You've got a fourteenth-century window with a mullion that's spalled down to the rubble core, and the glass is still in it. The stonemason who built it is dead. The quarry that supplied the stone has been a housing estate for four hundred years. And the faculty committee — or the vestry, or the heritage officer — wants a drawing by Friday.
Carving replacement stones for Gothic tracery windows is not restoration in the sense most people imagine. You are not repairing a broken object. You are making a new object that has to sit inside a ruin, accept loads from glass that was cut for a different piece of stone, and look like it has always been there. Get the mineral wrong and the new piece will eat the old one within a decade. Get the profile wrong by three millimetres and the window won't close.
The process is slower, stranger and more forensic than the finished result suggests. Here's how it actually goes, including the places where it goes wrong.
Key Takeaways
- Survey before you carve. A replacement tracery stone is only as good as the record you take of the original — photogrammetry or a contact mould, not a tape measure and a sketch.
- Stone selection is the decision that outlives you. Porosity, frost resistance and the way a new stone weathers against old are what determine whether the repair lasts thirty years or fails in eight.
- Carve wet, fix dry, but never use cement-rich mortar against historic limestone — the salts and the rigidity will do the damage.
- Nobody carves a mullion from a photograph. The profile template matters more than the drawing.
- Budget for joinery, not just stone. Refitting leaded glass and ironwork is typically a bigger line item than the carving itself.
Why Gothic tracery is a different beast to a plain block
A quoin is a box. A coping is a ramp. Tracery is a skeleton.
Look closely at a Decorated-period window and you'll see that the stonework is doing three jobs at once: carrying the weight of the wall above, holding the glass and leading in place against wind, and reading as a pattern from twenty metres away. The section of a mullion is not a rectangle — it's a moulded profile with hollows, fillets and rolls that catch light and shadow in a particular sequence. Change one of those rolls and the window stops looking medieval and starts looking like a 1970s pastiche.
The profiles you cannot fake
Each period has its own grammar. Early English lancets favour bold, simple rounds with deep hollows. Decorated work gets fussier, with intersecting patterns and a fillet that runs continuously around the curve. Perpendicular goes all vertical, with long straight mullions and shallow, almost sketchy mouldings. If you are carving a replacement stone for a window of a particular date, the profile has to belong to that grammar.
The trap is that most windows aren't pure. I once worked on a chancel window where the sill was fourteenth century, three mullion bases were replaced in the 1890s with something the Victorian mason had invented, and the head tracery had been patched in the 1950s in a stone that had since gone quite orange. Which profile did we carve? The client assumed the Victorian work.
We went back to the pre-restoration drawings in the diocesan archive and found the 1890s mason had got it wrong. We carved to the original, not to the surviving evidence.
Surveying the original before you touch anything
Here is the step that separates a repair that holds from a repair that's back on the scaffold in nine years. You need a record of the existing stone, not the imagined original.
How do you measure a tracery stone accurately?
For most pieces, photogrammetry plus a physical profile gauge gets you there. Photogrammetry — hundreds of overlapping photographs processed into a dense mesh — gives you the overall geometry, the twist in a mullion, the way a transom has sagged. It will not give you the fine detail of a moulding to the millimetre, because the surface is eroded and speckled with lichen. So you take a profile gauge or a contact mould in silicone or plaster across the section at three or four points along the stone, and you combine the two.
On one job I insisted we take a plaster contact mould of a badly weathered window head. It took half a day and cost the client a few hundred pounds. It also revealed that the original mullion section was asymmetric — the outer roll was larger than the inner, a detail that had been completely lost on the crumbled face. Without the mould, we would have carved a symmetrical profile and the window would have looked subtly wrong from the road. Subtly wrong is the worst kind.
- Total station or photogrammetry for overall geometry and set-out
- Profile gauges and silicone moulds for section detail
- Condition survey with a hammer — tap the stone and listen. Dull thud means delamination under the surface, even if it looks sound
- Photographs of the archive drawings, not just of the stone
- And a moisture reading, because you need to know what the wall is doing
Choosing the replacement stone
This is the decision almost nobody gets excited about, and the one that determines the outcome. You can carve a perfect stone and ruin the job in the quarry.
What stone should you use to replace Gothic tracery?
The short answer: as close as you can get to the original in mineralogy, porosity and coefficient of thermal expansion — not just colour. A match that looks right on the day of fixing and is wrong mineralogically will fail faster than a slightly mismatched stone that behaves similarly.
