Research Misconduct

What it is

Three states get confused with each other and are handled differently here.

A retracted paper has been formally withdrawn by the journal that published it. A paper compromised by authorship may carry no retraction notice of its own, but comes from a body of work an institution has said it cannot vouch for. A commercially conflicted paper is funded or produced by a party with a stake in the answer, which is a disclosure problem rather than an integrity finding. A conflicted paper is usable with the conflict stated. A compromised one is not usable at all.

The case this rule was written for is a laboratory whose institution stated in September 2018 that it could not vouch for the validity of a set of its papers and could not produce the original data (Cornell University statement, 20 September 2018), after which the journals concerned retracted them (JAMA Network retraction notices, 19 September 2018). See Brian Wansink for the record.

The tension in this page is real and is not resolved here. The publication's governing principle, settled after an audit of a much more famous author, is that status attaches to claims and never to authors. That principle is what stops the site dismissing work by reputation. Setting aside a body of research because of where it came from sits awkwardly next to it, and this page says so.

The reading offered here is that a misconduct finding is an application of the principle rather than an exception to it. Where a laboratory's data cannot be produced and the institution has withdrawn its assurance, the claim has no evidence attached to it any more. There is nothing left to assess. That is a statement about the claim, and it is the only version of this rule that does not become a rule about people.

In effect

The finding that was re-earned

One result from that laboratory was not dropped. It was replicated. An independent, preregistered team ran it again with 464 participants against the original 54 and found the effect at about half the size. This site therefore cites Alejandra Lopez and colleagues (2024) and names the original only as the thing being replicated. The claim survived because somebody else went and got evidence for it, which is the only route back.

The claim that does not survive

Plate size has no such replication. The strongest version of the belief runs through the compromised laboratory, the cleaner work has not found a consistent effect, and so the claim simply does not get made. The page exists to record the gap rather than to fill it.

Three working rules

Flag every retracted or compromised source in bold, in the note and on the page, so it cannot be picked up later by accident. Distinguish retracted, compromised by authorship and commercially conflicted, because they are three different states with three different handlings. And record what could not be established, including when a specific paper carries neither a retraction notice nor a clean confirmation.

What it does not say

It does not say a compromised body of work was wrong about everything. Some of the claims may well be true. What is missing is evidence for them, which is a different problem and one that only replication fixes.

It does not license dismissing work by reputation. The rule is triggered by an institutional finding, not by a reader's suspicion, a bad-tempered review or a failed replication.

It does not make ordinary weakness into misconduct. A small sample, a questionable analytic choice or a result that does not reproduce are normal features of science and are recorded as such throughout this site.

It does not resolve its own tension. The principle that status attaches to claims and never to authors, and a standing flag attached to a named laboratory, pull against each other. That is recorded here rather than tidied away.

---