only partly correct, in a way that turns out to say more about the difference between recognizing a pattern and assigning a category than almost any other case in ornithology.

The phenomenon behind that expectation has a name and a long history outside birds entirely. Unrelated lineages arriving independently at the same solution to the same ecological problem is common enough across the tree of life that biologists rarely treat any single case as remarkable on its own. Ichthyosaurs, sharks, and dolphins (a reptile, a fish, and a mammal, separated from one another by hundreds of millions of years of independent history) converged on nearly the same streamlined, finned, torpedo-shaped body because water imposes the same hydrodynamic penalties on any fast-swimming predator regardless of what kind of animal is doing the swimming. The extinct Tasmanian thylacine, a marsupial, converged so closely on the skull shape and general build of a placental wolf that early taxonomists occasionally mistook the resemblance for kinship, before pouched and placental mammals were understood to have been evolving on separate continents since the age of dinosaurs. What makes convergence this common is not coincidence piling up case after case; it is that a comparatively small number of physical and ecological problems (moving efficiently through water, catching fast prey on open ground, finding food that cannot be defended) recur constantly across an enormous diversity of lineages, and a comparatively small number of workable solutions to each of those problems tends to exist. Vultures turn out to be a textbook instance of a very general rule, not an exception to it.

Classification, in its modern form, makes a specific claim, not a vague gesture toward similarity. A shared name at any rank above the individual is supposed to indicate shared descent: two organisms placed in the same family are being asserted, by that placement, to share a more recent common ancestor with each other than either shares with anything outside the family. This is what makes taxonomy more than a filing system. It is also what makes taxonomy falsifiable, in principle, by any new evidence bearing on ancestry, which is exactly what happened to the vultures. Molecular studies sequencing the mitochondrial cytochrome b gene across both New World and Old World vulture-like birds found that the resemblance connecting them is not a family resemblance at all. Carrion-feeding, framed as a general strategy along with the particular suite of traits that tends to accompany it (a bare head resistant to bacterial fouling, a hooked bill suited to tearing rather than killing, broad wings built for cheap, sustained soaring over large foraging ranges) had evolved independently at least three separate times among diurnal birds of prey: once in the New World vultures, and twice more within separate Old World lineages that are not each other's closest relatives either. Three unrelated solutions to the same problem, each one converging, from a different starting point in the tree of life, on nearly the same visible answer.

The story does not stop there, and its second half is the more instructive one. Once it became clear that New World vultures were not close kin of the Old World vultures they resemble, an alternative hypothesis followed close behind: perhaps their true relatives were storks instead. New World vultures share a genuine set of traits with storks that Old World vultures do not: a similar skull structure, comparable arrangements of certain muscles, and, most strikingly, a shared method of thermoregulation in which both groups cool themselves by defecating onto their own legs and letting the evaporation carry heat away. A large-scale DNA hybridization study in the late 1980s took these shared traits seriously enough to reclassify New World vultures as a subfamily of storks, and a subsequent mitochondrial DNA study appeared, at first, to lend the idea further support. It took another decade and a half of more careful, multi-locus genomic work to unwind the stork hypothesis and establish that New World vultures form their own distinct, ancient lineage, related to birds of prey more broadly but not tucked inside either the storks or the Old World vultures, a placement current genomic evidence continues to refine rather than one settled once and permanently. Two separate, serious scientific hypotheses, each generated by genuinely shared traits, and each one wrong about what those shared traits actually meant.

This is worth sitting with, because it cuts against a tidy version of the argument this essay is about to make. The naive lesson from the first half of the vulture story looks like a straightforward vindication of pattern over category: three unrelated groups converged on the same functional design, and no amount of taxonomic labeling would have predicted or explained that design nearly as well as recognizing the recurring problem (locating and processing carrion efficiently, at low energetic cost, over large areas) that the design solves. But the second half of the story shows the same recognition machinery misfiring in the opposite direction: traits that were treated as evidence of shared ancestry with storks turned out to be either their own separate case of convergence or an artifact of the particular molecular method used to detect them. Pattern-recognition explained the vulture body plan beautifully and mangled the vulture family tree in the same breath, using structurally the same cognitive move both times. Whatever this essay ultimately argues about the relative power of patterns and categories, it cannot argue that patterns are simply more trustworthy. It has to argue for something more specific: that patterns and categories are answers to different questions, and that most of the trouble in this story came from letting an answer to one question stand in for an answer to the other.

