The cost of a number

For roughly half a century, one number governed the design of museum environments across Europe and North America. Fifty percent relative humidity, held constant, became the thing a building was judged against, the thing a loan agreement specified, and the thing a great deal of expensive plant was procured to deliver.

It was never derived from an experiment that established fifty as the optimum. It emerged from mid-century practice at a moment when air conditioning was newly available and the reasonable working assumption was that a constant environment must be a safe one. That assumption was defensible. What happened next was not: a working assumption hardened into a specification, and the specification acquired a precision the underlying evidence never had.

What it bought

It is worth being fair about the benefits. The move to controlled environments in the mid-twentieth century coincided with a genuine reduction in the kind of catastrophic climate-driven damage that had been routine. Buildings that had swung wildly stopped swinging wildly, and collections benefited.

The problem was not that control was worthless. It was that the cost curve is very steep at the tight end, and nobody was asked whether the last few percentage points were worth what they cost.

What it cost

Three things, and the third is the one that matters most now.

Energy. Holding a narrow band in a leaky historic building means running plant continuously against the weather. The carbon cost of that has become impossible to defend, and it was the proximate reason the major lending institutions revisited their conditions.

Buildings. Delivering tight control requires ducts, plant rooms and sealed envelopes, all of which have to be inserted into structures that are themselves the collection in many cases. A number of historic interiors were damaged in the course of protecting the objects inside them.

Attention. This is the least discussed and probably the largest. An institution that spends its capital budget on air handling has spent it. The monitoring traps that would have caught the moth infestation cost almost nothing and were not bought, because the environmental problem had been defined as a humidity problem and solved with plant.

What the revision actually says

The current position is not that humidity does not matter. It is narrower and more useful than that.

Rate of change matters more than absolute value across the middle of the range. Objects that have been in equilibrium with a set of conditions for a long time are usually safest kept there. There are genuine thresholds at the extremes, above which mould becomes a real risk and below which mechanical damage does. And between those thresholds, the precise figure is far less important than whether it moves.

None of that is a licence to stop caring. It is a redistribution of where the caring goes.

What it means for a small collection

The revision is unambiguously good news for institutions that could never have afforded the tight specification in the first place, and the practical consequences are worth stating plainly.

A building that sits stably at an unremarkable value is not failing. It was never failing; it was being measured against a specification that was itself an assumption. Documenting what it does, and being able to explain why that is acceptable for the collection it holds, is a stronger position than aspiring to a class the building cannot reach.

Money is better spent on the cheap interventions that address the agents nobody measured. Traps. Boxes. A logger. Blinds on the south windows. Fixing the gutter above the store. Every one of those returns more per pound than a percentage point of humidity control, and every one of them was available all along.

And the tightest control available should go to the smallest volume that needs it. A sealed case with conditioned silica gel delivers, for one cubic metre, what a plant room delivers for a thousand, at a fraction of the cost and with no risk of failing at three in the morning over a bank holiday.

The general lesson

The useful thing to take from this is not a new number. It is a suspicion of any single number presented without its evidence and without its cost.

When a specification arrives, the questions worth asking are what it is protecting against, how strong the evidence is, what it costs to deliver, and what else that money could have prevented. Those questions would have been worth asking in 1975, and they are worth asking of whatever replaces the current consensus.