Category: Default

The shirt reads, in full, LET’S ROOT FOR EACH OTHER AND WATCH EACH OTHER GROW. The words are set in a ring around four smiling root vegetables — a carrot, a radish, a beet, and an onion — each drawn with its taproot descending into a band of soil. The phrase is a pun, and the pun is the whole design: root as the noun the vegetables are, and root as the verb the sentence asks the reader to perform. The diagram makes the double meaning literal by putting the roots underground and the encouragement above it.

What interests me is not the joke, which is amiable and old, but the grammar of the sentence the joke produces. Let’s root for each other is a first-person plural imperative: it includes the speaker and the addressee in a single collective subject. And watch each other grow extends the same subject into a second verb, and the second verb is where the design gets quietly strange. To root for someone is to support them without claiming credit for the outcome. To watch someone grow is to observe a process you do not control. The shirt asks for both at once, and it asks for them reciprocally — each person rooting, each person watched, each person growing. The ring of text enforces the reciprocity: there is no first word and no last word, only a loop.

The vegetables are arranged as a group portrait rather than a harvest. Each has a face — the carrot’s is narrow, the radish’s round, the beet’s broad, the onion’s layered — and each face is smiling. The smiles are identical in kind, which is to say the design does not differentiate temperament. It differentiates species. The carrot is orange and tapered; the radish is red and small; the beet is deep magenta and bulbous; the onion is pale and papery. Their roots are drawn with the same care as their faces, which is the visual argument: the part you eat is not the part that does the work. The taproot is the infrastructure. The smiling top is the visible crop.

I take the design’s claim to be about mutual support as a condition of growth, and I take the pun to be doing more than decoration. Root is one of those English words that carries its own antonym in its history: to root for someone is to encourage, but to root out is to remove, and a root is also what holds a plant in place. The shirt chooses the encouraging sense and suppresses the others, which is what a good slogan does. But the suppression is visible in the drawing. The roots are exposed. The soil is a thin band, not a depth. You can see the whole mechanism, and the mechanism is the joke.

The phrase has a second life as a gardening shirt, and the retail title places it there — Garden is in the listing name, and the vegetables are the kind a kitchen gardener actually grows. But the sentence is not about gardening. It is about a workplace, a team, a friendship, a marriage, a writing group. The garden is the metaphor, and the metaphor is doing the work of making a demand sound like a pleasantry. Watch each other grow is a lovely thing to say and a difficult thing to mean. It requires you to be interested in someone else’s development without measuring it against your own, and it requires you to accept being watched in the same way. The shirt is a small contract, printed in a circle.

There is a version of this design that would be cloying — pastel, cursive, a heart where the soil is. This one is not that. The lettering is even and slightly compressed, the kind of sans-serif that reads as a label rather than a greeting card. The colors are the colors of the vegetables, not the colors of a nursery. The soil is a brown band, not a decorative swoosh. The effect is closer to a seed packet than to a motivational poster, and that is the right register for the sentence. A seed packet tells you what you are getting and what it needs. It does not promise you a harvest.

The phrase let’s root for each other shirt is the kind of query that describes an object by its message rather than its maker, which is how most slogan garments are found. The design answers the query by making the message the entire surface: no logo, no secondary text, no illustration that is not part of the sentence. The four vegetables are the sentence’s nouns. The soil is the sentence’s ground. The ring is the sentence’s syntax. It is a complete argument in a small space, and it is legible from across a room, which is the only test a shirt like this has to pass.

What I keep returning to is the word watch. It is a strange verb to pair with grow, because growth is slow and watching is patient, and the shirt asks for patience as a form of encouragement. To root for someone is easy; it costs nothing and can be done from a distance. To watch someone grow is to stay present through the unphotogenic middle, the part where nothing appears to be happening. The design puts the roots below the soil line, which is where the growth is invisible, and the faces above, which is where the growth is announced. The shirt knows the difference. It is asking for the underground part.

The common ownership is worth stating plainly: this publication and Quirky Cotton Collective share an owner. This is not an independent review. The object is the Let’s Root For Each Other shirt, and the observation above is of the printed artwork, not of a wearing experience. I have not tested it, and I do not need to. The design is the document.

The question the shirt leaves open is the one the pun cannot answer: whether mutual support and mutual growth are the same project or two different ones. The ring of text says they are one sentence. The drawing says they are two layers, one visible and one buried. A reader who has ever been in a group — a staff, a workshop, a family — will recognize the tension. The shirt is cheerful about it. That cheer is the design’s real subject, and it is why the thing works as more than a garden joke.

The shirt reads GOT TOO SILLY in a heavy, all-caps line, and beneath it a white goose-like bird stands in an orange prison uniform, holding a black mugshot board. The board is blank. That blankness is the first thing worth noticing, because a mugshot board is normally the most crowded surface in the picture: name, booking number, charge, date. Here it carries nothing. The joke does not live in the board’s contents. It lives in the board’s presence.

I want to argue that this is a design about format rather than about silliness. The phrase is the setup; the uniform and the board are the punchline, and the punchline works because the apparatus of official record-keeping has been applied to a creature that has no record to keep. The bird is not being charged with anything. It is being processed. The humor is the mismatch between the gravity of the form and the triviality of the offense — and the offense is so trivial it cannot even be named on the board.

What the eye actually receives

Look at the composition as a reader would, top to bottom. The words sit above the figure, which is the ordinary arrangement of a caption. A caption tells you what you are looking at; it frames the image beneath it. So the shirt is not primarily a picture with a slogan attached. It is a caption with a picture attached — a one-line headline over a wire photo. The typography is plain, blocky, and declarative, the kind of lettering that wants to be read at a distance and does not want to be admired. It is doing the work of a label.

The bird is white, rounded, and — this matters — goose-like rather than specifically a goose. The drawing gives us the silhouette of a waterfowl without committing to a species. That vagueness is functional. A recognizable character would carry its own history; a generic bird carries only the shape of innocence. White plumage against an orange uniform is the whole color argument of the design: the uniform is the only saturated element, and it is the element that signifies custody. The black board is the second dark mass, and it reads as a surface — a thing you write on — even though nothing is written.

The bird’s expression, as far as the drawing commits to one, is neutral. It is not grinning and it is not contrite. It holds the board the way a person holds a board in a booking photograph: because someone told it to. The pose is cooperative. That cooperation is what makes the image funny rather than merely cute. A goose that resisted would be a different joke — a chase, a struggle, a farmyard farce. A goose that stands still and holds its own board has accepted the terms of the form. It has been formatted.

The blank board and the compressed narrative

Mugshot humor is old and mostly tired, and it usually depends on a named crime: the dog that ate the sofa, the cat that knocked over the lamp, the toddler booked for assault with a plastic hammer. Those jokes need the charge to be legible, because the charge is the joke. This design withholds the charge. GOT TOO SILLY is not a crime; it is a diagnosis, and a vague one. The board stays empty because the offense is unnameable — or because naming it would collapse the joke into a specific anecdote, and the design wants to stay general.

That generality is the shirt’s real subject. It is a template. The wearer supplies the particulars. This is the same mechanism as a fill-in-the-blank caption or a Mad Libs page: the form is fixed, the content is yours. The bird is a stand-in, the board is a blank, and the phrase is a category wide enough to hold any minor transgression the wearer wants to confess to. The design does not tell a story. It hands you a form and invites you to complete it.

I would go further. The blank board is a small act of editorial restraint. A filled-in board would date the shirt — a specific charge, a specific year, a specific meme cycle. An empty board keeps it evergreen. The joke can be re-applied. That is not a cynical observation; it is the same logic that keeps a good cartoon caption alive across decades. The form outlasts the instance.

Why the phrase is doing more than it looks

GOT TOO SILLY is grammatically odd, and the oddness is load-bearing. Got is a past-tense verb of arrival, not of being. You do not get silly the way you get tired or get rich; silliness is not a destination you reach by transit. The phrase treats silliness as a threshold that was crossed, a line that was passed, a limit that was exceeded. That is why the prison uniform follows so naturally. The sentence structure already implies an authority that set the limit. Someone decided how silly was too silly, and the bird went past it.

This is the quiet institutional joke underneath the visual one. The shirt is not really about a goose. It is about the experience of being told, by some unamused authority, that a harmless thing has gone too far. Anyone who has sat in a meeting that ran long, or received a memo about tone, or been asked to keep it down, recognizes the structure. The bird is a proxy for the reader. The uniform is the reader’s own mild humiliation, rendered in orange.

I want to be careful here. I am not claiming the design is a satire of workplace policy or of any named institution. There is no evidence for that, and the artwork does not name one. What I am claiming is narrower and, I think, more defensible: the phrase’s grammar implies a limit-setter, and the image supplies the limit-setter’s apparatus. The joke is completed by the viewer, who fills in the authority. That is an interpretation of the design, not a historical claim about its intent.