Taxonomy answers a question about history: what shares descent with what. It does this with enormous rigor once enough independent genetic evidence is in hand, and an earlier chapter in this collection already made the case that this kind of disciplined division of nature into named, bounded units is one of biology's genuine achievements, not an obstacle to understanding it. That achievement is not diminished by anything in the vulture story; if anything, the vulture story is a demonstration of taxonomy working exactly as it should, correcting itself twice over as better evidence arrived, rather than settling comfortably on whichever hypothesis first felt intuitive. A classification system that could not be revised by new genetic evidence would not be a stronger system for its stubbornness; it would simply be a pattern wearing the authority of a category without having earned it. The category's real value lies precisely in the fact that it is answerable to a kind of evidence (shared, inherited genetic sequence) that a resemblance in silhouette or behavior can never supply on its own, however striking that resemblance is to look at from the ground.

But descent is not the only question a naturalist standing under a kettle of vultures actually wants answered. Knowing that Cathartidae and the Old World Gyps vultures are only distantly related tells a visitor nothing about why both groups share a bald head, why both soar rather than flap, why both tend to arrive at a carcass in loose aggregations rather than as solitary hunters, or why both are exceptionally vulnerable to poisoning when the carrion they depend on has been contaminated, a vulnerability that has driven catastrophic population collapses on more than one continent in recent decades. Those questions are answered by the recurring functional pattern (obligate scavenger, dependent on an unpredictable and ephemeral resource too widely scattered to defend by force, unable to profitably kill its own food), not by the family tree. The pattern is what explains the design. The category is what explains where the design's carriers came from. Both are true and necessary, and neither substitutes for the other.

A cleaner demonstration of the pattern's explanatory reach, uncomplicated by any taxonomic confusion, comes from a very different set of animals on a very different set of islands. On each of the four largest islands of the Greater Antilles, Anolis lizards have radiated into a strikingly similar set of habitat specialists (a species with short legs and a long tail that lives out on thin twigs, a species with long legs built for sprinting across open ground, a species with expanded toepads that clings high in the tree crown, and several others in between), a set of forms researchers call ecomorphs. What makes the case so useful here is not simply that the same forms recur, but that they recur independently: careful reconstruction of anole family trees shows that the twig specialists on different islands are not one another's closest relatives, and neither are the trunk-crown specialists or any of the others. Each island generated its own version of nearly the same lineup, starting from a different founding population and evolving toward the same set of solutions by a different genealogical route. Once a lizard's ecomorph is known, an enormous amount about its biology follows immediately and reliably (its typical perch height and diameter, its limb proportions, its sprinting versus clinging performance, even aspects of its social display), and this holds regardless of which of the roughly four hundred Anolis species is under discussion or which island it comes from. Knowing the species name alone, without the ecomorph, predicts almost none of this. The pattern, once again, is doing the explanatory work that the category cannot do on its own.

The same structure appears in Costa Rica's forests without needing an island archipelago to reveal it, in the guise of what ecologists call a guild: a group of species that exploit the same kind of resource in the same general way, regardless of how distantly related they are to one another. A nectarivore guild in a single patch of forest can include hummingbirds, which are birds; certain bats that hover at flowers at night; and sphinx moths, which are insects: three lineages separated by hundreds of millions of years of independent evolution, converging on long, narrow mouthparts or tongues, hovering flight, and a diet built around the same energetically concentrated resource. A visitor who understands the guild as a pattern (an organism built to extract nectar efficiently from a flower it cannot afford to damage, competing against every other nectar-feeder in the same forest regardless of what kingdom, class, or order they happen to belong to) understands something that transfers to the next forest, the next continent, and the next guild encountered there. A visitor who has only memorized that hummingbirds drink nectar has learned a fact about one lineage that tells them nothing about the bat or the moth working the same flowers an hour later.

