So how long do researchers have to keep records of experiments?
We had that question while reading the retraction notice for a 2007 paper in the Journal of Clinical Investigation:
It has come to our attention that some of the panels in Figure 5 appear to be duplications of each other. The authors are unable to provide the original source files that were used to generate these data. In the absence of the original data that verify the integrity of the images, the JCI Editorial Board has decided to withdraw Figure 5 from the scientific record. No issues have been raised in regard to any of the other data in this manuscript.
The authors concur with this course of action.
The paper — which has been cited 55 times, according to Thomson Scientific’s Web of Knowledge — was the subject of an earlier unrelated correction, for a misspelled author’s name.
Now back to our question: It’s not exactly an answer, but the Office of Research Integrity requires that allegations be brought forward within six years, a time frame that at least one publisher follows. We imagine different universities have different standards, so we’d welcome input from Retraction Watch readers.
According to the notice, only Figure 5 appears to be retracted, not the whole paper as your blog entry title suggests.
Indeed. But the journal is calling this a retraction rather than a correction… Not even ‘partial retraction’, as I believe some journals have dared use.
but they also clearly state the rest of the data is not affected, making it sound as if the rest is not retracted.
I was about to comment on that as well…
“the JCI Editorial Board has decided to withdraw Figure 5 from the scientific record. No issues have been raised in regard to any of the other data in this manuscript.”
How many lanes are there in the blots in Figure 6 and Figure 8?
Reblogged this on lab ant and commented:
This is an important question! In my opinion data must be published (a matter of transparency and getting the maximum out of data) and archived forever using one of the services available e.g http://www.datadryad.org/ I posted this question on Research Gate as well. Lets see what comes out of this! https://www.researchgate.net/post/How_long_do_researchers_have_to_keep_records_original_data_of_experiments
thank you for this great question. I think today there is no excuse any more to “loose” or delete data.
Reblogged this on The Firewall.
I can see some glaring problem with Figure 3, 6 and 9
In Figure 3, two of the three bands in the lowest lane are pasted. If you adjust the contrast/brightness they become apparent.
In figure 6B lane 4 (+iPTH), there is a cut mark corresponding to lane 16
Figure 9 is the result of three or four different blots. If you take the blot to photoshop (or a comparable software) and reduce the brightness and maximize the contrast you will see three distinct boxes.
Schmuck, you are correct.
The authors have had some problems in the past:
Pages F838–F843. Weinman EJ, Mohanlal V, Stoycheff N, Wang F, Steplock D, Shenolikar S, and Cunningham R. Longitudinal study of urinary excretion of phosphate, calcium, and uric acid in mutant NHERF-1 null mice. Am J Physiol Renal Physiol 290: F838–F843, 2006 (doi:10.1152/ajprenal.00374.2005).
On p. 840, Fig. 1 was printed incorrectly. The two panels in C were duplicates of the graphs in A. The correct figure appears here with its original legend.
I can’t find the problems with Figs 3 and 6. Fig 9 clearly has three boxes. This seems very serious and certainly intentional – is there any innocent explanation?
Michael, the other 2 figures (3 & 6) are less clear as somebody (or something) created boxes in figure 3 on heavy background (not the clear one of figure 9). As for figure 6, you have to look for a straight line and empty background in that lane.
“Office of Research Integrity requires that allegations be brought forward within six years”.
I believe that research misconduct is a continuing offense and therefore the statute of limitation is extended. Furthermore, if public money and public interest are involved, allegations can be brought any time.
In the UK,the BBSRC requires that we maintain data for 10 years at least. I threw out a couple of months ago some data from ~22 years ago – print out from an instrument, no longer readable and the paper was crumbling – time to go!
But one should really keep data for as long as possible. it is sometimes useful – I used some 20 year old data in supplementary recently, to answer a reviewer’s question. An experiment done a long time ago, but never published came in handy!
