Relational Linking Is a Distinct Organizing Family From LATCH's Attributes

# Relational Linking Is a Distinct Organizing Family From LATCH's Attributes Richard Saul Wurman's LATCH framework claims that information can be organized in only five ways — by **L**ocation, **A**lphabet, **T**ime, **C**ategory, or **H**ierarchy (PDF-02-03). The "only five" is the book's most repeated and most quotable organizing claim, and it is wrong in an instructive way. All five LATCH modes share a hidden common structure: each sorts items by an **intrinsic attribute** of the item — where it is, how it's spelled, when it happened, what kind it is, how big it ranks. They are all answers to the question "what property do I sort on?" But there is a different question entirely — "what is this item *connected to*?" — and the orders it produces are not reducible to any LATCH attribute without emptying the word "Category" of meaning. So the honest correction is not "there's a sixth way" but "there are at least three *families* of organizing, and LATCH is only one of them." The **attribute family** is LATCH: items are placed by a property they carry. The **relational family** is associative linking: items are placed by their edges to other items — the many-to-many, non-hierarchical graph of a Zettelkasten or a wiki, where a single note can sit in arbitrarily many neighborhoods at once and the structure is emergent rather than assigned. The **computed family** is order generated on demand: relevance ranking, full-text search, recommendation, even deliberate randomization — orders that don't exist until you ask for them and that no static filing scheme contains. Wurman could not quite see the relational family because in 1989–2000 the dominant information container was the printed page and the filing cabinet, where one item lives in one place; the graph as a *usable everyday* container arrived with hypertext and matured with tools like Obsidian. This matters because it dissolves a confusion that haunts personal knowledge management. People trained on LATCH-style thinking keep trying to find the *one right folder* for a note, and they experience the failure to find it as a personal organizing deficiency. But the failure is structural: an idea worth keeping usually belongs to several neighborhoods at once, which the attribute family physically cannot express and the relational family expresses natively. This is the same insight that [[The Slip-Box Thinks With You Not For You]] makes from the practice side — the slip-box's value is in the links you draw, not the boxes you sort into. The two notes are the same claim from two directions: that note is about *who does the synthesis work*; this one is about *which organizing family even permits the connection to be recorded*.  ## Why This Matters For **PKM practice**, the implication is permission to stop folder-agonizing. Use the attribute family (LATCH, folders, tags, dates) for *retrieval of known items* — the things you look up by a property you remember. Use the relational family (links, MOCs, backlinks) for *thinking* — the connections that produce new claims. The two are complementary, not competing; the mistake is forcing thinking-work into an attribute container. This is why my own vault "works" in a way LATCH-only filing never did: the value lives in the link graph, and LATCH survives only as a retrieval convenience. For **DevRel and documentation**, the implication is that a docs site organized purely by category (the LATCH default) systematically hides the relational structure developers actually traverse — "this error connects to that config connects to this concept." Good docs need the relational family (see-also links, concept graphs) layered over the attribute family (the nav tree), not the nav tree alone. For **information architecture as a discipline**, the implication is that Wurman's foundational slogan undercounts the design space by two-thirds, and that much of the post-2000 web (search, recommendation, social graphs) lives in the families he didn't name. ## Counterpoints to Hold - **"Category" can be stretched to swallow links.** One can call a link "membership in an ad-hoc category." But this is the elastic-category rescue that makes "only five" unfalsifiable — if every relation is a category, the taxonomy explains nothing. The families are worth distinguishing precisely because collapsing them hides design choices. - **LATCH was never meant as a theory of cognition.** Wurman offered it as a practical designer's checklist, and as a checklist for *displaying* a finite dataset it is genuinely useful. The critique bites only against the universal "only five," not against the tool. - **Relational systems have real costs.** Link graphs can become orphan-note swamps; the attribute family's rigidity is sometimes exactly what a novice needs. "Distinct family" is not "superior family." - **The computed family may be eating the others.** If relevance-ranked search is good enough, explicit organizing of either kind may matter less than Wurman or I assume. The note's framing is pre-AI-search in spirit and should be revisited against IA2-P4. ## Links - [[3. Resources/Permanent Notes/Information Overload Is at Least as Much a Structure Problem as a Volume Problem]] — applies-in: relational links provide the missing structural family that reduces overload when attribute-based filing alone cannot expose meaning. - [[The Slip-Box Thinks With You Not For You]] — the practice-side twin: links are where synthesis happens. - [[Information Capture Without Synthesis Is a Form of Externalization]] — why collecting in any family without linking fails. - [[How to Organize Your Notes in Obsidian the LATCH Method]] — the attribute-family tool, correctly scoped to retrieval. - [[Methods for Organizing Notes in Obsidian]] — the folders-links-tags-metadata inventory this note separates into attribute and relational families. - [[When Answers Are Cheap Understanding Becomes the Scarce Good]] — the computed family under AI abundance. - [[Personal Knowledge Management MOC]] — cluster context. - [[3. Resources/Permanent Notes/You Understand the New Only by Anchoring It to the Known]] — applies-in: relational links preserve the explicit prior-knowledge anchors through which new information becomes understandable. ## Open Questions - Is "computed order" truly a third family, or a dynamic way of producing attribute/relational orders on demand? The taxonomy may need only two families plus a generation mechanism. - What is the right *ratio* of attribute to relational structure in a working vault, and does it shift with vault size? - Can the relational family be taught to LATCH-trained users without them re-importing the one-right-place instinct into their link graph? ## Source - [[4. Archives/Book reviews/Information Anxiety 2]] — §8 (C-009 repair), §12.3 IA2-P1 - [[4. Archives/Book reviews/Information Anxiety 2 — Source Passages]] — PDF-02-03 (LATCH) - Wurman, *Information Anxiety 2* (2000), Ch 2