Walk into any public or academic library, and you’ll see rows of spines arranged with quiet precision. Each book has been assigned a number, a letter, or a string of both, its address in the collection. That address is not neutral. It is the outcome of a long negotiation between logic and legacy, between what fits neatly into categories and what resists them. Library classification systems are not abstract taxonomies; they are spatial solutions to a physical problem: how to fit thousands of distinct objects onto linear shelves, each with fixed width and height, in a way that supports discovery without requiring a map for every visit. The decision is never just about subject, it is about priority, proximity, and precedent.
This constraint shapes everything. A history of botany might sit beside works on soil chemistry under one system, but next to treatises on Renaissance art under another, not because the topics are equivalent, but because the system’s architecture forces adjacency where none was intended. Classification does not describe knowledge so much as it constructs pathways through it. And when those pathways narrow, curve unexpectedly, or end abruptly, the gaps tell their own story.
The shelf as a line of thought
Every classification scheme begins with an assumption about how knowledge unfolds. The Dewey Decimal Classification (DDC), first published in 1876, treats knowledge as a branching tree rooted in ten broad classes, from philosophy at 100 to technology at 600. Its decimal subdivisions allow fine-grained placement: 581.34 is for ferns, 581.35 for horsetails, 581.36 for clubmosses. This structure suits general collections well. In a small-town library, a reader browsing 581 will find field guides, gardening manuals, and botanical monographs all within arm’s reach. But DDC’s hierarchy also compresses disciplines that evolved separately. Linguistics appears at 410, tucked beneath ‘language’, a category that includes grammar handbooks and phrasebooks, but sits far from anthropology (306) or cognitive science (153), even though all three fields routinely draw on one another’s methods and findings. The shelf becomes a corridor with locked doors: adjacent rooms share walls, but no door connects them.
Dewey vs Library of Congress: different logics, different adjacencies
The Library of Congress Classification (LCC), developed in the early twentieth century for the US national collection, takes a different approach. Rather than a single tree, LCC uses twenty-one alphanumeric schedules, each built around a discipline’s internal structure. History (Class D) splits geographically: D covers Europe, DA Britain, DC France, DD Germany. Within DA, works on Tudor England appear before those on Victorian Britain, not by chronology, but by the order in which the Library of Congress acquired its earliest materials. This makes LCC highly adaptable for research libraries, where users often seek deep, focused runs: a scholar of eighteenth-century French literature can locate primary texts (PQ), literary criticism (PN), and contemporary political pamphlets (DC) in adjacent sections, even if the call numbers differ in their first letters. But that adaptability comes at a cost. LCC’s lack of a unified numeric spine means that related subjects, say, ethics (BJ) and political theory (JC), are housed in separate wings, sometimes on different floors. The physical distance mirrors a disciplinary divide that may no longer hold in practice.
What breaks when a subject does not fit
No scheme handles interdisciplinary work gracefully. Consider climate migration: a topic drawing on geography, law, demography, and environmental science. Under DDC, it might land in 304.2 (human ecology), 344.04 (environmental law), or 363.7 (environmental health), depending on the author’s emphasis. In LCC, it could appear in GF (geography of human systems), KF (US law), or TD (environmental technology), again shaped by the publisher’s framing. Neither system has a dedicated class for the phenomenon itself, because neither was designed to accommodate phenomena that emerge at intersections. Instead, the book is assigned to the most ‘dominant’ discipline, and the rest of its relevance is relegated to subject headings in the catalogue: invisible on the shelf, recoverable only through search. This mismatch produces real friction. A student looking for policy responses to sea-level rise may scan the 363s and miss key legal analyses shelved in KF. A researcher tracking demographic shifts might walk past relevant sociological studies filed under HM because the book’s cover highlights its GIS methodology instead.
When the shelf fails, we add layers
Libraries respond to these fractures not by abandoning classification, but by adding interpretive layers. Subject headings, controlled vocabulary terms like ‘Refugees, Government policy, United States’ or ‘Coral reefs, Conservation’, act as conceptual bridges across physical distances. Call numbers get supplemented with coloured dots, shelf tags, or QR codes linking to digital reading lists. Some academic libraries now group high-use interdisciplinary titles in ‘floating collections’, shelved by theme rather than class number, climate change, decolonisation, urban futures, regardless of their official call number. These are not replacements for classification, but acknowledgements that the shelf alone cannot hold complexity. They also reveal something subtle: classification is rarely used in isolation. Patrons consult the online catalogue first, then follow a path to the shelf. The call number is a waypoint, not the destination. Its job is less to explain than to locate, and to do so reliably, even when the logic behind it feels arbitrary.
Other schemes, other priorities
Not all libraries use DDC or LCC. The Universal Decimal Classification (UDC), widely adopted in Europe and parts of Asia, extends Dewey with special symbols (+ for combination: for relation, / for range) to express compound subjects more flexibly. A UDC number like 330.42:551.521 represents ‘economic impacts of climate change’, explicitly linking economics and meteorology. But this expressiveness demands training: few patrons read UDC strings intuitively, and the notation doesn’t translate to shelf layout as cleanly as Dewey’s decimals. Then there’s the Bliss Bibliographic Classification, developed in the UK in the 1930s, which organises knowledge by function rather than discipline, grouping ‘sources of information’ (bibliographies, archives, databases) together regardless of subject. It never achieved wide adoption, partly because its logic resists visual scanning: a patron cannot glance at a row of spines and infer scope the way they can with Dewey’s 500s or LCC’s QK-QR ranges. Its failure wasn’t intellectual, it was ergonomic.
Why some books stay unclassified
A small but telling category exists outside all major schemes: local history collections, rare manuscripts, and archival boxes. These are often assigned accession numbers rather than classification numbers, shelved in closed stacks or climate-controlled vaults. Their exclusion isn’t oversight, it’s recognition that their value lies in provenance and uniqueness, not comparability. To force a seventeenth-century land deed into the 333s (economics) or 974.7 (regional history) would flatten its evidentiary role. Here, classification gives way to custody: the priority shifts from arrangement for access to preservation for integrity. The book no longer lives on the open shelf; it lives in a record, a box, a controlled temperature.
None of this is accidental. Every time a librarian assigns a call number, they enact a choice: to privilege disciplinary lineage over thematic resonance, chronological sequence over conceptual kinship, or current usage over historical context. These choices accumulate over decades, becoming visible not in manifestos but in the wear on certain shelf edges, the frequency of certain interlibrary loan requests, the annotations pencilled beside call numbers in staff manuals. A library’s classification is its silent curriculum, a set of assumptions about what belongs together, what deserves prominence, and what can be deferred to a footnote. It is easier to change a catalogue record than to reshelve ten thousand volumes. So systems persist, not because they are perfect, but because they are legible, durable, and deeply embedded in the architecture of retrieval. The shelf remains a line of thought, not because it traces a single argument, but because every gap, every adjacency, every forced compromise tells us something about how knowledge is held, how it is used, and how it is, inevitably, incomplete.