The Signal Is the Scatter

In one line: In 1943 Salvador Luria and Max Delbrück proved that bacteria mutate by blind chance — before they ever meet the threat, not in response to it — without watching a single mutation happen, by noticing that if mutations arise at random during growth the colony counts across parallel cultures must scatter wildly, and then finding exactly that scatter.

---

The world then

Pour a deadly bacteriophage onto a lawn of bacteria and nearly all of them die — but a few survive, and their offspring are resistant forever. It looked as though the virus had taught the bacteria to resist: an adaptation summoned by the threat itself. A stranger idea was on the table too — that resistance was a random mutation already present in a few cells before the virus ever arrived, and the virus merely exposed it. The trouble was that the survivors look identical either way. The phenotype alone could not decide between "induced adaptation" and "pre-existing mutation."

The question

Does the virus create the resistance, or merely reveal a mutation that was already there? You cannot tell from the survivors, and — this is the crux — you cannot tell from their number either. On average, both pictures predict the very same count of resistant cells. The two hypotheses agree on the mean. Where they disagree is on the variance: how much the counts jump around from one culture to the next.

The design

Luria's insight came, famously, at a faculty dance in February 1943, watching a slot machine pay out a jackpot: don't count the average — watch the scatter. The two hypotheses make opposite predictions about it.

• If the virus induces resistance, each cell takes its own independent long shot at the moment of exposure. Grow many separate cultures and each should turn up about the same small number of survivors — counts clustered tightly around the mean, a Poisson spread where the variance roughly equals the mean.
• If resistance is a spontaneous mutation that strikes at random as the culture grows — before any virus — then timing sets the count. A mutation late in growth leaves only a few resistant cells; a mutation early gets copied, and copied again, into an enormous resistant clone — a jackpot. The distribution is wildly skewed: most cultures with a handful of survivors, a rare few with hundreds.

The test almost runs itself. Grow many tiny cultures side by side, each on its own, with no virus at all. Let them grow up. Then pour each whole culture onto a virus plate and count the resistant colonies. Two pictures, two predictions: tight and even — or wildly uneven, with jackpots. This is the fluctuation test.

The result

The cultures did not agree. Most gave a few resistant colonies, but some erupted into jackpots hundreds strong. The scatter between cultures was far larger than the mean — far beyond the tight, even spread the "induced" idea demanded (for which variance would merely equal the mean). The jackpots gave it away: the mutations had already been there, arising blindly as the cells divided, long before the virus arrived. Crucially, this was a statistical inference from the shape of the fluctuation, not a direct sighting of a pre-existing mutant — that direct confirmation came later, from Newcombe's spreading experiment (1949) and the Lederbergs' replica plating (1952).

What we learned, and why it's beautiful

Mutations arise on their own — at random, without purpose, and before any need for them. The virus doesn't teach resistance; it only exposes the mutants that chance had already made. In one stroke the result showed that bacteria have genes and mutate like everything else, founding the field of bacterial genetics. The beauty is that it measured the invisible. No one saw a single mutation happen. Yet by reading pure chance — the shape of the scatter, not any one number — Luria and Delbrück caught the timing of unseen events, and turned a question that sounded like philosophy into a fact you could count.

Sources

Full claim-by-claim evidence is in references.md. Primary anchors:

• Luria, S. E. & Delbrück, M. "Mutations of Bacteria from Virus Sensitivity to Virus Resistance." Genetics 28(6), 491–511 (1943). — the fluctuation test.
• Newcombe, H. B. "Origin of Bacterial Variants." Nature 164, 150 (1949). — direct confirmation by spreading.
• Lederberg, J. & Lederberg, E. M. "Replica Plating and Indirect Selection of Bacterial Mutants." J. Bacteriol. 63(3), 399–406 (1952). — direct confirmation by replica plating.
• Luria, S. E. A Slot Machine, a Broken Test Tube: An Autobiography. Harper & Row (1984). — the slot-machine origin.
• The Nobel Prize in Physiology or Medicine 1969 (Delbrück, Hershey, Luria).

Accuracy notes: The two hypotheses predict the same mean number of resistant cells; only the variance/scatter across parallel cultures decides between them. For the "induced" (Poisson) null the benchmark is variance ≈ mean — the mutation signature is variance ≫ mean, with jackpots. The 1943 result is a statistical inference from that fluctuation, not a direct observation of pre-existing mutants; the direct sightings came from Newcombe (1949) and Lederberg replica plating (1952). Scope matters: this shows that phage-resistance mutations are random with respect to selection and pre-exist it — it is not a blanket disproof of all directed change. Host organism was E. coli strain B and the selective agent bacteriophage α (T1). Finally, the 1969 Nobel Prize went to Delbrück, Alfred Hershey and Luria "for their discoveries concerning the replication mechanism and the genetic structure of viruses" — the broader phage-genetics program, not the fluctuation test alone.