The 100 most cited papers of all time show how science really gets done

Far from being a list of the most brilliant research, the list reveals a more conflicted structure than many scientists would like to admit.

I really hate the “…and you won’t believe who’s on it!” headline style, but in this case it was almost warranted. Nature Magazine, one of the few surviving great-grandaddies of science publishing, has compiled a list of the 100 most cited scientific articles of all time — and you won’t believe who’s on it.

Far from being a list of the most impactful, famous, or even brilliant research, the kings of the social meta-game that is science these days reveal a more complex and conflicted structure than many scientists would like to admit. Does science proceed according to pure, high-minded adherence to Bertrand Russell’s idea of the scientific ideal? Honestly, it’s a bit hard to tell.

First off, as Nature itself notes in its own breakdown, you won’t find Watson and Crick’s initial paper on the double helical structure of DNA, nor the paper laying out high-temperature super-conductivity. In reality — and this makes sense when you think about it — many of the most cited papers describe practical, physical lab techniques. Probably the most famous paper with a top spot, with more than 65,000 citations, is Nature‘s number four ranking, one of the original DNA sequencing papers that still underlies tons of work today. And, added together, all the top-100 papers about genome-searching algorithms and X-ray crystallography can be grouped to show a truly astonishing level of referencing going on.

The all-out number one most cited one paper of all time, according to Nature? An unassuming study about protein concentration measurement, which has over 300,000 citations so far.

Nature takes a certain view of these results, which is that they show how “researchers rely on relatively unsung papers to describe experimental methods, databases and software.” This is directly at odds with the magazine’s own analysis of the overall dataset, which is described as a “paper mountain.” If all the papers assessed (about 58 million in total) were stacked vertically so the stack reached to the top of Mount Kilimanjaro, the top 100 papers would be just the top centimeter of space. More to the point, those papers that have just one citation, or even no citations at all, make up roughly half the climb overall — the magazine refers to theses as the “foothills.”

Click for the full image.

As the creator of Nature‘s number one article himself admits, saying, “I really know it’s not a great paper,” it’s not necessarily quality that leads to a higher citation count. Nor is it even really everyday usefulness, as there are certainly many more ways to assay protein content than described in the number-one paper. The top tier of cited research is so disproportionately popular, and the vast fields of un- or under-read material so disproportionately vast, that it’s odd to see this list used as evidence for the principled non-conformism of modern scientists.

Rather, I think this shows that the basic practical and psychological considerations that shape every human endeavor are no less powerful in science, or among scientists. The number-one citation wasn’t buoyed by a high-school wish to cite the coolest paper around, but by the honest but still possibly damaging effects of visibility and ease of access. The near-vertical face of this citation mountain proves that scientists are as prone as anyone else to think that the world they see is the extent of the world that exists — the statistics prove that all too few are able to take the time to read widely in their own fields. I say, “able to” because I don’t want to imply that it would necessarily be possible to keep up with the size and pace of modern research; remember, the paper mountain is more than 50 million articles tall.

I’m making the assumption here that the 50% of all research which gets basically no attention is, to a large extent, more worthwhile than its impact would seem to imply — but that’s a fair assumption, I think. Additionally, this list shows the enormous body of knowledge now accepted literally without question — citations being answers to the implicit question, “How do I know you’re right about that?” When doing a calculation based on the free-fall acceleration at the Earth’s surface, nobody goes back to cite the ancient philosophers who first measured it; when talking about DNA, it’s no longer necessary to justify the assumption that it will be stored in a reliable helical structure in most cases. That’s fine, of course — we want scientists to be able to get things done in the real world after all — but it does seem to imply that citation count tracks the middle of the impact spectrum. Get too popular, and you’re no longer worth citing at all.

Important research, but perhaps too successful to make this list.

In the gap between monocle-popping, Francis Bacon-style adherence to principle and heedless rushing toward progress without caution, we can find a workable midpoint that benefits humanity with both accuracy and efficiency. That’s the idea, at least, but this list shows that perhaps a largely unwatched citation system might not be directing the research world toward this midpoint as effectively as many had assumed. Belief in the inherent truth-finding power of the citation system is reminiscent of the idea of free markets in economics. That is, it has a lot of wisdom at its core, but shows a rather shocking naivety about human nature when applied to the real world with the assumption that the real world won’t muck with theory. We’ve been able to think of scientists as uniquely rational and socially unaffected actors in game-theory thought experiments because we’ve always rightly held scientists in high esteem — but even earned respect doesn’t make someone super-human.

It seems to me that this list belies a fundamental weakness, not with science, but with science publications. If we assume that scientists generally want to read and cite more holistically from all the research available, and that’s currently very difficult, then it’s those who organize and display the research who could most directly implement some solutions.

Regardless of how you look at it, this is a fascinating piece of reporting by Nature. Google has also done a list via the larger and less curated Web of Science, putting together a list that’s similar but includes books and other forms ignored by Nature. They both seem to tell the same basic story, however: while you can’t argue with science’s results over the past several decades, it’s hard not to wonder how much more might have gotten done in a system less tied to the unavoidable realities of human fallibility.