Robust evidence for modest diversity loss across the K/Pg in neoselachians: Response to Guinot et al.

This paper defends its findings of a modest K/Pg diversity decline in neoselachians against Guinot et al.'s criticisms by demonstrating that the original conclusions remain robust even when applying the critics' data treatments and sensitivity tests.

The Gist

Imagine the history of sharks (specifically a group called "neoselachians") as a long, winding road trip spanning 145 million years. Recently, a team of scientists (the authors of this paper) took a look at the map and said, "Hey, there was a small bump in the road during a major event called the K/Pg extinction (the asteroid impact that killed the dinosaurs), but the sharks didn't crash; they just slowed down a bit and kept driving."

Another group of scientists, Guinot et al., looked at the same map and shouted, "No way! That bump was a massive cliff! The sharks almost went extinct right then and there!" They claimed the first team's map was drawn with bad data and wrong methods.

This new paper is the first team's response. They are essentially saying: "We checked your complaints, we redrew the map using your rules, and guess what? It still looks like a small bump, not a cliff."

Here is a breakdown of their argument using simple analogies:

1. The "Recipe" Dispute (The Data)

Guinot et al. argued that the first team's "recipe" for counting shark fossils was flawed. They said, "You included too many questionable ingredients! Some of your fossil records don't have pictures, and some names are spelled weirdly. You should throw 40% of your data in the trash."

• The Authors' Reply: "We didn't throw anything away because we were lazy; we kept those records because they are like receipts. Just because a receipt doesn't have a photo of the burger doesn't mean the burger wasn't eaten. In paleontology, we need to count every single report of a shark ever found to get the full picture. Throwing away 40% of the data is like trying to count the crowd at a concert by only looking at the VIP section and ignoring the general admission."

2. The "Stress Test" (The Sensitivity Tests)

To prove they weren't just stubborn, the authors decided to play along. They took their data and applied Guinot's strict rules (throwing away the "questionable" records). They ran three different "stress tests" to see if the result would change.

• Test 1: They removed the records with weird names.
• Test 2: They built a special computer model to focus specifically on the extinction moment.
• Test 3: They removed everything Guinot complained about, including the records without pictures (the most extreme version).

The Result: Even with the most extreme data removal, the sharks didn't show a "catastrophic collapse." The diversity dropped, yes, but it was a modest dip, not a total wipeout. It's like running a car through a mud pit, a snowstorm, and a rock garden; the car gets dirty and slows down, but the engine doesn't blow up.

3. The "Turnover" Analogy (Why the numbers look different)

Guinot et al. focused only on how many sharks died. The authors argue that looking only at deaths is like judging a busy restaurant only by how many people leave the door.

• The Reality: While many sharks did die during the asteroid impact, new sharks were being born (or evolving) at an incredibly fast rate at the same time.
• The Metaphor: Imagine a crowded room where 10 people leave, but 12 new people walk in immediately. If you only count the people leaving, it looks like a disaster. But if you count the total number of people in the room, the crowd size barely changed. The authors found that high "birth rates" buffered the "death rates," keeping the shark population stable enough to survive the crisis.

4. The "Goalposts" Argument

The authors feel Guinot et al. are moving the goalposts. Guinot claims that simply counting diversity over time is a bad way to study evolution. The authors counter that this method has been the "gold standard" for decades and has helped us understand the rise and fall of dinosaurs, mammals, and plants.

They argue that while no single method is perfect, throwing out massive amounts of data just to fit a specific theory is dangerous. It's like saying, "We can't trust the weather forecast because we only looked at the sunny days," when in reality, we need to look at the rain, the clouds, and the sun to get the real picture.

The Bottom Line

This paper is a defense of robust science. The authors are saying:

1. We used a massive, transparent dataset.
2. We checked your criticisms and found they were mostly about how we organized the data, not the data itself.
3. Even if we use your strict rules, the conclusion stays the same: Sharks suffered a modest hit during the dinosaur extinction, but they didn't collapse.
4. To understand the past, we need to embrace all the evidence we have, not just the "perfect" pieces, and use smart math to sort out the noise.

In short: The sharks survived the asteroid impact with a few scratches, not a broken spine.

Technical Summary

1. Problem Statement

This paper serves as a direct rebuttal to a critique by Guinot et al. regarding a previous study (Gardiner et al.) on neoselachian (modern sharks and rays) diversity trajectories.

• The Conflict: Guinot et al. argued that the original findings—specifically the recovery of a modest decline in diversity across the Cretaceous-Paleogene (K/Pg) boundary—were flawed due to issues with the underlying fossil occurrence dataset and the methodology used to quantify extinction magnitude.
• The Critique: Guinot et al. suggested that the data contained significant errors (e.g., uncertain nomenclature, lack of evidence) and that the original analysis failed to account for these, potentially masking a catastrophic diversity collapse. They also questioned the validity of using diversity trajectory reconstruction as a primary method for understanding deep-time evolution.
• The Authors' Stance: Gardiner et al. contend that the alleged data issues are largely operational choices rather than fundamental errors, and that their original conclusion of a moderate decline (rather than a collapse) remains robust even under stricter data treatments.

