01 · The Question
How Close Is Too Close When You Paraphrase Someone Else's Writing?
You have cited the source. You have changed several words. Perhaps you have rearranged the sentence as well. Yet when you place your version beside the original, they still look remarkably similar.
Researchers often want a numerical boundary: five consecutive words, seven words, ten words, or perhaps a particular similarity percentage. Scholarly writing does not provide such a universal threshold. Determining whether wording is improperly borrowed requires looking at what was taken and how it is presented, not merely counting matching words.
02 · The Short Answer
There Is No Universal Number of Matching Words
In Brief
No universal rule says that a particular number of identical words, consecutive words, or percentage of textual similarity automatically becomes plagiarism. Wording becomes problematic when borrowed language is sufficiently substantial or distinctive that presenting it as your own misrepresents who authored the expression, especially when it is not properly identified and attributed.
Minor changes such as replacing synonyms, changing tense, switching active to passive voice, or rearranging phrases may still leave a passage too close to its source. At the same time, some identical technical or conventional phrases may legitimately occur across research papers.
03 · What You Need to Know
Why Plagiarism Cannot Be Reduced to a Word Count
There is no five-word, seven-word, or ten-word rule
A convenient numerical rule would make plagiarism easy to police, but it would also produce absurd results. A short, highly distinctive expression can carry more authorial originality than a longer conventional phrase. Meanwhile, researchers working in the same technical field may legitimately use identical terminology because there are only so many accurate ways to name a procedure, instrument, statistical test, or established phenomenon.
ORI's educational guidance explicitly explains why an operational rule based on a fixed number of consecutive borrowed words is impractical. It notes that stock phrases and highly technical descriptions can occur legitimately and that context matters when evaluating textual borrowing.
Substantial unattributed copying is more important than a magic number
ORI's working policy describes textual plagiarism in terms of substantial unattributed verbatim or nearly verbatim copying that materially misleads an ordinary reader about the author's contribution. The policy also notes that ORI generally does not pursue limited identical or nearly identical phrases describing commonly used methodology or previous research when such use is not substantially misleading or significant.
This does not create a license to copy methodological prose. It illustrates why the evaluation is qualitative as well as quantitative. The same number of matching words can have very different significance depending on what those words are and where they occur.
Changing individual words may leave the underlying sentence essentially copied
One common strategy is to keep the original sentence and replace enough words to make it look different. Another is to change the grammatical voice, move clauses around, or substitute synonyms while preserving the source's sequence and structure.
ORI describes such cosmetic modification as inappropriate paraphrasing. Its guidance specifically warns that changing one or two words, rearranging order, changing active to passive voice, or altering tense does not by itself create a proper paraphrase.
This kind of writing is often described as patchwriting or mosaic plagiarism when borrowed pieces of language remain woven into apparently new prose. The precise terminology varies, but the underlying concern is straightforward: the new passage remains excessively dependent on another author's expression.
A citation does not automatically make close copying acceptable
Researchers sometimes assume that once a citation appears at the end of the sentence, textual similarity no longer matters. That confuses attribution of ideas with identification of borrowed language.
ORI's examples of inappropriate paraphrasing demonstrate that a passage can credit the original source yet remain problematic because too much of the source's wording and structure has been preserved. If the words themselves are being reproduced, scholarly conventions may require quotation or another clear method of identifying them as borrowed text.
This is why citing a paraphrased source and paraphrasing it adequately are related but separate responsibilities.
Proper attribution
Tells the reader whose idea, evidence, or other contribution you are using.
Independent paraphrasing
Makes clear that the prose itself is your formulation rather than lightly modified language from the source.
Distinctive language deserves more caution than conventional language
Not all matching language carries the same evidentiary weight. A conventional phrase such as the name of a statistical procedure cannot sensibly be paraphrased merely to avoid matching another paper. Nor should researchers distort technical terminology to manufacture superficial originality.
By contrast, an unusual metaphor, distinctive explanation, memorable formulation, or idiosyncratic sequence of phrases may be strongly associated with a particular author. Reproducing such language can be consequential even when the borrowed passage is short.
The question is therefore not simply how many words match. Ask whether the matching language reflects unavoidable disciplinary terminology or another author's distinctive expression.
Similarity percentages do not provide the missing threshold
Text-matching systems can calculate overlap, but a similarity percentage is not a plagiarism probability. A report may contain legitimate matches from references, standard terminology, quoted material, methodological language, or previously disclosed material. It can also miss plagiarism that involves an idea rather than matching words .
Accordingly, there is no universal similarity percentage at which a paper suddenly becomes plagiarized. The matched passages have to be examined individually and in context.
Watch Out
A target such as "keep the similarity below 15%" can encourage the wrong behavior. Researchers may start rewriting merely to defeat a detector rather than ensuring that sources are accurately represented, genuinely paraphrased, and appropriately attributed.
