An idiom is a combination of words that has a figurative meaning owing to its common usage. An idiom's figurative meaning is separate from the literal meaning. There are thousands of idioms and they occur frequently in all languages. There are estimated to be at least twenty-five thousand idiomatic expressions in the English language.
Many idiomatic expressions, in their original use were not figurative but had literal meaning.
For instance: spill the beans meaning to let out a secret probably originates in a physical spilling of beans which are either being eaten or measured out. The point is that the spiller certainly does not want to lose any beans.
let the cat out of the bag : has a meaning similar to the former, but the secret revealed in this case will likely cause some problems. A cat was sometimes put in bags to keep it under control or to pretend that it was a more saleable animal, such as a pig or a rabbit. So, to let the cat out of the bag suggests either that the ruse is revealed or that the situation is out of control.
break a leg: meaning good luck in a performance/presentation etc. This common idiom comes from superstition. It was thought that there were gremlins or sprites, little fairy-like creatures, backstage in theaters who would do exactly the opposite of whatever they were told. To say break a leg was to ensure the sprites would not in fact do the performers any damage.
In linguistics, idioms are usually presumed to be figures of speech contradicting the principle of compositionality. This principle states that the meaning of a whole should be constructed from the meanings of the parts that make up the whole. In other words, one should be in a position to understand the whole if one understands the meanings of each of the parts that make up the whole. The following example is widely employed to illustrate the point:
Fred kicked the bucket.
Understood compositionally, Fred has literally kicked an actual, physical bucket. The much more likely idiomatic reading, however, is non-compositional: Fred is understood to have died. Arriving at the idiomatic reading from the literal reading is unlikely for most speakers. What this means is that the idiomatic reading is, rather, stored as a single lexical item that is now largely independent of the literal reading.
In phraseology, idioms are defined as a sub-type of phraseme, the meaning of which is not the regular sum of the meanings of its component parts. John Saeed defines an idiom as collocated words that became affixed to each other until metamorphosing into a fossilised term. This collocation of words redefines each component word in the word-group and becomes an idiomatic expression. Idioms usually do not translate well; in some cases, when an idiom is translated directly word-for-word into another language, either its meaning is changed or it is meaningless.
When two or three words are often used together in a particular sequence, the words are said to be irreversible binomials, or Siamese twins. Usage will prevent the words from being displaced or rearranged. For example, a person may be left "high and dry" but never "dry and high". This idiom in turn means that the person is left in their former condition rather than being assisted so that their condition improves. Not all Siamese twins are idioms, however. "Reading, writing, and arithmetic" is a frozen trinomial, but it is usually taken literally.
Dealing with non-compositionality
The non-compositionality of meaning of idioms challenges theories of syntax. The fixed words of many idioms do not qualify as constituents in any sense, e.g.
a. How do we get to the bottom of this situation?
The fixed words of this idiom (in bold) do not form a constituent in any theory's analysis of syntactic structure because the object of the preposition (here this situation) is not part of the idiom (but rather it is an argument of the idiom). One can know that it is not part of the idiom because it is variable, e.g. How do we get to the bottom of this situation / the claim / the phenomenon / her statement / etc. What this means is that theories of syntax that take the constituent to be the fundamental unit of syntactic analysis are challenged. The manner in which units of meaning are assigned to units of syntax remains unclear. This problem has motivated a tremendous amount of discussion and debate in linguistics circles and it is a primary motivator behind the Construction Grammar framework.
A relatively recent development in the syntactic analysis of idioms departs from a constituent-based account of syntactic structure, preferring instead the catena-based account. Any word or any combination of words that are linked together by dependencies qualifies as a catena. The words constituting idioms are stored as catenae in the lexicon, and as such, they are concrete units of syntax. The dependency grammar trees of a few sentences containing non-constituent idioms illustrate the point:
The fixed words of the idiom (in orange) in each case are linked together by dependencies; they form a catena. The material that is outside of the idiom (in normal black script) is not part of the idiom. The following two trees illustrate proverbs:
The fixed words of the proverbs (in orange) again form a catena each time. The adjective nitty-gritty and the adverb always are not part of the respective proverb and their appearance does not interrupt the fixed words of the proverb. A caveat concerning the catena-based analysis of idioms concerns their status in the lexicon. Idioms are lexical items, which means they are stored as catenae in the lexicon. In the actual syntax, however, some idioms can be broken up by various functional constructions.
The catena-based analysis of idioms provides a basis for an understanding of meaning compositionality. The Principle of Compositionality can in fact be maintained. Units of meaning are being assigned to catenae, whereby many of these catenae are not constituents.