Science Fictions: How Fraud, Bias, Negligence, and Hype Undermine the Search for Truth

by Stuart Ritchie

It is the peculiar and perpetual error of the human understanding to be more moved and excited by affirmatives than by negatives. Francis Bacon, Novum Organum (1620)

For a scientific finding to be worth taking seriously, it can’t be something that occurred because of random chance, or a glitch in the equipment, or because the scientist was cheating or dissembling. It has to have really happened. And if it did, then in principle I should be able to go out and find broadly the same results as yours. In many ways, that’s the essence of science, and something that sets it apart from other ways of knowing about the world: if it won’t replicate, then it’s hard to describe what you’ve done as scientific at all.

Perhaps Bem himself best summed up many scientists’ attitudes to replication, in an interview some years after his infamous study. ‘I’m all for rigor,’ he said, ‘but I don’t have the patience for it … If you looked at all my past experiments, they were always rhetorical devices. I gathered data to show how my point would be made. I used data as a point of persuasion, and I never really worried about, “Will this replicate or will this not?”’13 Worrying about whether results will replicate or not isn’t optional. It’s the basic spirit of science; a spirit that’s supposed to be made manifest in the system of peer review and journal publication, which acts as a bulwark against false findings, mistaken experiments and dodgy data. As this book will show, though, that system is badly broken.

Other books feature scientists taking the fight to a rogue’s gallery of pseudoscientists: creationists, homeopaths, flat-Earthers, astrologers, and their ilk, who misunderstand and abuse science – usually unwittingly, sometimes maliciously, and always irresponsibly.14 This book is different. It reveals a deep corruption within science itself: a corruption that affects the very culture in which research is practised and published. Science, the discipline in which we should find the harshest scepticism, the most pin-sharp rationality and the hardest-headed empiricism, has become home to a dizzying array of incompetence, delusion, lies and self-deception. In the process, the central purpose of science – to find our way ever closer to the truth – is being undermined.

doing science involves much more than just running experiments or testing hypotheses. Science is inherently a social thing, where you have to convince other people – other scientists – of what you’ve found. And since science is also a human thing, we know that any scientists will be prone to human characteristics, such as irrationality, biases, lapses in attention, in-group favouritism and outright cheating to get what they want.

Much can be learned from mistakes. On one of his albums, the musician Todd Rundgren has a spoken-word introduction encouraging the listener to play a little game he calls ‘Sounds of the Studio’. Rundgren describes all the missteps that can be made when recording music: hums, hisses, pops on the microphone whenever someone sings a word with a ‘p’ in it, choppy editing, and so on. He suggests that the reader listen for these mistakes in the songs that follow, and on other records. And just as a better understanding of recording studio slip-ups can give you a new insight into how music is made, learning about how science goes wrong can tell you a lot about the process by which we arrive at our knowledge.

Most people, including scientists, assume peer review has always been a crucial feature of scientific publication, but its history is more complicated. Although in the seventeenth century the Royal Society tended to ask some of its members whether they thought a paper was interesting enough to publish in Philosophical Transactions, requiring them to provide a written evaluation of each study wasn’t tried until at least 1831.15 Even then, the formal peer review system we know today didn’t become universal until well into the twentieth century (as you can tell from a letter Albert Einstein sent in 1936 to the editors of Physical Review, huffily announcing that he was withdrawing his paper from consideration at their journal because they had dared to send it to another physicist for comment).16 It took until the 1970s for all journals to adopt the modern model of sending out submissions to independent experts for peer review, giving them the gatekeeping role they have today.17 Peer reviewers are usually anonymous, which is both a blessing and a curse: a blessing because it allows them to speak their minds without concern about repercussions from the scientists whose work they’re criticising (a junior scientist can be truly honest about the flaws of a big-name professor’s work), but a curse because, well, it allows them to speak their minds without concern about repercussions from the scientists whose work they’re criticising. The following are genuine excerpts from peer revie...

