Plagiarism
in Research & Scientific Writing
Dr. Arun Mitra
Three Day workshop on Manuscript Preparation & Research Communication
3 - 5 June, 2026
Why this matters, before any definition
One paper. 1998. The Lancet. Andrew Wakefield linked the MMR vaccine to autism. The data were falsified and the conflicts of interest undisclosed.
Retracted in 2010. But the damage outlived the retraction: collapsing vaccine coverage and measles outbreaks that continue 25+ years later.
Closer to home, Indian Vice-Chancellors and senior faculty have been removed from office over plagiarism. The UGC wrote its 2018 law because of cases like these.
This is not only an ethics issue. It is a career, and a patient-safety, issue.
This is not abstract
It reaches senior faculty, in India, now.
A Vice-Chancellor sent to jail
A Vice-Chancellor sacked by the President
A Vice-Chancellor under CBI probe
Twenty-five papers from one university, retracted
And it is accelerating
Retractions from Indian institutions, 2016 to 2025
What you’ll be able to do
By the end of these 30 minutes:
- Recognise the forms of plagiarism, including the ones that don’t feel like plagiarism.
- Read your similarity report against the UGC 2018 thresholds and journal rules.
- Apply real paraphrasing, quoting and citing techniques to your own thesis & manuscripts.
- Navigate the grey zones: self-plagiarism, “common knowledge”, methods text, and AI tools.
What plagiarism actually is
Presenting someone else’s words or ideas as your own, without attribution.
The half residents forget is “or ideas.” You can plagiarise without copying a single sentence.
Two axes that cause all the confusion:
- Intent: deliberate · reckless · unintentional. all three are penalised
- Object: text · data · images · ideas / structure
“I didn’t mean to” is not a defence under the UGC regulations.
The taxonomy, from obvious to subtle
- Verbatim / copy-paste: the obvious one.
- Mosaic / patchwork: phrases stitched from several sources. The most common resident error.
- Close paraphrase: change a few words, keep the structure. Still plagiarism.
- Idea / structure: reuse an argument or framework, even fully reworded.
- Self-plagiarism & duplicate publication: reusing your own text or data without disclosure (“salami slicing”).
- Citation plagiarism: citing a paper you never read, copied from another’s reference list.
Spot the plagiarism
Four real-world patterns. Reveal the verdict, then the fix.
Example 1: Verbatim
Source (published, 2019)
“Anaemia in pregnancy remains a major public health challenge in India, contributing substantially to maternal morbidity and adverse perinatal outcomes.”
In the resident’s manuscript
Anaemia in pregnancy remains a major public health challenge in India, contributing substantially to maternal morbidity and adverse perinatal outcomes.
Verdict: Plagiarism. Word-for-word, no quotation marks, no citation. The similarity checker flags 100% of this sentence.
Fix: paraphrase and cite
In India, anaemia during pregnancy is a leading contributor to maternal illness and poor newborn outcomes.[12]
Example 2: Mosaic / patchwork
Two sources
A: “…iron deficiency accounts for nearly half of all anaemia cases among pregnant women in low-income settings.”
B: “…weekly iron–folic acid supplementation has been shown to reduce the prevalence of moderate anaemia.”
In the manuscript (one sentence, no marks)
Iron deficiency accounts for nearly half of all anaemia cases among pregnant women, and weekly iron–folic acid supplementation has been shown to reduce the prevalence of moderate anaemia.
Verdict: Plagiarism. Distinctive phrases lifted verbatim from two sources and stitched together. Adding the two citations does not fix it; the wording is still copied.
Fix: synthesise in your own words and cite both
Roughly half of anaemia in pregnant women is attributable to iron deficiency,[5] and weekly iron–folic acid dosing can lower moderate-anaemia rates.[8]
Example 3: Close paraphrase
Source
“Delayed initiation of antenatal care is strongly associated with poor pregnancy outcomes, particularly among women in rural and tribal communities.”