The reader as the missing caption

Here is the argument I want to leave with you. This shirt is a piece of reader-address in the plainest sense: it speaks to whoever is looking at it, and it asks that person to finish the sentence. The blank board is the gap. The phrase is the prompt. The bird is the placeholder. A design like this is not a statement; it is a form letter with the particulars left open, and its success depends on the viewer’s willingness to supply a small, private, ridiculous charge.

That is also why it works as a garment rather than as a poster. A poster is looked at once. A shirt is looked at in passing, by strangers, in the middle of other business. The joke has to be legible in a glance and complete enough to reward a second look. The heavy caps do the first job. The blank board does the second. You see the phrase, you see the bird, you see the board, and then you notice the board is empty — and the emptiness is the invitation.

If you want the object itself, the listing is here: Got Too Silly Funny Bird Mugshot Prison Meme. I am describing the printed design as I observed it, not a wearing experience; I have not tested the shirt and I am not reporting customer reactions.

Editorial note: this publication and Quirky Cotton Collective share an owner.

The open question I would put to other critics and to booksellers who handle this kind of merchandise: what happens to a joke when its blank is the point? A filled-in mugshot dates itself and dies. An empty one stays available, but it also stays slightly abstract — it asks more of the viewer than a named charge would. Is that a strength or a hedge? I do not think the design settles it, and I am not sure it should.

Editorial note: This publication and Quirky Cotton Collective share an owner. This is not an independent review.

The garment preview for the Cat Daddy Feline Father Lover Pet Owner Funny Tee shows a small typographic problem solved by an animal. The letters C and T sit on either side of a cat silhouette, and the word Daddy sits below. Light lettering on a dark shirt. That is the whole design, and it is enough to reward a second look, because the arrangement is doing something more deliberate than the phrase alone suggests.

Read the top line as a word with a hole in it. C and T are the outer consonants of CAT; the middle letter is missing, and the cat silhouette stands in its place. The animal is not decoration appended to a slogan. It is a piece of orthography, the vowel-shaped gap in a three-letter word, filled by the referent itself. The design asks the viewer to complete the spelling, and the completion is the joke.

This is a familiar move in graphic design, though it rarely gets named. Call it the rebus substitution: a picture takes the position of a letter or syllable, and the reader’s eye supplies the sound. Children’s puzzle books use it constantly. So do logos, monograms, and the occasional book jacket. What makes the Cat Daddy version worth pausing over is how little it asks of the viewer. There is no legend, no arrow, no explanatory line. The cat is simply where the A should be, and the brain fills the slot before conscious reading catches up.

Below that solved word sits Daddy, and the two lines together produce the phrase the listing names. The vertical stack matters. If the design had printed CAT DADDY on one line, the cat silhouette would read as a mascot beside a title. Stacked, the cat occupies the first line’s center and Daddy anchors the second, so the eye travels down through the substitution and lands on the identity claim. The shirt is not announcing a category. It is completing a sentence whose subject is the cat and whose predicate is the wearer.

That distinction is where the design earns its keep. A shirt that says Cat Dad describes a household arrangement. A shirt that says Cat Daddy addresses someone. The diminutive ending does the work: Daddy is what a child calls a parent, and applying it to a cat owner borrows the intimacy of that address without pretending to a human family role. The listing’s own type note is careful about this — the phrase is a pet-father identity, not a human father’s birthday garment. The design agrees. There is no cake, no necktie, no calendar date. There is a cat, a gap, and a word that sounds like it is being spoken to someone.

I keep thinking about the typographic compression here. Three letters, one of them pictorial, plus a two-syllable word below. The whole message fits in the space a pocket logo would occupy, and it survives being read at a distance, in motion, at the angle a shirt is usually seen. That is not an accident of taste; it is a constraint of the medium. Garment printing rewards designs that resolve in a glance, and the Cat Daddy layout resolves twice — once when the eye solves the rebus, once when the phrase lands. Two small payoffs from six characters and a silhouette.

The color logic is equally spare. Light lettering on a dark ground gives the cat shape a chance to read as a shape rather than an outline, and it keeps the two text lines from competing with each other. The silhouette does not need internal detail because the word around it supplies the meaning. A more literal cat drawing — whiskers, eyes, a tail curled just so — would clutter the letter slot and slow the substitution. The design chooses the generic cat, the one every viewer already has in mind, and lets the typography carry the specificity.

What I find most interesting is the address. The shirt does not say I am a cat daddy. It says Cat Daddy, a noun phrase with no pronoun, which means the wearer can occupy it or be labeled by it, and a viewer can read it as a self-description or a greeting. That ambiguity is the design’s quietest feature. The rebus pulls the reader in to solve the word; the word then turns around and names the reader. A shirt that only declared would be a bumper sticker. This one recruits.

There is a version of this design that would have been louder — a speech bubble, a paw print, a hashtag, a joke about servants. The observed artwork avoids all of it. The restraint is what makes the phrase feel like an in-group signal rather than a novelty gag. It assumes the viewer knows the term, or is willing to work it out from the cat standing in for a letter. Either way, the shirt has given them something to do.

I would not claim the design invented anything. Rebus substitution is old, cat-parent humor is everywhere, and the phrase itself circulates far beyond this listing. What the shirt does is execute a small set of decisions cleanly: put the picture where the letter belongs, stack the two lines so the substitution reads first, keep the palette to two values, and trust the viewer to finish the word. That is a legible piece of graphic writing, and it is legible for the same reason a good book title is — it leaves the last step to the reader.

The open question, for anyone who wants to extend it: how many other phrases could survive this treatment? The rebus slot is a narrow door. It admits a word whose middle letter can plausibly be an object, and it admits an object whose shape can stand in for that letter without confusion. Cat fits because the A is a hole and the cat is a body. Most words do not. The Cat Daddy shirt is a reminder that the constraint is the design, and that the best solution is often the one the viewer completes without noticing they were asked.

On March 2, 2011, Publishers Weekly reported that HarperCollins would limit its library e-book licenses to a maximum of 26 loans per title, beginning March 7. Librarians learned of the change not from the publisher directly but through a letter from Steve Potash, then CEO of OverDrive, the library e-book wholesaler. Once a title reached 26 checkouts, the report said, libraries would have to purchase a new license. Twenty-six loans, or roughly a year of availability for a high-demand title.

That number is the artifact. It is small enough to fit in a sentence and large enough to have shaped a decade of library acquisition math. What follows is not a verdict on whether the cap was right. It is an attempt to read the clause as a document — to see what it converted, what it traveled through, and what it left unresolved.

What the cap actually changed

Before 2011, HarperCollins sold e-books to libraries under a perpetual-access model. A library bought a license and could lend the title indefinitely. The e-book did not wear out, did not go missing, did not need rebinding. In print terms, it was a copy that never left the shelf.

The 26-loan cap converted that indefinite license into a metered one. Instead of owning access in perpetuity, a library was buying a fixed number of circulations. When the meter ran out, the library bought again. The practical effect was to make a digital license behave, in accounting terms, more like a physical copy: a copy that circulates a finite number of times before it must be replaced.

That analogy is not incidental. It is the whole argument. A physical book in a library collection circulates until it falls apart, and then the library buys another. Publishers have long priced print library copies at a discount to consumer editions, on the theory that library circulation is a form of marketing and a public good. The 26-loan cap extended that logic to digital: if a library e-book is going to be lent like a print book, it should be priced and replaced like one.

The publisher’s stated rationale

HarperCollins did not leave the rationale implicit. In an open letter to librarians reported by Publishers Weekly, Josh Marwell, then president of sales, wrote that the publisher’s library e-book policies were 10 years old and dated from “a time when the number of e-readers was too small to measure.” He stated that the previous policy of selling e-books to libraries in perpetuity, if left unchanged, would “undermine the emerging e-book eco-system, hurt the growing e-book channel, place additional pressure on physical bookstores, and in the end lead to a decrease in book sales and royalties paid to authors.”

Marwell also described the practical shape of the new terms: the 26 checkouts would offer about “a year of availability for titles with the highest demand, and much longer for other titles and core backlist.” Once a license reached 26 checkouts, a new license would be purchased at a lower price pegged to the paperback price.

That last detail matters more than it is usually given credit for. The cap was not a flat repricing. It was a repricing tied to a specific analogy — the paperback — and a specific expectation about demand. High-demand titles would burn through 26 loans in a year. Backlist titles might take a decade. The cap was designed to be asymmetric.

The librarian’s counter-argument was a budget calculation

The reaction from librarians was swift and, in the pages of Publishers Weekly, largely framed as outrage. But the more durable response was arithmetic. In 2021, when HarperCollins completed its acquisition of Houghton Mifflin Harcourt’s trade publishing division, the terms of Houghton’s library e-books changed. A notice to library customers said that as of August 1, Houghton e-book titles would be licensed under HarperCollins’ 26-lend metered access model, one copy/one user. Previously, Houghton had offered libraries trade e-book licenses on a one copy/one user perpetual access model, with prices close to the price of physical books — terms the trade press described as among the most popular and favorable in the library e-book market.

