* This is the seventh installment of “Problematica.” It is written by Max Dresow…
Here is a question. What do biological species, meteorites, cyclothems, and episodic memories have in common? Answer: they are natural groupings defined on the basis of historical properties.
Scientists often classify objects on the basis of their histories. Many of these classifications play important roles in successful scientific projects. And according to a popular tradition, classifications that play important roles in successful scientific projects come as close as we’re likely to get to “carving nature at its joints.”
So why have philosophers been slow to develop a general account of “historical kinds”? I can only speculate, but perhaps it has to do with the intuition that natural kind terms apply to what George Simpson calls the “immanent” as opposed to the “configurational” properties of the universe (Simpson 1964). Since historical properties are merely configurational, historical kinds are unlikely to capture any really natural divisions. They may carve nature, but not at its joints. A second possibility ascribes the oversight to a dearth of philosophical interest in geology. Had philosophers paid more attention to geology, then the need for an account of historical kinds may have been evident sooner. But philosophers have mostly ignored geology, so the need was not felt. (Incidentally, there is no love lost between geologists and philosophers. A famous geologist once remarked that whenever he heard a philosopher comment on his field, his response was to reach for his hammer. Sadly, history failed to record the philosophical abuse this man was subjected to.)
Whatever the case, philosophers have had relatively little to say about historical kinds as such. But the times they are a-changin’. For one thing, the neglect of geology seems to be over. During the last ten years, philosophical work on the earth sciences has exploded, including some promising work in the philosophy of geology. Over the same period, philosophical interest in historical kinds has begun to percolate. Especially noteworthy are Laura Franklin-Hall’s discussion of animal sexes as “historical explanatory kinds” (Franklin-Hall 2020), Adrian Currie’s remarks on historical kinds in Scientific Knowledge and the Deep Past, and Alisa Bokulich’s general account of scientific types, which draws its main examples from stratigraphy and taxonomy (Bokulich 2020).
But none of these accounts is the subject of this essay. Instead, I want to examine what is arguably the most capacious account of historical kinds to appear in recent years. This is Muhammad Ali Khalidi’s picture of historical (or as he prefers to say, “etiological”) kinds, which appeared last year. The account is refreshingly easy to describe. Khalidi defines a historical kind as “[a set of objects] whose members share an origin, history, or causal trajectory.” Following Franklin-Hall, he distinguishes between “type-historical kinds,” which categorize objects on the basis of the types of processes that produce them, and “token-historical kinds,” which categorize objects on the basis of shared token histories. This yields a sixfold taxonomy of historical kinds, which Khalidi illustrates with examples. Meteorites are a type-historical kind, since they share the property of having arrived on Earth from elsewhere in the universe. Ditto major rock types. An igneous rock is an igneous rock because it formed through a kind of process involving the cooling and solidification of magma. Biological species, by contrast, are token-historical kinds, since members of a species trace their ancestry to a particular historical event and not simply a type of event or process. (Token historical kinds can also be described as individuals with parts that share an origin, history, or causal trajectory.)
I am a philosopher of science with a weakness for pragmatism—I am not a metaphysician. As such, what interests me about Khalidi’s account is how it metabolizes actual scientific material. To get a handle on this, my strategy will be to force-feed it an example from the history of paleontology. This is Jack Sepkoski’s characterization of the three “great evolutionary faunas”: a seminal contribution to the literature on evolutionary paleobiology. I have chosen the example not because it is typical, but because it is challenging. It is a tough meal to digest. Still, I do not offer it in the usual spirit of a philosophical counterexample. The indigestion I’m after is a means to the ends of refinement and clarification. I come in peace, with a challenge.
* * *
Jack Sepkoski was arguably the most influential paleobiologist of the last quarter of the twentieth century. During this time he published a number of influential works on the pattern and dynamics of diversification in marine settings. I will focus on just one of these, which appeared in 1981 under the title “A factor analytic description of the Phanerozoic marine fossil record.” This was Sepkoski’s attempt to extract general patterns from his self-curated database of marine fossil taxa, and to see if these threw any light on the dynamics of evolution over many millions of years.*
[* Basically, Sepkoski had a database consisting of the names and stratigraphic ranges of all known marine fossil families. And he wanted to know whether there was anything general he could say about the construction of family-level diversity over the past half-billion years. He also wanted to know what was responsible for the patterns he uncovered. That is, he wanted to gain insight into the dynamics of diversification over very long time scales.]
As the title indicates, Sepkoski’s study employed a method called factor analysis. This is a way of accounting for correlations in datasets using a smaller number of underlying explanatory variables. In Sepkoski’s case, factor analysis was performed on a dataset consisting of all the known families of marine organisms along with their temporal ranges. The exact methodological details needn’t bother us; but what he found was that the data have a relatively simple structure. Only three factors were needed to account for over 90% of the data, which is to say, only three factors were needed to explain patterns of covariation for over 90% of marine taxa. This suggested that the history of life could be split into three broad assemblages. First was a trilobite-dominated “Cambrian fauna”; then came a “Paleozoic fauna” dominated by brachiopods and so-called rugose and tabulate corals; and finally there was a “Modern fauna” dominated by gastropods and bivalves, along with bony and cartilaginous fishes. Sepkoski labeled each of these associations a “great evolutionary fauna,” and proceeded to observe that they are “intimately associated with [particular phases] in the history of total marine diversity” (Sepkoski 1981, 36).
