A tissue cluster growing by cell division needs to know when to stop. But it has no ruler, no central counter, no growth hormone receptor with a programmed limit. Instead, cells move. Their motility generates mechanical stress at junctions between them. When a cluster grows large enough, the accumulated stress exceeds junction strength, and the cluster fractures. The size distribution of clusters depends on a single parameter: the ratio of breaking rate to growth rate. No external measurement needed. The breaking itself is the size control.
Scientific retractions are growing exponentially, doubling roughly every five years. At 0.12% of publications, the absolute rate is small. But retractions are to scientific integrity what fractures are to tissue size: a destructive process that, when functioning, serves as the system's primary control mechanism. A retraction removes a bad paper, corrects the record, and signals to the community that a finding was unreliable. The rate ratio that matters is retractions to misconduct — not retractions to publications.
The structural parallel: in both systems, the destructive process is not a failure. It is the regulatory mechanism. When it stops working, the system loses control.
In tissue biology, if fracture rates drop below the threshold needed to balance growth, clusters grow without bound. This is one model of tumor formation: not necessarily accelerated growth, but impaired fragmentation. The cells divide at the normal rate, but the breaking mechanism fails — junctions become too strong, motility decreases, or the mechanical coupling between cells changes. The growth rate hasn't increased. The control rate has decreased. And the system has lost its only size-regulation tool.
In science, if retraction rates drop below the level needed to match misconduct rates, the record accumulates unreliable findings without correction. This doesn't require an increase in fraud — just a decrease in detection. Overburdened journals, understaffed integrity offices, perverse incentives against retraction (reputational cost, legal risk) — all reduce the fracture rate without changing the growth rate. The system's only mechanism for removing bad work is weakened.
The ratio formulation makes this precise. The tissue's size distribution depends on k_break / k_grow. When this ratio is high, clusters stay small — fragments break off before they can grow large. When the ratio drops, the distribution shifts toward larger clusters. The control is continuous, not threshold-based: any reduction in the ratio produces larger average clusters.
For scientific retractions, the analogous ratio would be k_retract / k_fraudulent_publish. We don't know this ratio precisely because the denominator — the rate of fraudulent publication — is not directly observable. But we know the retraction rate (the numerator), and we know it has been growing. The critical question is whether it's growing faster or slower than the denominator. If slower, the system is losing control even though more retractions are happening.
The tissue model suggests a diagnostic: measure the size distribution. If fraudulent papers are persisting longer before retraction — if the “cluster size” of undetected misconduct is growing — then the breaking rate is falling behind. The destruction is necessary. The system cannot self-regulate without it. And the ratio, not the absolute rate of either process, determines whether control is maintained.