Objects
The card index and the search that had a physical shape
Before a search box, finding something meant knowing how a drawer had been organised. That knowledge was a skill, and it was widely held.
Treat the sections below as a sequence. With the card index, getting the early decisions right makes the later ones much easier.
Before you start
- A card index forced a single ordering decision that governed every later lookup.
- Cross-references were manual, so each one cost a card and a hand to write it.
- Full-text search removed the need to guess how somebody else had classified a thing.
The drawer was the database
A card index stored one record per card, sorted into a physical order that determined how quickly anything could be retrieved. Choosing that order was the entire design task, because a set of cards sorted by surname cannot also be sorted by subject.
Anything you wanted to find by a second route needed a second card, filed elsewhere, pointing back at the first. The size of an index therefore grew faster than the number of things in it, which is a real and underappreciated cost. Every organisation running one had somebody whose judgement about categories quietly governed what the rest of the staff could find.
Standard cards created an accidental network
Cards were made to common sizes, which meant cabinets, guides and trays from different suppliers could be mixed without any coordination. That standardisation let institutions expand an index by buying more drawers rather than by redesigning the system they already had.
It also allowed printed catalogue cards to be produced centrally and distributed, so many libraries filed identical descriptions of the same book. Shared cataloguing rules were an early and largely successful attempt at making separate collections behave as though they were one. The eventual move to shared electronic catalogues was less a break with that idea than its completion by other means.
Using one was a taught skill
A reader had to guess which word the cataloguer had chosen, then work outward through guide cards until the guess was confirmed or abandoned. Failure was informative in a way search results rarely are, because the neighbouring cards showed what the collection did contain. Browsing an alphabetical run put unrelated but adjacent subjects in front of people, and a good deal of accidental discovery came from that.
Learning the local conventions took an afternoon and then lasted for years, which made regular users considerably faster than occasional ones. Institutions ran short courses in it, which is a reminder that using an information system used to be assumed to require teaching.
Offices ran on the same idea
Sales departments kept a card per customer, and the card carried the whole relationship in handwriting that accumulated over years. Rotating desktop card files put a few hundred contacts within arm's reach and made the size of somebody's network physically obvious. Because a card was a single object, it could be removed, carried to a meeting and lost, which was the main failure mode.
There was no copy unless somebody had made one, so a fire or a departing employee could remove a decade of contact history.
The fragility was understood, and duplicate files were kept exactly where the loss would have been most expensive.
What the replacement does differently
Full-text search removes the requirement to guess the classifier's vocabulary, because the system will match words the record actually contains. That is a genuine and enormous gain, particularly for people who do not already know how a field organises its own subjects.
What it removes is the visible shape of the collection, since results arrive as a ranked list with no neighbours and no edges. A ranked list also conceals the size of the thing you are searching, which a drawer communicated the moment you opened it. Facets, tags and filters are attempts to give some of that structure back, and they work best where somebody has still done the classifying.
Memory is a poor archive, and it is most of what survives.
Where the cards ended up
Many were destroyed once catalogues were converted, and the cabinets were sold or repurposed as storage for small hardware and seeds. Some libraries kept theirs because the cards carried annotations, acquisition notes and provenance details that were never transferred into the electronic record. Those marginal notes have become genuinely valuable to historians, who use them to reconstruct how collections were built and used.
By the end, the lesson is that migration captures the data somebody thought was the data, and discards whatever accumulated around it. Every conversion since has repeated that pattern, usually without anyone noticing until the discarded layer is wanted.
The takeaway
A drawer told you how big the question was before you asked it. A results page never does.
The loss is small and cumulative, which is why it goes unremarked.
Questions readers ask
Was searching a card index really slower?
For a known item, often not, since an experienced user went straight to the drawer. For anything vague, or phrased in words the cataloguer had not used, it was far slower and frequently failed entirely.
Why do some libraries still keep old catalogue cards?
Because the cards often carry handwritten notes, dates and provenance details that were never copied into the electronic catalogue. The card can be a better record of the collection than the record that replaced it.
