Crosswords by Charlie Methven

Tips for setting cryptics

This page offers some advice for people starting to set their own cryptic crosswords. I assume rookie setters know the basics of how clues work, so these posts are mainly about highlighting some of the less obvious crosswording conventions and warning people about some of the pitfalls that other setters including me have cheerfully stumbled into in the past.

This is a partial guide in two senses: I've tried to keep things snappy without outlining every single convention, and I've leaned towards my own biases and opinions as a setter and solver. Reading some of the other guides linked at the bottom of the page will give you a more rounded view of different debates in crosswords.

Believe it or not, the few thousand words below are designed to help people have fun. Solvers don’t get much enjoyment out of solving puzzles which unknowingly ignore the conventions of crosswordland, and equally it’s not much fun for rookie setters to find themselves accused of breaking rules they might not even know about. My hope is that this guide will help rookie setters launch their setting careers with fewer rookie errors and more excellent clues.

On surface readings

One of the most important aspects of writing crossword clues is constructing good surfaces. The surface (or surface reading) of a clue is what the clue might be understood to mean if someone saw it written down in isolation, with no idea that it was a crossword clue with a hidden cryptic meaning. 

The best surfaces are linguistically smooth and narratively plausible. By “linguistically smooth” I mean the surface is grammatically correct and uses words in familiar, natural-sounding ways. A surface reading should sound like something you might overhear in a conversation, not like something you might get after putting some text through several rounds of Google Translate. 

By “narratively plausible” I mean the story or image conjured up by the clue makes some sort of sense regardless of whatever its secret cryptic meaning is. There's sometimes a stereotype that cryptic clues are bizarre, never before seen sequences of words reminiscent of Noam Chomsky’s invented sentence “Colourless green ideas sleep furiously”, but in fact the clues most admired by crossworders are the ones which reveal unexpected possibilities for wordplay in ordinary-sounding language. The best surfaces are not just plausible but apt, humorous, surprising, vivid or memorable.

Consider these clues for PIRATE:

Illegally copy a tripe terribly (6)

This is grammatically bizarre as well as meaningless. “Illegally copy” initially reads like an instruction to the reader, but “a tripe terribly” doesn’t mean anything, partly because “a tripe” is not a natural bit of English (tripe isn’t countable, so you can't have "a" tripe). The fact the first and last word in the clue are both adverbs means it doesn't read very naturally.

Buccaneer trained air pet (6)

This is better, because it makes grammatical sense, and the anagram indicator “trained” fits well with “pet” and works to conjure up some kind of story, but what are we to make of "air pet"? At a stretch we could say it's a weird way of describing a parrot, but ultimately, like "a tripe", it's just poorly disguised anagram fodder, so the clue isn't very elegant or satisfying.

Naughty sailor ate most of ripe bananas (6)

This clue, which asks solvers for an anagram (“bananas”) of ATE + most of the letters in RIP[e], is much better. The clue is not just grammatical, but idiomatic and natural-sounding (you can imagine someone telling a story about a "naughty sailor"), plausible as well as apt (sailors might well eat bananas) and mildly amusing (maybe the sailor found the bananas disappointing or got in trouble for eating most of them). A final flourish is that the real meaning of “naughty sailor” turns out to be slightly different from the surface-level meaning: we need a criminal who sails rather than just a sailor who misbehaves.

Think of writing good surfaces as the main service you’re providing solvers. It's pretty easy to whack together gibberish sets of wordplay instructions, and you could probably get a computer to do it, but (so far) only humans have the sensitivity and precision needed to write clues which combine natural-sounding language with familiar concepts in a satisfying way. Wordplay is fun, but we’d all get bored very quickly if it wasn't wrapped up in amusing, vivid or memorable surface readings.

Back to top

On matching parts of speech

One of the biggest giveaways that a setter is just starting out is difficulty following the rule that X can only fairly act as a definition of Y if they're the same "part of speech" (the same goes for X becoming Y as part of a wordplay operation). If you also got through school without hearing this term, parts of speech are the grammatical categories a word falls into based on its role in a sentence: nouns, verbs, adjectives, adverbs, prepositions, and a few others. If you're starting with a word which is only ever a verb, it can only define or turn into another verb, never a noun or adjective ("refurbish" can define "redecorate", but not "renovation" or "reconstructive"). Likewise a word which is only ever an adjective can define or turn into another adjective but never a noun ("studious" becomes "scholarly", never "scholar").

