Holons, hierarchy and scale
Is a holarchy a defensible structure for research, given the critiques of hierarchy and scale?
The literature does not support a holarchy as a description of a city, and it does not rule one out as a device for reading. Simon gives the strongest defence: nested systems are common because stable sub-assemblies are stable, and a level can be read on its own when links within it are much stronger than links across it. Salthe turns that into a working rule: a focal level is read against the level above and the level below. Against this, Alexander shows that cities are made of overlapping sets, and a tree of nested wholes cuts the overlaps; Potochnik and McGill argue that discrete levels rarely exist; Marston, Jones and Woodward drop vertical scale for a flat ontology; Tsing rejects precision-nested scales because they work by erasing difference; Latour's digital test treats the whole as narrower than its parts. The rule 'test a reading at another level' survives if three things hold: levels are defined by the kind of evidence admitted, not by size; nothing is aggregated upward; and the diagram carries overlaps as well as nests. It breaks where one document, one infrastructure or one model acts on every level at once, because the second level is then not independent. Holagram can keep the holon as a vantage, flat in being and nested in analysis, and treat near-decomposability as a hypothesis the first study is allowed to fail.
Sources
- Arthur Koestler (1967). The Ghost in the Machine. Hutchinson (UK); Macmillan (US)checked
The source term that Holagram builds on. A Koestler holon is a whole to its parts and a part of a larger whole, with a self-assertive and an integrative tendency, so no level is only a container or only a component. That supports Holagram's claim that each level carries its own evidence.
- Herbert A. Simon (1962). The Architecture of Complexity. Proceedings of the American Philosophical Society, vol. 106, pp. 467-482 (reprinted in The Sciences of the Artificial)checkedmust read
The strongest defence of hierarchy, and the source of the condition Holagram's rule depends on. Simon argues that complex systems are often hierarchic because stable intermediate forms evolve faster (the watchmaker parable), and that they are nearly decomposable: interactions within a subsystem are much stronger than interactions between subsystems. Only where that holds can a level be read on its own, so near-decomposability states when 'test a reading at another level' gives a check that is independent. For cities it is an empirical claim, not a given; Holagram should treat it as a hypothesis the first study can fail.
- Stanley N. Salthe (1985). Evolving Hierarchical Systems: Their Structure and Representation. Columbia University Pressnot re-checked
The closest thing in hierarchy theory to a working form of Holagram's rule. The basic triadic system says that to understand anything at a focal level you need the level above and the level below, and that the choice of focal level is a research decision. Read as a Test protocol: one check from below, one from above. It also treats levels as observer-relative, which fits the Reading register. Allen and Starr, Hierarchy: Perspectives for Ecological Complexity (Chicago, 1982; 2nd ed. 2017, seen in a catalogue record) is the neighbouring text if Salthe is too dense.
Read first: The chapter 'Representing a Dynamic System Hierarchically: The Basic Triadic System'.
- Lance H. Gunderson and C. S. Holling (eds) (2002). Panarchy: Understanding Transformations in Human and Natural Systems. Island Pressnot re-checked
Nested systems in which each level runs its own cycle at its own speed and the levels are coupled in both directions. For Holagram this offers typed cross-level links in place of a plain ladder: a slower level constrains a faster one, and a faster one can trigger change in a slower one. It comes from ecology, and its urban uses are mostly descriptive (see holarchy-05).
- Craig R. Allen, David G. Angeler, Ahjond S. Garmestani, Lance H. Gunderson, C. S. Holling (2014). Panarchy: Theory and Application. Ecosystems 17(4): 578-589checked
The originators audit their own concept. They report that its use in the literature is mostly descriptive and abstract, discuss limits on where it applies, and propose testable hypotheses. This is a model for how Holagram should handle its own holarchy: as a set of hypotheses the study exposes to failure, not as a description of the city.
