Barthes, the Death of the Author, and the Prompt

The first English publication of Roland Barthes's most quoted essay was not in a book. It was in a box. In 1967 the Irish artist and critic Brian O'Doherty edited a double issue of the magazine Aspen, numbered 5 and 6, which arrived as a small white box containing, in the Irish Times's inventory of it, "a book, four films, five records, eight cut-out boards and 10 printed texts". Contributors included Susan Sontag, John Cage, Robert Rauschenberg, Sol LeWitt, Marcel Duchamp and Samuel Beckett. The paper notes that these, "and Barthes's seminal essay The Death of the Author, were all first published here".

An essay arguing that a text has no single origin, issued as a set of loose parts in a box with no fixed reading order, by an editor who is quoted saying of Barthes that "he got it immediately", and connecting the essay to "my notion that art, writing etc was... a kind of anti-self". The form was the argument. It is a pity that almost everyone who cites the essay encountered it as a flat PDF.

A correction about the date

The essay is habitually cited as 1968. That is the date of the French publication. John Logie made the case for the earlier date in College English in May 2013, in an article titled "1967: The Birth of 'The Death of the Author'", volume 75, issue 5, pages 493–512. His argument is that the initial English publication occurred in 1967 rather than the commonly cited 1968, and that the essay's frequently invoked revolutionary character derives from an artistic rather than a political revolution, connected to multimedia experimentation that predates digital technology by decades.

The second half is the more interesting correction. The essay is routinely enlisted as a document of 1968, of the barricades. Logie places its English birth a year earlier, inside an avant-garde publishing experiment about media and form. Read that way it is a text about how works circulate, not a manifesto about who is allowed to make them.

What the essay actually argues

It opens with a specific sentence in a specific novella. In Richard Howard's translation, the one printed in the Aspen box, the first line is: "In his story Sarrasine, Balzac, speaking of a castrato disguised as a woman, writes this sentence..." Barthes then asks who is speaking that sentence. Balzac the man? Balzac the novelist professing literary ideas about femininity? Universal wisdom? Romantic psychology? His answer is that we cannot know, and that this is not a gap in our information. It is the condition of writing.

From that he draws the distinction the essay is actually built on, and which gets lost every time the title is quoted without the body. There is the Author, capitalised, a figure who precedes the text, nourishes it, stands to it as parent to child, and whose biography is supposed to explain it. And there is what Howard renders as the scriptor: "the modern writer (scriptor) is born simultaneously with his text; he is in no way supplied with a being".

The scriptor does not express an interior. The scriptor assembles. "The text is a tissue of citations, resulting from the thousand sources of culture." That sentence is the engine of the whole piece, and note what it does not say. It does not say the citations are unattributable, or that their sources do not exist, or that drawing on them is costless. It says a text is made of them.

The conclusion follows from the premise about reading, not about writing. In Stephen Heath's later translation, collected in Image-Music-Text (1977), the key sentence runs: "to give a text an Author is to impose a limit on that text, to furnish it with a final signified, to close the writing". The Author is a device for shutting down interpretation. Remove the device and interpretation opens.

Two famous last lines

Here is a detail worth having, because the essay's most quoted sentence exists in two materially different English forms and people quote them interchangeably.

Howard, 1967, in the Aspen box: "the birth of the reader must be ransomed by the death of the Author."

Heath, 1977, in Image-Music-Text: "the birth of the reader must be at the cost of the death of the Author."

Heath's is the version everyone knows. Howard's "ransomed" is stranger and, I think, better: a ransom is a price paid to recover something held captive, which is closer to what Barthes is describing than a neutral cost. Neither is the French.

An essay about the impossibility of a single origin has, in English, at least two of them.

What it is and is not a claim about

Almost every misuse of the essay comes from reading it as a claim about production when it is a claim about reading.

The fourth row is the one that matters for anything made by writing a prompt. Barthes is not opening a door, he is dismantling a room. A reading that arrives at "so I am the author after all, by a different route" has inverted it.

Foucault asked the better question

Two years later Michel Foucault gave a lecture to the Société française de philosophie, on 22 February 1969, published in the Bulletin de la Société française de philosophie, 63e année, number 3, pages 73–104. He delivered a modified version at the State University of New York at Buffalo in March 1970.

