The first person to use "diglossia" to refer to the phenomenon was Emmanuel Roidis in 1885, as recorded in his Parerga, a collection of his recent newspaper articles (Roidis, Emmanuel. 1885. Πάρεργα. Ed. Stamatopoulos, Dimitrios I. Athens: Ανδρέας Κορομηλάς. p. xvii). I'm grateful to Tasos Kaplanis who corrected the Common Wisdom of the West in a comment last post, and who scanned in the page in question (from a newer edition). As he notes, Psichari used the term in French the following year in his voluminous Essais de grammaire historique néo-grecque, citing Roidis. And that popularised the concept in the West.
The popularisation took some time though. The next early mention Ferguson cites, in his 1959 paper defining diglossia, was William Marçais in 1930, and the usage he cites from Psichari was in 1928. That's why the Common Wisdom of the West assumed Psichari coined the term: it's the earliest reference in Ferguson 1959. Psichari was certainly involved in generalising the term beyond Greek, as this paper trail shows. And Pernot, who readers of this blog know I'm a fan of, was involved as well.
Long quote, but there is a long-standing misapprehension to correct, and a Westerner has corrected the Common Wisdom of the West before.
It would seen that the oldest attestation of diglossia, or rather "diglossie", with the linguistic connotations it currently carries, is to be found in "Essais de grammaire historique néo-grecque" published in 1885 by the French hellenist J. Psichari. Psichari used this term to describe the Greek linguistic situation characterized by the coexistence of katharévousa and démotiki (Prudent, 1981). It also seems that Psichari was the first to extend the use of the term diglossie to refer to the coexistence of two varieties of Arabic, the "spoken" and the "written" ones, and this in an article written in 1928 (id.)
(Naccach, Albert F.H. On Some Implications of the Tadmuraean Aramaic/Arabic Diglossia. Proceedings of the International Conference "Palmyra and the Silk Road", Palmyra, April 1992. p.2.)
Merci infiniment, French government, for putting Prudent's paper online. This tells most of the story you need to know, though not the initial bit with Roidis:
If we are to believe Jardel & Valdman (1979), the term diglossia should be attributed to the Greek Hellenist Jean Psichari, who had popularised it in an article in Mercure de France in 1928. Tracking that reference (laconically mentioned by Ferguson in 1959), we go back to 1885, the date when the same Psichari published his Essais de grammaire historique néo-grecque, in which he says that he has taken the word diglossia from a certain Mr Roidis, who had published an article in the newspaper Acropolis a few months before. In dealing with the question of the evolution of Modern Greek, and the difficulty of reconstructing certain stages of the language, Psichari mentions twice (in 633 pages!) the "strange diglossia Greece is suffering from" [étrange diglossie dont souffre la Grèce].
It was left to Psichari's student Hubert Pernot to give the first consistent definition of the term. In his Grammarire grecque moderne of 1897, he dedicates his entire introduction to explaining the sociolinguistic situation prevalent in the country of Homer. Reprising the political and linguistic history of Greece, Pernot explains how this "strange diglossia" came about: on the one hand a "scholastic, scholarly or purist" language (Katharevousa), a primarily written language, but also spoken in official ceremonies and "some very infequent pedants". On the other hand the current or vulgar Greek (Demotic or Romaic), which is not taught at all, but is the only language in true use.
Twenty years later, Pernot refined his argument in the preface of his Grammaire du grec (langue officielle), which he authored with Camille Polack (1918):"Diglossia", or the duality of languages, is the main hurdle not only for foreigners learning Modern Greek, but also for Greeks, from their primary education.
[Goes on about children having to learn "a double lexical and grammatical system, doubtless related, but clearly distinct, and few of whose elements are interchangeable." Pernot acknowledged debt to Psichari]
Psichari's article in the 1928 Mercure de France went to a different audience than Pernot's preface and introduction. [...] Finally, because he extended diglossia to other Mediterranean domains, fixing its Asiatic origin and speculating on the future of such linguistic duality.[...] Diglossia does not consist solely of using a double vocabulary [...] Diglossia relates to the entire grammatical system. There are two ways of declining, two ways of conjugating, two ways of pronouncing; in a word, there are two languages, the spoken and the written, like one would say of Vulgar and Literary Arabic.
[Psichari's racism, "it's all the Asiatics' fault"]
Beyond the antagonism of systems and groups which he describes humorously [NN: Yup, that's Psichari alright], our innovator is concerned to relativise the declarations of traditional philology ("Descartes' French is gibberish compared to Old French. You could say the same about 1928 French compared to 1830 French", p. 78), in order to explain the notion of transitional language states and constant change. He also touches on the thorny problem of "mixed languages", which he integrates like Dauzat into the ordinary dynamics of language change. Finally we note a solid reflection on the literary problem of diglossia, with reference to Italian, Provençal, French and other societies.
