Chain indexing is one of the most elegant ideas in cataloguing. Instead of guessing which subject headings a reader might look for, you let the class number do the work. Once a document is classified, the indexer simply unwinds that class number digit by digit and translates it back into ordinary language. The result is a set of subject entries that point straight to the classified part of the catalogue. The method was developed by S.R. Ranganathan, but it was his close associate Professor G. Bhattacharyya who laid out the full procedure as a clear sequence of eleven steps. This guide walks through those steps and shows how to apply them to the Dewey Decimal Classification, especially the 19th edition.
Table of Contents
- Understanding the 11 steps
- Steps for analysing and naming the subject
- Steps for building the chain and writing entries
- Application to DDC 19th edition
- Special problems with Dewey
- Examples of real-world application
- A book on history
- A book on science
- Sorting and organizing headings
- Editing and refining index entries
- Use schedule terminology, not your own
- Add cross-references and qualifiers
- Handle the missing link
- Check coextensiveness and reverse rendering
Understanding the 11 steps
According to Professor G. Bhattacharyya (1936-2006), there are eleven steps involved in deriving subject headings from a class number and preparing class index entries. These steps cover everything from analysing the document right up to arranging the final entries. Some of them deal with synthesising the class number itself, while the rest deal with turning that number back into words. Here is the full sequence as explained in standard cataloguing literature on chain procedure.
Steps for analysing and naming the subject
1. Determination of the specific subject: Identify exactly what the document is about. This is done using the title, the table of contents, and a careful reading of the text. The aim is to arrive at the most specific subject, not just a broad area.
2. Expressive name of the subject: State that specific subject in natural language, clearly and fully, so nothing important is left out.
3. Kernel terms: Strip the expressive name down to its fundamental components. You remove all the auxiliary words (articles, prepositions, connectives) and keep only the meaningful kernel terms.
4. Analysed name of the subject: Decide the category of each kernel term according to a set of postulates and principles. In Ranganathan’s system this means sorting terms into fundamental categories such as Personality, Matter, Energy, Space, and Time.
5. Transformed name of the subject: Rearrange the components, if needed, to follow the rules of syntax. This puts the terms into the correct citation order.
6. Standard terms: Replace each term with the standard term used in the classification scheme you are working with. This keeps the index consistent with the schedule.
Steps for building the chain and writing entries
7. Determination of links and construction of the chain: Represent the class number as a chain. Each link has two parts, the class number and its translation into natural language, joined by an equals sign, with the links written one below the other.
8. Determination of the different kinds of links: Classify each link as a Sought Link, False Link, Unsought Link, or Missing Link. This step decides which links will actually become index entries, and it is the heart of the method.
9. Derivation of subject headings: Derive a subject heading from each sought link, working in reverse from the most specific link to the most general. This is called reverse rendering.
10. Preparation of cross-reference entries: Prepare subject reference entries to connect related or alternative terms, so a reader who looks under one word is guided to the right place.
11. Arrangement: Merge all the entries and arrange them in a single alphabetical sequence in the alphabetical part of the classified catalogue.
The four kinds of links in step 8 deserve a closer look, because most of the practical difficulty in chain indexing lies here. A Sought Link (SL) is a link a reader is likely to approach the document by. An Unsought Link (USL) is a valid concept that readers are unlikely to search under. A False Link (FL) does not represent a genuine concept on its own, and usually ends in a connecting symbol or indicator digit. A Missing Link (ML) corresponds to an isolate that exists conceptually but has no separate digit in the class number, creating a gap in the chain.
Application to DDC 19th edition
Chain indexing was originally designed for Colon Classification, a faceted scheme whose notation shows hierarchy clearly. It can be applied to any scheme whose notation follows a hierarchical pattern, and that includes Dewey. But Dewey is an enumerative scheme, not a fully faceted one, so it brings its own complications. The British National Bibliography, the British Catalogue of Music, and the Indian National Bibliography were major users of chain indexing in the 1950s and 1960s, and their experience with applying it to a less systematic scheme like DDC shaped many of its refinements.
The basic mechanics stay the same. You unwind the class number one digit at a time, and each step becomes one link. Each indicator digit, and any link drawn from the Time facet, corresponds to a false link, which is ignored. You translate each link using only the terms found in the DDC schedules, never your own wording, because this is what gives the index its consistency. Then you prepare a class index entry for each sought link, moving from the narrower heading to the broader one.
Special problems with Dewey
The 19th edition throws up a few specific issues. One is the problem of dual headings, where a single number carries two captions. For example, 597 is captioned both Cold-blooded Vertebrates and Pisces (Fishes), and 954 carries both South Asia and India. The indexer has to decide which caption to use as the heading. Another is the problem of alternative headings, where a number can be expressed in more than one acceptable way. These ambiguities were later cleaned up in newer editions, but in the 19th edition they require a deliberate choice by the indexer.
