Diagrams, notation and drawing
What makes a diagram a finding and not an illustration?
The literature gives a clear answer to the first question and a mixed one to the second. A diagram becomes a finding rather than an illustration when it can be worked on: Peirce's test, as Stjernfelt reads it, is whether manipulating the diagram shows more than it states. Latour adds the material condition: inscriptions that stay stable, travel and combine, so that unlike things can be compared on one surface. Bender and Marrinan, Larkin and Simon and the 2022 anthropology volume Working with Diagrams give the working detail, and Drucker shows how to draw interpretation and ambiguity without claiming neutrality. The closest precedents for an output made of linked entries are Forensic Architecture, the Research Catalogue's expositions, Latour's AIME platform, Lombardi's drawings and Sousanis's drawn dissertation. They earned legitimacy through outside forums that could contest them (courts, peer review, art institutions), not through the diagram itself. This supports Holagram's rule that no reading is tested by its author. It also threatens it. Jameson's remark that conspiracy theory is a degraded cognitive map describes the failure Holagram is built to avoid; Alexander's tree and semilattice challenge the nested levels; Star and Griesemer suggest boundary objects work by vagueness where Tests need precision; and Sommer's history of diagrams of human kinship suggests how long a contested reading can survive inside a diagram. Work on diagrams as research output in planning itself is thin.
Sources
- Bruno Latour (1986). Visualisation and cognition: drawing things together. Knowledge and Society: Studies in the Sociology of Culture Past and Present, vol. 6correctedmust read
Gives the shortest account of why a drawing can count as evidence: an inscription that stays stable while it travels and can be combined with others, so that many unlike things are compared on one surface. For Holagram this is the standard a Trace must meet (someone else can carry it and open it) and the reason a diagram of linked Entries can yield a finding at all. It also threatens the method: the same combining power builds centres that control what they gather, which is the totalising move Holagram says it resists, and a lived body-level experience is not an immutable mobile.
Audit: Replace the URL with Latour's own page or the Kuklick volume. The essay circulates under two titles: the 1986 Knowledge and Society 6 version ('Visualization and Cognition: Thinking with Eyes and Hands', pp. 1-40, from memory) and the 1990 revision 'Drawing Things Together' (Lynch and Woolgar, eds). The entry's title i
- Frederik Stjernfelt (2019). Dimensions of Peircean diagrammaticality. Semioticacheckedmust readopen access
Peirce's working definition of a diagram is operational: a sign you can manipulate to learn more about its object than the sign states. Stjernfelt separates corollarial from theorematic diagram reasoning, which in Peirce's sense means reading off what is already drawn versus adding something to the diagram and observing. Peirce's Existential Graphs (developed c. 1882 to 1914, with Alpha, Beta and Gamma systems) are a diagram system with explicit rules; Holagram's links have none yet. The 2007 book is the fuller treatment.
- Lukas Engelmann, Caroline Humphrey, Christos Lynteris (eds) (2022). Working with Diagrams. Berghahn Books (Studies in Social Analysis, vol. 14)checkedmust read
The closest recent scholarship to the thread's question. The editors look at what diagrams do in developing theory rather than defining them, and the chapters supply cases: a diagram that obscures what it cannot draw (Bird-David), the limits of visual coherence in anthropological writing (Candea), spatial diagrams as early epidemiological argument (Engelmann), and one diagram travelling between fields (Humphrey). Useful to Holagram for cases of diagrams doing argumentative work and for vocabulary on what a diagram leaves out.
Read first: The introduction, Diagrams beyond Mere Tools, then the chapters by Candea (visual coherence and excess) and Engelmann (spatial diagrams in early epidemiology).
- Johanna Drucker (2011). Humanities Approaches to Graphical Display. Digital Humanities Quarterly 5(1)not re-checkedmust readopen access
Argues that graphical display borrowed from the sciences presents interpretation as given data, and proposes drawing it as taken or constructed (capta), with redesigned charts, elastic timelines and maps distorted by experience. For Holagram it shows how to draw ambiguity and standpoint instead of hiding them, and it challenges the Trace register's claim to record what is there. Lavin's 2021 reply (DHQ 15:2, Why Digital Humanists Should Emphasize Situated Data over Capta) accepts the critique but prefers the term situated data, with the choices behind each datum documented: a sharper form for the source note on a Trace.
Read first: Data as capta: from information visualization to graphical expressions of interpretation; then Space as Spatiality.
