Generational Turnover in Science

Speculative If a field genuinely opens to outside ideas when its dominant figure dies — and in the one place this has been measured cleanly it does — then the prize is a lever on the rate of discovery that costs nothing to pull, because the deaths happen anyway and the only decision is whether the institutions around a field make the vacated space easy or hard for outsiders to enter. The condition comes first because the effect is measured in exactly one corner of science, and section 7 says what would show whether it generalises.

Established The magnitude, in the units the anchor study reports, is an 8.6% rise in the flow of articles from scientists who were previously outsiders to a subfield, appearing after the premature death of the star who dominated it, and those outside papers are disproportionately highly cited. Five years out, that inflow is large enough to fully offset the decline in output from the star's own collaborators — a decline the earlier study in this line put at 5–10% of quality-adjusted output and found permanent. The net effect on the field is not a hole where the star was; it is a partial turnover of who works there and which questions get asked.

Frontier The gain accrues to the field and to the outsiders who enter it, not to the incumbents, and that is the whole political difficulty of acting on it. The people who could lower the barriers — senior editors, grant panellists, department chairs — are the incumbents, and the beneficiaries are people not yet in the room. A field that made entry cheaper without waiting for a funeral would capture the same inflow a death produces, but nobody has shown a policy that does this, which is why the honest version of the upside is a measurement waiting to be taken rather than a programme ready to fund.

Handwave The condition that binds is that the effect has been established only in the life sciences, on PubMed data, and the strong reading of Planck's principle — that old science must die before new science can win — is not what the data show even there. What the data show is entry into a vacated space, not the conversion or defeat of the old guard; section 15 separates the measured claim from the slogan, and section 5 names why the physical sciences, where a rising Nobel-discovery age climbed 13.4 years in physics across the last century, remain untested.

Established Planck's principle is the claim that science changes not because its practitioners change their minds but because its practitioners are replaced. Max Planck wrote, in the autobiography published in 1950, that “a new scientific truth does not triumph by convincing its opponents and making them see the light, but rather because its opponents eventually die, and a new generation grows up that is familiar with it.” The pithy form everyone quotes — “science advances one funeral at a time” — is not Planck's; the economist Paul Samuelson coined “funeral by funeral” in a 1975 magazine column and it stuck to Planck afterwards. The distinction matters, because the slogan asserts more than Planck did and far more than anyone has measured.

Frontier This brief owns a single, unusually clean quasi-experimental question: when the dominant figure in a scientific subfield dies, does the subfield open to new people and new ideas, and by how much? That question is separable from the ones the neighbouring briefs already hold. Scientific Revolutions owns whether change is Kuhnian and abrupt or gradual and Lakatosian; Science of Science owns the discovery production function; Scientific Institutions Through History owns the referee corps and the throughput of the gate. None of them owns generational turnover as a measurable shock, which is what makes Planck's principle testable at all.

Established The reason the question is answerable is that death is exogenous, and the reason it is worth answering is that almost nothing else in the sociology of science is. Most claims about gatekeeping, cohorts and paradigm loyalty rest on correlations that a good seminar would demolish: senior scientists differ from junior ones in a hundred ways besides age. A premature, unexpected death holds the field fixed and removes one person, which is as close to a randomised deletion of an incumbent as the world offers. The anchor study of section 3 built its whole identification on that fact.

Established A note on sourcing. This brief was commissioned in September 2026. Its central estimates come from two peer-reviewed studies whose working-paper versions were fetched in full; the pre-2019 tests of Planck's principle are cited from the bibliographic record where the journal versions could not be re-fetched, and are flagged as the mixed and mostly negative evidence they are.

Established “Generational turnover” here means the turnover of who works in and leads a field, not the replacement of one cohort's beliefs by another's. The distinction is the load-bearing one for the whole brief: Frontier the measured effect is that new people arrive when an incumbent leaves, which is a claim about entry and mobility, whereas Planck's slogan is a claim about conversion and belief — and the two come apart in the evidence of section 3.

