When Science Corrects Itself: Retraction & Self-Correction

Scientific Methodology

Chapter 9 · When Science Corrects Itself: Retraction & Self-Correction

The sibling course Research Methodology used the Wakefield MMR fraud as its own retraction case — a real 12-year gap between publication (1998) and retraction (2010). This chapter uses a genuinely different, far faster case to show self-correction actually working close to the speed it's supposed to.

A Real, Fast Self-Correction: The Surgisphere Scandal

In May 2020, two major papers — one in The Lancet (22 May, on hydroxychloroquine and COVID-19 mortality), one in the New England Journal of Medicine (1 May, on cardiovascular drugs and COVID-19) — drew on a purported multinational hospital registry from a small analytics company, Surgisphere.

The Claimed Dataset

The Lancet paper claimed real data from 96,032 patients across 671 hospitals on six continents — an extraordinary registry no one in the research community had previously known to exist.

Independent researchers quickly flagged real implausibilities — patient and hospital counts, and demographic/dosing breakdowns, that didn't square with what was independently known about hospital records in the claimed regions. Surgisphere refused to grant independent auditors access to the underlying patient-level data or identify the participating hospitals. On 28 May 2020, over 180 researchers, clinicians, and statisticians signed a real open letter to the Lancet's editor stating the numbers "seem unlikely" and demanding an independent audit.

CasePublishedRetractedReal gap
Wakefield MMR (Research Methodology)19982010~12 years
Surgisphere — Lancet22 May 20204 June 2020~13 days
Surgisphere — NEJM1 May 20204 June 2020~34 days
Why the Contrast Matters
Both cases involved a real, serious flaw eventually caught and formally retracted — but the real gap between them is enormous. Wakefield's fraud survived over a decade of real, ongoing public and medical consequences before retraction; Surgisphere collapsed within weeks once independent scientists actually scrutinized the data. Self-correction isn't guaranteed to be fast, but this case shows it genuinely can be, especially when a claim reaches enough scrutiny quickly.

The Formal Escalation: Correction, Expression of Concern, Retraction

The Committee on Publication Ethics (COPE), founded in 1997 and followed by most major journals and publishers today, defines a real, three-tier framework for how a journal responds to a flawed paper:

Correction

An honest error that doesn't undermine the paper's core results or conclusions — a fix, not a retraction.

Expression of Concern

Real, unresolved doubt about a paper's reliability, often issued mid-investigation, without yet formally invalidating it.

Retraction

Reserved for errors or misconduct serious enough to genuinely invalidate the paper's own results and conclusions.

The Surgisphere case is a clean, real, worked example of this exact escalation. NEJM issued an Expression of Concern on 2 June 2020; The Lancet followed with its own on 3 June. Both papers were formally retracted the very next day, 4 June 2020 — a real two-step escalation, from suspicion to formal invalidation, completed in under 48 hours once independent verification of the underlying data proved impossible.

Retraction Watch: Tracking Self-Correction Itself

Retraction Watch was founded in August 2010 by science journalists Ivan Oransky and Adam Marcus, specifically to track and report on scientific retractions. Its structured Retraction Watch Database (launched 2015) now tracks over 30,000 retractions and counting — a real, running public record of exactly the kind of self-correction this chapter is about.

Is Science Getting Worse? A Real, More Nuanced Trend

A Real, Measured Increase

The retraction rate for European biomedical-science papers roughly quadrupled between 2000 and 2021, with research misconduct — data or image manipulation, authorship fraud — accounting for roughly two-thirds of retractions and a growing share of the total.

Why This Isn't Simply "Science Is Getting Worse"
A rising retraction rate could mean more misconduct is genuinely happening — or that more of it is being caught. Research-integrity specialists attribute the real rise to both: paper mills and image manipulation have genuinely grown at scale, but so has detection — tools like PubPeer (a post-publication scrutiny site launched in 2012), statistical forensics, and plagiarism/image-duplication software have made misconduct far harder to hide than it was decades ago. Most experts believe improved detection is doing a large share of the real work behind the rising number — which means, echoing this chapter's own theme, a rising retraction count can itself be a sign self-correction is working better, not proof science is failing.

Hands-On Exercises

Exercise 1

Explain, in your own words, what specifically allowed the Surgisphere case to be caught and retracted so much faster than Wakefield's — what real factor made the difference between roughly two weeks and roughly twelve years?

📄 View solution
Exercise 2

Explain, in your own words, why the Surgisphere case moved through an Expression of Concern before a full retraction rather than going straight to retraction — what real purpose does that intermediate step serve?

📄 View solution
Exercise 3

Explain, in your own words, why a rising retraction rate is genuinely ambiguous evidence about whether science is "getting worse" — what two real, competing explanations does it require you to weigh?

📄 View solution

Chapter 9 Quick Reference

  • The Surgisphere/hydroxychloroquine scandal (May-June 2020) saw two major papers retracted within ~13-34 days of publication, versus Wakefield's ~12-year gap
  • COPE's three-tier framework — correction, expression of concern, retraction — reflects increasing severity; Surgisphere moved from expression of concern to full retraction in under 48 hours
  • Retraction Watch (founded 2010, Oransky & Marcus) tracks over 30,000 real retractions in its public database
  • Biomedical retraction rates roughly quadrupled 2000-2021 — driven by both genuinely more misconduct and genuinely better detection tools (e.g., PubPeer, 2012)
  • A rising retraction rate is ambiguous evidence: it can reflect a worsening problem or a self-correction system working better than before