* This is the latest installment of “Problematica.” It is part of an Extinct double-feature on the subject of negative evidence. You can find the other part, by Matt Brewer, here. Problematica is written by Max Dresow…
When I was a precocious young reader, my parents bought me an unauthorized anthology of the adventures of Sherlock Holmes. I must have read it ten times, getting a bit more of the picture each time through. One of my favorite stories had the title of a spaghetti western: “The Adventure of Silver Blaze.” The plot is typical Holmesian fare. A champion racehorse has vanished and its trainer is dead. Holmes is summoned to Derbyshire to investigate. A bookmaker named Simpson has been arrested for the murder, but upon examination the case against him begins to fray. Further investigation only deepens the mystery. Holmes uncovers a variety of clues: a box of matches, a burnt candle, a bill of sale for a fancy dress. But what finally unlocks the case is what Holmes calls the “curious incident of the dog in the night-time.” The incident was in fact a non-incident. When Silver Blaze was abducted, the eponymous dog made nary a yip. But that was the curious thing, for if Simpson or some other intruder had been the culprit, why hadn’t the dog barked? Evidently, the visitor was someone the dog knew well. In the event, it was none other than the horse’s trainer, who was killed when the animal administered a panicked kick, inadvertently foiling his criminal designs.
Holmes and Watson on the train to Derbyshire
There is more to the adventure, but we can stop here. What makes these events memorable is not the revelation of the trainer’s motive (something about satisfying a mistress with expensive taste). It is rather how the entire performance hinges on an event that didn’t happen. A dog didn’t bark, so the culprit must have had a familiar face— probably the face of the horse trainer. Implicit in the inference is a background of expectation. Dogs bark at intruders in the night unless they know them. (Or unless the intruder is especially stealthy: something that’s hard to pull off when a horse is involved.) Any whiff of paradox quickly dissipates. Sometimes evidence takes the form of expectations fulfilled; other times it takes the form of expectations frustrated. In either case, it functions by constraining interpretations against a background supplied by other beliefs and expectations. (In aphoristic form: negative evidence is evidence.)
I was recently made to think of Silver Blaze by a paper in Palgrave Communications. Its author is Efraim Wallach, and it makes an interesting— and I think, provocative— claim. Wallach observes that “inferences from absence” are typically regarded as invalid or “at best [as] weak and inconclusive” in the natural sciences. However, in archeology, inferences from absence are relatively common. Wallach lists as examples
the absence of evidence for fire use in early hominins (taken to support a younger date for the technology);
chronological gaps in the sequence of fossils and artifacts outside of Africa (taken to support various hypotheses about the timing and dispersion of early humans); and
the absence of domesticated plant varieties from archeological deposits in the Nile and Indus Valleys (taken to support claims about the origins and dispersal of agricultural practices).
Other examples involve not inferences from absence, but from paucity, but the basic point is unaffected. In Wallach’s words, “[it] is a demonstrable fact that inference and reasoning from absence are common in archaeology, often enjoying a status on par with other empirical inferences” (Wallach 2019).
But why should negative evidence enjoy a privileged status in archeology? Wallach puts this down to two factors. First, he notes that humans leave a “strong and distinctive footprint” in the stratigraphic record because we are constantly producing and discarding recognizable artifacts. Indeed, according to an estimate by Foley and Lahr (2015), the average human living during the period of stone tool evolution produced “between 10 and a 100 stone tools… per year, with each production accompanied by a débitage of many detectable [flakes].” Because these objects are easily recognized, human culture produces a distinctive stratigraphic signature. But the signature is not just distinctive, it is also durable; and this is Wallach’s second factor. “Human artefacts produced from inorganic materials… have a [relatively high] chance of leaving material traces that will withstand the elements of nature and destruction by humans.” So, because humans are constantly making and shedding durable artifacts, inferences from absence are often possible in archeology:
A sedentary or semi-sedentary human presence can be expected to result in a large number of discernible material remains at their site of living. Therefore, when a systematic and extensive search fails to discover such remains (of a particular period), this constitutes plausible evidence from absence that the site was uninhabited during this period. (Wallach 2019)
Débitage from stone tool-making (from Stemp and Rosenswig 2022)
Wallach’s claim about “[the] distinctiveness of archaeology” is an eye-catcher. If true, it represents an interesting addition to the philosophy of the historical sciences, and one that deserves more attention than it is likely to receive in Palgrave Communications. Still, Wallach enters a caveat near the end of the paper that has the effect of significantly complicating his thesis. According to Wallach, two situations encountered outside of archeology also favor the uptake of negative evidence. The first is when the research question under investigation is “quite general and refers to a wide swath of time and/or geography.” The second is when, for narrower questions, an “undisturbed— or only lightly disturbed— [assemblage of trace evidence]” exists. I agree that both situations are favorable to the uptake of inferences from absence (and paucity). But Wallach treats them as uncommon, which is how he can maintain his claim that archeology is special in virtue of the distinctive material profligacy of human societies. I am not convinced. At least in geology and paleontology, situations like these seem relatively common. This suggests that archeology may not be as distinctive as Wallach seems to think.*
[* I should say that I am not sure how strong a claim Wallach intends to enter on behalf of archeology. Since other sciences sometimes employ inferences from absence (as Wallach admits), his claim cannot be that archeology is unique in its handling of negative evidence. It must rather be that, in comparison to the other natural sciences, archeology is unusual in its valuation of arguments from absence and paucity. Still, the impression one receives from Wallach’s paper is that archeology is an extreme outlier. Most natural sciences don’t care a hang about negative evidence; indeed, “[s]cientific inferences from absence are usually treated as problematic, if not outright fallacious.” In archeology, by contrast, things are different. Here negative evidence is routinely leveraged in evidential arguments for the simple reason that humans are constantly making and shedding durable artifacts.]
