At what career stage do scientists typically do their best work?

Anecdotally, many great scientists had their best ideas when young. For instance, in ecology, G. F. Gause wrote his classic The Struggle For Existence when he was an undergraduate (!) See this post and comments for some other “prodigies” in ecology and evolution. But also anecdotally, there are scientists who did tremendously important work late in their careers. Anecdotes about exceptional people are, well, exceptional. When does the typical researcher peak?

We now have an answer from economics (ht Marginal Revolution). According to the linked paper*, academic economists who obtain tenure exhibit peak productivity at the time they’re tenured, with productivity falling 25-30% in the next five years and declining steadily for a decade thereafter. That’s true whether productivity is measured as number of publications, or as proportion of publications that become highly cited. Conversely, the proportion of publications that end up being poorly cited rises steadily after tenure, while the proportion that end up being highly cited declines. And this does seem to be connected to tenure per se, as the authors show by exploiting among-economist variation in the timing of tenure.

I’m now very curious whether the same would be true in ecology. I’m not really sure–what do you think?

These data make me feel even better about my goal to be the Jamie Moyer of science, since apparently that would be a major accomplishment if I somehow managed it. 🙂 And I’m now even more impressed with scientists who’ve done outstanding work late in their careers.

So why do researchers peak around the time they get tenure? The authors speculate that obtaining tenure induces academics to rest on their laurels rather than taking intellectual risks. I think that’s only one of many non-mutually-exclusive explanations, and doubt it’s the main explanation in most cases. For instance, post-tenure academics have more administrative and service duties, leaving them both less time to write and less time to read and think. As Peter Adler has noted, it doesn’t get any easier after you get tenure–it gets harder. I suspect that the rate at which one comes up with good ideas post-tenure is limited much less by inclination to work hard and take intellectual risks and much more limited by the factors Peter identifies. Finally, many of the intellectual “risks” one might take post-tenure (where “risk” is broadly defined as “trying new things”) don’t result in papers. Writing a book, for instance.

So what can you do to make yourself an exception to the statistical rule that researchers peak at tenure? Here’s some advice from Hope Jahren, and here’s some from Ray Huey.

A final thought: the ecologists I can think of who did great, influential work late in their careers also did great, influential work early in their careers. Think David Tilman or Jim Brown, for instance. I’m not thinking of any “late bloomers”, ecologists who only did great, influential work late in their careers. Can you think of any? In evolutionary biology, how about John Maynard Smith? His career change from engineering to biology only partly accounts for the lateness of his great work on game theory and the major transitions in evolution. As far as I know, that late work is far more important than his earlier work on Drosophila. There’s George Price too, though he’s an unusual case because he bounced between jobs and fields for many years before doing the great work for which he’s now remembered. In a funny way, he was both a late bloomer (as a scientist) and a prodigy (as an evolutionary biologist specifically). A quick search reveals various suggestions for late bloomers in other scientific fields–see here, here, here, and here. But many of them seem to have been “late bloomers” only in the sense of “not prodigies”, which barely counts. Yes, it is unusual for someone who struggled as an undergrad to go on to a great scientific career, like, say, Craig Venter. But Venter started making an impact as soon as he got his first faculty position, so he didn’t bloom that late. For purposes of this post, I’m more interested in really late bloomers–people who spent many years as scientists before doing really outstanding work for the first time late in their careers. Bonus points if the great work wasn’t in large part an accident, as with Alexander Fleming’s discovery of penicillin.**

p.s. On another note: the data from the linked economics paper are difficult to reconcile with the view that senior academics easily get papers accepted and highly cited just by virtue of being senior.

*Which I’ve only skimmed, and so can’t fully vouch for. For instance, another commentator notes that the paper may not adequately allow for “Sleeping Beauties“–papers that only become highly cited after a long lag.

**Darwin doesn’t count. Yes, he was 50 when he published the Origin. But that was because he put off publishing for many years. He’d developed most of the major ideas and lines of evidence in the Origin by 1842, when he was only 34 and was just starting to make a name for himself in science.

35 thoughts on “At what career stage do scientists typically do their best work?

  1. > I’m now very curious whether the same would be true in ecology. I’m not really sure–what do you think?

    I think it’s probably highly field dependent. I don’t know enough about economics to compare to ecology, but I’d bet that ecology peaks later than, say, mathematics. What do YOU think?