[Then William Marçais uses the word "diglossia" in 1930 about Arabic, with no mention of Roidis, Psichari or Pernot, and with no italics or scare quotes]
(Prudent, L.-F., 1981. Diglossie et interlecte. Languages, vol. 61, pp. 13-38. Here pp. 15-16)
So, Roidis, then Psichari in passing, then Pernot more formally, then Psichari explicitly to a general audience and including Arabic, then Marçais applying it to Arabic, then Ferguson.
There are some catches in giving priority to Roidis though. First, Roidis was not the first to coin the word διγλωσσία "two-tongued-ness" at all; that honour belongs (as far as I can tell) to the Didache, a 1st Century Christian text (also cited in the Epistle of Barnabas). Not that the Didache is talking about diglossia or bilingualism. It's speaking about talking in forked tongues; the BDAG definition of the word is "doubleness of speech that conceals true intentions by deceitful words, duplicity, insincerity":
Οὐκ ἔσῃ διγνώμων οὐδὲ δίγλωσσος· παγὶς γὰρ θανάτου ἡ διγλωσσία.
You shall not be double-minded nor double-tongued [diglossos], for to be double-tongued [diglossia] is a snare of death. (Didache 2:4)
(If the diglossia wars were still raging in Greece, someone could have taken this up as a motto...)
Trapp's Byzantine dictonary also records it as used by the Abbot Isaias (no, I don't know either), and in the Opuscula of Eustathius of Thessalonica (as διγλωττία), in the same meaning.
So as often happened in 19th Century Greek, a term that was coined some time in Hellenistic Greek was re-coined later on, which a somewhat different meaning. Like καπνιστήριο "smoke place", which meant a steam bath a couple of millennia before it meant a smoking room. (And of course that doesn't prove as much continuity of the Greek language as some would claim, which is why the bold etymologist does not say "καπνιστήριο, From Hellenistic καπνιστήριον" but "καπνιστήριο, Calque of French salon de fumer; cf. Hellenistic καπνιστήριον.") The Didache was first published in 1883, so Roidis may have come across it and reused the term.
But he needn't have. In Modern Greek, διγλωσσία "two-language-ness" is the obvious way of saying "bilingualism", so the word would have been easily reinvented to mean that. That's "bilingualism", as in the state of an individual using two languages, rather than a society stratifying them. Diglossias require bilingualism to work, but they're not the same thing. (As you can imagine, this ambiguity has led to some confusion in Greek linguistics, and occasional differentiation between "personal διγλωσσία" and "societal διγλωσσία".)
Now, Roidis is describing diglossia alright, as you'll see; but his first use of the term διγλωσσία still arguably means "bilingualism", as what individuals do rather than what a society does. It's a diglossia-conditioned bilingualism, to be sure, but it's not the way we now use "diglossia". Roidis' second use of the term on the same page *does* refer to what a society does; but he's arguing that Demotic was not yielding to Puristic, and remained a separate language. The way that has happened is diglossia, and what he has just described is diglossia; but the term in that context means just "bilinguality, the existence of two discrete languages." In fact, that's how Pernot defined diglossia as well.
I suspect that in leaving the word untranslated as diglossie, Psichari was helping along its reinterpretation from an individual to a societal phenomenon: it was not mere bilingualisme. In fact, that nuance was presumably why he left it untranslated. The concept is there in Roidis, but the reanalysis to an explicitly social phenomenon is probably Psichari's ("the strange diglossia of Greece", not "of Greeks"), and more so his student Pernot's.
Judge for yourselves though. Roidis has as mordant a tongue as Psichari did—though he was still writing in Puristic, as everyone had to in Athens in 1885, so the tenor is different. (There's a simple reason Psichari was bold enough to write prose in Demotic, as has long been noted: he did not live in Greece.)
Some people regurgitate ad nauseam the notion that the spoken language is "progressing along with" the written language—that is, it is being archaised; and they bring up as evidence parliamentarians, public prosecutors, lawyers and preachers who speak in monologue. These people would be much more convincing, if they were kind enough to inform us what language those orators speak when they make merry with friends, when they endure the pains of surgery, when they scold a child breaking a glass, when they farewell an expiring relative, when they are trod on the street by a careless pedestrian, when they kneel before a woman or when they babble in a dream. This is not at all a matter of one language for the common people and another language for scholars. This is [diglossia] bilingualism in the same people, who have a living language through which they express all their sentiments and passions—but who are condemned to use another language in writing or speechifying, a language in which it is quite impossible to express any sentiment and any passion.