Established The anchor is Azoulay, Fons-Rosen and Graff Zivin, “Does Science Advance One Funeral at a Time?”, American Economic Review 109(8):2889–2920 (2019). The authors identified 452 elite life scientists, drawn from a population of 12,935, who died prematurely between 1975 and 2003, reconstructed each star's subfields from PubMed using the Related Citations Algorithm over Medical Subject Headings, and measured the flow of publications into those subfields before and after the death. The design is a difference-in-differences with the death as the shock and comparable still-living stars' fields as controls.

Established The headline result is that the flow of articles by non-collaborators into an affected subfield increases by 8.6% on average after the star dies, and those additional contributions are disproportionately likely to be highly cited. The increase is not a reshuffle of the existing hierarchy: the entrants are drawn from outside the subfield, are more likely to be scientists who were never active in the star's area, and are themselves disproportionately likely to become future stars. In the authors' own closing sentence, “once in control of the commanding heights of their fields, star scientists tend to hold on to their exalted position a bit too long.”

Established The other half of the effect is the collaborators, and it points the opposite way. The same line of work established first, in Azoulay, Graff Zivin and Wang's “Superstar Extinction” (Quarterly Journal of Economics 125(2):549–589, 2010), that the star's own coauthors suffer a 5–10% permanent decline in quality-adjusted output after the death — about 7.4% on a journal-impact-weighted count across 161 stars who died 1982–2003, with sudden deaths (roughly 10.5%) hitting harder than anticipated ones (roughly 4.7%). Established That decline is best explained by the loss of an irreplaceable source of ideas rather than by lost coordination, because it is not the most recent collaborations that drive it. So a death does two things at once: it silences an “invisible college” of insiders and it opens the door to outsiders, and the 2019 finding is that within five years the outsider inflow is large enough to fully offset the insider decline.

Established The identification rests on the deaths being unforeseeable, and the authors test exactly that. Of the 452, they separate the 185 who died suddenly and unexpectedly from the 229 who died after a protracted illness, and the outsider-entry effect appears in both groups, which is the check that matters: a field cannot have adjusted in advance to a sudden death. Established The controls are the subfields of comparable stars who did not die, and a difference-in-differences lives or dies on parallel pre-death trends between treated and control fields, which the authors show hold.

Established The collaborator decline is consequential precisely because science has become a team enterprise. Wuchty, Jones and Uzzi documented that teams displaced solo authors across the sciences over five decades and that team-authored work is more highly cited, so a star anchors an invisible college whose dissolution removes more than one person's hands. Frontier That is why the 2010 study reads the permanent decline as lost access to an irreplaceable source of ideas rather than as lost coordination — the loss is of specific knowledge that lived in the collaboration, which is why recent collaborations, where coordination frictions would bite hardest, are not what drive the effect.

Frontier The turnover question draws its stakes from a contested claim that science is consolidating. Park, Leahey and Funk reported in 2023 that papers and patents have grown steadily less disruptive since 1945 on a citation-based index; Speculative if that decline is real rather than an artefact of lengthening reference lists — a dispute Science of Science carries in full — then a mechanism that injects outsider ideas into settled fields matters more, not less. Established This brief takes no side on the disruption index; it observes only that the entry effect is one of the few measured antidotes to consolidation, whatever consolidation's true size.

Frontier The mechanism the authors document is a deterrent, not a conspiracy. They report that few of the deceased stars had served as journal editors or on the committees that award grants, which rules out direct administrative gatekeeping as the main channel in the life sciences. What remains is a mix of intellectual, social and resource barriers — the “Goliath's shadow” in which the prospect of challenging a living luminary deters outsiders from entering at all — together with the intellectual coherence of the subfield: outsiders enter only those subfields whose topology offers a less hostile landscape for the acceptance of foreign ideas.

Frontier Set against that clean result, the older literature testing Planck's principle directly is mixed and mostly negative on the strong version. Hull, Tessner and Diamond's 1978 study in Science, which examined the acceptance of Darwinism, found age to be at most a minor factor; Diamond's 1980 study of the acceptance of cliometrics found the age effect weak; and Messeri's 1988 study of geologists adopting plate tectonics found that older scientists adopted the new theory sooner than younger ones, the reverse of what Planck's principle predicts. Speculative The reconciliation this brief offers is that these studies tested conversion — did the old change their minds — while the anchor study tested entry — did new people arrive — and only the latter shows a strong, well-identified effect.