Now it seems clear to me that Sepkoski’s faunas have a plausible claim to be regarded as natural kinds, since we might reasonably think that these associations carve total marine diversity at its joints. (Here I want to shelf discussion of whether Sepkoski’s partitioning of diversity into faunas is the correct one. All I will assume is that some such partitioning is reasonable.) Moreover, they are historical natural kinds if they are natural kinds at all. The question thus becomes: where do the great evolutionary faunas fit within Khalidi’s etiological framework? In virtue of what are members of an evolutionary fauna members of that particular fauna?
Recall Khalidi’s central claim: that a historical kind is one whose members share a (token or type) origin, history, or causal trajectory. This provides six potential hooks to hang the great evolutionary faunas upon.
Begin with token origin. Is it the case that great evolutionary faunas “have members that all originate in the very same event”? They do not, since the event that gives rise to each evolutionary fauna is a prolonged episode of rapid diversification—but many families originate after this episode has wound down, and these families are no less members of the evolutionary fauna than those that originate in the initial burst. A trilobite family that originated in the Ordovician period, during which time the Cambrian fauna was in decline and the Paleozoic fauna was taking off, nonetheless counts as a member of the Cambrian fauna by Sepkoski’s criteria. So members of evolutionary faunas are not grouped together in virtue of a shared token origin.
Type origin, on the other hand, seems more promising. As noted, each great evolutionary fauna begins with a period of exponential growth in family numbers. The faunas are largely delineated on the basis of this diversification event, so evolutionary faunas can be said to share a type origin. But notice that here it is evolutionary faunas themselves that are in focus—not members of these faunas (particular taxa). Members of particular evolutionary faunas do not all share a type origin: only the evolutionary faunas themselves do. So the Cambrian evolutionary fauna is a member of the type EVOLUTIONARY FAUNA in part because it shares a type origin with other evolutionary faunas. But members of the Cambrian evolutionary fauna (particular taxa) are not members of the Cambrian fauna in virtue of sharing a type origin with one another.
There are two points here. The first is that members of an evolutionary fauna do not originate in a burst of diversification. It is faunas that do this, not taxa. The second is that not all taxa belonging to an evolutionary fauna originate during the initial burst; so they do not all originate in the same type of event anyway. There is also a complication. As Sepkoski states, the data suggest “that the [past half-billion years] can be divided into three basic intervals, each characterized by a different ‘style’ of diversification” (Sepkoski 1981, 49, emphasis added). Sepkoski understands “styles of diversification” in terms of the shape of diversity curves, which is different for each fauna. So depending on how you characterize the episodes of diversification inaugurating a fauna—as basically similar with some differences, or as distinct “styles of diversification”—it may or may not be the case that evolutionary faunas share a type origin.
I’m going to skip over shared history for a moment to discuss shared causal trajectories. Are evolutionary faunas distinguished by their members sharing a (token or type) causal trajectory? I think it is clear that they don’t share a token trajectory. Members of the evolutionary faunas share no obvious causal trajectory vis-à-vis the fauna as a whole, and certainly they are not assigned to evolutionary faunas on the basis of any such trajectory. What about a type causal trajectory? Again, this seems more promising, but only if we “frameshift,” so to speak, and consider members of the kind EVOLUTIONARY FAUNA as opposed to members of these faunas. Consider that each evolutionary fauna is associated with a (1) distinct phase of diversification (of variable length), (2) a subsequent plateau (again, of variable length) and (3) an eventual decline (excepting the Modern fauna, which is not yet in decline). The faunas are not delineated on the basis of these “life histories”—it is not the case that a diversity assemblage was identified as an evolutionary fauna because it underwent a life cycle of logistic growth followed by decline. Still, elements of the causal trajectory of an assemblage are clearly relevant for defining what an evolutionary fauna is, and in this sense, facts about a type of causal (or perhaps only phenomenological) trajectory are built into the concept of an EVOLUTIONARY FAUNA.
Finally, consider history. Are evolutionary faunas kinds whose members share “the selfsame [token or type] history” (p. 3)? I’m not sure how to answer this. By most accounts, history consists of an unrepeatable sequence of unique events, which suggests that “sharing a selfsame history” means participating in or being affected by the same set of unrepeatable or unique events. Presumably there are no discrete events that all members of an evolutionary fauna participate in or are affected by. This means (if I have understood the concept of “shared history” correctly) that members of a fauna cannot be said to share a token history. Yet recall Sepkoski’s claim that each evolutionary fauna is “intimately associated with a particular phase in the history of total marine diversity” (Sepkoski 1981, 36). This could be taken to mean that it is a taxon’s “association with a particular phase in the history of marine diversity” that sorts it into a fauna. The proposal is not altogether straightforward. As you can see in the figure (above), faunas overlap one another in time, so the mapping of faunas onto periods of history is not one-to-one. Still, as the names of the faunas indicate, there is something important about the temporal location of a fauna, such that what it is to be a particular fauna (and also, I take it, a member of that fauna) is partly a matter of being situated at a particular juncture in the history of marine diversity.