What ensures that the participants in the process just described – the researcher who submits the paper, the editor at the journal, the peers who review it – all conduct themselves with the honesty and integrity that trustworthy science requires? There’s no law requiring that everyone acts fairly and rationally when evaluating science, so what’s needed is a shared ethos, a set of values that aligns the scientists’ behaviour.19 The best-known attempt to write down these unwritten rules is that of the sociologist Robert Merton. In 1942, Merton set out four scientific values, now known as the ‘Mertonian Norms’. None of them have snappy names, but all of them are good aspirations for scientists. First, universalism: scientific knowledge is scientific knowledge, no matter who comes up with it – so long as their methods for finding that knowledge are sound. The race, sex, age, gender, sexuality, income, social background, nationality, popularity, or any other status of a scientist should have no bearing on how their factual claims are assessed. You also can’t judge someone’s research based on what a pleasant or unpleasant person they are – which should come as a relief for some of my more disagreeable colleagues. Second, and relatedly, disinterestedness: scientists aren’t in it for the money, for political or ideological reasons, or to enhance their own ego or reputation (or the reputation of their university, country, or anything else). They’re in it to advance our understandin...

Replication, then, has long been a key part of how science is supposed to work – and incidentally, it’s another of its social aspects, with results only being taken seriously after they’ve been corroborated by multiple observers. But somewhere along the way, between Boyle and modern academia, a great many scientists forgot about the importance of replication. In the collision of our Mertonian ideals with the realities of the scientific publication system – not to mention the realities of human nature – the ideals have proven the more fragile, leaving us with a scientific literature full of untrustworthy, unreliable, unreplicable studies that often do more to confuse than enlighten.

studies we saw above).32 In general, though, the effect on psychology has been devastating. This wasn’t just a case of fluffy, flashy research like priming and power posing being debunked: a great deal of far more ‘serious’ psychological research (like the Stanford Prison Experiment, and much else besides) was also thrown into doubt. And neither was it a matter of digging up some irrelevant antiques and performatively showing that they were bad – like when Pope Stephen VI, in the year 897, exhumed the corpse of one of his predecessors, Pope Formosus, and put it on trial (it was found guilty). The studies that failed to replicate continued to be routinely cited both by scientists and other writers: entire lines of research, and bestselling popular books, were being built on their foundation. ‘Crisis’ seems to be an apt description.

In economics, a miserable 0.1 per cent of all articles published were attempted replications of prior results; in psychology, the number was better, but still nowhere near good, with an attempted replication rate of just over 1 per cent.40 If everyone is constantly marching onwards to new findings without stopping to check if our previous knowledge is robust, is the above list of replication failures that much of a surprise?

You’d think that if you obtained the exact same dataset as was used in a published study, you’d be able to derive the exact same results that the study reported. Unfortunately, in many subjects, researchers have had terrible difficulty with this seemingly straightforward task. This is a problem sometimes described as a question of reproducibility, as opposed to replicability (the latter term being usually reserved to mean studies that ask the same questions of different data). How is it possible that some results can’t even be reproduced? Sometimes it’s due to errors in the original study. Other times, the original scientists weren’t clear enough with reporting their analysis: they took twists and turns with the statistics that weren’t declared in their scientific paper, and thus their exact steps can’t be retraced by independent researchers. When new scientists run the statistics in their own way, the results come out differently. Those studies are like a cookbook including mouth-watering photographs of meals but providing only the patchiest details of the ingredients and recipe needed to produce them.