Of course, it's very common for English words to belong to different parts of speech depending on context, and a word might fit into one grammatical category in the surface reading of the clue but secretly belong in another category in the cryptic reading. For example, a clue which uses "academic" to make you picture someone employed as an academic at the surface level might secretly need you to treat that word as an adjective meaning scholarly or relating to academia in the cryptic reading. Tricking the solver like this is not just fine but essential to spinning out interesting clues.

All of the above also applies to phrases, which are categorised based on their overall function within a sentence (even if the phrases themselves could be further broken down into different POS categories). So the noun phrase “studious person” can fairly lead to a noun like "scholar”, the verb phrase “writes scruffily” can lead to a verb like “scribbles”, the adjective phrase “famously bad” can lead to an adjective like “notorious”, and the adverbial phrase “in a slow manner” can lead to an adverb like “sluggishly”. Remember that you need to match other grammatical aspects of the word too (plurals go to other plurals; past tense verbs go to other past tense verbs; -er adjectives go to other -er adjectives etc).

Back to top

On indirect anagrams etc

If I could only give two bits of advice to new setters, both of them would be to never do indirect anagrams. These are anagram clues where the letters the solver is going to have to rearrange are not written down directly in the clue. For example, African country’s emblems organised (7) requires the solver to turn “emblems” into REGALIA and solve an anagram of that to get to ALGERIA. The expected and fair way to do this would be African country’s regalia organised (7), which gives the solver all the necessary letters up front. Indirect anagrams are universally regarded as unfair, but a lot of new setters launch into writing clues without anyone giving them this memo. 

It’s also worth being aware of other conventions on direct/indirect cluing for various wordplay devices. The table below shows the expectations solvers have for how various wordplay devices would be clued. Green ticks indicate standard practice, red crosses indicate things that'd be seen as unfair, and lemons indicate uses which are sort of fine but just a bit rubbish because they're too easy.

Back to top

On non-Ximenean indicators

You can read lots more about “Ximenean” standards elsewhere (e.g. here and here), but the gist is that these are a set of standards of precision and fairness popularised by an influential early cryptic setter known as Ximenes. When people say a clue is non-Ximenean today they normally mean it uses a wordplay indicator which fails to provide the solver with clear and grammatically correct instructions.

A typical example of a non-Ximenean indicator would be something like "second shelf" to mean the second letter of the word "shelf". The problem here is that there's no precedent in normal English usage for "second [X]” to mean the second part of X. Yes, a phrase like "second shelf" means the second shelf of some collection of shelves, but it would never mean the second part of that individual shelf. If you're in a bookshop and want to buy the second book along on a particular shelf, you wouldn't say "second shelf" to get this across, so you shouldn't ask a solver for "second shelf" if what you mean is the second bit of the word "shelf".

A fairer indicator than "second [X]" would be something like “[X]'s second", "second for [X]”, "second to [X]", "second from [X]" or "second of [X]". All of these can intuitively be interpreted as meaning the particular second thing which belongs to X: you just have to compare them to real-world phrases like "France's currency", "bill for dinner", "lyrics to the song", "captain from the team" or "mother of Mary", all of which mean the Y belonging or pertaining to some X. There would be more debate over something like "[X] top", as some people argue that this fails to indicate the top of X in a truly grammatical way. My two cents here is that we all know that "party leader" means the leader of a party, so we shouldn't have any difficulty understanding "party leader" to mean the first letter of the word "party" either.

Something like "redhead" meaning R (the head of "red") is even more likely to annoy people. I tend to avoid rocking the boat on this one but wouldn't be opposed to a mutiny. Any English speaker who's heard of mastheads, arrowheads or letterheads knows that "[X]head" can mean the topmost part of something, so I can't see how it's unreasonable to ask solvers to interpret "redhead" as the top bit of the word "red". The fact the real-world concept of red doesn't (and couldn't) have a top is no more relevant here than the fact that words can't get drunk is relevant when we're asked to see "drunk" as an anagram indicator. I suppose I'd be less convinced by "behead"=B or "overhead"=O (maybe because they're not nouns?), or god forbid "defend"=F or "extend"=T!, which are probably just too clever by half.

But on the likes of "top shelf" I'm firmly in the Ximenean camp, though I don’t particularly think it’s a question of fairness to the solver. After all, it's not a huge leap to guess that "first [X]" might be telling to you to do something with the first letter of X (check out some non-Ximenean clues on Minute Cryptic to find plenty of satisifed solvers who've made such leaps without injury [1] [2] [3]). I think the reason I dislike non-Ximenean indicators is really that they seem unfair to other setters. Part of the challenge and the fun of setting cryptics is that we're all trying to discover latent possibilities for wordplay within a language that - to put it mildly - was not designed with that in mind. In that sense it's a constrained form of writing not unlike a sonnet or a limerick. Fudging your cryptic grammar is a bit like having too many or too few syllables in your sonnet, or not quite nailing the rhythm and rhyme scheme of a limerick. I'm left po-facedly grumbling "sure, we can all write natural surfaces if we ditch the rules".