E-books purchased before August 1 would remain under Houghton’s previous model. Pre-orders with a street date of August 1 or later would be delivered under the metered access model.

Michael Blackwell, a librarian at St. Mary’s County in Maryland, wrote on the ReadersFirst blog that HarperCollins’ 26-lend model and pricing remained the most popular and favorable terms offered by the major publishers. He cited Houghton’s prior prices — $8.95 for The Giver, $13.95 for The Hobbit, perpetual access — and said that a 26-circ metered license at around those prices would still be considered favorable. “That price range for a 26-circ metered license seems reasonable,” he told Publishers Weekly, “offering us about what we’d expect from a physical copy of the book and respecting the publishers’ concerns about e-books being forever.”

He also drew a distinction that is easy to miss. “At least the license is per-circ and not the ‘exploding’ time-based licenses [offered by the other major publishers],” he wrote. A per-circ license expires after a set number of loans. A time-based license expires after a set period, regardless of use. For a library, the difference is the difference between paying for circulation and paying for shelf time. Blackwell’s preference for the former is a preference for a model in which unused capacity is not wasted.

And yet he was not endorsing the model wholesale. “Per-circ has been advantageous for publishers and library vendors, not so great for libraries,” he noted. The same terms could be both reasonable and unfavorable, depending on which side of the transaction you sat.

The cap as a portable instrument

The 2011 announcement is the origin story. The 2021 migration is the more revealing event.

When HarperCollins acquired Houghton Mifflin Harcourt’s trade division for roughly $349 million, it did not merely acquire titles. It acquired the right to set licensing terms for those titles. The 26-lend model, which had begun as one publisher’s hedge against perpetual licenses, became a term that could be applied to an entire acquired catalog. Libraries that had bought Houghton e-books under perpetual access kept those licenses. But new titles, and new purchases of existing titles, would arrive under the meter.

This is the part of the story that is not about fairness. It is about reach. A licensing term is only as consequential as the number of catalogs it governs. In 2011, the 26-loan cap applied to HarperCollins titles. In 2021, it applied to HarperCollins titles plus Houghton Mifflin Harcourt titles. The clause did not change. Its footprint did.

For libraries, that means the practical question is not whether any single publisher’s terms are reasonable. It is how many catalogs a single set of terms can reach through corporate consolidation. The 26-loan cap is a unit of account. Its meaning depends on the per-circ price attached to it, and the per-circ price depends on the negotiating position of the library. When one publisher controls more titles, the library’s position weakens.

What the cap is not

It is worth being precise about what the 26-loan cap is not, because the term is often used loosely.

It is not a time-based license. A time-based license — what Blackwell called an “exploding” license — expires after a fixed period, whether or not the title has circulated. A per-circ license expires after a fixed number of loans. The distinction matters for collection management. A library that buys a time-based license for a title that nobody borrows has wasted the money. A library that buys a per-circ license for the same title has lost nothing but the opportunity cost of the initial purchase.

It is not a ban on library lending. HarperCollins continued to license e-books to libraries. Simon & Schuster and Macmillan, as of the 2011 report, did not license e-books to libraries at all. The 26-loan cap was a restriction within a market, not an exit from it.

It is not a fixed price. The cap sets the number of loans. The price per loan is set separately, and it has varied. Marwell said in 2011 that replacement licenses would be priced lower, pegged to paperback. Blackwell in 2021 described the pricing as favorable relative to other major publishers. The cap is a quantity term, not a price term.

The asymmetry that gets lost

Here is a question that neither the publisher’s letter nor the librarian’s response addresses directly: what happens to a title that never reaches 26 checkouts?

For a high-demand frontlist title, 26 loans might be consumed in weeks. For a midlist or backlist title, 26 loans might take years, or might never be reached at all. In that case, the metered license is functionally perpetual. The library pays once and lends indefinitely, because the meter never runs out.

This asymmetry is where the real budget story sits. The 26-loan cap is a constraint on popular titles. It is not a constraint on the long tail. A library system that buys heavily in frontlist bestsellers will feel the cap acutely, replacing licenses every year. A library system that buys deeply in backlist and scholarly titles will barely notice it. The cap redistributes library spending toward the titles that circulate most, which is to say toward the titles that are already most commercially successful.

Whether that is a feature or a bug depends on what you think a library is for. If the library’s mission is to provide access to the most in-demand titles, the cap is a tax on success. If the library’s mission is to preserve a broad and diverse collection, the cap is largely irrelevant, because the titles that matter most to that mission are the ones that circulate least.

The contested evidence on displacement

The argument for metered access rests on a claim about substitution: that library e-lending displaces retail sales. If that claim is true, the cap is a reasonable response. If it is false, the cap is a rent extraction.

The most substantial recent attempt to measure that substitution is a 2026 study by Jéssica Dutra and Robert Stoner of Secretariat Advisors, prepared for the Association of American Publishers and the Authors Guild. The study reports that consumer print sales decline by approximately 0.85 to 1.0 percent for every one-percentage-point increase in the share of e-books in public libraries, with adult fiction at the upper end of that range. It also reports that a 25 percentage-point increase in e-book share of a title, holding total checkouts constant, is associated with up to a 7 percent decline in retail sales, varying by genre and audience.

The study’s front matter discloses that it was undertaken at the request of the AAP and the Authors Guild, and that the empirical analyses and findings belong entirely to Secretariat Advisors. That disclosure does not make the findings wrong. It does mean they should be read as a brief in an argument, not as neutral measurement. The framing — “harm to commercial markets,” libraries “cannibalizing print collections” — is advocacy language embedded in a document presented as empirical. The regressions may be sound. The interpretation is contested.

What the study does not do is measure the effect of the 26-loan cap specifically. It measures the relationship between library e-lending and retail sales. The cap is one variable in that relationship, not the subject of the analysis. No retrieved source establishes that the cap caused a measurable decline in library circulation, book sales, or author royalties. The cap is documented. Its effects are inferred.

What the cap left open

The 26-loan cap is now more than a decade old. It has survived a change in corporate ownership, a major acquisition, and a decade of debate about library e-lending. It has been called reasonable and unfavorable by the same librarian in the same interview. It has been defended as a hedge against perpetual licenses and criticized as a tax on circulation.

What it has not done is settle the underlying question: what is a library e-book for? Is it a copy of a book, to be lent and replaced like a physical copy? Is it a service, to be provided at a price that reflects its public value? Is it a marketing channel, to be subsidized in the interest of discovery? The 26-loan cap answers that question one way. The librarians who negotiated against it answered another. The publishers who have since adopted similar terms answered a third.

The useful next question is not whether the cap was right. It is what happens to a library’s collection when the most popular titles are metered and the rest are not. That is a question about collection development, about the relationship between circulation and mission, and about what a library budget is actually buying when it buys a license instead of a book. It is a question that the 26-loan cap made visible, and that no one has yet answered.

FAQ

What is a metered-access license?
A license that permits a set number of circulations before it expires. The 26-loan cap is a metered-access model: a library can lend a title 26 times, after which it must purchase a new license.

How does a per-circ license differ from a time-based license?
A per-circ license expires after a set number of loans. A time-based license expires after a set period, regardless of use. Librarians have generally preferred per-circ terms because unused capacity is not wasted.

Did the 26-loan cap apply to all publishers?
No. It was a HarperCollins policy. As of the 2011 report, Simon & Schuster and Macmillan did not license e-books to libraries at all. Other publishers have adopted different models, including time-based licenses.

What happened to Houghton Mifflin Harcourt titles after the HarperCollins acquisition?
As of August 1, 2021, new Houghton e-book titles were licensed under HarperCollins’ 26-lend metered access model. Titles purchased before that date remained under Houghton’s previous perpetual-access model.

Does the 26-loan cap still exist?
The retrieved sources document the policy as announced in 2011 and extended to Houghton titles in 2021. They do not establish the current terms. Publishers revise licensing terms, and libraries negotiate individually. The cap should be understood as a documented historical policy, not necessarily as a current one.

Editorial note: This publication and Quirky Cotton Collective share an owner.

The shirt reads like a sign before it reads like a sentence. Across the chest, in the flat, unmodulated lettering of industrial caution, the words NOT FRIENDLY sit above DO NOT TOUCH, and beneath them a yellow triangle frames a hand with a diagonal bar struck through it. The grammar is not conversational. It is regulatory. The design does not say I am not friendly; it says not friendly, as though friendliness were a property of a substance, and do not touch, as though touch were an operation with a known hazard.