Whatever viewpoint one examines the issue under, one always ends up with [diglossia] bilinguality. Those unwilling to admit that the written language is different from the spoken must nonetheless accept the split into a language of monologue and a language of dialogue, a language of emotion and a language of passionlessness [ataraxia], each having its own lexicon and a rather different grammar. According to Shakespeare it is evidence of strong emotion if one forgets to put on his tie and breeches; but to judge the psychological state of the modern-day Greek, a more certain indication is whether they use or omit reduplication.
Οἱ κατακόρως ἀναμασσῶντες ὄτι συμπροοδεύει, ἤτοι ἐξαρχαΐζεται, μετὰ τῆς γραπτῆς, καὶ ἡ λαλουμένη, καὶ ὡς παράδειγμα φέροντες βουλευτάς, εἰσαγγελεῖς, δικηγόρους καὶ ἱεροκύρηκας μονολογοῦντας, ἤθελον εἶναι πολὺ πειστικώτεροι, ἂν ηὐδόκουν νὰ πληροφορήσωσιν ἡμᾶς τίνα οἱ ρήτορες οὗτοι λαλοῦσι γλῶσσαν συνευθυμοῦντες μετὰ φίλων, ὑπομένοντες τοὺς πόνους χειρουργικῆς ἐγχειρήσεως, ἐπιπλήττοντες παῖδα θραύσαντα ποτήριον, ἀποχαιρετῶντες ἐκπνέοντα συγγενῆ, πατοῦμενοι καθ’ ὁδὸν ὑπὸ ἀπροσέκτου διαβάτου, γονατίζοντες ἐνώπιον γυναικὸς ἢ παραληροῦντες ἐν νείρῳ. Οὐδόλως ἐνταῦθα πρόκειται περὶ γλώσσης τοῦ λαοῦ καὶ γλώσσης τῶν λογίων, ἀλλᾶ περὶ διγλωσσίας τῶν αὐτῶν ἀνθρώπων, τῶν ἐχόντων γλώσσαν ζωντανήν, δι’ ἧς ἐκφράζουσι πάντα αὐτῶν τὰ αἰσθήματα καὶ τὰ πάθη, καὶ καταδικαζομένων νὰ μεταχειρίζωνται γράφοντες ἢ ἀγορεύοντες ἄλλην τινά, δι’ ἧς εἶναι ἀπολύτως ἀδύνατος ἡ ἔκφρασις παντὸς αἰσθήματος καὶ παντὸς πάθους.
Ὑφ’ οἱανδήποτε καὶ ἂν ἐξετάσῃ τις τὸ ζήτημα ἔποψιν, πάντοτε εἰς τὴν διγλωσσίαν καταντᾷ. Οἱ μὴ θέλοντες νὰ ὁμολογήσωσιν ὅτι ἄλλη εἶναι ἡ γραφομένη καὶ ἄλλη ἡ ὁμιλουμένη, πρέπει ἐξ ἅπαντος νὰ παραδεχθῶσι τὴν διχοτόμησιν εἰς γλῶσσαν μονολόγου και γλῶσσαν διαλόγου, εἰς γλῶσσαν συγκινήσεως καὶ γλῶσσαν ἀταραξίας, ἐχούσας ἰδιαίτερον ἑκάστη λεξικὸν καὶ ἔτι μᾶλλον διαφέρουσαν γραμματικήν. Κατὰ τὸν Σαιξπεῖρον τεκμήριον σφοδρᾶς συγκινήσεως εἶναι τὸ νὰ λησμονῇ τις ἐνδυόμενος τὸν λαιμοδέτην αὐτοῦ καὶ τὰς περικνημίδας· πρὸς ἐκτίμησιν ὅμως τῆς ψυχικῆς καταστάσεως τοῦ σήμερον Ἕλληνος φαίνεται πολὺ ἀσφαλέστερον γνώρισμα ἡ χρῆσις ἢ παράλειψις τοῦ ἀναδιπλασιασμοῦ.
(Some nice syntax there: "Under whichever i may examine j someone the issue j viewpoint i", "if they had good will that they should inform us which i the orators j these j do speak language i". Well, old hat if you're a Classicist, but that's the point: we don't speak in Classical syntax. Or German syntax, which Puristic was as indebted to.)
So, assembled wisdom of the internets and loyal readership: how shall we fix the Wikipedia citation for "diglossia"? Who coined the word diglossia as we now understand it—Roidis, Psichari, Pernot, or Ferguson?