How To Read A Retraction

Mar 05 2008 Published by under Scientific Misconduct

Nature published a retraction today of a paper from the laboratory of Nobel Prize winner Linda Buck. The study used a trans-synaptic tracing method to determine ascending axonal projection patterns in olfactory cortex. Here is the substance of the retraction:

During efforts to replicate and extend this work, we have been unable to reproduce the reported findings. Moreover, we have found inconsistencies between some of the figures and data published in the paper and the original data. We have therefore lost confidence in the reported conclusions.

Let's interpret the significance of the retraction, as well as an interesting comment made by someone in the field:


One thing that is very interesting about this retraction is that it includes a detailed "Author Contributions" section that never appeared in the original article:

Author Contributions L.B.B. and L.F.H. conceived the project, L.F.H. and J.-P.M. prepared gene-targeting constructs to generate the mice, S.S. trained Z.Z. in gene-targeting techniques, Z.Z. prepared and analysed the mice and provided all figures and data for the paper, and L.B.B. and Z.Z. wrote the paper. Correspondence and requests for materials should be addressed to L.B.B. (Email: lbuck@fhcrc.org)

Most Nature retractions I have looked at do not contain and "Author Contributions" section, and it's inclusion is very significant. It is also worth noting that, because there is no statement that any of the authors has refused to sign the retraction, all of the authors have agreed to the full content of the retraction. Can you see who's falling on their sword?
Now here's something that made me laugh. In a news piece on the retraction, another researcher in the field had this to say:

Neuroscientist Gilles Laurent of the California Institute of Technology in Pasadena, whose work on insect olfactory networks has occasionally conflicted with the results reported by Buck's lab, says that this [retraction] has not hindered his research. "These questions are sufficiently complex and require such large amounts of data at high resolution that I have never considered them convincingly resolved in any system," he says.

In other words, Laurent always thought the results were bullshit, and never relied on them in his own work.

29 responses so far

  • juniorprof says:

    This one really stinks if you ask me... Are they accusing without accusing by including this unprecedented author contribution line for a retraction?

  • Coturnix says:

    Z.Z. is in trouble.

  • DrugMonkey says:

    Oh this is a fun one. We're learning all kinds of stuff. Such as the fact that "training" someone in "gene-targeting techniques" gets you a Nature authorship. good to know.
    Question: is making constructs really worth a co-first authorship when someone else supposedly generates all the data that actually made up a paper?

  • Ewan says:

    Can you see who's falling on their sword?
    Well, yes. But he's already moved on to his faculty position elsewhere..
    ..I might also have expected a line to the effect that 'data forming the basis for other related publications have been examined and no further problems have been found.' Absent that, one has to wonder, I guess.

  • PhysioProf says:

    Question: is making constructs really worth a co-first authorship when someone else supposedly generates all the data that actually made up a paper?

    I would say, it depends. If you are just cloning a fucking 300 base pair EcoRI fragment into pBluescript, probably not. If it is a large transgenesis or gene targeting construct with many different elements, why not? It is extremely common, even traditional, in experiments with genetically modified mice that the person who makes the construct is the first author of the first paper that results.

  • DrugMonkey says:

    Well, yes. But he's already moved on to his faculty position elsewhere..
    Indeed.
    Not to mention the R03 and R21 awards.
    An appointment and two grants which were not awarded to non-data-faking peers.
    Are you getting pissed yet?

  • PhysioProf says:

    If that dude had listened to PhysioProf, he'd already have a fucking R01, and not just an R03 and an R21.

  • Wow. That is disgusting. Inasmuch as ol' ZZ has been either totally incompetent and/or less than honest, hanging him out to dry like that is cheap. Were these data not scrutinized and discussed in lab meetings at all?
    ---"Z.Z. prepared and analysed the mice and provided all figures and data for the paper"--- and we were happy to go along for the Nature ride without even bothering to look at the original data, but now that the shit has hit the fan, we'd like you to know it's not our fault......
    Come to think of it, if it is a case of incompetence, at least the senior author is equally culpable; if it is fraud, there needs to be a report to the OSI, an investigation etc.
    Any which way, it is distasteful to see that attempt at laundry. Great find, DM.
    [Ed- actually that was the PP that found this one]

  • ---"[Ed- actually that was the PP that found this one]"---
    D'oh! Sorry for the oversight. Props, PP.
    But then again, if my incorrect citation managed to get PP all worked up into a profane shiznit frenzy for a few minutes, then I guess some good would have out of it!

  • PhysioProf says:

    Dude, PhysioProf is always in a profane shiznit frenzy. There is no "working up to" required.