    Also, do you think it’s a shame that professors take on all that other commitment after tenure? I mean, if you have someone doing amazing work, wouldn’t it be better for science to try to get as much other stuff out of that person’s way and let them keep doing amazing work?

    • Yes, mathematics has a reputation as a young persons’ game, though even there there are exceptions like Paul Erdos. Like you, I suspect that ecologists peak much later than mathematicians. But even in mathematics, I’m reluctant to draw broad conclusions about the entire field from anecdotes about exceptional mathematicians. I bet somebody has looked at data on this in mathematics.

      It’s also possible that averaging across investigators conceals interesting among-investigator heterogeneity. For instance, I could imagine that in some fields there might be an association between the age as which someone does their best work and the excellence of that work. Maybe in mathematics the very best people tend to also be those who peak early, while most people peak lower and later?

      “Also, do you think it’s a shame that professors take on all that other commitment after tenure? ”

      Well, if the more senior profs don’t take on those teaching and admin commitments, then junior profs get stuck with more teaching and admin. That seems suboptimal.

      “I mean, if you have someone doing amazing work, wouldn’t it be better for science to try to get as much other stuff out of that person’s way and let them keep doing amazing work?”

      There are various ways this can happen. Research chairs that come with reduced teaching and admin loads. Programs like the Howard Hughes fellowships, the holders of which are released from teaching and admin. Independent research institutes with no students and so no teaching. The NSERC Discovery Grant program has something called Accelerator Supplements, which are big Discovery Grant top ups given to investigators whose research programs are judged to be poised for major advances. There are some universities that assign differential teaching loads to faculty based on their research productivity. And you can have some people specialize on research while others specialize on teaching. Harvard of course assigns a lot of the teaching to junior fellows–the equivalent of graduate TAs and postdocs–rather than faculty. And here at Calgary and various other universities, some of the faculty are teaching faculty, which means that on average the other faculty have lower teaching loads than they’d otherwise have.

      There are some challenges and drawbacks associated with each of these, of course. For instance, assigning differential teaching loads to people based on their research productivity runs a high risk of creating resentment within a department.

    • I wonder if it would partly be a function of how much knowledge and experience one needs to be insightful and creative in a field. Admittedly, I don’t know what qualities a mathematician needs to create highly progressive work, but I speculate you need more experience to produce that kind of work in ecology, given the complexity and variability in the systems ecologists work with.

    • Yes, in referring to Fleming’s discovery as accidental I freely admit I’m just repeating the standard potted history. But I wouldn’t be surprised to learn that the standard potted history is oversimplified. As you note, chance favors the prepared mind.

      • Yes, that’s what that piece about Fleming and penicillin (see citation 4, below) was getting at – chance favors the prepared mind. Also, being open to possibilities that aren’t anticipated, and several other key points.

        Re creativity and innovation in science, there’s a lot of really interesting literature and commentary out there that (generally speaking) indicates experience/expertise is only part of the equation. Rubbing shoulders with a wide range of ideas in one’s discipline, grappling with problems/questions at the limits of one’s knowledge (including applying expertise to peripherally related questions), and engaging deeply in another way of thinking all appear crucial (see citations 1-3, below).

        Selected references:
        1. Creativity/Innovation in science: A lot of the work done by Robert Root-Bernstein, such as his look at what predicts Nobel awards (“Arts foster scientific success.” Journal of Psychology of Science and Technology, 2008) or his chapter in the International Handbook on Innovation (“The Art of Innovation: Polymaths and Universality of the Creative Process.” Elsevier, 2003).
        2. Loehle, C. 1990. A Guide to Increased Creativity in Research: Inspiration or Perspiration? BioScience, Vol. 40, No. 2: 123-129. Includes recommendation to investigate something “new” approx. every 7 years to maintain high motivation/innovation/productivity, along with a number of other interesting prescriptions based in science history/research examples.
        3. Andreasen and Ramchandran. 2012. Creativity in art and science: are there two cultures? Diaologs in Clinical Neuroscience. Excerpt from abstract: “The findings give no support for the notion that the artists and scientists represent “two cultures.” Rather, they suggest that very gifted artists and scientists have association cortices that respond in similar ways.”
        4. The back story on Fleming and penicillin: Root-Bernstein, R. 1989. “How Scientists Really Think.” Perspectives in Biology and Medicine, vol. 32 473-488. Reprinted as chapter 7 in Smith, R. 2012. Scientific Work and Creativity: Advice from the Masters. Citizen Scientists League. Also reprinted in that book (as chapter 10 and 15) is another take on the penicillin discoveries and developments: Beveridge, W.I.B. 1957. The Art of Scientific Investigation. W.W. Norton: 27-40, 82-95.