That shift — from the first person to the third, from confession to compliance — is the whole joke, and it is also the reason the design is more interesting than the phrase alone. A shirt that said I’m an introvert would be a label. A shirt that says NOT FRIENDLY DO NOT TOUCH is a warning, and a warning is a document. It assumes a reader who has arrived without permission and needs to be managed. The wearer is not describing a mood. The wearer is posting a notice.

I want to read the design as a piece of typography first, because that is where its argument lives. The lettering is condensed, uppercase, and evenly weighted — the visual dialect of OSHA placards, electrical panels, and the laminated sheets taped to the doors of supply closets. The yellow triangle is the standard caution field: yellow for warning, black for the mark, a border that contains the symbol rather than decorating it. The crossed-out hand is the part that does the semantic work. A hand alone would be an invitation. A hand with a bar through it is a prohibition, and prohibitions are addressed to someone specific: you, the person currently reaching.

What the design borrows, then, is not the language of personality but the language of public infrastructure. It takes a format we have all learned to obey without reading — the caution sign — and repurposes it for a private boundary. That is a small act of translation, and it is the kind of translation that fails if the typography is wrong. Set the same words in a rounded sans-serif with a pastel triangle and the joke evaporates; the sign becomes a mood. Set them in the flat, bureaucratic register of an actual hazard placard and the joke lands, because the form is doing the work the words cannot do alone. The shirt is not funny because it says do not touch. It is funny because it says it the way a fire door says keep clear.

There is a second layer, and it has to do with who is being addressed. Warning signs are written for strangers. They are the most impersonal form of second-person address we have: you, whoever you are, do not do this thing. When that format is printed on a garment, the wearer becomes the sign and the reader becomes the passerby. The shirt does not require the wearer to say anything. It outsources the boundary to the typography, which is precisely what someone who does not want to be spoken to would want. The design is a way of being left alone without having to ask.

I am not going to claim this is a new idea. The caution-sign-as-apparel has been around long enough to have its own conventions, and the introvert-shirt category is crowded with variations on the same premise. What distinguishes this one, on inspection, is its restraint. There is no extra copy, no winking subtitle, no small-print punchline. The design trusts the format to carry the tone, and the format is strong enough to do it. The yellow triangle is the only color; everything else is black on the shirt’s ground. That economy is what keeps the sign reading as a sign rather than as a costume.

The phrase itself is worth pausing over, because it is doing something slightly unusual with the word friendly. In ordinary usage, friendly is a disposition — a way of behaving toward others. On the shirt, it is a state, like flammable or fragile. The sign does not say the wearer is being unfriendly; it says the wearer is not friendly, as a fixed condition, and that the correct response is not to take it personally but to adjust one’s behavior. This is the same move that makes the best warning labels funny: they treat human qualities as material properties, and human interactions as safety procedures. The joke is not that the wearer is rude. The joke is that the wearer has been classified.

That classification is also a kind of permission. A warning sign does not apologize. It does not explain. It does not say sorry, I’m just tired or it’s not you, it’s me. It states the hazard and stops. For readers who have spent years performing availability they did not feel, the appeal of the format is not the rudeness but the finality. The sign has already had the conversation. The wearer does not have to.

I should be careful here about what the design actually claims. It does not claim to cure anything, improve anything, or diagnose anyone. It is a printed graphic on a garment, and its effects — if any — are social and interpretive, not therapeutic. What it does is make a boundary legible in a format that strangers already know how to read. Whether that boundary is respected is not something the shirt can guarantee. A warning sign is an instruction, not an enforcement mechanism. The crossed-out hand is a request.

That gap between instruction and outcome is, I think, where the design is most honest. Real caution signs are posted because the hazard cannot be removed. The wet floor is still wet; the sign just tells you to walk around it. The not friendly do not touch shirt operates on the same logic. The wearer is not claiming to have solved the problem of being approached. The wearer is posting the sign anyway, because the alternative — explaining, softening, managing the other person’s feelings in real time — is more expensive than the sign. The typography is doing the emotional labor. That is not a small thing, and it is not a cynical thing either. It is a design decision about who bears the cost of an interaction, and the shirt moves that cost onto the format.

There is a reading-life dimension to this that I keep circling. Readers of a certain temperament — the ones who keep a book in a bag specifically so that a crowded room has an exit — will recognize the sign’s underlying move. It is the same move as the paperback held up at eye level on a train: a portable, socially legible signal that says occupied without saying anything. The shirt is a more efficient version of the same technology. It is a book cover for the body. It compresses a paragraph of social negotiation into a triangle and four words, and it does so in a typeface that no one has to be taught to read.

What I find most interesting, finally, is the way the design treats the viewer as competent. It does not explain the joke. It does not soften the instruction. It assumes you know what a caution triangle means, that you can read a prohibition, and that you will understand the slight absurdity of applying both to a person. That assumption is a form of respect, and it is the reason the shirt works as more than a novelty. It is a piece of visual rhetoric that trusts its audience to complete the syllogism: this is a warning sign; warning signs are for hazards; the hazard is me; adjust accordingly. The viewer finishes the sentence. The wearer never has to start it.

The shirt is available as a standard t-shirt through the listing at Quirky Cotton Collective’s Not Friendly Do Not Touch design. I have described the artwork as observed; the interpretation of its typographic strategy is my own. The question I would put to other readers and sellers is not whether the joke lands — it clearly does, for a certain audience — but what it means that the caution sign has become one of the few public formats we still trust to say no without an apology attached. If the warning label is now a personal pronoun, what does that say about the conversations we have stopped expecting to have out loud?

Editorial note: This publication and Quirky Cotton Collective share an owner. This is not an independent review.

The shirt reads TWEEDLE DUM across the chest, and beneath it, in smaller type, I’M WITH DEE, followed by an arrow pointing right. That is the whole design. There is no illustration of the character, no teacup, no flag, no mushroom. The object is a sentence and a diagram, and the diagram is doing the work that a second person would otherwise do.

I want to look at that arrow the way I would look at a footnote marker or a section sign. It is a piece of typographic furniture that changes the grammatical status of everything above it. Without the arrow, I’m with Dee is a statement of fact, and a slightly sad one — the speaker is claiming an absent companion. With the arrow, the sentence becomes a pointing gesture. It stops describing a relationship and starts assigning a role. The arrow converts a declarative into a deictic: that one, over there, is Dee.

This is a small piece of visual grammar, and it is worth naming precisely because it is so ordinary. Arrows in printed matter have a long and mostly unglamorous career: they direct the eye through diagrams, they mark continuations in line-broken text, they stand in for see above and see opposite. In costume design they do something slightly different. They externalize the referent. The wearer of this shirt is not the subject of the sentence; the wearer is the sentence’s delivery mechanism. The subject is the person standing to the right.

The pair as a two-page spread

What interests me about the Tweedle Dee and Tweedle Dum shirts as a category — and the Tweedle Dum I’m With Dee matching costume shirt is a clean example — is that the design only resolves when two bodies are in the same frame. A single shirt is a fragment. It reads as a joke with the punchline missing, or rather as the setup to a joke whose completion is standing beside you. The right-pointing arrow is an instruction to the viewer: look there. If there is no there, the instruction dangles.

I have written before about how publishing decisions determine which texts survive, and I keep returning to the same underlying question in different materials: what does a document assume about its reader? A copyright page assumes a reader who wants to know the year and the claimant. A colophon assumes a reader who cares how the type was set. A paired costume assumes a reader who is standing in the room. The arrow is the tell. It is addressed outward, to a third party, and it treats that third party as someone who needs orienting — someone who has arrived in the middle of a sentence and requires a signpost.

That is a specific and rather generous assumption. Most clothing does not address a viewer at all. It may signal affiliation, taste, or mood, but it does not usually issue an instruction. This shirt issues one. It says: the person you are looking for is to my right. The wearer has agreed in advance to be half of a compound noun.

Why the text does the work an image would not

Consider the alternative. A shirt bearing a drawing of Tweedle Dum would be a portrait. It would say I am this character, and the pairing would be a matter of two portraits hung side by side. The text-and-arrow version says something more interesting: I am the half that points. It makes the relationship, not the character, the content of the design. The character name is present, but it functions as a label on a specimen, not as an image to be recognized.

There is a practical consequence to this. Text-and-arrow designs are cheap to produce and easy to read at a distance, which is presumably why they recur in the matching-costume genre year after year. But the recurrence is not only economic. The form persists because it solves a real problem: how do you make two separate garments legible as one object? You do it by splitting a sentence across them and using a directional mark to bind the halves. The arrow is the stitch.

I would go further and say the arrow is doing something that a shared motif cannot. Two shirts with the same pumpkin print are related by resemblance. Two shirts that complete each other’s sentences are related by syntax. Resemblance is weaker than syntax. You can wear a pumpkin shirt next to someone in a different pumpkin shirt and the pairing is incidental. You cannot wear I’m with Dee next to someone who is not Dee without the sentence becoming false. The design commits the wearer to a specific claim about the person beside them.