It is worth asking, briefly, why convergence of this kind should be common at all rather than a curiosity confined to a handful of textbook cases. Part of the answer lies outside natural history entirely, in research on the predators whose hunting behavior helps generate many of these patterns in the first place. Controlled experiments with blue jays searching for camouflaged, computer-generated moths found that the birds' accuracy at spotting a given cryptic pattern rose with recent, repeated exposure to that pattern and fell when the pattern was rare or unfamiliar: evidence for what researchers call a search image, a perceptual template a predator builds up through experience that sharpens detection of familiar prey at some cost to noticing unfamiliar ones, a concept introduced earlier in this collection to describe how trained attention operates in general. Over successive simulated generations in these experiments, the prey population evolved toward greater crypticity and greater variability, precisely because a predator's search image rewards rarity and penalizes uniformity. A search image is a form of pattern recognition operating inside the predator, and it generates, as a side effect, selection pressure on the prey to defeat that recognition by resembling something the predator is not currently primed to notice: a background object, a different and less familiar prey type, or another member of a large and visually uniform population. Wherever this dynamic recurs across the natural world, in whatever lineage happens to be under pressure from a visually hunting predator, something like convergence on effective camouflage should be expected to recur along with it, for the same reason vulture-like scavenging recurs wherever a large-bodied carcass resource goes underexploited: not because the lineages involved are related, but because the underlying problem is the same problem, encountered again and again by organisms that have no way of consulting each other's family trees.

Guilds, ecomorphs, and search-image-driven crypsis are, in this sense, three faces of the same underlying claim rather than three separate observations loosely gathered under one essay. In each case, a selective pressure recurs across independent lineages (the geometry of moving fast through water, the physics of extracting nectar without damaging a flower, the cognitive habits of a visually hunting predator), and in each case the recurring pressure produces a recurring solution regardless of which branch of the tree of life happens to be facing it at the time. A category, drawn along the branches of that tree, will never by itself reveal this kind of repetition, because repetition across branches is exactly what a category, defined by descent within a single branch, is not built to track. Seeing the repetition requires a different kind of attention, oriented toward problems and solutions rather than toward ancestors and descendants, which is very close to what an earlier chapter in this collection meant in arguing that a forest's relationships carry more explanatory weight than its species list, extended here from relationships within a single system to solutions recurring across many unrelated ones.

This gives the essay's central claim a more precise shape than pattern good, category bad would suggest. A pattern, in the sense meant here, is a recurring configuration of form, behavior, or relationship that appears across independent lineages because it solves a recurring ecological problem, and its explanatory power comes specifically from that repetition: seeing the same solution arrived at more than once, by different routes, is what licenses the inference that the solution is doing real work rather than reflecting some arbitrary accident of one particular lineage's history. A category, in the taxonomic sense, is a claim about which lineages share which history, and its power comes from an entirely different source: the accumulated, cross-checked evidence connecting organisms through actual descent. Confusing the two produces exactly the vulture-and-stork problem, in either direction: mistaking a shared solution for shared history, or, just as easily, dismissing a genuine ancestral signal because it happens to be carried by traits that look, superficially, like they could just as well be convergent. Telling the two apart requires the same discipline in the field that it required in the vulture case in the laboratory: treating an observed similarity as a hypothesis to be tested against independent evidence, rather than as a conclusion the similarity has already supplied on its own.

That discipline matters practically, and not only to specialists revising a phylogeny. It was argued elsewhere in this collection that an interpreter's task is less about delivering facts than about installing, in someone else's attention, the frame needed to recognize a pattern the next time it appears, and that this kind of installed frame is unusually resistant to later correction because it operates as a lens rather than as a checkable claim. A visitor taught to recognize the pattern obligate scavenger (unable to defend a kill by force, dependent on an unpredictable resource, therefore social at carcasses and vulnerable to whatever poisons accumulate in them) carries something that will keep working correctly on a hyena in East Africa, a burying beetle in a temperate forest, or a stork-relative none of them have met, because the frame tracks the recurring problem rather than any particular lineage's membership card. A visitor taught only that turkey vultures and black vultures are the two vulture species found locally carries a fact bounded to this valley and these two names, true and useless anywhere else. But the vulture-and-stork history is also a warning about how easily the first kind of teaching can overreach: an interpreter confident that a resemblance reveals a pattern, without the patience to ask whether it might instead be coincidence or a trap laid by convergence in the opposite direction, is exactly as capable of installing a wrong frame as a right one, and the wrong frame will feel, from the inside, exactly as coherent and confirmed as the correct one would have.