    • I have a couple of times had a really fun lab meeting by reading a couple of papers on creativity (including your Loehle reference) and then talking about where scientific creativity comes from. We seem to always come around to how sleep and exercise factor into it too.

      When I was a PhD student at Arizona, Art Winfree taught a course on how to be creative. He claimed it was a trainable skill. I never took the course. I definitely regret not taking it.

  2. I’m no neurobiologist, but it seems reasonable to wonder whether there isn’t something physical/evolutionary playing at least some role. As a 20-something grad student, the brain is just firing away constantly, physical energy is high, so ideas flow and all-nighters to develop them before they flame out are no problem. The brain is no longer in adolescent overdrive (not conducive to deep reflection!), and enough knowledge and skills have been gained so you know what to do with an idea when it comes. If there’s anything to this, the explanation for the age of peak athletic performance is maybe not so different than the explanation for peak intellectual performance, even if the latter peak is a smidgen later. Until recently in evolutionary terms, human life expectancy didn’t extend much past tenure, so selection for post-tenure performance wouldn’t be too strong. The first instinct of we the post-tenured might be denial (I’m smarter than ever!), but denial is the well-known first response to loss…

    • Hmm…could be. Hard to distinguish from other possibilities like increasing admin and other obligations and so reduced time to read and think.

      Re: peak athletic performance, worth noting that professional chess players also peak before 40 on average.

    • Mark,

      Speaking as somebody who is over 50, ahem, I prefer to think of it as a trade-off. There are a number of studies that show cognitive function declines continuously from the 20s. Most of these are measuring fairly straight forward things like calculation speed, short term memory, etc.

      Yet if you really believe all aspects of cognition decline from the 20s on, shouldn’t most of the CEOs of major corporations be in their 20s (or 30s). Shouldn’t we be electing really young presidents? Shouldn’t people in intellectual fields like academics or programming retire earlier than people in rote jobs (when in fact the opposite happens).

      I think there has to be value behind experience – you could even call it wisdom – about thinking through complex situations that increases over time. Or at least so I tell myself.

      I think this could be the explanation why raw brainpower fields like math peak early but complex, domain knowledge required fields like ecology peak later.

      • I definitely agree that accumulated knowledge, broad vision, and wisdom accumulate with time, and that makes us better at some things, but something is also probably lost – that’s the tradeoff (totally agree). So, what are we better at? Maybe: guiding younger scientists to achieve their potential, synthesizing broad topics or across topics, leading big important projects (but maybe not the most field-altering), figuring out how to solve problems. What’s lost? Something that’s hard to put your finger on, but that generally leads to bold new ideas, angles on things that come out of “left field”, etc. Most CEOs are not in their 20s or 30s, but neither were Google, Apple, Facebook, or Microsoft created by someone in their 50s (or even 30s for that matter!). In some fields, I suspect that accumulated wisdom and experience counts for relatively more: a medical doctor’s ability to diagnose unusual diseases, a lawyer’s ability to draw on past cases, a politician or businessperson’s accumulated connections and capital.

        Also, many if not most new shiny and influential scientific contributions by young people no doubt were facilitated to a major degree by the wisdom and guidance of senior scientists (e.g., Hutchinson for MacArthur).

      • @Brian and Mark:

        I have nothing to add, I just want you both to keep talking. It’s a pleasure to follow along with this exchange.

      • I think cognitive decline may be partly related to how you spend your time. Many American adults have routine jobs, and watch TV after work. However, those in academic jobs who are constantly exposed to different types of problems, and are challenged in their thinking may not experience the same amount of cognitive decline.

        Then, F.S., Luck, T., Luppa, M., Arélin, K., Schroeter, M.L., Engel, C., Löffler, M., Thiery, J., Villringer, A. and Riedel-Heller, S.G., 2014. Association between mental demands at work and cognitive functioning in the general population–results of the health study of the Leipzig research center for civilization diseases (LIFE). Journal of occupational medicine and toxicology, 9(1), p.23.

