One Catalogue, Two Codes: An AACR2-to-RDA Consistency and Authority-Control Audit of a University Library OPAC

Most Indian university libraries computerised their catalogues in the 2000s under AACR2, the Anglo-American Cataloguing Rules, and have kept accessioning ever since. The international standard, however, moved on: RDA — Resource Description and Access — replaced AACR2 from 2010–13, and the Library of Congress, the British Library and most major cataloguing agencies implemented it in 2013. The result inside one typical OPAC is a hybrid catalogue: records transcribed under two different codes, with different transcription conventions, different access-point policies and different handling of the General Material Designation — sitting in one database with no systematic record of where the seams are.

This M.Lib.I.Sc dissertation audits that seam. Working at the Central Library of Vidyagiri State University (fictional), which runs SOUL and has catalogued continuously since 2004, the study draws a stratified random sample of 400 bibliographic records — 200 created before 2013 (the AACR2 era) and 200 after 2017 (the settled RDA era) — and audits each against a 12-element checklist built from AACR2 and RDA transcription rules and MARC 21 documentation.

The audit measures element-by-element consistency: presence of the GMD in 245 $h versus RDA's 336/337/338 content-media-carrier fields, abbreviations, ISBD punctuation, access-point practices, and whether name headings match their forms in the LC Name Authority File and VIAF. Chi-square tests compare the two year-bands; error and variant rates per element become a gap register, a one-page house style sheet, and a costed conversion checklist the library can actually execute — the dissertation's practical deliverable to the institution.

An Indian university library that automated in the 2000s and continues to accession today carries two cataloguing codes in one database — AACR2-era records and RDA-era records — with no inventory of what that heterogeneity costs users or the library. Transcription conventions differ: the GMD appears in old records and not in new ones; abbreviations such as 'p.' and 'ed.' persist in one era and not the other; access points follow different policies; and name headings drift from the forms recorded in LCNAF and VIAF, so the same author can hide under several headings. Because nobody has measured the mix, the library cannot answer basic management questions: what proportion of the catalogue is RDA-consistent, which elements diverge most, whether users face duplicate entries under variant headings, and what an upgrade to full RDA consistency would require in staff-days. Cataloguing literature debates the codes rule by rule but rarely audits a real hybrid OPAC record by record. This dissertation converts an invisible problem into a measured one — and into a remediation plan the library can budget for.

Abstract

A library that has catalogued continuously for two decades does not have one cataloguing code — it has layers. This dissertation audits the layering in the OPAC of a state university central library (SOUL, records since 2004) by comparing 400 randomly sampled records from the AACR2 era (pre-2013) and the settled RDA era (post-2017) across 12 transcription and access-point elements, verifying name headings against LCNAF and VIAF, and translating the measured inconsistencies into a house style sheet and a costed conversion checklist for the library.

Introduction

Cataloguing codes are not software; they do not upgrade in place. When RDA succeeded AACR2, libraries worldwide faced the same question Indian university libraries still face: what happens to two decades of records made under the old code? International literature (for example, studies of RDA implementation in large libraries) describes the decision process at well-funded institutions, but the typical Indian university library's situation — a single cataloguer, SOUL or Koha, decades of AACR2-era records, and copy cataloguing mixed with original cataloguing — is barely measured. Without measurement, upgrade plans are guesswork and cataloguing quality debates remain anecdotal.

Existing studies and the gap

Published comparisons of AACR2 and RDA explain rule-by-rule differences — GMD versus the 336/337/338 fields, abbreviation policy, relationship designators — and LIS journals carry awareness surveys of Indian cataloguers. What is missing is a record-level audit of a real hybrid catalogue that quantifies how much of each code actually sits in the database, how often headings drift from their authorised forms, and what a remediation plan would cost in staff-days. This study supplies that measurement and the plan.

The proposed study

A stratified random sample of 400 records — half created before 2013, half after 2017, stratified further across monographs, theses and e-resources — is exported in MARC 21 from SOUL and audited on a 12-element checklist: GMD presence and form; 245 transcription; edition, place, publisher and date transcription (abbreviated versus RDA-style); extent phrasing (pages versus pages, unnumbered sequence handling); 336/337/338 presence and correctness; ISBD punctuation integrity; main and added entry structure; relator terms; subject-heading form (LCSH consistency); series treatment; and authority verification of the first three name headings against LCNAF/VIAF.

Ten percent of the sample is re-audited independently by a classmate for intra/inter-auditor agreement. Element error and variant rates are computed per year-band; chi-square tests whether the two eras differ per element. Deliverables: a gap register, a house style sheet, a prioritised conversion checklist with staff-day estimates, and an authority-control workflow for SOUL and Koha.

Feasibility

All 400 records exist in the library's own database; export, audit and analysis need only Excel, SPSS and the free RDA Toolkit trial period and LC documentation. The audit sheet is piloted on 20 records. Permission is sought from the University Librarian; records contain no personal data. The 20-week schedule reserves six weeks for analysis and writing, so a slow permissions process cannot sink the timeline.

