How long should a postdoctoral experience be from a training perspective?

Dec 11 2014 Published by under Uncategorized

In light of the recent release of the National Recent Council report The Postdoctoral Experience Revisited as well as my recent analysis of the age distribution of New and Early Stage Investigators, I have been thinking again about optimal time for a postdoctoral experience. From an analysis that we did when I was at NIGMS and subsequently shared through Drugmonkey, the median postdoctoral period among NIGMS grantees who received R01s in FY2004-2006 was 5.0 years. It seems likely that this period has increased subsequently.

Many factors contribute to the length of a postdoctoral experience including training needs, career options, other life factors, and so on. Here, I would like to get thoughts on how long a postdoctoral experience purely FROM A TRAINING PERSPECTIVE. In other words, how much time in a postdoctoral experience does a young scientist need to learn new concepts, techniques, and so on without regard to the need to accumulate publications, etc. in order to be competitive for their desired career path?

38 responses so far

  • Drugmonkey says:

    Concepts and techniques are not what makes one competitive for a faculty job.

  • Established PI says:

    It all boils down to how long it takes a postdoc to accumulate publications that will get them a faculty position. That time has stretched ever longer as the demand for glamour pubs grows in parallel for the demands by glamour journals for ever more experiments in order to get published. That reality, rather than how long one needs to be properly trained, is what is dictating the length of postdoctoral fellowships. Part of the problem as I see it is how much power we senior scientists have handed over to the high-profile journals, most of which are run by professional editors rather than practicing scientists.

  • datahound says:

    DM and EPI: I am trying to disentangle training needs from career needs. I realize that career needs are more important from a practical perspective. Nonetheless, as the community (perhaps) tries to improve the situation for scientists in training, it would be helpful to get thoughts on training benefits.

    Full disclosure: In a different era, I did a short postdoc even through I dramatically changed fields. While I am sure I would have benefitted from more experience, I had learned ~80% of what I would have learned eventually. But becoming independent did allow me to pursue my own ideas and to get credit for work that I likely would have done either way.

  • DJMH says:

    3-4 years. You're a faster learner now than in grad school, but you also have to have enough time to figure out your field to the level where you can write a grant. (I don't know if "being ready for grantwriting" counts as a training need from your perspective, or not.)

  • qaz says:

    0 years. The skills that one learns as a postdoc do not generally translate to the skills one needs as a faculty member. Unless you are explicitly switching fields, where you need 2-3 years to learn a specific new technique.

    Many fields ran very successful science without postdocs until the faculty jobs filled up.

    • drugmonkey says:

      Many fields ran very successful science without postdocs until the faculty jobs filled up.

      Absolutely. They had jobs in their late 20s, fresh out of a 4 year grad program. And they did fine, after all, these are McKnight's non-riff-raff.

    • Anonymous says:

      "Many fields ran very successful science without postdocs until the faculty jobs filled up."

      Exactly! I'm doing bioengineering in an ECE Dept. IF I wanted to specialize in biomedical signal processing and join an ECE Dept., I could skip a postdoc altogether. But to be competitive in Bioeng, I'll need at least 2-3 yrs.

      I can see the value of a 2-3 yr postdoc to learn new techniques or branch out into a somewhat different specialty, but any more time than that just seems like a needless delay to me. This is why I think the K99 needs to be re-thought. There's no way that s.o. w/4yrs experience as a postdoc needs 2 more yrs of additional mentoring/training!

  • neuromusic says:

    "I am trying to disentangle training needs from career needs."


    I don't understand how one even attempts to disentangle these things. But the best way to isolate one variable is to set the other one to "zero" so...

    "How much time in a postdoctoral experience does a young scientist need to learn new concepts, techniques, and so on in order to be competitive for their desired career path [::cough:: R01] WITHOUT ACCUMULATING ANY PUBLICATIONS?"

    • datahound says:

      While I understand your point, I am certainly not arguing that doing a short postdoc and not publishing is a good plan either from a training or career perspective. To push your argument in the other direction, I would not argue for doing a postdoc that completely overlaps your graduate subfield. This would potentially enable considerable productivity in terms of publication, but would have limited training value.

