Although crosswords are now to be found in many of the world's languages, for the most part these are "quick" crosswords, whose clues generally consist of a simple synonym of the target answer (Connor, 2013b). The first quick crossword is generally agreed to have been published in New York in 1913; the British cryptic was an offshoot, first found in the 1930s (Arnot, 1981). The cryptic is largely a British phenomenon; Connor (2013a) suggests that this is partly attributable to a British taste for "imagery of the nudge-and-wink variety", toilet humour, and mild profanity, but also to a lack of other solvers from whom to learn the conventions of the cryptic clue. Learning how to solve cryptic crossword clues is a matter of practice, and of learning the rules by which words are encoded into clues (Stratmann, 1978). In order to understand the specific markup needs of the CCCP, it will therefore be necessary to take a brief tour of how target answers may be encoded by cryptic crossword setters.

A quick clue for the location of this conference might simply read "Capital city (6)" (the number in parentheses indicates how many letters are in the target answer). There are, of course, a number of capital cities with six letters; the quick-crossword solver relies on the intersecting words to provide some of the letters needed to narrow down the possibilities until the correct target is found. The cryptic-crossword solver, in contrast, should be left in no doubt of the correct answer, even if there are no intersecting letters available to confirm it. This is achieved by careful writing of clues that encode each target answer in two ways.

The commonest structure for a cryptic clue consists of the definition and the subsidiary indication. The definition provides some kind of synonym or narrative definition of the target; the subsidiary indication encodes the target in some other way. A frequent type of subsidiary indication is the charade, in which letters or groups of letters from the target are clued in sequence in the subsidiary indication. If we take our "capital city" example, a charade clue might read as follows: "Left Ontario with DeLillo for UK city" (6). The clue breaks down as follows: L, a common abbreviation for "left"; ON, the postal abbreviation for "Ontario"; DON, the first name of the American novelist DeLillo. Put together, these fragments provide the target LONDON. "UK city", of course, is the definition. Notice the words "with" and "for". These are not mere filler. Each one has meaning in constructing the clue: "with" indicates that the fragments can be placed together, and "for" is a common indicator that what follows is the definition of the target. I will call these types of words "metalanguage".

The function of the metalanguage is more prominent in this clue: "City bird around nine doves, initially (6)". Here, the clue's first word is the definition, and all that follows is the subsidiary indication. We are looking for a word meaning "bird" which can be placed around the initial letters of "nine doves". The bird is LOON, and if we place ND between its letters we again arrive at the target LONDON. Again, every single word in the clue serves a purpose. Finally, let us consider the following clue: "Melon donut in possession of Julie (6)". Here we need to know that Julie London was a famous singer, in order to understand that "Julie" provides the definition component of the clue. The subsidiary definition is what is known as a "container" type: the letters of the target are inside ("in possession of") the words meLON DONut.

These latter two clues show the ambiguity inherent in most cryptic clues, and how linguistic features are exploited to create that ambiguity. The apparent noun phrase "city bird" leads the reader away from the necessary understanding that the split between definition and subsidiary indication happens between the two words. We can see, also, that the cryptic crossword setter is permitted to play fast and loose with punctuation: a more honest punctuation of the clue as "City: bird around nine doves initially" would make the roles of the various words clearer. Similarly, the metalanguage in the final clue is most easily read as suggesting that the donut is in Julie's possession, whereas the solution requires understanding that "in possession of" works in reverse here. Either reading is perfectly syntactically allowable in English, but the more obvious reading is the wrong one for finding the target answer.

These examples, I hope, portray a distinctive characteristic of a well-written cryptic clue: the solver should be left in no doubt that her answer is correct (Pham, 2016). The subsidiary indication makes it clear that no other answer is possible, by providing specific (albeit heavily encoded) instructions about the physical makeup of the target answer (Coffey, 1998). Not all clues consist of a definition and subsidiary indication, however. So-called "double-definition" clues are also popular. These offer two definitions, incorporated into an apparently single syntactic unit, which approach the target answer from two different directions. An example from The Times will exemplify this type of clue: "Alert goalkeeper may dive thus (2, 3, 4)". The answer is "ON THE BALL", for which "alert" is a synonym, and "goalkeeper may dive thus" is a descriptive definition. Once the solver has hit upon this answer, it is quite clear that nothing else will fit the bill; there is no other phrase that is defined by both of the definitions. Another example, this time from the Daily Mail, shows the dangers of attempting this type of clue: "Guess there's no proof for it (10)". The answer, "CONJECTURE", is not particularly obvious. The two definitions, the synonym "Guess" and the descriptive "there's no proof for it" could apply to many similar words. It will be interesting to see whether intuitions about what makes for a good and a bad clue can be supported by linguistic evidence. For example, it seems possible that a good double-definition clue needs to have a certain amount of semantic distance between the two definitions. If these are too close, there's a risk (as in the Daily Mail example) that the clue ends up being too broad. It is the intersection of apparently quite different synonyms or definitions which both apply to a single target that makes for a good clue.