How Chain Indexing Works: A Step-by-Step Guide

When a classifier assigns a number like 954.03 to a book, that number quietly holds an entire structure of knowledge inside it. Chain indexing is the technique that unlocks this structure and turns it into subject headings that readers can actually search for. Developed by Dr. S.R. Ranganathan, it lets the cataloguer begin exactly where the classifier stopped, deriving subject entries from the class number through a near-mechanical procedure rather than guesswork. This guide walks through how the mechanism works, step by step, using the Dewey Decimal Classification (DDC) as the working example.

What a chain really means here 🔗

In ordinary speech a chain is a series of connected links. In indexing, the idea is almost the same. A chain is the structural representation of a subject, showing how its parts connect from the most general idea down to the most specific one. Each individual concept in that sequence is a link. A complete chain contains a link for every level of division lying between the broadest class and the exact subject of the document.

This hierarchy is the engine of the whole method. Because schemes like the DDC and Ranganathan’s Colon Classification use notation where each added digit signals a step of subordination, the class number itself is a coded chain. The job of the indexer is to read that chain in reverse and convert it into words the reader will recognise. The chain procedure is therefore described as a method to derive subject index entries directly from the class number, avoiding any duplication of the classifier’s work.

The steps in chain indexing 🔗

Unwinding a class number into meaningful headings follows a fixed sequence. The standard procedure can be broken into a few clear stages, which makes the method easy to apply consistently.

Start with the class number 🔗

The first stage is already done by the classifier. Suppose a book is titled A History of British Rule in India and carries the class number 954.03 in DDC. The indexer does not re-analyse the subject from scratch. The number is the raw material for everything that follows.

Write out the chain digit by digit 🔗

Next, the class number is unwound one digit at a time, and each stage is given the verbal term taken from the DDC schedule. The procedure consists in unwinding the chain one digit at a time, where each step translates into one link. For 954.03 the chain reads:

9 → History (900)
95 → Asia, History (950)
954 → India, History
954.03 → British Period, India, History

Classify each link 🔗

Once the chain is written, every link is examined and labelled as a sought, unsought, or false link, with missing links inserted where the scheme has a gap. This judgement is the one genuinely intellectual part of an otherwise mechanical process. The types of links are explained in detail in the next section.

Reverse render to get the subject heading 🔗

The specific subject heading is built by reading the sought links backwards, from the most specific link upward. Starting at 954.03 and moving up gives the heading British Period, India, History. This is called reverse rendering, and it produces a heading that is co-extensive with, meaning exactly as broad as, the subject of the book.

Prepare class index entries 🔗

Finally, a class index entry is prepared for each sought link, pointing the reader from the verbal heading back to the class number. A complete class index entry has three parts: the subject heading, a directional element guiding the reader to the classified file, and the class number itself. One entry is generated for every sought heading, so the number of entries depends on how many sought links the chain contains.

The types of links in a chain 🔗

Not every link earns a place in the index. Sorting the links correctly is what separates a clean, usable index from a cluttered one. There are four kinds the indexer must recognise.

Sought links 🔗

A sought link represents a concept that a reader is likely to use as an access point when searching. These are the links that deserve their own entries. The first link in a chain is always a sought heading, and the last link is usually sought too, provided it does not end in a time facet. In the example above, History, Asia History, India History, and British Period are all sought, because readers genuinely look for documents under each.

Unsought links 🔗

An unsought link represents a concept that exists in the hierarchy but that no reader is realistically going to search under. As the classic teaching example shows, a long chain might contain eleven links yet yield only a handful of significant ones, with several treated as unsought and simply dropped. Recognising these saves the catalogue from being swamped with useless entries.

False links 🔗

A false link carries no real subject meaning at all. It is usually a connecting symbol, an indicator digit, the decimal point, or a place-holding zero. In DDC, the zeros used as fillers and the dot before the decimals are false links, since they signal the type of facet to follow rather than a subject. A link is also treated as false if it ends with a connecting symbol or a relation device. False links are always ignored and never generate an entry. Importantly, in practice a false link is simply a meaningless notational element, not the same thing as a mistaken connection between two unrelated subjects.

Upper and lower links 🔗

The terms upper link and lower link describe relative position rather than a fixed category. An upper link is the broader concept; the lower link is the narrower one sitting directly next to it. The relationship is relative, so a single link can be an upper link for one neighbour and a lower link for another. Taking the chain 900 → 950 → 954: the link 950 (Asia, History) is the lower link of 900 (History) but the upper link of 954 (India, History). The last link in any chain has no lower link beneath it.

One more category is worth knowing. A missing link is a concept that the document needs but that the classification scheme does not provide. The indexer supplies it through verbal extension at the point of the gap, so the chain stays unbroken.

Working through a DDC example 🔗

Bringing the steps and link types together makes the mechanism concrete. Take a book on the history of India under the Moguls, classified at 954.025. Unwinding the number produces a chain something like:

9 → History
95 → Asia, History
954 → India, History
954.0 → (false link: the zero is only a place-holder)
954.02 → Medieval period, India, History
954.025 → Mogul period, India, History

Reading the sought links upward gives the specific subject heading Mogul Period, India, History. The indexer then prepares class index entries for each sought heading, moving from broad to specific: History (900), Asia History (950), India History (954), and Mogul Period India History (954.025). The zero at 954.0 is dropped as a false link, contributing nothing to the index. Notice how each entry guides a reader who starts from a different level of specificity straight to the same shelf location.

This is exactly why chain indexing was prized for national bibliographies. The British National Bibliography adopted the technique in 1951, layering it onto the DDC notation, and used it through the 1950s and 1960s before later moving to the PRECIS system.

How to decide which links are useful 🔗

The hardest skill is judging the boundary between sought and unsought links, because this is where the indexer’s discretion enters an otherwise automatic process. A few practical tests help.

Test for reader approach: Ask whether a real reader would plausibly look for the document under this term. If yes, it is sought; if the level is too abstract or too incidental to matter, it is unsought.

Watch the facet endings: A link that ends in a time facet or a form division is often not the access point readers want, and a link ending in a connecting symbol is false by definition. These should be filtered out before any entry is made.

Avoid redundancy: Class index entries for the broader sought links, such as History or Asia History, are prepared only once for the whole library. If they already exist from an earlier document, there is no need to repeat them; only the new, more specific heading needs a fresh entry. This is what makes the method economical, since each term is dropped after it has been indexed once.

Applied carefully, these checks ensure the index is co-extensive with the subject, free of notational noise, and built without duplicating either the classifier’s effort or entries already in the catalogue. That balance of precision and economy is what made chain indexing such an enduring contribution to subject cataloguing.

What do you think? If you had to index a book on Drug Addiction among Youth in India, which links in its chain would you treat as sought and which as merely unsought? And as catalogues move into fully digital, keyword-driven search, does a hierarchical method like chain indexing still hold value, or has automated indexing made it largely a historical lesson?

References

1. https://www.librarianshipstudies.com/2017/04/chain-indexing.html
2. https://egyankosh.ac.in/bitstream/123456789/38416/5/Unit-14.pdf
3. https://www.studocu.com/in/document/university-of-delhi/library-science/unit-14-pdf/89625416
4. https://ebooks.inflibnet.ac.in/lisp3/chapter/subject-cataloguing-chain-procedure-popsi-and-precis/
5. https://en.wikipedia.org/wiki/British_National_Bibliography