The keystone over the door had lost its face. Not cracked — gone, weathered back about an inch and a half until the carved scroll was a suggestion rather than a shape. Next door, four houses down a row built in the same year by the same hands, the identical keystone sat there sharp enough to cast a shadow. That is the quiet gift of a speculative terrace: the missing piece usually still exists somewhere on the block, and replicating it is an exercise in surveying the neighbours rather than inventing a detail.

Carved, cast, or pressed — tell them apart first

Three things get called stone on a row house front and they behave differently. Carved natural stone shows bedding planes and tool marks, and its weathering follows the grain of the bed. Cast stone is a precast concrete with a fine facing mix, usually with a visible aggregate where it has chipped and a crisp, repeated profile because it came from a mould. Pressed terracotta is hollow-backed, glazed or slipped on the face, and rings differently under a hammer. Getting this wrong at survey stage sends the repair down the wrong road entirely, because a patch mortar that suits soft sandstone will fail on a dense cast unit, and a mould-made replacement inserted among hand-carved neighbours will read as a photocopy.

Taking the pattern off a surviving unit

Where a good example survives nearby, the sequence is to clean a test area gently, take a silicone squeeze of the profile, and back it with a plaster mother mould so the rubber holds its shape. For a low-relief unit a profile gauge and a set of square-on photographs with a scale rule will get you most of the way. The awkward cases are units that have weathered asymmetrically, where the surviving half has to be mirrored and the judgement call is whether the original was symmetrical at all — on hand-carved work it often was not, and copying the irregularity is what stops the new piece looking mechanical. I keep the mould after the job, labelled with the street and the year, because the same block tends to come back.

Choosing the replacement material

The governing rule in older districts is compatibility: the new unit should be no stronger and no less permeable than the material around it. Dense, high-cement replacements next to soft original stone push moisture and salt into the old material and spall it from behind, which is a slow failure that only becomes visible five or ten years later. For a single missing keystone in a carved sandstone front, hand-carving a matched stone is usually the right answer despite the cost, because one unit does not justify a mould. For a run of repeating brackets, a cast unit in a lime-rich mix with a matched aggregate gives a better result per dollar. On the older streets in the District the local practice leans toward matching the original bedding and jointing exactly, and a shop like Masonry DC will generally want the mortar analysis in hand before the replacement mix is signed off.

Setting the piece so it lasts

Installation is where good replication gets undone. A carved stone must be set on its natural bed, the same way it lay in the quarry, or it will delaminate in sheets within a couple of decades. Anchors should be stainless steel — the rusting ferrous cramp is the single most common cause of split stonework on nineteenth-century fronts, and reusing the original pocket without replacing the metal repeats the failure. Joints want to be the same width as the surrounding historic joints, typically 1/4 to 3/8 in on a fine ashlar front and wider on rubble work, and pointed with a mortar softer than the stone. Anything that puts a hard, impermeable line around a soft unit turns the joint into a dam.

What to expect on cost and programme

A hand-carved replacement unit in matched stone is bench work, typically several days per unit plus the survey and templating, and it prices accordingly — expect the survey, template and single carved unit for a modest keystone or console to run into the low thousands. Cast replication drops sharply per unit once the mould exists: the mould is most of the cost, and the second through the tenth casting are comparatively cheap. Historic district review, where it applies, adds weeks rather than days, and it is far smoother when the application includes the profile drawings and the mortar analysis rather than a photograph and a promise.

Frequently asked questions

Can a badly weathered carved detail be built back up rather than replaced? Often yes. A lime-based repair mortar, built in layers over stainless armature where depth demands it, is standard conservation practice for losses up to an inch or two and is reversible in a way that a replacement unit is not.

Does a replacement need to match the colour exactly on day one? No, and trying to force a match usually means pigmenting the mix, which ages badly. A correctly chosen material weathers toward its neighbours over a few seasons.

Is cleaning the facade necessary before replicating a unit? A local test clean is, because soiling hides both the true colour and the surface texture you are trying to match. Whole-facade cleaning is a separate decision with its own risks to soft stone.