According to another meta-analysis published in October, scientists are becoming less likely to admit to fabrication, falsification or plagiarism. That study also found that researchers see plagiarism as more heinous than making results up. They are more likely to report a colleague they catch in an act of plagiarism than one fabricating or falsifying data.

How can this be so, when honesty is supposedly such an essential attribute? Because it gets the job done. Raymond De Vries at the University of Michigan and colleagues have argued that data manipulation based on intuition of what a result should look like is “normal misbehaviour”. They see such common misbehaviours as having “a useful and irreplaceable role” in science. Why? Because of “the ambiguities and everyday demands of scientific research”.

In other words, data isn’t often as clean as you would like. According to Frederick Grinnell, an ethicist at the University of Texas, intuition is “an important, and perhaps in the end a researcher’s best, guide to distinguishing between data and noise”. Sometimes you just know that data point was an anomaly to be ignored.