- Eyal Weizman (2017). Forensic Architecture: Violence at the Threshold of Detectability. Zone Bookscheckedmust read
The best documented case of a research output assembled from linked traces (smartphone video, site surveys, models, testimony, remote sensing) and made credible in forums that can contest it: courts, parliamentary inquiries, UN meetings, then galleries. Method for Holagram: cross-referencing independent kinds of source, models that place each trace in space and time, and testimony that the model helps recall. It also shows what legitimacy depended on (adversaries with the power to reject the work), which a lone researcher with an internal Adversary lacks.
- John Bender, Michael Marrinan (2010). The Culture of Diagram. Stanford University Pressnot re-checked
A history of the diagram as a way of thinking, from Diderot's Encyclopedie to quantum physics, built on correlation: a diagram invites the viewer to leap across gaps in data. That describes both why a diagram of Entries can show something and why the Reading register is risky, since the leap is the argument and also the apophenia. A book worth owning for the long case that diagrams are arguments, not pictures.
Read first: The chapter titled Diagram (second of six; chapter titles seen in a search result).
- Jill H. Larkin, Herbert A. Simon (1987). Why a Diagram is (Sometimes) Worth Ten Thousand Words. Cognitive Science 11(1), pp. 65-99not re-checked
Compares diagrams and sentences carrying the same information: a diagram is indexed by location, so related items sit together and an inference needs little search. That gives a test for Holagram's layout. The diagram earns its place over the log only if position puts together what an inference needs, for example a Reading beside its Trace and its Test. A force-directed graph of Entries with no meaningful position would not obviously have such an advantage. A 2022 critique by Waszek, seen in a search result and not opened, argues the paper reduces representational differences to computational efficiency.
- Gilles Deleuze (1986). Foucault. Editions de Minuit; English translation by Sean Hand, University of Minnesota Press, 1988not re-checked
The diagram as a map of relations between forces that works inside the social field, with the Panopticon as the usual example. This is the theoretical source for the diagram half of the name Holagram. It also carries a warning for the method: a diagram of this kind is not a picture of a state but part of the machinery that produces it, so a Holagram of a place acts on the place it reads, and the Reading register needs a rule for that.
- Marianne Sommer (2024). Diagrammatics of 'Race': Visualizing Human Relatedness in the History of Physical, Evolutionary, and Genetic Anthropology, ca. 1770-2020. Open Book Publisherschecked
A recent history of diagrams used since the 1770s to show or deny kinship among human groups, including craniometry and the human family tree. For Holagram it bears on the failure log and the rule that no reading is tested at its own level, and its question whether a network models relatedness better than a tree bears on the holarchy.
- Rob Kitchin, Martin Dodge (2007). Rethinking maps. Progress in Human Geography 31(3), pp. 331-344checkedopen access
Argues that a map is not a finished representation but comes into being each time people map, so mapping is a continuing process. For Holagram this supports treating the diagram as versions in a log rather than one finished figure, and it bridges applied mapping and the critical view of maps as power. The threat is the converse: if the diagram is always becoming, it is hard to say what has been found.
- James Corner (1999). The Agency of Mapping: Speculation, Critique and Invention. In Denis Cosgrove (ed.), Mappings, Reaktion Books, chapter 10corrected
A planner-facing argument that mapping is an act that remakes the world it draws, with three operations (field, extraction, plotting), each involving choices. Holagram can use the three operations as headings for the choice log of each diagram version. One critique reported in the same results says the account leaves out the body, which is where Holagram's lowest level starts.
Audit: Keep the citation to Cosgrove (ed.), Mappings (Reaktion, 1999), but replace the URL, and note the 2014 reprint in The Landscape Imagination. Page numbers should be checked against the book.
- Susan Leigh Star, James R. Griesemer (1989). Institutional Ecology, 'Translations' and Boundary Objects: Amateurs and Professionals in Berkeley's Museum of Vertebrate Zoology, 1907-39. Social Studies of Science 19(3), pp. 387-420checked
Names the kind of thing a diagram of linked Entries has to be if planners, residents and machine agents are all to work from it: readable differently by each group, stable enough to stay the same object. It sharpens the design of the diagram, and it names a conflict with the method, because boundary objects work through interpretive flexibility while the Test register tries to remove it. Star's later reflection warns that the concept has been stretched too far.
- Fredric Jameson (1988). Cognitive Mapping. In Cary Nelson and Lawrence Grossberg (eds), Marxism and the Interpretation of Culture, University of Illinois Press, pp. 347-60corrected
Cognitive mapping is the attempt to give the individual a representation of a social totality too large to be lived directly, borrowed from Lynch's study of how people form mental images of a city. It is the nearest existing name for what Holagram wants a diagram to do. The conspiracy remark is the direct threat: Jameson reads conspiracy as a failed attempt to think the same totality, which resembles the self-confirming reading in Sellars that Holagram's safeguards are built against.