  • no name says:

    "Oh this is a fun one. We're learning all kinds of stuff. Such as the fact that "training" someone in "gene-targeting techniques" gets you a Nature authorship. good to know."
    I don't know about the politics of that particular paper, but it may be that SS actually did a bit more than "training in gene targeting techniques". Given the authorship break down coming as it does retrospectively, it's damage control for poor SS's career (and indeed, for all I know SS is entirely innocent and deserving of such protection).
    "Question: is making constructs really worth a co-first authorship when someone else supposedly generates all the data that actually made up a paper?"
    If it makes you feel any better, you can work for years in Dr. Bucks laboratory as a tech and only end up on the "acknowledgements" of a few papers. (My labmate is such a person. To be entirely clear- he mostly seems fine with it- I don't mean to imply anything bad about the lab in those respects. I just happen to think it unlikely that this lab should be known for handing out authorships like candy).
    As far as meriting joint first authorship- it says quite clearly that LFH conceived of the project with LBB. Now I know that PIs tend to believe all ideas spring forth fully-formed from the PI's head (and *it goes without saying* that no one other than an R01-holding 'professorial' status individual can ever have any ideas), but it sounds like LFH contributed to planning of the experiments as well as construction of the constructs- which would meet NIH criteria for authorship when combined with making the constructs. First authorship de facto depends partially on whether both people devoted comparable amounts of time, and given how evil some constructs can be, it could easily have taken as much time as collecting the actual data.

  • Sven DiMIlo says:

    Man, I feel for you Big Science guys with your R01s and your labs full of postdocs and graduate slaves and your Nature papers and retractions thereof. In my so-old-timey-its-damn-near-dinosaurian field of little science, it's never so tough. The first author is the one who, you know, wrote the paper.

  • Barn Owl says:

    Author Contributions: L.B.B. and L.F.H. conceived the project, L.F.H. and J.-P.M. prepared gene-targeting constructs to generate the mice, S.S. trained Z.Z. in gene-targeting techniques, Z.Z. prepared and analysed the mice and provided all figures and data for the paper, and L.B.B. and Z.Z. wrote the paper.
    At least on the abstract page you linked, there's an error in the numbering for the "authors who contributed equally" (only S.S. has the superscript 3). And if S.S. trained Z.Z. in "gene-targeting techniques", why did L.F.H. and J.-P.M. have to prepare the gene-targeting constructs...unless S.S. trained Z.Z. to do the electroporations and blastocyst injections, using the constructs that the other two generated? Very strange.
    Genetically engineered mice are so commonplace these days that you have do a lot of novel and dramatic experiments with them, once generated, in order to get a publication in a top tier journal. Even back in 2001, that was the case, so undoubtedly there was a great deal of pressure on the person(s) who spent the time and \(\)$ generating the GEM, to produce a knock-yer-sox-off set of results. If you don't have a human disease model, then you'd better have loads of colorful cells and some fancy neural circuitry. Or tentacles. Tentacles on a GEM would get you a top tier pub. As would opposable thumbs or a cockatoo crest.
    I wouldn't envy Z.Z. his faculty position, necessarily. Back in 2006, UTMB fired over 1000 employees, including some clinical and research faculty, and the long-term fate of the university as a whole is not clear.

  • Pinko Punko says:

    I wouldn't read that much into Laurent's comment about whether he believed this particular work or not. It sounded like he was saying that no one paper is definitive in this field. I would like to know whether Nature going forward is going to publish "author contributions". I believe that are preferred but not mandatory for regular contributions.
    As for "hanging ZZ out to dry"- I don't think so. They stood by the work while comfortable with it, and rightly pulled it when they were not. If it is a case of fraud, I'm not certain what any of the other authors could have done, especially in a paper where the majority of the data are all micrographs. The lead author could have doctored/been selective with all images at every stage of presentation.
    These things are always quite sad. I hope they get to the bottom of this

  • PhysioProf says:

    I wouldn't read that much into Laurent's comment about whether he believed this particular work or not.

    Let's just say I have good reason to feel confident in my interpretation of his comment.

  • Pinko Punko says:

    Ha! Awesome.
    I hear ya.
    I had some long protracted comment debates with some internet trolls concerning particular scientific controversies, and I kept having to resort to statements like yours to maintain any level of anonymity. Trolls never take your word for it.