In the manuscript (cited!)
Late starting of antenatal care is highly linked with bad pregnancy results, especially among women in rural and tribal populations.[9]
Verdict: Plagiarism. Same sentence skeleton, synonyms swapped in slots. A citation does not rescue it. This is the classic “synonym-swap” failure, and modern checkers catch the structure.
Fix: rewrite from understanding
Women who book antenatal care late tend to have worse outcomes, and this gap is widest in rural and tribal areas.[9] (Read it, look away, write it, then cite.)
Example 4: Self-plagiarism
Your OWN earlier paper (2023)
“We conducted a community-based cross-sectional survey across 12 sub-centres using a two-stage cluster sampling design; households were selected by probability proportional to size.”
Pasted into the new manuscript, unchanged, no note
We conducted a community-based cross-sectional survey across 12 sub-centres using a two-stage cluster sampling design; households were selected by probability proportional to size.
Verdict: Self-plagiarism. Reusing your own published text without disclosure. The checker matches it against your prior paper, and editors treat it as a red flag.
Fix: rewrite and signpost the prior work
The sampling and survey procedures followed those of our earlier study.[3] Briefly, households across 12 sub-centres were enrolled using two-stage cluster sampling with probability proportional to size.
The Indian rulebook: UGC 2018
UGC (Promotion of Academic Integrity & Prevention of Plagiarism in HEIs) Regulations, 2018 set out a graded penalty system based on your similarity index:
| Level | Similarity | Consequence (student / researcher; illustrative) |
|---|---|---|
| 0 | ≤ 10% | Minor, no penalty |
| 1 | > 10–40% | Revise & resubmit (usually within 6 months) |
| 2 | > 40–60% | Debarred from resubmission for ~1 year |
| 3 | > 60% | Registration / programme can be cancelled |
Your thesis submission requires a similarity report (Turnitin / DrillBit–ShodhShuddhi). Journals run a second, independent check at submission (iThenticate / Crossref Similarity Check).
Three myths that get residents into trouble
Myth: “Under 10% means I’m safe.” 9% lifted verbatim from one source is still plagiarism. The index is a screen, not an absolution.
Myth: “My references and quotes count against me.” A properly configured report excludes the bibliography and correctly quoted matter. Learn to read, and to set up, the report.
Myth: “My guide is responsible, not me.” The UGC penalises the author. Supervisors carry separate, additional liability; it does not transfer off you.
The toolkit: how to NOT plagiarise
- Paraphrase by understanding: read, look away, write, then cite. Never synonym-swap with the source open.
- Quote sparingly: only for distinctive wording, definitions, contested claims, with quotation marks and citation.
- Cite ideas, not just words: every borrowed claim gets a citation, even when fully reworded.
- “Common-knowledge” test: would any expert state this without a source? If unsure, cite.
- Methods grey zone: rewrite, cite your prior work, disclose: “as previously described [ref].”
- Workflow hygiene: reference manager (Zotero) from day one; tag verbatim notes instantly; run your own similarity check before you submit.
The AI frontier, 2026
LLMs (ChatGPT and others) now both write and get caught in manuscripts. Journals flag tell-tale phrasing and fabricated references.
Where policy stands (and it’s moving):
- UGC has not issued definitive guidance yet.
- AICTE restricts unacknowledged AI use.
- Most institutions & journals now treat undisclosed AI-generated text as misconduct.
- ICMJE / COPE: disclose AI use; AI cannot be an author.
AI may assist (grammar, structure) with disclosure. It must never generate claims, data, or citations you haven’t verified.
Three rules to leave with
Attribute everything borrowed.
Verify every reference.
Disclose every tool.
Thank you · Questions
Quick references for your desk:
- UGC Regulations 2018, your institutional plagiarism policy & similarity-report SOP
- COPE flowcharts, what editors do when misconduct is suspected
- ICMJE Recommendations, authorship & AI-disclosure criteria
Run your own similarity check before submission, not after rejection.