Scientists today don’t publish any more than they did in 1900

No, that’s not a #slatepitch or deliberate provocation. It’s hard data.

A little while ago I compared the cv’s of this year’s ASN Jasper Loftus-Hill Young Investigator Award applicants to those of past winners. To my surprise, I found that top young ecologists and evolutionary biologists today are not publishing more papers than they did decades ago. Rather, they’re just publishing with more co-authors.

Turns out that my findings generalize to all scientists since 1900! In the linked paper (open access), Fanelli & Larivière analyzed WoS data for 40,000 scientists across all fields of science from 1900 to 2013. If you correct for increases in average number of co-authors per paper, or only count first-authored papers, scientists today are not publishing any more on a per-capita basis during their first 15 years of publication activity than they were a century ago. And no, the answer doesn’t really change if you only look at post-1980 data, or if you only look at scientists from certain countries. To which, wow. I would not have guessed that.

I’ve been trying and failing to think of a reason the results could be a totally misleading artifact. I haven’t come up with anything. Can you think of anything? (Note: please read the paper before answering. There’s no point in speculating on the answer to this question based only on my bare-bones summary or the paper’s abstract.)

It is true that more scientists are publishing at least occasional papers now than they used to. Back before the 1960s or so, only a minority of American college and university faculty ever published. Today a substantial majority do. So it’s possible that the variance of the distribution has increased, even though the central tendency hasn’t budged. Maybe these days there are more people publishing only occasional papers (rather than none at all), and also more people publishing scads of papers. It’s hard to tell if this is true from eyeballing the figures in the paper, but it looks like it could be. Which if so, might explain the widespread mistaken impression that everybody is publishing more these days. Maybe we all pay attention to those rare people who crank out scads of papers, and then mistakenly think of those people as representative.

I’m also curious about the extent to which changes in number of co-authors per paper represent changes in authorship practices, vs. changes in how science is actually done. Are we all truly more collaborative these days than we used to be, and doing work that people could not have done on their own back in less-collaborative times? Or was everybody collaborating just as much back in the day, but not giving one another co-authorships for it? Here’s one data point from Meg, suggesting that the only thing that’s changed is authorship practices.

So, if people aren’t actually publishing more these days, how come everybody thinks they are, and thinks pressures to publish are increasing? My “misinterpreting increased variance” hypothesis is one possibility. Another possibility: maybe everyone mistakenly assumes that, because people now have stronger incentives to publish more*, that they actually are publishing more. I’ve made that mistake myself in the past–assuming that, because people have strong incentives to behave in a certain way, that they actually do behave in that way. The authors suggest a few other hypotheses. For instance, perhaps scientists today have increasing demands on their time, and so feel pressure to maintain the same publication rate while allocating less time to research and writing. Or, perhaps the amount of time and effort required to produce a paper’s worth of science has increased, so that scientists are allocating more time to research and writing than they used to but only producing papers at the same rate.**

Anyway, consider “everyone is publishing more these days” another busted academic myth. A reminder, if one were needed, that things everybody “knows” often aren’t true.

*Do they, though? For instance, do you now need many more publications to get a grant than used to be the case? Do you now need to have more publications than you used to in order to have good odds of getting hired for a tenure-track job? Do your odds of being tenured or promoted, or your salary, increase more steeply with your number of publications than used to be the case? I dunno, maybe the answers are “yes”, “yes”, and “yes”. But I wouldn’t assume that, and I’d be curious to see data. I’m sure it’s out there.

**Maybe in part because people are increasingly wasting time submitting to, and getting rejected from, highly selective journals? Then again, maybe not: 75% of all published papers are submitted first to the journal that eventually publishes them. Now, maybe that number was even higher in the past. But even if it was I find it hard to believe that time spent revising and resubmitting rejected papers is a big reason why scientists today are not publishing more than they did decades ago.

21 thoughts on “Scientists today don’t publish any more than they did in 1900

  1. Interesting and counterintuitive – as a personal example, my father, who was a plant pathologist (research and extension) only published five papers in his career – he retired in 1980. My publication list overtook his in 1981 (the year after I had finished my PhD).

    • Yes, counterintuitive indeed. I’m still wondering if there’s not some artifact. Somehow, data from the early decades of the 20th century only include the scientists who published at the highest rate, while today’s data include everybody? But I can’t see a reason why that would be the case. Especially not if you restrict attention to data since 1980.

  2. Something’s got to give here. According to something cited in one of Brian’s posts, the doubling time of the literature (total # papers I assume) is about 10 years. That has nothing to do with number of authors per paper. It is necessarily true that:

    Total # papers = # first authors * (# papers / first author)

    Is it possible the number of people writing papers doubles every 10 years? Seems extremely unlikely, but if # papers / first author is not on the rise, then it must be true.