      A major motivation for this post is arguments that I have been involved in that postdoc periods are long now because of the complexity of science and training needs rather than career considerations. I tend to disagree with this, but I wanted perspectives from others closer to their graduate and postdoc experiences.

      • drugmonkey says:

        This would potentially enable considerable productivity in terms of publication, but would have limited training value.

        But again, this depends on the career goal. Some jobs will not care that you are a one-trick pony, they want you to be a demonstrated world expert in that one trick. More productivity is better and alternate training is a waste of time, in that case.

        arguments that I have been involved in that postdoc periods are long now because of the complexity of science and training needs rather than career considerations.

        HAHAHHAAHAHAH!!!!!! C'mon. You know better. This is just the excuse the greybearded make for their little exploitation scheme. Just like making grad students stay around for 7 years to get that Nature paper they "need". (gee, it's in your own best interest because people just like us over there at that U won't hire you unless you have one).

  • Noncoding Arenay says:

    Agree with qaz. But with the current situation as it is, doing a postdoc can rarely be avoided. That said, three years should be plenty. Four if one moves to a totally new field for their postdoc.

  • lurker says:

    For someone of your stature, and to be so in tune with the blogosphre and twitterati, you still show your age, Datahound, with such a rhetorical question of what's an appropriate length for postdoc training today. Your "Full Disclosure" had me ROTFLOL!

    Of course your cohort, the McKnight generation, had it great, what with your 2-yr postdocs! My PD mentor who is also of your cohort had a faculty position LINED UP for him as he was recruited during his PhD defense, and he "opted" to do a 1-year postdoc in a Nobel laureate's lab across the street from where his new lab was being built. To compare your PD training to today is like pulling a McKnight on how study sections used to be.

    Moreover, to equate the PD as a training period is a farce for those who work in BSD labs, which will soon be the only ones who can afford PD's if their salaries are starting $50K and they are the only ones getting their grants renewed. Only if your PI is Drugmonkey, or Potnia, or Odyssey (but definitely not CPP) might one get real training in how the sausage is made.

    As a former PD from a BSD's lab, there was no real "training". I floundered on my own, made countless rookie mistakes (that I learned on my own) and by brownian motion somehow figured out how to doggy paddle myself to a glamour publication. Had I just not figured it out myself, I would have been like the bulk of other PD's that didn't "train" themselves, and sank into "alternative careers" (not academia).

    I only escaped from the PD-dome because by luck one of my hunch experiments actually paid off and I managed to corral as many supporting workers (middle authors) to float up a glamor pub that eventually helped me land my own funds which is the only reason a school gave me a TT position. The true "training boost" I got from my BSD mentor was the pedigree and his cachet with the glamour pub editor.

    You are delusional to think in this day you can partition "PD training" from career advancement, when the game is all about funding and glamour that is 99% tied to number of pubs with high enough IFs. It's also delusional to set a "time limit" for PD training as proposed by the National Academy's report (yeah, more McKnight'ers commenting what they think is good for us), the perverse effect will be a new expectation that PD's now must do TWO 5-yr stints to get enough glamour pubs and cachet to move up to the next level, because the Greybeards refuse to leave and Want MORE, MORE, MORE (grant funds, supplementary data, and new Postdoc blood)!

    • drugmonkey says:

      You mad, bro?

    • datahound says:

      Lurker: I provided my "Full disclosure" to reveal some of the sources of my biases, not because I feel that my experience is transferable to the present day. Even in my day, I was very fortunate. I did my postdoc in a lab of a more-or-less starting but up and coming Assistant Professor in a field very different from what I had done my Ph.D. in. The lab was small (1 grad student and two other postdocs) and the PI spent considerable time in the lab. I learned a lot, about conceptual, technical, and (to some extent) career issues. I was fortunate to get a faculty position based largely on my graduate and earlier work which had led to a large number of publications.

      The point of my question was that discussions about postdoctoral issues tend to intermix training and career aspects in ways that make it seem that they are the same. From my perspective (in 2014), they are not the same.