  • neurlover says:

    I work in a field where I feel like Laurent (a different one mind you), so I know what you mean, physioprof.
    Something similar happened in Hood's lab at Caltech years ago now -- a post-doctoral fellow faked data. Hood wasn't responsible directly, in the sense that he certainly didn't fake the data. In addition, it's clear that in many fields, it's awfully hard to check up on the faking, depending on how it's done. Much data that we see is processed already, and can be faked before we see it.
    But, I have a problem with how this type of malfeasance is being dealt with (both here, and in the Hood case). If we create a system where the PI suffers no consequences when a post-doc in the lab fakes data, we create no incentive for them to put safeguards in place to prevent the faking of the data. As others have alluded, there's clearly significant pressure on the junior investigators to produce the "result." It's not surprising that some will take the route of faking data (especially when the alternative is being thrown back on the street in china, disgraced). Hood & Buck have an obligation to put safeguards in place to prevent this problem, and we (as a society? scientific community? )have to incentivise them to do so.
    There's no secret that this kind of fraud occurs in big mega-labs with very high profiles, I would suspect, at least partially, because they are immune to the other consequences (like other people not replicating your work, and thus deciding not to fund your future grants).
    Ugh.
    (PS, real conspiracy theorists would say that one picks the least favored member of the group, and throws them on the sword. I am not a conspiracy theorist, so I don't believe that -- in any of these instances. But, I do believe that systems that fail need to be investigated)

  • According to this report in The Scientist, Harvard is investigating the retraction and LBB has asked the Hutch to look into two subsequent papers with ZZ as first author.
    http://www.the-scientist.com/blog/display/54406/. The comment following the piece is interesting (and, to me, a bit illuminating).
    And Pinko, I do feel bad for a senior author who gets caught in such situations. It is not always easy to be on top of all the data that come out of one's lab. But that is one's responsibility as le grand fromage. You are ultimately responsible for what comes out of your lab, especially if you sign off as PI-equivalent on it; tough crap.
    The author contribution (listed only in the retraction and not in the original) is the combination of a plaintive "How could I know he faked the data...I trust my guys...I feel so betrayed...poor put-upon me" and "He did it. It wasn't me... It was him, He's the rotten egg.....It's not my fault really" and is pathetic. Even if it was done in the spirit of protecting the other authors, this is not the way or place to do it. If ZZ is accountable, he should be held so by a committee that investigates the matter, not be sacrificed in the public domain by his ex-boss, who happens to have the clout of a Nobel Prize to prop up her credibility. The "hanging out to dry" comment is not meant to exculpate ZZ. It is meant to point out that leaders need to take more responsibility, especially when the shit hits the fan. There's plenty of time to assign culpability once the matter has been reviewed.

  • DrugMonkey says:

    There's no secret that this kind of fraud occurs in big mega-labs with very high profiles, I would suspect, at least partially, because they are immune to the other consequences (like other people not replicating your work,
    BINGO! And there is a structural problem of science here which I frequently attack. The fact that we now prioritize first-report sensationalism over replication enhances the chances that those who could replicate the work will simply choose not to because it wouldn't lead to a sufficiently high IF pub.
    It is not always easy to be on top of all the data that come out of one's lab. But that is one's responsibility as le grand fromage. You are ultimately responsible for what comes out of your lab, especially if you sign off as PI-equivalent on it
    The PI cannot always fact-check the data. As I've argued, it is not really feasible or necessary that the PI be intimately familiar with each type of data being generated in the laboratory. Making this a requirement would dismantle the large-group science that produces what we consider to be the best science, i.e. broad-ranging papers, these days. I have my criticisms but I think in some ways it would be a loss to dismantle these types of groups.
    What I do hold the PI responsible for is credulousness. The PI should be the most critical backstop of the lab, not the naif who accepts anything than supports his or her hypotheses. The PI should most certainly not be favoring the postdoc who can do no wrong, who's experiments always go well, who's results always match the PIs hypothesis, etc.
    That is why in these "investigations" I'd like to see more weight placed on investigating lab culture. One disgruntled trainee bitching about the PI is one thing. But in the bad lab cultures which encourage faking, one is going to hear the story from more than one individual.
    I'll also suggest y'all go read the Hellinga case discussion over at writedit's place if you haven't seen it yet.

  • DrugMonkey says:

    Nature Neuroscience's Action Potential blog is trying to get some discussion going on this.
    also, SciGuy has an un-cited statement attributed to UTMB, Zou's institution:

    Although Dr. Zou (an assistant professor in our department of neuroscience and cell biology) agreed to the retraction request, he says he's disappointed that the published work in Nature has been retracted. He's currently reviewing files and more than a thousand slides to assist in a review of the experimental work by Harvard University. Dr. Zou has not admitted any wrongdoing and is confident that the experimental results can be reproduced at UTMB. He hopes to begin that work as soon as possible. The original experiments were conducted in 1997 through 2001.

