In the last installation of our ongoing discussion of the obligations of scientists, I said the next post in the series would take up scientists' positive duties (i.e., duties to actually do particular kinds of things). I've decided to amend that plan to say just a bit more about scientists' negative duties (i.e., duties to refrain from doing particular kinds of things).

Here, I want to examine a certain minimalist view of scientists' duties (or of scientists' negative duties) that is roughly analogous to the old Google motto, "Don't be evil." For scientists, the motto would be "Don't commit scientific misconduct." The premise is that if X isn't scientific misconduct, then X is acceptable conduct -- at least, acceptable conduct within the context of doing science.

The next question, if you're trying to avoid committing scientific misconduct, is how scientific misconduct is defined. For scientists in the U.S., a good place to look is to the federal agencies that provide funding for scientific research and training.

Here's the Office of Research Integrity's definition of misconduct:

Research misconduct means fabrication, falsification, or plagiarism in proposing, performing, or reviewing research, or in reporting research results. …

Research misconduct does not include honest error or differences of opinion.

Here's the National Science Foundation's definition of misconduct:

Research misconduct means fabrication, falsification, or plagiarism in proposing or performing research funded by NSF, reviewing research proposals submitted to NSF, or in reporting research results funded by NSF. …

Research misconduct does not include honest error or differences of opinion.

These definitions are quite similar, although NSF restricts its definition to actions that are part of a scientist's interaction with NSF -- giving the impression that the same actions committed in a scientist's interaction with NIH would not be scientific misconduct. I'm fairly certain that NSF officials view all scientific plagiarism as bad. However, when the plagiarism is committed in connection with NIH funding, NSF leaves it to the ORI to pursue sanctions. This is a matter of jurisdiction for enforcement.

It's worth thinking about why federal funders define (and forbid) scientific misconduct in the first place rather than leaving it to scientists as a professional community to police. One stated goal is to ensure that the money they are distributing to support scientific research and training is not being misused -- and to have a mechanism with which they can cut off scientists who have proven themselves to be bad actors from further funding. Another stated goal is to protect the quality of the scientific record -- that is, to ensure that the published results of the funded research reflect honest reporting of good scientific work rather than lies.

The upshot here is that public money for science comes with strings attached, and that one of those strings is that the money be used to conduct actual science.

Ensuring the proper use of the funding and protecting the integrity of the scientific record needn't be the only goals of federal funding agencies in the U.S. in their interactions with scientists or in the way they frame their definitions of scientific misconduct, but at present these are the goals in the foreground in discussions of why federally funded scientists should avoid scientific misconduct.

Let's consider the three high crimes identified in these definitions of scientific misconduct.

Fabrication is making up data or results rather than actually collecting them from observation or experimentation. Obviously, fabrication undermines the project of building a reliable body of knowledge about the world – faked data can’t be counted on to give us an accurate picture of what the world is really like.

A close cousin of fabrication is falsification. Here, rather than making up data out of whole cloth, falsification involves “adjusting” real data – changing the values, adding some data points, omitting other data points. As with fabrication, falsification is lying about your empirical data, representing the falsified data as an honest report of what you observed when it isn’t.

The third high crime is plagiarism, misrepresenting the words or ideas (or, for that matter, data or computer code, for example) of others as your own. Like fabrication and falsification, plagiarism is a variety of dishonesty.

Observation and experimentation are central in establishing the relevant facts about the phenomena scientists are trying to understand. Establishing such relevant facts requires truthfulness about what is observed or measured and under what conditions. Deception, therefore, undermines this aim of science. So at a minimum, scientists must embrace the norm of truthfulness or abandon the goal of building accurate pictures of reality. This doesn’t mean that honest scientists never make mistakes in setting up their experiments, making their measurements, performing data analysis, or reporting what they found to other scientists. However, when honest scientists discover these mistakes, they do what they can to correct them, so that they don’t mislead their fellow scientists even accidentally.

The importance of reliable empirical data, whether as the source of or a test of one’s theory, is why fabrication and falsification of data are rightly regarded as cardinal sins against science. Made-up data are no kind of reliable indicator of what the world is like or whether a particular theory is a good one. Similarly, “cooking” data sets to better support particular hypotheses amounts to ignoring the reality of what has actually been measured. The scientific rules of engagement with phenomena hold the scientist to account for what has actually been observed. While the scientist is always permitted to get additional data about the object of study, one cannot willfully ignore facts one finds puzzling or inconvenient. Even if these facts are not explained, they must be acknowledged.

