Introduction
Begin with the divergence that organises what follows. Sound has never been better instrumented, and the evidence that chronic exposure to it damages health has never been better established. Yet instrument and evidence are both pulling away from the governance meant to sit between them. The sources of noise are changing. The energy transition rearranges the soundscape at the doorstep, and the digital transition concentrates a new low-frequency hum in infrastructure. The instruments are changing too — citizen sensing, smartphone dosimetry, AI-based soundscape analysis. And the regulatory architecture, built decades ago around a single weighted metric, is changing slowest of all. What follows is an attempt to hold those three rates of change in one frame.
From nuisance to risk
The older category for noise was nuisance: a matter of amenity and annoyance, of one person’s sound intruding on another’s peace, adjudicated where it could be and tolerated where it could not. The reclassification that motivates this work is the passage of noise out of that category and into another — risk factor, in the epidemiological sense, standing beside the familiar cardiovascular hazards rather than trailing them as a quality-of-life footnote (Münzel et al., 2024). The warrant for the move is both mechanistic and graded. Chronic exposure acts on the autonomic system, below the level of deliberation, so its damage does not require the sound to be loud enough to annoy. The dose–response it produces is smooth and without an evident floor: for road-traffic noise and ischaemic heart disease, high-quality pooled evidence puts the relative risk at about 1.08 for each additional ten decibels of \(L_\text{den}\), and finds no safe threshold within the range the studies observed (Kempen et al., 2018). Nor is the harm confined to the heart, since the same switched-on stress response reaches metabolic outcomes, type-2 diabetes among them (Münzel et al., 2025). A single-organ nuisance behaves, on this evidence, like a systemic risk factor.
Scaled across a continent, a per-person increment becomes a public-health total. On the European Environment Agency’s health-impact assessment, transport noise is associated with some 66,000 premature deaths, 50,000 new cardiovascular cases and 23,000 cases of type-2 diabetes a year. That places environmental noise among the largest environmental health burdens in Europe, second only to air pollution (European Environment Agency, 2025). The figures are estimates, and are treated here as such rather than as a register of named victims. The exercise that produces them is a modelled one, and its uncertainties are kept visible throughout. The direction of travel, though, is not in doubt, and stark enough to open on.
The spine, stated once
Three chapters carry the argument distinctive to this work, and giving it plainly here spares the chapters the need to stop and explain themselves. The A-weighted decibel underwriting European noise policy is an epistemic decision, not a neutral fact: a metric that, by construction, discounts the low-frequency and tonal content that most reliably drives complaint (2 Measuring the invisible). The newest and fastest-growing sources, data centres and AI-scale compute among them, produce precisely that content (7 The machine hum). A governance built on that metric therefore fails those sources while appearing, on its own terms, to comply (11 Governing sound): the home is unliveable and the source is lawful, and both hold at once because the instrument defining compliance cannot register what makes the dwelling intolerable. Philosophy of measurement meets agnotology on a material, measurable object — what an instrument renders visible or invisible, and who is served by the invisibility. Beyond those three chapters it is not named again, only relied upon.
How to read what follows
The three parts move from the object to the sources to the politics, from perception, measurement and the body, through a soundscape in rapid change, to distribution, cost, governance and design, ending on the constructive turn from abating sound to composing it. Two commitments hold across all of them. The first is evidential. No figure appears without a cited source and an access date, no causal claim is pressed past what the evidence will bear, and the aim throughout is to diagnose rather than to prescribe — to draw the components of a current philosophy of noise without issuing regulatory verdicts as certainties. The second is that the interactive layer is part of the argument, not an ornament upon it. Each simulator computes on the parameters the reader enters, so that moving a control becomes a way of seeing where the evidence holds and where, at the low end or the margins, it runs out. Granada and Andalucía carry their own numbers wherever the published data allow, beside the European and global aggregates.