Established Around the anchor sits a robust descriptive record that science has aged. Jones and Weinberg, “Age dynamics in scientific creativity” (PNAS 108(47):18910, 2011), found across 525 Nobel laureates that the mean age of prize-winning work rose 7.4 years in medicine, 10.2 years in chemistry and 13.4 years in physics between the pre-1905 and post-1985 cohorts, with theorists making their great contributions about 4.4 years earlier than empiricists. Established By the 21st century, only about 19% of prize-winning physics discoveries were made by scientists under 40, against roughly 60% in the early period, and no chemistry laureate under 40 has been recognised for 21st-century work. The population of working scientists has aged in step: the median age at first NIH R01-equivalent award sits near 42 as of FY2021–2025, having risen through the 1990s and 2000s before plateauing.

Frontier The one direct extension into technology reverses the sign, and that is a finding, not a footnote. Esposito's “Age, Death, and the Nucleus of Invention” (2023 working paper) linked 54,673 surviving USPTO inventors to 13,112 deceased collaborators over 1976–2018 and found that losing an early-career collaborator, not a senior one, is what hurts: inventors who lose a young partner produce about 14% fewer patents relative to the median and combine ideas about 2.5 years older afterwards. Speculative In invention, the young are the source of fresh human capital that flows upward, so removing an incumbent is not obviously generative — which is the first hard evidence that the funeral effect is domain-specific rather than a law of science.

Frontier The first open question is whether the 8.6% effect survives outside the life sciences at all. The anchor is built entirely on PubMed, whose field structure — many small subfields organised around techniques, organisms and diseases — is unusually hospitable to the design. Whether a field with a single dominant theoretical core, as Scientific Revolutions describes for much of physics, would show the same entry response, a larger one, or none, is unknown, and section 7 makes it the decisive test.

Frontier The second is which field topologies open and which stay closed. The anchor found that outsiders enter only subfields whose intellectual landscape is less hostile to foreign ideas, but “less hostile” is measured after the fact from who entered. A prospective metric — predicting from a subfield's citation and concept structure how open it will be to a post-death influx — does not exist, and building one is the difference between a retrospective story and a usable instrument.

Speculative The third is whether conversion ever matters, or only entry. Planck's own claim was about opponents changing (or failing to change) their minds; the modern evidence measures new people arriving. Messeri's plate-tectonics reversal suggests conversion can run the opposite way to the slogan. Whether there exists any real episode in which a specific truth waited on the literal deaths of its named opponents, rather than on the field filling with newcomers indifferent to the old fight, is genuinely unsettled and may be untestable at the level of individual beliefs.

Frontier The fourth is the optimal tenure of a paradigm. The anchor concedes that gatekeeping can help a young field cohere through shared assumptions before it helps to have those assumptions challenged, so the welfare-relevant quantity is not whether incumbency slows entry but when the cost of coherence starts to exceed its benefit. Nobody has estimated that crossover point for any field, and it is the number a funder would actually want.

Frontier The fifth open question is whether the shadow outlives the star. The deterrent is partly a living presence, yet the surviving collaborators keep their grants, editorships and referee slots; Speculative so the barrier may persist after death through the surviving invisible college, and the roughly five-year lag before outsiders fully offset the decline is at least consistent with a shadow that fades slowly rather than lifting at the funeral. No study yet separates the living deterrent from the inherited one.

Frontier The sixth is the functional form: how the effect scales with the star's dominance. The anchor reports that especially accomplished stars are a stronger deterrent to entry — the Goliath's-shadow finding — so the effect is not flat across stars. Speculative But whether entry responds linearly to a star's citations, grant share or prizes, or only past some threshold of dominance, is unmeasured, and it is the shape a policy aimed at concentration would need.

Established The first bottleneck is that the clean design exists in one dataset and one discipline. PubMed's coverage, its algorithmic subfield delineation and the density of biomedical stars are what make the 452-death sample large enough to identify an 8.6% effect. The physical sciences have no equally clean, equally complete, algorithmically partitioned corpus with a comparable population of premature elite deaths, so the most important replication is blocked less by method than by the absence of an assembled dataset.

