First Latin scholar: Crispin, I have some fresh lime. Would you decline a tequila?
Second Latin scholar: Certainly. Tequil-a, tequil-am, tequil-ae, tequil-ae, tequil-ā. Tequil-ae, tequil-ās, tequil-ārum, tequil-īs, tequil-īs.

That’s a very old joke, referring to Latin first-declension feminine nouns. (My, what a laugh we used to have at the expense of those poor Classics students.)

But the joke’s relevant here because, as a follow-up to my previous post about the plurals of English nouns ending in -us, I’m going to write about English nouns ending in -a. Many of these are derived from Latin first-declension feminine nouns which (as Crispin points out) take -ae in the plural.

The first problem is deciding how to pronounce -ae. The Oxford English Dictionary favours /iː/, to rhyme with “ski”. But there’s a body of opinion, in British English at least, that favours /eɪ/, to rhyme with “say”, and even a minority view that champions /aɪ/, to rhyme with “sky”.

But however you say it, we have alga/algae, antenna/antennae, formula/formulae, lacuna/lacunae, larva/larvae, nebula/nebulae, vertebra/vertebrae. And the female equivalent of alumnus is alumna/alumnae.

As with all unusual plural forms, there’s a tendency to regularize. Here’s the Google Ngram for the two plural versions of formula:

In the nineteenth century, formulae was the more popular; now formulas is winning. In the Google corpus, antennas achieved dominance over antennae as recently as the 1980s, but nebulae is still dominant over nebulas (though perhaps because the Google corpus contains rather a lot of scientific literature). And some first-declension feminine nouns have almost never taken the Latin plural form: area/areas, arena/arenas and era/eras. (The Oxford English Dictionary gives the plural of verruca as verrucae, with a soft “c”—but that just seems like a guarantee of incomprehension from your listeners.)

Then there are the Latin-derived English nouns ending in -a that are not first-declension feminine singular, and which therefore cannot take an -ae plural ending. Opera is already a Latin plural, of the third-declension opus, “work”, so the English plural has to be regular—operas. And we have Saturnalia/Saturnalias for the same reason—Saturnalia is the plural of third-declension Saturnalis, “something pertaining to Saturn”. Candelabra is the plural of second-declension candelabrum, “candlestick”, and so becomes candelabras in the plural. Likewise, agenda is the plural of second-declension agendum, “task to be done”, and becomes agendas. (Though when committee meetings become really tedious, there is some diversion to be had from referring to a single agenda item as an agendum. Or is that just me?) Insignia is the plural of third-declension insigne, “emblem”—it’s still widely used as a plural noun in English, with insigne as its rare singular; but the singular “an insignia” is increasingly common, with an associated plural insignias. And propaganda is not a Latin noun—it derives from the Latin Congregatio de Propaganda Fide, “Congregation for the Propagation of the Faith”, a missionary wing of the Catholic Church. In current English it tends to be a mass noun, with no plural—but Google Ngram records a couple of interesting spikes in the usage of the plural propagandas:

Can it be merely a coincidence that the spikes correspond to the rise of communism and fascism?

The singular “a bacteria” (with plural bacterias) is sometimes sighted in the wild, but is still considered an error—bacteria is the plural of bacterium. The same can be said of strata, which has been making sporadic appearances as a singular noun since the eighteenth century. (The Merriam-Webster Concise Dictionary of English Usage wearily describes it as “a Latin plural with ambitions to become an English singular”.) But despite its long history, the singular use of strata is still considered erroneous—the singular is stratum. And then there’s datum/data. Some of us still say “the data are …” but “the data is …” has been around for a century or more, and is now the dominant usage. The singular data is generally treated as a mass noun, however, so there’s no need to reach for datas.

Then there are the English nouns ending in -a that have Greek, rather than Latin, origins. These would take the Greek plural ending -ata, but that usage is fading fast, and all can be safely regularized. Stigma/stigmata is still standard in a religious context, but stigmas is more common in general use. Miasma/miasmata is likely to be interpreted as rather affected, as will dogma/dogmata. And diploma/diplomata, dilemma/dilemmata and drama/dramata have never been popular, with diplomas, dilemmas and dramas having always been the dominant usage. (Though the marvellous phrase pedagogical dramata will turn you up a few Google hits—it means “educational computer games”, apparently.) Both diorama/dioramas and phantasmagoria/phantasmagorias are Greek in origin but came to us via French—they’ve never taken the plural -ata ending in English.

A couple of Greek plurals ending in -a have been trying to edge their way into singular use—but criteria and phenomena are the plurals of criterion and phenomenon, and treating them as a singular is still considered an error.

And finally, there are all those nouns ending in -a that have come to us from other languages, which are simply too numerous to list. But we have credenza/credenzas (Italian), pavlova/pavlovas (Russian), yarmulka/yarmulkas (Yiddish), quagga/quaggas (Nama), jerboa/jerboas (Arabic), bazooka/bazookas (nonsense) and … well … tequila/tequilas (Spanish).