Why is data falsification reported to be more usual in some fields than others?

Question

This big info-graphic (at the end of the question) explains that biomedicine and psychology have greater rate of data fabrication/falsification/mishandling than other fields. A similar effect was reported a PLOS ONE paper:

Once these factors were controlled for, surveys conducted among clinical, medical and pharmacological researchers appeared to yield higher rates of misconduct than surveys in other fields or in mixed samples.

and

it suggests that misconduct in clinical, pharmacological and medical research is more widespread than in other fields. This would support growing fears that the large financial interests that often drive medical research are severely biasing it.

However, the paper does not offer any further discussion of the field dependence of data manipulation. So, have there been explanations proposed about why these fields are "special"?

---

   

Answer 1

Your quotes from the Plos One meta analysis highlight problems in clinical, medical and pharmacological, but your question is about biomedicine and psychology. The meta analysis paper found no evidence for increased rates of reporting misconduct in either biomedicine or the social sciences suggesting that those fields are not "special". The authors then present a reasonable conjecture about why the reports of misconduct in medicine might be higher

However, as all survey-based data, this finding is open to the alternative interpretation that respondents in the medical profession are simply more aware of the problem and more willing to report it. This could indeed be the case, because medical research is a preferred target of research and training programs in scientific integrity, and because the severe social and legal consequences of misconduct in medical research might motivate respondents to report it.

I think base on the meta analysis I would argue that data falsification is NOT more usual in some fields than others, but simply that reports and sensitivity to it is more usual when a field specifically trains individuals to recognize it.

Answer 2

I haven't read the paper you mention, but I can definitely see why falsification could be more common in these fields.

I think the factor that affects data falsification the most is not ethical differences between fields, but rather simply how difficult it is to identify the falsification.

In fields where experiments are simple to replicate, false results will be discovered easily. In clinical/pharmacological research, for example when testing an effect of a medical treatment, studies on human subjects are extremely costly to perform, so they are difficult to replicate. Even if they are replicated, there is such intrinsic large variance in the results that it would be difficult to conclude that results from a previous study were purposefully falsified even if they are different.

Answer 3

I'd be careful with the assertion that some fields are more prone to scientific misconduct than others. To make that claim I would like to see a study comparing different fields, while keeping, e.g., the stage of the researchers (PhD student, PostDoc, Professor) comparable. And let's hope that this study about data falsification does not contain data falsification.

Then there's the stage of the discussion about scientific misconduct. Psychology and other disciplines had a couple of highly visible cases of scientific misconduct (e.g., Stapel). They put the discipline, or at least the subdiscipline (social psychology) in question. So now there are people actively looking for ways to detect misconduct, e.g., by looking for statistical inconsistencies that happen when people fake data. So it might just be that a field is more sensitive to the issue and thus more prone to detect cases. The current "house-cleaning" might expose the misconduct that still lies hidden in other domains.

But there also might be differences between fields because it might be easier to get away with these forms of misconduct in some compared to others. Looking at the fraud triangle (see posting here, my blog), you need a motive/pressure, a rationalization and the opportunity to commit fraud.

• Motive: I know no scientific field that isn't dominated by publish or perish, so there's a motive. In some fields you can make a lot of money or the alternatives to an academic career look especially bleak (after all, it's impossible for all PhDs/PostDocs to stay in Academia, even if all were excellent researchers). And while all research is risky (no-one can guarantee that your experiments work), some domains may have a higher failure rate. So motive might differ.
• Rationalization: The rationalization might be higher in more fuzzy/soft domains, where it might be easier to convince yourself that the theory is sound, but the data is skewed (so many possibly confounding variables). There might also be "personal experience" as a bias, e.g., when it comes to evaluating therapies ("But I know that it works, I have seen it.").
• Opportunity: Also the opportunity might be higher in some fields where you alone have access to the data and it's virtually impossible to reproduce exactly the same results (again, confounding variables). Not to mention that replications are rare, esp. with more "interesting but not so relevant" topics.

So while it's an interesting question, it's also a highly complex issue. And there might even be the desire to point to other disciplines to avoid addressing the issue oneself ("misconduct never happens in science ... in our domain ... in our sub-domain ... at our institute ... in our workgroup ..."). Instead, science needs better controls for scientific misconduct, no matter the discipline.

Edit: P.S.: As for the flyer highlighting misconduct in (clinical) psychology -- well, consider the source. ;-)