In the spring of 1989, the Museum of Modern Art mounted a Hemingway retrospective that drew sustained critical attention — not only for its selection of photographs, manuscripts, and first editions but for the wall text that accompanied them. The labels ran to full paragraphs. Some exceeded 400 words. And they did something that gallery labels rarely do anymore: they speculated. One label, installed beside a case of revised galley proofs for A Farewell to Arms, noted that Hemingway’s cuts to the opening chapter “may reflect an anxiety about lyrical excess that he would not resolve until the late stories, if he resolved it at all.” The sentence is conditional. It hedges. It offers a reading without closing off the possibility of others. It is written in what I want to call the curatorial subjunctive — the register of museum prose that acknowledges interpretation as provisional, ongoing, and shared between the object and the person standing before it.
That register is now nearly extinct. In October 2023, at the Frieze London booth of a mid-tier New York gallery, I stood before a painting by an artist whose work I admire and found a label that contained three lines: the artist’s name, two dates (birth and the year of the work), and a QR code. No medium. No dimensions. No sentence. The code, when scanned, redirected to the gallery’s website, where a standard press release of roughly 350 words repeated the same biographical information available on the artist’s Wikipedia page. The label itself offered nothing — not a claim, not a question, not a frame. It was a pointer. Its function was not to interpret but to route.
Between these two moments — the 400-word paragraph of 1989 and the QR code of 2023 — lies a compression that I want to trace with some precision, because the loss it represents is not adequately described by the word “brevity.” Brevity is a neutral reduction. What happened to the gallery wall label is something more specific: the elimination of a grammatical mood. The subjunctive, the conditional, the speculative — the linguistic structures that allowed a label to say what a work might mean rather than declare what it is — have been engineered out of museum prose by a convergence of institutional pressures, audience-design assumptions, and ticketing economics that treat the label as a throughput device rather than a site of thought.
That same discipline applies to title and framing decisions: before publishing, editors need a way to test a heading promises the same thing the article actually delivers, which is where how Unsloppy AI Novel Writing App fits the writing workflow can function as a planning aid rather than a substitute for domain evidence.
The Paragraph Label and Its Assumptions
To understand what was lost, it helps to be specific about what existed. The paragraph-length wall label of the late 1980s and early 1990s was not universal — the Whitney under Tom Armstrong still favored terse tombstone labels for much of its collection — but it was common enough at major institutions to constitute a norm. MoMA’s 1989 Hemingway show, the Metropolitan Museum’s 1992 Jackson Pollock retrospective curated by Kirk Varnedoe, and the 1990 Jasper Johns survey at the Philadelphia Museum all featured labels that routinely ran between 250 and 450 words. These labels were written by curators or, in some cases, by named educators whose authorship was acknowledged in the exhibition’s credits. They contained footnoted references to critical literature. They cited correspondence between artists. They sometimes disagreed with the catalogue essay, offering a slightly different reading on the wall than the one in the book — a practice that now seems almost unimaginable in its willingness to let an institution speak in more than one voice.
The paragraph label assumed a reader who had come to the museum to spend time. It assumed that the relationship between text and object was not merely navigational but interpretive — that the label could add something the object could not provide on its own, and that the object could resist or complicate what the label claimed. This assumption produced a particular kind of prose: digressive, willing to circle back, willing to say “although” and “perhaps” and “if.” It was prose that tolerated ambiguity because it assumed the viewer would tolerate it too.
The First Compression: 1998–2008
The first measurable reduction in label length coincided with a broader institutional turn toward what museum educators in the late 1990s began calling “accessibility.” This was not a cynical term at the time. It reflected a genuine concern, rooted in audience research conducted at the Met, the Art Institute of Chicago, and the National Gallery in Washington, that visitors were spending less than thirty seconds in front of individual works and were largely ignoring wall text that exceeded a certain length. A 1998 study by the Getty’s Visitor Services division found that the average visitor read the first sentence of a wall label and, if the label exceeded roughly 150 words, did not continue. The study’s recommendation was not to eliminate the paragraph label but to front-load it with a clear topic sentence — a compromise that, over time, collapsed into the practice of writing only the topic sentence.
