Fiddlers on the proof

Some scientists use questionable practices when it comes to interpreting their research and proving its validity.

The scientist Robert Millikan won a Nobel Prize for his research into the electron - findings that were later found to be flowed.
Powered by automated translation

Dodgy counting, fiddled results, outright fraud: we may expect little else from financiers and politicians, but now scientists have been shown to be up to the same tricks. A controversial new study has found that as many as one in three scientists admits to questionable research practices, while one in 50 confesses to the ultimate sin of fiddling their results. With opinion polls routinely putting scientists among the most trusted members of society, the findings - published in the online science journal PLoS ONE - cast a pall over the public image of scientists as dispassionate seekers after truth. Yet anyone who's actually done scientific research is likely to agree with the study's author Dr Daniele Fanelli of the University of Edinburgh, Scotland, that the figures are probably on the low side.

Even those whose involvement in science ended in the school lab will know that only the very simplest experiments give neat and tidy outcomes. All too often messy data "outliers" ruin an otherwise nice, smooth graph or clear conclusion. Within the profession itself, there is a standard joke about how "representative data" presented at conferences really means "hand-picked stuff that backs our conclusions".

It is a joke that highlights a reality that researchers quickly discover during their career: success sometimes demands an unwholesome mix of gut instinct and cold logic. Some scientists have paid a heavy price for playing strictly by the rules. Take the case of Felix Ehrenhaft, a physicist at the University of Vienna who carried out pioneering studies of the properties of the electron. As so often with experiments, Ehrenhaft's results were a mess: sometimes they hinted that the electron is the basic unit of electrostatic charge, and sometimes that the electron isn't fundamental at all. But like a good scientist, Ehrenhaft ignored his own prejudices and focused on gathering the data to reveal the truth.

Unfortunately, he was not alone in his quest. At the University of Chicago, physicist Robert Millikan was attempting the same measurements, and getting similarly messy results. But he had a much simpler way of dealing with them: he just ignored them. His data notebooks are littered with remarks like "Very low - something wrong" and "This is almost exactly right". Precisely how Millikan could tell which data were right isn't clear - after all, the whole point of the experiment was to establish the correct value. What is clear is his determination to get results that were both self-consistent and confirmed his view that the electron really is a fundamental particle.

As a strategy for convincing other scientists, it paid off handsomely. While Ehrenhaft struggled to get a consistent result out of his experiment, Millikan published his neat and tidy result - and won a Nobel Prize. Only decades after his death did his questionable practices come to light, along with the fact that his results weren't actually correct. Perhaps most disturbing of all is the way other researchers confirmed Millikan's results until after the great man's death, at which point they began to shift towards today's value - which is incompatible with his own.

Was Millikan a scientific crook? He was certainly ambitious: he persuaded his young assistant to give him sole credit for the electron studies in return for a nice reference. On the other hand, Millikan was also capable of setting aside his own prejudices: he later provided definitive experimental support for Einstein's theories, despite his scepticism about them. Similar concerns surround the work of the father of modern genetics, the Austrian monk Gregor Mendel. In 1865 he unveiled the results of plant breeding experiments apparently proving that their traits followed strict mathematical laws. These are now known to be a direct result of the existence of genes - "packets" of inheritable information that are passed down the generations.

As with Millikan, after Mendel's death questions began to be asked about his astonishingly consistent raw data. In 1936, the renowned geneticist and statistician Ronald Fisher of the University of London argued that the match between Mendel's results and genetic theory was simply too good to be true. According to Fisher, someone must have altered the raw data: he suggested that one of Mendel's assistants may have fiddled the results to please his master.

Not everyone suspects foul play, however. Some statisticians have claimed that Fisher's analysis is flawed, while others have pointed out that Mendel may have fallen prey to unconscious bias while analysing his results. His experiment involved adding up the relative numbers of different types of hybrid plant - a tedious, error-prone task. Yet if Mendel deemed only results that disagreed with his theory to be worth rechecking, he could easily have ended up with amazingly consistent results.

Some of Mendel's defenders have a much simpler explanation: his results agreed with his theory because his theory was right. It's hard to credit such post hoc reasoning - and not a little dangerous. Imagine, say, medical scientists deciding they "know" that some new treatment works, and simply ignored evidence to the contrary - resulting in useless or even harmful drugs getting onto the market. The good news is that medical researchers have led the field in trying to stamp this out. New drugs are routinely tested in "blinded" clinical trials, where the researchers aren't allowed to know which patients received the new drug until after the results are in.

The bad news is that the system isn't foolproof: it is often possible to work out which patients are getting the new drug, and fiddle the results. And the even worse news comes from Dr Fanelli's new study of scientific misconduct: the highest reported rates of malpractice appear to be in clinical medicine. As Dr Fanelli points out, there is a charitable explanation for this: medical scientists may just be more willing than others to report misconduct by colleagues. That certainly seems more plausible than the alternative, with lab-coated fraudsters fiddling the result of drugs trials. After all, unlike electrons or hybrid plants, dodgy data on drugs can lead to the deaths of thousands of innocent people. And no scientist would countenance that - would they?

Robert Matthews is a Visiting Reader in Science at Aston University, Birmingham, England