The quote that Edison probably never said this way
“I have not failed. I’ve just found 10,000 ways that won’t work.” The line appears on posters, mugs and slide decks, always credited to Thomas Edison. The trail behind it is thinner than the posters suggest. The earliest version that quote researchers have found appears in a 1910 biography by Frank Dyer and Thomas Martin. It describes a visitor asking about the long search for a better storage battery, and Edison answering that he had gotten “a lot of results” and knew “several thousand things that won’t work.” That account itself rests on a witness’s memory. The tidy figure of 10,000 shows up in sources only much later, in a 1982 talk. So we should treat the famous wording as a paraphrase of a real attitude, not a verified quotation.
The deeper problem with the poster version is what it implies: that failure is automatically valuable if you tolerate enough of it. Case histories suggest something more specific. Failures contributed to invention when someone was set up to learn from them.
Case 1: failure on purpose, with instruments
In 1901 Wilbur and Orville Wright took a larger glider to Kitty Hawk, and it performed badly. According to the Smithsonian’s National Air and Space Museum, it produced roughly one-third of the lift their calculations predicted, and its controls behaved erratically; the brothers were “utterly baffled.” They did not simply try again. They suspected the published aerodynamic data they had relied on, built a small bicycle-mounted device and then a wind tunnel, and measured lift and drag themselves. They ran preliminary tests on as many as 200 model wing shapes and recorded formal data on nearly 50. Among other things, they recalculated a long-used air-pressure coefficient, 0.005, and arrived at about 0.0033.
Their 1902 glider, designed from those results, reached full three-axis control, and they logged hundreds of glides. The lesson is not that the 1901 failure was “good.” It is that they turned a bad result into numbers, and numbers could be compared. They also relied on others, including a mechanic named Charlie Taylor, who helped design the engine and did nearly all of the machine work.
- State the expectation. The Wrights had a predicted lift. Without a prediction, a miss is only a disappointment.
- Measure the gap. They built instruments to turn “it didn’t work” into numbers.
- Question the inputs. They checked the borrowed data, not just their own work.
- Change one thing at a time. Wing shapes were varied in a controlled way.
- Validate at full scale. A new glider tested whether model results held up.
Case 2: the “lucky” mould
Penicillin is often told as a story of pure luck. In 1928 Alexander Fleming noticed that a Penicillium mould contaminating a bacterial culture had stopped bacteria growing nearby. A spore had probably drifted into a dish that was left on a bench rather than in the incubator, which let the mould grow. That chance observation was real, and noticing it took a trained eye. But Fleming could not isolate the active substance, and in the 1930s it stayed a laboratory curiosity.
It became a medicine only after years of further work. At Oxford in the late 1930s, a team led by Howard Florey and including Ernst Chain and Norman Heatley found ways to purify it. A small mouse experiment in 1940 showed protection against infection, and in 1941 a policeman named Albert Alexander, severely ill with an infection, improved when treated but died a month later after the drug ran out. Researchers in the United States then improved yields by deep fermentation, and by 1943 there was enough for Allied forces. Fleming, Florey and Chain shared the 1945 Nobel Prize. The “accident” was the start of a story, and most of it was patient, institutional work, much of it dealing with setbacks such as scarcity and instability.
Case 3: the glue with no use
A similar thing happened at 3M. In 1968 the chemist Spencer Silver set out to make a strong adhesive and produced a weak, reusable one. For about five years he sought a use for it. Then Art Fry, a colleague who wanted a better bookmark for his church hymnal, saw that it could make paper slightly sticky without tearing. Sources differ on whether Fry’s idea came in 1973 or 1974. Others at the company, including Henry Courtney and Roger Merrill, developed ways to coat single sheets. Early market tests in 1977 gave mixed results, and a free-sample campaign in Boise, Idaho, in 1978 changed the picture; the notes reached stores nationally in 1980.
It is fair to say the adhesive was an unplanned result of a different aim, and that it looked like a failure for years. It is less fair to say it was “an accident” and nothing else, because turning it into a product took a person who connected it to a problem, colleagues who built a process, and a company willing to keep testing.
The failures we never hear about
Every example above comes from a project that eventually worked. That is survivorship bias in action: we tell stories about failures that led somewhere, and we rarely hear about the vastly larger number of dead ends that led nowhere, left no records or hurt people. Failures also have costs that fall unevenly: a well-funded laboratory can afford thousands of experiments, while a small business or a patient in a trial may not. The glamorous phrase “fail fast” can overlook that.
For related reading, how humans learn to do what once seemed impossible looks at how feedback shapes skill, and the hidden teams behind famous breakthroughs shows how many people stand behind any single success.