Examples of real-world application
Examples make the procedure click, so let us trace two, one from history and one from science.
A book on history
Take a classic textbook example: a book titled History of India during the Mughal king Shah Jahan. In DDC this is classified at 954.0257. Unwinding the number gives a chain that runs from the broad subject down to the specific one:
9 = History. 95 = Asia, history. 954 = India, history. 954.02 = Muslim period (medieval India), history. 954.025 = the Mughal era. 954.0257 = the reign of Shah Jahan. The decimal point and the digits that signal period subdivisions function as false links and are dropped.
Working backwards through the sought links, you derive a set of class index entries such as SHAH JAHAN, INDIA, HISTORY; MUGHALS, INDIA, HISTORY; MUSLIM PERIOD, INDIA, HISTORY; INDIA, HISTORY; and ASIA, HISTORY. Each of these refers the reader to the same spot in the classified part of the catalogue. Notice how a reader who only knows the broad term “Asia” and one who knows the very specific term “Shah Jahan” both reach the right place.
A book on science
Now take a science example to show how the same logic handles a different discipline. In DDC, technology and applied science sit in the 600s, with sub-divisions like 620 for engineering, 621 for applied physics and mechanical engineering, and 004 for computer science under the 000s. Suppose you have a book on a narrow topic within mechanical engineering. You build the chain from 600 (Technology) down through the relevant sub-class to the specific number assigned to the book. You then mark which links are sought. A reader is very likely to look under “Mechanical engineering,” so that is a sought link. An intermediate link that no reader would naturally search under becomes an unsought link and generates no entry. The indicator points and any non-meaningful steps become false links and are dropped. The surviving sought links are reverse-rendered into headings.
The pattern is identical across subjects. You only ever prepare one fully specific entry for the document; the broader entries are generated by dropping the rightmost term step by step. This is what makes the method economical, because it gives both generic and specific access without permuting every component.
Sorting and organizing headings
Once you have your raw chain, the real judgement begins. Sorting headings is about separating the links that earn an entry from the ones that do not.
Two rules anchor this work. The first link is always a sought heading. It represents the main discipline and is a natural access point. The last link is also a sought heading, unless it ends with a Time facet, in which case it is treated as a false link. Everything in between has to be judged individually.
The decision about whether a link is sought or unsought is partly arbitrary, and that is by design. It depends on the library, the collection it holds, the readers it serves, and the kind of service it wants to offer. A specialist research library and a small college library might mark the same intermediate link differently, and both can be correct for their context. The guiding question is simple: is a reader in this library likely to approach the document through this concept?
After marking the links, you order the entries from the narrowest sought heading to the broadest. Each sought heading produces exactly one class index entry, so the number of entries equals the number of sought links. Writing the chain vertically, with the class number on the left and the verbal translation on the right, makes the whole structure visible and reduces mistakes.
Editing and refining index entries
The last stage is polishing. A chain that is mechanically correct can still be clumsy for the reader, so editing focuses on precision and usefulness.
Use schedule terminology, not your own
Every term in an entry should come from the classification schedule, not from your own vocabulary. This is what keeps headings uniform no matter who does the cataloguing. If two indexers follow the procedure honestly, they should arrive at the same headings, which is one of the recognised strengths of chain procedure over hand-picked subject heading lists.
Add cross-references and qualifiers
Where a reader might use a different word for the same idea, add a cross-reference so they are guided to the preferred heading. Qualifiers help here too. Pairing a general term with a more specific one, for instance linking a plain “history” entry to “India, history,” removes ambiguity and tells the reader exactly which slice of the subject they have reached.
Handle the missing link
Sometimes a concept that readers genuinely want has no separate digit in the class number, which produces the problem of the disappearing or missing chain. When this happens, the indexer inserts the missing isolate through verbal extension at the point of the gap, so the chain still offers an entry for that sought concept. This is one of the trickier editing tasks, and it is exactly the weakness that later systems like POPSI, also developed by Bhattacharyya, were designed to overcome.
Check coextensiveness and reverse rendering
Finally, confirm that the most specific entry is genuinely coextensive with the subject of the book, meaning it represents the whole subject and nothing more. Then read the reverse-rendered headings as a reader would. Reverse rendering can occasionally produce a word order that feels unnatural, and spotting these is part of refining a good index into a great one.
Done carefully, these five stages turn a string of digits into a reliable, alphabetical key to the classified catalogue. The procedure is semi-mechanical, but the judgement at the sorting and editing stages is what separates a competent index from an excellent one.
What do you think? If you were cataloguing for a small college library versus a large research institute, which intermediate links would you mark as sought, and how would your choices differ? And when a sought concept has no matching digit in the DDC schedule, do you think inserting a missing link is a fair fix, or does it stretch the “mechanical” promise of chain indexing too far?
Leave a Reply