How to Reconcile Two Coauthors' Reference Libraries

Two coauthors can each hold a defensible reference library and still produce a merged bibliography that is wrong. The usual failure is not missing papers; it is duplicate records that look different, identifiers that point to the wrong version, and citations that were never checked against the sentence they support. Reconciliation is therefore a verification task before it is a formatting task.

The practical answer: treat reconciliation as a short, staged audit. Freeze both libraries, normalize identifiers, resolve every DOI, compare bibliographic fields against the source text, screen the merged BibTeX, and only then agree on a single canonical file. This article gives that sequence with checkpoints, a worked example, a reusable checklist, and stop rules for when to escalate.

Libraries diverge because the same work can be described in many valid ways. Crossref exposes publisher-deposited metadata through its REST API, including titles, authors, venues, dates, identifiers, licenses, funding, ORCID and ROR identifiers, abstracts when deposited, and post-publication updates from trusted sources. That richness is useful, but it also means two records for one paper may differ in author formatting, venue naming, or date while both are traceable to the same deposit.

A second source of divergence is versioning. A DOI can be resolved through doi.org to the current destination registered for that identifier. Resolution helps locate a work, while bibliographic fields and the source text still need to be checked for identity and claim support. In other words, a working DOI link does not by itself prove that the record's authors, year, or venue are correct, and it does not prove the paper supports the claim it is attached to.

A third source is search behavior. Google Scholar documents searching by author and title, date sorting and date limits, Cited by and Related articles, citation export, library access links, email alerts, and ways to locate alternative versions of a work. Coauthors who searched differently, or who exported from different versions, will arrive with records that need to be matched rather than assumed identical.

Before touching either file, establish what you actually have. Most reconciliation projects go wrong because the team merges first and diagnoses later.

Ask each coauthor to report, in writing, three things: how the library was built (manual entry, database export, or search-tool export), whether DOIs are present for most records, and whether any records were added from memory or secondary sources. A library with DOIs throughout behaves very differently from one assembled by hand.

Export both libraries to a fixed format and record the export date. From this point, all work happens on copies. This prevents a coauthor from silently editing the source while reconciliation is underway and makes disagreements traceable to a specific snapshot.

Decide which fields matter for your output: author list, year, venue, DOI, and the specific claim each citation supports. A reference that is merely "in the library" is not yet a citation; it becomes one when it is attached to a sentence. Write down the target style and the fields you will compare.

Identifiers are the cheapest way to collapse duplicates. Sort both libraries by DOI and by title. Records sharing a DOI are candidates for merging; records with different DOIs but near-identical titles need closer inspection, because they may be different versions or genuinely different works.

For each DOI, resolve it and confirm that the destination matches the record's title and authors. Where a record has no DOI, search by exact title and first author. Google Scholar's documented author-and-title search and its alternative-version lookup are useful here for locating a canonical record.

If you need to search for missing or ambiguous records, use a scholarly search interface rather than a general web search. WisPaper's scholarly search interface is built for finding academic papers, which keeps the search step inside the same environment where you will later store and query the merged set.

Do not resolve duplicates by intuition. Build a table with one row per candidate record and one column per field. This is the artifact that makes the reconciliation auditable and lets a third coauthor review the decisions.

Fill the "Evidence used" column with what you actually checked: a resolved DOI, a Crossref record, or the paper's own text. Leave it blank rather than guessing. A blank cell is a signal to investigate; a guessed cell is a future error.

These are two different questions and they fail in different ways.

Identity: Is this record the work we think it is? Resolve the DOI, compare title and authors against the deposit, and confirm the year and venue. Crossref's inclusion of ORCID and ROR identifiers can help disambiguate authors and institutions when names are common.

Claim support: Does this work actually support the sentence it is attached to? This requires reading the relevant passage in the source text. A resolved DOI tells you where the work lives; it does not tell you what the work says.

Keep these checks in separate columns of your table. A record can pass identity and fail claim support, and that is the more dangerous failure because the citation looks perfect.

Once duplicates are collapsed, export the merged library to BibTeX and screen it. This is where hallucinated or malformed entries surface. TrueCite is designed for exactly this step: it checks BibTeX against real academic databases to flag hallucinated references. Run the merged file through TrueCite's BibTeX citation screening and treat every flag as a record to investigate, not a record to delete automatically.

Screening is a checkpoint, not a finish line. A clean screen means the entries correspond to real works; it does not mean the entries are attached to the right claims. Return to your comparison table for that.