  • DJMH says:

    I know that history is filled with successful scientists who didn't do postdocs, or anything like it, but I'd say a 3-4 yr postdoc is, for most people, a good thing. Switching directions is good for flexible thinking, perspective, and frankly maturity.

    You might ask, what has happened with the various people who have done the modern day version, gone from being grad students or very early stage postdocs to fellows at Janelia (and isn't there a Berkeley equivalent?) Not sure any of those people are soaring past their peers, especially after dividing by the denominator ($).

  • Rheophile says:

    I think the answer is, in many cases, zero. I have learned quite a bit about both science and academic sausage-making as a postdoc - but I probably could have learned lots of it as a fresh PhD in a faculty position, too. People who had bad graduate experiences or are switching fields can definitely benefit from 2-4 years of postdoc.

    The problem is that "training" level and the job market are inextricably linked. The question is, what standard are we going to hold new professors to? If faculty members learned some of what they cover in postdocs on-the-job, they would not accomplish as much pre-tenure. Right now, it seems like new faculty are expected to have plans to submit large numbers of solid grant proposals within their first year [with the expectations for "solid" increasing all the time]. This isn't a rule of nature - it's a statement about how long new faculty have been working in the field pre-appointment.

  • Morgan Price says:

    I think the majority of new PhDs don't have enough experience or a broad enough perspective to be fully independent scientists, and many of them will develop that after a couple years of further work. Or ometimes longer. But it isn't really about the supposed training, it's about the experience.

  • Established PI says:

    DH, there are really two separate issues here. One is how long a newly-minted Ph.D. needs to acquire the necessary skills to run their own lab. These skills include techniques, learning about a new field and system, and gaining additional experience or education in grant-writing, lab management, etc. While I know you would agree your PD was unusually short, my guess is that the typical postdoc would need 2-3 years for the training part and a little more to write up papers. There is no way anyone would need more than 5 years of training. Anyone working beyond that should be reclassified as a research associate and paid accordingly. The problem with the proposed shortened timelines is that one sometimes need a bit of luck to have everything fall into place in a five-year time period (or a little longer, while being paid as a research associate).

    A separate question is, what does a PD need to accomplish in order to get a job? For mid- to high-ranked universities or labs, this means getting one knock-it-out-of-the park glamour first-authors pub or a few very good first-authors pubs in very good journals. Having funding (K award, private) matters more the lower the university rank is (it is only a minor plus factor where I am and most new hires don't have it).

    The publication issue is huge since we have all somehow made ourselves slaves to the CNS journals (except for Michael Eisen), whose spiraling demands can delay publication for a year or more. Postdocs with fantastic work have to linger an extra year just waiting to satisfy unreasonable reviewer and editor demands so that they can get that ticket to a job.

    The cohort of serial postdocs includes people like Lurker (above), who I will assume are talented scientists who did not receive good mentoring, as well as a much larger group of people who will never, ever be competitive for faculty positions, either in the U.S. or (if foreign) in their home countries. We are not always honest with the latter and just pass them off to the next lab who needs cheap labor to keep the working flowing. This is the situation the proposed limit on postdoctoral training years is intended to address.

  • Joe says:

    2 years is enough to learn the techniques and get to know the people in the field. However, the real training they need is in how to write good and exciting papers and how to write fundable grants. That takes longer.

  • K99er says:

    I think a postdoc should seriously evaluate their position around year 2 or 3. It takes about that amount of time for the PI to decide if you’re one of their favorites who will be supported in terms of grant writing, networking/conferences, and learning the hottest technologies. By this time you should also know the answer to the questions, “will my boss fight for my project to get into a glamour journal?” and “will my boss write an awesome letter for me and make phone calls to search committees?” If you have this ideal relationship, then you could easily stay for a few more years and continue to build your competitiveness. If you don’t have this, then you should move on, because it’s probably not the right environment to help you get a faculty position.

  • Dave says:

    A separate question is, what does a PD need to accomplish in order to get a job?

    It's not a separate question. It's the only question. Nobody really gives a shit about training. Training is over once a PD gets a CNS paper out.

