Wang was eighteen. He wanted to be a clarinettist. He went to the music library at his university, pulled a Beethoven string quartet from the shelf, and pulled a Mahler symphony from the shelf. He sat down with blank sheet music. He started copying.1
Not playing. Not analyzing. Not reading-and-thinking-about. Copying. Note by note, instrument by instrument, every articulation marking, every dynamic change, every rest. The way an apprentice in a Renaissance workshop copied a master's drawing — not to produce a saleable artifact but to absorb the underlying decisions through the body's own labor.
Several years later, no longer pursuing music professionally, Wang was trying to figure out how to write better. The instinct most writers have at this point is to read more carefully, study craft books, take a class. Wang did something else. He pulled an article from The New Yorker that he liked. He sat down with a blank page. He started typing the article out, sentence by sentence, exactly as printed.1
This is the technique. Not metaphorical copying. Not "internalizing" or "absorbing the rhythm." Literally retyping the sentences as if you were transcribing them, while paying full attention to what is happening on the page in front of you.
It sounds, as Wang himself admits, a bit like a silly exercise.1 It is not. It is the most direct way available to train the specific cognition that produces voice.
Here is what is actually happening when you copy a Beethoven string quartet by hand.
You write the first note. You see it is a quarter note. The next note is a quarter note also, but pitched a third lower. The next is a half note that ties into the next bar. By the time you have written six notes, your hand has executed a melodic decision that you, as the copier, did not make. You are enacting a decision Beethoven made. The hand-eye-page coordination is forcing your nervous system through Beethoven's actual sequence of choices in slow motion.
You keep going. Four more bars. The viola part starts. You realize the viola has been silent for the last six bars. You write its first note. Why did Beethoven hold the viola back? The question would not have surfaced if you had only listened to the piece. The act of inscribing the viola's silence — having to leave the staff blank for six bars and then put a note on it — makes the silence visible as a decision rather than as a default.
You continue. The second violin enters. You write its part. Why this particular interval against the cello? Why not a third higher? Why an offbeat entry rather than on the downbeat? Each question is a place where Beethoven could have done something different and did not. The copying process surfaces those choice-points one by one, in the order they were originally made, and forces your attention onto them.
Wang's word for this is exact: you really get into the mind of a composer.1 Not metaphorically. The copying process is a partial reconstruction of the composer's actual decision-sequence, and your nervous system is being trained on that sequence by the labor of inscribing it.
Sentence-level copying does the same thing for prose. When you retype a New Yorker sentence by sentence, the same surfacing of decisions happens.
You type: "It was the kind of evening when even the small things felt enormous." You notice you typed kind rather than type. You notice the rhythm of small things felt enormous — the long-vowel close at the end versus the short-vowel opening. You wonder why the sentence is fifteen words and not eight or thirty. Could this have been three short sentences instead? The answer is partly aesthetic and partly structural — short sentences would have removed the unspooling quality that the long sentence is doing — but the question would not have arisen if you had merely read the sentence. Reading is too fast for the questions to surface.
Now you type the next sentence. The next sentence begins with He. You realize the previous sentence had no human subject — it was about an evening — and the writer has chosen to ground the next sentence in a person rather than continue with atmosphere. Why this transition rather than another? You type the rest of the paragraph. By the time you have copied a full paragraph, you have noticed, in your hand and in your eye, twenty or thirty decisions the writer made that the writer's reader was supposed to glide past without consciously registering. The writer's invisible labor has become visible to you.
Wang's claim about what this produces: you start thinking about the trade-offs they make. You start thinking about the choices that they make. You start thinking, at this point maybe I could have used a different adjective. At this other point, maybe my sentence should have flown in a different direction. And then you really start understanding their intentionality, their choices, and what exactly that they really tried to do.1
The phrase create intentionality within yourself is the destination. The exercise is not about imitating any particular writer; it is about building the cognitive habit of treating your own sentences as decisions rather than as defaults. Once you have been forced to see another writer's sentences as a sequence of choices, you cannot unsee your own sentences the same way. Voice emerges from this — not from finding it inside, but from the practiced habit of intentionality you have trained by copying other people's intentionality first.
Wang adds a quick caveat. Maybe The New Yorker is not the pinnacle of writing. Maybe we should be looking through different authors instead.1 He is right to flag this. The technique is method-neutral; the quality of the trained intuition depends on the quality of the source.