- Christopher Alexander (1965). A City Is Not a Tree. Architectural Forum 122(1), April 1965, and 122(2), May 1965not re-checkedmust read
The sharpest attack on a nested structure of the city. A tree is a collection of sets in which any two are nested or disjoint; a semilattice allows overlap. Alexander argues that planned cities are trees and that real urban life depends on the overlaps the tree removes. Holagram has to say whether its levels are disjoint nests or overlapping sets. The essay is also a method model: the argument is carried by diagrams. Mattern (holarchy-18) carries the argument into algorithmic urbanism.
Read first: Both parts; each is about five pages. The definition of tree and semilattice comes early; the diagrams of planned cities follow.
- F. Harary and J. Rockey (1976). A City Is Not a Semilattice Either. Environment and Planning A 8(4): 375-384checked
The reply that closes the easy exit. The authors revisit Alexander's graph-theoretic argument, refute some of it, and conclude that a city is neither a tree nor a semilattice; they suggest that negative descriptions may serve better than positive ones. For Holagram: swapping a tree for a semilattice does not make the diagram true to the city. Any diagram is a reading device, which matches the concept's position that spatial facts resist totalising description.
- Sallie A. Marston, John Paul Jones III, Keith Woodward (2005). Human Geography without Scale. Transactions of the Institute of British Geographers 30(4): 416-432checkedmust read
The central attack on hierarchical scale in geography. The authors argue that the vertical, nested view of scale cannot be repaired by adding networks, and propose dropping scale for a flat, site-based ontology; the paper closes on political implications. Holagram's body-to-territory ladder is the target. The defensible reply is to use levels as positions for testing, not as layers of reality, and to say so. Read 'Situating Flatness' and the Prytherch reply to see how the authors answer critics.
- Anna Lowenhaupt Tsing (2012). On Nonscalability: The Living World Is Not Amenable to Precision-Nested Scales. Common Knowledge 18(3): 505-524 (reprinted 2019, vol. 25)checkedmust read
Names the exact target: precision-nested scales. Scalability, as she uses it, is the ability of a project to grow without changing its framework, and it is achieved by erasing local difference (examples: sugarcane clones, standardised labour units, pixels). A holarchy of body, street, site, institution and territory looks like a precision-nested scale. Holagram's answer has to be structural: entries must not aggregate or generalise upward, and a reading at one level is retested elsewhere, never treated as a sample of it.
- Bruno Latour, Pablo Jensen, Tommaso Venturini, Sebastian Grauwin, Dominique Boullier (2012). 'The whole is always smaller than its parts': a digital test of Gabriel Tarde's monads. British Journal of Sociology 63(4): 590-615checkedopen access
A test of the idea that an individual is richer than any aggregate that describes it, done by moving between individual traces and aggregated profiles. For Holagram this reframes the higher levels: territory and institution are vantages assembled from selected Traces, not containers that hold the street. An entry at those levels should state which Traces it was built from. The paper is also a worked example of a digital test, and it is open access.
- Angela Potochnik and Brian McGill (2012). The Limitations of Hierarchical Organization. Philosophy of Science 79(1): 120-140checkedmust readopen access
The most direct critique of levels as such. Using mainly ecological examples, the authors argue that discrete hierarchical levels face many difficulties, and propose scale, with causal processes studied at different scales, as a way to keep quasi-levels. For Holagram: the five levels cannot be assumed to be discrete layers, and a Test should say which causal process at which scale it checks. The 2021 follow-up pushes further and is worth reading second.
- William C. Wimsatt (1994). The Ontology of Complex Systems: Levels of Organization, Perspectives, and Causal Thickets. Canadian Journal of Philosophyrecalled, unverified
From memory: Wimsatt treats levels as local peaks of regularity and predictability, with messier causal thickets in between, and treats them as perspectives. If right, it gives a way to decide which of Holagram's five levels earn the name and where the gaps are.
- David W. Cash, W. Neil Adger, Fikret Berkes, Po Garden, Louis Lebel, Per Olsson, Lowell Pritchard, Oran Young (2006). Scale and Cross-Scale Dynamics: Governance and Information in a Multilevel World. Ecology and Society 11(2): art. 8checkedopen access
Vocabulary for the level question. Holagram's five levels mix size (body, street, site, territory) and institutional type (institution), so they are not positions on one scale; this paper helps name that. The scale/level definitions are recalled, not seen.
