Since the traditional major dictionaries are very carefully compiled and edited, ghost words are rare in the top tier of reference works. But dictionary writers have been known to copy entries from one another without looking for citations, which means that some ghost words can be propagated. (See this Language Hat post on how the ghost word pornial, a mistake for primal appeared first in the Century Dictionary and then was copied into Funk & Wagnalls.)
Apart from traditional dictionaries, there are a whole host of less scrupulously compiled word lists designed to feed people's appetite for odd or unusual words. A mistake, or even a misleading definition, can definitely propagate, particularly with the Internet, where odd words get passed around.
Janet McConnaughey over at the Books and Writers Community sent me a message asking about an obscure word that has apparently been floating around the Internet: agenhina. Although the word isn't completely invented from whole cloth, the form in which it is being transmitted is problematic enough that it seems plausible to call it a ghost word.
Here's the citation:
agenhina (n.) — a guest at an inn who, after having stayed for three nights, was considered one of the family (Saxon law)
Googling for the term only turns up various sites quoting this definition, mostly without attribution, although at least one gives the immediate source: Mrs. Byrne's Dictionary of Unusual, Obscure, and Preposteroous Words by Josefa Heifetz Byrne. (I've checked this source, and the definition is there word-for-word, so that appears to be the immediate source. I have not, however, looked to see where she derived her information from.)
I'd never heard the term before the query, but on the surface it looks suspicious. Two things catch my eye immediately. First, this is supposed to be "Saxon" law. Does the definition writer mean Anglo-Saxon law? That seems more likely for a putatively English word, but that makes the reference to "inn" somewhat anachronistic. Inns, in the sense of "a public house for the lodging and entertainment of travellers" (OED), don't appear until the 15th century, to judge by the entries in the OED. Before that time, inn simply means a dwelling place. Also, the whole situation seems bizarre. A guest at an inn becomes part of the family? The apparent exoticism of the situation probably accounts for why netizens have been sharing this word with each other. (Gee, they sure did things strangely back then!)
Etymologically, "agen hine" makes sense as either Old English or Early Middle English for "own family", or "own servant" (the words for "family" in this period include servants as well as blood relations) but where does the rest of the definition come from?
In such situations, the first place to turn is the OED. Searching the CD-ROM edition for agenhine takes us to the word hoghenhine (agenhine is listed as a variant)
The OED gives a much simpler definition than Byrne does, one that is much closer to the etymology:
Barbarous forms, handed down in the Law books, of early ME o3en hine lit. own domestic (hind), member of one's own family (see hind n.2 2)."
[I'm using '3' for yogh for the benefit of those who don't have full unicode fonts]
Clearly, this definition isn't the immediate source for Byrne. If we look at the original law books, though, we can get some idea of what happened:
The OED cites passages from the Laws of Edward the Confessor (after 1115) and Bracton (ca. 1250):
12.. Laws of Edw. Conf. c. 23 (Schmid) Habeat eum ad rectum tanquam de propria familia, quod Angli dicunt 'tuua nicte geste þe þridde nicte a3en hine' [Holkham Ms. tuo niht gest þe þridde o3en hine; Hoveden, Tvain nithes gest thrid nith hawan man, Lambard, Twa ni3ht 3est, þrid ni3ht a3en hine.]
Translation: "Let him hold him to justice as if of the same family, as the English say, 'two nights guest, the third night, own family.'"
The variant manuscripts give different versions of the Middle English, but they all mean the same thing. In all of them, the expression a3en hine seems to be an open phrase, rather than a single word. Note in particular the Hoveden version, which has hawan man (own man), so substitution of an equivalent phrase was possible.
The Bracton passage seems to show that the expression has become lexicalized (i.e., a word in its own right) and is diverging from a transparent etymology:
c1250 Bracton III. II. x, Prima nocte dici poterit uncuth, secunda vero gust, tertia nocte hog~henehyne.
Translation: "On the first night he shall be said to be unknown, on the second, a guest, on the third night, hoghenhyne."
