Fossil-based analyses of clades' diversification patterns require taxonomic expertise and appropriate methodology

This paper argues that the conflicting findings of Gardiner et al. regarding neoselachian diversification patterns, particularly concerning the K/Pg extinction and recent diversity decline, stem from a combination of incomplete and erroneous fossil data coupled with inappropriate analytical methodologies, thereby underscoring the critical need for taxonomic expertise and robust methods in fossil-based macroevolutionary studies.

The Gist

Imagine you are trying to write the history of a famous family, like the "Shark Dynasty," by looking through a giant, dusty attic filled with old photos, letters, and artifacts. This is what paleontologists do when they study how sharks have evolved over millions of years.

Recently, a group of researchers (let's call them Team GEA) tried to write this history using a super-smart computer program (a "deep-learning" model) to scan a massive digital collection of shark fossils. They claimed to have found a surprising story: that sharks barely suffered during the asteroid impact that killed the dinosaurs (the K/Pg extinction), and that their numbers have been slowly dropping ever since.

The authors of this new paper (Guinot and colleagues) are saying: "Stop the presses! The story is wrong, the photos are blurry, and the computer was fed bad data."

Here is the breakdown of their critique using simple analogies:

1. The "Bad Photo Album" Problem (Data Quality)

Team GEA used a digital library of shark records. The authors of this new paper say this library is like a photo album where someone glued the wrong captions to the pictures.

• The "Maybe" Mistake: In science, when a fossil looks sort of like a specific shark but isn't 100% sure, experts label it "Maybe Shark X" (using terms like cf. or aff.). Team GEA's computer just ignored the "Maybe" and decided, "Okay, it's definitely Shark X!"
  • Analogy: Imagine you see a blurry photo of a dog that looks a bit like a Golden Retriever. Instead of saying "It's probably a dog," you label it "This is definitely a Golden Retriever named Buster." If you do this for 1,800 blurry photos, you end up with a fake list of 1,800 Golden Retrievers that never actually existed.
• The "Ghost" Records: About 39% of the records in Team GEA's database were "ghosts." These were entries with no photos, no specimen numbers, and no descriptions. They were just names on a list.
  • Analogy: It's like counting the population of a city by including people who are just rumors. You can't verify they exist, so they shouldn't be in the census.
• The Wrong Dates: Some fossils were dated incorrectly. A shark that lived before the asteroid hit was accidentally labeled as living after it.
  • Analogy: It's like finding a photo of a person in a 1920s flapper dress and accidentally labeling it as a photo from the 2020s. Suddenly, it looks like they lived for 100 years!

The Result: Because of these errors, the computer thought many sharks that should have died out with the dinosaurs actually survived. This made the "extinction" look tiny (only 10% loss) when it was actually massive.

2. The "Wrong Math" Problem (Methodology)

Even if the data were perfect, the authors argue that Team GEA used the wrong way to do the math.

• The "Headcount" vs. "Turnover" Analogy:
  • Team GEA just counted how many sharks were in the "Before" box and how many were in the "After" box, then subtracted the two.
  • Analogy: Imagine a classroom. In the morning, there are 20 students. In the afternoon, there are 15. Team GEA says, "Oh, only 5 students left!"
  • The Reality: What actually happened is that 15 students left, but 10 new ones arrived. The net change is -5, but the extinction was actually 75% (15 out of 20).
  • By ignoring the "new arrivals" (new species evolving) and the "departures" (extinctions), Team GEA's method hid the true scale of the disaster.

3. The Real Story

When the authors of this paper cleaned up the "photo album" (fixed the labels, removed the ghosts, and corrected the dates) and used the right math, the story changed completely:

• The Extinction: Instead of a tiny 10% loss, the shark family actually lost over 90% of its species during the dinosaur-killing asteroid event.
• The Recovery: Sharks didn't just "bounce back" easily; they had to rebuild their family tree from almost scratch.
• The Recent Decline: The idea that shark numbers are naturally dropping today is likely an illusion caused by the fact that we haven't found as many recent fossils yet, not because the sharks are naturally dying out.

The Big Takeaway

This paper is a warning to scientists: You cannot trust a computer to tell you the history of life if you feed it a messy, unverified dataset and use a simple calculator.

Just because a high-tech AI says something is true doesn't mean it is. You still need human experts (taxonomists) to look at the fossils, check the dates, and make sure the names are right. If we don't do this, we might think sharks are tougher than they really are, which could be dangerous for conservation efforts today. We need to know the truth about how fragile they really are.