- Liesbet Hooghe and Gary Marks (2003). Unraveling the Central State, but How? Types of Multi-Level Governance. American Political Science Review 97(2): 233-243 (IHS Political Science Series 87)checkedopen access
The governance version of Alexander's argument. For the first study, a public planning document sits in both kinds of structure, so the institution level cannot be a clean nest. Record the overlaps as entries.
- Carole L. Crumley (1995). Heterarchy and the Analysis of Complex Societies. Archeological Papers of the American Anthropological Association 6(1): 1-5not re-checked
The standard counter-term to hierarchy: relations among elements that are unranked or can be ranked in more than one way. Crumley brought it into archaeology to show that hierarchy and heterarchy coexist in societies. For Holagram: within a level, entries should be linked without rank, with the holarchy kept for the ordering of levels. McCulloch (1945) is the origin of the term in neuroscience.
- Manuel DeLanda (2016). Assemblage Theory. Edinburgh University Press (Speculative Realism series)not re-checked
The nearest philosophical cousin of the holon without Koestler's ladder: wholes are assemblages whose components keep their autonomy, and nothing is privileged ontologically, yet analysis can still move between smaller and larger assemblages. That is the position Holagram can take: flat in being, nested in analysis. Whether flat-ontology authors accept this is not checked (see holarchy-08).
- Benjamin H. Bratton (2015). The Stack: On Software and Sovereignty. MIT Pressnot re-checked
Planetary-scale computation as a six-layer stack in which the City is one layer. That threatens the independence of a Test run at 'another level' if the same model sits under both. It is also a totalising diagram, which Holagram wants to resist. The reading of the layers as function-based is mine, not checked against the text.
- Shannon Mattern (2021). A City Is Not a Computer: Other Urban Intelligences. Princeton University Press (Places Books)checked
Carries Alexander's tree argument into the city-as-computer metaphor, which she says reduces place-based knowledge to information processing. Holagram plans dialogue with computational agents and puts lived experience at the centre, so this is the closest published argument against the risk in the concept.
Read first: The Introduction ('Cities, Trees & Algorithms') and chapter 2.
- Alex Hanna and Tina M. Park (2020). Against Scale: Provocations and Resistances to Scale Thinking. arXiv:2010.08850 (workshop paper, CSCW '20, 'Reconsidering Scale and Scaling in CSCW Research')not re-checkedopen access
Scale thinking in computing, seen from outside the technology: the drive for efficient growth shapes which problems look worth solving. The authors draw on mutual aid networks and offer questions for evaluating collaborative systems and finding resistance to scale. Holagram's synthetic agents can produce Fiction and Tests in bulk, which is scale thinking arriving through the back door. Use their questions as a checklist for the first study, and cap what the agents produce per entry.
- Ole B. Jensen (2021). Rethinking Scale: Relationality, Place, and Critical Zone. Nordes 2021: Matters of Scale (Kolding), Full Paperscheckedopen access
A short, recent, open paper arguing that scale is not neutral and should be read through fluidity, mobility, networks and continuums, with the Critical Zone as the planetary reference. A quick way into the recent scale debate from a design-research direction, and the proceedings, themed 'Matters of Scale', are a place to look for more.
- Stewart Brand (1994). How Buildings Learn: What Happens After They're Built (pace layers, developed in The Clock of the Long Now, 1999). Viking (1994); Penguin (1995)not re-checked
Practitioner writing that orders layers by rate of change, not size. It gives Holagram a second axis beside body-to-territory: each Trace has a pace, and a Reading about a slow layer built on fast Traces is weak. Wikipedia says the model is rooted in hierarchical ecosystem theory, so it sits close to panarchy. Caution: it is an ordered list that invites a ladder reading, and it is a popular schema, not tested evidence.
