On February 24, 1988, Richard Lenski at Michigan State University inoculated twelve flasks with identical populations of Escherichia coli. Thirty-seven years have now passed. In 2024, the experiment surpassed 80,000 generations — the equivalent of roughly 1.6 million years of human time.
Richard Dawkins called it "a beautiful demonstration of evolution in action" in The Greatest Show on Earth.2 Jerry Coyne called it "a poke in the eye for creationists."3 A Veritasium video with tens of millions of views declared it "one of the most direct demonstrations of Darwinian adaptation by natural selection you can imagine." Amateur bloggers announced that macroevolution had been proven and mocked creationists for their silence.
So after 80,000 generations, what has E. coli become?
"E. coli."
1. Let's Correct the Misunderstanding — Creationism Predicted This
Evolutionists frequently say that Lenski's experiment makes creationists uncomfortable. This is not true. What makes creationists uncomfortable is not the experiment itself. It is the exaggerated interpretation placed upon it.
Creationism predicts microevolution — variation and adaptation within existing genetic information — from the outset. If a Creator designed organisms with the flexibility to adapt to changing environments, it follows naturally that E. coli would optimize for a glucose-limited laboratory flask. What creationists contest is not the existence of natural selection. What they contest is the claim that natural selection can generate entirely new functional information from nothing.
For macroevolution — from single cell to human — to be explained, one must observe the spontaneous emergence of new protein folds, new biochemical pathways, and new developmental programs. Lenski's experiment has failed to demonstrate any of these. That is the crux of the debate.
Creationists accepted microevolution and natural selection before Lenski's experiment, not as a concession but as a prediction of their model. The real question is whether microevolution can be legitimately extrapolated to macroevolution.
2. LTEE Timeline: What Actually Happened Over 37 Years
12 E. coli populations, glucose-limited minimal medium. Daily 1% serial transfer. Frozen samples every 500 generations.
Increased cell size, faster growth, adjusted DNA supercoiling. All 12 populations show the same changes — all within existing regulatory range.
In half the populations, DNA repair enzymes are damaged, causing mutation rates to skyrocket. Evolutionists call this "adaptation." It is more accurately described as a loss of function.
(reported 2008)
One of 12 populations evolves the ability to feed on citrate under aerobic conditions. Lenski calls it a "presumptive speciation event." Dawkins and Coyne celebrate.
Same Cit+ phenotype reproduced in as few as 12 generations under direct selection. Fellow evolutionary scientists acknowledge the need to "reinterpret" Lenski's claims.
As Lenski prepares to retire, the experiment is handed to former postdoc Jeffrey Barrick at UT Austin.
All-time peak fitness recorded. "Proto-gene" transcription events identified in LTEE genomes — but function remains unconfirmed.
Barrick appointed Hannah Distinguished Professor at Michigan State. LTEE returns to Michigan. Experiment continues.
3. Evaluating the Observed Changes Honestly
Increased Cell Size
All 12 populations produced larger cells. Larger cells can accommodate more ribosomes, accelerating protein synthesis before division. But the regulation of cell size is an existing capability of E. coli. Expression levels of existing genes were adjusted. No new gene was created. Interestingly, a 2022 PNAS study found that cell size and metabolic efficiency "decoupled" in unexpected ways — an exception to known biological scaling laws, not the emergence of new information.
Improved Growth Rate and Diminishing Returns
Fitness improved rapidly at first, then decelerated. The "all-time peak fitness" recorded in 2024 represents fine-tuning within a highly constrained laboratory environment. No new metabolic pathway was generated.
Hypermutator Lineages
In six of twelve populations, DNA repair enzymes were damaged, causing mutation rates to increase hundreds-fold. This is described as "adaptive" because it generates more potentially beneficial mutations in the short term. But the damage of a genome surveillance system is not a mechanism for generating new information. It is the accelerated consumption of existing information.
Almost every beneficial mutation observed in the LTEE involves the loss or relaxation of existing gene function — switching off metabolic pathways that are unnecessary in the glucose-only environment. How does this relate to the generation of the new functional information that macroevolution requires?
4. Dissecting the Citrate Affair — The Most Famous "Evidence"
The LTEE's most celebrated finding is the acquisition of aerobic citrate metabolism at approximately generation 31,500. A population of E. coli — which cannot normally use citrate as a carbon source under aerobic conditions — evolved this capability. Lenski described it as a "key innovation" and a "presumptive speciation event."
Three facts must be faced directly.
Fact 1: E. coli Already Metabolizes Citrate
E. coli already uses citrate as an energy source under anaerobic conditions. The TCA cycle, of which citrate is an intermediate, is a core metabolic pathway. What changed is that the suppression of the citrate transporter gene (citT) under aerobic conditions was lifted. No new gene was created. The Lenski team's own paper describes it as a change in gene regulation or minor modification of protein structure.4
"The switch was not built from scratch. An existing switch was turned on."
Fact 2: It Happens in 12 Generations — Van Hofwegen (2016)
In 2016, Van Hofwegen, Hovde, and Minnich at the University of Idaho published a direct rebuttal in the Journal of Bacteriology.6 When direct selection pressure was applied — with citrate as the sole carbon source — the identical Cit+ mutant appeared in as few as 12 generations. Forty-six independent mutants were produced. Genomic sequencing confirmed the same mechanism in every case: promoter capture of the existing citT gene.
