The Break That Lasts

Yesterday, Nature reported on a study that complicates a warning from a few months back.¹ In May, a paper in Science² found that as they advance through their careers, scientists become steadily less likely to produce »disruptive« work, the kind that overturns rather than builds on what came before. Covering the study back then, The Economist analysed its sharper implication – countries with younger scientific workforces produce a higher share of disruptive research than countries with older ones – and read that as a warning that ageing countries risk being out-innovated by younger ones.³ The study published yesterday in the Proceedings of the National Academy of Sciences doesn’t dispute that finding. It asks a different question of the same data: of the disruptions that do happen, which ones last?⁴

»An idea’s survival is not decided purely by the idea«

The answer runs the opposite way from the previous conclusion. Through a citation-counting exercise at a scale no human team could do by hand, the new study draws on more than 11.2 million papers by 206,130 scientists with careers spanning at least forty years. Using SciSciNet, a large-scale open bibliographic data lake for science-of-science research, the researchers could distinguish disruption from persistent disruption.

The difference is roughly this: if an idea disrupts a field, people stop citing what came before and instead cite the new finding. But then that new finding might itself become old news and get swept aside the same way by some other idea a few years later. Persistent disruption is when the new idea disrupts the field and then stays the reference point: later work keeps citing it rather than routing around it. A persistent disruption breaks with the past and survives, rather than being quickly superseded itself. The probability of the first kind falls with experience, exactly as the May paper found. But the probability of the second rises, from around one per cent early in a career to roughly 2.2 per cent by the fortieth year.

The two findings aren’t really in tension. Fewer disruptions occur as a scientific career lengthens – the earlier paper already got that right, and this one cites it directly rather than quietly disagreeing. What it adds is that the disruptions that are still being produced are increasingly built to last. But it’s worth noticing how differently the same underlying metric sounds after five months. In May, this was a warning about national competitiveness, young countries edging out old ones in a race for scientific advantage. In October, it has become a quiet reassurance that a career’s best work might still lie ahead of its scientist.

The sociologist of science Theodore Porter spent a career studying exactly this mechanism – how a metric like »the probability of persistent disruption rises to roughly 2.2 per cent by a career’s fortieth year«, once it exists, becomes a settled fact that can be carried into almost any argument, detached from the reasoning that produced it.⁵ The citation-counting index behind both papers hasn’t changed between May and October. What changed is which story it was asked to carry.

Does this mean that wisdom accrues with age? That would be a comforting read, but the study gives reason to hesitate. Of the papers judged equally disruptive, those written by more experienced researchers drew more citations, more attention on social media and in the news, and more research funding per paper. An idea’s survival is not decided purely by the idea. Management scholar Phil Rosenzweig⁶ found the same phenomenon in business writing and named it »the halo effect« – once a company succeeds, what it did in the past gets read as having caused the success even in cases when it clearly didn’t. The halo of the scientist behind a paper might have an impact on how much it gets cited.

The paper was reviewed for publication by Mark Granovetter⁷, the sociologist best known for showing that a person’s opportunities often depend less on their own merit than on their position within a network of connections. It seems fitting that a paper making its way past him argues something similar about ideas – that the fate of an idea, like that of a person, is never decided by the idea alone.

References