- Georges Perec (trans. John Sturrock) (1974). Species of Spaces and Other Pieces (Espèces d'espaces, 1974). Galilée (French original); Penguin Twentieth-Century Classics (English, 1997; one listing says 1998)not re-checked
An unconventional model of nested spaces: a literary holarchy that moves from the page to the bed, the apartment building, the street and out to space, attending to what is done in each place instead of treating it as a container. Useful as a format for the Trace register (inventories, descriptions of the ordinary) and for the Fiction register, and as a reminder that levels can be written from the inside. It is a literary work, not a method.
- Anna Lowenhaupt Tsing, Jennifer Deger, Alder Keleman Saxena, Feifei Zhou (eds) (2020). Feral Atlas: The More-Than-Human Anthropocene. Stanford University Press (digital project)checkedopen access
Nonscalability in practice. A digital atlas of field reports, essays, videos and artworks, assembled by observation in place and organised by what it documents, not by nesting. It is a working precedent for a diagram of linked entries that mixes registers and does not total. Look at how entries are cross-linked and how authorship is shown.
- Brian J. Robertson (2015). Holacracy: The New Management System for a Rapidly Changing World. Henry Holtnot re-checked
The best-known deployed holarchy, in management. It shows what the word does when it reaches organisations: roles in nested circles, a rulebook, and the objection that hierarchy remains. Holagram's holarchy is a reading device, not an organising scheme, and readers who know Holacracy or Wilber's integral theory will expect otherwise. State the difference early.
- Jianmei Yang (2025). Near decomposability of complex networks based on Simon's theory of complexity architecture: conceptions, approaches and an application. arXiv:2501.12748 (preprint)checkedopen access
A recent attempt to make Simon's near-decomposability measurable in networks: it separates a near-decomposable hierarchical structure from near-decomposability in behaviour and properties, and proposes a way to find the first in real networks by splitting adjacency matrices. For Holagram it suggests a check on the finished diagram: are links denser within levels than across them?
Ideas to take
Define each level by the kind of Trace it admits (body: diary, sensing, count; street: photograph, observation, count; site: plan, permit, record; institution: document; territory: map, statistic), and let each Test name which scale of causal process it checks.
Changes: Levels stop being a claim about the city and become a rule about evidence. That answers the flat-ontology and anti-levels critiques and keeps 'test at another level' usable.
Risk: The levels may then reflect only the one document and one place chosen. 'Institution' still mixes type and size. Wimsatt-style local peaks of regularity may not show up in a three-week study.
Make the Test a triad: for a Reading at a focal level, one check from the level below and one from the level above. Allow a declared skip-level Test where a document or an infrastructure acts directly across levels.
Changes: 'Another level' becomes two named checks with different evidence, and the cases where independence fails are written down instead of ignored.
Risk: Doubles the checking in a three-week study. The two checks can share one source (a single planning document speaks for institution and territory), so independence is only formal.
Run Simon's check on Holagram's own log: after the study, read the diagram of linked entries as a network and compare links within a level to links across levels. A diagram where cross-level links dominate is a failed holarchy and goes in the failure log.
Changes: Adds a 'level audit' entry to the failure log and makes the holarchy falsifiable, which is what the panarchy authors ask of their own theory.
Risk: A few dozen entries from one place carry no statistical weight, so the result is descriptive.
Type the links in the diagram (part-of, overlaps, tests, contradicts). Let an entry sit under two parents and record overlaps as entries of their own; within a level use unranked links. In the first study, place the planning document among general-purpose nested jurisdictions and task-specific overlapping ones.
Changes: The holarchy becomes an ordering of levels, not a claim that the city is a tree. It answers Alexander directly.
Risk: Harary and Rockey show that a semilattice is also a formal claim the city does not meet, so the typed diagram must not be presented as a model of the city. More edge types can make the diagram unreadable to the outside reader.
Adopt a no-aggregation rule: a Reading at a higher level is never a summary of entries below, but a new Reading from that level's own Traces, and it states its vantage and its Traces. Cap synthetic output: Alter and Adversary work one entry at a time, with no bulk generation.