  • CC says:

    There's no secret that this kind of fraud occurs in big mega-labs with very high profiles, I would suspect, at least partially, because they are immune to the other consequences (like other people not replicating your work, and thus deciding not to fund your future grants).
    I'd have speculated the opposite -- that high-profile labs have results that are more likely to be tested and used in other labs, and more likely to be reportable when replication fails. Less prominent labs produce a lot of stuff that no one ever bothers to follow up, and even when someone does and can't get it to work, they're more likely to shrug and move on, or to berate their own grad student for incompetence.
    Plus, the mega-labs do strongly select for particularly, errr, driven students and postdocs...
    As for "hanging ZZ out to dry"- I don't think so. They stood by the work while comfortable with it, and rightly pulled it when they were not.
    Perhaps, but they certainly invented a new method of accusing one guy of fraud and exonerating another, without producing any evidence for either.

  • neurolover says:

    "Less prominent labs produce a lot of stuff that no one ever bothers to follow up, and even when someone does and can't get it to work, they're more likely to shrug and move on, or to berate their own grad student for incompetence."
    but, a condition of "this kind of fraud" is publication in Nature, which I'd guess has a reasonable probability of being followed up on. I'm arguing that "high profile" labs are more immune to consequences of failure of replication, presuming that both have published in Nature. And, that's 'cause it's pretty unlikely that a PI in a small lab would be able to survive the fraud of a junior investigator they worked with, whereas a Nobel Laureate can.
    I mean, I guess retracting papers is a consequence. But it seems like there should be more. For example, an event like this should be followed by an investigation of the "culture" of the lab, and a plan going forward of how a similar bad apple will be detected.
    don't know that all megalabs have this flaw, but, there are certainly meglabs where the PI's general attitude is ""Will none of the knaves eating my bread rid me of this turbulent priest?" (or this annoying control experiment that didn't work).

  • Pinko Punko says:

    Upon reflection, if it were not fraud, and it was sloppiness, or just skittishness at reproducibility, then LB should have been more up front about being responsible. The thing is there are a bajillion irreproducible papers out there that never get retracted, and the labs know it. This is just not good.

  • bayman says:

    "These questions are sufficiently complex and require such large amounts of data at high resolution that I have never considered them convincingly resolved in any system,"
    I heard a nearly identical comment on several nature papers that were retracted in a different field. The data was all faked and the guy basically said it would have no impact on whether he agreed with the overall conclusions of the paper or on his view of the field.
    Hmmmmm...What does this tell you? This "big science" from these big labs is shit science to begin with, retracted or not. It's sensational, career-manufacturing garbage. The data have no influence on what people think about anything, and as another commenter pointed out above, little of it can be put to rigorous testing by others lab in the field.

  • whimple says:

    Even if there's no harm because nobody believes the data, these labs still suck up a hugely disproportionate share of the funding resources.

  • DrugMonkey says:

    OMG! The first author, ZZ, apparently has jumped into the comments over at Nature:
    http://www.nature.com/news/2008/080305/full/452013a.html#comments
    you'll have to page down, they don't seem to have specific comment links...
    he hasn't heard of "lawyering up" apparently...
    he also takes a little swipe at his co-first-author...

  • PhysioProf says:

    ZZ:

    Yes, I signed the retraction letter and hope every scientist who is aware of problems with a paper will take similar actions immediately. However, I stand behind the conclusions of the paper and believe the experiments can be repeated. I am planning to do so. This will undoubtedly be daunting to a struggling junior faculty, and no one can guarantee success. I agree with the view that everyone who is on a publication should take full responsibility. Otherwise, stay in the acknowledgement.

    All I can say is, "Holy Shit!"
    This is even better than the Spitzer imbroglio!

  • Linc says:

    Imagine if Dr. Zou was not caught like many others! Years later, he probably will be the authority in the field. At that time, he would have power to put his name in the paper without taking any responsibility.
    I don't believe that Dr. Zou and Dr. Buck's other papers are clean. I don't believe that there is no misconduct. I don't believe that Dr. Zou is innocent. I don't believe Dr. Buck is innocent. Buck likely is the head of the conspiracy. Likely Dr. Zou is just a genuflector to the fallacious ecosystem of bioscience. Due to his bending, he got the reward from the boss.
    However, later on Harvard committee probably will cover all things up for Dr. Buck. To the system, Dr. Zou is dispensable, but not Dr. Buck. Let us watch.

  • Flameout says:

    "In other words, Laurent always thought the results were bullshit, and never relied on them in his own work."
    This is not specific to faked papers or high-profile papers. We should regard ALL published literature with a healthy dose of skepticism and a critical eye. Working models are constantly being verified and falsified - not because people are faking data, but because people are learning new things that need to be taken into account. I could say what Laurent said about every paper I have ever read - skepticism is in a scientist's nature, and should not itself be taken as an indication that a given paper "bullshit" or that the scientific literature, high- or low-profile, is worthless.

Leave a Reply