Audit: Citation is right; pages are 347-357, not 347-60. Replace the SSRN URL, which points to a commentary.
- kollektiv orangotango+ (eds) (2018). This Is Not an Atlas: A Global Collection of Counter-Cartographies. transcript Verlagnot re-checkedopen access
Forty-plus documented counter-mapping projects, written up as a how-to for people who want to make their own. It gives Holagram practice cases where mapping is done with those mapped and used to contest power, and a model of one unit per project that parallels the Entry. It also exposes a gap in the method: Holagram has one outside reader but gives residents no part in the diagram. The Cartographic Perspectives review criticises the collection's thin coverage of some regions.
- Shannon Mattern (2017). Mapping's Intelligent Agents. Places Journal; reprinted as an open access chapter in Mapping and Politics in the Digital Age (Routledge, 2018)not re-checkedopen access
Asks what ways of seeing machine mapping builds in and what it leaves out, and argues that mapping agents include local and Indigenous communities and other species. This bears directly on Holagram's synthetic hands: what Alter and Adversary bracket out, and whose mapping the diagram counts. Short enough to read in an hour.
- Shannon Mattern (2015). Deep Mapping the Media City. University of Minnesota Press (Forerunners)not re-checked
The passages that matter here: deep maps as intentionally fragile and temporary and not claiming to be objective; balancing GIS precision with the fuzziness of humanistic data; the montage form of Benjamin's Arcades Project; and the deep section as a way to show layers beneath a project. It is a ready model of a layered, non-authoritative map and sits close to Holagram's levels.
- Christopher Alexander (1965). A City is Not a Tree. Architectural Forum 122(1) and 122(2); the author's archive also carries itnot re-checkedopen access
An essay whose argument is carried by diagrams: a tree is a structure whose units never overlap, a semilattice allows overlap, and planned cities are drawn as trees against how cities work. It is the sharpest planning-side attack on nested whole-and-part schemes, so it tests Holagram's five levels, and it is also an example of a diagram doing argumentative work in a planning text. The Harary and Rockey reply (a city is not a semilattice either) shows the formal claim is itself disputed.
- Werner Kunz, Horst W. J. Rittel (1970). Issues as Elements of Information Systems. Working paper, Institute of Urban and Regional Development, University of California, Berkeleychecked
A planning-born notation for a diagram of linked entries in which each argument for or against a position is a node. Rittel is the planner of wicked problems, so this is a precedent from inside the field. The Test register could borrow the issue, position and argument types, and the history of revivals in the 1980s and mid-2000s is evidence on what keeps such a notation in use. IBIS records deliberation, not evidence, so it does not replace the Trace.
- Mark Lombardi; catalogue edited by Robert Hobbs (2003). Mark Lombardi: Global Networks. Independent Curators International (travelling exhibition and catalogue)not re-checked
Close to Holagram in material: public sources, an index-card archive of linked facts, and drawn networks of scandals. It shows a diagram built from Traces with the card archive as its audit trail, and the staged versions as a record of how the drawing changed. An unconventional item that suits the style.
- Nick Sousanis (2015). Unflattening. Harvard University Pressnot re-checked
A precedent for a doctoral argument that is a drawing. Legitimacy came by three routes: the institution accepting the form, a university press publishing it, and prizes from the comics field. It supports the PhD route for a drawn output and shows its limit, since it is one case in an education faculty and its acceptance by readers of comics is not a research test. An unconventional item.
- Society for Artistic Research (n.d.). Journal for Artistic Research and the Research Catalogue: submission and review criteria for expositions. Research Catalogue (researchcatalogue.net)correctedopen access
The established route by which a non-linear, multimedia research output is peer reviewed. The criteria give Holagram a checklist, and the Research Catalogue is a possible home for the diagram plus log as an exposition, with the conventional write-up alongside. For a practice-based doctorate at an art academy this is the closest institutional precedent for the diagram being the research.
Read first: The submission page's description of expositions and the review criteria.
Audit: Cite as: M. Schwab and B. Drabble, 'Submitting to JAR', JAR webinar slides, Research Catalogue, 2025. For a stable source use the journal's own submission guidelines page.