Although we’ve seen poor replicability in some weightier areas, such as economic theory, how much difference can it make to our lives if a bunch of academics end up disagreeing over whether power posing works, or whether alpha-male sparrows have more black patches of feathers? There are two responses to make. The first is that there’s a wider principle at stake here: science is crucial to our society, and we mustn’t let any of it be compromised by low-quality, unreplicable studies. If we let standards slip in any part of science, we risk tarnishing the reputation of the scientific enterprise more generally. The second response is to look at a scientific field we haven’t yet considered, where the immediate consequences of a lack of replicability are indisputable. That field is, of course, medical research.

a more comprehensive study took a random sample of 268 biomedical papers, including clinical trials, and found that all but one of them failed to report their full protocol. Meaning that, once again, you’d need additional details beyond the paper even to try to replicate the study.53 Another analysis found that 54 per cent of biomedical studies didn’t even fully describe what kind of animals, chemicals or cells they used in their experiment.54 Let’s take a moment to think about how odd this is. If a paper only provides a superficial description of a study, with necessary details only appearing after months of emailing with the authors (or possibly being lost forever), what was the point of writing it in the first place? Going back at least to Robert Boyle in the seventeenth century, recall that the original, fundamental motivation for scientists reporting all the specifics of their studies was so that others could scrutinise, and try to replicate, their research. The papers here failed that elementary test, just as the journals that published them also failed to perform their basic critical function.

Replication problems in medicine haven’t only affected laboratory based, preclinical research: they can have direct effects on the treatments doctors give to their patients. Common treatments, it turns out, are often based on low-quality research; instead of being solidly grounded in evidence, accepted medical wisdom is instead regularly contradicted by new studies. This phenomenon occurs so frequently that the medical scientists Vinay Prasad and Adam Cifu have a name for it: ‘medical reversal’.57

It’s understandable that doctors, and those who write the guidelines for medical treatments, sometimes find themselves relying on low-quality evidence. Often the alternative is no evidence at all, and their job is to help patients who need treatment right now. And it’s inevitable that advances in technology, methodology, and funding allow scientists to do better research today than was possible a few years ago: that’s normal scientific progress. But scientists have let doctors and patients down by creating such a constant state of flux in the medical literature, running and publishing poor-quality studies that even students in undergraduate classes on research design would recognise as inadequate. Even at the time many of these original studies were published, we knew how to do better – and yet we didn’t.

A 2016 survey of over 1,500 researchers – though admittedly not a properly representative one, since it just involved those who filled in a questionnaire on the website of the journal Nature – found that 52 per cent thought there was a ‘significant crisis’ of replicability. A further 38 per cent believed there was, in a somewhat peculiar turn of phrase, at least a ‘slight crisis’.69 Nearly 90 per cent of chemists said that they’d had the experience of failing to replicate another researcher’s result; nearly 80 per cent of biologists said the same, as did almost 70 per cent of physicists, engineers, and medical scientists. Only a slightly lower percentage of scientists said they’d had trouble in replicating their own results. Because the study wasn’t a proper poll, and scientists who were already worried about replicability were probably more likely to complete it, these numbers will be somewhat exaggerated. But they do tell us that there are widespread concerns about just how much of the scientific literature, including even the research we’ve done ourselves, we can trust.

What led to such a drawn-out process, where even the horrific, painful deaths of multiple patients hadn’t moved the university to fire, or even censure, the man responsible? Indeed, what led to the university lashing out against the very people who blew the whistle on the falsified results?33 Macchiarini clearly used his ‘breakthrough’ operations to burnish his reputation and fame. But he was also an asset to the Karolinska Institute and its planned international expansion. It has been suggested that the university’s association with the superstar surgeon helped to grease the wheels of its new regenerative medicine centre that was being set up in Hong Kong.34 Sheer panic and embarrassment were likely also a factor in their wanting to cover up the case: a university with as high a standing as the Karolinska would hardly be keen to admit to themselves, let alone the public, that they’d let a lethally dangerous fraudster loose on vulnerable patients.

The fact that the scientific community so proudly cherishes an image of itself as objective and scrupulously honest – a system where fraud is complete anathema – might, perversely, be what prevents it from spotting the bad actors in its midst. The very idea of villains like Macchiarini existing in science is so abhorrent that many adopt a see-no-evil attitude, overlooking even the most glaring signs of scientific misconduct. Others are in denial about the prevalence and the effects of fraud. But as we’ll see in this chapter, fraud in science is not the vanishingly rare scenario that we desperately hope it to be. In fact, it’s a distressingly common one.