In scope

• The bibliographic database of one university central library (SOUL), records created 2004–2025.
• 400 records: stratified by era (200 pre-2013, 200 post-2017) and format (monographs, theses, e-resources).
• A 12-element record-level audit built from AACR2, RDA and MARC 21 documentation, plus LCNAF/VIAF verification of three headings per record.
• Statistical comparison of the two eras, a gap register, a house style sheet, a costed conversion checklist and an authority-control workflow proposal.

Out of scope

• Retrospective conversion of legacy card data or record clean-up execution — this study audits and plans; it does not edit the live database.
• Subject-heading depth analysis beyond LCSH form consistency, and classification practice (DDC numbers) except where it affects the audit sheet.
• Interoperability testing with union catalogues, and user-side search-behaviour studies.
• Any claim about other libraries; findings generalise to the audited catalogue and suggest method, not universals.

Design. A record-level audit study — quantitative, descriptive and comparative — executed in four phases: instrument development, sampling, audit, analysis. The audit sheet treats each bibliographic record as a case with 12 binary and categorical element variables plus three authority-check outcomes.

flowchart TD
  A["Permissions: University Librarian and department ethics note"] --> B["Build audit sheet from AACR2, RDA and MARC 21 rules"]
  B --> C["Pilot on 20 records; refine decision rules"]
  C --> D["Stratified random sample: 200 pre-2013, 200 post-2017"]
  D --> E["Export MARC 21 from SOUL; audit 400 records"]
  E --> F["Reliability: 10 percent re-audited independently"]
  F --> G["Analysis: error rates per element, chi-square by era band"]
  G --> H["Deliverables: gap register, style sheet, conversion checklist, authority workflow"]

Sampling. The population is all bib records with creation dates in SOUL. Two strata by era (pre-2013, post-2017) are drawn at 200 each with a random start and fixed interval from date-sorted lists; within each era, allocation across monographs, theses and e-resources is proportional. The sampling plan, with its random start, prints in the appendix.

Instrument. The 12-element audit sheet — each element with its AACR2 rule, its RDA counterpart, the MARC 21 field affected, and the coded verdict (compliant, variant, absent, hybrid) — is derived from the codes themselves and from LC MARC 21 documentation, then piloted on 20 records until every decision rule is unambiguous.

Reliability. A classmate re-audits a random 10 percent (40 records); element-level agreement and Cohen's kappa are reported, and any element under 0.70 agreement is re-defined and re-audited before analysis proceeds.

Analysis. In Excel and SPSS: element error/variant rates per era band; chi-square tests of era × element verdict; authority-verification match rate; a hybrid-record count (records carrying neither GMD nor 336/337/338). Findings convert directly into the conversion checklist, each item costed in staff-days from the audit's own rates.

As a dissertation, the 'architecture' is the study design — how a live catalogue becomes an auditable dataset — and the audit sheet's logic, which maps every cataloguing-code difference to the MARC 21 field where it is observable.

flowchart TD
  CAT["SOUL catalogue: records 2004-2025"] --> EXP["MARC 21 export, date-sorted"]
  EXP --> S1["Stratum A: 200 records, pre-2013 (AACR2 era)"]
  EXP --> S2["Stratum B: 200 records, post-2017 (settled RDA era)"]
  S1 --> AUD["12-element audit sheet + LCNAF / VIAF checks"]
  S2 --> AUD
  AUD --> KAP["Reliability: 10 percent re-audit, Cohen's kappa"]
  KAP --> ST["SPSS: chi-square by era, element error rates"]
  ST --> OUT["Gap register, house style sheet, costed conversion checklist"]

flowchart LR
  E["Element audited"] --> G["GMD vs 336/337/338 — fields 245 h, 336, 337, 338"]
  E --> T["Transcription — fields 245, 250, 260/264, 300"]
  E --> AP["Access points — fields 100, 110, 700, 710, relators in 100 and 700"]
  E --> SH["Subject form — field 650, LCSH practice"]
  E --> AU["Authority — heading vs LCNAF and VIAF form"]

The second diagram is the dissertation's conceptual spine: every AACR2-versus-RDA difference that matters is observable in a named MARC 21 field, which is what makes the audit objective — an auditor codes the field as it stands, not the cataloguer's intent. The same mapping drives the conversion checklist, since each remediation item names its field.

• Execute the checklist: a follow-on project that applies the conversion checklist to one collection (say, the theses) and measures the before–after consistency shift.
• Scale the audit: replicate in three libraries of different automation vintage to test whether the hybrid-pattern findings hold across institutions.
• BIBFRAME horizon: extend the audit's logic to examine how the catalogue's data would map toward BIBFRAME as linked-data migration becomes practical.
• Authority control automation: pilot heading auto-verification against VIAF through a script or Koha's tools, costed against the manual rates measured in this study.
• User-side validation: pair the metadata findings with a search-task study measuring whether heading drift actually costs users retrieval time.