Gravitational wave astronomy works by detecting individual events — two black holes merging, two neutron stars colliding. Each event must stand out above the detector's noise floor. The louder the event relative to the noise, the more precisely you can extract its properties: mass, distance, spin. The entire detection pipeline is built around this principle. Signal-to-noise ratio is the currency. Events below threshold are discarded. They are noise, by definition.
But every merger below the detection threshold still produces gravitational waves. Billions of mergers across cosmic history — too faint to individually resolve, too distant to characterize — contribute a collective background hum. This astrophysical gravitational-wave background is the sum of everything the detector can't see. It has always been present in the data. The detection pipeline treats it as contamination — part of the noise floor that individual events must exceed.
Cousins, Holz, and colleagues at Illinois and Chicago recognized that this background carries cosmological information that individual detections do not. At lower values of the Hubble constant, the universe's volume is smaller, so the density of mergers within a given region is higher, and the background hum is louder. The intensity of the aggregate signal — the sum of everything individually undetectable — constrains the expansion rate of the universe. They call this the stochastic siren method, published in Physical Review Letters in March 2026.
The insight is structurally precise. The standard siren method uses individual bright mergers to measure cosmological distances. The stochastic siren method uses the collective behavior of mergers too faint to individually identify. The two methods draw on the same physical population but from opposite ends of the brightness distribution. What one method requires (a detectable event rising above the background), the other method discards (the event's individuality is irrelevant; only the aggregate matters).
The general pattern: the aggregate of what falls below a detection threshold can contain information that individual detections above it do not. This is not a statement about noise being “secretly useful” or about finding signal where none was expected. The background is genuinely signal — it encodes cosmological geometry in its intensity. The detection framework made it noise by defining the threshold. Events below threshold don't stop carrying information. They stop being counted. The framework's cutoff determines what registers, and everything below it pools into a collective that is invisible to the framework's accounting but not to the universe's physics.