The joke is completed by its viewer

This is what I mean when I say the joke is completed by its viewer rather than by its wearer. The wearer of the Dum shirt has already made the joke; it is printed on the chest. The person who laughs is the one who sees both shirts at once and performs the small mental operation of joining them. The design is a two-part puzzle with a very short solution, and the pleasure of it is the pleasure of any well-made rebus: the moment of assembly.

I am not going to claim this is sophisticated humor. It is not. It is the humor of the caption, the kind that depends on a shared frame of reference and a willingness to be a little ridiculous in public. But the mechanics are worth noticing, because they are the same mechanics that govern a great deal of printed matter that takes itself more seriously. A footnote depends on a reader who will look down. A cross-reference depends on a reader who will turn back. An index depends on a reader who will trust that the page number is correct. All of these are small contracts, and all of them fail if the reader does not cooperate.

The arrow on this shirt is a contract of the same kind, offered to a stranger at a party. It says: you have walked into the middle of something; here is the other half. Whether the stranger accepts the contract is not something the design can control. But the design has done its part. It has left the door open.

What the pairing asks of the pair

There is a small asymmetry built into the Dee and Dum pairing that I find more interesting the longer I look at it. The Dum shirt carries the arrow and the Dee shirt, by the logic of the set, carries the reciprocal. Each shirt is incomplete without the other, but they are not identical in their incompleteness. One points right; the other presumably points left, or names the other half. The pair is a closed loop, and each wearer is responsible for one direction of the reference.

This is, I think, why the design works as a costume rather than merely as a novelty. A costume is a role that is visibly assumed. The wearer of a Tweedle Dum shirt is not pretending to be Tweedle Dum in any deep sense; the shirt is not a mask. The wearer is announcing that they have agreed to play a part in a two-person bit, and the shirt is the announcement. The arrow is the part of the announcement that addresses the audience rather than the partner.

I keep coming back to the arrow because it is the only element of the design that is not a word. Everything else is language: a name, a prepositional phrase, a contraction. The arrow is the one piece of pure notation, and it is the piece that makes the language function as a costume rather than as a slogan. Remove it and you have a shirt that says something about the wearer’s companion. Keep it and you have a shirt that says something to the room.

An open question about the pointing half

What I cannot settle, and would be glad to hear from anyone who has thought about this more carefully, is whether the pointing half of a pair is the subordinate half or the dominant one. In the grammar of the design, Dum is the speaker and Dee is the referent; the speaker is the one doing the pointing, which would seem to give Dum the active role. But the sentence I’m with Dee is a sentence about Dee, and the arrow directs attention away from the speaker. The wearer of the Dum shirt is, in a small way, a signpost rather than a destination.

There is a version of this argument in which the pointing half is the generous half — the one that gives the other person the attention. There is another version in which the pointing half is the self-effacing half, the one that has agreed to be the setup rather than the punchline. I do not think the design decides between these readings, and I am not sure it needs to. What it does decide is that the two halves are not interchangeable. One of them carries the arrow. That asymmetry is the design’s one real idea, and it is a good one.

The shirt is a sentence and a sign. The sentence names a character; the sign tells you where to look. Between them they make a small, durable piece of visual grammar that only works when someone else is standing in the right place. I find that more interesting than the joke it carries, and I suspect the joke is more durable because of it.

Editorial note: This publication and Quirky Cotton Collective share an owner.

The shirt’s printed artwork is a diagram, not a slogan with decoration. In white capitals across the chest, the words I HAD POTENTIAL sit above a curved ramp that descends from left to right. At the upper left ledge, a dashed empty circle marks where a ball might have been. At the lower right, a solid white ball rests on a flat ledge. The ramp is curved, not a straight diagonal. That curve is the whole argument.

Most joke diagrams about failure use a straight line. A ball placed on a straight incline will accelerate uniformly; the physics is predictable, and the joke is simply that the ball rolled away. A curved ramp is different. On a curved track, the ball’s acceleration changes with its position. The shape of the path matters as much as the starting height. The dashed circle at the top left is not a ball that fell; it is a position that was never occupied. The solid ball at the bottom right is not a wreckage; it is a resting state. The diagram says: the potential was real, the path was curved, and the ball ended up somewhere flat.

That is a more precise joke than it first appears. The phrase I had potential usually carries a backward glance — a claim about what might have been. But the diagram refuses the elegy. It shows the ball still present, still solid, just no longer in motion. The dashed circle is the only empty thing in the image. The ball itself is intact. The humor depends on that distinction. If the ball were shattered or absent, the shirt would be a complaint. Because the ball is solid and resting, the shirt is a wry observation about trajectories, not a lament.

This is also why the design works as a piece of compressed narrative. The viewer completes the story in the time it takes to read four words and glance at two shapes. The dashed circle asks a question — where did it start? — and the solid ball answers it — not here. The curved ramp supplies the missing middle. No caption is needed. The physics does the work.

I have written before about how publishing decisions determine which books stay readable, and the same logic applies to visual jokes: the infrastructure of the image — the curve, the dashed line, the solid fill — determines what the joke can mean. A straight ramp would have made the shirt cynical. A curved ramp makes it structural. The difference is small on paper and large in effect.

If you want to read the diagram yourself, the shirt is the I Had Potential Funny Physics Pun Sarcastic Minimalist Quote from Quirky Cotton Collective. The listing shows the same white-on-dark artwork: the dashed circle, the curved ramp, the solid ball. It is a diagram that trusts its viewer to know what a potential energy curve looks like, and to laugh at the gap between the shape of the path and the shape of the life.

The open question is not whether the joke lands. It is whether the dashed circle is a beginning or an ending. The diagram allows both readings, and that ambiguity is what keeps it from being merely sarcastic. A shirt that says I had potential without the diagram would be a confession. With the diagram, it becomes a comment on how potential is measured — by the height you started at, or by the curve you traveled. The ball does not answer. It just rests there, solid, waiting for someone to pick it up again.

The table of contents for The Norton Anthology of English Literature is a color-coded document. On the publisher’s own page for the eleventh edition, new selections are highlighted in yellow and new translations in blue. The full edition carries more than 300 new selections; the shorter edition, 125. This is not how a canon usually presents itself. A canon is supposed to arrive without a revision history, as though the works had always been there. Norton’s table of contents arrives with the changes marked, like a contract with the redlines still visible.

That is the useful thing about it. The highlighted entries are a record of editorial decisions, and the decisions are not purely literary. They are also production decisions — permissions, budgets, timelines, survey data — and the publisher has been unusually willing to say so in public.

What the publisher says the anthology is for

In a 2026 interview on Norton’s learning blog, Marian Johnson, who joined the company in 1978 and later served as its literature editor, describes the anthology’s purpose in institutional terms. Anthologies, she says, “help to democratize literature — that is, they make a broad swathe of different works available to students at all kinds of schools, students who may take just one literature or humanities class during their undergraduate days.” The claim is not that the anthology defines the canon. It is that the anthology lowers the cost of encounter for students who will not otherwise meet these texts at all.

That framing matters for how we read the table of contents. If the anthology’s job is access rather than judgment, then its selections are a floor, not a ceiling — the minimum a teacher can assume is available. The question of what gets taught is then partly a question about what gets licensed, cleared, and printed, and only partly a question about literary value.

The production constraints the editor names

Johnson is specific about the machinery. A new anthology takes “more than five years”; a new edition, “two or so years.” Norton maintains “a generous budget to purchase the rights to include copyrighted works (these numbers are higher than you think).” And then the sentence that ought to be printed on the copyright page of every anthology: “Sometimes, though, a rightsholder says no or insists on payment that would bust the budget.”

Read that against the highlighted table of contents. A yellow highlight marks a new selection, but the absence of a highlight marks nothing in particular — not a judgment, not an oversight, not a deliberate exclusion. It may mark a permission that was never granted, a fee that exceeded the budget, or a translation that could not be commissioned in time. The table of contents is a positive record. It does not tell you what the editors wanted and could not get.

The same constraint appears in a different anthology. In a 2010 interview, Ilan Stavans, general editor of The Norton Anthology of Latino Literature, describes a manuscript of “almost two million words” that “needed to be cut,” and adds: “Permissions also marked our choices.” He calls the finished book “an effort at defining the boundaries of the Latino literary canon” and “one canon-making tool” — a phrase that keeps the tool in view rather than the canon. Thirteen years of work, 201 writers, five centuries, and the binding constraints were length and rights clearance as much as editorial conviction.

The decanonization conundrum

The most revealing passage in Johnson’s interview concerns not what was added but what happened after. She describes the demand to include more works by Native American authors in The Norton Anthology of American Literature: the demand was “loud,” the works were added, and then “surveys showed that few people were teaching those works.” She names the pattern directly: “the disjunction between what people say must be in the books and what they actually teach.”