Both measurement error and sampling error are unpredictable, but they’re predictably unpredictable. You can always expect data from different samples, measures or groups to have somewhat different characteristics – in terms of the averages, the highest and lowest scores, and practically everything else. So even though they’re normally a nuisance, measurement error and sampling error can be useful as a means of spotting fraudulent data. If a dataset looks too neat, too tidily similar across different groups, something strange might be afoot. As the geneticist J. B. S. Haldane put it, ‘man is an orderly animal’ who ‘finds it very hard to imitate the disorder of nature’, and that goes for fraudsters as much as for the rest of us.68

The immunologist and Nobelist Sir Peter Medawar has argued, perhaps counter-intuitively, that scientists who commit fraud care too much about the truth, but that their idea of what’s true has become disconnected from reality. ‘I believe,’ he wrote, ‘that the most important incentive to scientific fraud is a passionate belief in the truth and significance of a theory or hypothesis which is disregarded or frankly not believed by the majority of scientists – colleagues who must accordingly be shocked into recognition of what the offending scientist believes to be a self-evident truth.’103 The physicist David Goodstein agrees: ‘Injecting falsehoods into the body of science is rarely, if ever, the purpose of those who perpetrate fraud,’ he suggests. ‘They almost always believe that they are injecting a truth into the scientific record … but without going through all the trouble that the real scientific method demands.’104

Particularly interesting for our hunt for the motives of fraudsters, though, is what comes at the end of the lengthy investigation report: a short reply from Schön himself. He pleads that, although he ‘made mistakes’ and ‘realise[s] there is a lack of credibility’ he also ‘truly believe[s] that the reported scientific effects are real, exciting, and worth working for’.110 We should of course be extremely wary about taking fraudsters at their word. But Schön’s apparent faith in his theory, even after so much of his work had been expunged from the scientific literature (he’s currently fifteenth on the Retraction Watch Leaderboard, with thirty-two retractions) speaks to the kind of delusion and self-deception that Medawar and Goodstein noted above.111 It’s impossible to know for sure, but Schön might genuinely have believed in his incredible transistors and seen his rule-breaking as a necessary evil in his quest to bring them to the world’s attention.

The obesity researcher Eric Poehlman, for example, the first scientific fraudster to be jailed in America, wasted millions of dollars of taxpayer money in grants from the US government in producing a decade’s worth of useless, fabricated data.114 And how many perfectly innocent scientists have wasted their own research grant money trying to follow or replicate his work, or that of other fraudsters? Aside from the waste, fraud has a terribly demoralising effect on scientists.

The acceptance of Wakefield’s study in as prominent an outlet as the Lancet will go down as one of the worst decisions in the history of scientific publication. There could hardly be a more concrete example of the importance of reliable science for society’s wellbeing, and of the failure of the peer-review system to screen out bad research. It brings us back once again to the issue of trust – this time, the public’s trust in science. Having your child vaccinated is an act of commission: you’re actively having something done to them, trusting that the medics are right that it’s safe.143 If a study in a famous journal with the seal of approval of scientific peer review implies that it’s not, it’s only rational to take notice. For years after the publication, people genuinely didn’t know who to trust about vaccines. Many still don’t.144

An adopted hypothesis gives us lynx-eyes for everything that confirms it and makes us blind to everything that contradicts it. Arthur Schopenhauer, The World as Will and Presentation (1818)

Science … commits suicide when it adopts a creed. T. H. Huxley, ‘The Darwin Memorial’ (1885)

There’s an age-old philosophical question that goes: ‘Why is there something rather than nothing?’ We can pose a similar query about the scientific process: ‘Why do studies always find something rather than nothing?’