Because I’m not sure how strong Wallach intends his claim to be, I’ll avoid coming down too hard on the thesis of archeological distinctiveness. Still, I think Wallach underestimates the importance of inferences from absence outside of archeology— specifically, in the other historical sciences. I will flesh out my skepticism with three examples, which span the areas of historical geology, stratigraphic paleobiology, and comparative biology/experimental taphonomy. The first two I develop at some length, because I happen to have published papers on them. The last one I run through more quickly. Together, they illustrate that historical scientists make use of negative evidence in diverse settings, and increasingly with tools designed to illuminate when evidential absences are to be expected. All of which significantly complicates the picture drawn by Wallach, with its special regard for archeology.
Case 1: HOMOTAXIS AND SYNCHRONICITY
In 1862, T. H. Huxley rose to address the Geological Society of London. His remarks, offered on behalf of the outgoing president, Leonard Horner, were characteristically lucid and provocative. He began with some High Victorian throat-clearing. Merchants, he observed, “occasionally go through a wholesome, though troublesome and not always satisfactory, process which they term taking stock.” After “the excitement of speculation, the pleasure of gain, and the pain of loss, the trader makes up his mind to face facts and to learn the exact quantity and quality of his solid and reliable possessions.” Scientists, Huxley proposed, should occasionally imitate this practice; and what better occasion than an anniversary address to the Geological Society? He thus proposed a theme for his address: “an inquiry… into the nature and value of the present results of paleontological investigation.”
Thomas Henry Huxley, who was a sprightly 37 years old in 1862
First, an inventory. What had paleontology achieved since it “came into full day… with Cuvier”? Some 40,000 species had been added to the taxonomic roll call, “a living population equivalent to that of a new continent in mere number… [and] to a new hemisphere, if we take into account the small population of insects as yet found fossil, and the large proportion and peculiar organization of many of the Vertebrata.” But that was just the beginning. Paleontology had given an immense spur to comparative anatomy— Huxley’s own field— as well as to botany, zoology, and osteology. In addition, it had established “two laws of inestimable importance.” The first was “that one and the same area of the earth's surface has been successively occupied by very different kinds of living beings”; the second “that the order of succession established in one locality holds good, approximately, in all.” For Huxley, these laws constituted the major achievement of paleontology during the past sixty years: the really philosophical results. So it was here he proposed to focus his attention.
He began with law number two, that local successions mirror global ones. From this it followed “that a peculiar relation frequently subsists between series of strata, containing organic remains, in different localities.” Call it resemblance:
The series resemble one another, not only in virtue of a general resemblance of the organic remains in the two, but also in virtue of a resemblance in the order and character of the serial succession in each. There is a resemblance of arrangement; so that the separate terms of each series, as well as the whole series, exhibit a correspondence.
Now, Huxley observed, in geology succession implies time. To say that two forms of life “succeed” one another in the geological column is to imply that the overlying form succeeded the underlying one temporally, not merely that it resides higher in a vertical stack of rock. The inference may not always be correct, especially in tectonically active areas. Still, in a general sense the reasoning is sound. So, “it was no wonder that correspondence in succession came to be looked upon as a correspondence in age, or contemporaneity.” Indeed, “so long as relative age only is spoken of, correspondence in succession is correspondence in age; it is relative contemporaneity.” (“Relative age” refers to an ordering of events in terms of “older-than” and “younger-than” relations, rather than in terms of a standardized unit of duration.)
