The Setting Room The bits that make up a cryptic crossword

Any abbreviation that is in the latest edition of Chambers can be used in a barred puzzle. For blocked puzzles, some of the more obscure abbreviations given by Chambers (eg D = deserted) are unlikely to be allowed by editors, but some of the glaring omissions, such as S = small, are generally accepted (and will also be allowed in some barred puzzles). There is no definitive list for blocked puzzles, but the list of abbreviations allowed in Telegraph puzzles will give you a good idea of what is available.

The guaranteed way to indicate an abbreviation is to use exactly the word or words which are abbreviated, eg 'learner' for L. Where the dictionary gives a Latin word and its English equivalent, either can be used, so 'recipe' and 'take' are both valid for R. Beyond that, it becomes more complicated, and editors are likely to be influenced more by established practice than any other consideration, particularly for abbreviations of two or more letters, so 'artist' for RA will never be rejected, and I have seen (for instance) 'from the orient' used to indicate E (ie Eastern); that seems to me a bridge too far. In general, though, abbreviations which 'have a life of their own' (in other words they are likely to be used 'as is' in written or spoken English) offer greater flexibility. So 'record' for LP ("I have that LP") and 'gym' for PE ("I never enjoyed PE") are fine, while 'lot of power' for MW (megawatt) would be questionable at best.

In one sense, yes. Where the abbreviation represents a word always spelt without an initial capital, it is considered rather weak to indicate them using a word beginning with a capital. SI units give the most problems in this regard: sometimes we'd like to use 'Henry' for H, but its actually an abbreviation of 'henry'; 'stokes' is similar. The neat way round is to put the word at the start of a sentence, eg "Henry consumes roasts" for HEATS. However, since in general it is allowable to unnecessarily capitalize a word, a clue like "See Stokes with bag" for SPOT is unlikely to be rejected.

Unless the words abbreviated are proper nouns, as in LA for Los Angeles, it is always acceptable to use uncapitalized indications, irrespective of whether Chambers shows the word(s) abbreviated as starting with capitals. So "Foremost tree on street" is fine for FIRST, although St is only ever seen as an abbreviation for Street (with a capital) in the names of individual streets. Since in general it is unacceptable to remove the initial capital from a proper name (eg to use 'nice' when you mean 'Nice'), this is actually something of an anomaly.

The British Territorial Army ihas been known since 2013 as the Volunteer Reserve. The abbreviation TA is in Chambers (and will surely remain there), so indicating TA with 'Territorial Army' is legitimate. However, the use of 'volunteers' on its own for TA is now considered inappropriate; it would need to be 'volunteers once' or the like. The same applies to 'teachers' for NUT, that union having ceased to exist in 2017, and several other Chambers entries.

The rules vary from puzzle to puzzle, but I suggest that in general if you can avoid using abbreviations (eg PCSO, ICBM) as answers you should do so. In puzzles where there is no requirement to advise solvers in the notes that, for instance, "The Chambers Dictionary (2016) is recommended; the answer at 31 is an abbreviation", you should give the solver as much help as possible in the clue, so RSVP might be 'Short request to reply".

Note that Chambers and Collins have widely divergent views when it comes to abbreviations which have become part of the language, such as AWOL or RADA; Collins considers these (and many others) to be words, while Chambers shows them as abbreviations. This means that in blocked crosswords such as the Times puzzle, where Collins is the primary reference, AWOL would be a perfectly valid answer, enumerated as '(4)'; in a barred puzzle where Chambers was the primary reference, it would be enumerated as '(4)' but solvers would have to be advised that the answer to that particular clue was an abbreviation. This does not fundamentally affect the use of these abbreviations in wordplays, but while 'drama school' would unequivocally be valid for RADA in a 'Collins' puzzle, the editor of a 'Chambers' puzzle might consider that only 'Royal Academy of Dramatic Art' was allowable, although I would hope that common sense would prevail when it comes to any puzzle where 'artist' is accepted for RA.

There are a few acronyms which are shown by Chambers as nouns, examples being 'MASH' and 'NIMBY'.

'New' is universally acceptable for N and 'second' or 'seconds' for S. But what about 'news' for NN or 'seconds' for SS? These fall broadly into the 'whimsical' category, along with, say, 'detailed' as a last letter reduction indicator, which means that they are unlikely to be accepted for puzzles which insist on a high degree of accuracy in terms of the English language, but may well be allowed in puzzles where a little whimsy is embraced.

Something like 'time after time' is clearly valid for TT, and 'repeatedly caught' is surely acceptable for CC. 'Two days' is perhaps slightly better than 'days' for DD, since it specifies the number of repetitions, but the real issue is that D is an abbreviation of 'day' but not 'days'; 'two seconds' could perhaps be interpreted as 'two instances of the abbreviation for seconds', and thus yield SS, but the difference between this and 'two days' is very small indeed.

A clue is ambiguous when it could legitimately lead to two or more answers of the stated length. An example would be "Character knocked back gin (4)" - this is a valid clue for TRAP (PART reversed), but is equally good for PART (TRAP reversed). In general, ambiguity in clues should be avoided, as it can be frustrating to solvers, but it only becomes a major problem when it can't be resolved by getting all the available checkers. So for a four-letter light with all except the third letter checked, as soon as any crossing clue was solved they would know whether TRAP or PART was the intended answer. But the clue "Level increase reported (4)" is a different matter; even when we have RA?E, we don't know whether the homophone for 'raise' is RASE or RAZE, since they are equally valid. The solver can only guess at the intended answer, and the clue is therefore faulty.

There are two situations which account for nearly all ambiguities. These are reversal and homophone clues of the form <definition> <indicator> <definition>, although with homophones the problem only arises if both defined words are the same length. So "Character knocked back gin (4)'"could lead to PART or TRAP, and "Man picked up post (4)" is valid for both MALE and MAIL; "View broadcast location (4)", however, can only be a clue for SITE, since SIGHT is the wrong length.

Note that the ambiguity only occurs with reversals where both 'definitions' lead to single words. "Second character knocked back gins (5)" can only be a clue for TRAPS.

Ambiguous clues can usually be rephrased in a way that eliminates the problem. So "Character knocked back gin" could become "Gin character knocked back" for TRAP or "Mouth reflects character" for PART (the subject of the indicative verb can only be the word preceding it). "Post man picked up" is a clue only for MAIL, while in "Post picked up for man" the link word 'for' means that it must be the post that is being 'picked up' so the answer is MALE. Based on questions that get asked on help forums, some solvers get confused about which bit of a homophone clue is which, but as long as you have done your job properly as a setter then all is well.

An anagram consists of an anagram indicator (also known as an 'anagrind') and anagram fodder, which is normally visible in the clue. At the simplest level, this would be something like 'disturbed night' for THING, or "he's not upset" for HONEST. However the fodder can include letter selections, so 'disturbed sleep initially annoying' would be fine for PLEASE. It is also acceptable to use conjunctions to link elements of fodder, as in 'mixed tea and rum' for MATURE or 'cat playing with tin' for INTACT; note that in the second example one part of the fodder comes before the anagrind and another part after it.

The anagram can of course be used in combination with other devices, as in 'son wearing loose coat' for ASCOT.

There is no hard and fast rule here. In general, single letter abbreviations are accepted, so 'second name changed' would usually be allowed for MANSE. Abbreviations of two or more letters, such as 'Welsh artist confused' for WHALERS (fodder = WELSH + RA), are distinctly questionable

An indirect anagram involves fodder which is not visible in the clue. "Directed nice film (4)" is a perfectly fair clue for CINE, but "Directed good film" involves an indirect anagram - the solver must first turn 'good' into 'nice' and then rearrange the letters of the latter. This type of clue was quite commonly seen in the 1960s and 1970s, but has been outlawed for many years, at least in the 'good'/CINE form. One could argue that an anagram where the fodder includes an abbreviation is indirect, since not all of the fodder is on view, but a certain leeway is given when the possibilities for the indirect element are severely limited - in "English chap boiled fruit (5)" for PEACH, it is pretty clear that English = E will deliver the fifth letter required for the fodder. The clue "French capital staggers couples (5)" for PAIRS involves an indirect anagram of sorts, but there is only one capital of France and a clue like this would be allowed by some editors.

In a subtractive anagram there are two groups of fodder, with one  group (y) being removed from the other (x). Convention dictates that when the letters to be removed are not to be found consecutively, and in the correct sequence, within x, then two anagram indicators are required. So while "Hurts deplorable partisan dismissing art (5)" for PAINS, an anagram of P(art)ISAN, requires only one anagrind, giving (x-y)*, "Face awful tirades after dismissing art carelessly (4)" for SIDE requires two anagrinds, giving (x* - y*). The logic behind this is slightly questionable, since a rearrangement of TIRADES could give SIDEART, from which ART can be removed 'as is', but that's how it works. The length of the answer in a subtractive anagram will be (length(x) - length(y)).

A composite anagram is not unlike a subtractive anagram, except that instead of the 'equation' leading to the answer, z, ie z = x* - y*, it is (z + y)* = x*. Those who know their algebra will say that this comes to the same thing, and they would be right, except that in a composite anagram the definition is mixed in with the wordplay. So 'Master could be confusing this brilliant person with me (4)' is a composite anagram for STAR, since the solution ('this brilliant person') plus ME could be rearranged to form MASTER. Composite anagrams often have only one anagrind, but (as here) that is combined with an indication of potentiality, such as 'could be'. The important thing is that solvers need to be given a pointer towards the correct parsing, so 'Master confused this brilliant person with me' would not be acceptable. The composite anagram often lends itself to '&lit' clues: 'Fancy that clues could make me act (6)' is a composite anagram &lit, where  an anagram ('fancy') of THAT CLUES could produce SLEUTH ('me', ie the answer) ACT.

Anagram indicators are usually adjectives (eg 'bad') or inflections of verbs ('shifts', 'moving', 'damaged'). Adverbs (eg 'carelessly') are acceptable, presumably on the basis that the solver can infer that the fodder is the subject of a verb such as 'behaving'. Noun anagrinds require careful handling; the construction '<fodder> <noun anagrind>' is generally considered unacceptable, so 'Eton mess' is not a valid wordplay for TONE, since 'Eton mess' is not the same as a 'mess of Eton'. Typically a conjunction is required, so 'Eton in mess' would do, as would 'Disturbance in Eton'. The question that you must ask yourself as a setter is whether the expression that you have written could reasonably be understood to mean that the fodder is to undergo rearrangement.

A look at the relevant Clinical Data section will give you a good idea what inflections are commonly used. Present participles of intransitive verbs (eg 'struggling') are extremely flexible when it comes to the grammar of the surface and cryptic readings, as are past participles of transitive verbs (eg 'damaged'). Present indicatives (eg 'blunders') are also absolutely fine.

Two verb forms which suggest that something is going to happen - presumably soon - are usually allowed. The infinitive (eg 'to change') is one, so 'Bros to reform' is valid for ROBS; the other is the future tense (eg 'will evolve), so 'Meal to cook' would work for LAME.

Past tenses which suggest that the disturbance occurred at some point but there is no reason to believe that its effects are still visible are not allowed (eg 'danced', 'was modified', 'had buckled', 'had been altered').