There’s a straightforward, but devastating, reason for this persistent positivity: scientists choose whether to publish studies based on their results. In a perfect world, the methodology of a study would be all that matters: if everyone agrees it’s a good test of its hypothesis, from a well-designed piece of research, it gets published. This would be a true expression of the Mertonian norm of disinterestedness, where scientists are supposed to care not about their specific results (the very idea of having a ‘pet theory’ is an affront to this norm) but just the rigour with which they’re investigated. That’s far from the reality, however. Results that support a theory are written up and submitted to journals with a flourish; disappointing ‘failures’ (which is how null results are often seen) are quietly dropped, the scientists moving on to the next study. And it isn’t just the researchers themselves: journal editors and reviewers also make the decision to accept and publish papers according to how interesting the findings appear, not necessarily how meticulous the researchers have been in discovering them. This feeds back to the scientists themselves and the whole thing becomes a vicious circle: why bother submitting your null paper for publication if it has a negligible chance of being accepted? This is publication bias. It’s also known, now anachronistically, as the ‘file-drawer problem’ – since that’s where scientists are said to be keeping all their null results, hidden...

Despite being one of the most commonly used statistics in science, the p-value has a notoriously tricky definition. A recent audit found that a stunning 89 per cent of a sample of introductory psychology textbooks got the definition wrong; I’ll try to avoid making the same mistake here.16 The p-value is the probability that your results would look the way they look, or would seem to show an even bigger effect, if the effect you’re interested in weren’t actually present.17 Notably, the p-value doesn’t tell you the probability that your result is true (whatever that might mean), nor how important it is. It just answers the question: ‘in a world where your hypothesis isn’t true, how likely is it that pure noise would give you results like the ones you have, or ones with an even larger effect?’18

Outside of exceptions such as the Higgs Boson, though, the 0.05 threshold remains, through conformity, tradition and inertia, the most widely used criterion today. It has scientists feverishly rifling through their statistical tables, checking for p-values lower than 0.05 so that they can report their results as being statistically significant. It’s easy to forget the arbitrariness. Richard Dawkins has bemoaned the ‘discontinuous mind’: our human tendency to think in terms of distinct, sharply defined categories rather than the messy, blurry, ambiguous way the world really is.25 One example is the debate around abortion, which often fixates on the question of when an embryo or foetus becomes ‘a person’, as if there could ever be a bright line by which we could make that decision. In Dawkins’s own field of evolutionary biology, pinpointing the exact moment when one species evolves into another is likewise a fool’s errand, no matter how satisfied we might feel if we could do so. It’s the same for p-values: the 0.05 cut-off for statistical significance encourages researchers to think of results below it as somehow ‘real’, and those above it as hopelessly ‘null’. But 0.05 is as much a convention as the 17-degree taps-aff rule – or, slightly more seriously, the societal decision that someone legally becomes an adult precisely on a particular birthday.

Perhaps the scientists who ran these studies thought something like: ‘well, it was only a small study, and the small effect I found is probably just due to noisy data. Come to think of it, I was silly even to expect to find an effect here! There’s no point in trying to publish this.’ Crucially, though, this post hoc rationalisation wouldn’t have occurred to them if the same small-sample study, with its potentially noisy data, happened to show a large effect: they’d have eagerly sent off their positive results to a journal. This double standard, based on the entrenched human tendency towards confirmation bias (interpreting evidence in the way that fits our pre-existing beliefs and desires), is what’s at the root of publication bias. If you consider the overall conclusions of a meta-analysis based on Figure 2B rather than 2A, you can see how publication bias deranges the scientific literature. If the studies with small effects have been removed from the funnel shape, the overall effect that shows up in the meta-analysis will by definition be larger than is justified. We get an exaggerated view of the importance of the effects and can be misled into believing something exists when it doesn’t. By failing to publish null or ambiguous studies, researchers force blinkers onto anyone who reads the scientific literature.

In business and politics, one of the great villains is the Yes Man. Countless books tell aspiring managers and leaders to be careful never to surround themselves with people who’ll just nod along with all their decisions, even the bad ones. Winston Churchill proclaimed that ‘the temptation to tell a chief in a great position the things he most likes to hear is one of the commonest explanations of mistaken policy. Thus the outlook of the leader on whose decisions fateful events depend is usually far more sanguine than the brutal facts admit.’42 In science, publication bias makes Yes Men out of the articles that do get published: we can see all the positives, but we don’t get to see the brutal null results. Making decisions on such partial information is a recipe for disaster.