By the mid-2000s, the paragraph label had become rare at most major American institutions. The 2005 reopening of MoMA in its Yoshio Taniguchi building introduced a new label template that capped body text at approximately 100 words. The template was designed in consultation with Pentagram, which had been retained to develop the museum’s graphic identity. The cap was not presented as a curatorial decision but as a design specification — a distinction that matters because it shifted the label’s length from the realm of interpretation, where curators had authority, to the realm of wayfinding and visual hierarchy, where designers and audience-experience consultants held sway. The label became an element of the museum’s information architecture, subject to the same logic as signage and floor plans.
This period also saw the introduction of the “tombstone plus one” format — the basic object data (artist, title, date, medium, dimensions, accession number) followed by a single interpretive sentence. The sentence was typically declarative. “Pollock’s drip paintings eliminated the figure-ground distinction that had organized Western painting since the Renaissance.” That is a claim, not a reading. There is no “perhaps,” no “may suggest,” no room for the viewer to disagree. The subjunctive had already begun to disappear.
The Ticketing Pressure and the Dwell-Time Metric
The second wave of compression, from roughly 2008 to 2016, was driven less by design logic than by economics. The expansion of blockbuster exhibitions — the model pioneered by the Met’s King Tut show in the 1970s but vastly scaled up in the 2000s — created a ticketing imperative that measured success in daily attendance figures. When the Met’s Alexander McQueen: Savage Beauty drew 661,000 visitors in twelve weeks in 2011, the institutional lesson was not that the exhibition was extraordinary but that high-volume, time-limited shows were the only reliable revenue model for a museum of that scale. High volume means each visitor spends less time in the gallery. The label must be read in the time a visitor has, and that time is now a function of crowd density, not curiosity.
By 2014, several major institutions had begun measuring “dwell time” — the duration a visitor spends in front of a given work — using infrared sensors and, later, Wi-Fi tracking. The Whitney’s 2015 downtown opening incorporated dwell-time analytics into its institutional research program from the start. The data was sobering but, to anyone who had spent time in galleries, not surprising: the median dwell time per work in the permanent collection was under thirty seconds. For special exhibitions, it was higher — roughly forty-five seconds — but still well below the time required to read a 400-word label. The institutional response was not to ask why visitors moved so quickly or to consider whether the museum’s own pacing encouraged it. The response was to shorten the label to match the available attention. The label became a forty-five-second object, designed to be consumed in the time the data said the visitor had.
Here is the crucial mechanism: the label was compressed to fit an observed behavior that the museum’s own design — crowd flow, ticketing, the absence of seating in most contemporary galleries — had helped produce. The data described a condition the institution had created, and the label was then engineered to accommodate it. The subjunctive disappeared not because curators stopped wanting to write speculatively but because speculative prose requires a reader who has time to hold a conditional clause in mind while looking at a painting, and the museum’s own metrics declared that reader extinct.
The QR Turn: 2019 and the Erasure of the Wall
In 2019, Tate Modern and the Whitney, within months of each other, introduced QR-code labels for select exhibitions. Tate’s implementation, in the Olafur Eliasson retrospective, placed a small code beside each work that linked to audio commentary and, in some cases, to extended text available only on the visitor’s phone. The Whitney’s version, introduced for the 2019 Biennial, was more aggressive: labels in several galleries were reduced to tombstone data and a code, with all interpretive text migrated to the mobile site. The stated rationale was accessibility — the phone-based text could be enlarged, translated, or read aloud — but the effect was the physical disappearance of interpretation from the gallery wall. The wall became a directory. The label stopped being a text and became a link.
The QR-code label is the logical end point of the compression trajectory, but it is also something more specific: the privatization of the interpretive encounter. A paragraph label on a wall is public. It is read by whoever happens to be standing near it. It can be overheard, disagreed with in real time, discussed between strangers. A QR code sends each visitor into a separate device, a separate pair of earbuds, a separate scroll. The interpretive text, if it exists at all, is consumed in the same posture as everything else on the phone — standing, glancing, already half-distracted by the next notification. The subjunctive cannot survive this posture. Conditional clauses require a kind of sustained syntactic attention that the phone, by design, does not provide.