    For mid- to high-ranked universities or labs, this means getting one knock-it-out-of-the park glamour first-authors pub or a few very good first-authors pubs in very good journals

    That's original. Define 'very good journals'. I'm 99% sure you are referring to multiple (e.g. > 2) CNS second-tier publications with double-digit impact factors during a 3 - 5 year post-doc, as a minimum requirement.

  • Established PI says:

    Dave, the ideal is that the PD receive training that helps them obtain and then succeed in an independent position, not that any future employer care per se about whether they received formal training. They are looking at the end result. A CNS pub without clear evidence of the PDs ability to write well, formulate a competitive grant, give a good talk and be a good mentor and colleague still won't get you a good job.

    As for journal quality, the requirements depend on the institution and the field. You are, alas, correct that second-tier CNS pubs are generally expected at high-ranked institutions, but I have seen people get jobs at research universities with papers in other journals (PNAS, EMBO, Genes Dev, to name a few). A set of good but not glamorous pubs are also sufficient for a job at many undergrad-focused institutions, as long as they are coupled with strong evidence of commitment to teaching and conducting research with undergrads.

  • Pure training? Depends on the lab you're in, because the sky is the limit. I'd go with 3-4 years for the typical environment. In addition to having enough time to learn new experimental techniques, you really do need time to write fellowship and other proposals and to have the time to go through the review process.

    My postdoc was nearly 5 years- I could have left a year or two earlier, but I loved every minute of it and figured I could be a PI for the rest of my life- you cannot go back and be a postdoc. I also think I matured substantially in 5 years- and that maturity is something that had helped me handle a lot of the more challenging aspects of this job that no one ever trained you for...

  • AcademicLurker says:

    On the perennial glamorpubs issue, some subfields seem to be more glam obsessed than others. I landed my tt job at a highly ranked department (# 2 in the country in terms of NIH $$ at the time I joined) with no CNS papers, first author or otherwise.

    This is probably a topic for a separate post, but it would interesting to see a breakdown of "you absolutely must have a CNS paper" sentiment by subfield.

  • Anonymous says:

    Seriously, the NIH should consider grants specifically for those w/in 5 yrs of their PhD. This might encourage universities to take a chance on really promising early-career folks. But unlike the K99, the unis here absorb all the risk, none of the shenanigans reported by some K99 awardees would ensue.

  • drugmonkey says:

    There would have to be some stick along with that carrot.

    • Anonymous says:

      You really think that unis would not be willing to bet on promising very early career folks if they knew there was a pot of money that only they had access to? You and others have said that early career folks without K99s are getting hired, so how is what I'm proposing not a better deal?

      • Drugmonkey says:

        i think the Unis would still see it as a risk, with no change in their evaluatin of the chances of the untried youngster versus the mid career with evidence of funding.

  • Ola says:

    I'd say it totally depends on the grad' school experience. For those coming from "big labs" where there's a lab manager who does ordering, and the PI travels a lot, and the students are essentially being taught by the post-doc's, when those folks get their PhD they're frickin' clueless about what it takes to be a PI, so they'll have to learn all that stuff as a post-doc (and they'd better not pick another glam-lab and repeat the cycle of non-exposure/non-learning). The same applies for folks coming into a post-doc' on a J1 or H1B visa from overseas, who have no clue how the "system" works here in the US. All these folks are better served by doing at least 2 post-doc's, preferably at least 3 years each, with a minimum of 5-6 years. That's what it takes to learn the ropes.

    The exceptions are the kids who had excellent pre-doc' mentoring which included some of the stuff you're supposed to learn as a post-doc. How to supervise an underling, how to think up your own projects and craft grant proposals, how to "play nice with others". How to teach. I don't see many incoming post-docs who have all these skills already, but the ones that do tend to come from smaller labs with younger PIs who spend more time in the lab' and involve their students in the day to day business of running the lab. These exceptions typically might require only a single post-doc' of 3-4 years before they're ready to move on up.

  • Established PI says:

    Given the vast oversupply of Ph.D.s chasing PI positions, we really don't need remedial postdoc positions. Students need to be smart about acquiring skills and knowledge and then double down as soon as they become postdocs. It is sometimes necessary to switch postdoc labs early on (bad fit, personal reasons) but doing two postdocs should be the exception, not the norm.