Wansink wrote a blog post about how he’d encouraged one of his graduate students to analyse a dataset he’d collected in a New York pizza restaurant.46 The original hypothesis, he said, had ‘failed’, but instead of publishing those null results or file-drawering the whole thing, he told the student that ‘there’s got to be something [in the dataset] we can salvage’. She complied, and ‘every day … came back with puzzling new results … and every day … c[a]me up with another way to reanalyse the data’. By openly admitting to dredging through his datasets looking for anything that was ‘significant’, Wansink had unintentionally revealed a major flaw in the way he, and unfortunately many thousands of other scientists, conducted research. That flaw has been dubbed ‘p-hacking’.

This latter type of p-hacking is known as HARKing, or Hypothesising After the Results are Known. It’s nicely summed up by the oft-repeated analogy of the ‘Texas sharpshooter’, who takes his revolver and randomly riddles the side of a barn with gunshots, then swaggers over to paint a bullseye around the few bullet holes that happen to be near to one another, claiming that’s where he was aiming all along.50

Did Carney receive a torrent of vile abuse, or lose her job, for admitting all this? No. In fact, the reaction was the precise opposite. A search of Twitter, often maligned as a haven for online bullies, reveals other researchers (rightly) calling Carney’s statement ‘brave’, ‘impressive’, ‘admirable’, ‘the way forward’, an example of ‘how to deal with failure to replicate one’s work’ and a ‘remarkable demonstration of scientific integrity’. One neuroscientist called Carney an ‘intellectual/academic hero’.64 I couldn’t find a single negative reaction – except the one from Amy Cuddy herself, who took the opportunity to distance herself from the p-hacking: ‘I … cannot contest the first author’s [that is, Carney’s] recollections of how the data were collected and analyzed, as she led both.’65

Similarly disheartening findings come from surveys in other fields. A survey of biomedical statisticians in 2018 found that 30 per cent had been asked by their scientist clients to ‘interpret the statistical findings on the basis of expectations, not the actual results’ while 55 per cent had been asked to ‘stress only significant findings but underreport nonsignificant ones’.67 32 per cent of economists in another survey admitted to ‘present[ing] empirical findings selectively so that they confirm[ed their] argument’, while 37 per cent said they had ‘stopped statistical analysis when they had [their] result’ – that is, whenever p happened to drop below 0.05, even if it might’ve done so by chance.68 If you collect together all the p-values from published papers and graph them by size, you see a strange, sudden bump just under 0.05 – there are a lot more 0.04s, 0.045s, 0.049s, and so on, than you’d expect by chance. This isn’t knockdown proof, but it’s a red flag for p-hacking: scientists appear to tweak studies just enough to limbo their results under the 0.05 line, then send them off for publication.69

if you already believe your hypothesis is true before testing it, it can seem eminently reasonable to give any uncertain results a push in the right direction. And whereas the true fraudster knows that they’re acting unethically, everyday researchers who p-hack often don’t.

Endless choices offer endless opportunities for scientists who begin their analysis without a clear idea of what they’re looking for. But as should now be clear, more analyses mean more chances for false-positive results. As the data scientists Tal Yarkoni and Jake Westfall explain, ‘The more flexible a[n] … investigator is willing to be – that is, the wider the range of patterns they are willing to ‘see’ in the data – the greater the risk of hallucinating a pattern that is not there at all.’71

Even without the trial-and-error of classic p-hacking, then, scientists who don’t come to their data with a proper plan can end up analysing themselves into an unreplicable corner. Why unreplicable? Because when they reach each fork in the path, the scientist is being strung along by the data: making a choice that looks like it might lead to p < 0.05 in that dataset, but that won’t necessarily do the same in others. This is the trouble with all kinds of p-hacking, whether explicit or otherwise: they cause the analysis to – using the technical term – overfit the data.73 In other words, the analysis might describe the patterns in that specific dataset well, but those patterns could just be noisy quirks and idiosyncrasies that won’t generalise to other data, or to the real world. This is useless. Most of the time we’re not interested in the workings of one particular dataset (we don’t want to know ‘what is the link between taking antipsychotic drugs and schizophrenia symptoms measured in this specific sample of 203 people between April and May 2019 in Denver, Colorado?’) – we’re looking for generalisable facts about the world (‘what is the link between taking antipsychotic drugs and schizophrenia symptoms in humans in general?’).