An idealized view of a stratigraphic succession, with four kinds of shells succeeding one another in the stratigraphic column and by inference, in time (from Bjornerud 2018)
Still, Huxley thought, it would have been better if geologists had abstained from admitting “so loose and ambiguous a word” as contemporaneous into their thinking. In comparative anatomy, the term “homology” was used to denote an abstract relationship of correspondence between parts. (This was a multitasking term. As Toby Appel notes, “[n]aturalists sought homologies between the parts of different animals, homologies between the parts in a single animal, homologies between the structures in the fetus of higher animals and the adult form of lower animals, and homologies between structures in so-called monsters and those in normal animals” (Apel 1987, 4).) What geologists needed was a similar term “in order to express an essentially similar idea.” Huxley proposed “homotaxis,” or similarity of order. As an apologia for the new bit of jargon, he turned to a concrete example:
The Lias of England and the Lias of Germany, the Cretaceous rocks of Britain and the Cretaceous rocks of Southern India, are termed by geologists "contemporaneous" formations; but whenever any thoughtful geologist is asked whether he means to say that they were deposited synchronously, he says, “No,— only within the same great epoch.” And if, in pursuing the inquiry, he is asked what may be the approximate value in time of a “great epoch”— whether it means a hundred years, or a thousand, or a million, or ten million years— his reply is, “I cannot tell.”
Indeed, “[if] the further question be put, whether physical geology is in possession of any method by which the actual synchrony (or the reverse) of any two distant deposits can be ascertained, no such method can be heard of; … [since] neither similarity of mineral composition, nor of physical character, nor even direct continuity of stratum, are absolute proofs of the synchronism of even approximated sedimentary strata: while, for distant deposits, there seems to be no kind of physical evidence attainable of a nature competent to decide whether such deposits were formed simultaneously, or whether they possess any given difference of antiquity.”
This is where paleontology entered the picture, and indeed it was among the triumphs of the preceding generation of geologists to show that fossils provide a reliable criterion for stratigraphic correlation across large distances (Rudwick 1985). The matter once stirred ferocious debate; however, by 1862 most everyone accepted “that deposits containing similar organic remains are synchronous… [at least] in a broad sense.” Huxley demurred. “In point of fact,” he declared, “similarity of organic contents cannot possibly afford any proof of the synchrony of the deposits which contain them; on the contrary, it is demonstrably compatible with the lapse of the most prodigious intervals of time, and with the interposition of vast changes in the organic and inorganic worlds, between the epochs in which such deposits were formed.” Huxley’s point was that there is no simple relationship between similarities in organic and similarities in age. Any attempt to infer contemporaneity from organic resemblance depends on a raft of auxiliary assumptions that need to be established on independent evidence. We should like to know, for example, whether the animal populations of past eras were as different from one another as are the current populations of Europe, South America, and Australia. But to know this, we would need to know that we are comparing apples with apples— that is, the fossil contents of contemporaneous paleocontinents. And that is exactly what we are at a loss to establish.
For anything that geology or paleontology are able to show to the contrary, a Devonian fauna and flora in the British Islands may have been contemporaneous with Silurian life in North America, and with a Carboniferous fauna and flora in Africa. Geographical provinces and zones may have been as distinctly marked in the Paleozoic epoch as at present, and those seemingly sudden appearances of new genera and species, which we ascribe to new creation, may be simple results of migration.
An idealized illustration of fossil-based correlation between two geographically-separated stratigraphic successions. (By “correlation,” what I mean is the establishment of a [temporal] correspondence between geographically separated exposures or sedimentary units.) The situation Huxley is inviting his audience to consider is one in which the dashed lines are not “time lines” indicating synchronicity between depositional events. Instead, Huxley is highlighting the possibility that the events connected by these lines (the beginning of sedimentary deposition here and there) may be offset in time
There is apparently “no escape from the admission that neither physical geology, nor paleontology, possesses any method by which the absolute synchronism of two strata can be demonstrated.” All they can prove is “local order of succession [i.e., homotaxis].” The problem is especially vicious when the strata to be correlated are far apart, perhaps even on different continents, as Huxley made clear:
It is mathematically certain that, in any given vertical linear section of an undisturbed series of sedimentary deposits, the bed which lies lowest is the oldest. In many other vertical linear sections of the same series, of course, corresponding beds will occur in a similar order; but, however great may be the probability, no man can say with absolute certainty that the beds in the two sections were synchronously deposited. For areas of moderate extent, it is doubtless true that no practical evil is likely to result from assuming the corresponding beds to be synchronous or strictly contemporaneous; and there are multitudes of accessory circumstances which may fully justify the assumption of such synchrony. But the moment the geologist has to deal with large areas, or with completely separated deposits, the mischief of confounding that "homotaxis" or "similarity of arrangement," which 'can' be demonstrated, with "synchrony" or "identity of date," for which there is not a shadow of proof, under the one common term of "contemporaneity" becomes incalculable, and proves the constant source of gratuitous speculations.
