I’ve covered dozens of studies on this blog about how AI complicates assessment, about detection tools that don’t work, about the need to redesign assignments from the ground up. But a 1993 paper by Rebecca Moore Howard, “A Plagiarism Pentimento,” might be the most important thing I’ve read on this subject, and it predates generative AI by three decades.
Howard offered a concept, patchwriting, that explains what students are doing when they borrow language from sources, and she argued it’s a composing strategy, not a crime.
This paper was published long before ChatGPT, before Turnitin, before the entire academic integrity industry existed. And yet it reads like a direct challenge to every AI detection debate we’re having in 2026. If we misunderstood what students were doing with source texts in 1993, we’re making the same mistake now with AI-generated text, just at a larger scale.
What Howard Found at Colgate
The origin story is specific and grounded. In 1986, Howard assigned first-year students at Colgate University to read Davidson’s Genesis 1-11 and write about it. Nine of twenty-six students submitted papers that borrowed Davidson’s language extensively, swapping synonyms, deleting phrases, tweaking grammar, but keeping the structure and much of the wording intact. Howard gave them all F’s and delivered a lecture on source attribution, the standard response most of us would default to.
But one-third of a class at a competitive liberal-arts college had all done the same thing. That nagged at her. Howard describes a slow intellectual reckoning: she moved from a judicio-moral reading (these students cheated) to a cognitivist one (they lack the skills), and finally to a social constructionist interpretation. The paper calls this shift a “plagiarism pentimento,” borrowing from Lillian Hellman’s metaphor of old paint layers showing through new ones. Each framework didn’t replace the last; it layered over it.
What she landed on is the concept that now carries her name in composition studies. Howard calls what her students did “patchwriting”: copying from a source and making surface-level changes. Deletion was the most frequent move (22 instances), followed by synonym substitution (18 instances), grammar and syntax changes (11), and rearrangement (just once). The students weren’t copying verbatim. They were working on the language. The work just happened at the word and phrase level, not at the level of ideas.
Patchwriting and Plagiarism
Howard draws on research from Brown, Day, and Jones, who studied how students learn to summarize. Their findings show a “copy-delete” strategy common among fifth and sixth graders: read the text, decide what to keep, and reproduce selected units close to verbatim.
Mature summary, the kind that involves genuine rearrangement and recasting of ideas, turns out to be a late-developing skill. Brown, Day, and Jones found that even college students were still refining it. Carol Sherrard’s research confirmed this: all ten undergraduates in her study resorted to copy-deletion when working with expository texts.
Howard connects these cognitive findings to her Colgate classroom. Her students weren’t lazy or dishonest. They were facing unfamiliar academic discourse and doing what developmentally makes sense: patching together the source’s language as a way to participate in a conversation they didn’t yet have the vocabulary to join on their own terms.
I find this argument deeply persuasive, and I’ve seen it confirmed over and over in my own work. I wrote about Johnson-Eilola and Selber’s (2007) concept of plagiarism as assemblage, the idea that all writing involves remixing and recombining existing texts, and Howard’s patchwriting concept is the developmental version of the same insight. Students patch because they’re learning, not because they’re stealing.
Why the Plagiarism Rules Don’t Fix the Problem
Howard makes an interesting point here. Teaching citation rules, the mechanics of quoting, paraphrasing, and building a Works Cited page, doesn’t respond to the actual motivation behind patchwriting. Students who patchwrite aren’t failing to cite. Many of Howard’s students did cite their sources. The issue is that they couldn’t yet produce the source’s ideas in their own language. Attribution rules address documentation; they don’t address comprehension.
Howard frames patchwriting as a “membership application.” Students are outsiders trying to gain entry to a discourse community. The rules of plagiarism mark the boundaries of that community and identify who belongs, but they don’t give the outsider a way in.
I’ve covered Eaton’s (2023) postplagiarism framework, which argues that the binary of plagiarism/not-plagiarism has outlived its usefulness, and Howard’s 1993 paper was already saying something similar from a different angle: the category itself is the problem.
This connects to the AI conversation in a way that’s hard to ignore. When students use ChatGPT to generate or polish academic text, are they cheating, or are they doing a technologically updated version of what Howard’s 1986 students did with Davidson’s prose? The answer isn’t always one or the other, and that ambiguity is exactly what Howard’s framework helps us think through.
I wrote about Chan’s (2023) concept of AI-giarism and student perceptions of AI misconduct, and the confusion Chan documented among students maps almost perfectly onto the confusion Howard found among faculty in 1993. Nobody agrees on where the line is because the line was always drawn in the wrong place.
The Pedagogical Response Howard Proposes
Howard’s solution is summary-writing instruction. She has students practice oral collaborative summaries, where the class reconstructs assigned readings from memory on the board while she acts as secretary. The passages they’ve collectively forgotten, the silences, turn out to be exactly the places where patchwriting shows up in written assignments. That exercise makes the difficulty visible and turns it into a shared teaching moment.
She also assigns individual written summaries with a time delay from the original reading, and she teaches the specific skills Brown, Day, and Jones identify for mature summary: selecting, condensing, rearranging, and restating. The goal is to move students through patchwriting toward genuine ownership of new vocabulary and ideas.
I covered Vosen’s (2008) approach to teaching plagiarism through Bloom’s taxonomy, and there’s a strong resonance here. Both Howard and Vosen argue that plagiarism instruction fails when it stays at the knowledge and comprehension levels. Students need to work at the application and synthesis levels, and that requires pedagogy, not policing.
Reading Howard in 2026
This paper is over thirty years old. The academic world it describes, before Turnitin, before the internet made source-finding trivial, before AI could generate entire essays, feels distant. But the core argument hasn’t aged at all. If anything, the stakes are higher now.
We have more tools for producing text and more tools for detecting it, and we’ve invested enormous energy in both sides of that arms race. What we haven’t done is invest seriously in the pedagogical middle ground that Howard was mapping in 1993: teaching students how to work with sources at the level of ideas, not just at the level of citation mechanics.
The technology has changed. The student hasn’t. Patchwriting is still what happens when a learner faces discourse they haven’t yet made their own. The question for educators in 2026 is the same one Howard asked in 1993: will we punish students for the struggle, or will we teach through it?
References
- Chan, C. K. Y. (2023). Is AI changing the rules of academic misconduct? An in-depth look at students’ perceptions of ‘AI-giarism.’ _International Journal of Educational Technology in Higher Education, 20_(1), 60. https://doi.org/10.1186/s41239-023-00408-z
- Eaton, S. E. (2023). Postplagiarism: Transdisciplinary ethics and integrity in the age of artificial intelligence and neurotechnology. _International Journal for Educational Integrity_, 19(23). https://doi.org/10.1007/s40979-023-00144-1
- Howard, Rebecca Moore. (1993). A plagiarism pentimento. Journal of Teaching Writing, 11, 233–246.
- Johnson-Eilola, J., & Selber, S. A. (2007). Plagiarism, originality, assemblage. Computers and Composition, 24(4), 375–403. https://doi.org/10.1016/j.compcom.2007.08.003
- Vosen, M. A. (2008). Using Bloom’s taxonomy to teach students about plagiarism. The English Journal, 97(6), 43-46. https://www.jstor.org/stable/40503410