This is what scientists are unwittingly doing when they p-hack: they’re making a big deal of what is often just random noise, counting it as part of the model instead of as nuisance variation that should be disregarded in favour of the real signal (if a signal exists in the first place). Woe betide anyone who takes a p-hacked, overfitted model and tries to replicate it in a different sample: its results were contingent on the specific forking paths its creators followed through their noisy data, so it’ll probably tell us very little about the world beyond that single dataset.

You can see why scientists are tempted by overfitting. If you focus only on your own data and forget that your job is to make general statements about the world, a perfectly fitting model like graph 3C is very alluring: there are no uncertainties, no messy datapoints that evade the line you’ve drawn. The neatness alone is not what makes it so compelling, though: simply connecting the dots in the graph doesn’t require any scientific knowledge. But a paper that sounds as if you had come up with the specific shape of the line (your theory) before collecting the data? Now you’ve got the scientific world’s attention – and as we know, a major goal in science is to convince other scientists that your model, theory or study is worth taking seriously.

Stephen Jay Gould referred to science as ‘a profession that awards status and power for clean and unambiguous discovery’ [my italics]. The social psychologist Roger Giner-Sorolla concurs: ‘In a head-to-head competition between papers … the paper with results that are all significant and consistent will be preferred over the equally well-conducted paper that reports the outcome, warts and all, to reach a more qualified conclusion.’74 Here we see how publication bias and p-hacking are two manifestations of the same phenomenon: a desire to erase results that don’t fit well with a preconceived theory.

The desire for attractive-looking results affects even the ‘hardest’ sciences. In her book Lost in Math, the physicist Sabine Hossenfelder argues that physicists have gotten high on their own supply, focusing on the elegance and beauty of models such as string theory at the expense of being able to test, in practice, whether they’re actually true.77 Although the lofty, mathematical work of these string theorists feels like it could hardly be further from the (almost literally) kitchen-sink science of Brian Wansink, both kinds of research can become saturated with the same kinds of all-too-human biases.

It’s impossible to know for sure how many patients have been given useless medical treatments – and false hope – because of p-hacked clinical trials, but the number is surely enormous.83 Think back to the meta-analyses we covered earlier. Even leaving aside the fact that some research is often missing due to publication bias, if the studies included in the meta-analysis are all themselves exaggerated by p-hacking, the overall combined effect – in what’s supposed to be an authoritative overview of all the knowledge on the topic – will end up far from reality.84 You might wonder how doctors and their patients are supposed to trust a medical literature that’s permeated with bias in this way, outside of a minority of clear, well-replicated findings. My response is that I have no idea.

In the US, where numbers are easily available, just over a third of registered medical trials in recent years were funded by the pharmaceutical industry.85 To what extent does this funding, from companies who plan to market the drug if it works, influence the results? The consensus of meta-scientific research on clinical trials is that industry-funded drug trials are indeed more likely to report positive results. In a recent review, for every positive trial funded by a government or non-profit source, there were 1.27 positive trials by drug companies.86