To understand what this cryptic instruction is talking about, it is helpful to look at a complete passage from one of these law books. The following comes from Book 1, chapter XIII, part 1 of Britton. I found the translation here.
In the next place, let inquiry be made of felons outlawed, and of such as have abjured the realm for felony, who have returned without our leave; and of those who knowingly receive them. And because it is needful that every one should know the danger of receiving such persons, our will is, that all who are of the age of fourteen years or upwards shall take an oath to us, that they will be faithful and loyal to us, and will neither be felons nor assenting to felons; and that every one be in some tithing and pledged by their tithingmen, except persons in religion, clerks, knights and their eldest sons, and women; and let the obligation of the pledge be this, that if they do not produce those for whom they are pledged, to be amenable to justice in our Court when required, the tithingmen with the tithings shall be in our mercy. With regard to clerks, knights, persons of religion, and women, our pleasure is that the head of every household be answerable for all his chief domestics, and that they answer for those under them. As to guests, we will that every one answer for his guest that he shall have harboured for more than two nights together, so that the first night he shall be deemed a stranger and uncouth,1 the second night a guest, and the third night a hoghenhine.2
1 Anglo-Saxon, uncuo [sic for "uncuð"], unknown.
2 Anglo-Saxon, agen hina, his own hind or domestic.
Now we can understand what the term applies to. This passage is talking about who is responsible for reporting outlaws and bringing felons to justice. Most adult males in England of this period would have belonged to a tithing, a group of men in the same community who were collectively responsible for ensuring that if any of their members broke the law, the offender would show up to face charges ["ad rectum"]. For those who weren't part of a regular tithing, the head of household wherever they lived was responsible for them in the same way. Since visitors are obviously out of their ordinary area, their host is responsible for all long-term guests. All it really says is that if a person stays with you for more than two nights, you are responsible in case he or she turns out to be a fugitive, just as you would be if you abetted a felon who was a regular member of your household.
In short, this is really placing an obligation on the heads of households rather than conferring special privileges on long-term guests.
As for the form agenhine, only one citation in the OED uses it:
1619 DALTON Country Just., The 3rd night is called an Hoghenhine or Agenhine..and if he offend the King's Peace his Oast must be answerable for him.
This seems to be an etymologizing back-formation, similar to the footnote in the translation of Britton, since agenhine never appears as a compound in the older texts.
So what is left of our putative word: as a single word, it's attested once, as a variant of hoghenhine and appears to be an etymological "tidying up" of a form that is actually attested.
When the definition writer claimed the term came from Saxon law, she was thinking of Anglo-Saxon law, probably because of the citation from the Laws of Edward the Confessor. The manuscript of those laws actually dates from after the Conquest, so it's not always possible to tell with certainty whether something mentioned there actually dates back to Anglo-Saxon law. Bracton and Britton are both 13th century. Bosworth-Toller, the Anglo-Saxon dictionary, has neither hoghenhine nor agenhina as a compound, although you can find the latter as an open expression in non-legal contexts. It does cite a slightly different version of the Laws of Edward the Confessor mentioned above, in the entry in the Supplement for gist, but the form there is "agen hewe." So the practice arguably may go back to pre-Conquest times, but it's certainly in operation in the 13th century, so it would be more accurate to say that this was a medieval English legal principal, from around the reign of Edward I.
When the definition writer states that the agenhina "became one of the family," she is being literally true, but incomplete, and therefore misleading. In Present-Day English, saying a person is one of the family carries with it a host of implications about inclusion into the social fabric of the nuclear family that are simply wrong in this context. The "own family" principle was narrow in scope, and directed at who was responsible for ensuring the legal accountability people who fell outside the normal rules.
And where does the notion of an inn come from? This seems wholly the creation of the definition writer, as nothing in the original sources justifies it.
Taken together, we have a word and definition mis-shaped from half-understood source material, and propagated by uncritical copiers. For me, that qualifies as a ghost word.