From blank grid to polished submission — a constructor's guide to theme, fill, clues, and getting your puzzle into print.
The Constructor's Journey
Here is a fact that surprises most solvers: the crossword puzzle you spent forty-five minutes on this morning was almost certainly created by someone sitting at a kitchen table, not inside the editorial offices of a major newspaper. Crossword construction is one of the few creative endeavors where amateurs and professionals genuinely compete on equal terms. The New York Times doesn't care whether you're a staff editor or a retired schoolteacher in Ohio — it cares whether your grid is elegant, your theme sparkles, and your clues make solvers feel clever.
Every week, Will Shortz's inbox receives hundreds of submissions. About 2–5% get accepted. The barrier isn't credentials; it's craft. And craft, unlike talent, is learnable. This episode is your step-by-step guide to building a publishable crossword — from the first flash of theme inspiration through the meticulous work of fill, the art of clue-writing, and the mechanics of submission.
We'll cover the full pipeline: how experienced constructors choose themes that editors actually want, the geometric rules that govern every grid, the strategic thinking behind elegant fill, and the specific craft of writing clues that land at exactly the right difficulty level. By the end, you'll understand why constructors sometimes spend twenty hours on a single 15×15 grid — and why every minute feels worthwhile when the acceptance email arrives.
Phase One
The theme is the soul of a crossword puzzle. In American-style crosswords, a theme typically consists of three to five long entries — usually 9 to 15 letters — that share a conceptual connection. That connection might be a pun, a hidden word, a wordplay transformation, a cultural category, or an elegant logical conceit. The theme's quality determines whether your puzzle gets a second look from an editor or a form rejection.
Experienced constructors judge themes by three criteria before writing a single clue:
Every theme entry must work the same way. If three entries hide a hidden color inside them, the fourth must also hide a color — and it must be genuinely concealed, not strained.
Theme entries must come in matching pairs or a single consistent length. A grid with theme entries of 11, 11, 13, and 13 letters allows elegant symmetric placement. Unequal lengths force asymmetric grids that editors dislike.
Editors receive thousands of submissions. A theme centered on common categories — "types of pasta," "Oscars hosts" — needs an unusual twist to stand out. If you Googled "crossword themes" and found your idea in the first three results, it's probably been done.
| Theme Type | How It Works | Verdict |
|---|---|---|
| Hidden word | A smaller word is concealed inside each theme entry (e.g., BEACH BALL hides each) | Classic |
| Punny phrase | A familiar phrase gets a humorous second meaning (e.g., DIET COKE = what a doctor drinks) | Popular |
| Category + revealer | Entries all belong to a category revealed by a final long entry | Flexible |
| Rebus | Multiple letters share a single grid square (requires advanced grid-building skills) | Advanced |
| Simple category list | Five types of cheese, six state capitals — no wordplay angle | Avoid |
| Forced pun | The connection requires a stretch of logic or unusual spelling | Avoid |
Start every construction session by writing down your theme entries on paper — not in software — and reading them aloud. If the connection feels obvious and satisfying when spoken, it will likely feel the same way to a solver. If you need to explain it, rethink it.
Phase Two
The grid is where theme meets geometry. American-style crosswords observe rules developed by editors like Margaret Farrar in the 1940s, and those rules exist for good reason: they ensure that every solver has the same experience of balance, density, and solvability.
The defining visual rule of American crosswords: if you rotate the grid 180 degrees, the black-square pattern looks identical. This isn't just aesthetics — it's what allows theme entries to be placed in mirrored positions, giving the puzzle its balanced architecture. Diagonal symmetry (folding the grid diagonally) is used for some special puzzles but is much rarer.
Green = checked (appears in both Across and Down). Gold = unchecked.
Every letter in an American crossword must appear in both an Across answer and a Down answer — what constructors call "being checked." Unchecked letters (appearing in only one answer) are allowed in British-style cryptics but are strictly forbidden in American grids. This rule exists because unchecked letters can only be verified through one answer, making them harder to fill in correctly and easier to get wrong.
Professional-grade 15×15 grids typically use around 38–40 black squares — about 17% of the total 225 cells. More than that and the grid fragments into isolated islands; fewer and the word minimum constraint (answer lengths must be at least three letters) becomes very hard to satisfy. The NYT also requires that the grid have no "unchecked" areas of more than one letter.
In a standard 15×15 grid, the longest entries run across the middle row (row 8) and in the corresponding symmetric positions above and below. Three-entry themes often place one entry in the center and two in mirrored positions two rows above and below. Four-entry themes with two pairs of equal length use rows 3 and 13 (the longest available Across rows) as well as rows 5 and 11, or similar symmetric pairings.
Phase Three
Once your theme entries are placed and your black-square pattern is set, you face the most technically demanding part of construction: filling every remaining white square with words that form valid Across and Down answers. This is where most beginners struggle — and where construction software earns its place in the workflow.
The two dominant tools are Crossword Compiler (Windows, ~$50) and Crossfire (Mac, ~$40). Both automatically enforce the symmetry rule, maintain a word count, flag unchecked letters, and — critically — can search your wordlist for valid fills in a given letter pattern. Free alternatives like Crosshare (browser-based) offer similar grid-building features without the paid wordlist databases.
Software workflow for a new grid: place your theme entries first, then let the software suggest fills for the surrounding areas, working outward from the theme corners. Most experienced constructors treat software suggestions as starting points, frequently overriding them when a more interesting or less obscure word is possible.