The textual situation is messier than the citations suggest, and Stuart Elden has documented it. There are three English translations: Josué Harari's in Textual Strategies (1979), pages 141–160; Donald Bouchard and Sherry Simon's in Language, Counter-Memory, Practice (1977), pages 113–138; and an earlier abridged one by Venit in Partisan Review, volume 42 number 4 (1975), pages 603–614, which omits the opening and the discussion. Harari's circulated most widely, through The Foucault Reader and the Essential Works, and later scholarship suggests it may be a hybrid rather than a faithful record of the Buffalo version. The discussion that followed the Paris lecture remains largely untranslated into English. Anyone quoting Foucault on the author is quoting one of several imperfectly reconciled texts.

Foucault opens by quoting Beckett: "What matter who's speaking, someone said, what matter who's speaking." He then does what Barthes did not. Rather than announcing that the author is gone, he asks what job the author was doing, and calls that job the author-function. His analysis gives it four characteristics: texts become objects of appropriation through legal codification, which ties authorship to property and to the possibility of transgression; the function is historically variable, so that scientific texts in the Middle Ages required attribution for credibility while literary works came to demand it only later; authors are not simply attributed to texts but constructed through a complex operation of comparison, extraction of traits and assignment of continuity; and the function "gives rise to a variety of egos and to a series of subjective positions" rather than to one unified consciousness. He also separates ordinary authors from initiators of discursive practices, his examples being Marx and Freud, who founded the possibility of discourse rather than producing a body of work within one.

The Wikipedia entry on the lecture preserves two formulations that carry the argument: "The Author is a certain functional principle by which, in our culture, one limits, excludes and chooses", and "The author is therefore the ideological figure by which one marks the manner in which we fear the proliferation of meaning." He closes by imagining a culture of pervasive anonymity in which the only question left is Beckett's: "What matter who's speaking?"

The reason Foucault is the more useful text is that a function can be examined empirically. You can ask who currently performs it, under what legal regime, with what consequences for money and blame. Barthes gives you a beautiful negation. Foucault gives you a question you can put to a contract.

Reading both against the prompt

So: work produced by describing it in language to a model. Where does the essay genuinely apply?

It applies to the output's meaning. A generated image has no interior to consult, and no biography behind it that would settle a disputed reading. Barthes's description of the scriptor, born simultaneously with the text and in no way supplied with a being, is an unnervingly exact description of what a model is at the moment of generation. If ever there were a text whose Author cannot be produced for questioning, this is it.

It applies to the tissue of citations. A model's output is, by construction, assembled from the thousand sources of its training corpus. The essay's description of how texts are made turns out to be a literal description of how these particular texts are made, which is either a vindication of Barthes or a coincidence so large it should make us suspicious of the vindication.

It applies to interpretation. Nobody sensible reads a generated image for what its maker meant, because the question does not parse. The Author as a device for closing interpretation has genuinely been removed here, which is what Barthes wanted.

Now the part that does not apply, and where I see the essay being used as cover.

The essay says nothing about labour. It is a piece of literary theory about where meaning is located in the act of reading. It has no position on who made the corpus, whether they consented, whether they were paid, or who should be named. Those are questions about production and distribution, and the essay does not address them because they are not its subject. Citing it as though it settled them is a category error dressed up as erudition.

Nor does the essay transfer authorship to the person at the keyboard. The common move runs: the Author is dead, therefore the old rules about credit do not apply, therefore whoever wrote the prompt is the maker. The first clause is Barthes, the second does not follow, the third contradicts him. He killed the Author. He did not hand out the estate.

The question the essay never addressed, and someone had to

Which brings us to where the author-function is currently being decided, in the consequential sense Foucault meant when he tied authorship to legal codification and property.

In January 2025 the United States Copyright Office published Copyright and Artificial Intelligence, Part 2: Copyrightability, a report of the Register of Copyrights. Its conclusion on prompts is direct: "Given current generally available technology, prompts alone do not provide sufficient human control to make users of an AI system the authors of the output." The reasoning is that "prompts essentially function as instructions that convey unprotectible ideas". The Office is not saying prompting is worthless. It is saying that instructions are ideas, and ideas have never been the thing copyright protects.

What the report does protect is stated in the same document: "Copyright protects the original expression in a work created by a human author, even if the work also includes AI-generated material", and "human authors are entitled to copyright in their works of authorship that are perceptible in AI-generated outputs, as well as the creative selection, coordination, or arrangement of material in the outputs, or creative modifications of the outputs."

