An interesting letter in this week's C&EN on scientific fraud:
As I read about the research fraud case at Columbia University (C&EN, July 11, page 4), my mind instantly went back to a similar case at Purdue University uncovered in 1964. I attended my graduation in May of that year and was awarded a Ph.D. in chemistry. A statement of retraction of a Ph.D. was printed in the graduation program. A student studying under the guidance of Robert A. Benkeser was charged with faking his research results, and his Ph.D. degree had been withdrawn.
The student was a very intelligent person, perhaps brilliant. He never seemed to study much and yet easily passed all of his course exams and degree qualifying examinations. His thesis was about the synthesis of silicocyclopentadiene or derivatives thereof. As in the case of Bengü Sezen, his chicanery was discovered when well-known researchers in the field reported to Benkeser that they could not repeat his student’s literature-reported experimental procedures.
Investigation of the student’s work revealed falsely constructed nuclear magnetic resonance and infrared spectra. Another of Benkeser’s students working in the same laboratory almost had his doctoral work discredited. Fortunately, timely and detailed detective work proved that the second student knew nothing of the fraud, and he was awarded his advanced degree. The student who faked the work allegedly graduated from Harvard University, but I believe the work there also proved fraudulent. Efforts to find the student who perpetrated the fraud were not successful. He was to have started postgraduate work with a famous researcher in Chicago but instead just disappeared.
Stephen E. FrenchI'm under the impression that it was probably a lot easier to fake data back then, but that's just a gut feeling. There were a lot fewer journals back then, and people seemed to pay a lot more attention to them. Sigh.
Rochester, N.Y.
It would have also been easier to "disappear" back then too.
ReplyDeleteI am in two minds. Tracking fakery was probably not as easy back then as it is now. On the other hand, science has gotten much more complex now and verifying results and detecting fakery is correspondingly more difficult and time-consuming (unless the fraud is blatant as in the Sezen or La Clair case). Unless there is an obvious reason, who would even try to verify the results of a complex genomics study, particle detection event or fifty step synthesis.
ReplyDeleteI'd imagine the amount of fakery is about the same now as it ever was. I think it's probably easier to detect now - the detection techniques are so much simpler and so many more people are trying to reproduce results.
ReplyDeleteI think pressure on the assistant prof on tenure track trying to fetch government funding while publishing in high-impact journals has gotten more vicious since then. But methodology with yield-reproducibility issues sometimes came from Corey group too and also (if i remember correctly) from Japanese groups. Whenever there is too much pressure to have the results to come out in certain way, you will find a person who gets ahead by trying to round the yields upwards a bit, for example by weighting the compound after a column but before drying it thoroughly on highvac overnight.
ReplyDeleteYou know, when I do that, I always get 240% yield. I suppose I should practice harder.
ReplyDeleteOK then, here is the deal: You dry your compound for 30 minutes to get a good yield but then you leave it overnight to get nice NMRs. Also, you need to mix and match the best yields for each step, from several runs. In that way when you multiply them you will get the best overall yield of multistep sequence.
ReplyDeleteWhile we are at it, it also it helps to have a "readily available" starting material that takes 6 steps to prepare from material that costs $250/100 mg from Aldrich and a key early step that requires 10 mol% of a proprietary dimeric catalyst with a molecular weight 1200.
I was always taught that a reaction had to work twice in your hands before you could put it in a paper, with similar yields each time. However, I have encountered lots of colleagues who take one good result and start writing up.
ReplyDeleteNot that this is necessarily fraud, per se, but it helps make "good" papers appear "great" when the yields are all over-the-top.