Frontier The second is that the outcome measures are all publication counts. Articles and citations are what PubMed exposes, so the anchor could rule out administrative gatekeeping only indirectly, by noting that few stars held editorships or grant-panel seats. Testing whether resource control is the channel elsewhere requires linking a death to grant-dollar share and journal-page share in the affected field — data held by funders and publishers, not by any bibliographic index, and released by neither.

Established The third is that no prospective registry turns anecdotes into base rates. As Anomaly Resolution in Physics argues for discrepancies, the difference between a pile of vivid stories about fields reborn after a death and a usable science is a registry that logs the field, the star, the topology and the outcome prospectively. The star deaths keep happening; nobody is recording them as they occur.

Frontier The fourth is causal identification once you leave the death design. Retirement, moves and demotions are all endogenous — a scientist retires partly because the field is changing — so the exogeneity that makes death usable is exactly what any policy-relevant version of the question (mandatory retirement, term limits on grants) lacks. The clean shock does not correspond to any lever an institution can pull.

Established A fifth bottleneck is that “star” is a threshold the analyst chooses. The anchor identified elite scientists using multiple markers — funding, citations, honours, patents — and the estimated effect depends on where that line is drawn. Frontier A physical-sciences replication cannot inherit the biomedical threshold; it must re-derive who counts as dominant in a field organised around facilities and collaborations rather than laboratories, and the answer is not obvious.

Established This brief consumes results from its siblings rather than re-deriving them. It takes the gradualist picture of conceptual change from Scientific Revolutions, whose 38-network measurement concluded that shifts occur as gradual Lakatosian modifications rather than abrupt Kuhnian ruptures; it takes the throughput-constant finding from Scientific Institutions Through History, where the Royal Society rejection rate held near 10–11% for a century while the author population changed completely; and it takes the instrument-validity discipline from Science of Science, that any pattern found in citations is first a pattern in citing.

Frontier Outside the corpus, two dependencies bind hard. The first is data access: the grant-dollar and journal-page-share series that would test the resource channel are held by funders and publishers with no obligation to share, the same politics that Scientific Funding Models shows has kept a decade of funding-lottery outcomes unpublished. Frontier The second is a validated field-topology metric — a way to score a subfield's openness to foreign ideas before a death rather than after — without which every prediction about which fields will open remains a post-hoc rationalisation.

Speculative A quieter dependency is a living cohort for the outsiders to come from. The entry effect presupposes that there are unattached, capable scientists positioned to move into a vacated field; if the early-career pipeline is thin or leaving, the door opens onto an empty corridor. A June 2026 survey of young biomedical scientists reported barely half now likely to remain in academia, which if sustained would attenuate exactly the inflow the anchor measures.

Frontier A third dependency is a partitionable corpus outside biology. PubMed's algorithmic subfield delineation is what made the anchor's field definitions defensible; the physical sciences have large open corpora such as arXiv and the APS journals, but nobody has partitioned them into subfields the way the Related Citations Algorithm partitions PubMed. Speculative Until that partition exists and is validated, a physics replication would be arguing about its field boundaries before it could argue about its effect.

Frontier The decisive test this brief would run is a pre-registered replication of the star-death design in the physical sciences, where the incumbent-theory structure differs from biology's, with grant-dollar share and journal-page share entered as outcomes alongside publication and citation counts. Assemble premature deaths of elite physicists and chemists from an equivalent corpus, delineate subfields, pre-register the entry hypothesis and the predicted direction, and measure the same 8.6%-style inflow — but add the two resource outcomes the anchor could only rule out indirectly. Established Every component has been demonstrated separately; what is missing is the assembled dataset and a funder willing to open the grant and page records.

Frontier The physical sciences are the right place to run it because the theory predicts divergence there. If a field's incumbent is a shared theoretical core rather than a person, a star's death should matter less, and the entry effect should shrink toward zero; if instead concentration of facilities, grants and journal space around a dominant figure is what deters outsiders, the effect should be as large or larger, and the grant-dollar and page-share outcomes should move where they did not in biology. Speculative The two hypotheses make opposite predictions, so a null result would be informative rather than empty — the rare meta-science test where either outcome is a finding.