The opportunities to use verbs in the imperative mood are relatively infrequent, since they offer very little flexibility in terms of the syntax of the wordplay, but they can lead to some neat (and deceptive) surface readings, in which they appear to be nouns (often in an attributive sense, modifying another noun, as in 'buffet car' as a cryptic indication of ARC). Particular favourites of setters are 'police' and 'school' - 'Police sergeant' would indicate an anagram of 'sergeant', 'School rules' likewise of 'rules'. Imperatives must be used with care, however - in practice, they will almost invariably need to be the first word in the wordplay.

When the anagram fodder is a single word, as in eg 'snow drifts' for OWNS, the verb is always in the singular. When the anagram fodder clearly consists of two or more parts, as in eg 'Ant and Dec rock' for CADENT or 'big, red boil' for BRIDGE, the verb must be in the plural. An issue arises when a verb is governed by a single string of two or more words, such as 'see car'. There is no doubt that a single string like this can take a singular verb, so 'see car crashes' would work for CREASE; however, there are those - including Azed - who believe that a string of words can also legitimately given a plural verb, eg 'see cars crash' for CREASES. I believe that this is at odds with 'real world' English, and should not be allowed. The use of the participle verb form usually offers a way round this issue, as in eg 'see cars crashing' for CREASES.

Note that words that are manipulated in clues are always treated as being in the third person, which can on occasion be counterintuitive. Hence 'meal I am cooking' might look ok for EMAIL, or 'tart you cook' for OUT-TRAY, but to be cryptically sound these would need to read 'meal I is cooking' and 'tart you cooks'. The usual workaround is to use the future tense, where person becomes irrelevant, ie 'meal I will cook' and 'tart you will cook'.

In other words, should setters try to avoid anagrams? Some setters do just that, and I have set puzzles in the past without anagrams.

Anagrams and letter selections offer ways to introduce words into a clue whose only contribution to the cryptic reading is the letters which they contain; this makes them very useful in terms of producing compelling surface readings, but arguably they score poorly in terms of technical merit. There is no doubt that anagrams and letter selections can be overdone, and a puzzle with a large number of obvious and simple anagrams is likely to provide limited entertainment. However, a few well-disguised anagrams add variety to a puzzle, and don't forget that they can be combined with other constructions, such as charades or insertions. It's also worth remembering that solvers need to be given some 'ways in' to a puzzle, and a couple of straightforward anagrams can be just the thing.

Custom dictates that each sentence within a clue should start with a capital letter. The only exception is where a pair of clues are joined by ellipses (...) - here the second clue will begin with an ellipsis, and the first word should be treated as though it were in the middle of a sentence, ie it will not have an initial capital unless it is a proper noun. Normal rules of capitalization apply to clues - words containing capital letters should appear just as they would in an English sentence (eg Orion, Rolls-Royce, eBay, ID card, IOU).

An example of this would be "Way to stroke Henry (4)" for PATH [PAT + H], where the name of the SI unit, henry, has been deceptively capitalized to suggest the name 'Henry'. Purists (among which I would count myself) consider this to be less than ideal, but it is generally considered acceptable. The argument usually put forward is that words which do not generally warrant an initial capital could on occasion be seen with one, eg at the start of a sentence or in a book title.

The clue "Victor is in nice undergarment (4)" for VEST [V + EST] requires that 'nice' be interpreted in the wordplay as 'Nice'. The wilful removal of a capital letter is generally deemed to be unfair to the solver and thus not acceptable. There is an exception to this - when using words which lead to an abbreviation, even when the abbreviation is for a capitalized word the capitals can be omitted from the indication, eg while R is an abbreviation for 'River' (on maps), 'river' is frequently seen in puzzles indicating R.

The expectation is that answers will always be entered entirely in capital letters (with accents, apostrophes, hyphens and spaces removed). Any deviation from this rule in a particular puzzle would have to be clearly identified in the preamble.

In terms of crossword clues, a classified word is one which is qualified in certain ways in the primary reference for the puzzle. Chambers Dictionary uses a considerable number of classifications, including 'obs' (obsolete), 'Scot' (Scottish) and 'poetic'; the most important ones are covered in the next part of this article. Some of these qualifiers, in particular 'obs', indicate words which are not in use in modern English, and in fairness to the solver these words should be flagged as such. Other qualifiers, such as 'inf' (informal), are usually applied to words which are in common use, particularly in conversational English.

Chambers uses a lot of classifications, some of them additionally qualified, eg 'esp N Am', meaning 'especially (but not exclusively) North American'. The important ones for crossword setters are listed in the following sections.

The classification may be applied to the headword, eg 'corsive (obs)', indicating that all entries and meaning under that headword are obsolete; to a particular entry underneath a headword (a subhead), eg 'Old Tom n (archaic)', indicating that all senses of that entry are archaic; or to a particular meaning (each individual meaning being separated by semicolons), eg 'felon¹ n a person guilty of a felony; a wicked person (obs)'.indicating that just the 'wicked person' sense of the entry is obsolete.

Classifications are additive, so a subhead shown as 'obs' under a 'Scot' headword is both Scottish and obsolete. Sometimes the classifications themselves are qualified, using abbreviations in italics such as  'esp' (especially) and 'orig' (originally), eg 'hornswoggle (inf; orig and esp US ). Since these imply that the word has been assimilated into British English, there is no need for setters to flag them, so HORNSWOGGLE can be defined simply as 'trick'.

These form a large group, dominated by 'Scot' (Scottish), 'US' (United States) and 'N Am' (North America).

When the answer to a clue carries one of these classifications, convention requires that it be flagged as such. This can be done in a variety of ways, either explicitly (eg 'Scots dispute' for CANGLE), or implicitly, either using a place (eg 'dispute at St Andrews') or a person (eg 'Sturgeon's dispute'). It can also be achieved by establishing an appropriate context in the clue, eg "introduction to Trump and his ruthless reduction of expenditure burden (3)" for TAX, 'ax' being the North American spelling of the word.

The geographical classification may be accompanied by 'dialect', eg 'N Eng dialect'; such a word should be indicated by reference to that particular area.

Chambers uses the classifications 'obs' (obsolete), 'archaic', 'rare' and 'hist' (historical) for words that are no longer (or never were) in common use. Those classified as obsolete are defunct as far as the language of today goes, and must always be flagged as such, typically using adjectives or adverbs such as 'ancient', 'former' or 'once' They can also be flagged implicitly, often through the use of a verb in the past tense, eg 'It imitated velvet' for MOCKADO. Expressions classified as 'rare' are flagged using qualifiers such as 'rare' or 'uncommonly'. Editors usually expect words shown as being archaic or historical to be flagged in a similar way to an obsolete word, although strictly speaking such words are still (just) part of the language.

Words are often attributed to a particular author, playwright or poet (although in many instances they were not used solely by that person). The most common examples are 'Shakesp' (Shakespeare) and 'Spenser', but there are many others. Convention dictates that such words must be flagged to the solver, but this can be done either by using an 'old' qualifier, as for obsolete or archaic words, or something slightly more personal - for Shakespearean words this could be, for instance, "poet's", "Bard's", or 'for Will'.

Where the person to whom the word is attributed is not English, eg Robert Burns or Walter Scott, the 'old' qualifier on its own cannot be used, because the word is, in effect, both obsolete and Scottish; typically this sort of word is qualified in definitions by 'for Walter' or the like.

Chambers classifies certain words as 'dialect' without specifying a locale, eg 'cauf (dialect)'. It is usual for such words to be flagged, typically through qualifiers such as 'local' or 'in places'.

Classifications which some editors require to be flagged are 'poetic', 'literary', 'facetious', and 'jocular', although there is no consensus on this. Some of the words or senses  which carry these classifications, such as 'garner', will be familiar to most solvers, although others, such as 'grot' for a grotto, will not. I think there is scope for applying discretion here, and using "poet's", 'in book', 'humorously' etc only where it seems appropriate.

It is generally accepted that words classified as 'slang' or 'inf' (informal) do not need to be flagged. Note, though, that a word shown as, say, 'old slang' needs to have its oldness indicated to the solver. The same applies to technical classifications such as 'archit' (architecture), 'phys' (physics), and 'milit' (military).

Chambers  gives certain words two related classifications, eg 'darg (Scot and  N Eng dialect)', or alternative classifications, eg 'morrow (archaic or poetic)'. In each instance, the setter can choose either classification, so darg could be 'hard work for Jock' or 'hard work for Geordie'.

Expressions classified by Chambers as 'offensive' or 'taboo' should be studiously avoided, whether as an answer or as part of a clue. Even the use of a different word with the same spelling as a word so classified is likely to provoke comment.

Setters should think very carefully before including words, or meanings of words, shown as 'derog' (derogatory); I will not use the derogatory sense a word in my own puzzles because of the potential for offence.

Some editors consider that a qualifier is required for definitions, but not for words which have a cryptic role in the wordplay – so ‘beyond’ as a definition of ‘without’ would require qualification (this sense of the latter word being shown as ‘archaic’), but ‘x without y’ to indicate containment of y by x would be acceptable, despite this sense of ‘without’ also being archaic. This strikes me as inconsistent, since the challenge for the solver is essentially the same.

Chambers contains a number of expressions that have their origins in other modern languages but which have been assimilated into English, such as 'risqué' and 'angst'. Others, however, have not achieved this level of integration. Chambers classifies words such as 'ami' (French) and 'echt' (German); these must be flagged in the same way as any other geographically-classified word would be.

Certain Latin words, typically those found in expressions, are also classified by Chambers as '(L)', eg 'ut' (meaning 'as'). These also should be flagged, using a qualifier such as  'from Rome'.

Common errors made by setters usually relate to the syntax of the wordplay. These can be hard to spot when the surface reading is sound and the cryptic reading 'looks' right.

1. Pronouns governing a verb in the wrong person

This most often occurs with the pronoun 'I'. The clue "I'm taken in by fellow chief (4)" looks good for MAIN (I in MAN), but in the cryptic reading 'I' is not a person but a letter, so the wordplay grammatically requires "I is taken in by fellow chief". The simple workaround is to use the future tense, eg "I will be taken in by fellow chief".

2. Plural nouns with verb indicators

Similar to (1), this issue is seen in a clue like "Ladies shape standards of perfection (6)" for IDEALS (anagram of LADIES). In the cryptic reading, 'ladies' is a singular word, irrespective of how it might appear. So the clue would need to be "Ladies shapes standards of perfection", which makes no sense. The present participle is the answer to this one, eg "Ladies shaping standards of perfection", catering for either singular or plural subjects.

3. Indicative tense with multiple manipulations

The clue "Fence breach holds nothing back (6)" might appear to be valid for PALING (GAP around NIL, all reversed), but the indicative form of 'hold' means that this can only mean 'GAP around (NIL reversed)' and not '(GAP around NIL) reversed'. As so often, the answer is the participle: "Fence breach holding nothing back" could lead to either PALING or GALINP, but the definition ensures there is no ambiguity.

4. Defining a transitive-only verb as an intransitive

A legitimate definition of EDIT could appear to be something like 'make changes'. This is one of those situations where the substitution test can be used - is 'edit the document' the same as 'make changes the document'? Clearly not, with 'edit' being transitive (it requires an object) and 'make changes' intransitive (it doesn't take an object). Possible definitions would be 'make changes to', or a transitive verb like 'correct', 'alter' or 'amend'.

