* This is the fourth installment of “Problematica.” It is written by Max Dresow…
Consider, for a moment, the horseshoe crab. The first thing to know is that it isn’t a crab. It isn’t even a crustacean. Horseshoe crabs belong to a group of arthropods that also includes ticks, mites, spiders, and harvestmen. Their closest relatives are the extinct “sea scorpions,” which were neither true scorpions nor, for the most part, denizens of the sea. But among the living fauna, horseshoe crabs are most closely related to arachnids—the group that includes the true scorpions. So tangled are the taxonomic webs we weave.
Horseshoe crabs are biological anachronisms. Richard Fortey calls them “survivors” because they preserve an ancient morphology more or less intact. The morphology is that of a miniature tank. The animal’s whole body is covered by a hard carapace that resembles nothing so well as a Stahlhelm, or old German infantry helmet. Lurching across beaches they present an unsettling aspect. Heavy and clumsy, they seem to have crawled out of the belly of time itself. Never has the designation “living fossil” seemed so apropos.
Atlantic horseshoe crabs cavorting on a beach
But here’s the thing. Horseshoe crabs are derived. They are not real throwbacks except in the sense of retaining the outline of an ancient morphology. This may qualify them as “living fossils,” but it doesn’t make them ancient. It doesn’t even necessarily make them “primitive.” The sense in which they are primitive is the same sense in which they are anachronisms—that is, living horseshoe crabs share a large number of features with earlier members of their group. We know this because the group has a fossil record dating all the way to the Ordovician, some 445 million years ago. And as early as the Triassic (ca. 250 ma) horseshoe crabs resembling living forms were scuttling across the ocean floor. Ergo, “living fossils.” But the designation is only appropriate (if it is) because fossil evidence suggests that a high degree of morphological stability has obtained in this lineage. It would be inappropriate to call them primitive on the basis of phylogenetic position alone, even if extant horseshoe crabs occupy a basal position within the chelicerates.
Why, then, do some phylogenetic studies refer to Xiophosura (the clade including all horseshoe crabs, living and extinct) as “an ancestral taxon in chelicerates as well as arthropods [more generally]” (Baek et al. 2014)? Just because Xiophosura occupies a basal position within Chelicerata does not mean that the taxon itself is ancestral to a more inclusive group. To be basally situated is not the same thing as to be an ancestor, which, if it means anything, must mean to be connected by direct descent to a more derived taxon or group of taxa. So, again, why refer to the clade as ancestral? Perhaps this is just an instance of benign linguistic slippage. But perhaps it is more than this: a linguistic betrayal of attitudes about basal taxa lacking a foundation in evolutionary theory.
* * *
It is issues like this that animate Ronald Jenner’s new book, Ancestors in Evolutionary Biology. Or, to be more precise, it is these issues that animate a small portion of his book, which follows a lengthy history of traditional phylogenetic practice. Jenner is a researcher at the Natural History Museum in London interested in the evolution of animal body plans. He did his PhD under the American paleontologist Frederick Schram, and a postdoc under Max Telford (a biologist who shares a name with an ultra-distance runner). By trade he is a practitioner of phylogenetic analysis whose recent work has concerned the evolution of venomous invertebrates. But by temperament he is a critic of scientific practice, prone to outbursts of common sense in the best tradition of his British colleagues. It was precisely these outbursts that led me to track down a copy of his new book, and I am happy to say that I was not disappointed.
This is not a review of Ancestors in Evolutionary Biology. It can’t be—I haven’t finished it yet. Rather, it is a review of a single chapter, which examines some bad habits in contemporary systematics. This is Chapter 10 if you’re following along at home, called “Phylogenetic Faux Pas and Narrative Ghosts in the Cladistic Machine.”