- Georges Didi-Huberman (translated by Shane B. Lillis) (2018). Atlas, or the Anxious Gay Science: How to Carry the World on One's Back?. University of Chicago Press (French original, Minuit, 2011)checked
A long meditation on Warburg's Mnemosyne Atlas, an argument made by arranging images in juxtapositions. It is the canonical precedent for an arrangement that argues, and for associative thinking as method. The Atlas was made by someone who was ill, and Didi-Huberman links knowing to suffering, which bears on Holagram's care rules for the researcher as much as the reader.
- Italo Calvino (1972). Le citta invisibili (Invisible Cities). Einaudinot re-checked
A book of short invented-city entries whose order is a pattern. It supports the Fiction register (about 150 words) and the short condensed-novel cases, and it shows an arrangement fixed in advance as a compositional rule. An unconventional item. The caution is that the pattern is partly the readers' reconstruction.
- Bruno Latour (with the AIME team) (2013). An Inquiry into Modes of Existence (book and AIME web platform). Harvard University Press; platform at modesofexistence.orgchecked
A research output deliberately issued as a linear book plus a navigable platform of linked documents, with readers invited to extend it. That matches Holagram's plan of a diagram of Entries with a conventional write-up produced from the log. It also reports what went wrong: collaborative features arrived late, and Fortun reads the platform's design as reflecting late industrial logic.
- Giada Peterle (2021). Comics as a Research Practice: Drawing Narrative Geographies Beyond the Frame. Routledge (Routledge Research in Culture, Space and Identity)not re-checked
A recent, Italian, geography-side account of making the drawn work the research, with the fieldwork steps described, including walking in urban neighbourhoods. Useful for the fieldwork blocks of the walkable-place study and as a second route to a drawn output beside the diagram. It gives method but, from the page alone, no account of how outsiders test the drawn claims.
Ideas to take
Write the diagram's rules before drawing it.
Changes: Add a short grammar to the concept: what may link to what (Trace to Reading to Test, adjacency of levels), which moves are allowed (merge, split, re-level, mark broken) and what each move is meant to show. Separate what can be read straight off the diagram from what only appears after changing it, for example moving an Entry to another level and seeing which Tests become possible. Place each Reading beside its Trace and its Test so position carries meaning.
Risk: Pseudo-formalism. Peirce's graphs encode logic exactly; Holagram's links are interpretive, so the inferences may be decoration. Try it on a handful of hand-made Entries before adopting.
Hold every Trace to the immutable-mobile standard, and record the choices behind it.
Changes: Give each Trace a card: a source someone can open, date, place, method of capture, what was left out, and whether a human or a machine captured it. A lived body-level experience enters as a described Trace with the conditions of its recording, not as if it could travel unchanged. Lavin's situated data gives the wording: document the decisions behind each datum.
Risk: Makes lived experience conform to what can be carried on paper, and ambiguity may be lost at the point of entry.
Draw ambiguity, hands and failure into the diagram, and keep dated versions.
Changes: Line style for held, broke and unchecked; hands shown on every node and link; the failure log as a visible layer, not an appendix; a snapshot of the diagram at the end of each week, as Lombardi's staged drawings do. Use Corner's three operations (field, extraction, plotting) as headings for the choice log of each version.
Risk: Visual noise, and uncertainty drawn as decoration is another kind of illustration. Lavin's point stands that a new label does not do the work; the log must.
Plan the legitimacy route at the start and write it down.
Changes: Name which outside forums will contest the output: the outside reader, a Research Catalogue exposition for the doctorate, and the conventional write-up made from the log. Keep a one-page memo that lists, for each part of the output, which review criterion it meets (for JAR: shows practice as research beyond documentation; uses the medium to support the argument). Borrow Forensic Architecture's cross-referencing of independent kinds of source and AIME's pairing of a linear text with a navigable platform.
Risk: AIME's collaborative features came late and drew criticism; Forensic Architecture's credibility rests on adversarial forums such as courts that Holagram will not have; Sousanis is a single case.
Add a conspiracy audit to every version of the diagram.
Changes: Report with each version: number of Traces, Readings and Tests; Readings with no Test; Readings checked only by their own hands; Readings whose links outnumber their Traces beyond a set ratio. Show the card archive beside the drawing, as Lombardi's show did. Review old versions for the story they have already fixed, which is the lesson Sommer's history seems to carry.
Risk: The ratio is arbitrary and can be gamed, and the audit cannot detect a self-confirming frame that decides in advance what counts as a Trace. It supplements the outside reader and does not replace her.
Treat the diagram as a boundary object with two views.
Changes: A Trace-only view for outside readers and people who use the place, and a full view with Readings and Fiction for the researcher. Alter and Adversary work through the same diagram, and each version records what the synthetic agents' tools leave out. Consider letting people from the place annotate Traces, in the counter-mapping manner.
