Walls Between Cells

Every interior wall belongs to two rooms and is owned by neither. That's the whole problem. Each room gets a say in what the wall has to be (its fire rating, how much sound it stops, what's inside it, how thick it is), but there's only one wall. When one side changes, the other side pays for it.

Who gives up area when a shared wall thickens

The essay's round numbers, not a wall-type specification: a 5 in partition is drawn, and the built assembly can be 5 to 12 in. Unit A's living room is drawn 14 ft clear and Unit B's bedroom 11 ft clear. Area given up = how far that room's face moves × the wall's length. Ratings, acoustics and the assembly itself are not modelled; the adjacency never changes.

The drawing is one such wall. Unit A's living room and Unit B's bedroom share a 30-foot party wall, drawn as a 5-inch partition. Set the thickness the built assembly actually needs, then choose the datum, the line that holds still while the wall grows: its centerline, Unit A's face or Unit B's face. The coral strips are the floor area each side gives up, shown at plan scale and again in an enlarged section through the wall. In every example the two rooms, their adjacency and the wall's length stay the same. Only the thickness and the datum change, and they decide who pays. The rest of this post is about why that decision belongs to the pair of rooms and not to the wall.

That's also the line between this thesis and #27, Walls as Edges: #27 treats the wall as an edge in the adjacency graph, where all that matters is that two rooms touch, while this one treats it as the physical thing they share, with a rating, an assembly, a thickness and two owners with opinions.

The pair decides the assembly

Every interior wall exists because two cells need to be kept apart. The bedroom wall exists because sleeping and living need separation: acoustic, visual, thermal. The corridor wall exists because circulation and occupancy have to be separated for egress and fire. The mechanical room wall exists because equipment noise and access have to stay out of the occupied space. Each wall exists because of the rooms on either side of it. On its own, it means nothing.

So you can't specify a wall by looking at it. You specify it by reading both of its faces. As #45 lays out, the wall between two bedrooms inside one dwelling unit typically needs no fire separation, a bedroom wall facing the common corridor usually needs a 1-hour assembly, and a wall between two units commonly needs 1 or 2 hours depending on construction type. A wall between two residential units needs a high STC (sound transmission class, the standard measure of how much airborne sound a wall stops) rating because privacy between neighbors is both a market requirement and a code requirement. Between a unit and a common corridor, fire separation dominates and the acoustic requirement may be different. Move the boundary a few feet and the specification can change completely, because the pair it sits between has changed. The wall doesn't define the boundary. The boundary defines the wall.

Who owns the thickness

Here's the part a graph can't see. Take two apartments side by side. Unit A's living room and Unit B's bedroom meet at a party wall 30 feet long. The plan was drawn with a plain partition there. Say it's about 5 inches thick; these are round numbers for the arithmetic, not a spec. Then someone notices it's a unit separation, and the rated, acoustically separated assembly comes in at, say, 10 inches. The wall just gained 5 inches. Where do they go?

• Hold the centerline. Each side gives up 2.5 inches. Over 30 feet, that's 2.5/12 × 30 = 6.25 ft² per unit, split evenly. Nobody's happy, but nobody got singled out.
• Hold Unit A's face (because A's kitchen run was already dimensioned to it). Unit B absorbs all 5 inches: 5/12 × 30 = 12.5 ft² out of one bedroom, whose clear width drops from 11 feet to 10 feet 7 inches. If that bedroom was drawn at the narrowest width you were willing to sell, it isn't that bedroom anymore.

Those are the drawing's first two examples; the third holds Unit B's face and sends the whole 12.5 ft² the other way. The total never changes: 5 extra inches over 30 feet is 12.5 ft² however you split it. The datum only decides whose plan it comes out of.

Neither unit's program changed. Neither unit's adjacency changed. In #27's graph, nothing happened: A still touches B. But under the second choice Unit B's bedroom gave up all 12.5 ft² of its width margin to a wall upgrade nothing in Unit B's plan asked for. That's what "boundary between cells" means in practice. The boundary has thickness, the thickness has to come out of somebody, and if you didn't decide whose side owns it, the last person to dimension a cabinet decided for you.

The same thing happens without any code trigger. Flip Unit A so its kitchen backs onto the party wall and that wall now carries plumbing, boxes and penetrations, every one of which has to be detailed so the rating and acoustic separation still hold. Unit B didn't move a thing, and its bedroom wall is now a wet wall. A change on one side of a boundary is a change on both sides, and nobody from B's side was in the room when it was decided.

The same boundary at other scales

Modular construction pushes this to its conclusion. A module is a cell that shows up on site with its boundary already built. The boundary between two modules isn't a wall anymore. It's a joint, a connection detail that handles structure, sound, fire and weather at the module interface. The question stops being “how do we build this wall?” and becomes “how do these two cells meet at their shared boundary?” Same idea, different physical form, and the thickness question usually gets answered at the factory, because each module tends to bring its own wall to the joint.

Open office plans are the cell without boundaries. The floor plate is one big cell, and leaving out interior walls is the point: you can reconfigure it, light reaches deeper, and it costs less to build. But open plans don't eliminate the need for boundaries; they pass it to furniture, acoustic panels and social convention. The building doesn't provide the boundary, so the tenant has to. That works until somebody needs a private phone call, and then the missing wall becomes a usability failure.

Energy models already think this way. Each enclosed cell is a thermal zone, and the wall between zones is a boundary condition. Its thermal properties (R-value, thermal mass, solar exposure) affect energy use on both sides. The wall's thermal performance matters because of the cells it separates, not because of the wall itself.

The judgment

It's common to treat wall type as a tag added late, in the construction documents (the CD set). I think that's backwards. Before a plan leaves schematic design, every boundary should carry three things: the pair it separates (which determines the assembly), a stated dimensioning datum (centerline or face), and which side gives up area when the assembly grows. A plan that only knows adjacency can tell you that A touches B. It can't tell you who loses five inches. A plan that knows its boundaries can.