Now, all this seems rather damning. Huxley went so far as to suggest that it squashed “all the grand hypotheses of the paleontologist respecting the general succession of life on the globe,” at least for the time being. Still, it is a remarkable fact that his remarks put hardly a dent in the confidence paleontologists reposed in their grand hypotheses (Dresow 2023a). Why?
I think the reason is that the problem of homotaxial correlation had already been effectively resolved by 1862, at least in the form Huxley was concerned with. The key to the solution was negative evidence (Dresow 2021). Consider Huxley’s example: how is it possible to know that “a Devonian fauna and flora in the British Islands [was not] contemporaneous with Silurian life in North America, and with a Carboniferous fauna and flora in Africa”? Part of the answer must be that we know of no sedimentary deposits where members of the Devonian fauna are found in association with members of either Silurian or Carboniferous ones. If such associations were discovered— or if Devonian and Carboniferous faunas were found interbedded at an exposure— then the jig would be up. (At least this would be the case if no other explanation could be provided for the anomaly.) But these sort of associations just aren’t observed, even with hardy cosmopolitan taxa like conodonts.* It follows that the absence of evidence of mixed-up faunas provides evidence of absence— the mixed up faunas probably didn't exist. With hordes of geologists combing exposures all over Europe and North America, the cumulative silence of the rocks spoke volumes. Indeed, I think it was decisive for answering lingering Huxleyan worries about massive asynchronism.
[* Of course, the faunas of successive periods do bleed into one another, which is just to say that geological periods are not separated by sharp discontinuities (with a few exceptions). Still, the bleeding is largely confined to the fuzzy edges of these periods, and does not threaten the argument from absence.]
Curiously, the subject of negative evidence does arise in Huxley’s address, in connection with the notion that “the commencement of the geological record is coeval with the commencement of life on the globe.” According to Huxley, this claim “rests entirely on negative evidence,” since only negative evidence can “prove the commencement of any series of phenomena.” Still, he cautioned, “the value of negative evidence depends entirely on the amount of positive corroboration it receives.”
If A. B. wishes to prove an alibi, it is of no use for him to get a thousand witnesses simply to swear that they did not see him in such and such a place, unless the witnesses are prepared to prove that they must have seen him had he been there. But the evidence that animal life commenced with the Lingula-flags, e.g., would seem to be exactly of this unsatisfactory uncorroborated sort. The Cambrian witnesses simply swear they "haven't seen anybody their way"; upon which the counsel for the other side immediately puts in ten or twelve thousand feet of Devonian sandstones to make oath they never saw a fish or a mollusk, though all the world knows there were plenty in their time. But then it is urged that, though the Devonian rocks in one part of the world exhibit no fossils, in another they do, while the lower Cambrian rocks nowhere exhibit fossils, and hence no living being could have existed in their epoch.
Some fossils from the Lingula flags, the first of which were described by the paleontologist John Salter in the 1850s
Regarding the last claim, Huxley was unconvinced: “the observational basis of the assertion that the lowest rocks are nowhere fossiliferous is an amazingly small one, seeing how very small an area, in comparison to that of the whole world, has yet been fully searched.” With so much of the world un-geologized, claims about the absence of life from the earliest rocks needed to be held in suspense.
Huxley might have had the same thought about synchronicity. (That is, if he realized that that case also turned on negative evidence.) Yet whether or not he realized it, there was an important difference between the two cases. In the latter case every fossiliferous section was the source of potential anomalies, not just early Cambrian ones. So, in the synchronicity case, absence of evidence provided much stronger evidence of absence than in the early-fossils case, even in 1862.
