6 ideas
Research can reproduce every visible ritual of science (apparatus, jargon, publications, statistics) while missing the essential ingredient, so the expected results never arrive. Like islanders building bamboo control towers and wooden headphones to summon cargo planes, practitioners follow the outward precepts but lack the self-critical integrity that makes the method actually work.
The first principle of science is not fooling yourself, because your own hopes, prior beliefs, and investment in a result bias you more subtly than any outside deception. Once you have stopped fooling yourself, being honest with others follows fairly easily. The hard discipline is internal.
Scientific integrity means bending over backwards to publish everything that might make your result wrong: other explanations you've ruled out and how, details that could cast doubt, and data that don't fit. Presenting only the evidence that supports a conclusion is advertising, not science. The listener needs the full picture to judge the claim for themselves.
After Millikan measured the electron's charge with a slightly wrong value, later measurements drifted upward gradually rather than jumping to the correct number. Researchers who got results far from Millikan's looked hard for errors and threw them out, but accepted results close to his without scrutiny. This shows how asymmetric error-hunting lets a field fool itself collectively.
In 1937 a researcher found that rats running a maze were cueing on subtle factors: the smell of the food, the sounds of the floor, and the light. He controlled for each of these until the rats truly had to learn the maze. Later rat researchers ignored his work and kept running uncontrolled experiments, because his findings made their experiments harder rather than generating new results.
Researchers rarely repeat a prior experiment before building on it, because funders and institutions reward novel findings, not confirmation. Without first reproducing the baseline condition, a new variation's results cannot be interpreted, so incentive structures quietly erode scientific validity. Feynman urged graduates to seek positions where they are free to keep this integrity even at the cost of funding or status.