Frontier A second experiment is the prospective registry. Register elite-scientist deaths as they occur across disciplines, record the field topology at the time of death, and follow entry and citation outcomes forward, converting the retrospective anchor into a base rate that Forecasting Replication could then attach probabilities to. Established It needs no new instrument, only a standing commitment to record what is otherwise studied only in hindsight.

Speculative A third experiment tests the topology hypothesis directly. Build a field-openness metric from citation and concept structure, pre-register a ranking of subfields by predicted openness, and check it against the entry that follows the next cohort of deaths. Frontier This is the step that would turn “outsiders enter less hostile fields” from a description into a prediction, and it is buildable now from public bibliometric data.

Speculative The experiment nobody can ethically run is the policy version. The clean shock is death; the lever an institution controls is retirement or term limits, which are endogenous and whose welfare effects the anchor's own concession about early-field coherence makes ambiguous. Handwave Any claim that mandatory retirement would speed discovery is an extrapolation from a mortality experiment to a personnel policy, and no design short of a randomised term limit — which will not be run — would settle it.

Established The engineering is record linkage, and it is modest by the standards of any physical-science brief. The anchor's machinery is a bibliographic corpus, an algorithm that partitions it into subfields, a list of deaths with dates, and a difference-in-differences estimator. Extending it to physics and chemistry, in the manner Scientific Revolutions and Anomaly Resolution in Physics would care about, means acquiring an equivalent open corpus and a defensible subfield partition, not building an instrument.

Frontier The expensive component is the resource data. Grant-dollar share by field and year, and journal-page allocation by subfield, exist inside funders and publishers as administrative records; federating them for research is a data-governance and privacy task of the kind Scientific Funding Models and Scientific Governance Models describe, not a technical one. Speculative A prospective death registry is the cheapest piece of all — a maintained table — and has never been built because no institution owns the question.

Established Timelines and horizons for this work are set in section 13. Nothing here waits on a new detector, a new material or a new computation; every dependency is a dataset or a decision to release one, which is why the honest engineering statement is that the bottleneck is access and will, not capability.

Established The death-date and career-reconstruction linkage is the one genuinely new piece of plumbing, and it has been demonstrated. Esposito's invention study built exactly this — linking inventors to birth and death records scraped from obituaries and public databases across 1976–2018 — so assembling a comparable death list for physicists and chemists is a data-collection task with a worked precedent. Frontier It is tedious rather than hard, which is why the absence of the dataset is a matter of nobody having been paid to build it.

Established The nearest neighbours share a subject and divide it cleanly. Scientific Revolutions holds the tempo-of-change question and the Kuhn-versus-Lakatos measurement; Scientific Institutions Through History holds the gate and its referees; Science of Science holds the production function and the instrument-validity traps; Scientific Funding Models holds the funder; and Scientific Governance Models holds the peer-review and integrity machinery a resource-channel test would plug into.

Frontier Two more sit one step out. Forecasting Replication is where the prospective registry's outputs would be scored as probabilities, and Innovation History holds the long-run patterns and the warning — inherited here — that a measured slowdown or speedup can be an artefact of the counting rather than a fact about the world. Established This brief measures turnover; it adjudicates no particular scientific dispute, which keeps its quarrels methodological.

Established The institutions that could act on the finding are the ones staffed by its subjects. The median age at first NIH R01-equivalent award near 42, up from lower values before the 1990s, is the demographic backdrop: the people who set budgets, edit journals and chair departments are, on average, later in their careers than their predecessors were, and the aging is measured, not asserted. A policy that lowered barriers to entry would be written by incumbents against their own field position.

Frontier The reflexive answer — make scientists younger by fiat — is not what the evidence supports. NIH's Early-Stage Investigator policies were designed to counter the aging of the workforce, and the median first-award age has plateaued rather than fallen; the anchor's own result says the generative event is the arrival of outsiders, not the removal of the old, and Scientific Funding Models records that funders run natural experiments on allocation and decline to publish the outcomes. Speculative The institutional lever that matches the finding is not retirement but entry: standing positions, small fast grants and journal space reserved for scientists new to a field.

