In 1910 the American physicist Robert Millikan determined the electronic charge carried by an electron through a series of experiments with falling oil-drops. The experimental setup was ingenious, but also prone to mistakes, and resent analysis of Millikan’s laboratory protocols have revealed that he filtered his raw data quite heavily and only included a subset of the experimental runs in his publications. The question is: Did Millikan do anything wrong?

When I started teaching research ethics to undergraduate students I included the Millikan story as part of a theme on research integrity. To my mind the case was clear: Millikan had deleted data points – sometimes based on prior analysis that a specific data point did not fit his hypothesis – and he had done so without transparency. This was a clear case of misconduct (at least from a moder point of view; how Millikan’s contemporaries would have judged the case is perhaps less clear). My students however did not see things the same way. Millikan received a Nobel prize for his effort, so his intuition was right all along, but more importantly the students pointed out that in real life, experimental work is often messy and difficult; sometimes you make mistakes, sometimes the equipment malfunctions and sometimes anomalies just prop up for no apparent reason. And of course, if you want to make a meaningful analysis you will have to remove the bad data. So, what I saw as illegitimate filtration of data, the students saw as necessary cleaning.

The experience made me realize that simple overall rules such as ‘do not filtrate data’ are of little use in research integrity training, as real-life lab practice (and real-life lab instructors) will often require students to remove anomalous data anyway. The question confronting students is rather: when is it justifiable to remove data points and when it is not? There is for instance a big difference between removing a data point because it is inconsistent with your hypothesis and removing it because you believe your equipment to be malfunctioning. And there is a big difference between deleting data in secret and doing it with full transparency.

The students that read the Millikan case lacked this deeper contextual understanding. They only saw the simple action ‘removing anomalous data points’ and recognized it as something normal and acceptable from their own laboratory practice. But you cannot make sound ethical evaluation of such an action without understanding the context and the justifications (or lack of same) for the action. In a sense, actions such as deleting data points are located in a grey zone, and students will have to develop sound judgement in order to act with integrity.

The handling of anomalous data is not the only grey zone students will have to learn to navigate. Similar grey zones arise in connection to the use of other’s work (when is it plagiarism and when is it not?) and collaboration with other students (when are we co-developing ideas and when is one of us copying?). Some of these grey zones are probably inevitable, some are products of student ignorance while others are created by teachers and institutional structures.

A central goal of research integrity is teaching students how to navigate such grey zones, but in order to reach that goal we – teachers and developers of research integrity courses –  have to understand the nature of the grey zones and how they arise. In our little corner of the INTEGRITY project (WP2) we are (among other things) probing these grey zones. It is highly interesting, and we can’t wait to tell you all what we have seen.