Case 2: Stratigraphic paleobiology
My first vignette involved reasoning about “a wide swath of time and/or geography.” This is one of the situations Wallach thinks favors the uptake of arguments from absence outside of archeology (although he seems to regard such situations as rare). The next case involves a practice that is not captured by this caveat. It comes from the recent history of paleobiology, but to understand it you need to know a thing or two about the branch of stratigraphic geology known as sequence stratigraphy. This necessitates a short (and kind of intense) digression. Sorry!
Sequence stratigraphy is not for the faint of heart. Nearly everything about it is demanding, not least its extensive and abstruse terminology. But distilled to its essence it is the study of “repetitive cycles of [sediment] accumulation followed by [gaps], at various time scales” (Miall 2015, 295). It is based on process models of sedimentary accumulation, which provide a framework for interpreting the stratigraphic record in terms of a small number of variables: things like rate of [eustatic] sea-level change, tectonic subsidence, and sediment supply. The variables are in turn related to each other, with tectonism and sea-level change controlling the space available for sedimentary accumulation (termed “accommodation [space]”), and changes in accommodation controlling the accumulation of sediments over tens of thousands to millions of years. (On shorter time scales, sedimentary accumulation is dominated by geologically instantaneous events like storms and submarine landslides.)
“Accomodation [space]” is the space available for potential sedimentary accumulation. It is determined by global sea-level changes and tectonic uplift and subsidence
This hopefully doesn’t sound too complicated. If the sea-floor goes up or down, or if global sea level rises or falls, the envelope of space available for sedimentary accumulation grows or shrinks. But how do sediments accumulate within this envelope? Here is where things get trickier.
The first thing to know is that sequence stratigraphy decomposes the stratigraphic record into a hierarchy of sedimentary units, the most important of which are termed depositional sequences. These are packages of sediment that can be hundreds of meters thick and typically represent millions of years of elapsed time. (Think of them as wedges of sediment that taper at their edges.) Depositional sequences are in turn composed of stratigraphic successions termed systems tracts, which are deposited during a particular phase of the relative sea-level cycle. The “low stand systems tract,” for example, includes deposits that accumulate after the onset of a relative sea-level rise. There is also a “high stand,” a “falling stage,” and a “transgressive” systems tract, not all of which are present in every depositional sequence. Finally, systems tracts are composed of sets of strata representing single episodes of sedimentary accumulation: parasequences. These are typically between one and ten meters thick and represent tens to hundreds of thousands of years of elapsed time. They are, so to speak, the building blocks from which systems tracts and depositional sequences are constructed.
The anatomy of a dispositional sequence. Notice the stacked parasequences bundled into systems tracts and separated by characteristic surfaces (to be discussed below). Also notice that this image shows two depositional sequences stacked one on top of the other. The colors on the diagram represent distinct depositional environments, or “facies”
Now, parasequences, systems tracts, and depositional sequences are so many packages of sediment. But just as important are the surfaces that bound and help to delineate them. In fact, it is these surfaces that supply the key to understanding the distribution of gaps in the stratigraphic record— an important feature of sequence stratigraphic analysis.
To raise the subject of surfaces, however, is to ask how sequence models depict the process of sedimentary accumulation in response to factors like sediment supply and accommodation space. Here I quote from my own recent paper (with apologies for repeating certain parts of the previous paragraph):
According to [a standard] model, when the rate of sediment supply exceeds the rate of increase in accommodation, sediment accumulates, forming packages of sediment called parasequences. These are successions of [non-gappy] strata bounded at their tops by “flooding surfaces” associated with deepening events (often periods of non-deposition [yielding gaps]). Parasequences are produced by oscillations in the balance between sediment supply and accommodation; but these oscillations are superimposed on longer-term changes that build the [larger] features of depositional sequences. The most important of these are called systems tracts, and consist of sets of parasequences bounded by surfaces of various kinds. The names of the systems tracts are not important here, but what is important is that they succeed one another in regular order, and that the collection of systems tracts is topped by a sequence boundary. This is the surface that separates one depositional sequence from another, and forms when sea-level falls, promoting the erosion of exposed sedimentary deposits. Sequence boundaries typically represent significant periods of time in which no sediment accumulates [so, big gaps]. But they are not the only chronostratigraphically significant surfaces in a sequence, and other surfaces, like the maximum flooding surface, are also associated with periods of highly reduced deposition.* (Dresow 2023b)
[* To say that a surface is “chronostratigraphically significant” is to say that all the rocks overlying it are everywhere younger than all the rocks underlying it, which is different from saying that the surface is isochronous, or that it represents a time line. But never mind: what matters is the big picture.]
