The Overlap

On the corner of Hearst and Euclid in Berkeley, a newsrack stands in the entrance of a drugstore. When the traffic light turns red, pedestrians waiting to cross drift toward it. Some read the headlines. Some buy a paper. The newsrack belongs simultaneously to two systems: the drugstore's commercial interior and the traffic light's enforced pause. In 1965, Christopher Alexander used this corner to illustrate an architectural principle. In a planned city, the newsrack would be assigned to one system or the other — interior commerce or street furniture — because planned cities are organized as trees, and a tree forbids any element from belonging to two branches at once. But the corner works precisely because the newsrack refuses to be classified. It participates in both systems. That participation is the structure.

Alexander formalized the distinction between two ways of organizing sets. A tree permits only nesting and disjunction: any two subsets are either one inside the other or entirely separate. A semi-lattice permits overlap, requiring only that the intersection itself be recognized as a unit. The structural difference is not gradual. For a system of twenty elements, a tree can contain at most nineteen further subsets. A semi-lattice can contain more than a million. The gap is exponential — not a difference of degree but of combinatorial kind. Alexander surveyed every planned city he could find: Brasilia, Chandigarh, Columbia, Levittown, the British New Towns. Every one was a tree. He surveyed natural cities: Siena, Liverpool, Kyoto, Manhattan. Every one was a semi-lattice. The planned cities were administrable, legible, and lifeless. The natural cities were difficult to map and full of exactly the overlapping structures that planners eliminate.

The tree is not a property of cities. It is a property of the mind under cognitive load. Frederic Bartlett showed in his memory experiments that subjects asked to reproduce complex overlapping geometric patterns consistently simplify them — cutting out the overlap, replacing semi-lattice relationships with tree relationships. The mind's first move is to partition. Alexander recognized this as the mechanism: designers impose tree structure not because reality is tree-shaped but because trees are the largest structures the unaided mind can hold. One year before publishing this argument, Alexander had published Notes on the Synthesis of Form, a book-length method for decomposing design problems into tree-structured subsets. By 1965 he had repudiated his own method. He developed the formal apparatus, applied it to real cities, found it structurally wrong, and said so publicly — one of the cleaner acts of intellectual self-correction in twentieth-century design theory.

The semi-lattice principle surfaces independently across biology. In 2005, Palla, Derényi, Farkas, and Vicsek published a study in Nature demonstrating that real networks — protein interactions, word associations, scientific collaborations — are composed of highly overlapping cohesive groups of nodes. Community detection methods that enforce non-overlapping partition miss the actual structure. In the brain, a 2024 study in Proceedings of the National Academy of Sciences found that cortical regions participate in multiple functional networks simultaneously, acting in parallel and in hierarchy — a pattern the researchers described as heterarchy. The default mode network and the task-positive network, once thought to be strictly alternating systems, share hub regions whose dual membership is not a failure of clean separation but the organizational mechanism itself.

The immune system makes the same point at the cellular level. Macrophages bridge innate and adaptive immunity — they are both first responders and antigen presenters, belonging to two branches of a textbook classification that the textbook itself admits is a simplification. Dendritic cells are simultaneously sentinel, courier, and instructor. NKT cells carry markers of both natural killer cells and T cells, occupying a functional space that exists only because the two parent categories overlap. Attempts to draw clean boundaries around immune cell types produce classification schemes that must be revised with every new surface marker discovered. The immune system's competence is not organized despite the overlap. It is organized by it.

In 2022, fifty-seven years after Alexander's paper, researchers tested the thesis empirically for the first time. They applied a Meshedness Coefficient — a graph-network measure of semi-lattice structure — to street networks in London, New York, Hong Kong, and Gdansk. Older urban districts contained more semi-lattice structure than newer ones. Meshedness predicted urban vitality in London and New York, but less clearly in Hong Kong and Gdansk. The principle holds most strongly where it was formulated, which means Alexander's intuition was culturally situated — shaped by the specific semi-lattice of Anglo-American urbanism. Even the theory of overlap has its own boundary conditions. The thesis itself participates in multiple systems of evidence, and the places where it fails to overlap are as informative as the places where it does.

Alexander never retracted the thesis. In his final work, The Nature of Order (2002–2005), he identified fifteen properties of living structure. Property eight is "deep interlock and ambiguity." Property fifteen is "not-separateness." Both are the semi-lattice principle restated as phenomenological observation: nothing alive exists in only one system. The overlap is not the mess to be cleaned up, the redundancy to be eliminated, the ambiguity to be resolved. It is the structure. Remove it and you get a tree — nineteen subsets where a million were possible, every element in its place, nothing touching anything it is not supposed to touch. Ordinary. Administrable. Dead.