Citation Management

Citation Management for Scientific Writing

Overview

Citation management is the systematic practice of collecting, organizing, annotating, and inserting bibliographic references in scientific documents. A well-chosen reference manager reduces manual formatting errors, keeps libraries synchronized across devices, and integrates cleanly with word processors and LaTeX. This guide covers tool selection, citation style conventions, DOI and persistent identifier workflows, citation tracking, and common pitfalls that account for a large share of post-publication corrections and retractions related to reference errors.

Key Concepts

1. Reference Manager Capabilities and Tool Comparison

All major reference managers share a common core: they store metadata (author, title, journal, year, DOI), generate formatted citations in a chosen style, and provide a browser extension for one-click capture from publisher pages. The differences lie in collaboration features, cloud storage, word processor plugin support, price, and data portability.

Key differences to know:

• Mendeley is owned by Elsevier. Desktop sync reliability has declined since acquisition; many labs have migrated to Zotero.
• EndNote has the deepest Microsoft Word integration and the largest style library, but high cost makes it unattractive for individual researchers.
• Paperpile's real-time Google Docs integration is unmatched, making it the best choice for teams that write primarily in Docs.
• Zotero is the most flexible: free, open-source, cross-platform, and extensible via plugins.

2. Citation Style Families

Citation styles are governed by the target journal's Instructions for Authors — the author does not choose independently. The two major families are author-date (parenthetical, e.g., "(Smith, 2023)") and citation-sequence (numbered, e.g., superscript¹ or bracketed [1]).

Journal name abbreviation: Vancouver uses ISO 4 abbreviations (e.g., "N. Engl. J. Med."); ACS uses Chemical Abstracts Service Source Index (CASSI); Nature uses PubMed/NLM abbreviations. These are not interchangeable. Your reference manager's built-in Citation Style Language (CSL) file handles this automatically — do not abbreviate manually.

3. DOIs and Persistent Identifiers

A Digital Object Identifier (DOI) is the canonical persistent identifier for a publication and should be included in every citation where available.

DOI format and resolution:

• All DOIs begin with 10. followed by a registrant code (e.g., 10.1038/) and a suffix
• Always resolve DOIs through https://doi.org/10.XXXX/suffix — a bare DOI is not a clickable URL
• Do not use http://dx.doi.org/ (old URL scheme); use https://doi.org/ (current)

Preprint DOIs: bioRxiv, medRxiv, and ChemRxiv assign DOIs to preprints that are distinct from the eventual journal article DOI. When a preprint is published, both DOIs are valid but refer to different versions. Always check whether a peer-reviewed version exists; if so, cite the journal article DOI, not the preprint DOI.

Other persistent identifiers:

• PMID (PubMed ID) and PMCID (PubMed Central ID): NLM-specific; include alongside DOI for medical literature; required by NIH for funded publications
• arXiv ID: physics, math, CS preprints; format arXiv:YYMM.NNNNN
• ISBN: books and book chapters — not resolvable as URL; use for archival purposes only
• Handle: institutional repository identifiers; less universal than DOI

4. Citation Tracking and Impact Metrics

Citation tracking serves two purposes: monitoring how often your own publications are cited (research impact) and discovering papers that have cited a key reference you found (forward citation chaining).

Setting up citation alerts: Google Scholar Alerts are the simplest way to receive email notifications when a paper is cited. Create an alert for your own papers, for key papers in your field, and for relevant author names.

Decision Framework

What is your primary need?
├── Set up or switch reference manager
│   ├── LaTeX + Overleaf as primary editor
│   │   └── → Zotero + Better BibTeX plugin (auto-syncing .bib)
│   ├── Google Docs as primary editor
│   │   └── → Paperpile (native Docs integration) or Zotero (Docs plugin)
│   ├── Microsoft Word, no institutional license
│   │   └── → Zotero (free, robust Word plugin)
│   ├── Microsoft Word, institutional license available
│   │   └── → EndNote (deepest Word integration) or Zotero
│   └── Collaborative team across institutions
│       └── → Zotero group library (free, open, no vendor lock-in)
│
├── Choose a citation style
│   ├── Medical / clinical journal → Vancouver (superscript numbered)
│   ├── Chemistry journal → ACS style
│   ├── Nature / Science family → Nature style
│   ├── Psychology / behavioral science → APA 7th
│   └── Engineering → IEEE
│   (Always check journal Instructions for Authors — do not guess)
│
├── Systematic review workflow
│   ├── Reference collection → Zotero or Mendeley
│   ├── Deduplication → reference manager built-in + manual check
│   ├── Title/abstract screening → Rayyan, Covidence, or Abstrackr
│   └── Full-text review → download PDFs into reference manager
│
└── Citation tracking
    ├── Free tracking → Google Scholar + iCite
    └── Institutional tracking (IF, h-index) → Web of Science or Scopus

Best Practices

1. Capture metadata at the point of discovery, using the browser extension from the publisher's page: Do not bookmark URLs to "add later" — metadata capture from publisher pages is far more accurate than from PDFs, which frequently garble author names with non-ASCII characters, truncate long author lists incorrectly, or misparse page ranges in old articles.

2. Always verify DOI resolution after import: After importing any reference, click the DOI link. Automated metadata import frequently fails for pre-2000 papers, conference proceedings, and book chapters. A broken or wrong DOI is a legitimate reason for a post-publication correction and will delay typesetting.

3. Store PDFs inside the reference manager, not in a parallel folder structure: Keeping PDFs as attachments within the reference manager (rather than external links to a file system location) ensures citations and full texts remain synchronized when you move, rename, or reorganize directories. External file links break silently on directory restructuring.

4. Use a consistent citation key format for LaTeX and never change it mid-project: In Zotero with Better BibTeX, configure a global citation key template (e.g., [auth:lower][year][veryshorttitle]) before importing your first reference. Changing the template mid-project regenerates all keys, orphaning every \cite{} command in your existing .tex files.

5. Organize by topic, not by manuscript: Create collections by research topic ("CRISPR mechanisms," "cardiac MRI methods") rather than by output document ("Chapter 3," "Paper 1"). Topic organization allows reference reuse across projects and prevents accumulating duplicate entries in separate project collections.

6. Run "Find Duplicates" before every manuscript submission: Duplicate entries produce references with inconsistent formatting for the same paper — one as "Smith et al." and one with the full author list — which editors and reviewers immediately flag. Most managers include automatic deduplication; run it at least once before generating the final reference list.

7. Export a static archive at submission: Export the reference list as .bib, .ris, or .enl at the moment of manuscript submission and archive it alongside the submitted PDF. This ensures you can reproduce the exact reference list even if your live library changes during the review period (which often lasts 6–18 months for high-impact journals).

8. For NIH grants, track PMCIDs separately: NIH requires that any publication resulting from NIH funding include the PubMed Central ID (PMCID) in grant progress reports and renewal applications. Maintain a spreadsheet of your publications with PMID, PMCID, and grant numbers to avoid last-minute compliance scrambles.