It’s currently a requirement at most journals that you disclose at the end of any published paper, in a conflict-of-interest section, any money you’ve received for, say, consulting for a pharmaceutical company.89 But other kinds of financial conflicts aren’t treated the same way. Many scientists, for example, forge lucrative careers based on their scientific results, producing bestselling books and regularly being paid five- or six-figure sums for lectures, business consulting and university commencement addresses.90 People should of course be allowed to pay whatever they like for book deals, speakers and consultants. But when a lucrative career rests on the truth of a certain theory, a scientist gains a new motivation in their day job: to publish only studies that support that theory (or p-hack them until they do). This is a financial conflict of interest like any other and one aggravated by the extra reputational concerns. One could argue that scientists in this situation should, for the sake of transparency, include a conflict-of-interest statement at the end of every relevant future study.91

the scientist means well and wants to feel like their research is beneficial. We could even call this the ‘meaning well bias’. It’s crushing when the trial you’ve designed to test your new treatment provides null results, meaning we’re no closer to helping sick people. It’s disheartening when you hypothesise about the link of some biological factor to a disease, and it turns out you’ve been barking up the wrong tree. Or, at least, it might feel that way if you have the wrong attitude to science. The currency of positive, statistically significant results in science is so strong that many researchers forget that null results matter too. To know that a treatment doesn’t work, or that a disease isn’t related to some bio-marker, is useful information: it means we might want to spend our time and money elsewhere in future. If it’s properly designed, a study should be of interest whether it produces positive or null results.

What about researchers who have a more ideological or political stake in the truth or falsity of a result? One of the more remarkable conflict-of-interest sections I’ve come across is in a public health paper on the so-called ‘Glasgow Effect’. This is the phenomenon whereby people from Glasgow, and Scotland more generally, die younger on average than those in other similar cities or countries, even after accounting for levels of poverty and deprivation. After reviewing the evidence on this effect, the paper concluded that the root of the unique problem was the ‘political attack’ on Scotland from Prime Minister Margaret Thatcher’s Conservative government in the 1980s, with its policies of deindustrialisation and pushback against organised labour. There are no financial conflicts listed in the conflict-of-interest section – instead, it’s noted that the lead author, Gerry McCartney, ‘is a member of the Scottish Socialist Party.’96 Good on him, I say, for this rare degree of honesty.97 My own field, psychology, is no stranger to scientists who identify as left-wing. The skew this way in psychology is very large indeed: surveys in the United States have found the ratio of liberals to conservatives to be around 10:1. (This isn’t found in other fields, such as engineering, business studies and computer science; in all of those, as in America as a whole, the ratio is far more equal.) Psychology, being about humans and their behaviour, is normally much more intertwined with politic...

The psychologist and philosopher Cordelia Fine – a perspicacious critic of shoddy, often p-hacked studies that purport to explain behavioural differences between the sexes by straightforwardly linking them to testosterone levels, in the process forgetting about social explanations – has also addressed the issue of males being treated as the ‘default’ in medical research, with females being considered a secondary concern or even an aberration.107 In a 2018 opinion piece in the Lancet, Fine suggested that ‘feminist science’ could redress the balance by highlighting these kinds of omissions. She acknowledges that some will be sceptical: ‘common thinking is that while feminism is all very well for Gender Studies, it should be kept away from science, lest political preferences of how women, men, and the world should be lead to distortion of scientific evidence as to how they actually are.’108 But, she argues, aren’t we all biased anyway? Indeed, she says, ‘everyone has a standpoint; no one enjoys a “view from nowhere”. Opening the door to feminist science doesn’t open the door to political bias. It just means we’re not all looking through the same blurred window.’109

Trying to correct for bias in science by injecting an equal and opposite dose of bias only compounds the problem, and potentially invites a vicious cycle of ever-increasing division between different ideological camps. Not only that but the suggestion that scientists should feel proud to let their ideological views impinge on their research seems to offend against both the Mertonian norms of disinterestedness (since it involves allowing non-scientific concerns to encroach on research) and universalism (since it might involve holding scientific arguments to a different standard depending on the political affiliation of their source). The reason that Gerry McCartney decided to declare his Socialist Party membership on his paper about Margaret Thatcher is because it would have been embarrassing – given the paper’s scientific conclusions, which were very much in line with Socialist Party ideology – if it had come out after publication. Even where they’re not as directly relevant to the study as in that case, scientists should regard merging their political beliefs with their science with the same sense of embarrassment.