The reason 31,500 generations were required in Lenski's experiment was not that this mutation is rare or difficult. It was that in a glucose-rich environment, there was no selection pressure to favor citrate utilization. Is that "a rare, innovative evolutionary event worthy of celebrating as incipient speciation"?
Fact 3: Without citT, It Never Happens
Van Hofwegen's team confirmed that in strains from which the citT gene had been deleted, the Cit+ phenotype never appeared under any condition and at any time point. This is decisive. The phenomenon is a change in the regulation of an existing gene. It occurred entirely within the existing design.
John Roth and Sophie Maisnier-Patin of UC Davis published a companion commentary acknowledging that Lenski's "historical contingency" interpretation may require reinterpretation.7 The creationist analysis was confirmed by evolutionary scientists themselves.
Lenski's Rebuttal and Its Limits
Lenski pushed back: "We already knew it could happen quickly. The experimental conditions were different." But this rebuttal undermines his original claim. Once he concedes that it happens in 12 generations when conditions are changed, the original characterization of "a rare, complex, historically contingent innovation" no longer stands.
5. The 2024 Proto-Gene Discovery — A Real Breakthrough?
A 2024 analysis of LTEE genomic data (Uz-Zaman et al., PLOS Biology) identified cases where previously unexpressed DNA regions had begun to be transcribed and translated.8 These were called "proto-genes." Some evolutionists described the finding as "evidence of de novo gene birth."
The finding is genuinely interesting. But an honest assessment shows the following:
| Criterion | Evolutionist Claim | Actual Data |
|---|---|---|
| New gene created? | "Evidence of de novo gene birth" | ✗ Existing promoters recruited |
| New regulatory info? | "Novel expression patterns" | ✗ Repositioning of existing sequences |
| Stable expression? | "Fixed in populations" | ✓ Confirmed |
| Function confirmed? | "Potential substrates for new genes" | ✗ Paper acknowledges "unknown" |
| Net increase in information? | "De novo gene origination" | ✗ Rearrangement of existing information |
A gene requires two things: stable expression and a beneficial function. This study demonstrated only the first. Scribbling a few notes on a page is not the birth of a symphony.
A 2025 PNAS study (Chihoub et al.) found that as E. coli adapts, its robustness to mutations in its current environment increases — but its fragility in novel environments also increases.9 Greater specialization comes at the cost of versatility. This demonstrates optimization within a constrained design, not the generation of new information.
6. The Silence of 80,000 Generations — What Evolution Must Still Answer
We acknowledge honestly what was observed in Lenski's experiment. But what matters even more is what was not observed.
Macroevolution — from single cell to human — requires the following:
- A new protein fold — not one in 80,000 generations
- A new biochemical pathway arising spontaneously — none
- A new cellular organelle — none
- A fully novel functional gene unrelated to pre-existing genes — none
- The beginning of multicellular organization — none
- A new gene regulatory network — none
- Increased cell size — within existing regulatory range
- Faster growth rate — optimization of existing pathways
- Aerobic citrate metabolism — regulatory change in an existing gene
- Proto-gene transcription — function unconfirmed
Thirty-seven years. The most controlled conditions possible. The fastest-reproducing organism available. And this is what the record shows. This is not silence. This is testimony.
7. The Information Problem Cannot Be Avoided
Here evolutionists raise a counter-argument: "The standard of 'net increase in functional information' is an ID-creationist construct. We don't accept that framing."
Even taking this objection seriously, the problem does not disappear. Evolutionary theory itself claims to explain "the origin of complex biological structures." Complex structures consist of molecules arranged in specific sequences. Explaining how those specific arrangements arose is precisely what evolutionary theory is supposed to do. Whether we call it "functional information" or "specified molecular sequence," the phenomenon requiring explanation is the same.
According to the experimental work of Douglas Axe, the probability of a 150-amino-acid functional protein arising by random processes is approximately 10-74. This exceeds by dozens of orders of magnitude the threshold set by statistical physicist Émile Borel — events with probability below 10-50 do not occur at the cosmic scale.
"As far as we know from experience, functionally specified complex information always originates from intelligence. Those who claim DNA is an exception bear the burden of proof."
Changing the vocabulary does not change what needs to be explained. And the LTEE — the most ambitious empirical test of evolutionary mechanisms ever attempted — produced 80,000 generations of record showing that no new protein structure and no new developmental program appeared. Not one.
Conclusion: The Experiment Continues — The Question Is Sharper Than Ever
Lenski's experiment continues. Barrick has returned it to Michigan. Perhaps something genuinely new will emerge someday, they say. But is that expectation a scientific prediction, or a statement of faith?
What 80,000 generations have made clear is this: natural selection filters existing information. Environmental pressure activates existing capability. But in all of that, entirely new functional information was never generated from nothing. Not once.
Evolutionists promoted the LTEE as "the most direct evidence for evolution." In doing so, they created the clearest record possible. Thirty-seven years. The most favorable conditions. The fastest organism. And the conclusion: no new information appeared.
Creationism predicted: "Natural selection can optimize within existing information but cannot generate new functional information." The LTEE across 80,000 generations is precisely consistent with that prediction. What thirty-seven years of experiment have taught us is this: the information inside E. coli was there from the beginning.
E. coli is still E. coli.