  • Crystaldoc says:

    I roll my eyes a little bit when I hear complaints like those of Lurker, bitching about the training environment of their Glamour lab PI. The postdoc training of value is often figuring out on your own how to lead a project, trouble-shoot it when things don't work as planned, direct a (mini)program, write up your work in a compelling way, and I'd bet that most who succeed in academia do a lot of that on their own as postdocs. That "floundering" IS the training. The benefit of the postdoc (as opposed to straight into PI-dom), is the chance to figure it out under a slightly less stressful environment where you don't yet have the continued employment of a lab riding on your shoulders. Also, the glamour-cachet of well-known advisor when it comes time to publish is nothing to sneeze at. It gives you a big leg up on competitors with otherwise similar skills and track records going for that same job.

    15 years ago in a synthetic chem/biochem lab, the chem postdocs were outta there in 1.5-2 yr and the bio ones in more like 3-5. This had not so much to do with training needs as for the opportunities at that time for synthetic chemists in industry. IMO 1-2 years should theoretically be enough training to learn the techniques kind of things if moving to a postdoc in a related field but with some new approaches/techniques. But, if you are leaving at that point, you are missing out on the 50-75% of the potential training that comes from writing it all up and shepherding the work through the publication process, possibly down a few tiers through multiple journals. More opportunities to work on writing grants and fellowships. More opportunities to give talks and seminars. More opportunities to write reviews. More opportunities to practice your supervisory skills on undergrads, grad students and techs. 10 years ago I completed a 6-year postdoc (experimental bioscience) and felt and still feel that every year of it was well-spent and involved training. And like Lurker, almost all of it was self-training.

    • Anonymous says:

      "More opportunities to work on writing grants and fellowships. More opportunities to give talks and seminars. More opportunities to write reviews. More opportunities to practice your supervisory skills on undergrads, grad students and techs"

      You have the rest of your life to do this as a PI. It's called on the job training. (With the exception of giving talks/seminars, which you should definitely have experienced to some extent as a grad student.)

  • Noncoding Arenay says:

    "More opportunities to work on writing grants and fellowships. More opportunities to give talks and seminars. More opportunities to write reviews. More opportunities to practice your supervisory skills on undergrads, grad students and techs."

    ^ this!

    I had excellent graduate mentoring where I did literally each of the above mentioned things several times. It is crucial that all grad students get this kind of exposure or your mentor is doing it wrong. A postdoc should entail doing "more" of the above and learning a few new techniques and/or branching out into a new subarea (~3-4 years total). I think beyond that no matter how long you stretch your postdoc, there's nothing more you "train" in except a repeat of the above and probably a few more publications. Essentially you enter a holding pattern at that point.

  • qaz says:

    The thing that separates successful PIs from unsuccessful is management - Are they able to actually manage the personnel in their lab? Are they able to manage their time? Very very few postdocs are getting training in these management skills. I love all the excuses about PDs being trained in "writing grants" and "opportunities to give talks and seminars". Ha! The grants being written by postdocs are fellowships and include a very large "mentor-quality" component and rarely translate well to R01s. Even K99s don't translate that well. Being able to give a great talk or seminar is not going to make or break PI success. (Yes, DH has shown that those who get K99s tend to do well as PIs, but I still don't see any evidence that the K99 is anything more than a marker of an on-track candidate.)

    The fact is that we have a lot of data over the last 100 years of which fields demanded postdocs when. There are three hypotheses:

    1. That postdocs are training for fields that cannot be taught in five years. This predicts that more complicated fields should have longer postdocs. It also predicts that fields should never shrink the length of the postdoc. (Fields don't get less complicated over time.)

    2. That its a time-issue. The culture simply says that postdocs are the way of the world. (Don't forget that one of the reasons Einstein did that patent thing was because there was no postdoc position for him to do while he was waiting for faculty jobs to open up.) This predicts that there should be a trend in postdocness. It predicts a strong correlation across fields in terms of length of postdoc.

    3. That the postdoc is a holding pattern for limited faculty spots. This predicts that the length of the postdoc should be correlated to the number of faculty positions. An interesting prediction of this is that the length of the postdoc should be negatively correlated to alternative career options.