Changes: Gives Holagram its own answer to nonscalability: entries do not scale. It also keeps the conventional write-up from bringing aggregation in unnoticed.
Risk: The territory level may stay thin. The final write-up has to generalise somewhere, so the rule will be broken at the last stage; say where.
Give every Trace a pace (fast: signage, behaviour, posts; slow: statute, structure, plan). Treat a Reading about a slow layer built only on fast Traces as weak. Name cross-level Tests as 'remember' (slow constrains fast) or 'revolt' (fast pushes slow).
Changes: A second axis beside size, which suits public documents and walkable places because they change at different rates.
Risk: Pace layers is a popular schema and panarchy is thinly tested. The ordered list invites a ladder reading.
Borrow forms: write the Trace register as an inventory of what is done at each level (sleep, cross, apply, file), after Perec, and give entries the look of field reports, after Feral Atlas. Keep Fiction labelled.
Changes: Concrete format models for the entry and for the condensed-novel short case, and a precedent for presenting without totalising.
Risk: Literary borrowing can hide weak evidence. Perec's book is literature, not a method; the Atlas was a large team project.
State in the first paragraph how 'holarchy' is used: nested self-regulating wholes with partial autonomy in Koestler's sense, flat in being and nested in analysis, and neither a management scheme nor a spiritual ladder.
Changes: One sentence that heads off the Holacracy and Wilber readings and the flat-ontology objection.
Risk: DeLanda's position is contested by flat-ontology authors (not checked here), and the sentence can read as evasion.
Counter-arguments
- Tsing (holarchy-09): a nested ladder of body, street, site, institution and territory is a precision-nested scale, and such scales work by erasing the difference they cannot hold. If an entry at street level is allowed to stand for a larger level, Holagram becomes a scalable pipeline.
- Marston, Jones and Woodward (holarchy-08) and Jensen (holarchy-20): scale is not given and is not neutral. A vertical ladder naturalises what is a practice and a politics, and a holarchy of levels will be read as a claim about the city unless it is declared a device.
- Alexander (holarchy-06) and Hooghe and Marks (holarchy-14): the first study joins a planning document (institution) and a walkable place (site). They do not nest. A planning document is not a bigger version of the place, and overlapping jurisdictions will cut across any tidy level.
- Potochnik and McGill (holarchy-11) and Wimsatt (holarchy-12, recalled): there may be no discrete levels, so the independence that makes a test at another level meaningful is not supplied by the world. Causes cross levels, and evidence from 'the other level' can share the cause of the reading.
- If cities are not nearly decomposable, because infrastructures, laws and software link all levels, then the rule has no ground.
- Latour and colleagues (holarchy-10): a higher-level whole holds less than its parts. A territory-level Trace (a map, a statistic) is thin, and a Test at that level may test only the aggregation.
- Harary and Rockey (holarchy-07): replacing the tree with a semilattice is not enough. Any formal structure is itself a totalising claim, and Holagram's diagram can fall into the fault it criticises.
- Bratton (holarchy-17), Mattern (holarchy-18), Hanna and Park (holarchy-19): synthetic agents sit across layers and run on scale logic. Dialogue with computational agents can reproduce the city-as-computer view, and the levels may end up supplied by the model's ontology, which is the self-confirming reading the method is built to prevent.
- Robertson (holarchy-24) and the Holacracy critics: in practice 'holarchy' has meant a management system that keeps hierarchy under another name. Readers may take Holagram's term to promise an organising scheme.
Gaps
- They are the next reading in the scale and planetary-urbanisation debate but are not listed.
- Recent work (2022 to 2026) is thin in this list: Hanna and Park 2020, Jensen 2021, Mattern 2021, Feral Atlas 2020, Yang 2025.
- No Italian, Finnish or Estonian item verified.
- Not found: literature on testing a reading across levels as a qualitative method (multi-level triangulation); the tangled-hierarchy and logical-types literature (Hofstadter, Bateson) that bears on the 'no reading tested at its own level' rule is recalled only and left for another thread.