A base needs an edge to point at

The tree method turned origami design from a craft into an algorithm, and its output is a uniaxial base: a flat, many-pointed arrangement with one flap per limb of the subject, of the right lengths, ready to be shaped.

Every one of those points is made of paper from the sheet’s boundary, and until now that has been a fact so obvious it was not worth writing.

Where a point comes from #

Take a folded base and look at the tip of one flap. It is a point of the folded object where several layers of paper come together, and following those layers back onto the flat sheet, each of them ends at the sheet’s edge.

That is not an accident of construction. A flap’s tip is where the paper stops, and the only places the paper stops are the boundary. An interior point of the sheet has paper all round it; it can be the apex of a cone or the centre of a twist, and it cannot be the end of a flap.

The circle argument says so already #

A flap costs a circle: a flap of length $L$ consumes every point within $L$ of its attachment, because folding is an isometry and the paper reaching the tip is drawn from a disc.

The disc is centred at the flap’s attachment — where the flap meets the body — and the flap’s tip is at distance $L$ from it, on the disc’s rim.

So the tip is at the boundary of the region the flap claims, and for the flap to be a point rather than a fold, that boundary has to coincide with the sheet’s own. A flap whose disc is entirely in the interior does not end in a point; it ends wherever the neighbouring flap’s paper starts, which is a crease rather than a tip.

Checking it with a base #

The claim is easy to test on anything already folded, and it is worth testing because it is the sort of thing that feels true and might not be.

Take any traditional base — a bird base from a square, a fish base, a preliminary base — and unfold it. Mark the four corners of the square with a pencil.

Now refold. Every point of the base has one of the marks at its tip, or is made from a piece of the square’s edge.

Try it on a crane. The head, the tail and the two wingtips are the square’s four corners; the body is the middle of the sheet, and the middle produces no point at all.

That is the whole of the claim, and doing it on two or three bases makes it hard to doubt. What the experiment cannot show is the general statement, which is that no interior point can ever be a tip — for that the argument about isometry is needed, and it takes two lines.

The argument, in two lines #

A flap’s tip is a point of the folded object at which the paper ends. Take a small circle round that point in the folded object and follow it back to the sheet: it maps to a set of arcs, one per layer, each lying on the sheet.

If the tip came from an interior point of the sheet, the small circle would pull back to a full circle of paper round that point, and a full circle of paper cannot fold into a point without the total angle being a multiple of a full turn — which is developability, and a vertex satisfying it folds into a fan rather than into a point.

If the tip came from a boundary point, the pull-back is arcs with ends, and the ends are the sheet’s edge. That folds into a point without contradiction.

So a tip is boundary, and the reason is the same reason a cone cannot be flattened: angle has to be accounted for, and only the boundary has any to spare.

What a hole would give #

The reverse operation is instructive, since it is the one design has actually used.

Cutting a hole in the middle of a sheet adds a boundary circle, and a boundary circle is somewhere a flap can end. So a holed sheet can have flaps pointing inward — attached near the hole, with their tips on the hole’s edge — which is a design space a plain square does not have.

That is part of why a hole is cheap paper: it removes the material that could not produce a point and adds an edge that can.

It is also why the operation is worth a designer’s attention and gluing is not. A hole adds the resource; a gluing removes it. And both add a parity condition, so the hole is the one that pays for its condition.

Where the boundary goes in a packing #

Once the boundary is understood as the resource, the packing argument reads slightly differently and it is worth walking through.

A packing places one circle per flap inside the sheet’s outline, with rivers between them. The circles must not overlap, because two flaps cannot claim the same paper.

What the outline is doing in that picture is two jobs at once. It bounds the area available, which is the constraint everybody states. And it is where the tips go, which is not stated and is why circles are placed touching the outline whenever possible — a designer packs to the edges not merely to save area but because a flap whose circle does not reach the boundary is not a flap.

That second job explains a habit that the area account does not: an efficient packing is one where the circles crowd the boundary, and an arrangement with the same total area but all the circles in the middle is not merely less efficient, it does not produce a base at all.

So the boundary is doing double duty in the argument, and separating the two duties is what makes it clear why a closed sheet produces nothing.

What the collection has and has not done #

Worth an accounting, since the essay makes a general claim about a method.

Not computed here: any packing on a closed sheet. Nothing has been attempted, because the output would be empty and an empty output is not a measurement.

Computed here: the counts and verdicts of glued sheets, which say what such a sheet is like combinatorially. Those are the measurements the rest of this phase rests on.

Argued: everything about tips and boundary. The argument is short, it is standard, and it is not a computation.

So the essay is a piece of reasoning about a method’s scope rather than a measurement of anything, which is the right kind of essay for the question and should be read as such.

What a closed sheet produces #

Nothing, in this sense.

A torus of paper has no boundary. Every point of it has a full turn of paper round it, no flap can end anywhere, and the uniaxial method’s output is empty. Not restricted, not inefficient — empty, because the object the method produces is a set of points and there is nowhere for one to be.

That is a stronger statement than the design method being awkward on a closed sheet. It is the method having no object to produce.

What a cylinder produces #

Points at its two ends and nowhere else.

A cylinder’s boundary is two circles, so the flaps can only be attached where their discs reach one of the circles, which means near the ends. A base on a cylinder is therefore a fringe at each end with a plain tube between them, and the tube is the body.

That is a real design space and it is a small one. The subject it describes is something with limbs at two ends and nothing in the middle, which is a narrow class, and it is not what the tree method was built for.