I am not sure whether this means that members of a fauna share a type history. But it strikes me that this is the most promising niche in Khalidi’s account for the great evolutionary faunas, assuming I have interpreted the categories correctly.
* * *
As I said before, these comments are not offered in the spirit of a counterexample. Instead, they are intended to show where the framework bulges when it is asked to digest a substantial and difficult scientific meal. Khalidi says that historical kinds are kinds whose members “share a (token or type) origin, history, or causal trajectory.” But in the present case, it doesn’t seem like the members of an evolutionary fauna share a token or type origin or causal trajectory, and it is questionable whether they can be said to share a “history.” By contrast, evolutionary faunas themselves share all or none of these depending on how the criteria are interpreted and the empirical phenomena characterized. Clarifying the criterion of shared history would obviously help in resolving these difficulties. But if I were to make a suggestion, it would be to consider a category of historical kinds whose members share a temporal location, or even a position in a temporal succession, as opposed to a “history” per se. This would go some way towards illuminating why members of an evolutionary fauna constitute a meaningful association, even though they do not share a (token) origin or causal trajectory.
I have so far ignored Khalidi’s distinction between “pure” and “impure” kinds: between kinds delineated solely with respect to historical properties and those only partly delimited on the basis of historical properties. But here it bears mentioning that evolutionary faunas are “impure kinds,” since they are delineated not only on the basis of their temporal position with respect to other faunas, but also in virtue of their members’ shared ecologies (Alroy 2004). This distributes the burden of accounting for kind-membership over a set of properties that includes both historical and non-historical ones. And this, I think, makes it more plausible to claim that fauna members share only a fairly thin historical property like temporal location. The historical property can be thin because it is not doing all the work of delineating the relevant kind. Shared ecology is at least as important.
I end with a word of advocacy. According to Khalidi’s general account of natural kinds, members of a kind are entities that occupy a shared node in the causal structure of the world (Khalidi 2018). This means that natural kinds “divide the world into individuals that share causal properties, enter into the same or similar causal relationships, and give rise to the same or similar causal processes.” In all this, explanatory considerations are paramount—Khalidi exhibits a pronounced hesitancy to embrace “truthful description” as an important goal of scientific inquiry, on a par with prediction and explanation. But in the historical sciences, truthful description is a weighty accomplishment indeed, and supplies the goal of many research projects (Dresow 2021; Dresow and Love 2022). This includes Sepkoski’s description of the great evolutionary faunas, whose primary aim is to reduce the chaos of the fossil record to something resembling order and simplicity.
We should not shrink from the implication, nor should we doubt the capacity of sophisticated descriptive practices to uncover the contours of natural groupings. While I am inclined to agree with Khalidi that prediction and explanation provide our best guide to nature’s divisions, sophisticated practices of characterization provide reliable guides as well.
References
Alroy, J. 2004. Are Sepkoski’s evolutionary faunas dynamically coherent? Evolutionary Ecology Research 6:1–32.
Bokulich, A. 2020. Understanding scientific types: holotypes, stratotypes, and measurement prototypes. Biology & Philosophy 35:1–28.
Currie, A.C. 2019. Scientific Knowledge and the Deep Past. Cambridge: Cambridge University Press.
Dresow, M. 2021. Explaining the apocalypse: the end-Permian mass extinction and the dynamics of explanation in geohistory.” Synthese. https://doi.org/10.1007/s11229-021-03254-w. [Despite the mention of “explanation” in the title, this paper is largely about the importance of descriptive or “characterizational” research in geohistory.]
Dresow, M., and Love, A.C. 2022. The interdisciplinary entanglement of characterization and explanation. The British Journal for the Philosophy of Science. https://doi.org/10.1086/720414. [This paper offers an updated account of scientific characterization for complex phenomena, focusing on the Cambrian Explosion.]
Franklin-Hall, L. 2020. The animal sexes as historical explanatory kinds. In S. Dasgupta, R. Dotan, B. Weslake (Eds.), Current Controversies in Philosophy of Science, 177–197. New York: Routledge.
Khalidi, M. 2018. Natural kinds as nodes in causal networks. Synthese 195:1379–1396.
Khalidi, M. 2022. Etiological kinds. Philosophy of Science 88:1–21.
Sepkoski, J.J., Jr. 1981. A factor analytic description of the Phanerozoic marine fossil record. Paleobiology 7:36–54.
Simpson, G.G. 1964. This View of Life: The World of an Evolutionist. New York: Harcourt, Brace, & World.