2. Methodology

The authors employed a multi-pronged approach to audit the critique and re-evaluate their results:

• Dataset Audit (FINS): They reviewed their "FINS" dataset (a transparent synthesis of neoselachian fossil occurrences) against the specific criticisms raised by Guinot et al. They categorized the alleged issues into:
  • Operational choices: e.g., inclusion of records "without evidence" (no illustrations/descriptions) or uncertain nomenclature.
  • Valid corrections: ~70% of the suggested corrections were deemed valid (minor taxonomic harmonizations).
  • Unfounded claims: ~30% of the suggested corrections were proven incorrect upon verification.
• Sensitivity Testing: To rigorously test the robustness of their original findings, they applied three distinct sensitivity tests using the DeepDive framework (a probabilistic database system for integrating data and modeling uncertainty):
  1. Test 1 (Nomenclature Filter): Removed all occurrences with uncertain nomenclature (removing ~8% of species-level and ~1% of genus-level occurrences).
  2. Test 2 (Explicit K/Pg Modeling): Trained the DeepDive model to explicitly model the K/Pg extinction event.
  3. Test 3 (Restrictive Filter): Applied the Test 1 filter plus the removal of all occurrences classified as "without evidence" (removing ~36% of species-level and ~39% of genus-level occurrences).
• Comparative Analysis: They compared the resulting diversity trajectories and extinction magnitudes from these tests against their original estimates and against independent studies using different datasets and methods (e.g., genus-level analyses and PyRate diversification rate modeling).

3. Key Contributions

• Validation of Data Integrity: The authors demonstrated that the "flaws" identified by Guinot et al. were largely matters of operational philosophy (e.g., whether to include unillustrated records) rather than data errors. They showed that excluding such data risks discarding a significant portion of the paleontological record (up to 38% of occurrences).
• Methodological Defense: The paper defends the use of diversity trajectory reconstruction as a valid macroevolutionary tool, arguing that while speciation/extinction rates can be non-identifiable, diversity trajectories remain key biological indicators.
• Robustness Proof: By applying the most restrictive data filters proposed by the critics, the authors proved that their core conclusion (modest decline vs. catastrophic collapse) holds true regardless of data treatment.

4. Key Results

The sensitivity tests yielded the following quantitative results regarding diversity change across the K/Pg boundary:

• Species-Level Analysis:

  • Original Estimate: -10% ± 9%.
  • Test 1 (Uncertain Nomenclature removed): -22.9% ± 1.1%.
  • Test 2 (Explicit K/Pg modeling): -29.7% ± 8.4%.
  • Test 3 (Most restrictive, "without evidence" removed): -33.1% ± 3.9%.
  • Interpretation: While the magnitude of decline increased slightly under stricter filters, it consistently remained a moderate decline, never approaching a catastrophic collapse.

• Genus-Level Analysis:

  • Original Estimate: +15% ± 4% (increase).
  • Test 1: +7.3% ± 5.7%.
  • Test 2: +7.2% ± 4.7%.
  • Test 3: -3.5% ± 3.7% (a small decline, but statistically consistent with zero change as the interval spans zero).
  • Interpretation: Even under the most restrictive conditions, the genus-level data does not support a significant diversity collapse.

• Synthesis: The results align with two other recent independent studies (using different datasets and genus-level data) and PyRate diversification analyses, which suggest that while extinction rates were high, they were buffered by elevated origination rates, leading to rapid turnover rather than net diversity loss.

5. Significance

• Re-evaluating the K/Pg Event for Neoselachians: The study challenges the narrative of a "mass extinction" for sharks and rays at the K/Pg boundary. Instead, it supports a model of resilience and rapid turnover, where high origination rates compensated for elevated extinction, resulting in only a modest net diversity loss.
• Philosophy of Paleobiological Data: The paper advocates for a "maximalist" approach to fossil data integration. It argues that excluding large portions of the fossil record (e.g., records without illustrations) to satisfy strict criteria can introduce bias and limit the scope of macroevolutionary research.
• Methodological Rigor: It demonstrates that modern statistical frameworks (like DeepDive) can effectively handle data uncertainty and that sensitivity analyses are crucial for validating paleontological conclusions against critiques.
• Future Direction: The authors conclude that the field advances by integrating the richness of the available fossil record with sophisticated quantitative methods that explicitly evaluate uncertainty, rather than by discarding data that does not meet arbitrary strictness thresholds.