5. Relying on the transitivity of meanings

This is covered in more detail in the Definitions section, but the mistake here is to assume that because A has a meaning of B, and B has a meaning of C, that A and C share a common meaning. For instance, 'grasp' means 'to seize and hold', so 'Fellow grasps English plan (4)" is fine for MEAN (MAN around E), and 'to understand' means 'to grasp'. However, 'understand' and 'seize and hold' are not synonymous, and the clue "Fellow understands English plan" for MEAN is unsound.

The rules for barred crosswords are pretty clear, and while an absence of symmetry (at least in themed puzzles) or the presence of entries with an unusually high proportion of checked letters will not get a puzzle rejected, there are two things that will:

1.  Two or more consecutive unchecked letters in a grid entry.

2. The inclusion of entries which break the Ximenean 'unch' rules, ie with more than 1 unchecked letter in a 4-letter or 5- letter entry, more than two unches in a 6-letter or 7-letter entry, more than three unches in an 8-letter, 9-letter, 10-letter or 11-letter entry, and more than four unches in a 12-letter or 13-letter entry.

In blocked puzzles, the errors which new setters are most likely to commit are:

1. Including more than two consecutive unches in an entry (two are usually allowed as long as at least half of the letters in the entry are checked).

2. Using a grid which does not have (at minimum) 180-degree rotational symmetry.

The answer seems pretty obvious, given that there are thousands of definitions in Chambers. But you'll find plenty of 'definitions' in puzzles that you won't find in any dictionary, and that's because the cruciverbal term 'definition' describes any element of a clue which leads the solver direct to the solution, in contrast to 'wordplay', which involves cryptic manipulations. So in crossword terms, a word such as SAILOR could legitimately be 'defined' as 'mariner' (a Chambers definition), 'tar' (an informal term), or 'main user' (an oblique definition). It could not, though, be 'defined' by 'last out', an anagram of SALT, which then leads to SAILOR; this involves a manipulation (in this instance an anagram) and therefore does not qualify as definition.

The clue "Deliver first-class service to American church (3)" for ACE (A + CE) contains a definition which could be described as 'oblique', but the clue still consists of a definition and a cryptic wordplay. A cryptic definition clue contains only a definition - this is designed to lead the solver in the wrong direction, but when interpreted correctly should lead unambiguously to the answer, at least when combined with the letter count. An example would be "Close it and you save the game (6)" for SEASON. This sort of clue is still allowed in some blocked puzzles, but since the mid-1970s 'cryptic def' clues have been outlawed in barred puzzles, where the rule is that every clue must offer at least two different routes to the answer. The definition in the SEASON clue could still be used, but it would have to be accompanied by a wordplay.

Consider these two, non-cryptic clues:

1. Dog (6), ‘C–L–I–’
2. Collie (3), ‘D–G’

The answer to (1) is definitely COLLIE. But can the answer to (2) be DOG? Well, for (1) we would ask ourselves whether a collie is a dog, and the answer is an unequivocal "yes". But when we come to (2), is a dog a collie? Clearly not all dogs are collies, and therein lies the problem. Collies constitute a subset of the class of ‘dogs’, and therefore a collie is an example of a dog. There is a school of thought which says that there must be something in the clue to show this, where clue (2) would become something like:

3. Collie, for instance (3), ‘D–G’

Clue (2) is an implicit ‘definition by example’, whilst clue (3) is an explicit one. Whilst it cannot be wrong to explicitly identify definitions by example, either by the use of qualifying words such as 'maybe' or 'perhaps', or by following the DBE with a question mark, many editors will not insist on it. The question of fairness to the solver should be the  paramount consideration: in the clue "Join girl with former partner (5)" for ANNEX (ANN + EX), girl = ANN is not a definition by example, but there are a great many girls' names from which to chose, while in "Delia goes after dismal forger (10)" for BLACKSMITH (BLACK + SMITH), 'Delia' is a definition by example, but how many Delias are sufficiently well-known to appear in a crossword clue?

Note, though, that Azed insists on definitions by example being flagged, so when writing clues for Azed competitions you must indicate them.

Consider the clue "Terrier frisky in car (5)" for CAIRN (anagram of IN CAR). Chambers gives one meaning of 'cairn' as 'a small variety of Scottish terrier', so this is a perfectly normal definition. But in the clue "Vera found in pedalo, exhausted (4)" for ALOE (hidden), we have a 'definition by association' and therefore a problem - 'Vera' is not an aloe, nor (on its own) is it a type of aloe, so even 'Vera, say', cannot indicate 'aloe'. The plant is called 'aloe vera', and if one were determined to involve Vera in a clue for 'aloe', it would have to be through the use of 'name given to type of aloe' or something along those lines.

Similarly, 'Stockholm' would have to be something like 'Capital  linked to syndrome' and 'Fibonacci' might be indicated by 'Numbers man'.  By the same token, 'French, say' would be valid for 'language' (a definition by example - 'the language of France'), but nor for 'dressing'.

The clue “Bully waste, light brown at the edges (7)" involves UREA (‘waste’) having TAN (‘light brown’) put around it (‘at the edges’), giving TAUREAN, The answer means ‘of or relating to a bull’, but there is no entry in Chambers for ‘bully’ under the headword ‘bull’ (in the bovine sense). The definition is what I usually term ‘fanciful’, and relies on the solver understanding what the word would mean if it did exist. The rather overused ‘flower’ to indicate a river falls into the same category. Strictly speaking, all clues that rely on these ‘made up’ words, whether in definition or wordplay, are unsound, but they add a bit of variety and fun to puzzles and I enjoy coming across them from time to time. There is, of course, a risk that the idea can be taken too far – ‘detailed’ indicating the removal of the last letter of a word strikes me as borderline, and the use, say, of ‘ingest’ to indicate containment in DEED seems too great a stretch.

In general, the answer and the definition in any clue need to be the same part of speech. So a noun such as SAILOR must be defined by a noun or noun phrase, eg 'mariner', while an adjective (eg FIERCE) must be defined by an adjective or adjectival phrase (eg 'savage'). Some setters, including Azed, make an exception specifically for nouns, allowing them to be the (missing) subject of a verb. The example that Azed has given in the past is 'barks and is man's best friend' for DOG, where the solver must infer a 'one' or similar at the beginning.

The clue "In hollow a dip through which river runs (4)” for WADI (hidden) involves just such a definition. The expanded ‘one through which river runs’ strikes me as a rather messy definition of WADI. There is another problem, though. ‘In hollow a dip’ cannot stand alone as wordplay, ie it doesn’t by itself deliver a substring from ‘hollow a dip’ – it needs a subject, as in eg ‘river valley in hollow a distraction’. That missing subject could perhaps follow, eg ‘In hollow a discernible river valley’, although I’d prefer to see a comma between the hiding place and the definition; but here we don’t have an implicit subject, which means that the clue has to be pre-processed to ‘In hollow a dip[, one] through which river runs’.

In algebra, transitivity applies: so if a=b and b=c then a=c. However, transitivity does not apply to the meanings of words and phrases. So Chambers gives one meaning of 'yield' as 'to surrender', and one meaning of 'afford' as 'to yield'. While one can deduce from this that 'yield' (b) can be considered synonymous with both 'afford' (a) and 'surrender', (c) it cannot be assumed that 'afford' and 'surrender' are thus synonymous. In this instance the words linked through 'yield' do not have a shared sense, so 'afford' cannot, for instance, be used as an expulsion indicator in the way that both 'yield' and 'surrender' could.

They are the bits in brackets at the end of each clue, indicating the length of the answer along with additional information such as the number of words involved.

This typically depends on the type of puzzle .

In blocked puzzles (eg UK back-pagers), the words that make up the answer are treated separately, so NOT HAVE A CLUE would be shown as (3,4,1,4).

In barred puzzles (eg the Listener), the total number of letters and the word count are given, so NOT HAVE A CLUE would be (12, 4 words) or (12, four words).

This depends on the type of puzzle.

In blocked puzzles, the length of the individual parts is indicated, so CLUED-UP would be shown as (5-2).

In barred puzzles, hyphens are generally ignored, so CLUED-UP would be shown as (7).

At one time it was customary to indicate apostrophes as part of the enumeration in blocked puzzles, so RUBIK'S CUBE would have been (5'1,4). Now, however, the standard treatment is to ignore apostrophes, so in a blocked puzzle RUBIK'S CUBE would be (6,4), and in a barred puzzle it would be (10, 2 words). In certain barred puzzles you might see it enumerated as (10, 2 words, apostrophe).

All other marks, such as accents, are ignored in enumerations.

Unless the instructions explicitly state otherwise (which may be true of occasional themed puzzles), every cell will contain a single capital letter and nothing else - no punctuation marks, no diacritical marks etc. Where a prize crossword entry is submitted on paper, the grid should be completed in ink.

In essence, any word which could be removed from the clue without affecting the cryptic interpretation is an 'extra word' in that it is technically superfluous. That doesn't mean that it is necessarily a bad thing, although such words tend to devalue a clue. In days gone by, clues of fifteen or more words were commonplace; the modern trend is towards much greater concision (driven both by taste and by space constraints).

Extra words are not allowed when they cannot be interpreted by the solver in a way that contributes to the cryptic sense of the clue, in other words the clue cannot be fully parsed while they remain. In the 'hidden' clue "Man found in French apartment (4)" for CHAP, the word 'found' is superfluous, but the clue is still entirely sound - 'found in' and 'in' cryptically come to exactly the same thing. By contrast, "Man in chic French apartment (4)" is unsound - we have a definition ('man'), an indicator ('in'), and a hiding place ('French apartment'), but what is the solver to make of that 'chic'? The clue cannot be satisfactorily solved while the word is there, and it must therefore be removed.

They come into several categories:

1.  Technically superfluous words in an indicator or a definition

This can often be a virtue rather than a fault. 'Parasitic worm' for CESTODE technically contains an extra word ('parasitic'), but this actually gives the solver an extra piece of information. Likewise 'moving about' as an anagram indicator could be reduced to either 'moving' or 'about', but it's absolutely fine as it is.

2. 'Neutral' indicators in wordplay

In the clue 'Boy with unknown woman (4)' for LADY (LAD + Y), the word 'with' has no cryptic effect; it is a 'neutral juxtaposition indicator', so when it appears in a wordplay between element A (here LAD) and element B (here Y), it indicates that A should be followed by B in the output. Words like 'with', 'and', 'before', 'ahead of', and 'leading' can all perform this role (as indeed can any of the 'before' indicators in the Juxtaposition Indicator list), and will often be used by setters to improve the surface reading of a clue.

3. Links between definition and wordplay

For the benefit of the surface reading, setters will frequently include links from definition to wordplay, or vice versa. An example would be 'King not having the French for tea (4)' for CHAR (CHARLES - LES). The word 'for' here sits between wordplay and definition, indicating that the wordplay leads to the word defined by 'tea'. In 'Bird from Sweden pale (4)' for SWAN (S + WAN), the 'from' indicates that the defined answer derives from the wordplay.

In themed puzzles, clue gimmicks are a way of providing the solver with thematically-linked information which can only be revealed by solving the clues. The clues thus fulfil a dual purpose, leading both to answers to be entered in the grid and to additional text which is assembled in a particular way (often in clue order) to produce some form of message or instruction. So each of the across clues in a puzzle might have a letter missing from the definition; these letters, in clue sequence, could spell out CHANGE EVERY D TO AN H.