What that says about the method #

The tree method is usually described as assuming uniaxiality — all the flaps lying along one line in the folded form — and the assumption is substantial and well known. A box, a curved shell or a tessellation is outside the method entirely.

There is a second assumption underneath it that is not usually stated: the sheet is a disc. The packing argument places circles inside an outline, the rivers run between them, and the construction that turns a packing into creases assumes the paper has one boundary curve enclosing everything.

Sheets with holes have been considered — a hole is cheap paper and the method handles them by treating the hole’s edge as more boundary. That works because a hole adds boundary.

A gluing removes it, and there is no version of the packing argument that removes boundary.

Two objects that look like counter-examples #

Both are worth handling, because a reader who folds will think of them.

A twist. The centre of a twist tessellation’s polygon looks like a point of the folded object, and it is an interior point of the sheet. It is not a flap’s tip: it is a small polygon of paper lying flat, with layers round it, and the folded object has area there rather than a point. Looking at it edge-on it may appear pointed, and measuring it shows a facet.

A sink. A sunk point is an interior point that has been pushed inside the model, and it produces something that looks like a tip pointing inward. Again it is not a tip in the sense here: the paper does not end, it turns round and comes back, and the object has layers rather than an ending.

Both are cases where the folded object has a feature at a place the sheet has no boundary, and neither is a flap. The distinction is between the paper stopping and the paper turning, and only the first makes a point.

That distinction is worth having explicitly because point in ordinary folding language means the visible shape and here it means where the material ends, and the two coincide on the bases everybody makes.

The method’s other assumptions #

Since one unstated assumption has been named, it is worth listing the stated ones so the omission is in proportion.

Uniaxial form. All the flaps lie along one line in the folded state. Stated everywhere, substantial, and it excludes boxes, curved shells and tessellations.

No cuts. The sheet is not cut, which is the whole subject’s convention rather than the method’s.

Circles do not overlap. The conservation argument, which is exact.

Rivers are not crossed. The connectivity constraint, which is what makes the packing genuinely hard.

The packing yields a pattern. Almost always true, with a standard construction, and not guaranteed.

The sheet is a disc. Not stated, invisible until there is a sheet that is not one.

Five stated and one not, and the unstated one is the only one that is a fact about the paper rather than about the method. That is a reasonable place for a hypothesis to hide.

Why this is not merely a limitation #

The interesting reading is not that the method fails on closed sheets. It is that the method’s output — a set of points — is a thing only a bounded sheet can have, and the boundary is therefore not an incidental feature of the design problem but the thing being carved up.

Read that way, the packing is a way of allocating the boundary rather than of allocating the paper. Each flap gets a piece of the sheet’s edge to end at, and the circles are the bookkeeping for how much interior each piece needs behind it.

That reframing is not new mathematics and it changes what the method looks like. A square has four units of boundary and the flaps divide them; a hexagon has six shorter ones; a cylinder has two circles; a torus has none.

The boundary as a budget #

Reading the boundary as the resource suggests a quantity nobody computes, and it is worth writing down even though nothing here measures it.

A sheet has a certain length of boundary. Each flap’s tip occupies some of it — how much depends on how thick the flap is at its end, which the tree method does not model, since its flaps are line segments with no width.

So there is a second conservation law lurking beside the circle argument: the flaps’ tips have to fit round the boundary, in order, without overlapping. For a square of side one the budget is four; for a disc of the same area it is about three and a half; for a cylinder made from that square it is two.

The tree method does not use it, because its flaps have no width and a zero-width tip occupies no boundary. A real flap has width, a real base’s tips are spread round the edge in the order the tree gives, and a designer who has ever run out of edge knows the constraint exists.

Nobody has priced it. It would need a model of flap width that the method does not have, and it is the second respect — after the tips themselves — in which the boundary is the thing being allocated.

Recorded as an observation rather than a result.

What a closed sheet is good at #

The complement is worth stating so the essay does not read as a list of absences.

A closed sheet is good at enclosing and at collapsing along an axis. A tube packs into a fraction of its length, holds a volume, and deploys — and none of those requires a point anywhere.

The engineered folded objects are almost all of that kind: booms, stents, bellows, packed antennae, airbags. They have no flaps and they were never going to be designed by a method whose output is a set of flaps.

So the two design traditions are about different sheets, and the division is not a matter of taste. A method that produces points needs a boundary to put them on, and a method that produces a collapsing tube needs a sheet that closes.

Why the assumption never bit #

A hypothesis stays unstated when nothing violates it, and it is worth asking what would have had to happen for anybody to notice.

Somebody would have had to try to design a base on a closed sheet. Nobody has, and there are two reasons.

The material reason: paper comes flat, and a designer works by folding a flat sheet and opening it out to look at the pattern. Taping a sheet closed first makes both of those impossible.

The subject reason: the things people design bases for are animals and insects, which have limbs, and limbs need points. A closed sheet produces no points, so a designer trying one would abandon it in the first minute for reasons they would describe as it does not work rather than as the method assumes a disc.

So the assumption was protected by the fact that violating it produces an immediate and obvious failure, which is the best possible protection and the worst possible way to learn what the assumption is.

It surfaces here only because the collection built closed sheets for an unrelated reason — asking what a boundary costs a search — and then had them lying around to ask other questions of.

The sentence that was missing #

The tree method assumes a uniaxial form, which everybody says, and a sheet with a boundary, which nobody does.

The second was invisible because every sheet anybody designed on had one. Naming it costs a clause and it clarifies why the method has the scope it does: not because closed forms are hard for it, but because its output is a set of points and a closed sheet has nowhere to put one.