A simplified diagram of how a depositional sequence is constructed in response to changes in relative sea-level. First the low-stand system tract (LST) builds out from the shore; then, when sea-level rises, the transgressive system tract (TST) builds shoreward; finally, the highstand system tract (HST) builds seaward. Notice the maximum flooding surface (MFS) in red, and the sequence boundary in green
Pause for beat. I know I’ve thrown a lot at you, but if you can remember that sequence stratigraphy decomposes the stratigraphic record into a set of sedimentary packages and intervening surfaces (interpreted in terms of a process model of sedimentary basin filling), then you’ve got the gist of it.
One last thing. In addition to illuminating the nature and distribution of gaps in the record, sequence stratigraphy is also informative about the distribution of depositional environments (“facies” in the lingo):
Consider parasequences. Each parasequence records the seaward movement of a shoreline. This means that if one examines a parasequence in cross-section, the facies within the succession will represent progressively shallower environments as one moves from the bottom of the succession to the top (until one reaches the flooding surface, at which point a deep water facies will give way to a shallow water one). But there are intelligible patterns at larger scales too. Parasequences typically occur in groups or sets that display consistent trends in their component facies and three-dimensional arrangement. The most conspicuous of these trends are stacking patterns, which, together with particular boundaries and the overall position of the set in the sequence, are used to define systems tracts. Depositional facies are in turn associated with particular stratigraphic positions, which reflect the sedimentological response to changes in sea-level, sediment supply, accommodation, and other factors. Because of this, sequence analysis can be used to predict the distribution of lithologies in basins: something that can help make sense of the apparently crazy-quilt pattern of sedimentary deposits in [the rock record]. (Dresow 2023b)
Now, what does this have to do with paleobiology and negative evidence? For as long as paleontologists have contemplated field data, they have confronted the question of how trustworthy the fossil record really is (Sepkoski 2012). Darwin famously referred to it as “a history of the world imperfectly kept,” and subsequent paleontologists agreed that field evidence was prone to mislead. But this did not settle the question of when field evidence was likely to be trustworthy and when researchers should turn a skeptical eye. For this, tools were required to dissect the incompleteness of the fossil record into component biases, or what comes to the same thing, to extract biological signals from the distorting effects of stratigraphic overprint.
It was perhaps not immediately obvious that sequence stratigraphic models would be useful in this connection. These models were developed in the context of petroleum exploration and only slowly made their way into more academic lines of research. Paleobiologists began using them during the 1980s to constrain past sea-level change and then to develop integrated models of depositional environments and paleoecology. However, by the mid-1990s, it was increasingly recognized that these models could be used to analyze the incompleteness of the fossil record in marine siliciclastic and carbonate systems. (Today, sequence stratigraphic models are applied to terrestrial siliciclastic systems as well.)
One person who recognized this was Steven Holland, an American paleontologist who completed his PhD under Susan Kidwell. In 1995, Holland published a first-of-its-kind modeling study, which used computer simulations of sedimentary basins to understand what controls the distribution of fossils. The details of the model aren’t relevant here, apart from the fact that it contained no interesting biology. Holland’s goal wasn’t to simulate interesting biology; it was to isolate the generic signature of stratigraphic overprint— those spikes in the first and last appearances of taxa that were due to changing depositional conditions (as opposed to competition or migration or whatever). If sedimentary processes imparted no structure to the fossil record, there would be no spikes at all. But there were spikes, and unsurprisingly, they came at just the points in the depositional sequence that you would expect: above (TST) flooding surfaces for first appearances, and below the sequence boundary and (TST) flooding surfaces for last appearances.
Results of Holland’s model, showing numbers of first and last occurrences for single stratigraphic sections throughout two depositional sequences. Time runs along the vertical axis. Systems tracts, corresponding to strata deposited during a phase of relative sea-level change, are indicated in the center of the figure. (Note that no sediment is deposited during the LST.) Major spikes in first occurrences happen just above the sequence boundary (the juncture between the unrecorded LST and the overlying TST) and at all TST flooding surfaces (contacts exhibiting evidence of sharp increases in water depth). Major spikes in last occurrences occur just below the sequence boundary and at all TST flooding surfaces
The upshot was clear. So long as paleobiologists were working within a sequence stratigraphic framework, they could pinpoint where in the sequence the pattern of first and last occurrences was likely to be misleading. So, if a paleobiologist were to observe an apparent extinction pulse just before a sequence boundary, it should be treated with care, since sequence boundaries reflect long periods of non-deposition. It’s even possible that none of the species that apparently went extinct at the boundary actually went extinct at the boundary. Similar counsels apply to the pulses associated with other surfaces, like the maximum flooding surface.