    (Unfortunately I did not take the advice of first reading the paper before weighing in – tall order on a busy day! – but I think the comment is pertinent nonetheless)

    • Mark my recollection is that the data is pretty iffy, but that the # of scientists is doubling about every 10 years, while the # papers is doubling more like every 7 years. Which suggests papers/scientist is increasing at a modest exponential rate. Of course that gap could be explained entirely by modest exponential rate of increase of authors/paper.

      • Wow! So, the expanding literature is largely about the expanding number of scientists. Not what I would have guessed. It’s like there’s a per capita law of constant final scientific yield.

    • Brian sort of beat me to it, but a few comments on what might have to give here:

      -The linked study is only counting people who published at least two papers, over at least a 15 year period (so if you published 27 papers, but the last one was 14 years after the first one, you don’t count). And for those people, it’s only counting papers published in that first 15 years of publishing. So total number of papers could be increasing in part because there are more and more people publishing these days who don’t meet those criteria. Say, more grad students and postdocs who publish for a time but who don’t get faculty positions and so don’t publish for at least 15 years. And maybe scientists these days are now more productive in the latter part of their careers (after the first 15 years) than they were decades ago.

      -The number of scientists publishing English-language papers (which is all that WoS indexes) has gone up since 1900. Remember that some of this growth is comes from growth in non-native speakers publishing in English. And some of the growth comes from the fact (noted in the post) that a greater fraction of academic scientists are publishing these days than used to be the case many decades ago.

      No idea about the relative importance of these various factors.

  3. I guess put me down in the unsurprised category. Not so much because I expected these results, but I just don’t think they’re that relevant to the conversations most people are having for two reasons:
    a) individual productivity is lognormally distrbuted. This means that the average is driven primarily by the level of productivity of the top few percent (most of the numerator in the average) and the number of individuals in the low productivity tail (the denominator in the average). In short the average says very little about the “typical” experience. In particular the fact that the average of # papers/first 15 years as first author has remained constant at about 10 really only says that the top performers have always done about 100/first 15 years and most people have done only 0-1-2 papers. There could be genuine trends in the middle without changing those numbers. And since readers of this blog are generally academically or at least research scientist focused I imagine they’re mostly in the middle. I don’t think the paper says much about the publication record of the typical reader who I would imagine typically fall in the 10-30 papers in first 15 years range.
    b) Authors ARE being asked to produce more papers, just not more first author papers. And remember that figure 1b that looks like a straight line trend is actually on a log scale so it is exponential growth in papers per person. I don’t think effort on a paper goes down as 1/# authors so that equates to growth in effort even if # of paper trends are driven by # coauthors. And they do have less time (one of my universities used to have one secretary per four professors and now they can barely keep one per department and the bureaucratic demands on time have definitely increased). Time pressure on publishing is almost certainly genuinely going up.

    • Oh and for those of us not lucky enough to live in Canada, pressure for grants and time spent hunting declining grants is a huge increase in effort that I suspect we may subconsciously attribute to research=writing papers.

    • Broader question Brian: I get the sense that you find the data in this paper to be not very informative (or even misleading?). That in your view, the feeling that “people are publishing more these days” is correct in some larger sense even if it’s false in some narrower sense. For instance because people are being asked to publish at the same rate while having less time.

      Which is in contrast to your view on the 80 hours/week myth, which you’ve been so instrumental in debunking. Everybody feels like they’re working 80 hours/week these days–and you called bullshit, based in part on hard data from people who track their time. Even though one could surely come up with reasons why the 80 hours/week myth is correct in some larger sense. After all, lots of people do *feel* really busy, and there’s got to be *some* reason for that!

      I’m not trying to catch you in a contradiction, I’m just genuinely curious. Do you see the two cases as analogous, and if so, why do you take different attitudes to the two cases? Or do you not see the two cases as analogous?

      • Fair questions. I guess two main thoughts.

        1) I think the data in the paper do show that people are publishing more. Just not more first-author papers. Not sure how to contextualize that in terms of work or importance. It gets to your question of whether the way we do science has changed. I know my own experience is involvement in 20 author papers (synthesis working groups and also large infrastructure development – e.g. databases) an those sure aren’t 1/20th the work of a regular paper. They’re often more work than a regular paper. And I suspect that for the main readers of this blog (positions at R1, SLAC, and probably even PUI) the number of papers/year may have gone up a bit more than the amount explained just by increased number of co-authors. But that is total speculation.

        2) And I think increased time pressure is very real. Namely declining support from universities (secretarial/administrative and lab tech support have plummeted in the US and from what I can tell most of the world). And in most of the world the amount of time spent chasing grants has gone up drastically for nearly everybody. So this leads to less time for research/publishing. But I’m not sure this is different from other jobs. Increased productivity (output/time) is a very generic phenomenon in economies. But it may be why people feel less relaxed and more insane.

