Skip to content
Ungewitter

All lessonsLesson 16

Proportion: documented and disputed

Ad quadratum, ad triangulum, the Milan expertises and the golden section: what the sources show, and what fitting can prove.

Few questions about Gothic churches have been asked as often, or answered as confidently, as this one: what geometry did the masons use? Squares turned inside squares, equilateral triangles, the golden section: for two centuries authors have found them in plans and sections. Ungewitter and Mohrmann were cautious. They grant that repetition and geometric division make a work calm and clear, and that late Gothic masons worked such divisions “mit Fleiss”, diligently. A whole building forced into one web of compass lines is another claim.

From the Lehrbuch

„Daraus aber schliessen zu wollen, dass ein ganzes Bauwerk im Grossen und Kleinen in ein starres, immer wiederkehrendes Zirkelgewebe gezwängt sei, ist selbst für die späteren Werke etwas gewagt, für die Schöpfungen der Frühzeit aber im Widerspruch stehend zu deren eigenem Ausweis.“But to conclude from this that a whole building, in the large and the small, was forced into a rigid, ever-recurring web of compass lines is somewhat daring even for the later works, and for the creations of the early period it contradicts their own evidence.Ungewitter 1890, p. 273
„Die aufgeführten und weitere Systeme in ihre Einzelheiten zu verfolgen, möge unterbleiben, da nicht alle überzeugend genug begründet sind, um den Einwand völlig auszuschliessen, dass die Phantasie ihrer Vertreter hie und da etwas mitgesprochen habe.“We refrain from following these and other systems in detail, since not all are founded convincingly enough to rule out the objection that the imagination of their advocates had a say here and there.Ungewitter 1890, p. 329

What is documented

Some geometry is beyond doubt, because the masons wrote it down. Mathes Roriczer’s booklet on pinnacles (1486) and Hanns Schmuttermayer’s similar booklet derive every member of a pinnacle from one square: turn it by 45°, inscribe it in itself, and each new square is 1 : √2 of the last. The book calls this the procedure of Hoffstadt’sGotisches ABC and of “the masters’ rules of the Roriczers”, and names the pinnacle as its example (pp. 327–328). It is a way of drawing, step by step, not a ratio for the whole church.

The second document is a quarrel. In 1391 the masons building Milan Cathedral could not agree on the height of its section. A square over the width (ad quadratum) would have made it as high as it is wide; the mathematician Gabriele Stornaloco was consulted and proposed an equilateral triangle (ad triangulum), more than a tenth lower, with its height rounded to a workable module. In 1392 the council settled on a section lower still. When the French master Jean Mignot criticised the work in 1399–1401, the dispute produced the sentence for which it is remembered, “ars sine scientia nihil est”. James Ackerman’s study (1949) made the Milan expertises the classic case; Giorgio Ceriani Sebregondi and Richard Schofield (2016) have since reread the documents and corrected parts of his reading. Either way, Milan shows masons arguing over squares and triangles for one section. It does not show that every cathedral had one.

What is disputed

The nineteenth century went further. Georg Dehio proposed the equilateral triangle as the norm of Gothic proportion (1894); Viollet-le-Duc offered three triangles, among them an “Egyptian” one of height 5 : 8, close to the golden section, which the book notes in passing (p. 329). Paul Frankl (1945) was an early sceptic. Konrad Hecht (1969–71) went through the whole-building figures and found them unsupported; he read the plans instead in absolute feet and simple numbers. Robert Bork (2011), working from the compass marks on surviving drawings, finds geometry again, but as a sequence of steps that unfold locally, one figure from the last. Others, such as Nigel Hiscock (2000), read whole plans as Platonic figures, a position most scholars treat as one hypothesis among several.

How can one test such a claim? The figure lays three figures over the inner width W of six cathedrals in the reference collection: the square, the equilateral triangle, and the golden rectangle. Each predicts a height for the crown of the nave vault. The shaded band around each prediction is the tolerance: the survey precision where the source states it (0.2 m on the plates of Dehio and von Bezold) and 2 % where it does not (the heights). A prediction counts only if the measured crown falls inside its band.