All clues in a puzzle may have the same gimmick, some may have the same gimmick while others are normal, or multiple gimmicks may be used. It is incumbent on the setter to clearly state in the preamble what gimmicks are in play, eg 'The definitions in twelve across clues are missing a letter which must be reinstated before solving; the definitions in twelve down clues contain an extra letter which must be removed before solving. Other clues are normal.'

Note that in all gimmicked clues, the surface reading of the clue as printed is expected to make reasonable sense; the surface reading of the corrected version can make no sense whatsoever.

Missing letter clues typically generate one 'thematic' letter per clue. The letter can be omitted from definitions only or from wordplays only; the preamble will mention this. Alternatively they can be omitted from wordplay or definition, and the preamble will simply say that a letter has been omitted from each clue.

The clue "Chap agitated, I bet (4)" is a missing letter clue for BITE, the corrected definition being CHAMP (missing letter M), while "Halt on summit (4)" gives STOP (S + TOP), with 'on' in the wordplay becoming SON (missing letter S).

It is implicit that the clue cannot properly be solved without the missing letter being reinstated, so in "Large bust majestic (5)" for REGAL the setter might have intended that 'bust' should be corrected to BURST (missing letter R), but the clue works perfectly well as it is written, and it is therefore no good. The same applies to a clue like "Bird, first to cause quarrel (4)" for CROW (C(ause) + ROW); there is no way that the solver can be expected to know that 'cause' should become CLAUSE (missing letter L).

This is very similar to the 'missing letter' gimmick, except that a letter must be removed before the clue can be properly solved, and all the observations about one type apply equally to the other. On occasion a sequence of consecutive letters may need to be removed, usually from a small number of clues - the preamble will always make this clear.

An example of an extra letter clue would be "Could give up crack (6)" for CLEAVE (C + LEAVE), where 'could' loses the letter U to become COLD.

The important thing about clues featuring the extra word gimmick is that they must not be solvable while the bonus word is present; this is vital, because there will almost certainly be other words whose presence or absence does not affect the cryptic reading. An example of a sound extra word clue would be "Man found in ditch outside a pub (4)" for CHAP (hidden in ditCH A Pub); the word 'outside' must be removed before this clue can be solved. The word 'found' is technically redundant, but the clue works both with and without it, so it is not an 'extra word' in the context of that gimmick.

In this type of clue, the definition leads to the answer while the wordplay leads to the answer all except one letter. An example would be "Bird near bananas (5)" for RAVEN (anagram of NEAR), where the anagram delivers all but the letter V, which becomes part of the message. Rightly or wrongly, the term 'wordplay' is considered to extend to a second definition, so "Bite fellow (5)" for CHAMP (CHAP), contributing the letter M, would be allowed.

There is a school of thought which says that this type of clue should not include link words between definition and wordplay, particularly those such as 'is' which indicate that the defined word and the output of the wordplay are the same, since the two are different. This rule may or may not be enforced by editors.

The same as the 'wordplay omits a letter' gimmick, except that here the wordplay delivers one more letter than is required to form the defined answer, with this letter then typically contributing to the thematic message (the preamble will explain what the solver must do with these letters). An example of this type of clue would be "Chap thought unhappy (6)" for FELLOW (FELT + LOW), where the extra letter is T.

Note that any double definition clue of the 'wordplay omits a letter' kind can also serve as a 'wordplay delivers extra letter' clue - so the "Bite fellow (5)" clue of the former type for CHAMP could become "Bite fellow (4)" for CHAP, delivering the extra letter M.

This is a rather different gimmick, in that the clue itself is normal. Once solved, a letter of the answer (typically the first or the last) provides, based on its position in the alphabet, an index into the clue itself. So if it is the last letters of the answers that are providing the indices, for an answer which ends in A, the solver must select the first letter of the clue as part of the thematic message; for an answer ending in B, the second letter of the clue etc. So in a 'last letter is index' clue "Caught husband with a quiet fellow (4)" for CHAP (C+H+A+P), the last letter of the answer is P, the sixteenth letter of the alphabet, and the thematic letter is the sixteenth letter in the clue, T.

There are a number of terms which are either specific to crosswords or which have a particular meaning in a cruciverbal context. The following section contains a list of the ones that I am aware of - please let me know if you come across any others which you would like me to explain or which you think should be added to the list.

Anagrind  Short for 'anagram indicator', often used in discussion of clues.

Biff  A verb derived from the acronym BIFD, "Bunged In From Definition". A 'biffed' entry is based on the definition and crossers, without an understanding of the wordplay, so it may or may not be correct.

Big Red Book or BRB  see Chambers.

Cell  One of the 'squares' in the grid.

Chambers  The Chambers Dictionary, the standard reference for most crosswords. The most recent version is the 2016 printing of the 13th edition. There are excellent electronic versions available for PC, iPad/iPhone and Android devices. It is not to be confused with the Chambers 21st Century Dictionary; this is available online but does not contain many of the more abstruse words to be found in the Chambers Dictionary.

Charade  A clue where the elements of the wordplay are to be placed one after another (resembling a game of charades), eg "Italian Romeo, ordinary fellow (5)" for ROMAN.

Checked Letter  A cell in the grid which is at the intersection of an across and a down entry, such that the letter entered is shared between the two entries and thus provides a check for the solver.

Classification  A category marker applied in a dictionary to a word or phrase, such as 'archaic' or 'dialect'.

CR  See Cryptic Reading.

Cryptic Reading  The way that a solver must interpret a clue in order to work out the answer (compare Surface Reading).

Definition  The part of a clue which directly (non-cryptically) indicates the answer.

Double Duty  A word in a clue is performing double duty when it is trying to fulfil two cryptic functions at once. An example would be the word 'of' in the clue "Just part of airport (4)" for FAIR, where it is required by both the hidden indicator 'part of' and the hiding place 'of airport'; such a clue is unsound.

Easter Egg  Something which the setter has concealed in a clue or in the grid but which has no bearing on the solving process. The original Easter Eggs were programmed into Google search (try googling "askew" and you'll find one of them), with the term being used to describe ‘an intentional hidden message, in-joke or feature in a work such as a computer program, web page, video game, movie, book or crossword.’ A Nina (qv), by contrast, is an unclued message which all solvers are likely to understand if they spot it, while a Ghost Theme (qv) typically links several clued answers.

Entry or Light  A group of cells running horizontally or vertically in the grid, separated only by thin lines. In a blocked crossword, the start and end of an entry will marked by a block or the edge of the grid; in a barred crossword, the delimiters are a solid line (a bar) or the edge of the grid.

Enumeration  The number in brackets following a clue - eg '(4)' - which tells you how many letters are in the answer. This will almost invariably match the number of cells available for the answer in the grid.

Fodder, Anagram Fodder, or Anagrist  The letters that need to be rearranged in an anagram, so in the clue "Old man cracking nut (6)" for ALMOND the fodder is OLD MAN.

FOI  'First one in', the first answer that a particular solver put into the grid.

Green Paint  A multi-word expression which has no significance beyond its component words and therefore cannot legitimately be used as an answer to a clue. An example would be GREEN PAINT, in contrast to, say, GREEN BELT or WAR PAINT.

GWIT  Acronym for 'guess what I'm thinking', applied pejoratively by commentators to a step in solving a themed puzzle which requires an unreasonable mental leap.

Ghost Theme  A thread that runs through a  crossword which is not explicitly themed, as in a plain puzzle where several of the entries are the surnames of British Prime Ministers.

Headword  A main entry in a dictionary, eg cross in Chambers.

Homonym  A word that is spelt the same as another, but has a different meaning and etymology, eg 'pine' (a tree/to waste away). The two words appear as (or under) different headwords in the dictionary.

Homophone  A word that is pronounced the same as another, but is different in spelling and meaning (eg 'fair'/'fare'). The two words appear as (or under) different headwords in the dictionary. A homophone used in a crossword is often described as a pun.

Libertarian  A term applied to setters who (deliberately or accidentally) produce puzzles which do not comply with certain of the principles set out by Ximenes (see Ximenean). This trait often manifests itself in clues where the grammar of the cryptic reading could be considered faulty. The back page puzzles of UK daily newspapers are generally constructed along Libertarian lines, while barred puzzles such as the Listener puzzle in The Times insist on Ximenean principles being upheld.

Lift and Separate  The term applied to an element of the clue which must be divided before the clue is solved. An example would be "Clubs copyright dance (5)" for CAPER, where 'copyright' needs to be split into 'copy' and 'right', giving APE + R.

Light  see Entry

LOI  'Last one in', the final answer that a particular solver put into the grid.

Lurker  Another name for a 'hidden' clue, such as "Man found in French apartment (4)" for CHAP.

Nina  The term 'Nina' refers to a message which appears in the filled grid but, like an Easter Egg or a Ghost Theme does not need to be identified in order to solve the puzzle. It takes its name from the name of American caricaturist Al Hirschfeld's daughter, which he frequently concealed within his drawings. They are typically found in blocked puzzles, where they might run round the perimeter or along a diagonal.

Pangram  A puzzle where the completed grid contains at least one instance of each of the letters A-Z. Two instances of each letter constitutes a double pangram, and so on.

PDM  The acronym for 'Penny Drop Moment', the point where an imaginary lightbulb illuminates above the head of the solver. A PDM can occur with an individual clue, but more often relates to the moment when one understands a particular stage of a themed puzzle.

Plain Puzzle  A standard puzzle with no gimmicks; the clues are solved and the grid is filled.

Preamble or Rubric The instructions which appear before the clues in a themed crossword, telling the solvers about such things as gimmicks in clues, modifications to answers before entry in the grid, changes to be made to the initially filled grid, cells that must be highlighted, or words to be written outside the grid.

Register  A Classification which relates to the meaning of a word in a particular context, eg 'law' or 'physics'.

Rekrul  The reverse of a Lurker, eg "Fellow over in Panama (3)" for MAN.

Rubric  see Preamble.

SR  See Surface Reading.

Subhead  A dictionary entry that appears underneath a headword, eg crossword under cross in Chambers.

Subsidiary Indication  See Wordplay.

Surface Reading  What a clue appears to say when read as a piece of English prose (compare Cryptic Reading).

Themed Puzzle  A puzzle where the solver is required to do more than simply solve the clues and enter the answers; such a puzzle will invariably be introduced by a preamble.

Unch  An unchecked cell, ie one that appears in a horizontal entry or a vertical entry but not both, so there is no additional check that the letter entered in that cell is correct.

Wordplay or Subsidiary Indication  The part of a clue which leads cryptically to the answer.

Write-in  A clue that one solves at the first time of asking.

Ximenean  This term supposedly describes puzzles which are constructed in compliance with the principles set out by Derrick Macnutt is his book Ximenes on the Art of the Crossword, as opposed to those described as Libertarian. In practice, it applies to setters (i) whose clues feature at least one definition and one complementary wordplay, contain no superfluous words which serve no purpose but to mislead, and are grammatically sound both in the surface and cryptic readings; and (ii) whose grids are broadly in line with the guidelines set out in the book. Things have changed a bit since Ximenes' time, and certain constructions which he would have accepted are no longer in favour; unfortunately, we are stuck with this rather arcane term until someone comes up with an alternative that is adopted by the crossword community.

The first of the Clue Movies provides an introduction to both blocked and barred grids. The section on grids starts around 2:19 into the video.

