In the first post of this series, we introduced a provocative idea: modern science is not just struggling — it is evolving in a direction that selects for weak, unreliable methods. This idea, central to Smaldino & McElreath’s influential 2016 paper The Natural Selection of Bad Science, rests on a powerful metaphor:

Scientific labs behave like organisms in an evolutionary system.
The methods they use are traits.
Their students are their progeny.
Publications are their fitness.

This metaphor isn’t just poetic. It is a rigorous conceptual framework allowing us to explain why poor research practices spread even when individuals have good intentions.

This post explores that evolutionary lens:

Why does academia behave like an ecosystem?
How do methods “reproduce”?
What is the unit of selection?
How do labs evolve over time?
Why do certain practices become dominant while others vanish?

We will also use stories from the history of science — from Darwin’s own notebooks to Thomas Kuhn to modern lab politics — to illustrate how culture, training, and incentives act as evolutionary forces.

1. The Central Analogy: Labs as Lineages

Smaldino & McElreath begin with a simple but profound observation:

Scientific practice is not created anew by each generation. It is inherited.

A PhD student or postdoc absorbs:

their advisor’s habits,
their lab’s methodological norms,
statistical preferences,
attitudes toward replication,
openness to data sharing,
norms about p-values,
and even meta-level beliefs about what “counts” as good science.

When these trainees move on to establish their own labs, they carry those inherited traits with them, modifying them slightly, recombining them with influences from other labs, but largely preserving the lineage.

This is cultural evolution — a well-studied field — but applied here to scientific methodology.

Examples of inherited scientific culture

The Cold Spring Harbor molecular biology lineage, which proliferated through shared summer courses and collaborative DNA work.
The Copenhagen School of quantum mechanics, where Bohr’s philosophical stance became a transmittable “method” for thinking about physics.
The Chicago School of economics, where rational-choice modeling spread through mentorship and institutional prestige.

Students didn’t just learn theories — they inherited methods, priorities, and epistemic values.

2. What Exactly Is Being Selected? “Traits” in Scientific Lineages

Traits = Research practices.

Examples include:

Sample sizes
Statistical thresholds
Willingness to preregister
Commitment to replication
How aggressively a lab chases significant results
The balance between carefulness and productivity
Whether negative results are ever written up
How much time is spent refining experiments

These traits are not innate; they are learned.

And crucially — some traits boost short-term success at the expense of long-term reliability.

This immediately sets up a tension:

Rigor is slow and costly.
Speed produces more publishable results.
Publishing more results increases grant success.
Therefore, speed boosts evolutionary fitness, even if it lowers rigor.

This is exactly the type of trade-off natural selection thrives on.

3. The Unit of Selection: The Lab, Not the Individual

Scientists often think of academic success as individual —
X scientists wins awards, publishes papers, secures grants.

But the evolutionary view shifts the focus:

The lab or research group is the unit of selection.

Why?

Because:

Labs recruit students.
Those students carry the lab’s practices elsewhere.
Successful labs breed more “descendants.”
Practices are copied and transmitted through mentoring lineages.

A lab that produces many successful students spreads its methods faster.
A lab that produces few students leaves little evolutionary footprint.

This is why certain laboratory cultures — good or bad — propagate with surprising persistence.

A historical anecdote

In early molecular biology, the Watson–Crick style of rapid, intuitive model-building spread widely, while Linus Pauling's more hierarchical and chemistry-heavy style faded.
Not because the former was inherently better, but because the labs carrying it produced more trainees in a rapidly expanding field.

Methods spread because trainees spread.

4. Selection Pressure: Publication Counts as Fitness

In biological evolution, fitness = reproductive success.
In academic evolution, fitness = publication success.

The publication record determines:

Who gets grants
Who gets tenure
Whose students get jobs
Who attracts new students
Which labs grow, split, and reproduce

Thus, labs with traits that maximize publication numbers reproduce more successfully in the academic ecosystem.

These traits may include:

running many small-N studies
chasing p < 0.05 results
favoring novelty over accuracy
avoiding replications
presenting exploratory findings as confirmatory
inflating claims
streamlining the path to publication

In other words, questionable research practices (QRPs) increase fitness.

This is not a moral accusation.

It is an evolutionary prediction.

