In 1856, Cambridge professor, Walter Skeat, coined “ghost-word.” Skeat defined a ghost-word as one that appears in dictionaries but doesn’t really exist.
These words get into dictionaries via typographical error, mistranslation, or, because a manuscript was misread.
The usual example cited whenever ghost-words are discussed is “dord.” Dord appeared in Merriam-Webster’s 1934 dictionary this way: “dord. n. density.” But dord isn’t a word and certainly isn’t a noun meaning “density.”
It entered the dictionary through an unlikely string of accidents. The intended entry was: “D. or d. Density.” It was slated for the section devoted to abbreviations. Somehow, it ended up on a list intended for the main body where it became dord.
This error was quickly spotted but, unbelievably, editors left it alone and dord remained through several subsequent printings. Scholars speculate that editors wanted to see how many users, if any, would notice and comment. There is no record regarding such queries.
Dord is a true example of a ghost-word, It was born because of a misunderstanding, was never used, and likely never will be. It doesn’t exist as a word and doesn’t haunt modern dictionaries.
Nevertheless, several ghost-words have become an accepted part of the lexicon. One example is “ye” as in names such as, “Ye Olde Tea Shoppe.”
Ye was never an ancient spelling of “the.” It arose through a misreading of the runic “thorn” (þ), a letter which signified “th.” The thorn still exists, albeit in but one living language — Icelandic.
Ye isn’t the only ghost-word to gain acceptance. Gravy is another. Gravy entered Middle English in 1591, a misreading of the Old Norse grane (grain). Yet earlier is scapegoat (1530), the product of an error in translation from Hebrew. Syllabus (1656) was a misreading of the Latin sittyba (little slip, label).
Everyone knows what tweed is. How many know it’s a trade name that should never have existed. Tweed (1847) is a misreading of tweel, the Scottish form of “twill,” a type of woolen cloth. Tweed’s acceptance was helped by the existence of the already named River Tweed.
An early label for language ghosts was “corrupt.” A text considered corrupt is untrue to its original. Such usage is ancient. It precedes printing and refers to errors made by scribes. We’ve no way to know of other mistakes incurred by scribes that ended up as respectable words.
Gutenberg’s printing press was introduced in 1450, almost 100 years prior to scapegoat’s entry to English.
A 16th-century term pertinent to this discussion is mumpsimus — “an established reading of a text that is patently wrong.” Mumpsimus grew from the story of a semi-literate medieval priest who garbled Latin and chanted the meaningless mumpsimus during mass when he should have used quod in ore sumpsimus (That we have taken up in the mouth).
He stubbornly ignored the correct wording even after his error was pointed out. Thus mumpsimus gained a second meaning: “Stubborn persistence in using a mispronunciation, misspelling, etc.”
Mumpsimus possibly explains some ghost-words. Carelessness is a more likely reason.