There are very few setters these days who use ruler, pen and paper to create a crossword grid. There are several pieces of software which will allow you produce all manner of grids, automatically enforcing whatever level of symmetry you select, and allowing you to export the empty or filled grid in a form that can be incorporated in various types of document, or published online for interactive solving. Most of these programs will also fill the grid for you, either completely or partially.

The Crossword Compiler program helps with all aspects of the setting process, from designing the grid to producing clues; it is user-friendly and works well, but it comes at a significant (recurring) cost. The Qxw program is free, and is an incredibly powerful tool for creating and filling grids of all shapes and sizes; it could not be described as user-friendly, but apart from that it is a great piece of software. Recommendations from readers concerning other similar programs will be gratefully received and reflected in this article.

Unchecked letters, or unches, are those in cells that feature in just one grid entry, so they are not 'checked' by a crossing entry.

In blocked puzzles, there are no hard and fast rules regarding unches. In general, entries that contain more unches than checked letters (eg five-letter entries with only two checked letters) are discouraged. Similarly, pairs of consecutive unches are considered undesirable, but are not forbidden; three consecutive unches in an entry, though, are not allowed.

The rules are much clearer for barred puzzles. No entry may contain consecutive unchecked letters. The maximum number of unchecked letters are as follows:

Entries of 3 letters: no unchecked letters
Entries of 4 or 5 letters: 1 unchecked letter
Entries of 6 or 7 letters: 2 unchecked letters
Entries of 8, 9, 10 or 11 letters: 3 unchecked letters
Entries of 12, 13, 14 or 15 letters : 4 unchecked letters

There is no minimum number of unches, although fully-checked entries of four or more letters in plain barred puzzles tend to be frowned on. The situation is a little different when a themed puzzle contains unclued entries; although technically the cells where these intersect a clued entry are 'checked', in reality if a clued four-letter entry intersects two clued entries and an unclued entry, solvers will only have two 'crossers' to help them solve the clue, so it is entirely acceptable - and may even be desirable - for such a light to intersect three clued entries along with the unclued entry.

There are two situations which can arise with barred puzzles:

(i) The pattern of bars at the centre of the grid grid can form a symbol which has become associated with Nazism. It is described by Chambers as 'ancient and worldwide', and editors will not in my experience reject a puzzle because of its presence. It is, though, something that occasionally draws negative comment on blog sites, and is best avoided where possible.

(ii) An undesirable word can be seen in the grid straddling the dividing bars. Most grid-filling software has the ability to check for this, but again it is not something that would cause a puzzle to be rejected unless the  presence of the 'message' could be construed as being intentional.

In a hidden clue, the answer is concealed within a word or words on view in the clue, either in the forward direction (a 'hidden' or 'lurker') or the reverse direction (a 'reverse hidden' or 'rekrul'). An example of a 'hidden' would be "Animal in bath or sea (5)" for HORSE, and of a reverse hidden "Woman over in Wensleydale (4)" for LADY.

The wordplay for hiddens typically takes one of three basic forms:

1. Where the answer is shown as coming from the hiding place, eg "Animal seen in conifer retreated (6)" for FERRET.

2. Where the hiding place is shown as containing the answer, eg "Magnificent rally holds key (7)" for CENTRAL.

3. Where the definition is separate from the wordplay, which explicitly indicates that a section of text should be extracted, eg "Good amount of golf in Europe (4)" for FINE.

There is an argument that the type 2 form with the definition at the end is not entirely sound (in the example above, the rally is the subject, not the key). The example would be better phrased with the definition as the subject, eg "Key held by magnificent rally".

A reverse hidden takes much the same form, except that a reversal indicator is required; either the extracted text or the entire hiding place can be reversed, eg "Party backed contribution to Conservative disco (4)" for SIDE or "Small sample of grain I muttered about (4)" for MINI.

A number of 'type 3' indicators can be found in the list of hidden indicators. When it comes to type 1/type 2 indicators, many Container and Contents Indicators can be used; the important thing is to select those which describe a steady state rather than a single action, so 'held by' is good, but 'grabbed by' is not. The clue "Man held by cha-cha partner (4)" works nicely for CHAP, while "Man grabbed by cha-cha partner (4)" doesn't - for as long as the CHA-CHA PARTNER has existed, the CHAP has been (held) there.

The hiding place should contain only words which contribute to the answer, so "Individual tucked into kippers one provided (6)" for PERSON is no good, since the word 'provided' plays no part in the clue ("Individual tucked into kippers once" would be fine). There is a specific exception to this rule, which is that the definite and indefinite articles can be included, so "Queen featuring in the Guardian newspaper (4)" for ANNE would be allowed.

The hiding place can contain punctuation marks as well as spaces - anything other than letters are ignored, with all letters being treated as unaccented capitals.

Yes, but the solver needs to be given sufficient information about the construction. So "Drinks linking West and East (4)" for TEAS would be fine, but "Drinks found in West and East" would not.

In general, yes, but a degree of leeway may be given. A reference to another entry is generally allowable, so if the answer to 5 is DREARY, then "Back in 5 (4)" would work for REAR, and if the answer to 8 were BOARDER, one might get away with the &lit clue "One of eight (3)" for OAR. Most editors would also, I think, accept "Vehicles lining Route 66 (4)" for UTES.

By their nature, hidden clues are generally easy to solve, although when well disguised they can still be tricky. Like simple anagrams, they have the virtue of giving solvers a way in to a puzzle, particularly a tough one. I would suggest that a couple of 'hiddens' in a puzzle is probably enough, and more than three would be considered excessive.

The names of certain letters of the alphabet, such as PEE (P) and SEE (C), can be useful to setters, as in the clue "See finished screen (5)" for COVER [C + OVER].

A 'reverse' translation' is also possible. In the clue “First character to perform admits unfortunate dressage manoeuvre (6)" for PESADE, the 'first to perform' yields PEE, which contains SAD ('unfortunate'). Similarly, 'first letter from cousin' could give SEE (it could, of course, also indicate C). The more concise 'first to perform' is certainly valid for the single letter P, but it is a moot point whether it can fairly indicate PEE, or whether ‘first to complain’ can give SEE. I believe that the solver must be pointed in the right direction through the use of ‘letter’ or ‘character’, thus "first character to perform' for PEE and 'first letter from cousin' for SEE.

The clue "I’ll be captivated by s-spiffy little bird on the river (8)" for DIDAPPER features a device which is quite often seen in Azed puzzles but rarely occurs elsewhere, and involves a word in the clue being modified in order to tell the solver that an analogous modification must be made to the word indicated. Like the ‘Cockney’ device, where for instance ‘ammer might indicate ‘it, here the modification suggests a stammer which must be similarly applied to the word derived (the result will not be a real word, but will form an element of the solution), so ‘remain b-blue’ could indicate ‘be l-low’ , and here ‘I’ is contained by ‘d-dapper’ to produce ‘didapper’. It is not a device that I’m particularly fond of.

This device involves the loss of an aspirated 'H' from the start of a word. It can be introduced explicitly, as in "Cockney passion" for (h)EAT, or implicitly, as in "'arass" for (h)ARRY. So the clue "Stroke quiet Cockney's headgear (3)" gives PAT [P + 'AT], and "Clubs 'ave bitter" yields COLD [C + 'OLD].

On occasion, certain words used in a clue can establish a context within which others are to be interpreted. So in the clue "Wake when honor students start putting calculators away (9)" for AFTERMATH, the American spelling 'honor' sets a context wherein the subject being studied would be MATH rather than MATHS.

They describe the role that a particular word plays in a sentence. Words can be divided into eight classes: adjectives (eg 'adorable'), adverbs (eg 'quickly'), conjunctions (eg 'and'), interjections (eg 'crikey'), nouns (eg 'saucepan'), pronouns (eg 'they'), prepositions (eg 'of'), and verbs (eg 'reimburse').

The examples above are words which can only function as a single part of speech, so 'adorable', for instance, can only qualify a noun or a pronoun. Many words, particularly common ones, can function as different parts of speech depending on the context in which they are used. Chambers groups the senses of words under the particular part of speech, ie 'adj' (adjective), 'adv' (adverb), 'conj' (conjunction), 'interj' (interjection), 'n' (noun), 'prep' (preposition), 'pronoun', 'vt' (transitive verb, one that has an object, such as 'disown'), and 'vi' (intransitive verb, one that has no object, such as 'ramble').

It is not uncommon to find words which can act as four different parts of speech (eg 'man'), and that is before any other words with the same spelling have been taken into account.

The most important weapon in the setter's armoury is deception. Producing a clue with a surface reading which appears to say one thing and a cryptic reading which means something very different, but legitimately leads to the answer, is a highly effective deception; unravelling such a clue can also give the solver considerable satisfaction.

Take the clue "Marine cross about chaps joining navy (6)". This appears to relate to a member of the Royal Marines being annoyed about the navy's recent intake, but actually it is a clue for SEAMAN, another word for a merman (ie 'marine cross'). In the wordplay, A ('about') divides ('chaps') SEAM ('joining') and N ('navy').

In the surface reading, 'Marine' appears to be a noun; in the cryptic reading it is an adjective. 'Cross' is not an adjective in the cryptic reading, but a noun. 'Chaps' would appear to be a noun, but it is a verb. And 'joining' looks like the participle of a verb, but actually it is a noun. 'About' and 'navy' are the same parts of speech (preposition, noun) in both readings, but they lead to single letter abbreviations in the cryptic version.

Using different parts of speech craftily (but fairly) is an essential element of the setter's art.

A proper noun is (almost, though not quite, always) one which starts with a capital letter, such as the names of people (Cher), places (Brisbane), organizations (Amazon), or branded products (Elastoplast), and the titles of works such as books (Emma), films (Titanic), or songs (Price Tag). Nouns which are not proper nouns are termed 'common nouns'; these are spelt without an initial capital.

The rules around including common nouns in puzzles (whether as answers or elements thereof) are the same as for verbs, adjectives etc - if they are in the dictionary which is the standard reference for that puzzle, they can be used freely.

Proper nouns can be used without constraint in the wordplays of puzzles when they are there simply to provide one or more letters, via eg anagrams, hiddens, or letter selection indicators. An example would be "Long leg back for Lawrence (4)" for PINE (PIN + (lawrenc)E), where it matters not whether the solver knows who Lawrence is.

When the proper noun is either the answer of part of the answer, things are a bit different. There is no problem with proper nouns that can be found as entries in the primary reference for the puzzle (usually Chambers, sometimes Collins or the Oxford Dictionary of English); examples from Chambers would be 'Cossack',  'Hoover', and 'Scottish'. Proper nouns which appear in the text or etymology of related items are also valid, for instance 'Circe' based on the definition of 'Circean'. Similarly those which appear as part of an entry, eg 'Nile' under 'Nile green', "a very pale green colour, thought of as the colour of the River Nile."

Beyond that, there is no need to inform the solver that non-dictionary proper nouns are being used, but the setter has to decide whether any such proper noun is likely to be either known to the solver or readily verifiable using, say, a basic atlas or the web. However, when we talk about the web, this means the sort of proper noun which at the very least would have its own (substantial) Wikipedia entry; if you choose any surname you like and google it together with 'poet', you will find a poet to match. What is vital is that the solver should be in no doubt that they have got the right name - ARNE and PURCELL are valid composers, while WOODHOUSE is not, even if there are several composers of that surname found by a web search.