Technical Summary

Technical Summary: Critique of Fossil-Based Diversification Analyses in Neoselachians

Problem Statement
The paper addresses a critical discrepancy in macroevolutionary studies regarding neoselachian (modern sharks and rays) diversification. Specifically, it critiques a recent study by Gardiner et al. (GEA), which utilized a deep-learning model and a large fossil-occurrence dataset to argue that neoselachian diversity was minimally impacted by the Cretaceous-Paleogene (K/Pg) mass extinction (~10% species loss) and has been declining since the mid-Eocene. The authors argue that these findings contradict the established scientific consensus, which posits a significant extinction event at the K/Pg boundary followed by a delayed recovery. The core problem identified is that GEA's conclusions stem from a combination of unverified, taxonomically incorrect fossil data and inappropriate analytical methodologies.

Methodology
The authors employed a multi-faceted approach to re-evaluate GEA's findings:

1. Data Re-examination: They scrutinized GEA's fossil-occurrence dataset (derived from the Paleobiology Database and Shark-References), focusing specifically on taxa crossing the K/Pg boundary.
2. Taxonomic Verification: They manually verified the taxonomy and stratigraphic age of "survivor" and "victim" taxa. This involved:
   • Identifying and correcting erroneous assignments of uncertain nomenclature (e.g., cf., aff., "?" designations) which GEA had automatically assigned to specific species.
   • Filtering out unverified occurrences lacking specimen numbers, illustrations, or descriptions.
   • Correcting outdated taxonomy and misattributions of genera to higher taxonomic levels.
3. Stratigraphic Correction: They identified errors in age assignments for specific localities (e.g., mixing Maastrichtian and Danian layers due to bioturbation) that artificially created "survivor" lineages.
4. Methodological Critique: They analyzed GEA's statistical approach, contrasting GEA's method of calculating "net change in species richness" against conventional extinction magnitude calculations (extinct species / total richness).
5. Re-calculation: Using the corrected dataset, they recalculated extinction magnitudes using both GEA's flawed methodology and conventional Bayesian-style metrics.

Key Contributions

• Identification of Systemic Data Flaws: The paper exposes that the GEA dataset contains severe errors, including:
  • 39% unverified occurrences: Nearly 10,000 records lacked evidence, making identification impossible.
  • 7.4% misassigned uncertain nomenclature: Tentative identifications were forced into specific species, creating artificial temporal ranges.
  • Taxonomic and Age Errors: Outdated names, incorrect higher-level classifications, and erroneous age assignments (e.g., treating mixed-age horizons as continuous) artificially inflated the number of K/Pg survivors.
• Methodological Critique: The authors demonstrate that GEA's reliance on "net richness change" between time bins is a naive metric that fails to capture true extinction dynamics. They argue that without modeling speciation and extinction rates, richness trajectories are uninformative regarding evolutionary processes.
• Correction of the K/Pg Narrative: By cleaning the data and applying standard extinction metrics, the authors provide a corrected view of the K/Pg event's impact on neoselachians.

Results

• Data Verification: Of the 81 species identified as K/Pg survivors in the GEA dataset, 83.95% (68 species) were found to be restricted to either the Paleogene or the Maastrichtian, meaning they did not actually survive the boundary.
• Impact on Extinction Estimates:
  • GEA's Original Claim: ~10% species loss (9.7% ± 9.1%).
  • Raw Data Re-calculation (using GEA's flawed data but conventional method): 64.32% extinction.
  • Corrected Data Re-calculation: After removing erroneous occurrences and fixing taxonomy, the extinction magnitude rises to 93.53% for species (and 65.29% for genera).
• Contradiction of GEA's Trends: The authors confirm that the perceived decline in diversity in the recent past is an artifact of poor fossil record availability and identification difficulties, not a biological reality.

Significance

• Scientific Integrity: The paper serves as a stark warning against the uncritical use of automated deep-learning models on fossil datasets without rigorous taxonomic expertise and manual verification. It highlights that "big data" in paleontology is only as reliable as the curation behind it.
• Conservation Implications: By underestimating the severity of the K/Pg extinction, GEA inadvertently suggested that neoselachians are resilient to mass extinction events. This paper corrects that narrative, emphasizing that these clades have historically suffered catastrophic losses, which is crucial for understanding their current vulnerability and resilience in the face of modern anthropogenic threats.
• Methodological Standard: The authors advocate for the integration of traditional taxonomic expertise with modern analytical tools, asserting that appropriate methodology is non-negotiable for accurate macroevolutionary inference. They strongly discourage the use of the GEA dataset for future diversification analyses until these fundamental errors are resolved.