        So I think it is much more accurate that people feel more pressured now than they did in the past than it is to say they are working more than in the past.

        And to me it is still an open question whether readers of this blog are publishing more papers/year compared to their peers 30 years ago beyond that explained by increased co-authors. Certainly the data in the paper has shifted my prior to make me think the truth of that statement is less probable than before I read it. But it hasn’t shrunk to something approaching certainty around no change.

    • Productivity may be log-normally distributed, but with a sigma small enough so that the mean still says something about typical publishing patterns. I would guess in ecology for ~90% of active researchers post-phd, the long-term publishing average for first-authored papers falls between 0.5 and 3 papers/yr. I’ve seen a couple people that do publish >> 3 first-authored papers per year, but seems so rare that it wouldn’t completely dominate the mean trends.

  4. Random aside for anybody silly enough to be skeptical of a paper just because it’s been critiqued on PubPeer. I just checked out discussion of this paper over there, and it’s largely bad so far. There’s somebody saying that the study is flawed because it didn’t count co-authored papers, which is laughably false.

    One shouldn’t infer from this that all PubPeer comments are this bad, of course. But increasingly often these days, I see people on social uncritically pointing to the fact that there are PubPeer comments on a paper as being a reliable indicator that the paper is flawed. No. PubPeer commenters are FAR from infallible.

  5. I do think there is pressure to produce more of any sort and that some people use dubious co-authorships to relieve some of that pressure (‘look I am publishing, see my long list’). I’ve turned down a few co-authorships and people have been surprised. I’ve also had to push a little to get a co-author on something I contributed a fair bit to.

    Where you are in the hierarchy/power structures certainly seems to make a difference to your net publication record in my experience.

    The first author thing is interesting but I wouldn’t extrapolate that to eg no extra pressure, no dubious publication maneuvering (which you may or may not be doing, I’m just emphasising the point). In fact I’d probably argue it confirms that people are no more productive these days despite trying to *appear* that they are.

  6. Wouldn’t combining first and last authorship for people with a 15-plus year publishing career make more sense? I think part of the pressure these days for early career scientists is to get their groups up and running so that grad students and postdocs can be generating papers. Not counting this as legitimate publish or perish pressure seems off (AFAIK some departments are now using number of last-authorship papers as specific metrics in tenure decisions).

  7. I’m curious whether we need to make a distinction here between ‘publishing’ and ‘trying to publish’: it may be that the average researcher is first author on more manuscripts, but this increase in productivity is masked by the decreased chances of getting each paper published. This would square with the observation that most of the complaints about the volume of papers ‘these days’ come from people involved in pre-publication assessment (i.e. editors and reviewers), rather than the readers of the published literature.

    • Increased rates of rejection (maybe particularly of desk rejection) could certainly help create an increasing sense of pressure to publish even though people aren’t publishing more on average. I recall that we talked about this a while back, and the data were mixed. For instance, I seem to recall that rejection rates at Ecology haven’t shown any long-term trend over the last decade or more, even if you include desk rejections. And I don’t think they’ve changed much at Oikos either. But the story might be different at other journals.

  8. The more I think about this, I think we’re over thinking it.

    The paper pretty clearly shows that:
    a) Total papers per author is going up at a nontrivial rate
    b) First author papers per author is constant
    a+b imply number of coauthored papers per author is going up.

    I think people tend to dismiss (a) by just saying well everybody hands out coauthorship like potato chips now and coauthorship is effortless so only (b) matters. I think this is the real myth that needs to be debunked. Once we acknowledge that most (lets say the average) coauthorship paper is real work. Then the conclusion is I am spending the same amount of time writing first author papers I did 100 years ago, but I’m now spending markedly more time on coauthored papers. Ergo, I am spending more time writing papers than I used to, just like everybody thought.(And per comments above, I have less of my day to spend on papers than I did 30 years ago).

    • And with a quick glance at the paper Figure 1B – it varies by field but total number of papers per author is up by 25-50% in 20 years. That’s not trivial.

      And to me the real message beyond # papers/author IS going up is that the social fabric of how we do science has changed really markedly.

    • Yes, I think that’s what it comes down to. It’s not clear to me how much of this is changing authorship practices vs. changes in how we do science. Probably some of both. You could be right that giving each co-author only 1/N worth of credit for co-authored papers is probably too little on average in this context. But giving each co-author full credit for co-authored papers would be way too much.

      I guess one thing to do would be to see what the data look like with different partitionings of co-authorship credit. 1/N, 2/N…N/N, say.

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 )

Twitter picture

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

Facebook photo

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

Google+ photo

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

Connecting to %s