If you are in any doubt, and the grid entry cannot be changed to avoid the need for the name, help the solver by qualifying the definition - make your playwright an 'Elizabethan playwright', your composer a 'writer of marches', or your town a 'Sussex town'. Also try to make the wordplay as clear as possible. If the questionable proper noun is being used as part of an answer, eg the 'composer' appears in the wordplay, just rewrite the wordplay.

By convention, a full stop is never used at the end of a clue. Otherwise the punctuation in a clue, whilst it may be tuned to support the surface reading, must not itself mislead the solver when it comes to the cryptic reading. The clue "Male guards, one leading (4)" for MAIN (MAN around I) is unfair because the cryptic reading requires 'male guards one', but the comma rules out this interpretation.

When it comes to groups of words in wordplays which are there simply to provide a series of letters, punctuation is essentially ignored. So the comma in the 'hidden' clue "Man tucked into peach, apple (4)" for CHAP is fine, as is "Man - old - crunched nut (6)" for ALMOND (anagram of MAN OLD); the punctuation in the latter clue is geared more to the surface reading than the cryptic one, but it doesn't unfairly mislead. Note that where a comma separates words which are part of a single wordplay element, such as the fodder for an anagram, that element cannot govern a single verb. hence "Man, old, cracks nut (6)" is not valid for ALMOND, but "Old man cracks nut" is, while "Men, old, crack stone table (6)" is fine for DOLMEN, but "Old men crack stone table" is not.

Question marks have certain special roles in cryptic clues beyond their normal usage in written English:

1. They can substitute for an expression such as 'perhaps' or 'for instance' to indicate a definition by example. The clue "Son left squash, perhaps (5)" for SPORT could also be written as "Son left squash?".

2. Similarly, they can indicate a definition which is a little out of the ordinary, as in "Criminal rang aunt's mother? (4)" for GRAN (anagram of RANG). This device is often used in &lit clues, such as "Are unfinished tiles transformed here? (7)" for ATELIER (anagram of ARE TILE(s)); this clue is fairer to the solver than the alternative "Unfinished tiles are transformed here".

3. They can tell the solver that the setter has done something slightly unusual, though not outrageous, as in "Dashes meaning something special after three dots? (11)" for SMATTERINGS ((MATTERING S) after S), where three dots is the letter S in Morse code.

4. They can be used to help the setter by splitting a clue into two parts, often with the definition preceding the question mark and the wordplay following, as in "Watery vapour? Nothing's left damp (4)" for MIST (M(o)IST).

Other than when required by the surface reading, normally for interjections (eg 'Gosh!"), exclamation marks should be reserved for one purpose, which is to indicate that the setter has done something very unusual. An example would be "What I am (in extreme form) gets me reverse of irate inside! (5)" for VEGAN (VEGETARIAN is the answer with IRATE< inside).

The temptation for setters is to use the shriek to announce to the world that they have done something very clever. If you have done something truly impressive, then the solver will appreciate it regardless of the exclamation mark; if they need to be told that it's terribly good, then it really isn't.

Ellipses are used to join a pair of consecutive clues in two situations:

1. To produce a single surface reading, although the two cryptic readings are entirely separate, as in this pair of clues:

People in Hispanic land... (4)
...are stopping to go round country (5)

for CLAN (hidden) and SPAIN (A in SPIN).

2. Where the clues are cryptically linked, and one of the clues cannot satisfactorily be solved without reference to the other, for example:

Low energy accepted by fellow... (4)
...occupying couch, a podge (4)

for MEAN (E in MAN) and CHAP (hidden), the definition in the second clue being the 'fellow' at the end of the first.

Apostrophes have no special meaning in clues, but they are very much the setter's friend, since they can indicate both possession and the contraction of 'is' or 'has' - and they can suggest one in the surface reading while actually meaning another in the cryptic reading. In the clue "Made aware of Andrew's blundering (6)" for WARNED, the apostrophe-s looks as though it indicates possession, but is actually the shortened form of 'is'.

Occasionally solvers suggest that, say, "journey's end" should indicate not Y but S. This is clearly wrong: "journey's end" is equivalent to 'the end of journey'; 'his rear' would not be a valid indication of S.

A single dash has no particular significance in a clue, but a continuous series of em dashes can be used to produce a blank, representing the answer. For example, in

"I.e.                    could be one busy with apron and beers (9)" for BARPERSON, the letters of IE + BARPERSON (the blank) are a possible rearrangement of I APRON BEERS.

This device is used almost exclusively in &lit clues, usually (as above) of the composite anagram type. It is typically employed when the answer can't legitimately be represented by alternatives like 'this' or 'such'.

Consider the clue “Backs possibly injured dropped from later game (5)". The  word ‘lame’ (‘possibly injured’, the ‘possibly’ being there because one can be lame for reasons other than injury) is ‘dropped’ from outside ‘later game’ to yield TERGA (‘Backs’). The clue would read more smoothly with a couple of commas, ie “Backs, possibly injured, dropped from later game”, but one of the principles of good clueing relates to respecting orthography – punctuation must not be used in the wordplay to mislead the solver unfairly. So the ‘all-in-one-breath’ expression LAME-dropped-from-LATER-GAME’ is fine, but in ‘LAME, dropped from LATER GAME’ the word LAME is clearly the subject and therefore the combination cannot reasonably lead to a modified form of LATER GAME. In general, punctuation should not be introduced to enhance the surface reading if it makes the cryptic reading less clear.

Most barred puzzles explicitly identify a specific edition of a particular dictionary as their primary reference. Blocked puzzles rarely have such a statement, but the Chambers Dictionary (2016) is the usual 'default' dictionary, and every setter should thus have a copy of it, either in printed or electronic form. The Oxford Dictionary of English (ODE) is another good, single-volume dictionary favoured by certain puzzles, while the free, online Collins Dictionary (https://www.collinsdictionary.com/dictionary/english) is a handy resource which tends to better reflect modern usage.

Thesauruses are the double-edged swords of the setter's weaponry. They can be very useful for identifying potential synonyms, but they can also serve up words which are far from synonymous (even extending to intransitive verbs as potential synonyms for transitive verbs, and vice versa). Roget's Thesaurus is a fine work, but not designed for crossword setters, while the Chambers Thesaurus is loose and outdated - only really useful for finding obscure (obsolete, Scots etc)  synonyms. The thesaurus which is part of the online Collins Dictionary (look up a word in the dictionary and click the 'Synonyms' tab) is pretty good, although some of the synonyms offered rely on modern senses of words not given by Chambers. My favourite thesaurus is part of the Oxford Dictionary Thesaurus, the pages of which are split into the dictionary (above) and the thesaurus (below) - it gives a few 'false positives', but a high proportion of the synonyms that it suggests will normally stand up to further scrutiny.

For a long time, the Oxford Dictionary of Quotations was a 'must' for both setters and solvers of themed puzzles. As a source of inspiration for setters it still has some value, but solvers cannot be expected to have the latest version (currently the 2014 8th edition) and the 'turnover' of quotations is far greater than that of words in a dictionary. Personally, I find the more recent editions deeply disappointing and find the web a better source of information when setting and when solving.

Brewer's Dictionary of Phrase and Fable is a splendid source of arcane information, and can be of assistance to a setter seeking inspiration. Solvers of the Spectator crossword are likely to have a copy, but is otherwise not a standard solving reference.

Different publications have different 'house styles', and a puzzle isn't going to be rejected because you've put 'organize' when they like 'organize', but if you have any doubt about whether the 80s should be the 80's or the Times should be The Times, then New Hart's Rules will give you an answer.

Not that long ago I thought perhaps I ought to get a copy of Bradford's Crossword Solver's Dictionary, but although I can see how it might be of value to new solvers, it very rarely leaves my bookshelf.

The Chambers Biographical Dictionary is a cracking piece of work, but for the crossword setter it has no obvious advantages over Wikipedia, while the Chambers Slang Dictionary is fun to dip into but of little or no value to setters. The Times Desktop Atlas of the World fits nicely on a bookshelf and a good way of finding anywhere that's anywhere.

The web has rendered all printed references for solvers, except for dictionaries (which can also be obtained in electronic or printed form), largely unnecessary. The nature of the web means that it's not a matter of working through one repository after another for the information that you need, but rather of adopting a 'search and click' strategy. The web is similarly a boon for setters, but you have to make sure that you double-check any information before incorporating it in a puzzle.

A superb resource for themed puzzle setters is Dave Hennings' database at xwdb.info - if you are thinking about setting a puzzle on a particular theme, a few clicks will tell you whether it's been done before, when, where, by whom, and how the subject was handled. For competitors in the Azed clue writing competitions (or anyone who wants to enjoy some classic clues from 50 years of Azed comps), the &lit archive is indispensable. Sadly it is no longer being updated, but if you want to know whether Azed will accept, say, 'stalling' as a containment indicator, you'll find out here.

There is no rule that says that similar answers cannot appear in the same grid, and occasionally puzzles deliberately repeat exactly the same word or phrase. This is done with intent, however, and in general setters should avoid including words that closely resemble each other (eg BOUGH and BOUGHT), particularly those which appear under the same headword in the dictionary (eg FAIR and FAIRNESS), or alternative spellings of the same word (eg PROA and PRAHU). Automated grid filling programs typically have the option to avoid answers which share a substring of a specified minimum length - if this were set to 4, then the answers PREVIOUS and REVISION would not be allowed to appear together. Personally, I feel that strings which appear either at the start of two answers or the end of two answers are far less desirable than, say, PREVIOUS and REVISION or POUND and ROUNDERS.

Most editors require that no indicator should be used more than once with the same cryptic meaning in a single puzzle. So 'holding' appearing in two separate clues as a containment indicator would be unacceptable, as would 'son' appearing twice as a pointer to the abbreviation S. This stricture would extend to inflections of the same verb (so 'keeps' and 'keeping' as containment indicators would constitute an unwanted repetition), and to multiple alternative ways of indicating the same element, eg 'love' and 'nil' for O.

The repetition of the same indicator but with a different meaning is not an issue - theoretically, 'about' could be used to indicate a reversal, a containment, an anagram, the letter C, and the letters RE in the same puzzle. Words which have no cryptic meaning can be used multiple times, eg 'stake' to define 'post' and 'risk' in separate clues. Common words such as 'and' and 'in' are generally exempt from restriction even when used cryptically unless they lead more than once to the same abbreviation (eg 'with' for W).

Note that these rules have nothing to do with fairness to the solver and everything to do with the overall impression made by the crossword. Hence the setter should where possible avoid other obvious repetitions (even though they may be allowed by the editor), particularly in consecutive clues - including a pair of clues which start with the same three words, even if the cryptic meanings are completely different, looks careless.

Special clues fall into two categories:

1. Those which are intended to present a different challenge to the solver from the conventional 'definition plus wordplay' clues. Forms such as Printer's Devilry, Spoonerisms, and Definition + Letter Mixture (DLM) fall into this category. Puzzles may have clues of just one type, or a mixture of several types (sometimes described as a 'gallimaufry').
2. Those which are based on the conventional form, but are designed to provide not just an answer to be put in the grid but additional information, typically one or more letters, which assist the solver of a themed puzzle in some way. Examples are clues containing misprints, clues where the wordplay delivers an extra letter not used in the answer, and clues which contain an extra word; the additional letters or words thus generated often produce an instruction to the solver. Clues such as this are often described as being 'gimmicked'. The preamble of the puzzle will state what gimmicks are in use, and will typically tell the solver which clues are affected. Individual gimmicks are described in their own section.

