The Cathedral and the Bazaar

Erich S. Reimann

Presented at the First International Congress of Vernacular and Sacred Construction, Bruges, 1997

Revised edition published by O’Reilly & Associates, Sebastopol, California, 1999

I had been building things the cathedral way for most of my career — carefully, slowly, with blueprints approved months in advance and materials ordered from a single supplier — before I understood that there was another way, and that it was better.

The insight came, as insights so often do, not from theory but from a building. In 1996, I acquired a small decommissioned chapel outside Malvern, Pennsylvania, with the intention of converting it into a farmers’ market. The chapel — a modest Carpenter Gothic structure dating from 1891, with a leaking roof and admirable bones — had been abandoned by its congregation for the usual reasons: demographic shift, a newer facility on the highway, the slow gravitational pull of the suburbs. I bought it for less than the price of a good car. What I did with it changed my understanding of how buildings are made.

But to explain what I did, I must first explain what I was reacting against.

The Cathedral Model

The construction of a cathedral is the most complete expression of what we might call the authoritarian school of architecture. A single vision — the master builder’s, or in later periods the architect’s — governs every detail. The craftsmen are skilled, sometimes brilliantly so, but they are anonymous. Their names are not recorded. Their individual judgments are subordinated to the plan. Abbot Suger at Saint-Denis, Villard de Honnecourt with his portfolio of templates, Viollet-le-Duc restoring (or reinventing) the Gothic at Carcassonne — these are men who believed that a great building requires a great mind at the top, issuing specifications downward through a hierarchy of execution.

The results are, one must concede, sometimes magnificent. Chartres is magnificent. But Chartres took sixty-six years to build, and the south tower doesn’t match the north tower, and nobody involved in its completion had met anybody involved in its commencement. The cathedral at Cologne was begun in 1248 and completed in 1880. Six hundred and thirty-two years. One shudders to think of the schedule overruns.

The cathedral model has several defining characteristics. First, the plan precedes the construction entirely. The building is complete in the mind of the architect before the first stone is laid. Modifications are not iterations; they are failures. Second, the craftsmen are interchangeable. A mason is a mason. If one dies — and on a project spanning decades, many will — another takes his place and executes the same specification. Third, the timeline is, by any commercial standard, absurd. The cathedral builder is not building for the market. He is building for eternity, which is a client that never complains but also never pays on time.

The fundamental assumption of the cathedral school is that complexity requires central control. A building as intricate as a Gothic cathedral — with its flying buttresses distributing thrust, its ribbed vaults channelling weight, its clerestory windows calibrated to the angle of seasonal light — cannot emerge from the uncoordinated efforts of independent workers. It must be designed. And designed means: one mind, or at most a small priesthood of minds, holding the complete vision.

This assumption is, I will argue, wrong.

The Bazaar Model

Consider, by contrast, the bazaar.

The Grand Bazaar of Istanbul — the Kapalıçarşı — covers sixty-one streets and contains over four thousand shops. It was not designed. It accumulated. The original structure was a small warehouse built in 1461, shortly after the Ottoman conquest. Stallholders attached themselves to its walls. The stalls became shops. The shops acquired roofs. The roofs connected to each other. Alleys formed between them. The alleys were paved. The paved alleys attracted more shops. Over five centuries, through fires, earthquakes, and renovations that were really demolitions that were really expansions, the Bazaar became the largest covered market in the world.

No architect planned it. No master builder held the complete vision. The Bazaar is the product of thousands of independent decisions made by people whose primary concern was not architectural coherence but commercial viability: where should I put my stall so that customers walking from the cloth district to the spice district will pass my copperware? The answers to ten thousand such questions, accumulated over centuries, produced a structure of extraordinary navigational complexity and functional excellence.

The bazaar model is characterised by the following principles. First, construction and occupation are simultaneous. Stalls open before they are finished. The first revenue is earned while the roof is still being nailed on. I have taken to calling this principle release early, release often — open the stall as soon as it can hold goods, improve it while trading. The feedback from actual commerce is worth more than any amount of planning.

Second, structural decisions are distributed. Each stallholder is responsible for his own load-bearing walls. If a wall is weak, the neighbouring stallholder will notice, because it is also his wall. This leads to a principle I have named after Linus Pauling, the Nobel laureate chemist, who during a brief consultancy on crystalline stress patterns in masonry observed that in a structure with enough independently motivated builders, failures are detected and corrected almost immediately: given enough masons, all cracks are shallow. The reasoning is straightforward. In a cathedral, a crack in an interior column may go unnoticed for years, because the masons responsible for that column finished their work and moved on; the column belongs to the plan, not to any individual. In a bazaar, a crack in a wall is noticed by the man whose livelihood depends on that wall not collapsing onto his merchandise.