Those who commit falsification and fabrication undermine the goal of science by knowingly introducing unreliable data into, or holding back relevant data from, the formulation and testing of theories. They sin by not holding themselves accountable to reality as observed in scientific experiments. When they falsify or fabricate in reports of research, they undermine the integrity of the scientific record. When they do it in grant proposals, they are attempting to secure funding under false pretenses.

Plagiarism, the third of the cardinal sins against responsible science, is dishonesty of another sort, namely, dishonesty about the source of words, ideas, methods, or results. A number of people who think hard about research ethics and scientific misconduct view plagiarism as importantly different in its effects from fabrication and falsification. For example, Donald E. Buzzelli (1999) writes:

[P]lagiarism is an instance of robbing a scientific worker of the credit for his or her work, not a matter of corrupting the record. (p. 278)

Kenneth D, Pimple (2002) writes:

One ideal of science, identified by Robert Merton as “disinterestedness,” holds that what matters is the finding, not who makes the finding. Under this norm, scientists do not judge each other’s work by reference to the race, religion, gender, prestige, or any other incidental characteristic of the researcher; the work is judged by the work, not the worker. No harm would be done to the Theory of Relativity if we discovered Einstein had plagiarized it…

[P]lagiarism … is an offense against the community of scientists, rather than against science itself. Who makes a particular finding will not matter to science in one hundred years, but today it matters deeply to the community of scientists. Plagiarism is a way of stealing credit, of gaining credit where credit is not due, and credit, typically in the form of authorship, is the coin of the realm in science. An offense against scientists qua scientists is an offense against science, and in its way plagiarism is as deep an offense against scientists as falsification and fabrication are offenses against science. (p. 196)

In fact, I think we can make a good argument that plagiarism does threaten the integrity of the scientific record (although I'll save that argument for a separate post). However, I agree with both Buzzelli and Pimple that plagiarism is also a problem because it embodies a particular kind of unfairness within scientific practice. That federal funders include plagiarism by name in their definitions of scientific misconduct suggests that their goals extend further than merely protecting the integrity of the scientific record.

Fabrication, falsification, and plagiarism are clearly instances of scientific misconduct, but the misconduct definitions of the United States Public Health Service (whose umbrella includes NIH) and NSF used to define scientific misconduct as fabrication, falsification, plagiarism, and other serious deviations from accepted research practices. The “other serious deviations” clause was controversial, with a panel of the National Academy of Sciences (among others) arguing that this language was ambiguous enough that it shouldn’t be part of an official misconduct definition. Maybe, the panel worried, “serious deviations from accepted research practices” might be interpreted to include cutting-edge methodological innovations, meaning that scientific innovation would count as misconduct.

In his article 1993 article, “The Definition of Misconduct in Science: A View from NSF,” Buzzelli claimed that there was no evidence that the broader definitions of misconduct had been used to lodge this kind of misconduct complaint. Since then, however, there there have been instances where definitions of scientific misconduct containing an "other serious deviations" clause could be argued to take advantage of the ambiguity of the clause to go after a scientist for political reasons.

If the “other serious deviations” clause isn’t meant to keep scientists from innovating, what kinds of misconduct is it supposed to cover? These include things like sabotaging other scientists’ experiments or equipment, falsifying colleagues’ data, violating agreements about sharing important research materials like cultures and reagents, making misrepresentations in grant proposals, and violating the confidentiality of the peer review process. None of these activities is necessarily covered by fabrication, falsification, or plagiarism, but each of these activities can be seriously harmful to scientific knowledge-building.

Buzzelli (1993) discusses a particular deviation from accepted research practices that the NSF judged as misconduct, one where a principal investigator directing an undergraduate primatology research experience funded by an NSF grant sexually harassed student researchers and graduate assistants. Buzzelli writes:

In carrying out this project, the senior researcher was accused of a range of coercive sexual offenses against various female undergraduate students and research assistants, up to and including rape. … He rationed out access to the research data and the computer on which they were stored and analyzed, as well as his own assistance, so they were only available to students who accepted his advances. He was also accused of threatening to blackball some of the graduate students in the professional community and to damage their careers if they reported his activities. (p. 585)

Even opponents of the "other serious deviations" clause would be unlikely to argue that this PI was not behaving very badly. However, they did argue that this PI's misconduct was not scientific misconduct -- that it should be handled by criminal or civil authorities rather than funding agencies, and that it was not conduct that did harm to science per se.