Printer's Devilry (PD) clues are highly unusual in that they contain no definition of the answer. Rather, a piece of conventional English prose has a sequence of letters which form a word omitted, the resulting gap closed, and the punctuation altered as necessary (anywhere in the clue) in order to produce a 'devilled' clue which makes at least a modicum of sense. There is no 'subsidiary indication' in a PD clue, so it is vital that the undevilled (full) version must read smoothly, such that the solver can be confident that they have correctly reconstructed it.

An example of a simple PD clue would be "Outraged reader, fitter, off to the press! (6)" for RESALE, the undevilled version being "Outraged reader fires a letter off to the press", with the break coming at 'fi/tter'. Only minor changes to the punctuation have been made here, but in this clue for PESTO, "Vaulting above tunnel entrance appears inane, scary", not only does the answer need to be inserted into 'sc/ary', but the word 'inane' also needs to be actively involved in the re-punctuation, giving "Vaulting over tunnel entrance appears in an escape story." The modern trend is away from clues involving pyrotechnic revisions, as exemplified by another Azed cup-winning clue, this time for MINARET, "Bunter-whine starts with jaw open: “Cease – condone – Wharton, please!”, which becomes "Bunter whines, "Tarts with jam in are twopence. A second one, Wharton, please."

Note that it is considered poor form to have a word break coinciding with the start or end of the inserted word, so WAS being 'undevilled' as WASH OVER [answer HOVER] is less than ideal, although WAR is fine for WASH OVER [answer SHOVE].

Spoonerisms can appear in normal puzzles, where they must be flagged, eg "Spooner's raced horseman" for RED SPIDER ("sped rider"). In a spoonerisms puzzle, though, the preamble will describe the clues, and each spoonerism will not be individually flagged. The clues generally fall into two categories:

(i) Spoonerized definition. Here the clue is normal, except that the definition part of the clue has been spoonerized. In "What bucks, after sighting, will be bagged by rifle aimed? (4)", the definition must be un-spoonerized to produce "What sucks after biting", a definition of FLEA, which is concealed within the wordplay. A change in punctuation within the definition, as seen here, is allowable.

(ii) Spoonerized entry. The wordplay yields the grid entry, while the definition part of the clue leads to a word or words (if more than one, then they do not need to have any connection), which must be spoonerized prior to entry in the grid. An example would be "Dapper coterie, a point accepted by street (7)" for SATINET, where "Dapper coterie" leads to 'natty set', while the wordplay gives ((A TINE) in ST). Note that in this form of clue there is no definition of the grid entry itself.

Spoonerisms can take two forms as far as these clues are concerned: consonantal, such as "wed late" for LEAD WEIGHT, or vocalic, such as "core part" for CARPORT. Those of the latter type generally present more issues for both setters and solvers.

DLM clues have a definition, but the wordplay is replaced by a series of words in which is hidden a jumble of the letters that make up the answer. So "Gift of pens tremendously well received (7)" would be a DLM clue for PRESENT ('Gift'), a jumble of the answer appearing in PENS TREmendously. There are two things to note here: firstly, that the non-definition part of the clue can contain words which are there simply to improve the surface reading (making DLM clues very easy to write); and secondly, that the letter mixture generally begins or ends at the beginning or end of a word in the clue, so "Gift of printer especially useful to student (7)" for PRESENT would not usually be considered acceptable, although the preamble will typically state how the DLMs work in a particular puzzle.

It is very rare these days to see DLM clues other than in themed crosswords, where they lend themselves to being combined with other non-standard clueing methods.

In a Letters Latent puzzle, the definition part of each clue leads to an answer from which all occurrences of a specific letter must be removed prior to entry in the grid, while the wordplay leads to the grid entry, which is typically a non-word. The letters removed, taken in clue order, will spell out a message of some kind. The enumeration in the clue usually refers to the length of the grid entry. An example of a Letters Latent clue would be "Gets boat wrecked on island (6)', where the definition leads to OBTAINS while the wordplay yields (BOAT* + IS). The grid entry is therefore OBTAIS, while the letter which contributes to the message is an N. In the clue "Dismissal, number being involved in corruption (4)", the definition is of RUN-OUT, while the wordplay gives (N in ROT), RONT; both of the Us in the defined word have been removed, but only a single U forms part of the message.

There are many different types of cryptic clue, but the vast majority use a few basic constructions, either alone or in combination.

A clue typically includes exactly one definition and one subsidiary indication (now usually called the wordplay). Sometimes clues will have more than one definition as well as a wordplay, and very occasionally they will include more than one wordplay in addition to the definition. When the clue can be read as both definition and wordplay, the clue is called an &lit ('and literally so'); when the clue contains only a definition which must be interpreted in a particular way to discover the answer, it is called a 'cryptic definition' clue.

The basic clue types are covered in the following sections.

Anagrams involve the rearrangement of text which is normally on plain view in the clue. For many more details about anagrams, please see the section here which is dedicated to them.

Anagrams can be combined with most other constructions.

A charade involves two or more elements being assembled in sequence. The elements may be complete words or groups of letters such as abbreviations; they can simply follow one after another in the clue, or they can be interspersed with juxtaposition indicators (eg 'with', 'and'). The following are examples of basic charades:

"Close main line (4)" SEAL (SEA L)
"Conflict about new underground network (6)" - WARREN (WAR RE N)
"Cheer trick with fish (7)" CONSOLE (CON SOLE)

The order of the elements in a charade can be altered using a juxtaposition indicator, eg

"Left after second game (5)" SPORT (PORT after S)

The individual elements in a charade may themselves be other constructions, as in

"Consume contents of six-pack brother brought round (6)" ABSORB (ABS + BRO<)

In a containment clue, one element is wrapped around another. Either a containment indicator (eg 'holding') or an insertion indicator (eg 'filling') is required. The point in the container where the insertion takes place must be determined by the solver. Examples of simple containment and insertion would be:

"Times contains exclusively drivel (7)" - BALONEY (BY around ALONE)
"In uniform, entering area of Tower Hamlets (7)" - POPULAR (U inserted into POPLAR)

In addition to the form 'X <indicator> Y', the construction can also be phrased as 'Y X <indicator>', where the indicator is an active verb, eg

"Complaint old chap bears (4)" - MOAN (MAN around O)
"Despicable fellows one has bored (4)" - MEAN (A in MEN)

Two or more elements can be contained, and other constructions can be incorporated, as in

"Dream about English Prime Minister upset with European restraint (10)" - TEMPERANCE (TRANCE around (E + PM< + E))

A rare variation involves only partial containment, for instance

"Bar, last piece in grip of obese child (6)" - INFANT (INN with the last letter contained by FAT)

The deletion construction involves part of one wordplay element being eliminated. This can be a specified sequence of letters, using a deletion indicator,

"Colour shot of car belonging to setter (4)" - MINE (CARMINE - CAR),

or a letter or letters specified by their position,

"Leaderless group fighting (6)" - ACTION (FACTION missing the first letter).

The focus of the construction is usually the string of which part is to be deleted, but it can be the other way round,

"Tolerate son abandoning post (4)" - TAKE (S leaving STAKE).

Where there is potential ambiguity, it is not essential that the setter should resolve it, but it is certainly desirable.

"Young Belgian adventurer's ultimately discharged in cast (4)" - TINT (TINTIN - IN). The 'ultimately' tells the solver that the IN to be removed from TINTIN is the one at the end.

A double definition clue guides the solver to the answer by giving two separate definitions of it. Ideally these should be of two different headwords in the dictionary (so the definitions are not 'cognate'), as in

"Put up with pain in the neck (4)" for BORE [past tense of 'bear'; noun]

but as long as the two meanings are different enough to pin down the answer, that is fine. An example that would not pass muster would be

"Live with bear (6)" for SUFFER [verb; verb], which could equally well be a clue for ENDURE or STAND, but

"Screen cricket match (4)" for TEST [verb; noun = 'test match'] is fine.

The main potential for misdirection in a double definition clue relates to the parts of speech involved, so setters should take advantage of ambiguity in this area wherever possible. So "Grasping plan (4)" is a better clue for MEAN than "Unkind to suggest".

Link words can be used between the two definitions, although these should be ones which suggest equality rather than one 'producing' the other, eg

Contemptuous indifference is trifling (6)" for SLIGHT [noun; adjective].

Generally the definitions are taken directly from the dictionary, or are synonyms of the sort that would be found in a thesaurus. Informal meanings are acceptable, but the use of 'oblique' definitions is normally avoided, as in something like

"I'm trying to sort out a puncture (4)" for BORE [noun; verb]

This type of clue can be extended to include three or more definitions, as in

Complete accurate model (7)" for PERFECT [verb; adjective; adjective]

and sometimes setters provide a cryptic wordplay in addition to the two definitions, as in

"Due to participate in second race (5)" for SHARE [noun; verb; S + HARE]

In this situation, the clue is effectively a standard definition + wordplay clue with a bonus definition, so the definitions can take any form that would be acceptable in such a clue.

In days gone by, the alternative route to the answer was termed a 'subsidiary indication', which would clearly include a second definition. This has now been superseded by 'wordplay', which suggests a cryptic construction, but when reference is made in the preamble of a themed puzzle to 'wordplay', this conventionally extends to either of the definitions in a double definition clue.

The drag and drop construction involves part of a word or phrase either being moved to a different position within it or swapped with another 'chunk'.

The chunks to be moved can be indicated either explicitly or by position:

"Mark getting son to finish bangers? (4)" for CARS [SCAR with S moved to end]
"First to finish in race practises (6)" for TRAINS [STRAIN with first letter moved to end]

A selection of drag and drop indicators can be found in the Clinical data section, along with further information about the construction.

In hidden clues the answer is concealed in plain sight within the wordplay, eg

"Sample from Tesco perfume range (5)" for SCOPE [hidden]
"Lamented element of 'hidden' I perspired over (7)" for REPINED [hidden reversed]

These clues are discussed in detail in their own section.

Perhaps the most controversial type of clue, the answer to a homophone or pun sounds like the indicated word or words when they are spoken.

The essential element of a homophone clue is a 'sounds like' indicator; there are many beyond the classic 'we hear', a selection being available in the relevant part of the Clinical Data section.

The most basic homophone is the 1:1 type, eg "Reasonable price for trip announced (4)" for FAIR [sounds like "fare"]. The requirement here is that the words should (according to the dictionary) be pronounced the same, or where the dictionary gives alternative pronunciations there should be at least one in common (eg 'whale' and 'wail'). It is essential that the homophone indicator forms part of the wordplay (so it isn't the 'wrong side' of the definition - "Announced reasonable price for trip" is a clue for FARE, not FAIR), but the only type of homophone where ambiguity is possible is a 1:1 pun where the two words are the same length, so

"Price for trip reportedly reasonable (4)" is ambiguous, but "Episode reportedly viewed" is not - if the answer is four letters, then it is SEEN; if five letters, it must be SCENE.

The more entertaining puns involve multiple words, often in a 2:1 ratio, eg

"Setter's amphibian sounded tiny (6)" for MINUTE ["my newt"]. A degree of latitude in pronunciation is usually allowed in multi-word puns, so

"Wickedness unusual, by the sound of it, in ecclesiastical council (5)" for SYNOD ["sin odd"] where the dictionary pronunciations of 'odd' and the second syllable of 'synod' differ slightly.