The chapter begins at the end of history, so to speak. For about a hundred years following the publication of the Origin, phylogenetic reconstruction was dominated by attempts to explain the origin and evolution of traits by linking them to precursors in hypothetical ancestors. (Jenner’s name for this is “narrative phylogenetics.”) However, beginning in the 1970s, this approach began to be supplanted by a new one, variously called “phylogenetic” or “cladistic analysis.” This was mainly a method for inferring systematic relationships using shared derived characters. But it also provided new tools for thinking about ancestors. In cladistic analysis, ancestors are not simply postulated based on their explanatory potential as precursors of living taxa. Instead, they are reconstructed based on the distribution of character states in phylogenetic trees: a procedure that clips the wings of those speculative flights central to narrative phylogenetics.
Jenner is quick to praise these methodological innovations. Yet his account is not about the triumph of ancestral state reconstruction over narrative fancy. It is rather about the persistence of elements of narrative phylogenetics in cladistic practice: “narrative ghosts… in the cladistic machine.” Forget ghosts for now. Instead, notice that it is exactly the machine-like nature of cladistic analysis that makes it such an appealing methodology for biologists. Rather than hinging on intuition or expert judgement (or whatever), cladistic analysis uses standardized procedures for inferring evolutionary relationships between taxa. This makes it more transparent—and in that respect, more “objective”—than traditional phylogenetic methods. Still, cladistic analysis provides ample scope for subjective preferences and even theoretical ideas to leave their marks on phylogenetic trees. (Boo!) In fact, it provides so much scope that researchers in the early 2000s were led to scuttle the entire project of morphological cladistic analysis, at least as an attempt to work out the relationships between major animal groups.
At this point Jenner switches into the first-person register, because as it turns out, he had a role to play in hastening the demise of this project. During his postdoc, he “read hundreds of articles and books to discover to what extent the many thousands of entries in cladistic data matrices were reliable reflections of observed character variation, or just speculations and assumptions” (Jenner 2022, 284). What he found was that there was enough of the latter to deep six the whole enterprise. To begin, many studies failed to include all the relevant characters for testing the hypotheses they set out to test. Jenner mentions a study on Myzostomida (small marine worms) which “confidently concluded in its title that ‘Myzostomida are not annelids’” (285). “Yet, with the exception of a character coding for parapodia, their morphological dataset included not a single character that could indicate the [annelid] affinities of myzostomids.” (Myzostomids are now thought to be annelids.) “Consciously or not, [these authors] had constructed their dataset with the same unequal eye that narrative phylogenetics had used to build scenarios.”
Myzostoma fuscomaculatum, a marine annelid that parasitizes crinoids (“sea lilies”)
There were other problems too. Some researchers scored characters for living taxa “based on assumed ground patterns of primitive character states rather than carefully inferred ancestral states” (285). More often than not, these reflected assumptions about evolutionary transitions “that were unsupported by observations about organisms.” Jenner lists as examples “coding a character that defines the gonocoels of molluscs and arthropods as secondarily reduced coeloms, which reflects the previously common assumption that these taxa evolved from annelid-like ancestors,” and “scoring chordates as having a trimeric body plan, which reflects the previously widespread view that they evolved from ancestors with three sets of coeloms.” At the same time, ancestral assumptions were often baked into the selection of terminal taxa. Consider that to recover the right relationship between taxa A, B, and C, it cannot be the case that either B or C belongs within taxon A. If B or C lies within A, a phylogenetic analysis that treats A, B, and C as terminal taxa will not reveal this. Only an analysis that splits A into constituent taxa can do the trick. Precisely this problem arose in Claus Nielsen’s (2001) analysis of animal phyla, in which Sipuncula was included as an independent phylum when in fact it belongs among the annelids. Since Nielsen did not split up the annelids, he was unable to correctly resolve the phylogenetic placement of Sipuncula; but his decision to separate Sipuncula from the annelids was not unmotivated. Instead, it was motivated by assumptions about ancestors.