What does this have to do with negative evidence? Simply this. In order to interpret an absence in the fossil record as a genuine absence, it is useful to know whether the absence is likely to have been the result of depositional factors. Sequence stratigraphy aids greatly with these inferences (Patzkowsky and Holland 2012). Above a sequence boundary, the absence of a taxon provides only weak evidence that the taxon went extinct at the boundary, precisely because the taxon could have been thriving during the interval represented by the hiatus. Likewise, if a taxon has a strong environmental preference, evidence for its absence will be weak if, above a certain horizon, that environment is no longer represented in the succession.* Backing up: the ability of paleobiologists to infer a genuine absence— like an extinction— from an absence of evidence— missing fossils— is enhanced by tools that tell them whether the absence is likely to be an artifact of preservation. These tools are now available for many stratigraphic environments. And so the power of negative evidence in paleontology has been greatly augmented.
[* Recall Huxley’s point that the testimony of witnesses to an absence should only be trusted when the witnesses are prepared to testify, not only that they did not see the suspect, but that they would have seen him if he had been there. Translating: the stratigraphic record can only be regarded as a trustworthy witness when, for any taxon, it is likely to have recorded evidence of that taxa when it was present.]
Case 3: “URBI”
The last case can be handled more quickly.
A very long time ago, perhaps as much as 700 Ma, a humble creature swam the world’s oceans. It was the urbilaterian, the last common ancestor of the protostome and deuterostome bilateral animals, and as its name suggests, it was bilaterally symmetrical: its right half was a mirror image of its left half. The urbilaterian must have existed. This is implied by the topology of the evolutionary tree of animals. Yet it left no known traces in the fossil record, including “traces” in the technical sense of ichnofossils or structures produced by biological activity. The challenge is to reconstruct it. What can we know about a creature that lived so long ago and that (apparently) left no record of its anatomy or activities?
The answer is “more than nothing, and possibly quite a bit”! The reason is that comparative phylogenetics allows researchers to infer the tool-kit of developmental genes present in the urbilaterian, which in turn permits an inference about its morphology. The work commenced in the 1990s, and picked up steam in the early 2000s. From the outset it was claimed that the urbilaterian was probably a complex (as opposed to a simplified) animal:
Urbilateria had a Hox gene complex consisting of at least 7 Hox genes flanked by up to 8 additional Antennapedia-type homeobox genes forming a Super-Hox complex (Butts et al., 2008). The Hox genes stayed together during evolution due to spatial, temporal, and transcriptional requirements (Darbellay et al., 2019; Duboule, 2022). When the colinear expression of Hox complexes in both the Drosophila and mammalian A-P axes was discovered, it became widely accepted that the gene system had to have been present in their last common ancestor. Such an elaborate system, including two regulatory microRNAs within the complex, could not have been assembled multiple times independently by convergent evolution (De Robertis, 2008). (De Robertis and Tejeda-Munoz 2022)
But here is a puzzle. If the urbilaterian was such a complicated organism, and presumably “large” by Precambrian standards, then where is it? Why has no trace of it been found in rocks of Cryogenian or Ediacaran age? The Neoproterozoic fossil record contains abundant and diverse trace fossils, and as Erwin and Davidson (2002) write, “a complex, benthic, vermiform [urbilaterian] equipped with appendages and segments would almost certainly have been detected, either as a trace or body fossil in Neoproterozoic deposits.” Such an animal, apparently, did not exist. But the urbilaterian existed. So— the inference goes— it was not “a complex, benthic, vermiform equipped with appendages and segments.”
Two reconstructions of the urbilaterian, one “complex” and one simple (from Wikipedia)
Here is Erwin and Davidson’s view. According to them, “the [urbilaterian] possessed the essential bilaterian toolkit for morphogenetic pattern formation, and it deployed many of the differentiation gene batteries that its modern descendants continue to rely on.” But this does not require an inference to morphological complexity. “The complexity of the [urbilaterian] may have been a great advance on its predecessors, but the safest assumption is that its morphology was unprepossessing. It had an AP axis, a two-ended gut, mesodermal layers and a central and peripheral nervous system with sensory cell types.” Possibly it was also “very small… compared to most modern bilaterians,” a feature that would help to account for the absence of a trace fossil record, with or without architectural complexity.