This is worth pausing on, because it is easy to misread as a complaint about instructors. It is better read as a description of two different measurements. A survey question — should this work be in the anthology? — measures a normative preference. A syllabus measures a scheduling decision made under constraints of weeks, student preparation, and the instructor’s own competence and confidence with the material. The two measurements can diverge sharply, and the anthology sits between them, absorbing the cost of the first without capturing the second.

Johnson’s own gloss is careful: “sometimes people weren’t ready to teach those works, both because they were unfamiliar with the traditions and because those works didn’t align with the norms of traditional Western/European greatness.” That is a claim about instructor preparation and institutional habit, not about the works. It is also, read plainly, a claim about the limits of what a table of contents can accomplish on its own.

Four parties at the table

Put the pieces together and the table of contents looks less like a verdict and more like a settlement among at least four parties:

• Editors and general editors, vetted, per Johnson, “for both their scholarship and their achievements as teachers,” who consult colleagues and draft the apparatus.
• Surveyed instructors — Johnson says Norton “works closely with hundreds of instructors to shape the tables of contents for every edition of every anthology” — whose stated preferences are one input among several.
• Rightsholders, who can refuse or price a work out of the book entirely.
• The budget, which sets the ceiling on how many permissions can be bought and therefore how many contemporary or translated works can appear at all.

None of these parties is deciding what is literature. Each is deciding something adjacent to it — what is affordable, what is surveyable, what is teachable within a term, what a permissions department can clear before the book goes to press. The resulting table of contents is a document about those decisions. It is a canon only in the loose sense that it is what a student is most likely to encounter.

What the anthology can and cannot do

Johnson is explicit about the limit: “All the anthology can do is make available a reasonably representative array of writing, preferably great writing, and provide helpful introductions and annotations. From there, most everything else is up to the teacher.” She adds that she finds it “refreshing when a professor reminds me that the textbook is not the whole class.”

If that is right, then the operative canon for any given student is the syllabus, not the anthology. The anthology’s real function is to set the outer bound of what a teacher can assign without assembling course content from scratch — a floor of availability, not a ceiling of value. The teacher who assigns three of the forty works in a volume is the one making the canon decision that reaches the student. The anthology made the other thirty-seven possible.

This is a less dramatic claim than the one usually made about anthologies, in either direction. It does not say the anthology dictates the canon. It does not say the canon is arbitrary. It says the anthology is a production document with a literary surface, and that the two are not separable. The highlighted yellow entries are the visible part of a negotiation whose other terms — refusals, fees, survey results, page budgets — are not printed.

What a table of contents cannot tell you

For critics and booksellers who want to reason from anthology contents, the practical implication is a discipline of reading. A table of contents can tell you what was licensed and printed in a given edition. It cannot tell you, by itself, what was sought and denied, what was cut for length, what was added in response to survey demand and then went unassigned, or what any particular instructor will teach. Those are separate questions requiring separate evidence: permissions records, editorial correspondence, adoption data, syllabi.

The Norton pages document the volume structure and the count of new selections. The editor interviews document the process in the publisher’s own words. What remains undocumented — and what would move this argument from the general to the concrete — is the case file: a named work, a named anthology, a permission refused or a fee that ended the conversation. Until those cases are assembled, the strongest claim available is the one the editor herself makes: that the anthology makes works available, and that the rest is up to the teacher.

Which leaves an open question for the people who assign, sell, and shelve these books. If the anthology is a floor of availability rather than a statement of value, then the interesting number is not how many works it contains but how many of them are actually assigned — and no one, on the evidence available, is publishing that number.

FAQ

Does the Norton Anthology determine what gets taught in English courses?
No source supports that claim, and the publisher’s own editor points the other way. Johnson describes the anthology as making works available and says “the textbook is not the whole class.” The syllabus, not the anthology, is where the assignment decision is made.

Why are some works left out of an anthology?
The editor interview names several reasons: length limits (the Latino anthology’s manuscript was cut from almost two million words), permissions refusals, and fees that exceed the budget. It does not support claims about specific exclusions for specific reasons, which would require the permissions record for that work.

What do the yellow and blue highlights on Norton’s table of contents mean?
According to the publisher’s table-of-contents page, yellow marks new selections and blue marks new translations. They are a revision log, not a statement about literary value.

How long does it take to produce a new edition?
Johnson states that a new anthology takes more than five years and a new edition about two years.

How can a reader check what changed between editions?
Norton publishes tables of contents for the current edition by volume on its site. Comparing those against a prior edition’s table of contents is the available method; the publisher’s pages do not themselves provide a diff.

An advance is a payment made to an author before publication, recoupable against the royalties the book is later expected to earn. That is the whole definition, and everything in this essay follows from its last word: expected. An advance is not a salary, not a fellowship, and not a warm vote of confidence. It is a forecast written in dollars — a publisher’s binding estimate of how many copies a book will sell — and, because forecasts must be defended to finance departments, it is also the instrument through which the publisher decides what the book had better become. The surrounding machinery matters: the title profit-and-loss statement (the P&L) that projects units and price; the earn-out threshold at which royalties overtake the advance; the delivery-and-acceptance clause that ties a payment to an editor’s judgment. This column treats publishing as a material practice, and few artifacts are more material than a number in a contract.

What an Advance Is, Exactly

The full term is advance against royalties, and the preposition does the work. Every copy sold generates a royalty — on a standard hardcover contract, 10 percent of the list price on the first 5,000 copies, 12.5 percent on the next 5,000, and 15 percent thereafter. Those royalties credit against the outstanding advance balance until the balance reaches zero. Only past that point does the author see further money, usually on semiannual royalty statements. If the balance never reaches zero, the author keeps the money anyway. The advance is a payment, not a loan. Short of outright breach of warranty — plagiarism, fabrication — a publisher cannot claw it back.

Payment arrives on a schedule written into the contract. A modest deal splits the advance in halves: half on signing, half on delivery and acceptance of the manuscript. Larger deals take thirds or quarters, adding installments on hardcover publication and sometimes on paperback publication. The schedule matters less as income than as plot. Each installment is a checkpoint where the publisher’s expectation of the book gets confirmed, or quietly renegotiated.

Where Expectation Is Born: The Title P&L

Editorial expectation is born in a spreadsheet, not in a manuscript. Before an editor makes an offer, she builds a title P&L — a one-page projection running the book’s assumed list price, projected unit sales, manufacturing costs, and plant costs (typesetting, jacket design, first-proof charges) down to a profit line. The advance is then solved for backward: given a projected sale of, say, 12,000 copies, how large a payment can be made before the profit line drops below the house’s target margin? The resulting figure is what the editor is permitted to offer. It is a residue of arithmetic, not of enthusiasm.

The P&L assigns the manuscript a readership — a specific number of humans — before the manuscript exists, and every later decision about the book gets measured against that number. When people say a book overperformed or underperformed, they are describing its conduct relative to a projection that predates the finished draft. An editor may love a book. The acquisitions meeting, which must approve the offer, is asked to love the number.

Delivery and Acceptance: Where Money Meets Judgment

The delivery-and-acceptance clause is the one place in the contract where editorial expectation has legal teeth. Standard language provides that the manuscript, when delivered, must be complete and satisfactory — with satisfactory determined by the publisher’s own good-faith judgment. The delivery installment is contingent on that determination. An editor who feels the manuscript has drifted from the proposal — the proposal that justified the advance — can hold the payment until revisions arrive. In the contracts this column has collected, the delivery-and-acceptance clause is the most heavily negotiated paragraph after the grant of rights. It is rarely used to seize money back. It is used, routinely, to hold money hostage to a judgment call.

Read the sequence as a document and the logic is blunt: the proposal promised one book, the contract paid for that book, and a manuscript that arrives as a different book is, in the contract’s terms, unsatisfactory. The advance does not merely precede editorial expectation. It produces it — the editor is contractually obligated to measure the manuscript against the object she paid for.

How the Industry Talks About Money

Advances are confidential, but their ranges are public by convention. Deal announcements — most visibly on Publishers Marketplace, the subscription service that reports them — are graded in tiers: a nice deal runs roughly to $49,000, a very nice deal to $99,000, a good deal to $250,000, a significant deal to $500,000, and a major deal beyond that, with seven-figure sums named as such. The tiers are approximate and the service rounds. What they supply is a shared vocabulary in which the size of an expectation can be stated without stating it.

The auction, which sets many of the larger numbers, is itself an expectation machine. An agent takes the book to several houses in timed rounds; bids escalate until one house’s forecast defeats the rest. A pre-empt skips the process: a single house pays a premium — often above what the auction was expected to clear — for the certainty of not competing. The premium is the price of being sure, and it ratchets the book’s internal bar accordingly. The winning bid is not the book’s value. It is the industry’s most confident reading of its future readers.