In practice, historic tracery in Britain and much of northern Europe is a fine-grained limestone or a magnesian limestone, occasionally a chalk, sometimes a sandstone. Indiana limestone gets used widely on American church work because it's consistent and carves cleanly. Portland stone is a common substitute in southern England. Each behaves differently when wet.
| Stone | Porosity | Frost behaviour | Where it goes wrong |
|---|---|---|---|
| Fine-grained limestone (close to original) | Moderate | Good if it's a frost-resistant bed | Quarry variation between beds — always test the actual block |
| Indiana limestone | Moderate | Generally reliable | Colour can read too cool and too even against weathered old stone |
| Portland stone | Moderate to high | Good in exposed conditions | Harder than many historic limestones — will not wear at the same rate, so the joint stays proud |
| Chalk or soft limestone | High | Poor in freeze-thaw | Absorbs water, spalls, and takes salts from the wall |
| Sandstone | Variable | Very dependent on bedding direction | Carved across the bed, it delaminates |
The frost test matters more than most people realise. A stone can look perfect and have a pore structure that traps water; three winters later the face has gone. Ask the quarry for the actual block's properties, not the general specification for the bed. I have seen a mason carve a beautiful tracery light in a stone that was technically correct and visually wrong because the particular block was harder and paler than the surrounding work.
Carve a sample first. A test piece the size of your hand, in the actual stone, finished and left outside for a winter, is worth more than any datasheet.
The carving itself
Tracery carving is done in the workshop wherever possible, on a bench, with the profile template in front of you and a drawing of the whole window on the wall. On-site carving happens only where the piece is too big or too fragile to move — a large window head, for instance — and it's slower and more expensive.
The sequence is unglamorous:
- Dress the block square and check it for faults — a hidden shake or a hard nodule will wreck the profile halfway through
- Set out from the profile template at both ends, so that any error is distributed rather than revealed at one end
- Rough out with a point and a bolster, leaving a generous margin
- Work the moulding with a series of moulding planes or chisels, guided by the template at regular intervals
- Finish with a drag or abrasive to a surface that is slightly less crisp than the eroded original will read from a distance — a razor-sharp virgin moulding next to a worn one shouts "new"
The last point is contested and I'll defend it anyway. Over-crisping a replacement stone is the most common aesthetic error I see. The eye picks up the difference in edge hardness long before it picks up the difference in colour.
Anchoring, fitting and the joint that betrays you
You've got a beautiful piece of carved stone and now you have to put it in a wall, next to old stone, and hold it there for a century.
Do not use cement-rich mortar. Historic limestone and cement are a bad marriage: the cement is harder and less permeable than the stone, so moisture gets trapped behind the joint and drives salt crystallisation into the masonry. That's how you get a spalled face two years after a beautiful repair. Use a lime-based mortar matched in strength, permeability and colour to the surrounding work. Lime putty, or a natural hydraulic lime of low strength, is the standard answer.
Anchoring is where modern methods and conservation principles argue. Stainless steel dowels and pins are entirely normal and appropriate — they must be positioned so that they don't create a hard point that concentrates stress, and so that the stone can move slightly with thermal cycling. Rigid fixing of a piece that wants to expand will crack it.
And the joint: a hairline butt joint with lime mortar, not a wide bed of cement. If your new stone is set with a fat mortar joint, the joint will weather proud and the whole repair will read as a patch, which is exactly what the conservation officer will object to.
What it costs and what it takes
Nobody likes the numbers, so here's the honest shape of them.
A single replacement tracery light — a cusped opening within a window — takes an experienced carver between two and five days in the workshop, depending on how many orders of moulding and how much fine carving there is in the cusps. A whole mullion with a moulded section, several days. A window head with intersecting tracery and carved cusps: weeks, not days.
Site time is a completely separate budget. Removing the old stone without destroying the glass, recording everything, preparing the opening, fitting, pointing, and refitting the leadwork — that is often the larger share of the job. I've seen projects where the carving was quoted at four figures and the total cost of the replacement was an order of magnitude higher.
The real cost of getting it wrong, though, is not the invoice. It's the scaffold you put up again in eight years, and the stone you can't source the second time.
The mistake I made, and the one you should avoid
My first serious tracery repair, I got the profile wrong. Not badly — by about four millimetres on one roll. It was set out from the drawing because the original was too eroded to measure properly. The piece fitted, the glass went back, and everyone was happy until the architect photographed it against the matching mullion next door. You could see it. Once you saw it, you couldn't unsee it.
We recarved it. The client was gracious about it, which I did not deserve.
The lesson is not "measure more carefully" — that's trivial. The lesson is that if the original is too eroded to measure, you must find another source: an archival drawing, a matching window elsewhere in the same building, a fragment of the same profile in another window. There is almost always another source. What there isn't, is time to look for it if you didn't build that into the programme.
So build it in. And when the committee asks why the survey stage takes a quarter of the calendar, tell them it takes a quarter of the calendar so that the next survey is in fifty years, not fifteen. That's the actual argument, and it's a good one.
