Running Hand

Running Hand

Fair Copy, the sheet before this one, starts from a typeface and works backwards: it rasterises a Mincho glyph, blurs it, and redraws the contour where that blurred field crosses a wandering threshold. It is a good trick and it produces a bad forgery, for one reason that no amount of extra distortion fixes — a warped outline is still an outline. The skeleton underneath is a designer’s, not a writer’s. The stems of an n stay identical twins; the terminals are still the flat cuts a punchcutter made; the stroke is thick and thin in the places the typeface decided, not in the places a hand would have slowed down.

This sheet throws the outline away and keeps the journey.

A letter here is stored as the path the pen takes through it — an ordered list of strokes, each a handful of points, in a box ruled with an ascender line, an x-height, a baseline and a descender line. Nothing is filled. The mark you see is built by walking that path and asking, at every step of about a fiftieth of a letter, how fast the nib is going and how wide a mark that leaves.

Three things come out of the walking, and between them they are most of what reads as handwritten:

The pen slows into every curve. Human drawing obeys a law with a rather grand name — the two-thirds power law — which says that the speed of the hand falls off as roughly the cube root of the radius of the curve it is describing. Tight turn, slow pen. Since a slow pen leaves a wider mark, the thick parts of every letter land where the writing turns, not where a type designer put a stem. Turn Speed down and watch the bowls of a, o and g swell at their turns while the straight run of an l goes thin.

Every stroke starts and ends from rest. A hand accelerates out of a standstill, peaks a little before the middle of the stroke, and coasts down again — the lognormal velocity profile that the kinematic theory of rapid movements predicts. So each stroke tapers in and tapers out on its own, which is the thing a printed terminal can never do.

The error is remembered. Each letter is also pushed out of shape as a whole — four corner offsets, interpolated across its box, so it comes out a different but still plausible letter rather than a ragged one. The wander along a stroke is smooth noise sampled by distance travelled, not a fresh random number per point, and the writer’s baseline, slant and size are carried forward by a decaying walk: today’s error is mostly yesterday’s error. That is the difference between drifting off the line — which people do, in long sags that correct themselves half a word later — and vibrating around it, which nobody does. Drift sets how far the writer wanders; Fatigue adds the one-way part, the sinking and shrinking that happens down a page.

Four controls set how far apart things sit, because in a hand these really are four different habits. Tracking is the mean gap between roman letters and Kana gap the mean gap between Japanese ones; Word gap is the space between words, which people keep far more loosely than they keep letter gaps; and Rhythm is how unevenly all of it is kept. Rhythm is the one that matters most here — it is drawn from a walk that barely remembers itself, so each gap is nearly a fresh error, and a line of text gets the uneven colour that set text never has. Turn it to zero and the spacing goes metronomic, which reads as type however good the letters are.

Then the pen keeps moving between letters. Where a joining hand would carry on, a light undercurve is drawn from where the last letter finished to where the next one starts, dipping to the line in between; letters that finish high — o, v, w, b, r — hand over at x-height instead, and the ones that drop below the line start the next letter cold. Which pairs get joined is partly the writer’s decision and partly the Speed: a hurried hand joins more.

Japanese

Only the roman alphabet, the figures and the marks are authored here as pen paths. Kana and kanji get their route a different way, and it is worth being plain about the difference: the glyph is rasterised once into a small buffer and thinned by Zhang–Suen erosion down to the one-pixel ridge running along the middle of every stroke. Thinning throws the typeface’s modulation away and keeps its skeleton — so the route through a kanji here is still, honestly, a typeface’s. What happens to that route afterwards is not.

Every stroke is finished the way a copybook says it should be. A Japanese stroke is not an even line with tapered ends the way a roman one is. It has three named parts — 起筆 the entry, 送筆 the travel, 収筆 the finish — and the finish is one of three things: 止め, stopped dead with a press; 払い, swept away to nothing; 跳ね, flicked off in a hook. Each traced stroke is sorted into 横画, 縦画, 左払い, 右払い, 跳ね or 点 from its own geometry, and then given the weight profile that belongs to it. That is why a 横画 here comes out pressed at both ends and thin in the middle, and why the long right-falling stroke of 水 or 入 swells before it leaves the paper. Brush 筆意 is how much of that profile to apply; at zero every stroke is an even line, which is what the sheet looked like before this existed.