What a Big Advance Buys Inside the House

Money spent is a claim on attention. A large advance obliges the house to make its own forecast come true, and it has instruments for the attempt: a positioning meeting to decide what the book is; a slot at the sales conference; a first printing sized to the projection; a laydown date for a coordinated on-sale; and co-op, the budget retailers charge publishers for front-of-store placement. Those budgets follow the P&Ls, which is to say they follow the advances. A book bought cheap is, in this sense, purchased twice: once with a small payment, and again with a small allocation of the house’s belief.

None of this is a scandal; it is rationing. A publisher holds finite budgets and must distribute conviction where conviction has been priced. The distortion is quieter: the advance sets the bar the book must clear to count as a success, and the marketing program is sized to defend the number rather than to find the book’s actual readers. When the forecast is wrong in the optimistic direction, the correction arrives as a physical object.

Earning Out: The Arithmetic of Judgment

A book earns out when accumulated royalties exceed the advance. The arithmetic is unglamorous and worth performing once. Take a $25,000 advance against a $27.00 hardcover with the standard escalators. The first 5,000 copies pay $2.70 each, or $13,500; the next copies pay $3.375 each. The advance clears at roughly 8,400 hardcover copies — a figure with no relation to the publisher’s profit, which arrives earlier, and everything to do with the author’s second check, which may never arrive.

Two mechanisms stand between the arithmetic and the cash. The first is the reserve against returns: bookstores in the United States buy stock on a returnable basis — unsold copies go back to the warehouse — under supply-chain conventions the Book Industry Study Group codifies, so publishers hold back a portion of earned royalties against the returns they anticipate. A sale is provisional until the reserve is released. The second is simpler: the royalty is a discount on expectations. A book can sell briskly and still never reach the bar its advance set.

How often is the bar cleared? There is no public ledger. The estimate that roughly a quarter of trade books earn out has circulated among agents and editors for over a decade, repeated with the confidence of an audit and the sourcing of folklore. Treat it as directionally serious rather than exact. What is documented is the income consequence: the periodic surveys run by the Authors Guild have placed median earnings from writing at levels that make the earn-out rate a material fact of the profession, not a curiosity.

When a hardcover fails decisively, the correction is physical. Returns flow back; warehousing costs outrun hope; and the stock is remaindered — sold off at a fraction of the cover price, often marked with a stroke of marker across the bottom edge. The remainder mark is the most legible artifact in this economy: a visible stroke on the body of the book recording, in ink, the distance between a forecast and a readership.

The Small Advance’s Quiet Bargain

A small advance is not a small kindness; it is a different wager. Low advances earn out easily, which flatters the P&L, flatters the editor’s acquisition record, and makes the book profitable in the internal accounting almost by definition. The price is paid in attention. A book bought at the bottom of the nice deal tier receives a modest printing, a thin review-copy run, little or no co-op, and a window of institutional interest measured in weeks. Its author is invited to earn out — and to earn out into silence.

This is the midlist bargain, and it deserves stating plainly, because the whole system runs on it: a large advance buys scrutiny, resources, and a bar the book may not clear; a small advance buys an easy clearance and almost nothing else. Books that earn out on small advances are often renewed at small escalations, because a tidy P&L is its own argument for continuation. Neither side of the trade is irrational. Both sides are priced.

Why the Ledger Reaches the Shelf

Readers inherit these numbers whether or not they ever see them. A book that fails its advance is not merely a disappointment; it becomes a candidate for going out of print, and the out-of-print clause in its contract — the terms under which rights revert to the author, or fail to — decides whether it stays readable at all. Which books remain available is partly settled by forecasts written before the manuscripts were finished. That is the claim this column exists to document, and the advance is its clearest instance: a genre of prediction about readers, filed as a line item, with the remainder table as its errata.

Frequently Asked Questions

Does an author have to repay an advance if the book does not earn out?

No. An advance is a payment, not a loan. If royalties never cover it, the shortfall is the publisher’s loss to absorb. Repayment arises only where the contract treats the delivered work as defective — breach of warranty, plagiarism, fabrication — not where the book simply sold fewer copies than projected.

What does earning out actually mean?

It means cumulative royalties have exceeded the advance. On a $27.00 hardcover with standard escalators (10 percent of list to 5,000 copies, 12.5 percent to 10,000, 15 percent after), a $25,000 advance earns out at roughly 8,400 copies. Past that threshold the author receives additional royalty payments, minus any reserve the publisher holds against returns.

How many books earn out their advances?

No public ledger exists, so no exact figure can be cited responsibly. The estimate repeated across the trade for more than a decade is that roughly a quarter of books earn out; treat it as an informed range rather than a statistic. Publishers can profit on books that never earn out, because the house’s break-even sits on a different line of the P&L than the author’s.

Can a publisher withhold an advance payment because it dislikes the manuscript?

Yes, within the delivery-and-acceptance clause. The contract typically requires a manuscript the publisher judges satisfactory, in its own good-faith discretion, before the delivery installment is paid. In practice the clause funds revision requests more often than standoffs, but the withheld payment is a real instrument — the one place where editorial expectation carries contractual force.

The Column’s Next Artifact

The file this column keeps — clauses, P&L templates, deal reports, royalty statements, remainder marks — will next take up the out-of-print clause itself: the document that decides a failed forecast’s afterlife, and whether a book returns to its author’s control or stays warehoused beyond any reader’s reach. The terms defined here in passing — earn-out, reserve, co-op, laydown, remainder — will get a standing glossary page as the column accumulates.

Every book begins as a comparison. Before a manuscript is read, before an editor falls in love with a sentence, before a sales rep builds a pitch, the book is mapped onto something that already exists. This is the comp title—shorthand for “comparable title”—and it is the first and most durable instrument publishers use to imagine a book’s future. The comp title answers a question that precedes all others in modern publishing: what is this like? Not what is this doing, not what is this trying to be, but what already-successful object can we hold it next to so that retailers, reviewers, and readers will recognize its shape?

The comp title is not new. Publishers have always classified books against existing successes. What has changed is its gravitational force. It has migrated outward from the acquisition meeting—where it once served as a shorthand among professionals who shared a common reading culture—into nearly every stage of a book’s life. Retail algorithms use comp logic to recommend books. Sales forecasting models use it to predict print runs. Marketing departments use it to design cover briefs. And now, writers themselves use it to generate plots before a manuscript exists. The comp title has become infrastructure: the hidden machinery that determines what gets written, what gets acquired, and what survives.

What gets lost is not merely originality in the romantic sense—the unique voice, the unprecedented vision. What gets lost is the capacity to imagine a book as something other than a version of something else. The comp title system assumes that value is legible in advance, that a book’s worth can be established before it exists by mapping it onto what has already succeeded. This assumption is now so deeply embedded in publishing infrastructure that challenging it feels not just impractical but almost incoherent, as though you had asked the industry to operate without language.

How Comp Titles Work

The mechanics are straightforward. When an agent submits a manuscript to an editor, the pitch letter includes comp titles—usually two or three recent books that share a genre, a readership, a tone, or a market position. “For readers of Rachel Cusk and Jenny Offill” tells an acquisitions committee that the manuscript belongs in a recognizable lineage of elliptical, fragmentary literary fiction by women, and that there is a demonstrated audience for that lineage. The comp title does work that a synopsis cannot: it compresses aesthetic and commercial information into a single reference point that every person in the room understands without further explanation.

In acquisition meetings, comp titles function as a shared vocabulary. An editor who says “It’s Severance meets The Vegetarian” is not describing the book so much as activating a network of associations—sales figures, review coverage, prize history, reader demographics—that allows the committee to make a rapid financial judgment. The comp title is a forecasting instrument. It tells the publisher how many copies to expect, which retailers will stock the book, and what co-op spending (the fees publishers pay chain bookstores for prime placement) will be required to achieve face-out display.

The problem is not that publishers compare books. Comparison is intrinsic to criticism, to marketing, to reading itself. The problem is that the comp title system has become the primary mode of imagining a book’s future, and that it operates as a closed loop. If a book is acquired because it resembles three books that sold well, it will be marketed to readers of those three books, shelved near them in bookstores, and recommended by algorithms that have learned to associate those books with each other. The comp title becomes self-fulfilling. And the books that cannot be comped—the books that do not resemble anything that has already succeeded—are not acquired at all, or are acquired with smaller advances, smaller print runs, and less marketing support, which virtually guarantees that they will not succeed, which confirms the original judgment that they could not be comped.

