Strictly speaking, a coined word is a word which has been invented (i.e. did not previously exist in any language) and defined by its inventor. For example, the words immoment and fuddle duddle did not exist prior to their being coined by William Shakespeare and former Canadian Prime Minister Pierre Elliot Trudeau respectively.

Words are also sometimes said to have been "coined" (re-coined?) when an existing word is given a new definition which is quite unrelated to its previous meaning or meanings. For example, the pre-existing words unit, elaboration and mode were given new definitions by the authors of the Algol 68 language specification. Former U.S. President Bill Clinton managed to get into a fair bit of trouble for "re-coining" a few words during his term in office.

Words are coined for a variety of reasons including:

  • the word has been coined to provide a shortened or alternative way of referring to another concept. Examples of this type of coined word include immoment (coined by William Shakespeare in Antony and Cleopatra to express unmomentous or of no moment or importance), fuddle duddle (coined by Pierre Elliott Trudeau as a polite reference to a very impolite term) and robotics (coined by Isaac Asimov to refer to the design and building of robots).

  • the word has been deliberately coined to unambiguously describe something quite specific. Examples of this type of coined word include essentially all latin-form species names (e.g. acanthaster planci, coryphaeschna viriditas or even homo sapiens) and many technical terms like the Algol 68 terms unit (roughly equivalent to a statement or expression), elaboration (roughly equivalent to "execution" as in "a unit is elaborated") and mode (roughly equivalent to "data type")1.

  • the word has been coined to make it more difficult for "others" to comprehend what is being discussed if it should be overheard or intercepted. Many classic examples of this type of coined word can be found in Cockney Rhyming Slang which many suggest was invented to allow street vendors to talk without the customer knowing what was being said and/or to allow criminals to talk without needing to worry about being overheard by the police.
Most if not all of the words in the second category allow experts within the appropriate field to have unambiguous conversations without having to constantly stop and clarify the terms which they're using. For example, a myocardial infarction is the death of some of the heart's cells due to a lack of oxygen and other nutrients whereas a cardiac arrest is the stoppage of the heart. The distinction between these two concepts could easily be quite important in a discussion between medical experts and yet the distinction is almost certainly irrelevant in a discussion between random members of the general public.

Finally, it should be pointed out that the phrase "coined term" is often used to refer to a term which has been coined (i.e. a series of one or more words like the species name "acanthaster planci"). One could and almost certainly should argue that the terms "coined word" and "coined term" are, for most practical purposes, synonyms of each other and differ on to the limited extent that the words (terms?) "word" and "term" differ in meaning.

1 Let there be no doubt that the terms "unit", "elaboration" and "mode" are quite carefully defined in the Algol 68 language specification (the colloquial name of the Algol 68 language specification document within the Algol 68 community is "the Algol 68 report" or just "the Report" and the full name is Report on the Algorithmic Language ALGOL 68). i.e. there is nothing "rough" about the meaning of these terms within the Algol 68 context. In fact, the reason that terms like "unit", "elaboration" and "mode" were defined and used within the Report is exactly because the roughly equivalent terms "statement or expression", "execution" and "data type" didn't already have unambiguous definitions. The alternative of specifying unambiguous definitions for the more common terms "statement or expression", "execution" and "data type" in the Report and then using the terms within the Report was rejected as a reader of the Report might not have realized that their understanding of the common term wasn't correct in the context that they were reading it in.