Science Puzzle
The Cure That Only Works on Mice
A team reports that a compound shrinks tumours in mice. The result is real, well-conducted, and replicated. The newspaper headline reads: "Scientists find cure for cancer."
The research itself is sound. Where exactly does the reasoning break, and why do the overwhelming majority of such compounds never become treatments?
The Answer
The headline has extrapolated far outside the range of what was tested. The study establishes one thing: this compound shrinks these tumours in these mice under these conditions. Every word of that is doing work, and the headline has quietly deleted all of it.
A finding is valid for the population and conditions actually studied. Extending it further is a new claim requiring new evidence, and the gap here is enormous. Mice are not small humans: they differ in metabolism, immune response, lifespan, and dosing, and their tumours are usually induced deliberately and are genetically uniform in a way that real human cancers are not. Roughly ninety percent of compounds that work in animals fail in human trials, which is not a scandal but the ordinary attrition of the process working as intended.
The generalisation trap is not confined to species. It bites whenever results are stretched past the tested range: a drug trialled on adults being assumed safe in children, a dosage tested from 1 to 10 mg being assumed to keep scaling at 100 mg, a psychology result from undergraduates being assumed to describe humanity. Each of those is the same move, which is why it is worth naming as a single error rather than a list of separate mistakes.
Notice also what the option about sample size is doing. More mice would make the mouse result more certain. It would do absolutely nothing to make the result apply to humans, because that is a question about which population you studied, not about how many of them. This is a distinction worth holding onto: precision within a population and generalisability across populations are different problems, and quantity only ever helps with the first.
The habit to build is to read every finding with its conditions stapled to it. Not "compound X shrinks tumours" but "compound X shrank induced tumours in inbred mice at this dose over this period." Once you are in the practice of restoring the missing clause, the leap in the headline becomes visible immediately, and so does the amount of work still standing between a promising result and a treatment.
The principle: Extrapolation and generalisability. A result holds for the population and conditions actually tested. Extending it to a different species, dose range, or group is a new claim needing new evidence, and more data within the original population does not help.