    I think the data is very strongly in support of number 3. If one looks at computer science, which has gone through several boom and bust cycles of alternative careers, one can see that the ability of top people to go straight to faculty is negatively correlated to the availability of other options.

    The fact is that the speed at which a new faculty member finds its feet is not correlated to the length of the postdoc that the person has done. From a straight training perspective, postdoc is unnecessary, unless you need to get a new technique under your belt.

  • Newish Professor says:

    As described above, there are several things that go into making a potentially successful PI. But I think that depending on the field, there are different things needed. I don't know about other fields, but in Biological Sciences I think that a postdoc is a necessary step. And I personally believe that training is a field or sub-field different than you were in as a grad student is absolutely recommended.

    Grad school should be where you build your toolbox of techniques and knowledge, so you can walk into any other lab and get started with the basics, right away. Sure there are specialized things you will need to learn in your postdoc, and you should learn them, but you should have a technique and assay basis walking in the door to get you started.

    On the postdoc side, I think that 3-5 years is needed. A combination of:

    1. gaining more scientific skills
    2. learning the background and current nature of the field you are in
    3. knowing the people in the field and making sure they know you (a critical thing for your future as a PI)
    4. confidence in your scientific direction

    As a postdoc, you are not going to learn how to run a lab, how to do a budget, how to write an approvable IACUC or IBC protocol, deal with the psychology of frustrated grad students, managing multiple projects by several people all at the same time, dealing with colleagues and collaborators and dealing with the stress that comes with being responsible for the well-being of people in your lab.

    You may think that you are ready for all of that but its very different when you start your own lab and your last name is on all the glassware.

    The postdoc time should be enough time to get yourself comfortable in a field, know and be known by the people in the field, and accrue enough data and directions to get you going in your own lab. I don't think you should have to have a K or other grant to get a job (even though my institution won't hire you if you don't). But I think you should have a model that you can work over to answer a set of interesting questions. Then once you get your name on the bottles, its up to you to get to the next steps.

    As someone who got the job without any Glamour pubs and is at a top tier medical research university and has been on our department search committee for 3 years, I can tell you more than anything, the step from postdoc to getting the job is more about the model you are working on and "do I want this person in my department".


  • FemalePhysioProf says:

    I am reading this thread with great interest and worry. My worry is that the question centers around 'time elapsed' instead of knowledge and skills gained. The reason for the this worry relates to another often discussed issue, which is that prime child-bearing years often overlap with postdoctoral training. So, if we focus on time elapsed without consideration temporary parental leaves of absence, we risk (inadvertently) reinforcing the idea that being a biomedical scientist is incompatible with being a parent. We may also amplify the influence of unconcious bias on decisions made by trainees (to postdoc or not to postdoc) and by mentors (if i've only got this great scientist for 5 years and 1 of them is may be at reduced productivity due to becoming a parent).

    So, DataHound, I challenge you to reframe the question or risk a severe unintended consequence of decreasing the proportion of working research scientists who are women. How about this? What are the knowledge and skills that differentiate a scientist with postdoctoral training from a newly minted PhD?

    • datahound says:

      Thanks for your comment. The issues with regard to parenting are important additional concerns with regard to the increase in the duration of the "training" period. I agree with your reframing, but as you saw from many of the comments, career issues appear to dominate knowledge and skill issues in the minds of many.

    • qaz says:

      This is a very important point that FPP brings up.

      A lot of the programs that are time-limited include opportunities to "explain" or "excuse" gaps as a way of trying to accommodate parenthood. But like the word "alternative" in "alternative career", that only serves to emphasize the non-standard nature of these explanations and can serve as a marker of differentness. We really need to abandon all of these time-limited issues (ESI, K99, etc.) and address people by their position in the pipeline (postdocs, junior faculty, etc.) Then people could progress at their own pace and contribute in their own way.

      Unfortunately, in response to your knowledge and skills argument, as I noted above, the evidence is that postdoc is a holding pattern in which people try to get more skills, not a necessary training program that happens to take a certain amount of time.

Leave a Reply