How a caution became a cudgel
In the early 1950s, epidemiologists on both sides of the Atlantic produced striking findings. Richard Doll and Austin Bradford Hill in the United Kingdom, and Ernst Wynder and Evarts Graham in the United States, reported that lung cancer patients were overwhelmingly smokers. The associations were large, consistent, and biologically plausible, pointing squarely at tobacco.
Fisher, already a towering figure in statistics, urged restraint. He argued that an observed association might be driven by hidden variables, reverse causality, or sheer coincidence—powerful reminders in an era when randomized trials were becoming the gold standard. He proposed a “constitutional hypothesis,” speculating that a genetic predisposition might simultaneously incline some people to smoke and make them more susceptible to cancer.
It was a tidy, testable alternative—and a critical intellectual service. Science advances when credible hypotheses compete. But cigarettes were not a candidate for randomized trials, and as new evidence mounted, the balance tipped decisively. Prospective cohort studies, particularly the decade-spanning British Doctors Study, revealed stark dose-response patterns and declining risks after quitting—signatures much harder to explain away.
“Cigarette smoking is causally related to lung cancer in men; the magnitude of the effect of cigarette smoking far outweighs all other factors.”
U.S. Surgeon General’s Advisory Committee on Smoking and Health, 1964
When doubt becomes a strategy
Fisher’s statistical cautions were real and important. Yet the tobacco industry recognized that uncertainty—no matter how slender—could be weaponized. At minimum, it bought time. At maximum, it reshaped public perception by turning a scientific debate into a rhetorical fog.
“Doubt is our product since it is the best means of competing with the ‘body of fact’ that exists in the mind of the general public.”
Brown & Williamson internal memo, 1969
That blunt corporate strategy did not require disproving the evidence; it only had to keep the conversation stuck on the word imply. Fisher, who publicly disputed the smoking-cancer link and published “Smoking: The Cancer Controversy” in 1959, brought formidable authority to the table. His arguments, motivated by a commitment to rigorous inference and, critics noted, colored by his own affection for tobacco and work with industry groups, helped make “correlation does not imply causation” a common refrain.
There is a lesson here that transcends personalities. The mantra is a vital guardrail in science. But stripped of context and wielded as a catchphrase, it can morph from a call for careful reasoning into a license for permanent doubt. The difference lies in what comes next: investigation or inertia.
How researchers proved a cause without a trial
Proving that smoking causes cancer demanded more than a single correlation and far more than a clever hypothesis. Epidemiologists built a cumulative case using multiple lines of evidence that, together, formed a coherent narrative. They showed temporality (smoking preceded disease), dose-response (heavier smoking meant higher risk), consistency (across populations, designs, and decades), coherence with biology (tobacco smoke contains carcinogens), and reversibility (risk drops after quitting).
In 1965, Austin Bradford Hill articulated a set of pragmatic “viewpoints” to evaluate when an association might reasonably be judged causal. His framework—now taught to generations of students—was an answer to exactly the kind of impasse Fisher highlighted, offering principles for judgment in the real world where randomized trials are not always possible.
“None of my nine viewpoints can bring indisputable evidence for or against the cause-and-effect hypothesis and none can be required as a sine qua non.”
Austin Bradford Hill, 1965
Hill’s point remains profoundly modern: causation in public health is rarely a single dramatic experiment; it is a mosaic. By the late 1960s, the tiles had been set. Laboratory studies demonstrated how tobacco-specific nitrosamines damage DNA. Large cohorts followed smokers and non-smokers for years, measuring risks with precision. Pathologists saw tumor patterns that mapped to exposure. No plausible third variable could simultaneously explain the breadth and depth of the evidence.
What the maxim should—and should not—do
“Correlation does not imply causation” should stop us from leaping to conclusions, not from looking for answers. It is an invitation to probe mechanisms, rule out confounding, and stress-test ideas with better designs and new data. In practice, that means asking whether the effect precedes the outcome, whether more exposure leads to more effect, whether multiple studies agree, and whether a credible mechanism connects the dots.
Modern causal inference has expanded those tools. Directed acyclic graphs help researchers make assumptions explicit and identify which variables to control for—and which to leave alone. Natural experiments, instrumental variables, and regression discontinuities can isolate causal signals where randomized trials cannot. The ethic is not reflexive doubt; it is disciplined curiosity.
The smoking saga also warns how easily sound skepticism can be co-opted. Industries facing inconvenient evidence—from fossil fuels to ultra-processed foods—have borrowed the tobacco playbook, spotlighting uncertainties while downplaying the weight of the whole. Responsible skepticism asks, “What evidence would change our minds?” Motivated skepticism asks, “What uncertainty can we highlight?” The difference is not semantic; it is societal.
“It is difficult to get a man to understand something, when his salary depends upon his not understanding it.”
Upton Sinclair
Fisher’s legacy, reclaimed with nuance
R. A. Fisher helped invent the statistical machinery that underpins modern science. His caution about causation—like David Hume’s philosophical doubts two centuries earlier—forced researchers to earn their conclusions. He was also, in the case of smoking, wrong.
Both truths can stand. The job of science is not to put geniuses on pedestals or to wield slogans as shields. It is to keep asking better questions, tightening the inferences, and refusing to confuse the absence of perfect proof with the absence of proof.
Today, as debates rage over everything from vaping to climate risks to nutrition advice, we could do worse than remember the full story behind those five famous words. Correlation is a flashlight in a dark room. What matters is whether we use it to find the switch—or to argue we cannot see.