This lab has a long record of dubious handling of figures. Here’s another in JBC that they had to amend.
http://www.jbc.org/content/283/7/4459.full.pdf (“During publication of this manuscript, gaps separating individual
panels to signify a composite figure were eliminated or omitted in some figures.”
nice one Pablo…
Apologies for what may be a long comment. Six years is a fairly good guideline range. First, consider ordinary electronic, digital data. People change institutions and lose control of their data. The equipment which generated the data fails or becomes obsolete, and some of the little details of the procedures, or even reagents, are lost. Think of the histological preparations from the era of microscope cytology, for example. Also, digital data can be lost or corrupted in a technology change. I’ve done paleontology research, where the key articles were supposedly available — but in a microfilm format for which there were no longer readers, or in an old thesis in which the photographs have been bleached white or have glued themselves together — all dutifully, if meaninglessly, reproduced digitally as grey smears.
But not all data is digital. Consider type specimens from zoology and paleontology. They’re supposed to be kept forever. Unfortunately, most pachypleurosaur types were housed in urban museums in Germany, during World War II, and were blown to dust by allied bombing. In the same era, an important collection of mycological type specimens was housed at the University of Hiroshima. Think of all the important specimens and data for archeology and anthropology housed today in places like Addis Abbaba, Damascus, or Baghdad, or already intentionally destroyed in Afghanistan or Yemen.
More often, institutions and researchers simply lose funding, lose interest, or die. I recently saw an effort to preserve and publish a mass of polar climatology data from the 1950’s. This is moderately useful and relevant stuff today, but a considerable amount has been lost. The Economist is running a short article this week on the uncertain status and fate of electronic data when the account owner dies. A number of state museums and research facilities closed during the recent recession. Most of their data is probably recoverable, but “most” and “probably” cover a multitude of sins.
Don’t wring your hands over all this too much. Things happen. We’re not, and don’t want to be, a bunch of medieval monks buried in an avalanche of our own history. To the contrary: in the social sciences and medicine, in particular, the incremental benefit of preserving all the data indefinitely is probably outweighed by the incremental danger of accidentally disclosing personal information buried in old data.
In business and law, experience has shown that retention periods of 5-10 years are usually enough for routine documents. That, in part, a reflects typical legal statutes of limitation; but those statutes are also a reflection of social experience. Even for the most significant stuff, the old common law rule of thumb, an adult lifetime (“the memory of man runneth not to the contrary”) or about 50 years, is more than enough. Science is supposed to be repeatable. Recently, someone attempted to show that Eddington’s photographs of the transit of Mercury in 1918 (IIRC) didn’t actually confirm general relativity. This was a pointless exercise. Even if the charge were true, general relativity is established physics today, just as Eddington’s tendency to confuse reality with his personal intuition is established history (read any biography of Chandrasekar).
Keeping data is a good thing. For a reasonable period of time, say six years, it ought to be required. Beyond that, old data is nice; but an absolute requirement rapidly becomes burdensome and unreasonable.
Again, sorry for the length.
I’ve looked into this a bit and the minimum required data retention time is 3 years from the submission of the final report on the grant if you’re federally funded in the US. Of course, retractions like this and other retention policies mean that 3 years is not long enough.
Current policies are all over the place when it comes to retention times. I’ve seen universities with 3-7 year post-grant retention requirements and even requirements from funders in the UK to retain data 10 years after last access (http://www.epsrc.ac.uk/about/standards/researchdata/Pages/expectations.aspx), meaning 10 years after the last time a third-party requests the data!
There doesn’t seem to be a clear answer on how long to keep data. My best guess is *at least* 6-10 years post-publication and that number is likely to increase.
I ended up writing a whole blog post in response to this RW post, so more information on retention policies and sources for all of my information can be found here: http://dataabinitio.com/?p=172
Someone forward this notice to duke.fact.checker ASAP! Outstanding blog on all things Duke that spent quite a bit of time on Potti-gate.