A writer who copies New Yorker articles will absorb New Yorker intuitions. The intonations of high-end magazine prose. A specific sentence-length distribution. A specific kind of irony. A specific posture toward the reader. None of these is bad. But none of them is universal — and a writer trained exclusively on this corpus will produce sentences that are recognizably of-this-corpus, regardless of subject matter.
The implication is that the technique amplifies rather than corrects the source. If you copy mediocre writing, you will train mediocre intuitions. If you copy from a single tradition, you will inherit that tradition's blind spots. The exercise produces internalized intentionality, but the kind of intentionality depends on whose sentences you put your hand through. The choice of source is therefore the strategic decision; the copying is the execution.
For most writers, the practical recommendation is: choose two or three writers who solve different problems, and copy from each of them. Not in series. In parallel. A passage from each, alternating, over weeks. The mind that emerges from copying Stendhal and Hitchens and Joan Didion in rotation will not be any one of them — and will not have any one of them's blind spots.
The reason the technique sounds silly is that it is easily confused with imitation, and imitation is widely (and rightly) discouraged. Don't write like other writers; find your own voice. That advice is descriptively wrong, but the alternative — imitate other writers — is descriptively even more wrong.
Manuscript copying is neither. It is internalization through embodied labor. The output is not a Wang sentence imitating Stendhal; the output is a Wang sentence written by a writer whose nervous system has been trained on Stendhal's intentionality patterns. The reader of the Wang sentence does not see Stendhal. They see Wang. But Wang's choices — the choices that aggregate into what readers will recognize as his voice — were trained by the copying. The technique is one or two layers below imitation, at the level where decisions are formed before they are made.
This is why copying once is useless and copying often is transformative. A single copied paragraph trains nothing durable. A copied paragraph a day for six months reshapes the way you make sentences. The threshold is in the body, not in the mind, and bodies require repetition.
The setup
The exercise
The practice
What changes
The discipline
1 Wang's account of the technique runs from lines 760-774. The Beethoven and Mahler copying is at lines 762-766. The transfer to New Yorker articles is at lines 766-768. The "silly exercise" admission and "pinnacle of writing" caveat are at line 768. The "into the mind of a writer" claim and the trade-offs/choices/intentionality unpacking are at lines 770-774. The destination phrase — create intentionality within yourself such that writing does not feel like a totally strange and random process — is at line 774.
The technique pulls against the Romantic mythology of voice-as-discovery that still dominates much American craft pedagogy. Find your own voice is the hegemonic instruction, and it implies that voice is internal and waiting to be uncovered. The copying technique implies the opposite: voice is constructed through deliberate exposure to other writers' decisions, and the writer who tries to find voice without absorbing models will produce idiosyncrasy rather than voice. The two positions are genuinely incompatible. Wang's framing wins on the empirical question — voices that readers experience as distinctive almost always have identifiable structural debts to specific predecessors — but the cultural mythology resists this for reasons that are partly about American individualism and partly about the awkwardness of admitting your sentences are not entirely your own.
The manuscript-copying technique converges with the broader voice-cultivation-through-stylistic-models page already in the vault, but adds a specific embodied-practice layer that the existing page describes more conceptually. Where voice-cultivation describes what Wang does (study Stendhal and Mozart for structural problems they solved), this page describes how Wang does it at the level of daily practice (copying sheet music, copying New Yorker articles). The two pages should be read together: voice-cultivation is the strategy; manuscript-copying is the operation.
The technique has a structural cousin in the apprentice-copies-master tradition that runs through visual art, calligraphy, and traditional craft pedagogy generally. Renaissance art apprentices copied master drawings. Chinese calligraphy students copied classical scripts. Carpentry apprentices reproduced their masters' joints before designing their own. The convergence with Wang's practice is exact: the embodied labor of reproduction trains the nervous system on choices the apprentice would not have made and could not have understood through observation alone. The divergence in Wang's case is one of context — manuscript copying for writers is a private, optional, individual practice rather than an institutional rite. The cost of this is that most writers never adopt it; the benefit is that the writers who do are individually selecting models that suit them rather than inheriting whatever the institution prescribes.
The technique is creative-practice, but the underlying mechanism — embodied reproduction as a way to internalize a master's intentionality — operates across multiple domains in this vault.