On the other hand, dropping/adding syllables or entire lines is exactly what Shakespeare did with some of his sonnets, and a healthy disregard for certain Ximenean principles didn't stop Araucaria becoming one of the most beloved setters ever. But on a third hand, hardly anyone has ever been Shakespeare or Araucaria, so for more aesthetically pleasing clues and fewer disgruntled solvers, a safer bet is to follow Ximenean principles wherever possible (except for all the ones I ignore, of course).

Back to top

On cryptic time

Cryptic clues have to be very short stories and instruction manuals at the same time, but time itself works differently in the two levels. The surface reading is like a story, and this story can be set in the past, present, future or have no real temporal setting at all. The cryptic reading, on the other hand, is the solver's instruction manual (or maybe a product spec), so I think most setters would say it should really be written in some sort of present tense.

The cryptic reading should either describe the wordplay which is patiently waiting to be discovered by the solver at this very moment or give the solver a set of instructions so they can create the answer as if for the first time. This means the cryptic reading should never be "set in the past", as if it happened another day and the solver missed it. To be less abstract, here's a table showing how the cryptic grammar of a clue might be rewritten in various ways (the answer is MAIN):

Back to top

On etymological overlap

Ideally, clues should lead the solver up the garden path before safely depositing them at the solution. Clues can be unsatisfying if any of the words used in the definition or as part of the wordplay are so etymologically close to the solution that it seems like the setter's made things too easy. You want to give solvers the breadcrumbs, not big chunks of bread. Consider these clues (definitions are underlined). 

Someone going flying ruined daft talent night (6,9) FLIGHT ATTENDANT
Anagram ("ruined") of DAFTTALENTNIGHT 

"Flight" in the solution is so obviously derived from the same root as "flying" that the solver doesn’t get much of a sense of achievement from cracking this clue. “Someone doing routine in aisles” would define the solution without any etymological overlap as well as neatly misdirecting the solver into thinking of aisles in a theatre.

Political plans about oil and spice (8) POLICIES
Anagram (“about”) of OIL+SPICE

It's the same with “political” and “policies” here. “Legislative plans” would work just as well without giving too much away to the solver.

Somewhere to work with excellent sort of computer (7) DESKTOP
DESK ("Somewhere to work") + TOP (“excellent”)

Etymological overlap can tarnish the wordplay too. Everybody knows that "desktop" combines "desk" and "top", so it's not very satisfying for a cryptic clue to serve up those original pieces and simply ask you to fit them back together again. It's much more interesting to find some new way of looking at the word.

Sort of coffee from French teashop unfinished (5) DECAF
DE ("from French") + CAF[e] ("teashop" "unfinished")

This one's doubly bad: the "caf" in "decaf" is obviously related to "café", so the wordplay is underwhelming, but "decaf" and "caffeine" ultimately derive from "coffee" too, so using "coffee" in the definition is a bit rubbish as well.

Eco-friendly Windsor possibly establishing garden structure (10) GREENHOUSE
GREEN (“environmentally-conscious”) + HOUSE (“Windsor possibly” - as in the royal house)

Note that this one is much more creative and interesting to solve. That's because the secondary senses of “green” and “house” used in the wordplay are one or two semantic steps away from the senses of these words which are relevant to the meaning of "greenhouse".

It’s worth paying particular attention to this sort of thing when writing double definitions. Obviously many double definitions involve two senses of basically the same word, but it’s good to make sure there’s some healthy semantic distance between the senses you're using. Something like Imitate tropical bird (6) is a bit rubbish for PARROT, because the sense of "parrot" which means "imitate" is just a familiar metaphorical usage of the literal, birdy sense. Quick popstar (5) is more satisfying for SWIFT, because although these two senses are etymologically connected if you go back far enough, they at least seem far apart. The further apart two senses, the more fun it is to find them lurking in a double definition. The closer together the senses, the more underwhelming the clue.

On a similar topic, I always think that if you're cluing a solution which contains a small grammatical function word like a preposition, article etc (e.g. OVER THE MOON, GIVES IN or BUBBLE AND SQUEAK) you should make sure you don't use that word as part of the clue (e.g. if the solution contains IN, don't use "in" anywhere in your clue). To me, that just feels tighter and more disciplined.