I will mention one last problem, which was perhaps the most damaging to the project of morphological cladistic analysis. As Jenner writes,
The vast majority of morphological cladistic analyses were coded as binary absence/presence characters [either a taxon had a character (1) or it lacked it (0)]. Although presence character states were generally carefully delineated, absence states were typically treated as a default that could be scored for any taxon lacking the presence state, irrespective of its morphology. As a result, the zeros scored for absence states in datasets reveal[ed] absolutely nothing about the morphology of the taxa involved. (286)
The practice also led to difficulties when characters reversed to undefined absence states, especially when such reversals led to groupings based on obviously non-homologous characters. Again, Jenner’s example is Nielsen (2001), who included among his “four unambiguous [derived characters] supporting the monophyly of Bilateria… the reversal of a character that scores the presence of an adult brain derived from, or associated with, the larval apical organ” (286). The problem is that “this reversal groups taxa on the basis of clearly nonhomologous nervous system configurations,” so the support it gives to the hypothesis of bilaterian monophyly is spurious.
Jenner concludes this discussion a philosophical note. Traditional narrative phylogenetics was based on the linking of ancestral and descendent forms in a continuous sequence: a practice that supported narrative explanations based on the transformation of particular ancestral structures into derived ones. But this approach cannot work when antecedent states are not carefully specified, or when (in a phylogenetic analysis) they are simply scored as a string of zeros in a character matrix. As Jenner writes, “[Derived character states] that emerge from empirically empty absence states cannot be captured in an evolutionary narrative of cause and effect… [they] can provide evidence for the synchronic explanation of sister group hypotheses between collateral relatives, but they lack the truly diachronic explanatory power of historical hypotheses” (287). Put more simply: if ancestral character states are not robustly reconstructed there can be no question of framing narrative explanations of derived character states, for want of a place to start.
* * *
I have not forgotten the horseshoe crabs with which I began this essay, but before returning to the link between basal position and ancestrality, I have one more stop to make. This is Jenner’s discussion of fossils in phylogenetic analysis (“wringing water from stones,” as he puts it). To get here I have bypassed his brief discussion of molecular phylogenetic analysis. Suffice it to say that incomplete sampling caused serious problems in the early going, providing a smokescreen behind which traditional assumptions about animal relationships could be smuggled in. But as Jenner observes, these practices were to some extent unavoidable in the 1980s and ‘90s, and have been mostly remedied in recent publications.
Jenner begins his discussion of fossils by criticizing some harmless instances of linguistic sloppiness, like calling Anomalocaris a “fossil ancestor,” when in fact it is merely a stem arthropod (Ortega-Hernández et al. 2019). I mention this because it is important to see why calling Anomalocaris a “fossil ancestor” is different from calling Xiophosura (the horseshoe crab clade) an “ancestral taxon.” Xiophosura is an extant clade. It thus belongs to the arthropod crown: to the collection of taxa comprising the most recent common ancestor of living arthropods and all the descendants of this ancestor, living and extinct. Anomalocaris, by contrast, is an extinct group located deep in the arthropod stem: in the collection of all arthropod taxa falling outside the crown-group. While no anomalocarid is a direct ancestor of any crown-group arthropod, anomalocarids do occupy an intermediate position between the last common ancestor of the total arthropod clade and the last common ancestor of the crown-group (see the figure, below). As a consequence, anomalocarids contain some but not all of the features that characterize the crown, and are thus informative about the sequence of events leading to the emergence of the crown-group body plan. Horseshoe crabs, by contrast, are members of the arthropod crown-group, and so are not intermediate between anything. (More simply: if we interpret “fossil ancestor” to mean a member of a clade lying outside the crown, then anomalocarids are “ancestral” arthropods while xiophosurans are not.)
The stem-group, crown-group, and total-group (stem + crown). Anomalocarids occupy a position deep in the arthropod stem-group
In any case, Jenner is not overly concerned with this form of linguistic slippage, at least when it is confined to the pages of scientific journals. More worrisome is the tendency he detects among paleontologists to regard the oldest fossil belonging to a group as a literal ancestor of that group. That is what he says, anyway. His real concern seems to be paleontologists’ tendency to treat ancient fossils as primitive without an accompanying phylogenetic analysis. This is a different sin from regarding these fossils as direct ancestors of a group. Jenner’s point is simply that age is no guarantee of primitiveness even if there is reason to expect these things to be positively correlated.
