Tracking retractions as a window into the scientific process
Physics journal pulls two papers for data shortcuts
A publisher is retracting two papers today by a team of physicists who took a short cut in reporting their data.
The papers present a method for imaging very small things — like biological processes on a molecular scale — that could be an alternative to electron microscopy, as the authors explain in a video. But after the papers were published in the New Journal of Physics, last author Ulf Leonhardt, now based at the Weizmann Institute of Science, found out that some of the data
were pixel-by-pixel mirror-symmetric, which is impossible for genuine experimental data.
One of the researchers co-authored a subsequent paper that acknowledges one of the papers incorrectly assumed the data were symmetrical, and could therefore be extrapolated from one side to the other. A representative of the publisher told us they have not seen any signs of misconduct, and the problem seemed to result from a “series of apparent miscommunications between the authors.”
IOP Publishing Limited is retracting this paper because its conclusions cannot be validated due to the lack of measured data on the right-hand side of figure 4. In the measured curves plotted in figure 4, the data on one side of the device were measured first, and the data on the other side were obtained by mirroring, under the symmetry assumption arising from the device structure.
The 2011 paper has been cited 38 times, according to Thomson Reuters Web of Science.
IOP Publishing Limited is retracting this paper because its conclusions cannot be validated due to the lack of measured data on the right-hand side of figure 7.
On page 12, for the measured curves plotted in figures 7(b) and (c), the data on one side of the device were measured first, and the data on the other side were obtained by mirroring, under the symmetry assumption arising from the device structure.
That 2012 paper has been cited 11 times.
Take a look at an example of the symmetry:
Figures 7(b) and (c) from “Subwavelength imaging with materials of in-principle arbitrarily low index contrast”
Here’s the imaging device used in both papers:
Figure 6: The device, from “Subwavelength imaging with materials of in-principle arbitrarily low index contrast”
Leonhardt told us he has been trying to have the papers flagged for a couple years, along with co-authors Sahar Sahebdivan, a PhD student at the University of St Andrews at the time of the work, and Tomas Tyc at Masaryk University in the Czech Republic. According to Leonhardt:
On 19 August 2014 we submitted an Expression of Concern to New Journal of Physics. We had found out that the experimental data of the figures in question were pixel-by-pixel mirror-symmetric, which is impossible for genuine experimental data. My student Alex Kogan at Weizmann Institute of Science had discovered this as part of his MSc project, and we informed NJP about it.
Tim Smith, the associate director of the Institute of Physics, which publishes the journal, told us that after the authors’ request for an EoC,
a long and complex investigation was carried out by the journal, the priority of which was to facilitate dialogue with all parties in order to maintain, and where necessary, correct the scientific record. The complex nature of the case revealed a series of apparent miscommunications between the authors involved however all authors agreed to the retractions in the interest of correcting the scientific record.
According to Leonhardt, data collection was the responsibility of the first author on both papers, Ma Yungui, now at the Zhejiang University in China.
We also note that the experiment assumed that the power absorbed by the scanning cable on the left and the right sides of the drain array is symmetric is not correct for the experiment reported in [New J. Phys. 13 (2011) 033016], as the drain array itself is not symmetric. The highly non-symmetric distribution of the absorbed power is also verified by our simulation and experimental results. The experimental “result” of resolving two image peaks could potentially be recovered using only a single image peak, which demonstrates the wrong assumption of mirror symmetry.
Leonhardt told us:
We contacted the journal on 11 June 2015 and asked the NJP to take action, as the case was clear now. We stated that “of course we would like to retract the two NJP papers, but we would do it in such a way that it becomes clear we are innocent. In fact, we have been the ones that were manipulated most of all.”
Co-author Sahebdivan added:
Regarding these two papers, every one [agreed] that an incorrect experimental process in the data collection which was based on a very vague assumption [resulted] in an incorrect conclusion.
Smith noted:
To the best of our knowledge there is no sign of misconduct.
When we asked Smith to explain further, he told us:
The question of whether any form of scientific misconduct has occurred is not something IOP Publishing can be unequivocal about based on the information that has been supplied by the authors during the investigation. For this reason it would not be appropriate to speculate in response to your question. We are satisfied however that the correct resolution to the situation has now been achieved in terms of the duty we have as a publisher to correct the scientific record.
We reached out to Yungui for his side of the story, and will update this post with anything else we learn.
A publisher is retracting two papers today by a team of physicists who took a short cut in reporting their data.