Back to top

Other quick tips

Redundant words

One Ximenean principle that most crossworders would endorse is that every word in a clue needs to earn its place there as part of the wordplay or the definition. Any additional words which are not vital for either of these are at best unnecessary and at worst actively unhelpful to the solver, so should be cut. In the clues below, the underlined words do nothing except make the surface more meaningful. Such words should be cut even if it means binning entire clues. Pay particular attention to how you use articles (“a”, “an” and “the”) – these are almost always unnecessary in cryptic reading of the clue so in my view should only be included if they make the surface reading clearer or more natural. 

Boss gets drunk – he cries (4) SOBS
Anagram ("gets drunk") of BOSS. Definition='cries'

Characters from Sudan despondent about mountain range (5) ANDES
Hidden ("characters from") in "sudAN DESpondent". Definition="mountain range"

If you're sharing online or submitting to an editor you'll probably want to provide explanations for your clues. I recommend breaking down your clues with some consistent annotation system like the one I've used above. That helps you identify any words that aren't doing anything important, and it's also a good way to catch typos or faulty wordplay.

Back to top

Link words

Most clues will consist of a definition and a bit of wordplay. The definition must be at the very beginning or the very end of the clue, so usually the wordplay is adjacent to the definition, with an invisible dividing line which solvers must sniff out. It's sometimes acceptable to plant a signpost in this gap in the form of a link word, as long as it joins the wordplay and definition in a plausible way. It's especially important to think about which “direction” the link is going in. Imagine replacing the link word with either an equals sign or an arrow joining the wordplay and definition - remembering that the definition essentially represents the answer the solver should end up with - and check that the link makes some logical sense. Remember that setters and solvers have different views on which links work, which is one reason it's best to do without them if possible (links, that is).

This table might help (Viresh Ratnakar has done a similar but much more comprehensive one here):

Back to top

Definitions by example

Definitions are often imprecisely described as synonyms of their solutions, but a word for a category of things can also define a more specific category within that category, or individual members of that category (e.g. "book" can define a more specific category like NOVEL or an example of a book like EMMA). Similarly, verbs and adjectives can define a solution which means a more precise kind of action or description (e.g. "perform" and "patterned" might define SING and STRIPEY respectively).

This is probably all fairly obvious, but it might be less obvious that an example of a category is not considered a fair definition of that category (so "Emma" on its own can't define NOVEL), and that a more precise verb or adjective is not a sufficient description of a less precise verb or adjective ("sing" and "stripey" on their own don't really define PERFORM and PATTERNED). If you want to use a specific instance to define a category, the convention is to use a definition-by-example indicator ("Emma, perhaps", "possibly sing", "stripey maybe", "Emma, for one" etc) to let the solver know the solution they're looking for is something broader in meaning than the definition rather than narrower or equivalent in meaning as would normally be the case. A question mark directly adjacent to the example ("Emma?") does the job with minimal disruption to the flow of the clue.

It's worth saying most setters and solvers expect definition-by-example indicators whenever a clue uses the name of a famous person (e.g. if you use "Tom" to define CRUISE or vice versa). In my eyes Tom Cruise is a member of the category of Toms and a member of the category of Cruises just as he's a member of the category of actors, so I don't feel any strong need to provide a DBE indicator, not least because such an indicator doesn't actually make the solver's life any easier. "Dog, perhaps" warns you that you need a parent category of "dog" rather than a synonym of it, but is "Anne, perhaps" really any more informative or helpful to the solver than "Anne"? What else could Anne mean except an individual with that name? All a DBE indicator is going to do here is clutter up your surface. That said, I generally include DBE indicators for names because I know people are used to them.

Back to top

Punctuation and capital letters

The first thing to say is don't put full stops at the end of your clues - I've seen a few new setters do this, but it always looks weird. Other punctuation marks can be used to improve the surface reading of the clue without necessarily having any function in the cryptic reading. It's generally alright to expect solvers to mentally delete such punctuation marks when deciphering the clue, although clues which don't require this step are often more elegant.

One exception to the above is possessive apostrophes, because these change the meaning of a word, not just its relationship to the words around it. If a definition (or synonym for part of the wordplay) includes a possessive apostrophe, the solver is entitled to take it at face value. Using "Trump's" (rather than "trumps") to define a solution like BEATS or CONQUERS is a step too far. The reverse is also true: using "dogs call" to define WOOF is unfair, as solvers should not have to mentally insert an apostrophe in order to turn this into "dog's call".

For some reason the convention with capital letters is that it's OK to sneak capital letters into words which wouldn't normally have them (you can use the verb "march" but write it "March" to make solvers think of the month), but not OK to withhold capital letters from words which would normally have them (you can't write "polish" when you really mean "Polish"). Note that the first letter of a clue will always be capitalised, so you can sneakily put "march" or "polish" at the beginning of a clue in order to gently mislead solvers within the rules of the game.