I promised a quicker treatment of this case, so I will limit myself to one final remark. A recent study has tried to make sense of the (absence of a) fossil record for early animal evolution by studying the distribution of fossilization processes over time (Anderson et al. 2023). The authors note that most early animals are preserved as carbonaceous remains in fully marine, fine-grained, siliciclastic rocks (“Burgess Shale-type preservation”). However, in the few assemblages characterized by this mode of preservation dated to 789 Ma or earlier, there is no evidence for metazoans. (Here it is metazoans that are in focus, not the urbilaterian specifically.) “Thus, taphonomic evidence argues for a maximum constraint of 788.72 ± 0.29 Ma on crown animal antiquity.” By characterizing the distribution of fossilization processes over time, and by leveraging background knowledge about the kinds of processes likely to preserve metazoans, scientists can make inference about the likely presence or absence of animal ancestors in fossil biotas.
* * *
A final observation to conclude. As Matt Brewer has shown, the expression “absence of evidence is not evidence of absence” has a geological provenance. It was first used in the waning years of the nineteenth century by scientists arguing about glacial phenomena, and from there entered sciences like astronomy and archeology. It makes sense that a geoscientist would have originated this expression of caution. Many absences in the rock record are indeed the byproducts of incomplete preservation. And yet the originator of the expression, Dugald Bell, in his geological practice, actually appealed to the absence of evidence as evidence of absence. This is because, even in geology with its patchwork records, some absences are genuinely unexpected.
It was like that in the nineteenth century. And it’s like that now, but more so. The major difference is that today, researchers are in a much better position to say where bits of trace evidence are to be expected. Whether it’s models of sedimentary basin filling or empirical studies of the distribution of fossilization processes over time, negative evidence in the historical sciences has never been more richly scaffolded. Claims of archeological distinctiveness are thus to be taken with a healthy pinch of salt.
References
Anderson, R. P., Woltz, C. R., Tosca, N. J., Porter, S. M., and Briggs, D. E. G. 2023. Fossilisation processes and our reading of antiquity. Trends in Ecology and Evolution 38:1060–1071.
Bjornerud, M. 2018. Timefulness. How Thinking Like a Geologist Can Help Save the World. Princeton: Princeton University Press.
Apel, T. A. 1987. The Cuvier-Geoffroy Debate: French Biology in the Decades Before Darwin. Oxford: Oxford University Press.
Dresow, M. 2021. Measuring time with fossils: a start-up problem in scientific practice. Philosophy of Science, https://doi.org/10.1086/714855.
Dresow, M. 2023a. Deep time: historical and philosophical perspectives. In Time and Science, ed. P. Harris and R. Lestienne. World Scientific Publishing Co.
Dresow, M. 2023b. Biased, spasmodic and ridiculously incomplete: sequence stratigraphy and the emergence of a new approach to stratigraphic complexity in paleobiology, 1973–1995. Journal of the History of Biology. https://link.springer.com/article/10.1007/s10739-023-09720-0.
Erwin, D. H. and Davidson, E. H. 2002. The last common bilateral ancestor. Development 129:3021–3032.
Foley, R. A. and Lahr, M. M. 2015. Lithic landscapes, early human impact from stone tool production on the central Saharan environment. PLoS One, 10(3):e0116482.
Holland, S. M. 1995. The stratigraphic distribution of fossils. Paleobiology 21:92–109.
Huxley, T. H. 1862. Geological contemporaeity and persistent types of life. https://www.gutenberg.org/files/2936/2936-h/2936-h.htm.
Miall, A. D. 2015. Making stratigraphy respectable: from stamp collecting to astronomical calibration. Geoscience Canada 42:271–302.
Patzkowsky, M. E. and Holland, S. M. 2012. Stratigraphic Paleobiology: Understanding the Distribution of Fossil Taxa in Time and Space. Chicago: University of Chicago Press.
Rudwick, M. J. S. 1985. The Great Devonian Controversy: The Shaping of Scientific Knowledge Among Gentlemanly Specialists. Chicago: University of Chicago Press.
Sepkoski, D. 2012. Rereading the Fossil Record: The Growth of Paleobiology as an Evolutionary Discipline. Chicago: University of Chicago Press.
Stemp, W. J. and Rosenswig, R. W. 2022. Archaic period lithic technology, sedentism, and subsistence in Northern Belize: what can debitage at Caye Coco and Fred Smith tell us? Latin American Antiquity 33:1–20.
Wallach, E. 2019. Inference from absence: the case of archaeology. Palgrave Communications. https://doi.org/10.1057/s41599-019-0307-9.