This gives the interpreter a narrower, more specific obligation than simply favoring patterns over categories, and it is worth stating plainly rather than leaving implicit. A pattern earns the confidence it is taught with only to the degree that its recurrence has actually been checked against independent cases: the way the vulture pattern was checked against genetic sequence, or the ecomorph pattern was checked against independently reconstructed family trees on four separate islands. A resemblance noticed once, in one place, on one walk, is a hypothesis about a pattern, not yet a pattern in the sense this essay has been defending, and teaching it with the same confidence as the checked cases collapses exactly the distinction the vulture-and-stork history exists to preserve. The interpreter's task, on this account, is not to prefer the more elegant explanation on sight but to know, and to signal honestly, which side of that checking process a given claim currently sits on.

I don't know, standing under a kettle of vultures with no laboratory anywhere nearby, how a field observer is supposed to tell, in the moment, whether a resemblance in front of them is the kind that held up under a cytochrome b sequence or the kind that didn't. The genetic test that eventually settled the vulture question is not a tool available on a trail, and most of the patterns worth noticing in a forest will never receive that kind of adjudication at all. A naturalist has cruder instruments in its place (independent recurrence across more than one unrelated lineage, a plausible account of what selective problem the pattern would be solving, and agreement with what is already reliably known about the groups involved), and each of these can be applied on a trail with nothing more than attention, comparison, and patience. But none of them carries the finality of a laboratory result, and the vulture case is a reminder that careful, well-informed observers, working from real rather than imagined evidence, spent the better part of a century assembling confident and mutually incompatible answers before a better instrument arrived to adjudicate between them.

What, short of a sequencer, is the naturalist’s actual method for separating a pattern that genuinely explains from one that merely appears to? And if there is no decisive method, does every relational frame amount to a provisional interpretation that the interpreter and the recipient can’t fully verify at the time?

Recommended scientific references

Seibold, I., & Helbig, A. J. (1995). Phylogeny of Old and New World vultures (Aves: Accipitridae and Cathartidae) inferred from nucleotide sequences of the mitochondrial cytochrome b gene. Zeitschrift für Naturforschung C, 50(11–12), 868–882.
Avise, J. C., Nelson, W. S., & Sibley, C. G. (1994). DNA sequence support for a close phylogenetic relationship between some storks and New World vultures. Proceedings of the National Academy of Sciences, 91(11), 5173–5177.
Losos, J. B., Jackman, T. R., Larson, A., de Queiroz, K., & Rodríguez-Schettino, L. (1998). Contingency and determinism in replicated adaptive radiations of island lizards. Science, 279(5359), 2115–2118.
Bond, A. B., & Kamil, A. C. (2002). Visual predators select for crypticity and polymorphism in virtual prey. Nature, 415(6872), 609–613.
Pietrewicz, A. T., & Kamil, A. C. (1979). Search image formation in the blue jay (Cyanocitta cristata). Science, 204(4399), 1332–1333.

Author's note

The scientific content of this essay, including the molecular phylogeny separating New World and Old World vultures, the earlier stork hypothesis and its overturning, the Anolis ecomorph research, and the search-image experiments in blue jays, reflects established, cited findings, including the genuinely unresolved edges of the vulture phylogeny, which this essay has tried not to flatten into more certainty than current evidence supports.

What is not an established finding, and belongs to The Neotropic Philosophy of Interpretation as this collection's own proposal, is the essay's central claim: that recurring ecological patterns generally carry more explanatory power for a naturalist's purposes than taxonomic categories do, and that responsible interpretation therefore means teaching patterns rather than classifications wherever the two can be told apart. The vulture-and-stork history was chosen deliberately because it also argues against a cruder version of this same claim: it shows pattern-recognition failing as readily as succeeding, and offers no field-ready method for telling the two cases apart in advance. The essay's proposal is therefore held, by its own logic, with the same caution it asks an interpreter to extend to any other pattern: as the more useful frame available at present, not as a claim beyond revision.

Patterns, Not Categories

​

Recommended scientific references

Suggested further reading

Author's note