Technical language creates genuine borderline cases
Some material is difficult to paraphrase without losing precision. ORI specifically discusses highly technical language as an area where close phrasing can arise because terminology and accurate descriptions may constrain the available wording.
That does not mean technical writing is exempt from plagiarism standards. Rather, it means mechanical rules become particularly unreliable. If distinctive source language truly needs to be preserved, quotation and citation may sometimes be clearer, subject to disciplinary conventions. If the language is conventional and unavoidable, similarity alone may tell you very little.
Close wording and idea plagiarism are different questions
You can move far away from the original wording and still fail to acknowledge its intellectual contribution. Conversely, two passages may share conventional terminology without one author stealing the other's idea.
The broader definition of plagiarism in research therefore cannot be reduced to textual resemblance. Wording is one dimension of attribution, not the whole concept.
04 · A Practical Example
Why Counting Changed Words Does Not Tell You Whether a Paraphrase Is Safe
Hypothetical Example
One source, three attempted paraphrases
Imagine that a source contains this hypothetical sentence: "Frequent low-stakes feedback allowed participants to identify misconceptions before those misconceptions became embedded in later tasks."
Version A: Cosmetic substitution
"Regular low-stakes feedback enabled participants to recognize misconceptions before those misconceptions became established in subsequent tasks." The writer has changed several words, but the sequence, structure, and much of the expression remain closely tied to the hypothetical original.
Version B: Rearrangement
"Before misconceptions became embedded in later tasks, participants could identify them through frequent low-stakes feedback." More has moved, but much of the original phrasing and conceptual sequence remains visible. Simply reversing the sentence does not establish that it is an independent paraphrase.
Version C: Independent restatement with attribution
"Providing repeated opportunities for formative feedback may help learners detect misunderstandings early enough to address them before subsequent activities build on those errors (Author, Year)." Here the writer has reconstructed the point rather than editing the original sentence word by word, while still acknowledging its intellectual source.
The sentence and source are hypothetical. There is no useful calculation such as "Version C changed 73% of the words, therefore it is acceptable." The relevant differences involve dependence on the original expression, accurate preservation of meaning, and appropriate attribution.
06 · What This Means for You
Stop Counting Words and Examine How the Passage Was Written
If you are asking whether you changed "enough words," that may itself be a useful warning sign. Strong paraphrasing usually begins with understanding the source and reconstructing its relevant meaning for your own argument, rather than editing the original sentence until a detector stops highlighting it.
A simple decision framework
If you need the author's exact distinctive wording
Use an appropriate quotation or other discipline-accepted method of identifying borrowed wording, together with attribution.
If you need the idea but not the original wording
Understand the point first, reconstruct it independently in the context of your argument, and cite its source.
If your version follows the source sentence clause by clause
Compare the passages carefully. Replacing synonyms or rearranging clauses may not have produced a genuine paraphrase.
If the matching language consists of unavoidable technical terminology
Preserve accuracy rather than inventing awkward synonyms, while ensuring that any attributable intellectual contribution is properly credited.
If you are relying on a similarity percentage to decide
Inspect the actual matched passages and their contexts instead of applying a universal cutoff.
A practical technique is to read the source until you understand it, set it aside, write the relevant point for the purpose of your own argument, and then return to the original to check both accuracy and accidental linguistic dependence. This does not guarantee an acceptable paraphrase, but it is generally more defensible than rewriting with the source sentence continuously visible.
07 · A Quick Checklist
Check Whether Your Paraphrase Is Still Too Close
Before keeping a close paraphrase, check:
Did I reconstruct the passage from my understanding, or merely edit the source sentence word by word?
Have I preserved distinctive phrases that should instead be identified as borrowed wording?
Does my sentence follow substantially the same syntax, clause order, and rhetorical structure as the source?
Are the matching words genuinely unavoidable technical terminology or conventional expressions?
Have I cited the source of the idea even if my wording is now completely independent?
If I used text-matching software, have I examined each consequential match rather than relying on the overall percentage?
Would a reasonable reader understand which language and intellectual contribution came from the source and which came from me?
09 · The Bottom Line
There Is No Numerical Boundary Between Paraphrasing and Plagiarism
The Bottom Line
No universal number of matching words or similarity percentage determines when wording becomes plagiarism; the real issue is whether borrowed language is presented in a way that misrepresents its authorship or intellectual source.
Do not paraphrase by counting how many words you have changed. Understand the source, formulate the relevant point independently, preserve necessary technical terminology, attribute the underlying contribution, and clearly identify distinctive wording when you need to reproduce it.
11 · Cite this Guide
How to Cite This Guide
This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.
Recommended (Field Guide)
APA
MLA
Chicago
Copy Citation