* * *
The final section concerns what Jenner calls “widely used ancestral shortcuts” in taxonomic language (291). These are strategies used to “evoke an ancestral aura around favorite organisms,” and are generally compressed into adjectives like “ancient,” “early,” “archaic,” “prototypical,” “classic,” “basal,” “early branching,” “generalized,” and lastly (a noun) “living fossil” (292). In Jenner’s pithy synopsis: “These words are ‘FOR SALE’ signs and what they sell is ancestors.” Especially significant is “basal,” which has come gain wide currency as a synonym for “ancestral” or “primitive.” Jenner gives many examples. Hydra is a basal (primitive) metazoan. So are corals, sea anemones, sponges, ctenophores, placozoans, and platyhelminths. Zebrafish are “basal vertebrates.” So are lampreys, cartilaginous fishes like elephant sharks, and amphibians like Xenopus. And so forth.
“That one label can be attached to such diverse casts of taxa suggests it is imprecise and imprecisely used” (292). Indeed, Jenner thinks this linguistic sloppiness is diagnostic of “a widespread and persistent misunderstanding of a crucial topic in evolutionary biology.” This is the relationship between the phylogenetic position of a taxon and the nature of its character states. Biologists tend to presuppose that basal phylogenetic positions are associated with primitive character states. (Philosophers too. According to Jessica Bolker (1995, 454), “more basal lineages within a clade have (by definition) fewer derived characters, and more characters retaining ancestral states that may be generalized to other groups.”) But “[in] a tree of extant taxa, all terminals are the tips of evolving lineages that have evolved for the same amount of time” (293). This means that there is no guarantee that basal taxa will have primitive character states. It is worth quoting Jenner here at some length:
Phylogenetic position just records the pattern of lineage splitting events in a tree. Nodes or terminal taxa can be labeled as more or less basal with respect to a common ancestral node in the tree, but in the absence of character state information about these taxa, this phylogenetic information has no predictive value about whether they are more or less likely to have retained [ancestral] character states. Even when the tree is unbalanced and fully pectinate (comb-like), and character states are known for all taxa except the earliest diverging one, one cannot predict what character state it is likely to have. Early diverging or basal lineages, which branch off from the other lineages closer to the root of a clade, continue to evolve just like the other lineages, and their characters may have diverged just as much or more as those of any of the lineages that separated later in time. (293–294)
To condense all this into a sentence: phylogenetic theory provides no reason to think that an unstudied taxon occupying a basal position in a tree will have primitive character states.*
[* Nor does evolutionary theory. Jenner notes that if the model of punctuated equilibria is right, then taxa are likely to have more primitive character states when there are less nodes located along their lineages. But it seems increasingly unlikely that evolutionary change is cramped into speciation events in the way that punctuated equilibria requires, and if it is not, then “there may [be no] correlation between the number of speciation events along a lineage and the amount of [evolutionary change] that it has accumulated” (294).]
It is important to realize what Jenner is not saying. He is not saying that basal taxa never retain primitive character states. What he is saying is that if all you know about a taxon is that it occupies a basal position in a phylogenetic tree, then you have no reason to suspect it will retain ancestral character states. In short, there are no “ancestral positions” in the crown of a phylogenetic tree. If horseshoe crabs, say, are primitive, this is something that needs to be established, not something that can be read off the family tree of arthropods.
A horseshoe crab from the Mesozoic
But if phylogenetic position is not predictive of morphological primitiveness, why do so many biologists speak as if it is? Jenner offers two explanations. The first points to a shortcoming of language: “In the absence of a vocabulary expressly designed to facilitate precise communication about lineages, we are forced to use the language of taxonomy, which was created for the realm of collateral relatives” (295). This creates problems when an investigator wants to frame an evolutionary narrative because of the mismatch between the temporally flat language of systematics and the need to temporalize relationships to fit a narrative structure. (Jenner’s examples are mostly instances of biologists designating one or another sister taxon as “basal” or “ancestral,” which he describes as a “phylogenetic faux pas.”)