Frontier The pipeline is a precondition, and it is under strain. The entry effect requires a cohort of capable outsiders able to move; a June 2026 survey of young biomedical scientists reported the share likely to remain in academia falling by more than 20 percentage points over six months after a funding-policy shock, and the share likely to remain in the United States falling comparably. Established Turnover cannot rejuvenate a field if the next generation has left, which ties this brief's mechanism to the workforce questions Science of Science tracks.

Speculative The policy that matches the mechanism is subsidised entry, not forced exit. Small, fast, low-barrier grants and journal space reserved for scientists moving into a field new to them would, in principle, capture the same inflow a death produces without the permanent knowledge loss. Frontier None has been evaluated against the 8.6% benchmark, so the claim that entry can be manufactured deliberately remains a hypothesis with a specific number to beat.

Established The deeper institutional problem is that the clean shock and the available levers do not correspond. Death is exogenous; retirement, term limits and hiring lines are endogenous and blunt, and the plateau in first-award age shows how slowly the levers move even when a funder pushes. Frontier An institution cannot ethically or practically reproduce the experiment that produced the finding, which is the gap between knowing the effect and being able to use it.

Established The finding is one sentence away from a weapon, and the sentence is a mistranslation. “Science advances one funeral at a time” reads as licence for ageism — push the old out and discovery accelerates — but the measured effect is entry by outsiders into a vacated field, and the same research line shows that losing a scientist permanently destroys a specific body of knowledge their collaborators cannot replace. The honest reading indicts incumbency structures, not old people.

Frontier Using the result to justify mandatory retirement inverts its own evidence. Esposito's invention data show that losing the young harms the senior, that human capital flows upward as well as down, and the anchor concedes that gatekeeping helps a young field cohere before it hurts. Speculative A policy that treated every senior scientist as a barrier would discard the specific-human-capital gains the 2010 study measured while betting on an entry effect established only in one discipline.

Frontier The distributional point is that the gains and the costs fall on different people. The outsiders who enter a vacated field capture the citations and, the anchor finds, disproportionately become future stars; the collaborators who lose a mentor bear the permanent output decline; and the incumbent whose death produced the opening is, obviously, past caring. Handwave A published claim that a particular group's incumbency is holding a field back will be read as an allocation instruction long before its confounds are audited, and the field's own history of turning descriptive metrics into evaluation weapons is the precedent to design against.

Established Consent and attribution bind the extension more than the anchor. The anchor works from public publication records, but the resource-channel test of section 7 would touch grant files that name living and dead individuals, and the governance for using identifiable career and funding data at that scale does not yet exist. Frontier A test designed to expose incumbency must not itself become a dossier on named scientists.

Frontier The stakes are set by the aging of discovery itself. If the mean age of great achievement has risen more than a decade in physics over a century, and if the burden of knowledge means each generation reaches the frontier later, then the rate at which fields turn over their leadership becomes a live parameter in the long-run growth of knowledge rather than a curiosity. A civilisation that wanted to hold its discovery rate constant against a rising training burden would need either more scientists or faster turnover of the questions being asked.

Frontier The anchor makes turnover a candidate free variable, but only a candidate. An 8.6% entry effect, fully offsetting the collaborator decline within five years and producing disproportionately cited work, is a real number attached to a real mechanism — and it is the only such number, measured in one discipline, on one dataset. Speculative Whether lowering entry barriers deliberately could capture the same inflow without waiting for a death, and whether the effect exists at all where a unified theory rather than a person holds the commanding heights, are the two unknowns on which any civilizational claim rests.

Handwave The strongest claim in circulation — that clearing out senior scientists would rejuvenate science across the board — is not supported by anything on this page. It generalises a life-sciences entry effect into a personnel policy, ignores the permanent knowledge loss the same authors measured, and runs opposite to the invention evidence; it works by assertion from a slogan Planck did not write to a conclusion the data do not reach.

Established The concrete version of the stakes is arithmetic on the aging numbers. If the mean age of prize-winning physics work rose 13.4 years across a century while the working population aged in step, then holding a field's discovery rate constant requires either more scientists reaching the frontier or faster turnover of the questions the field inherits. Frontier Generational turnover is the second of those two dials, and this brief's whole contribution is to have put one measured number — 8.6%, in one discipline — on how much that dial can move.