      • I’m just going to add that maybe it’s not age at all, but familiarity with the subject. I think that once you’ve been seeped in a topic/discipline/career for too long, it’s harder to see new angles, to be creative, to make connections others haven’t seen. So Brian’s age doesn’t matter a smidge; it’s only how new to ecology he is that matters. This might explain why those in some fields tend to peak earlier. Mathematics, for example, starts being taught in grade school. Most ecologists don’t learn a thing about it until college (or later).

      • @Margaret:

        I don’t know whether I’d say age per se doesn’t matter at all. But I agree that there may well be some optimal intermediate level of familiarity with the field. You need to know enough about the field to contribute, but not be so steeped in it that you can’t contribute creatively. That’s why Ray Huey recommends that people stay fresh by switching topics.

    • The bio-based idea leads to a testable-ish hypothesis (as much as you can put numbers on any of this): women should tend to peak earlier than men.

    • Presumably that’s why the analysis in the linked econ paper is restricted to economists who obtain an academic position. And unless I misunderstood or missed something, the analysis is restricted to people who got tenure.

      • Ah, ok, I’m with you now.

        Yes, insofar as new faculty have already bloomed to some extent, they’re starting from a high base and so have less scope to bloom later. On the other hand, they also have less scope to bloom around the time of tenure. So I don’t know that this is a full explanation for the lack of late bloomers.

        I suspect that late bloomers tend to be people who switch fields or bounce from topic to topic. One line of evidence for this is that people who make important contributions throughout their careers–who peak early and then remain at a high peak–often are people who switch from topic to topic, or who maintain several quite different lines of research. Jim Brown and Dave Tilman, for instance. On this view, the only difference between a late bloomer and someone who peaks early and then maintains for a long time, is whether or not their early work happens to be outstanding work.

      • @Margaret:

        “+1 Survivor bias”

        Except that, as I commented earlier, I don’t think that’s the full explanation here. As I understand it, the linked econ paper is looking at variation in peak timing among the survivors. How does survivorship bias–some people don’t get academic jobs, or don’t get tenure–explain why people who do get academic jobs and get tenure peak around the time of tenure as opposed to earlier or later? Am I being very dense and misunderstanding what you and standingoutinmyfield mean by survivor bias? (if so, apologies)

      • I think you’re right about the econ paper. It’s looking specifically at peak timing among survivors. That part’s fine. You just posit in your post about why you don’t have people who just peak late. And I think survivor bias probably answers a good part of that question. If you’re going to peak late, you probably won’t make it past the initial hurdles to even get to tenure. Maybe it’s interesting that very few people do quite well (and get tenure) and then plateau and then “peak” later in their careers. But I don’t think it’s *that* interesting, as it’s a very specific trajectory and there’s no reason to think that people would follow it very frequently.

        So I think we’re all agreeing here, just using different words.

      • Ok, thank you for the clarification Margaret, I’m with you now, this all seems reasonable to me.

        Though getting back to your earlier comment about variation in career trajectories among fields, I suspect there are fields in which late bloomers are more common than in ecology. So I’m perhaps a bit less inclined than you to write off the rarity of late bloomers in ecology as not all *that* interesting. I don’t think it’s an *inherently* or *obviously* rare or odd career trajectory.

        For instance, Wikipedia tells me that founding father of sociology Max Weber’s late work was his most influential (though his earlier work was influential too–he seems to have started from a high base and then kept climbing):

        The optimal intermediate level of familiarity with one’s field may well vary considerably between fields…

  3. I suspect that following tenure, some academics feel less pressure to do highly citable work. The emergent prediction is that quality measured by citation metrics falls, but quality measured by more thoughtful metrics remains high or increases.

  4. In terms of ecologists who did much more transformative work later in their careers (well past tenure), two people come to mind from the paleo side of things: Steve Jackson and Jim Clark.

    Steve’s earlier works were solid but classic paleo studies, while is more recent efforts have been a series of wonderful, synthetic, theory papers. Even Jackson & Overpeck (2000) came five years after he hit tenure, which I’d argue was the launch of his most exciting works.

    Jim Clark, similarly, got tenure back when he was still doing mostly Quaternary fire work, and much later went on to do some really different — and influential — work on global change and disturbance ecology, including on Bayesian modeling and incorporating individual variability.

    • Interesting suggestions. I don’t know Jackson’s work at all, so can’t comment on him. Jim Clark seems like a good suggestion of an unusually “late bloomer” to me.

  5. Pingback: Friday links: statistical significance vs. statistical “clarity”, philosophy of science vs. cell biology, and more | Dynamic Ecology

Leave a Comment

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.