Third, the overall plan is emergent rather than prescribed. No one decides where the spice district ends and the textile district begins. The boundary forms through a process of commercial natural selection. Spice merchants cluster near other spice merchants because customers expect to find them there; the clustering reinforces the expectation; the expectation attracts new spice merchants. The structure is the residue of a million transactions.

The Chapel Conversion

My chapel in Malvern was a test of these ideas on a modest scale. The cathedral approach would have been to hire an architect, draw plans, obtain permits for the complete renovation, and open the market when everything was finished. I chose instead to apply bazaar principles.

I opened the market with four vendors — a baker, a vegetable grower, a cheese maker, and a woman who sold cut flowers — before the renovation was half complete. The north transept still had scaffolding. The vestry was being used for storage. The parking area was unpaved. But the market was open, and the vendors were trading, and their revenue funded the next phase of construction. More importantly, their presence revealed requirements that no architect sitting at a drafting table could have anticipated. The cheese maker needed a cooler corner than I had allocated. The flower seller needed to be near the entrance, because customers bought flowers on impulse at arrival, not on reflection at departure. The vegetable grower needed more floor drainage than the original chapel had ever provided, for reasons that anyone who has handled root vegetables in November will understand.

Each of these discoveries would have been a change order in the cathedral model — an expensive modification to an approved plan, requiring the architect’s sign-off and the contractor’s revised estimate. In the bazaar model, they were just Thursday. The cheese maker moved to the north aisle, which was cooler. The flower seller colonised a table near the door. I installed a drain. The building adapted to its occupants rather than requiring its occupants to adapt to the building.

Within eighteen months, the market had twenty-two vendors, a small café in the former vestry, and a waiting list. The renovation had cost roughly a third of what the architect’s original estimate had projected, because two-thirds of the architect’s plan had turned out to be unnecessary — solutions to problems that the actual vendors did not have, finishes that the actual customers did not notice, structural modifications that the actual loads did not require.

I do not wish to overstate the significance of a farmers’ market in suburban Pennsylvania. But I believe the principle it illustrates is general. The cathedral school assumes that the architect’s vision is the irreducible prerequisite of a good building. The bazaar school proposes that the collective intelligence of motivated occupants, expressed through continuous small adjustments, produces buildings that are more functional, more adaptable, and more humane than anything a single mind could design — and at a fraction of the cost.

Objections

The cathedral school will object that bazaar-built structures are ugly. This is sometimes true. The Grand Bazaar of Istanbul is not beautiful in the way that Chartres is beautiful. It is not meant to be. It is meant to be useful, and it is useful in a way that Chartres, for all its magnificence, is not. You cannot buy cinnamon at Chartres.

They will object that bazaar-built structures are unsafe. This is more serious, but the evidence does not support it. The Bazaar has survived multiple earthquakes. It has burned and been rebuilt. It persists. Cathedral construction, meanwhile, has its own structural failures — the collapse of the crossing tower at Beauvais in 1284, the perpetual anxiety about the foundations at Pisa, the calamitous fire at Notre-Dame in 2019 that revealed the cathedral’s medieval timber roof to have been, in structural engineering terms, an enormous bonfire waiting for a spark. The cathedral’s advantage in structural integrity is largely mythological, sustained by survivorship bias: we see the cathedrals that survived and infer that the method was sound, forgetting the dozens that partially collapsed, were never completed, or were so extensively rebuilt over the centuries that the original structure is a philosophical fiction.

They will object that great architecture requires a unified vision. On this point I can only report my experience: the most functional buildings I have entered were not designed by visionaries. They were designed by commerce.

Conclusion

Suger would not have approved of my farmers’ market. Viollet-le-Duc would have found it chaotic. But the cheese is excellent, the flowers are fresh, and the building is paid for. I leave it to the reader to determine which model of construction is, on the whole, better suited to a world where people need places to buy vegetables more than they need places to contemplate the divine — while acknowledging that both needs are real, and that the divine, if it takes an interest in architecture at all, may prefer the building where something is actually happening.

Erich S. Reimann is an architect and architectural critic based in Malvern, Pennsylvania. He is a fellow of the Vernacular Architecture Forum and the author of The New Hacker’s Dictionary of Architectural Terms (MIT Press, 1996). He does not have a degree in architecture, a fact he considers an advantage.