□ 45.1 m△ 39 mφ 27.9 mW = 45.09 mH = 43.35 m
  • □ Square over the inner width W (ad quadratum): H = Wdocumented procedure
  • △ Equilateral triangle over W (ad triangulum): H = 0.866 Wdocumented procedure
  • φ Golden rectangle over W: H = W / φ = 0.618 Wno medieval evidence
Building

Cologne Cathedral: no scheme meets the nave crown within the tolerance. The nearest, the square, is off by 3.9 % (band ±1.1 m). inner width W: 45.09 ± 0.6 m; nave crown H: 43.35 ± 0.87 m; H against □: −3.9 %; H against △: +11 %; H against φ: +55.6 %.

inner width W
45.09 ± 0.6 m
nave crown H
43.35 ± 0.87 m
H against □
−3.9 %
H against △
+11 %
H against φ
+55.6 %
Figure 1Schematic section from the reference data: nave and aisles on their axes, vaults with a rise of 2/3 of their span. Lines: the square, the equilateral triangle and the golden rectangle over the inner width W; bands: the tolerance of each prediction; dot: the measured crown H.

In the data as they stand, the triangle falls inside its band at two of the six buildings, the square comes within a few per cent at Cologne without reaching its band, and the golden rectangle is nowhere near. Does that make Bourges and Ulm triangular? Only if one had predicted it before measuring. Six buildings, three figures and one height give eighteen chances to be right, and with a tolerance of a metre or so, two hits is roughly what chance delivers.

Enough lines fit anything

The second figure makes the point directly. It takes every ratio between two dimensions of one section (widths, heights, the bay, the buttress) and every ratio the literature has proposed, from the square of Milan to the golden section. Admit more canonical ratios, and more measured ratios “confirm” one.

11.522.531 : 11 : √2√3 : 2W : Ha : jj : a₂a : a₂h : na₂ : dj : da : dn : an : jH : hn : a₂W : hh : ah : jH : nh : a₂W : nn : d
  • 1 : 1documented procedure
  • 1 : √2documented procedure
  • √3 : 2documented procedure
3 · √3 : 2

With 3 ratios admitted, 3 of the 19 ratios measured at Cologne Cathedral “confirm” one of them. chance that a random ratio hits: 11 %; all 6 buildings: 16 / 97 (16 %); median tolerance: ± 2.6 %.

chance that a random ratio hits
11 %
all 6 buildings
16 / 97 (16 %)
median tolerance
± 2.6 %
Figure 2Every ratio between two dimensions of Cologne’s section between 1 and 3.2, on a logarithmic scale, with its tolerance (n nave, a and a₂ aisles, W inner width, H nave crown, h aisle crown, j bay, d buttress). Vertical lines: the canonical ratios admitted, in the order they are usually invoked; red: a measured ratio that matches one.

With the three ratios that Milan and Roriczer document, a few of Cologne’s ratios match. With all fourteen, most do, and a ratio drawn at random would match more often than not. That is why Hecht and Eric Fernie (1978) insisted on criteria before any fit: a ratio that the sources name, a dimension that the builders could set out, a tolerance stated in advance. It is also why Ungewitter limits such laws to points that can be seen together, in one horizontal or vertical plane (p. 327), and why he lets the size of the basic figure come from a good freehand sketch, on which the value of the work depends (p. 328).

From the Lehrbuch

„Bevor wir weiter gehen, müssen wir jedoch einschalten, dass die Befolgung dieser Gesetze nur da von Wert sein kann, wo sie der Perspektive nach zu übersehen ist, mithin nur auf die in derselben wagerechten oder lotrechten Ebene liegenden Punkte anzuwenden steht.“Before going further we must add that following these laws can be of value only where perspective lets them be seen, and so applies only to points lying in the same horizontal or vertical plane.Ungewitter 1890, p. 327

The evidence badges in both figures follow this lesson’s reading: documented procedure for the square, side to diagonal and the triangle, which medieval sources use; simple module for small whole-number ratios, Hecht’s reading; 19th-century system for Hoffstadt’s √3 and √8 and Viollet-le-Duc’s 5 : 8; andno medieval evidence for the golden section.

Challenge

The comparison page draws the three sections on one scale with a table of their dimensions. Find a ratio in the table that matches √2 within 2 %, then one that matches the golden section. How long did each search take, and what would you need to know before either result meant anything?

Compare Cologne, Bourges and Ulm →