Buzzelli (who, I should mention, was writing as a senior scientist in the Office of the Inspector General in the National Science Foundation) disagreed with this assessment. He argued that NSF had to get involved in this sexual harassment case in order to protect the integrity of its research funds. The PI in question, operating with NSF funds designated to provide an undergraduate training experience, used his power as a research director and mentor to make sexual demands of his undergraduate trainees. The only way for the undergraduate trainees to receive the training, mentoring, and even access to their own data that they were meant to receive in this research experience at a remote field site was for them to submit to the PI’s demands. In other words, while the PI’s behavior may not have directly compromised the shared body of scientific knowledge, it undermined the other central job of the tribe of science: the training of new scientists. Buzzelli writes:

These demands and assaults, plus the professional blackmail mentioned earlier, were an integral part of the subject’s performance as a research mentor and director and ethically compromised that performance. Hence, they seriously deviated from the practices accepted in the scientific community. (p. 647)

Buzzelli makes the case for an understanding of scientific misconduct as practices that do harm to science. Thus, practices that damage the integrity of training and supervision of associates and students – an important element of the research process – would count as misconduct. Indeed, in his 1999 article, he notes that the first official NIH definition of scientific misconduct (in 1986) used the phrase "serious deviations, such as fabrication, falsification, or plagiarism, from accepted practices in carrying out research or in reporting the results of research." (p. 276) This language shifted in subsequent statements of the definition of scientific misconduct, for example "fabrication, falsification, plagiarism, and other serious deviations from accepted practices" in the NSF definition that was in place in 1999.

Reordering the words this way might not seem like a big shift, but as Buzzelli points out, it conveys the impression that "other serious deviations" is a fourth item in the list after the clearly enumerated fabrication, falsification, and plagiarism, an ill-defined catch-all meant to cover cases too fuzzy to enumerate in advance. The original NIH wording, in contrast, suggests that the essence of scientific misconduct is that it is an ethical deviation from accepted scientific practice. In this framing of the definition, fabrication, falsification, and plagiarism are offered as three examples of the kind of deviation that counts as scientific misconduct, but there is no claim that these three examples are the only deviations that count as scientific misconduct.

To those still worried by the imprecision of this definition, Buzzelli offers the following:

[T]he ethical import of "serious deviations from accepted practices" has escaped some critics, who have taken it to refer instead to such things as doing creative and novel research, exhibiting personality quirks, or deviating from some artificial ideal of scientific method. They consider the language of the present definition to be excessively broad because it would supposedly allow misconduct findings to be made against scientists for these inappropriate reasons.

However, the real import of "accepted practices" is that is makes the ethical standards held by the scientific community itself the regulatory standard that a federal agency will use in considering a case of misconduct against a scientist. (p. 277)

In other words, Buzzelli is arguing that a definition of scientific misconduct that is centered on practices that the scientific community finds harmful to knowledge-building is better for ensuring the proper use of research funding and protecting the integrity of the scientific record than a definition that restricts scientific misconduct to fabrication, falsification, and plagiarism. Refraining from fabrication, falsification, and plagiarism, then, would not suffice to fulfill the negative duties of a scientist.

We'll continue our discussion of the duties of scientists with a sidebar discussion on what kind of harm I claim plagiarism does to scientific knowledge-building. From there, we will press on to discuss what the positive duties of scientists might be, as well as the sources of these duties.


Buzzelli, D. E. (1993). The definition of misconduct in science: a view from NSF. Science, 259(5095), 584-648.

Buzzelli, D. (1999). Serious deviation from accepted practices. Science and Engineering Ethics, 5(2), 275-282.

Pimple, K. D. (2002). Six domains of research ethics. Science and Engineering Ethics, 8(2), 191-205.


Posts in this series:

Questions for the non-scientists in the audience.

Questions for the scientists in the audience.

What do we owe you, and who’s “we” anyway? Obligations of scientists (part 1)

Scientists’ powers and ways they shouldn’t use them: Obligations of scientists (part 2)

Don’t be evil: Obligations of scientists (part 3)

How plagiarism hurts knowledge-building: Obligations of scientists (part 4)

What scientists ought to do for non-scientists, and why: Obligations of scientists (part 5)

What do I owe society for my scientific training? Obligations of scientists (part 6)

Are you saying I can't go home until we cure cancer? Obligations of scientists (part 7)