Eastern Spirituality: Sadhana Practice Hub — daily ritual practice (japa, mantra repetition, prescribed niyasa sequences) shares the structural feature of embodied reproduction as transformation mechanism. The sadhana practitioner does not perform japa in order to "imitate" the lineage's prior teachers; they perform japa to become the kind of being who has been trained by this exact sequence of repetitions. The convergence with manuscript copying is the underlying claim that embodied repetition reshapes the practitioner in ways that pure cognitive understanding cannot. The handshake produces a generalizable principle: certain kinds of cognitive transformation require physical-temporal practice and cannot be shortcut by understanding the practice. The writer who reads a craft book and skips the copying gets the conceptual content; only the writer who copies gets the cognitive transformation. The bhakta who reads about japa and skips the chanting gets the conceptual content; only the bhakta who chants gets the cognitive transformation.
AI-Collaboration: Transformer Architecture and Language Model Mechanics — the way large language models acquire stylistic capacities through next-token prediction on vast corpora is, at a structural level, an automated version of the manuscript-copying technique. The model is repeatedly forced to predict the next token given the prior context — which is the computational analog of "what would the master have written next?" The convergence reveals a non-obvious symmetry: the practice that trains human voice and the training process that trains LLM voice operate through the same underlying mechanism — repeated exposure to next-decision points within a master's sequence. The divergence is that the LLM has no embodiment, no temporal limitation, and no selective attention; it absorbs the entire training corpus indiscriminately. The handshake produces an actionable claim: a writer who wants to be distinguishable from LLM-generated prose should not train themselves on the same broad corpus the LLM trains on (a generic web corpus or "good general writing"). They should train on a narrow, deliberately chosen corpus that the LLM has averaged into invisibility. Manuscript copying with strategic source selection is the pre-industrial version of fine-tuning, and it produces the kind of voice the broad-trained LLM cannot reproduce because the LLM's training was averaged across too many sources.
Behavioral-Mechanics: Behavioral Mechanics Hub — the internalization-through-embodied-practice mechanism appears in behavioral-mechanics literature as the underlying principle of skill acquisition through deliberate practice (Ericsson) and through state-dependent learning more broadly. The convergence is on the underlying claim that cognitive habits cannot be installed cognitively; they require physical-temporal repetition that recruits the body's learning systems. The handshake reveals a generalizable principle for behavioral transformation: anyone who wants to change a cognitive habit (whether writing voice, decision-making style, conversational pattern, or strategic intuition) needs an embodied-practice protocol that runs through the actual decision-sequence repeatedly, not just an understanding of the desired endpoint.
Psychology: Identity Architecture and Defense Hub — the construction of writerly voice through copying is structurally similar to the construction of any aspect of identity through embodied imitation of admired others. Children acquire posture, intonation, and mannerism through unconscious copying of caregivers; adolescents acquire identity through conscious copying of admired peers; adults acquire professional voice through copying of mentors. The handshake reveals: identity is partly an accumulation of embodied imitations, and the writer's voice is one specialized instance of a much broader phenomenon. The implication is non-obvious — who you copy is who you become — and it cuts both ways. The writer choosing manuscript-copying sources is also, in a small way, choosing future versions of themselves.
The four handshakes converge on a single claim: embodied repetition is the universal mechanism for the transformation of cognitive habits, and the manuscript-copying technique is one application of a much broader principle that operates in spiritual practice, AI training, skill acquisition, and identity formation.
The Sharpest Implication
If voice is constructed through embodied labor on chosen models, then most contemporary writing instruction is operating at the wrong layer. The writing classroom that teaches "show don't tell" and "vary sentence length" and "earn your epiphanies" is teaching the what of good writing without giving students access to the how — the underlying intentionality muscle that makes the rules feel obvious from inside the practice rather than arbitrary from outside it. Students learn to identify violations of the rules in their own drafts but never develop the cognitive capacity to make the choices that the rules are trying to encode. The result is a generation of writers who can tell you what is wrong with a sentence but cannot reliably write a better one. The cure is unfashionable — copy other writers by hand, every day, for months — and the unfashionability is not accidental. The technique is too direct to support a teaching industry. There is no curriculum, no certification, no progressive sequence; there is only the writer alone with a paragraph by Joan Didion and the willingness to retype it for ten minutes a day. The implication for the writer who takes this seriously is that the work of building voice is mostly invisible, mostly solitary, mostly slow, and mostly already done before any class begins. The class can polish what the copying built. It cannot replace the copying.
Generative Questions