The horizontals run uphill. 右上がり — a Japanese hand carries its horizontal strokes up to the right by a few degrees while leaving its verticals upright. That is a shear, not a rotation, and it is one of the strongest signals that a page was written rather than set.

The three sizes. A kana is written smaller than a kanji, a small kana smaller again, and 、。 smaller still. Getting these wrong is the first thing that makes set Japanese read as type; every character is also centred in its own cell rather than hung off the left of it.

Its own gap. Japanese is not tracked out the way roman letters are, so Kana gap 和文字間 is a separate control from Tracking — move one and the other stays where it was. On 原稿用紙 neither applies: the squares are printed, and every character sits in one. Line breaking follows 禁則処理 — a closing mark never begins a line, it hangs past the edge instead, and an opening bracket never ends one.

This is also the tier where the old sheet broke outright: Japanese has no spaces, so treating a paragraph as one unbreakable word ran the whole of it off the edge of the page. Text now breaks between characters, as it should.

Control	Effect
Text to write	Anything up to 420 characters, in roman or in Japanese; line breaks are kept, long lines wrap between words or, in Japanese, between characters
Specimen	A letter, a pangram with figures, a note to self, a sheet of repeated letters for comparing one a against the next, and three Japanese ones — 手紙, いろは, 俳句
Whose hand	Seven complete writers — every control below at once, because a hand is a way of moving a pen rather than a style laid over one. 楷書 sets the Japanese one: brush, sumi, 原稿用紙, columns
Paper · Ink	An exercise book with a red margin, laid writing paper with its chain lines, section paper, plain bank, 原稿用紙. Ink multiplies onto the sheet, so crossings darken
Size · Ruling	Writing size as a fraction of the sheet; whether the printed rules show
Speed	How fast the nib travels. Faster means less braking into curves, so flatter contrast; more corner-cutting; more slant; more joining
Join	How much of a running hand this is. At zero the pen lifts between every letter; at full it joins wherever a hand would
Slant	Mean lean. Negative reads as left-handed, and the Left-handed preset pairs it with a steeper nib angle
Drift	How far the writer’s baseline, slant and size wander — as a slow walk, not as jitter
Fatigue	The one-way part of that walk: sinking below the line and shrinking, down the page
Wobble	Amplitude of the tremor along each stroke, applied across the direction of travel
Roundness	How much of a corner the hand cuts. A hurried writer cannot turn a sharp corner, so its v and z come out as curves
Stretch	Length of ascenders and descenders against a fixed x-height — the most personal ratio in a hand
Rise 右上がり	How far the horizontal strokes of a Japanese character run uphill to the right. Roman letters ignore it
Tracking	Mean gap between roman letters
Kana gap 和文字間	Mean gap between Japanese characters, kept separate from Tracking because a hand keeps the two differently. Ignored on 原稿用紙
Word gap	Width of the space between words — the loosest measurement on the sheet
Rhythm	How unevenly every one of those gaps is kept. At zero the spacing is metronomic; nothing else on this page gives away a machine faster
Line	Distance between the ruled lines
Nib	Fountain, ballpoint, pencil, felt tip, 筆, broad italic. Each is a different relation between speed and mark width, and a different thing to do at a stop
Brush 筆意	How strongly the 楷書 weight profiles are applied to Japanese strokes. At zero they are even lines; at full, every 止め and 払い is there
Nib size · Pressure	Base width of the mark, and how strongly speed is allowed to change it. At zero pressure the line is a constant width, like a felt tip
Flow	Ink supply. Low enough and the pen skips: the ribbon actually breaks where it ran dry
Bleed	Halo of ink soaking outwards into the paper
Nib angle	Only the italic nib has an edge. This is the angle it is held at, and it decides which diagonal comes out thick
Another writer	A fresh writer: every hand and pen control redealt, within ranges that still read as writing
Write it again	The same writer, the same settings, copying the same text a second time
Save sheet	The sheet at twice the size, rules and slug and all, as a PNG

While a slider is moving the sheet is written at a coarser step — fewer samples along every stroke, no bleed, no pooled stroke-ends — and letting go writes it out properly. Nothing on this page loads a font file, and nothing draws a glyph.