In a lecture in 1963 the Nobel laureate biologist Peter Medawar asked “Is the scientific paper a fraud?” He did not mean to imply that scientists were making up their results (although occasionally that happens). Rather, Medawar pointed out that published papers invariably present a curated story – typically one in which the authors present a hypothesis or theory and then show that their experiments support or contradict that idea.
It’s rarely thus, Medawar argued. Scientific hypotheses often evolve as the work proceeds, and “they are imaginative and inspirational in character”, not the result of dry logic. Historians studying the lab notebooks in which great scientific discoveries were recorded often find that the stories the celebrated discoverers tell in later life don’t reflect what really happened but are a tidied-up, romanticised version. What we don’t tend to be told about in papers or in memoirs are the mistakes, the false leads, the experiments that didn’t work.
It’s not terribly edifying, perhaps, to hear of failed equipment, broken test tubes, or experiments that were simply too messy and puzzling to permit any interpretation. But the result is that science becomes seen as a triumphant march of success and discovery, rather than – as all researchers know – often a litany of failure and frustration.
In this sense, then, a recent meeting held at Imperial College London on “failure in science” was breaching a taboo. There are many ways to fail, of which the notion of a “failed experiment” is a familiar but also a complicated one. Does an experiment fail when it doesn’t produce the result you expected or hoped for – perhaps showing that your theory is wrong? According to the philosopher Karl Popper, it’s precisely through such “failures” that science progresses: falsification of hypotheses by experiment is the only route to reliable knowledge.
Although Popperian falsification is fetishised by scientists, philosophers of science today recognise that’s only a part of how science works. Still, one systemic problem is that negative results – showing that some theory, technique, or drug doesn’t work as hoped – don’t tend to get published as easily as positive ones that announced new discoveries. It’s just not as exciting. There are consequences, though. We might get a distorted picture of the evidence for and against a drug’s efficacy, for example (pharmaceutical companies are notoriously quiet about drug trials that fail). Time and resources get wasted by researchers trying things that others have already shown won’t work. And the media rarely reports on work disproving some flashy finding that made headlines last year.
Failure in science should be a normal part of how it works, but gets disincentivised, even punished. Significant breakthroughs often come from “high-risk” research, which sets ambitious targets or examines bold new ideas. But by definition such research should have a high failure rate. When failed research is seen as a black mark, scientists are likely to retreat into making small, timid steps, promising only what they know already they can deliver. Grant agencies aren’t terribly good at supporting high-risk projects: assessment panels might conclude that the chances of success are too slender and give the money instead to a safer bet. It’s not obvious how to crack that problem – how do you know whether an idea is bold or just bonkers? But unless we try, genuine innovation and paradigm-shifting discovery might become more rare, as some have suggested is already the case.
Suggested Reading
Hantavirus: the disease with no vaccine
For individual scientists, especially in the early stages of their career, fear of failure can be overwhelming. Often it’s a failure to secure funding. Some agencies now reject 90% or so of the applications they receive, many of which are likely to be perfectly good. Getting used to grant rejections is often seen as a necessary part of the training, but it can be demoralising when it happens repeatedly, and hugely wasteful of the time and effort involved in assembling a team and writing the application. But arguably even more debilitating is the fear of a “failed” paper shown by later work to be flawed, of a poorly received talk, of not knowing the answer, of being scooped by rivals, of criticism from peers or senior researchers.
All this works to reduce the efficiency of the research enterprise, not least because it discourages some young scientists from taking risks or staying in the game at all. As an editorial in Nature (which supported the meeting) put it, “There must be a better way to learn from failure, and not to see it as a burden – to make failure a normal part of the scientific process. That begins with talking about it.”