These horizons track when the turnover question acquires instruments and answers, not when science itself speeds up.

• 10 yr: Frontier The star-death design is replicated in at least one physical science with a pre-registered protocol; a prospective death registry exists for at least one discipline; a first field-openness metric is published and scored against subsequent entry.
• 25 yr: Speculative Grant-dollar and journal-page-share outcomes have been linked to elite deaths in at least one field, settling whether the resource channel binds outside biology; the crossover point at which a paradigm's coherence stops paying is estimated for a handful of fields.
• 50 yr: Speculative Whether generational turnover is a usable lever on discovery rate — or an artefact of one discipline's data — is an answered historical question; entry-promoting institutions, if they work, are routine rather than experimental.
• 100 / 250+ yr: Handwave Deliberate civilizational management of field turnover against measured openness curves; nothing on this page grounds a forecast at this range.

• Depends on the gradualist picture of conceptual change from Scientific Revolutions, the gate-throughput record from Scientific Institutions Through History, and the instrument-validity discipline of Science of Science — all existing briefs this one builds on rather than re-derives.
• Requires (not on this map) five constraints. A pre-registered replication of the star-death design in the physical sciences, this brief's decisive test; the linked grant-dollar and journal-page-share data that would test the resource channel the anchor could only rule out indirectly; a decision by funders and publishers to release the allocation and page records that carry it; a validated field-openness metric that predicts entry before a death rather than describing it after; and a cohort of early-career positions large enough that outsiders exist to fill the fields a death vacates.
• Enables a base rate for whether generational turnover accelerates discovery, which would let a funder choose entry-promoting policies on evidence rather than on the slogan; and a prospective registry whose probabilities Forecasting Replication could score.
• Adjacent Scientific Funding Models for the allocation records, Scientific Governance Models for the review machinery, and Anomaly Resolution in Physics for the registry-versus-anecdote argument.

Frontier “Science advances one funeral at a time — old scientists have to die before new truths can win.” Planck wrote something weaker and Samuelson made it pithy; the strong version says opponents must literally die, and the evidence for that is thin. The pre-2019 tests found age a minor factor in the acceptance of Darwinism, weak for cliometrics, and reversed for plate tectonics, where older geologists adopted the new theory sooner. What the clean modern study measures is not conversion but entry — new people arriving in a vacated field — which is a real 8.6% effect and a different claim.

Speculative “The funeral effect is a general law of science.” It is established in the life sciences, on PubMed, for 452 premature deaths. The one direct extension into invention found the opposite sign — losing young collaborators, not senior ones, is what depresses output — and the physical sciences are untested. A single well-identified result in one discipline is a strong finding and a weak law.

Established “It works because senior stars sit on the journals and grant panels and block outsiders.” The anchor's authors checked and report that few of the deceased stars held editorships or grant-committee seats, which is why they attribute the effect to a deterrent — the prospect of challenging a living luminary — and to the intellectual coherence of the subfield, rather than to administrative gatekeeping. Whether the resource channel binds in fields where facilities and grants are more concentrated is exactly what section 7 would test.

Handwave “So mandatory retirement would speed up science.” This extrapolates from a mortality experiment to a personnel policy the mortality experiment cannot license. The same research line shows the death permanently destroys specific human capital the collaborators cannot replace; the invention evidence runs the other way; and the anchor concedes gatekeeping can help a young field cohere. The measured lever is entry, not exit, and no design short of a randomised term limit would settle the policy claim.

Frontier “Older scientists resist new ideas because they can no longer grasp them.” The age-of-achievement literature is about lengthening training and the burden of knowledge — the mean age of prize-winning physics work rose 13.4 years over a century — not about cognitive rigidity, and the plate-tectonics case has the old adopting sooner. The barrier the data support is structural incumbency and its shadow, not senescence.

Frontier “So the death of a star is good for science.” The net effect on a field is a partial turnover, not a windfall: the same death that lets outsiders in also silences the star's invisible college permanently, a 5–10% output loss the collaborators never recover. Speculative The measured claim is that fields would benefit from lower entry barriers while their stars are alive, not that anyone is better off dead — and the humane version of the policy is the one section 10 names, subsidised entry rather than attrition.