The Sound A Building Makes

AFew buildings are designed to be heard, and almost all of them are. A classroom in which a teacher repeats every third sentence, an office in which a neighbour's telephone call cannot be ignored: these are failures of a building, and far harder to put right once the concrete has set. Since the 1990s, when several countries first wrote sound into their building regulations, the field has divided over what a good room actually does. One camp holds that the answer is a property of the room itself, measurable before anyone moves in. A second holds that what decides the matter is not the room but the relation between wanted and unwanted sound reaching a particular pair of ears. A third holds that both measure the wrong thing, since what makes a space bearable is settled by the occupants and their control over it. National standards are written out of this disagreement.

BThe oldest position is the easiest to test, and Ingrid Halvorsen, who runs the room acoustics laboratory at the Lindqvist Institute in Trondheim, has spent a career testing it. Reverberation time — the seconds a sound needs to decay by sixty decibels after its source stops — has measured rooms since Wallace Sabine derived it in 1898, and an empty building yields the figure in an afternoon. In a study published in 2017, Halvorsen's group fitted absorbent ceilings in 128 primary classrooms, cutting mean reverberation time from 1.1 seconds to 0.5. Word-recognition scores among pupils aged seven to nine, tested against a recorded voice at a fixed level, rose from 62 per cent correct to 81. Halvorsen is careful about what the number carries. The children heard recordings rather than teachers, and a word list is not a lesson. What the trial demonstrates, she says, is that the room can be made to stop obstructing speech; whether the lesson then improves is a separate question her data cannot answer.

CTomás Reguera, an environmental psychologist at the Institute of the Built Environment in Valparaíso, arrives elsewhere, and arrives from what occupants say rather than from what rooms do. Between 2014 and 2020 his group gathered 9,400 questionnaires in 63 open-plan offices across eleven countries. Loudness predicted very little: the floors with the highest measured levels were not the floors people objected to. What predicted objection was intelligibility, whether a neighbour's words could be made out. Sixty-eight per cent of those reporting serious disturbance named distinguishable conversation as its source, and the worst-rated offices were often the quiet ones, where a single voice carried unmasked across forty desks. Reguera's remedy is counter-intuitive: add sound. A low, steady hiss tuned to the frequencies of speech lifts the floor of the room and leaves the neighbour unintelligible without making the office loud. Reguera accepts that his case at present rests on annoyance rather than on output. No study, his own included, has yet shown that a masked office produces more or better work.

DWendy Okonjo, of the Marchmont Building Science Unit in Auckland, argues that both camps have mistaken the room for the experience of it. Between 2008 and 2019 she audited 52 offices and schools refurbished specifically to bring them inside the national reverberation limit; the measured quantity improved in every one, and the rate of noise complaints fell in none. The reverse case exists too: she has recorded workplaces two or three times over the limit where complaints are effectively absent, and these are consistently the ones where occupants choose their own desks and may move when the room turns against them. If the number can be mended without the complaints shifting, and the complaints can disappear while the number stays bad, the number is not the thing occupants are responding to. Okonjo's own proposal is deliberately modest: the room fixes the outer limit of what is possible, but inside that limit what counts as bearable is settled by control, expectation and who is making the noise.

EIn 2022 Halvorsen and Okonjo ran a trial together. Seventy-four meeting rooms in four buildings were surveyed; absorbent panels went into half of them, chosen at random, and reverberation time there fell from a mean of 0.9 seconds to 0.45. Occupants sat a standard speech-recognition test, then answered a satisfaction questionnaire six months later. Test scores were 11 percentage points higher where the panels had gone in. Satisfaction ratings were not: treated and untreated rooms were rated alike, to within the width of the measurement. Halvorsen reads the outcome as showing that the treatment does precisely what it is sold as doing; Okonjo reads the same outcome as showing that a gain which never reaches the occupant is not the gain the regulation was written to buy. Reguera, whose argument concerns open floors rather than enclosed rooms, considers the trial beside the point, and has said so in print.

FBeneath the argument about rooms lies an older one about evidence. Everything known about acoustic annoyance arrives through somebody's report of it; a meter registers pressure at a point in the air, never the irritation of the listener standing at that point. The limitation falls unevenly. It weighs least on Halvorsen, whose claims concern rooms and word scores and need no access to feeling, and most on Reguera, whose entire case is built from what occupants say. There is a quieter dispute about the word 'standard'. Regulators, and Halvorsen with them, use it to mean a numerical threshold a room must meet before it is occupied and can be written into a contract. Okonjo uses it to mean the disturbance a population of occupants will in fact tolerate, which nobody can know until the building is full. Until those two meanings are pulled apart, a measurement that looks conclusive to one camp will keep arriving in the other as an irrelevance.