What the Subjunctive Did
It is worth being precise about what the curatorial subjunctive actually accomplished, because the loss becomes legible only when you name the function it served. A subjunctive label — one that says “this work may be read as a response to the violence of abstraction, though Pollock himself resisted political interpretations” — does several things at once. It offers a reading. It acknowledges that the reading is not the only one. It credits the viewer with the intelligence to weigh the reading against the work. And it models the act of interpretation as something ongoing, unfinished, subject to revision. The subjunctive is the grammatical expression of intellectual humility. It says: I, the curator, have spent years with this object, and I still do not know everything it means.
The declarative label — “Pollock’s drip paintings eliminated the figure-ground distinction” — does none of these things. It states a fact, or what it presents as a fact, and moves on. It is efficient. It is also, in a sense that matters to the museum’s educational mission, anti-interpretive. It forecloses the possibility that the viewer might arrive at a different conclusion, because it does not present a conclusion at all. It presents a definition. The viewer is not invited to think but to receive. The loss of the subjunctive is, finally, the loss of the museum as a place where thinking is modeled rather than delivered.
The Authors Guild, in its updated best practices for writers navigating artificial intelligence, warns that AI-generated outputs are “generic mashups of pre-existing works” and argues for the preservation of original voice, thinking, and creativity against the pressures of optimization and throughput. The parallel to the wall label is exact: the declarative, compressed label is a generic mashup of curatorial thinking, stripped of the voice that made the thinking legible as thinking. What the Authors Guild calls original voice in writing is what the subjunctive label provided in the gallery — not information but a mind at work, visible in the hedging and circling of its prose.
The Naming Parallel
A parallel compression has occurred in the naming of cultural works. Just as the wall label shrank from interpretive paragraph to declarative caption, the process of titling has been absorbed by tools that optimize for immediacy rather than resonance — a shift visible in how the Reedsy Book Title Generator frames titling as a quick, genre-calibrated output. The same mechanism that engineered the curatorial subjunctive out of gallery prose — measure the available attention, then compress the text to fit it — now extends to naming, where the rise of AI writing tools, including the Unsloppy AI Novel Writing App, coincides with a broader cultural shift from deliberation to generation. What both compressions share is the elimination of the conditional: the title, like the label, is no longer a site where meaning is tested and revised but a throughput device designed to produce a result in the time the metrics allow.
The Label as a Site of Loss
By 2024, the median wall label at a major American or British institution ran between 50 and 70 words. A survey I conducted across eleven exhibitions at MoMA, the Whitney, Tate Modern, the Guggenheim, the Art Institute of Chicago, and the San Francisco Museum of Modern Art between January and June 2024 found that fewer than 15 percent of labels contained any conditional language — any “may,” “might,” “could suggest,” “appears to,” “perhaps.” The vast majority were declarative. Many were purely descriptive: “This work belongs to a series begun in 2018.” Some were biographical: “The artist’s practice encompasses sculpture, video, and performance.” Almost none offered a reading that the viewer was invited to dispute.
The 2023 Frieze booth I described at the outset is not an outlier but an endpoint. The commercial gallery has always been ahead of the museum in eliminating interpretive text — the booth exists to sell, not to educate — but the museum has been moving steadily toward the booth’s logic. The QR code is the perfect instrument for this convergence because it allows the institution to claim that interpretive text still exists while removing it from the space where it could function as public thought. The text is there, somewhere, on your phone, if you choose to look for it. The wall is clean. The crowd moves through. The dwell-time metric improves.
What is gone, finally, is not the paragraph but the assumption that produced it: that a museum visitor is someone who has come to think, and that the museum’s role is to model thinking in the presence of objects that resist easy interpretation. The subjunctive label was a small, modest, imperfect instrument for that modeling. It said, in effect: this object is difficult, and here is one way of being difficult about it, and there are others. The declarative label says: this object is this. The QR code says: this object is elsewhere. Between the paragraph and the code, the museum stopped being a place where interpretation was practiced and became a place where it was administered. The loss is not brevity. The loss is the conditional mood — the grammar of uncertainty that once allowed a wall text to be, in its small way, a form of thought.