It is perfectly acceptable in an n:1 homophone for a word indicated by the wordplay to transfer directly into the answer, since every word sounds like itself! An example would be

"Daft in conversation to exclude setter" for BARMY ["bar me"].

Some setters take the n:1 homophone a stage further into the partial homophone. Here the pun produces not the answer but a piece of the answer. In some instances, this piece is itself a word, as in

"Picked up main article in rage (6)" for SEETHE ["sea" + THE]. Clearly the element THE cannot be part of the pun, because the pronunciation is completely wrong. Sometimes the pun leads to a non-word, for instance

"Record on the radio by US singer is something remarkable (8)" for SCORCHER ["score" + CHER]. My personal view is that this is a bridge much too far.

Homophone clues often draw criticism from solvers because of regional variations in pronunciation. "Past eel" might sound like PASTILLE in some parts of the UK, but in other parts it certainly wouldn't.

A spoonerisms is a variation on the homophone clue - see the next section for details.

Spoonerism clues are a type of homophone clue, which rely on clergyman and Oxford don William Spooner's supposed tendency to exchange the syllables of two words in a phrase. Sometimes themed puzzles involve a number of spoonerism clues of different types (see the section on 'Special Clues'), but when one appears in a conventional puzzle, the definition leads to the answer while the wordplay leads to a combination of words (usually two, likely to be unconnected with each other) which must be spoonerized to produce the answer. The fact that a spoonerism is required must be flagged to the solver; this is typically done by including something like "Spooner's" or "for Spooner", although sometimes setters come up with rather more obscure, and less sound, indications.

In the clue "Head off Spooner's attempted  plundering (9)", the 'attempted plundering' gives 'tried sack', a spoonerism of SIDETRACK. Often the answer will be a two-word phrase, as in 'drain mag' for MAIN DRAG. Spoonerisms can be consonantal (the usual kind), such as 'sow press' for PROCESS, or vocalic, as in 'lay coat' for LOCATE. Setters are generally allowed a little more licence with spoonerisms than with standard homophones, where identical pronunciations (as given by Chambers) are required.

Letter selections typically form part of a wordplay, often making up a shortfall of one or two letters, rather than being the only device employed; the exceptions are likely to be intermittent selections (see below), or the selection of a series of first or last letters, for example "Fellow made angry noises initially (3)" for MAN [first letters]. As in this clue, the 'fodder' from which the letter(s) are selected must be in plain view; there are no exceptions to this rule.

The relevant Clinical Data section contains a range of letter selection indicators. In most instances, the indicators apply to a single word in the clue, so "start of meeting" is M and "start of our meeting" is O (not OM). The exceptions to this rule are adverbs (eg 'initially' or 'ultimately') and adjectival phrases (eg 'on vacation') which can apply (as in the preceding example) to any number of consecutive words in the clue. With noun indicators, a plural form must be used if single letters from two or more words are to be selected, as in "Fellow starts to make angry noises (3)" for MAN.

Where first and last letters are being selected, an indicator which is already in the plural applies only to a single word, so "borders of large region" would be LE, not LERN; this behaviour can be modified by the inclusion of 'and', so "borders of France and Monaco" would be FEMO (not FE or any other combination). Where the indicator is in the singular, the plural form applies to multiple words, so "shells of large molluscs" would be LEMS.

With noun indicators (eg head, rear) it is necessary to include a preposition or a possessive apostrophe-s to complete the construction, as with 'starts of' in the example above, or "close to home" for E. Setters must decide whether a preposition which fits naturally in the surface reading is valid in the cryptic reading, for instance does "back to work" adequately indicate "the rear part of [the word] work"? It might appear that "squadron leader", say, would be valid for S, but although in the real world it describes the leader of [a] squadron, in the cryptic reading 'squadron' is just a series of letters, and the word 'leader' doesn't work without a preposition or a possessive. Occasionally solvers suggest that "journey's end", say, should lead to S. This is incorrect - "journey's end" is "the end of journey", and gives the letter Y.

Indicators that are adjectives probably ought to apply only to single words, but if "primarily for effect" can give FE, then "all options exhausted" can give ALOS.

Intermittent selectors are listed separately in the Clinical Data section; these may provide a complete answer, as in

"Bird regularly behind deer (5)" for EIDER [alternate letters], or "Bird regularly behind yew with rook (5)" for EIDER [alternate letters + R].

In a replacement clue, one part of a word is replaced with something else. The outgoing chunk can be specified explicitly or by position; the incoming chunk is always specified explicitly. Examples would be:

"Tease fellow swapping America for Italy (4)" for CHIP [CHAP with I replacing A] and "Sample spread, having time for starter (5)" for TASTE [PASTE with T replacing the first letter]

The relevant Clinical Data section contains a range of replacement indicators.

Reversal involves the complete inversion of a wordplay element, which can be a single word, as in "Returned school register (4)" for NOTE [ETON<], or multiple words, as in "Trouble a German about royal privileges (7)" for REGALIA [(AIL A GER)<].

A list of reversal indicators can be found on this site. Many, such as 'lifted' and 'rampant' can only be used in down clues; some editors will not accept certain reversal indicators, such as 'backing', in down clues. I find this strange, given that the clues are written horizontally, so it would seem logical that the manipulation could be carried out in the same plane prior to entry in the grid. I would accept that something like 'from the east' seems counterintuitive for reversal in a down clue, which is why a few similar indicators are shown as 'Across only'.

Reversals are often combined in clues with other cryptic operations, as in "Observed returning hotel laundry (4)" for WASH [SAW< + H].

Where a clue consists simply of the reversal of one word to form another, there is considerable scope for ambiguity - see the Ambiguity section for information on how to deal with this.

Occasionally, a setter may want to include an anagram where the rearrangement is in fact a reversal, as in "Stop running around in stations (6)" for POINTS (STOP* around IN). While a clue like this cannot be considered unsound, I would advise using a reversal indicator whenever a reversal is involved, ideally using a synonym of the word being reversed rather than the word itself, eg "Arrest over drinking in stations (6)" [STOP< around IN].

Many conventional clues feature definitions that will not be found in the dictionary, such as 'stiff examination' for AUTOPSY, but the cryptic definition clue consists solely of a definition, the idea being that there is sufficient information provided for the solver to be able to confidently identify the answer without any supporting indication.

The clue "Naughty type of Limerick (8)" for SPALPEEN is an example, but this depends on the solver knowing that 'spalpeen' is an Irish word for a rascal, since there is no other way to get to the answer. It is for this reason that cryptic definition clues have been outlawed in barred puzzles since the mid-1970s. They are still occasionally seen in blocked puzzles, often as clues for phrases, eg "Sporting event one's bound to compete in (5-6,4)" for THREE-LEGGED RACE. The combination of a familiar expression and the unusual enumeration means that a clue like this could hardly be seen as unfair. "This cylinder is jammed" for SWISS ROLL with its (5,4) enumeration probably clears the fairness bar, but "Even when frayed, it should be kept (6)" for TEMPER strikes me as more of a riddle than a clue.

There are few, if any, types of clue that divide opinion more than the &lit, or all-in-one, clue - some consider it the pinnacle of the setter's art, others see it as showy flimflam. The whole of an &lit clue serves both as the wordplay and the definition - although the term 'definition' in the context of a cryptic clue often describes something which is not a dictionary definition (eg "man's best friend" for DOG), the &lit clue can take this to a whole new level, where 'indication' becomes more accurate than 'definition'. This is one of the areas that detractors will pick up on, and on occasions 'indication' veers towards 'suggestion',

A well-constructed &lit can be a fine thing, as with Colin Dexter's "Item gran arranged family slides in? (5,7)" for MAGIC LANTERN [CLAN in (ITEM GRAN)*].

Sometimes the whole clue will represent the definition but only part will constitute the wordplay. This is a 'semi &lit' or 'offshoot &lit', the latter being the term coined by Ximenes to describe a clue such as "What a bishop may have had before getting a crook (7)" for PREBEND [PRE BEND], where the first six words do not contribute to the wordplay.

There are some clues which could be viewed as having the standard wordplay+definition structure, but where the wordplay part enhances the definition. An example would be "What AZ has to go through and measure efforts (7)" for ENTRIES [EN + TRIES], where 'What AZ has to go through' would be a reasonable definition, but the remaining words serve to enhance it, such that this clue could also be considered a 'semi &lit'.

In a self-referencing clue, part of the clue refers to something that appears earlier in the same clue. An example would be “One unseating e.g. Herod rent him in pieces (9)”. While the solution here is clearly DETHRONER, derived from the letters of HEROD RENT, it may not be immediately obvious where the ‘e.g’ and ‘him’ fit in. Clues may legitimately be self-referential, for instance by including a pronoun which refers to a noun appearing earlier in the clue. This requires the solver to essentially pre-process the clue, replacing the reference with the item referenced – so here ‘him’ must replaced by ‘Herod’, making the normalised clue “One unseating e.g. Herod rent Herod in pieces”. where the definition is ‘One unseating e.g. Herod’ and the wordplay is ‘rent Herod in pieces’. Note that because the pre-processing takes place before the clue is solved in the conventional definition/wordplay sense, the internal reference must ‘work’ in the context of the surface reading. The clue “Compass point, one to avoid (4)” for SNUB [S + NUB] is unfair – it must expand in line with the surface reading as “Compass point, compass point to avoid” (not “Compass point, point to avoid”), and the indication of NUB is no good.  ‘”Compass point, one most up-to-date (4)”, pre-processed to “Compass point, compass point most up-to-date”, would probably be valid for NEWEST.

In the clue "Foreign knife, its head buried in dog now dead (4)”, the ‘its’ is intended to refer back to ‘knife’, and it is the first letter (‘head’) of this word which must be removed from KUKRI to produce KURI, the ‘dog now dead’. The clue raises two points: firstly, can ‘its’ legitimately refer to ‘knife’ alone? In my view, no – the result of pre-processing the clue is “Foreign knife, foreign knife’s head buried in dog now dead” and the wordplay fails; “Knife, its head buried in dog now dead” would work very nicely. Secondly, could ‘its head’ alternatively have indicated I, the first letter of ‘its’? No, that would have demanded the use of “it’s head” (which makes no sense) – ‘its head’ does not mean ‘the head of the word it’ (or of the word ‘its’). A possessive pronoun can never (fairly) be used in wordplay with a letter selection indicator alone in order to deliver a letter from the pronoun itself – ‘their leader’ is not T, nor ‘his conclusion’ S.

In answer-referencing clues, the answer itself is used cryptically within the clue, sometimes to produce an &lit. An example would be "What’s this, involving a bit of rapacity?" for AVARICE - the answer to the question "What's this?" when applied to the answer is A VICE, and when those letters are put around ('involving') A and the first letter ('bit') of 'rapacity', the word AVARICE emerges. Sometimes the placeholder for the answer is italicized, so here the clue could be "What’s this, involving a bit of rapacity?"

Often the requirement is to break down the answer and insert it into the wordplay. For example, in "Treat changes to seat with this cement" for TRASS, the answer needs to be broken up and used to replace 'this' in the wordplay, which becomes "TREAT changes to SEAT with TR AS S". Here the 'before' and 'after' words are explicitly stated, but often one or the other (though not both, which would make the clue unreasonably tough) will be indicated by a synonym.