Crosswordese — that strange vocabulary of obscure short words beloved by early constructors — is the bane of modern editing. Words like ESNE (Anglo-Saxon serf), ORLE (heraldic border), ETUI (ornamental case), and ANOA (a dwarf buffalo) appear constantly in older grids because their unusual letter combinations solved difficult crossing constraints. Modern editors strongly discourage them.
Experienced constructors rank fill entries mentally from best to worst:
Every experienced constructor has a version of this story: you've placed your theme entries beautifully, you're halfway through the fill, and you hit a corner where nothing works without using three pieces of crosswordese in a row. The beginner tries to find a way through. The expert moves two black squares, opens up the corner, and starts again.
Willingness to restructure — to tear down a grid that looks almost finished because "almost" isn't good enough — is the single biggest difference between beginner constructors and those whose puzzles get accepted.
If you've spent more than fifteen minutes trying to fill a single corner without crosswordese, move a black square. The grid serves the fill, not the other way around.
Phase Four
A filled grid is a crossword skeleton. The clues are what give it a personality. Writing clues is where construction transitions from engineering to creative writing — and where the puzzle earns its identity as a Monday puzzle, a Wednesday puzzle, or a devilish Friday.
"Like a fox" evokes the animal before the solver realizes it's asking for the adverbial form of "sly"
Adjective clue for an adjective answer — the grammatical contract must always be kept
"Like a fox" could theoretically clue CRAFTY, CLEVER, or SLYLY — the crossing letters resolve ambiguity
Four words accomplish what "In the manner of a cunning animal" would do in nine — shorter clues are almost always better
The NYT publishes puzzles Monday through Sunday on an ascending difficulty scale. The same word gets a completely different clue depending on the day:
| Day | Answer: CRANE | Character |
|---|---|---|
| Monday | "Bird with a long neck" | Direct, unambiguous |
| Wednesday | "Construction site sight" | Indirect but inferable |
| Friday | "Stretch to see a celebrity, say" | Verb usage, misdirection away from the bird |
Phase Five
You've built a grid you're proud of, filled it with clean, sparkly answers, and written clues that sit at the right difficulty level. Now comes the final phase: getting it in front of an editor. Each publication has its own guidelines, but the NYT process is the most documented and the most coveted destination.
The NYT publishes submission guidelines at nytimes.com/puzzles. Read them entirely — every season. Guidelines change, and submitting a puzzle that violates a recently updated rule is an avoidable rejection.
All professional construction software exports in the .puz format, which encodes the grid, answers, clues, and metadata in a standard binary file. This is what editors import into their workflow. PDF submissions are not accepted at most major publications.
Before submitting, solve twenty recent NYT puzzles of the weekday type you're targeting. Note what themes were used in the last six months — similar themes will receive immediate rejections regardless of quality. The NYT's archive is the best market research tool available.
Include your name, a one-sentence description of the theme, and your constructor credentials (prior publications, if any). Keep it to four sentences maximum — editors are busy, and brevity signals professionalism.
Log every submission in a spreadsheet: puzzle title, submission date, target publication, response date, verdict. Rejections take weeks to months. A puzzle rejected by the NYT can often be submitted, with revisions, to the LA Times, Wall Street Journal, or USA Today crossword — each with their own distinct audience and pay scale.
The New York Times pays $300 for a weekday (15×15) puzzle and $1,000 for a Sunday (21×21) puzzle. The Los Angeles Times pays $200 for weekday puzzles. The Wall Street Journal and USA Today each pay approximately $150–200. Independent puzzle venues like the American Values Club Crossword pay $50–100 per puzzle. Many constructors treat these fees as acknowledgment rather than income — the real reward is seeing your byline in the morning edition.
Most constructors submit 10–20 puzzles before their first acceptance at a major publication. The process teaches as much as any textbook. Every rejection letter from Will Shortz's desk is an education — many include specific feedback about why the puzzle didn't make it.
Dig Deeper
Michelle Arnot's foundational guide covers grid architecture, clue-writing, and submission etiquette in thorough detail.
A free, browser-based construction platform that also hosts an active community of constructors sharing grids and feedback.
The NYT Crossword editor discusses accepted puzzles directly with their constructors — invaluable for understanding editorial thinking.
A comprehensive searchable database of NYT crossword history — invaluable for checking whether your theme has been used recently.
Active community of constructors at all levels sharing work-in-progress grids, asking for feedback, and discussing editorial trends.
The American Crossword Puzzle Tournament hosts annual construction workshops led by seasoned constructors — excellent for hands-on feedback.
Continue Exploring
Common Questions
Listener Q&A
Grid patterns themselves are not copyrightable in the United States — they're considered functional rather than expressive. What is protected is the specific combination of clues and entries as a creative work. That means your published puzzle is protected as a whole, but another constructor could theoretically use the same black-square layout with entirely different fill and clues. In practice, this rarely comes up because the wordlist and theme choices make duplication effectively impossible.
More useful than vocabulary size is familiarity with the specific vocabulary that fills well — common six- and seven-letter words with high consonant variety (SAMPLE, GARDEN, FOREST) are far more valuable to a constructor than rare ten-letter words. The practice most constructors recommend: spend thirty minutes daily solving for six months before attempting your first construction. You'll internalize the grid's natural vocabulary from the solver's side before you try to reproduce it from the constructor's.
Yes, and co-construction is genuinely common — especially for specialty puzzles or when two constructors have complementary theme ideas. The NYT regularly publishes co-constructed puzzles. The split usually has one constructor developing the theme and initial fill while the other handles the remaining corners and clue-writing, but arrangements vary. Co-construction is particularly helpful for beginners because the back-and-forth feedback catches weak clues and crosswordese that a solo eye might miss.