Tracing a leak

Why does the wet patch on a ceiling so rarely sit under the actual hole?

A brown ring appears on the bedroom ceiling. The obvious move is to go into the loft, stand directly above it, and look for the hole. Almost everyone does this, and almost everyone finds nothing there.

That failure is so reliable it deserves a proper explanation. Water falls straight down; the stain is where the water came out; so why is the entry point not overhead? The answer turns out to be less about plumbing than about what a single observation can and cannot tell you — and once that is clear, the whole business of finding leaks changes shape from searching to testing.

Follow one drop of water in from a defect in a roof. It lands on the underside of a tile, or on the sarking felt beneath, and now it is on a surface that is not horizontal. Gravity pulls it down that surface, not through it. Surface tension holds it against the material rather than letting it fall free, so it runs — along the felt, sideways to a batten, down a rafter, along the top edge of a purlin. Water in a building behaves like water on a windowpane: it travels far more horizontally than you expect before it ever drops.

Where does it finally leave? At the first place it can. That means a discontinuity: the end of a batten, a nail point, a joint in the felt, a screw head, the low point of a sagging cable. It drips there and hits the plasterboard, which absorbs it and spreads it further by capillary action through the paper and gypsum until the wet area is broad and its centre no longer marks anything in particular.

So look at what the stain actually is. It is the last point in a path, and it is often several metres from the first. Worse, plumbing does the same thing with more geometry available: a leaking joint under a bathroom floor will find a pipe or a joist and run along it, so the water can appear in a different room entirely, and pressurised supply pipes can push water some distance before it emerges at all.

Now state the epistemic shape of the problem, because that is what makes leaks hard rather than merely inconvenient. Many different entry points can produce the same stain. The mapping from cause to observation is many-to-one, and a many-to-one mapping cannot be inverted — no amount of staring at the ceiling will recover which of the candidate entry points was responsible, because they all predict roughly the same thing you are looking at.

What do you do when an observation cannot be inverted? You stop treating it as a search and start treating it as a test. If passive observation cannot distinguish the hypotheses, you have to make them predict different things, which means intervening. This is why the professional method for a suspected roof leak is a hose test conducted from the lowest suspect point upward, wetting one small area at a time, with someone watching inside, waiting several minutes between areas. Working upward matters: soak the ridge first and you wet everything below it, and every hypothesis is confirmed at once, which is the same as none being tested.

Notice the structure of that. Each section wetted is a single hypothesis given the chance to falsify itself, and the order is chosen so that a positive result implicates one candidate rather than a whole region. The waiting is not politeness; it is allowing for the travel time along the path, which for a slow leak can be considerable.

The other instruments follow the same logic of forcing a difference. A moisture meter maps the extent of damp behind an intact surface, so the shape of the wet area can be read rather than guessed, and the highest damp point often lies nearer the source than the visible stain. Thermal imaging does not see water — a common misconception — but sees temperature, and damp material differs from dry material both in evaporative cooling and in how quickly it changes temperature, so it shows the wet region indirectly. Fluorescent tracer dye put into a suspect system answers a different question again: not where the water goes, but which system it came from. Acoustic detection listens for the noise a pressurised pipe makes escaping, which works only because the pipe is pressurised.

One correction belongs here, because it prevents a whole class of fruitless hose tests. Not every ceiling stain is a leak. Interstitial condensation deposits water inside a cold roof space when warm moist indoor air reaches a cold surface, and it produces damp patches, mould and drips that look convincingly like ingress. The diagnostic is timing: rain-driven ingress correlates with rain, and condensation correlates with cold, still weather and with what is happening inside the house. Checking that correlation before climbing anywhere is close to free.

Three things follow that are easy to get wrong.

First, the stain's position is not useless — it is a weak constraint rather than no constraint. The source is almost always above it and usually not far away in plan, so it narrows the search space considerably. It simply cannot pick a point.

Second, several defects can feed one stain, and this is common on old roofs. Finding a defect and fixing it is not the same as finding the defect, which is why a leak that improves but does not stop is a normal outcome rather than evidence of a bad repair.

Third, the arrival of water at the ceiling is not proportional to the size of the entry. A pinhole feeding a path that concentrates and channels can deliver more visible water than a wide gap that merely dampens a large area, so the size of the stain is a poor guide to the size of the hole.

• How pressure testing isolates a leaking circuit in a heating system.
• Why cold-roof condensation is designed against with ventilation rather than sealing.
• What thermal imaging can and cannot establish about moisture behind a surface.

Every term the collection defines is gathered in the glossary.