5. Cultural Evolution in Action: Famous Examples

Example 1: Mendel vs. Fisher — Methodological Divergence

Mendel’s experiments are sometimes criticized for being too perfect.
Fisher famously suggested low-variance results indicated bias or over-tidying of data.

But more interesting is how Mendel’s meticulous methods did not spread.
His successors were not trained in his exacting style, and the field evolved into very different methodological norms.

Why?
Because the evolutionary environment around genetics changed.
Speed of discovery mattered more than perfection.

Example 2: The Replication Crisis in Psychology

For decades, psychology labs that produced:

surprising effects
clever paradigms
small studies
publishable results

…thrived.
These labs trained many students.

Meanwhile, labs that insisted on:

high power
robust replication
slow, careful experimentation

…produced fewer papers and trained fewer students.

Over time, the field evolved toward flashy, unreliable results.

Example 3: Biomedical “breakthrough culture”

Preclinical cancer biology is notorious for irreproducibility.
Amgen reported in 2012 that they could reproduce only 6 of 53 “landmark” studies.

Why?
Because the labs producing “breakthroughs” got the funding.
They reproduced themselves.
Their methods spread.

Labs doing slow, confirmatory research did not grow.

Evolutionary selection at work.

6. Why Good Practices Often Lose

In evolution, “good” = survival-enhancing, not morally good.

In academia:

High-power studies (good for truth)
require money, time, effort.
Low-power studies (bad for truth)
allow more experiments → more publications.

Thus:

Low power has higher fitness than high power.

This single insight explains much of the replication crisis.

Bad methods win not because they are bad, but because they:

are cheaper,
produce publishable results faster,
generate more “offspring labs.”

Rigor is selected against.

This parallels biological evolution:

Peacocks evolve burdensome tails because sexual selection rewards flashiness.
Labs evolve burdensome statistical habits because academic selection rewards flashiness.

7. Transmission: How Practices Spread Through Academic Pedigrees

Humans are cultural animals.
We copy behaviors with high social payoffs.
Science is no different.

The main transmission pathways:

Advisor → student
Collaborator → collaborator
Postdoc → new institution
Hiring committee → new faculty (selecting for “productive” candidates)
Grant panel → funded lab

This is similar to the transmission of:

languages
tool use in primates
religious norms
social rituals
craft techniques

Scientific methodology is a cultural artifact.

8. Drift, Mutation, Selection — All Present in Science

Smaldino & McElreath’s insight allows us to map biological evolutionary features directly onto academia.

Mutation: method innovations

Statistical innovations (Bayesian models, preregistration) are mutations.
Some spread, some die off.

Drift: accidental shifts

A charismatic advisor or influential journal editor can cause random swings in norms, independent of method quality.

Selection: survival of the most publishable

Traits that maximize output proliferate.

9. A Closer Look: Why Evolutionary Thinking Helps Us Understand Scientific Decline

The evolutionary perspective clarifies several mysteries:

❓ Why do bad practices persist even though everyone agrees they are harmful?

➡ Because they increase fitness under current incentives.

❓ Why don’t reforms (e.g., “use bigger samples”) stick?

➡ Because selection pressure overrides idealistic norms.

❓ Why do some labs produce generations of similarly unreliable work?

➡ Because success breeds reproduction of method lineages.

❓ Why do fields diverge so much in reliability?

➡ Because each field has a different ecological niche:

neuroscientists face expensive data collection → chronic low power
social psychologists have flexible experiments → high p-hacking incentives
physics has strong mathematical constraints → low methodological drift

Evolutionary environments differ.

10. Conclusion: Science Evolves — but Who Directs That Evolution?

Viewing science as an evolutionary system reveals something uncomfortable:

We have created a selection environment in which poor methods thrive.

This doesn’t mean the scientists themselves are bad.
It means the system rewards the wrong traits.

Until we redesign the incentive structure, science will continue to evolve toward:

more questionable practices
more flashy but unreliable findings
lower average rigor
higher false positive rates
declining public trust

In the next post, we’ll dive into the mathematical and computational models that Smaldino & McElreath use to demonstrate how bad science wins under current selection pressures.

Nagarjun's blog

Saturday, January 31, 2026

Post 2 — Evolutionary Theory Meets Academia: How Selection Shapes Scientific Methods