After screening, load the canonical set into a single shared library. In WisPaper, My Library is where you save or upload papers and organize them, which makes it the natural home for the reconciled set. Upload the canonical PDFs and records there so both coauthors work from one source rather than two exports.

Once the set is in one place, you can ask questions grounded in those papers. Library QA lets you ask questions based on papers uploaded to My Library or added from search results. Use it to check whether a specific paper in the merged set addresses a specific claim before you attach the citation.

Suppose two coauthors are preparing a brief on a defined clinical question in a specific population, with a stated outcome. Coauthor A built a library from database exports with DOIs; Coauthor B added papers by hand from reading lists, some without DOIs.

Freeze and classify. Both export on the same date. A reports DOIs for most records; B reports several manual entries.

Normalize. Sorting by DOI collapses the obvious duplicates. Three records remain ambiguous: two share a near-identical title but different DOIs, and one has no DOI at all.

Resolve. The two near-identical records resolve to different destinations, so they are treated as separate works until the source text shows otherwise. The DOI-less record is located by exact title and first author search.

Compare. The table shows one record with a mismatched author order and one with a year that differs between deposit and issue date. Both are corrected against the publisher deposit, with the evidence noted.

Screen. The merged BibTeX goes through TrueCite; flagged entries are investigated against the resolved records.

Claim check. For each citation in the draft, the coauthors confirm the source text addresses the stated population and outcome. Where a paper addresses a related but different population, the citation is either narrowed in the sentence or replaced.

Consolidate. The canonical set is uploaded to My Library, and remaining questions about specific papers are checked through Library QA.

Reconciliation is also an evidence-appraisal task, because a merged library can quietly mix studies of very different strength. Keep three judgments separate in your notes.

• Evidence level: what kind of study is this, and how directly does its design address the question? Record the design alongside the citation.
• Applicability: does the study population and setting match the population in your brief? A paper can be high quality and still not apply.
• Indirectness: does the outcome measured match the outcome you care about, or is it a proxy? Note the gap explicitly.
• Limitations: record what the study itself cannot establish, as stated in the source.

These judgments belong in the comparison table, not in a separate document that drifts out of sync. When a coauthor disputes a citation, the table shows whether the disagreement is about identity, evidence level, applicability, or claim support — four different conversations.

This article describes a reconciliation workflow for the team's own evidence brief. The clinical content of that brief follows the evidence standards the team has set for the population and outcome it addresses.

Use this checklist for each reconciliation round. It is deliberately short enough to complete in one sitting.

1. Both libraries exported and dated; working copies created.
2. Merge target fields and citation style agreed in writing.
3. Libraries sorted by DOI and by title.
4. Every DOI resolved and destination confirmed against title and authors.
5. DOI-less records located by exact title and first author.
6. Comparison table completed for every disputed record, with evidence noted.
7. Identity checks and claim checks recorded in separate columns.
8. Merged BibTeX exported and screened; every flag investigated.
9. Canonical set uploaded to one shared library.
10. Every citation in the draft confirmed against the source text for population and outcome.
11. Evidence level, applicability, indirectness, and limitations recorded per citation.
12. Final file frozen and both coauthors confirm they are citing from it.

Know when to pause rather than push through.

Stop and escalate when a DOI resolves to a different work than the record describes. Do not silently correct the record; flag it, because the error may have propagated into the draft.

Stop when two records cannot be distinguished by title, author, and DOI. Escalate to reading both source texts before deciding whether they are one work or two.

Stop when a citation's claim support is disputed. This is a content disagreement, not a formatting one, and it needs the coauthors to read the passage together.

Escalate when the merged library changes the evidence base of the brief. If reconciliation adds or removes studies that shift the overall picture, the brief's conclusions need review, not just its bibliography.

Escalate when a coauthor's library cannot be reconstructed. If records were added from memory with no traceable source, treat them as unverified until located.

The reconciliation is ready when four conditions hold. First, every record in the canonical library resolves to a work whose title and authors match the record. Second, the merged BibTeX has been screened and every flag resolved. Third, every citation in the draft has been checked against the source text for the population and outcome it claims to support. Fourth, both coauthors are citing from the same frozen file and can point to the comparison table for any disputed entry.

If any of these is missing, the library is merged but not reconciled. The distinction matters: a merged library looks finished, while a reconciled one can be defended.