Final thoughts on punctuation: you can use question marks to indicate definitions by example (see above) or to warn solvers your clue is dubious or cheeky in some way. Similarly, exclamation marks can be used to flag up something unusual, e.g. to show the clue is an &lit, but too many exclamation marks can start to look like you've given your favourite clues a big tick rather than letting them speak for themselves. For this reason I try to ration exclamation marks unless the clue I'm writing contains a word or phrase like "that does it!" or "blimey!" which would just look odd without one.

Back to top

Dictionaries

When setting a new puzzle, it's often worth checking your understanding of what a word means against a proper dictionary definition of it (this is also a good way to find useful synonyms etc). Historically, many crossword editors and solvers have expected that any word sense involved in a cryptic clue should be verifiable using Chambers dictionary (a free but incomplete version is available at chambers.co.uk). In particular, many solvers get tetchy if a clue uses an abbreviation not backed up by Chambers.

Personally I think treating "the big red book" as gospel is a bit outdated, and leads to a slightly absurd situation where some incredibly obscure “zombie” abbreviations continue to lurch around while other abbreviations which would be far more familiar to the average person alive today are off-limits. I feel I've done my job properly if I use a word sense or abbreviation that I reckon most people can reasonably be expected to know, Chambers or no Chambers, whereas I feel I've let solvers down if I ever rely too much on fossilised obscurities which only seem to exist in the pages of Chambers. Part of this is that I think it's important to attract new solvers, and I'm more worried about putting off novice solvers by expecting them to own or buy a particular dictionary than I am about winding up established solvers by occasionally trusting them to Google something.

I do appreciate the argument that having Chambers as the source of truth is the only real way to be fair to solvers: a standard list of established definitions and abbreviations which anyone can consult would be the only way to put all solvers on a roughly level playing field, and this is more of an issue if a crossword has a cash prize or is a more advanced puzzle demanding significant time from solvers. But I'd guess many new solvers have never heard of Chambers, and there's often nothing in the newspaper to point them towards it, so in practice this universal list can easily become a mysterious and slow-to-update resource for those already in the know. For this reason I think that when it comes to daily puzzles at least, editors, setters and solvers can all use some common sense in letting some newer abbreviations in from time to time, and I'm glad to say that some editors are more relaxed on this kind of thing, especially if another well-known dictionary like the OED or Collins supports something which Chambers doesn't.

Back to top

Fairness

You should always be asking whether you’re giving the solver a fair chance, although that doesn’t mean you can never be devious or tough - some setters or puzzles are just more difficult than others, after all (and some end up with a reputation for difficulty without meaning to). But whatever the overall difficulty of a puzzle, its most obscure or unfamiliar solutions should generally be given its most generous or straightforward clues. It's a bit unkind to use an anagram if the solution is difficult to spell. If a word is so obscure that many solvers will never have heard of it, you can be especially generous by making it a hidden clue so it’s right under their noses. The reverse kind of holds true as well: if a definition is likely to be a dead giveaway, then you’ll probably be forgiven for tightening the screws a bit more in the wordplay. It's very difficult to see your clues the way a solver would see them, so it's a good idea to have someone test-solve your puzzles before publication if possible.

Back to top

Other links and resources

Michael Callaghan's guide to the construction of cryptic crossword clues
A comprehensive guide to constructing crosswords written by Michael Callaghan aka Prolixic as part of his involvement in Rookie Corner at Big Dave's crossword site (see below).

Alberich Crosswords
This site collects writings by Neil Shepherd aka Alberich which were available on his own website before his death in 2022. There's a lot more detail on Ximenean cluing to be found there.

Cryptic Grammar by Gussalufz
This is a very clear article on cryptic grammar by Viresh Ratnakar aka Guffalufz who sets for the Hindu Times. Guffalufz also set up Exet, a site for creating crosswords online (including 3D ones).

Fifteensquared / Big Dave's Crossword Blog / Times for the Times
Reading what the bloggers make of each day's puzzles on these sites (and what solvers think of them in the comments below) is probably the single best way to get a thorough understanding of the various conventions followed by most UK setters working today. Big Dave's site also includes Rookie Corner, where budding setters can submit a puzzle for detailed feedback from bloggers and commenters.

MyCrossword
Tom Blackwell aka Raider's site is a great place to publish crosswords online because as well as having a stylish and easy to use interface the website has created a real community of puzzle enthusiasts who are happy to provide feedback in comments.