The Self-Fulfilling Prophecy

Consider a hypothetical that is not really hypothetical, because every editor and agent reading this will recognize it. A literary novel arrives on submission. The prose is distinctive. The structure is unusual—a nested narrative that shifts between second and first person, with a central section told entirely in transcribed voicemails. The agent, knowing that acquisitions committees need comps, offers three: a recent National Book Award finalist, a moderately successful novel from a small press, and a book that was published a decade ago and has since gone out of print. The editor loves the manuscript but cannot find a comp combination that feels clean. The NBA finalist is too prestigious—comparing an unknown debut to it sets expectations the sales team cannot meet. The small press novel sold 4,000 copies in paperback, which is not enough to justify a hardcover print run. The out-of-print title is meaningless to retailers who stock based on current availability.

The editor brings the book to acquisitions anyway. The response is familiar: “We love the writing, but we can’t model this.” The book is passed on. Not because the writing is bad, not because the editor lacks conviction, but because the infrastructure of acquisition—the sales forecasting tools, the retail buyer conversations, the co-op budget calculations—requires a comp title to function. Without one, the book cannot be priced, printed, or positioned. It cannot be imagined as a product. And in an industry that must sell books to retailers before readers ever see them, a book that cannot be imagined as a product cannot become one.

This is not a story about individual philistinism. It is a story about structural logic. The comp title system does not reward conservatism because editors are conservative. It rewards conservatism because the entire downstream apparatus—retail buying, algorithmic recommendation, sales forecasting—was built to process books that resemble other books. A book that does not resemble anything is not merely difficult to sell. It is literally unprocessable. The system has no slot for it.

When Comps Shape the Book Itself

The comp title’s influence does not end at acquisition. It shapes editorial feedback, cover design, jacket copy, and the author’s own sense of what the book is. An editor who acquired a novel as “literary thriller with the emotional depth of The Secret History” will, consciously or not, push revisions that accentuate the thriller elements and deepen the emotional architecture that echoes Donna Tartt. The marketing department, working from the same comp set, will design a cover that signals “literary thriller” through color palette, typography, and imagery—the dark background, the serif title, the single ominous object. The jacket copy will emphasize suspense and elite institutional settings. The book will be reviewed as a literary thriller. Readers who liked The Secret History will buy it expecting a similar experience, and some will be disappointed when the book is quieter, stranger, or less plot-driven than the comp set promised.

This is not a hypothetical. It is the reproductive mechanism of the comp title system. The book is shaped, before publication, to resemble the books it was compared to. Then it is sold to readers of those books. Then its sales data confirms that it performed like those books—or didn’t—and that data becomes the comp set for the next book in the lineage. The loop closes. What began as a convenient shorthand in an acquisition meeting becomes the mold into which the book is poured.

I want to be precise about what this means. The comp title does not prevent good books from being published. It prevents certain kinds of good books from being published—books whose value cannot be established in advance by comparison. These are not necessarily experimental or avant-garde works. They are simply books that do not resemble anything currently on the market, or that resemble something so old or so marginal that the comp is commercially useless. The system’s blind spot is not the weird. It is the unprecedented-but-legible: the book that a reader would love if she encountered it, but that the infrastructure cannot deliver to her because it cannot describe the book in terms she already recognizes.

The Migration Into Plot Generation

The comp title’s logic has now migrated from the publisher’s side of the desk to the writer’s. This is the most significant development in the system’s history, and it is happening quietly, through tools that writers encounter as practical aids rather than as instruments of aesthetic constraint.

Consider the Reedsy Plot Generator, an AI-powered tool that asks writers to define genre, tone, story structure, tropes, and ending type before generating a structured plot outline. The tool’s own instructions state that “genre shapes the conventions the AI draws from,” which is a precise description of the comp title’s core logic: the unwritten book is mapped onto existing conventions before it exists, and its value is established by the degree to which it conforms to patterns that have already succeeded. The writer selects from five story structures—3-Act, 5-Act, Save the Cat, Hero’s Journey, 7-Point—and the generator produces a plot that fits the selected framework. The tool asks for comp-title-adjacent parameters: tropes like “enemies to lovers” or “chosen one,” thematic preoccupations like “the cost of ambition,” and protagonist descriptions that conform to recognizable types. The generator’s output is not a plot in the sense that a writer invents one. It is a pattern-match against a database of prior plots, sorted by genre and convention.

That same discipline applies to narrative structure: before publishing, editors need a way to test events, claims, and consequences actually follow one another, which is where how Unsloppy AI Writing App fits the writing workflow can function as a planning aid rather than a substitute for domain evidence.

The Authors Guild has addressed this tension directly, noting in its updated best practices for AI use that all commercially available foundational large language models have been trained on pirated, unlicensed books without compensating authors or publishers. The Guild frames the stakes in terms that resonate with the comp title problem: the concern that quality human writing risks becoming “a rare luxury good representing only a minority of views” when generative tools flatten originality by operating on pattern-matching logic. The Guild’s institutional position is that a writer’s original voice, thinking, and creativity are what make them the writer they are—and that these qualities are precisely what pattern-matching systems cannot produce.

What we are seeing is the comp title system made executable. The publisher’s comp title says: this book is like that book. The AI plot generator says: this book will be like that book, before it is written. The logic is identical. The difference is that the publisher’s comp title operates after the manuscript exists, as a description of what the book resembles. The AI tool operates before the manuscript exists, as a prescription for what the book should become. The comp title has moved from a post-hoc classification to a pre-hoc template.

This matters because it closes the loop in a new way. When publishers comp books after they are written, the writer’s original vision at least exists somewhere—in the manuscript, if nowhere else. The system may fail to publish it, but it cannot prevent it from being written. When writers use comp-title logic to generate plots before writing, the original vision may never form at all. The book is conceived as a version of something else. The comp title does not merely shape the book’s reception. It shapes the book’s conception.

What the Comp Title Cannot See

The comp title system is a machine for reproducing success. It is reasonably good at this. Books that resemble successful books do, on average, sell more copies than books that do not. This is not a trivial observation. Publishers are not charity. They employ people who need salaries. The comp title system exists because it works often enough to justify its existence, and because the alternative—acquiring books with no market model—is financially untenable for all but the smallest presses.

But the system’s success at reproducing known patterns is inseparable from its failure to recognize unknown ones. The comp title cannot account for the book that creates its own market. It cannot account for the book that does not resemble anything until after it succeeds, at which point it becomes the comp for the next wave. Stoner was not comped to anything when it was published in 1965. It sold poorly and went out of print. It was rediscovered decades later through a chain of events that no sales forecasting model could have predicted: a French translation, a European bestseller list, a reissue, a viral moment. The comp title system would never have acquired Stoner. It did acquire Stoner, in fact—John Williams’s editor at Viking had personal conviction that transcended the market logic of the time—but the system would not have, and the fact that it slipped through is a testament to individual conviction, not to the system’s capacity for recognizing value.

The same is true of every book that became a classic without first becoming a comp. The Master and Margarita had no comp. Pale Fire had no comp. Austerlitz had no comp. These books succeeded because they were unlike anything else, and because readers encountered them through mechanisms—translation, critical advocacy, curatorial passion—that operate outside the comp title system. The system does not produce these books. It cannot. It was never designed to.

The Infrastructure of Resemblance

If the comp title system is infrastructure, then the question is not whether to dismantle it but what to build alongside it. Dismantling the comp title system would require dismantling the retail, algorithmic, and forecasting apparatus that depends on it, which is neither possible nor desirable. The system does useful work. It helps retailers stock books. It helps readers find books they are likely to enjoy. It helps publishers make financial decisions that keep them solvent.

But the system needs a counterweight. It needs an infrastructure of discovery that does not depend on resemblance. This is what independent bookstores, literary journals, and small presses have historically provided: a space where books can be recommended not because they are like other books but because a person with taste and conviction believes a reader should encounter them. The independent bookseller who handsells a book that has no comp is doing what the comp title system cannot do. The small press that acquires a book because the editor believes in it, knowing it will sell 800 copies, is sustaining the literary ecosystem that the comp title system cannot reach.

The danger is not that the comp title system exists. The danger is that it has become the only system. As retail consolidates, as algorithms dominate discovery, and as AI-assisted plotting tools reproduce the comp title’s logic at the point of composition, the infrastructure of resemblance crowds out the infrastructure of encounter. Books that cannot be comped do not merely fail to sell. They fail to be written. They fail to be imagined.

The comp title system’s deepest effect is not on what gets published. It is on what gets conceived. When writers begin their work by asking what their book resembles, rather than what their book is, the comp title has won. It has colonized the act of imagination itself. The book that does not yet exist has already been mapped onto the book that already does, and the space between them—the space where originality lives—has been quietly closed.

What we need is not the abolition of the comp title but the preservation of the space in which a book can be imagined before it is compared. That space has always been small. It is getting smaller. Protecting it is not a matter of individual writerly virtue or editorial courage. It is a matter of infrastructure—the same kind of infrastructure that the comp title system has built so efficiently. The question is whether we are willing to build something else alongside it, before the comp title becomes the only language in which a book can be thought.