Set that next to Foucault's first characteristic of the author-function, that texts become objects of appropriation through legal codification, and the two documents are doing the same work fifty-six years apart. The Office is drawing the boundary of a function: what has to be true of a human contribution before the legal system will attach a name and a property right to it. Foucault's point was that this boundary has always been drawn, that it has moved before, and that its movements tell you more about a culture than any theory of genius does.

Note also how narrow the protected zone is, and how it maps onto craft rather than intention. Selection, coordination, arrangement, modification. Those are editing verbs. The function has not been abolished. It has been relocated from the moment of conception to the moment of choosing, which is roughly where a working designer's contribution has always actually sat, and which Barthes would probably have found funny.

What I take from this

I have used the essay lazily. Not in print, but in conversation, in exactly the way described above: as a licence, as a way of making a question about credit sound naive. That is a misuse, and the essay is good enough that misusing it is a waste.

The honest position is that Barthes gives us one true and limited thing about prompt-made work, which is that there is no interior to consult and no biography that will settle a reading. Everything else that people want from the essay, permission, absolution, a theory of why attribution is unnecessary, is not in it and was never in it.

For the rest, Foucault's question is the one to carry: not whether the author is dead, but what the author-function is currently doing, who performs it, and who benefits from the arrangement. In 2025 part of that answer was written down by a government office in a report about prompts. That is not a philosophical defeat. It is exactly what Foucault predicted the answer would look like.

Sources

• Roland Barthes, "The Death of the Author", translated by Richard Howard, in Aspen no. 5+6, edited by Brian O'Doherty, 1967. Text hosted by UbuWeb. Source of the Howard renderings quoted here, including "the birth of the reader must be ransomed by the death of the Author". ubu.com/aspen/aspen5and6/threeEssays.html

• Roland Barthes, "The Death of the Author", in Image-Music-Text, translated by Stephen Heath, Hill and Wang, 1977, pages 142–148. Source of "to give a text an Author is to impose a limit on that text, to furnish it with a final signified, to close the writing" and "the birth of the reader must be at the cost of the death of the Author", as reproduced by New Learning Online. newlearningonline.com/literacies/chapter-8/roland-barthes-on-the-death-of-the-author

• John Logie, "1967: The Birth of 'The Death of the Author'", College English, volume 75, issue 5, May 2013, pages 493–512, ISSN 0010-0994. Argues for 1967 rather than the commonly cited 1968 as the date of first English publication, and locates the essay's radicalism in artistic rather than political revolution. ERIC accession number EJ1008739, eric.ed.gov

• "Modern Ireland in 100 Artworks: 1967, Aspen 5+6, by Brian O'Doherty", The Irish Times. Description of the white box and its contents, the contributor list, and O'Doherty's remarks about Barthes. irishtimes.com/culture/modern-ireland-in-100-artworks-1967-aspen-5-6-by-brian-o-doherty-1.2419975

• Michel Foucault, "What is an Author?", lecture to the Société française de philosophie, 22 February 1969. English text as distributed by The Open University. Source of the Beckett epigraph, the four characteristics of the author-function, the initiators of discursive practices, and the closing question. open.edu/openlearn/ocw/pluginfile.php/624849/mod_resource/content/1/a840_1_michel_foucault.pdf

• Stuart Elden, "The textual issues around Foucault's 'What is an Author?'", Progressive Geographies, 11 April 2017. Publication details: Bulletin de la Société française de philosophie, 63e année, no. 3, pages 73–104; the Buffalo version of March 1970; and the three English translations by Venit (1975), Bouchard and Simon (1977) and Harari (1979). progressivegeographies.com/2017/04/11/the-textual-issues-around-foucaults-what-is-an-author/

• "What Is an Author?", Wikipedia. Source of the two Foucault formulations quoted: the Author as "a certain functional principle by which, in our culture, one limits, excludes and chooses", and as "the ideological figure by which one marks the manner in which we fear the proliferation of meaning". en.wikipedia.org/wiki/What_Is_an_Author%3F

• United States Copyright Office, Copyright and Artificial Intelligence, Part 2: Copyrightability, Report of the Register of Copyrights, January 2025. Source of the conclusion that prompts alone do not provide sufficient human control, and of the statements on protectable human contribution in the executive summary. copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-2-Copyrightability-Report.pdf

Related reading

• Aura, and the Generated Image

• The Contrast Formula, Its Known Flaws, and What Comes Next

• Flusser: The Photographer Inside the Programme

• Between Noise and Silence

More from the Moc Mallen Studio journal