The second explanation is epistemological. “Ancestral state reconstruction discovers character evolution by finding character state differences between sister taxa in a tree… Consequently, because each node in a tree offers a chance to sample character state differences between sister taxa, the probability of discovering character evolution along a lineage is positively correlated with the number of nodes along it” (296). Now, since basal taxa are separated from ancestral taxa by fewer nodes than less basal taxa, there is a risk that inferred character evolution along a basal lineage will be underestimated. This is termed a “node density artifact,” and its result is to exaggerate the amount of phenotypic conservatism in basal taxa.
In addition, “it is easy to get the false impression that more basal taxa are more likely to retain [ancestral] morphologies than less basal taxa” (296). To illustrate, Jenner conjures a pectinate tree composed of ten extant species. He observes that character evolution has the same probability of occurring along each of the 10 lineages, as each represents the same amount of elapsed time since the last common ancestor. “However, the chance that character evolution occurs in each of the eight supraspecific clades [i.e., the clades formed by excluding, sequentially, the most basal lineage]...is higher than for each of their sister taxa [i.e., the most basal lineage excluded], which are represented only by a single species, because these [supraspecific] clades encompass a greater total of evolutionary time.” As Jenner observes, “Such trees give us the correct impression that different amounts of character evolution have happened in sister lineages of unequal size… Yet they [give] us the wrong impression that more basally diverging lineages are more likely to retain ancestral character states than individual lineages that diverge higher up in the tree” (296–297). So basal position has no special association with morphological “ancestrality.”
On the left is what is known as a “balanced” tree. On the right is a “pectinate,” or comb-like, tree. The most basal taxon in the pectinate tree is labeled “B.” No terminal taxon in the balanced tree is more basal than any other
* * *
Near the end of his tenth chapter, Jenner describes a “paradox,” and it is with this paradox that I will conclude this essay.
The paradox is this. As it happens, basal unbranched lineages have strong effects on the reconstruction of ancestral states at the basal nodes of trees. This presents no problems on the assumption that these lineages are especially likely to retain ancestral character states. But, Jenner thinks, it is precisely these lineages that researchers should most distrust as avatars of ancestral forms. The reason is that “the less branched a lineage is, the less chance there is to infer character change along it, and the more likely it is that derived character states [will be] mistaken for primitive states” (297). Unbranched basal lineages are phylogenetic black boxes, and given that we cannot see inside them, there is an especial risk that we will fill them up with our own prejudices.
I am not sure why Jenner calls this a “paradox,” but nevermind that. The moral once again is that there is nothing ancestral about extant taxa occupying basal positions in a phylogenetic tree. (Consider again the horseshoe crab.) Every extant lineage in a clade has been evolving for the same amount of time, so every lineage should be treated as equally informative about the morphology of the last common ancestor. Students of body plan evolution disregard this advice at their own peril.
References
Baek, S.Y., et al. 2014. Complete mitochondrial genomes of Carcinoscorpius rotundicauda and Tachypleus tridentatus (Xiphosura, Arthropoda) and implications for chelicerate phylogenetic studies. International Journal of Biological Sciences 16:479–489.
Bolker, J.A. 1995. Model systems in developmental biology. BioEssays 17:451–455.
Fortey, R.A. 2011. Survivors: The Animals and Plants that Time has Left Behind. London: HarperCollins.
Jenner, R.A. 2022. Ancestors in Evolutionary Biology: Linear Thinking About Branching Trees. Cambridge: Cambridge University Press.
Nielsen, C. 2001. Animal Evolution: Interrelationships of the Living Phyla [Second Edition]. Oxford: Oxford University Press.
Ortega-Hernández, J., Janssen, R., Budd, G.E. 2019. The last common ancestor of Ecdysozoa had an adult terminal mouth. Arthropod Structure & Development 49:155–158.