Risk: Boundary objects work through interpretive flexibility and Tests need the opposite. Residents' annotation raises consent and care questions beyond a three-week study from public material.
Test the five levels against the tree and the semilattice.
Changes: Let an Entry link across levels and sit at more than one. After the study, count cross-level links against nested ones and report that count as a finding on whether the holarchy holds in this place. Consider IBIS-style node types (issue, position, argument) for Tests.
Risk: If nearly everything overlaps, the five levels turn out to be an imposed tree, which may be the useful result but changes the concept's frame. Harary and Rockey's reply shows the formal claim is itself disputed.
Fix a combinatorial plan for the short cases and the fieldwork blocks.
Changes: Show Traces beside Fiction in Warburg-style panels, and write a Peterle-style protocol for the walking blocks. Empty cells are then visible and count as findings.
Risk: A fixed grid can force Entries into cells that do not fit. Warburg made the Atlas while ill, so the care rules apply to the researcher too.
Counter-arguments
- Latour's inscriptions gain power by being gathered in one place; a diagram of Entries becomes a centre of calculation that decides what counts, and the body level is not an immutable mobile. Holagram would need to show what it leaves out (diagram-01).
- Peirce's criterion is strict: a diagram is a finding only if there are rules for working on it. Without them a diagram of linked Entries is a mind map or an illustration, however carefully it is made (diagram-02, diagram-07).
- Jameson reads conspiracy theory as a failed cognitive map. A method that licenses strong, even paranoid, Readings and links them into a network is open to that description unless the Tests are independent in fact, not only in procedure (diagram-13, diagram-19).
- Bender and Marrinan describe the diagram's persuasion as correlation that invites leaps across gaps; Sommer's history, as described, follows diagrams of human relatedness over two and a half centuries and looks like a case of such leaps surviving inside respected diagrams (to be checked). A tidy diagram can launder a weak Reading (diagram-06, diagram-09).
- Alexander would say nested levels are a tree, and that a city's structure lies in the overlaps; a five-level holarchy may impose the tidiness it claims to resist (diagram-17).
- Drucker would say the Trace is taken, not given. The register called what is there starts from a claim of innocence that her argument, and Lavin's reply, say must be replaced by a record of choices (diagram-04).
- Star and Griesemer: objects that travel between groups work by vagueness. The Test register asks for the opposite, so the diagram may serve one purpose well and not both (diagram-12).
- Counter-mapping and critical cartography would ask who the diagram is for and who draws it. One outside reader and a synthetic Adversary are not the people whose place is mapped (diagram-14, diagram-15).
- Forensic Architecture's credibility rests on adversaries who can reject it (courts, states, peers). A lone researcher with an internal Adversary does not have that, and the debate over proof versus art in that work is not settled (diagram-05).
- The art-research route (JAR, Sousanis, Lombardi) may be judged on its form and its argument about knowledge rather than on whether readings were checked. A Holagram presented that way may be judged as art and its Tests overlooked (diagram-21, diagram-20, diagram-19).
Gaps
- What exists: Mattern, Deep Mapping the Media City (listed as diagram-16); Kenneth Field, Thematic Mapping (the modifiable areal unit problem: where boundaries are drawn controls the pattern, a ready example for a Test); Orit Halpern, Beautiful Data (history of data display and objectivity); Tonkiss, Cities by Design (cites Latour and Hermant's Paris ville invisible, 1998); Jack Self, The Ownership Loop (an article built on a diagram); a note on Kevin Lynch in Notes Archive/phd notes; Eisenman mentions in Attunement and Architecture Words 1.
- Search budget.
- Architecture's diagram debate (Eisenman, Stan Allen, Somol and Whiting, van Berkel and Bos) is covered only at bibliographic level. Those are the first things to check.
- Peirce's own texts were not opened. Wikipedia's page lists the primary sources (the 1906 Prolegomena to an Apology for Pragmaticism in The Monist, the 1903 Lowell Lectures, the 1908 Some Amazing Mazes) and points to the Prolegomena for the claim that reasoning proceeds by experiment on a diagram. The Stjernfelt PDF returned only metadata.
- The nearest are Alexander, Rittel's IBIS, Geddes, and the art-research and forensic cases, so the legitimacy route is borrowed from outside the field.
- Finnish and Estonian items: none verified. Hannula, Suoranta and Vaden, Artistic Research Methodology (Peter Lang, 2014) returned nothing in the catalogue and stays recalled only. Italian items: Calvino, Peterle (Padua), and a lead to Farinelli listed under cross_leads.