Why a church is not a concert hall for amplified speech

By Joey Calkins · Section Sound

Both buildings are large, both sound impressive when empty, and both are routinely booked for the same evening of spoken word. They are not interchangeable, and the reason is not taste. A concert hall is designed around what unamplified music needs. A church is designed around what a congregation needs, which historically meant volume, height and stone. Neither was designed around a play.

Start with the arithmetic. Reverberation time rises with volume and falls with absorption, so what matters is not the size of a room but the cubic volume per person and the softness of the surfaces those people sit on. A church nave gives every occupant a great deal of air and very little upholstery. Add a congregation of forty to a space built for four hundred and the absorption barely moves while the volume stays exactly the same.

Concert halls are tuned, not merely large

A hall built for symphonic repertoire is aiming for a long but well-behaved decay, typically somewhere near two seconds when occupied, with the early reflections arriving from the side rather than from overhead. The academic literature on hall design, most prominently the survey work of Leo Beranek, treats early lateral energy and the initial time delay gap as the parameters that separate a celebrated hall from a merely loud one. That work is scholarly rather than commercial, which is why it is worth naming.

Two design habits follow from it and both matter here. Diffusers, coffering and irregular surface relief are put in deliberately so that reflections scatter rather than concentrate. And concave surfaces are avoided or broken up, because a curve focuses.

What a vault does that a ceiling does not

Apses, domes and barrel vaults are concave, and concave surfaces behave like mirrors with a focal point. Energy that strikes them does not scatter evenly across the room; it converges on particular seats and skips others. The audible result is a hall where one row hears a slap that the row behind does not, and where a performer moving two metres upstage changes the balance for a quarter of the house.

Long tails and focusing together produce the characteristic problem of the converted chapel: a room that flatters a hymn and destroys a sentence. Sustained sung tone benefits from a tail that keeps the previous note alive underneath the current one. Speech needs the previous syllable to get out of the way.

The remedy is distance, not power

Since the ratio of direct to reflected sound cannot be improved by turning the system up, the only lever that scales is putting a source closer to the listener. That is what distributed systems do. Instead of one loud position at the front, several quieter positions are placed down the length of the room, each covering the rows near it.

This works because of the precedence effect: when the same sound arrives twice within a short window, a listener localises it to the first arrival and hears the second as part of the same event rather than as an echo. Delaying each downstream loudspeaker so that the sound from the stage still arrives first keeps the performance apparently on the stage while the nearby box does the work of intelligibility. Get the delay wrong and the audience hears the loudspeaker beside them instead of the actor in front of them.

Column arrangements address the vault directly. A tall, narrow source produces a narrow vertical pattern, which can be aimed down onto the seated audience and away from the ceiling. Energy that never reaches the vault never comes back from it. This is why so many churches end up with slim vertical loudspeakers on the piers rather than a conventional pair at the front, and the choice is acoustic rather than aesthetic.

Noise sets the floor

Reverberation gets the attention, but the steady background level in a room decides how much of the decay actually matters. Ventilation plant, a boiler running under the floor, traffic through single glazing and the hum of older lighting all raise that floor, and every decibel of it eats a decibel of the quiet consonants the tail was already threatening. A room measured empty on a Sunday morning and a room occupied on a Friday night with the heating on are different rooms.

This is why the cheapest intelligibility improvement in many buildings is not acoustic at all. Turning the ventilation off for the duration of a performance costs nothing and routinely buys more than a larger loudspeaker would.

What none of it fixes

Distributed systems raise intelligibility and complicate everything else. Each position needs its own delay and level, which means the design must be commissioned rather than merely installed. Sightlines suffer. Cabling in a listed building is frequently the hardest part of the job, and in many cases the fabric cannot be drilled at all.

The unamplified performer gains nothing from any of it. A recitalist, a string quartet or an unmiked chorus still faces the room as built, and in a heavily reverberant nave the answer is repertoire and tempo rather than equipment.

What we cannot verify

Reverberation figures for historic buildings are often quoted from surveys that did not state occupancy, source position or measurement method, and such numbers cannot be compared with one another. We have not measured any specific building, and nothing here should be read as a description of one. Design guidance published by loudspeaker manufacturers on distributed systems is produced by parties selling the loudspeakers, and while the underlying physics is not in dispute, the recommended quantity of product is not independent advice.