A few years ago I watched a BBC historical documentary about ancient York and I was startled to hear the cultured-sounding narrator (obviously an actor trained to sound cultured) pronounce the second name of the emperor Septimius Severus as though were “sever us” (/’sɛvərəs/), since every reference gives the pronunciation as /sə’vɪərəs/ (“severe us”). I wondered where that had come from.
I am a long-time participant in international folk-dancing, and among the dances I liked to do are many from the Republic of Macedonia, with names such as kostursko, nevestinsko and ovčepolsko oro. The boldface syllables are the ones that are stressed, since stress in Macedonian is very regular: in native words of three or more syllables, the stress is always proparoxytone, that is, on the antepenultimate syllable. And yet almost everyone in the folk dance community says these words as paroxytone, that is, stressed on the penultimate, as if they were Polish (another language with regular stress, but with a different rule); this includes people who learned these dances from the famous Macedonian dance teacher and musician Pece Atanasovski, whose name they usually say as Atanasovski.
English is a language with notoriously irregular stress. Not only is not obvious from the written form of a polysyllabic word where the stress might be, but often there are different stress patterns for the same word. Sometimes the stress is different when the word is a noun and a verb (such as protest); sometimes the difference is transatlantic (as in café or contribute, or names such as Bernard, Maurice, Barnett or Burrell); and sometimes there is simply no agreement for a given word, such as address, detail, insurance and many others.
But when it comes to Latin words — by which I mean words taken directly from Latin, with no change in spelling — there is an old rule that such words, even if pronounced in the traditional “Anglo-Latin” way (with letters read according to their usual English values — the way well-known Latin names, words and phrases are usually pronounced, as well as Latin terms in law, medicine and astronomy), the original Latin stress is preserved. There is a caveat here: there are words of Latin origin that came into Middle English from Old or Middle French and in the evolution into modern English recovered their Latin spelling. Examples include senator and liquor, which in Middle English were senatour and licour, oxytone (stressed on the last syllable) as in French. U.S. English has a great many such words — favor, honor, vigor and the like — which in British English have kept their medieval -our ending. Such words, if they have more than two syllables, don’t necessarily follow the rule.
According to this rule, then, there is no excuse for Severus. But the rule seems to have given way, in recent times — I don’t know how recent — to another one, applied as a kind of default to foreign-looking words that are relatively unfamiliar, namely that words ending in a consonant are read as proparoxytone, and those ending in a vowel are paroxytone.
The first part of this rule (which should be called a misrule) explains not only Severus but also such common words as abdomen (Latin abdōmen) or acumen (Latin acūmen); Donald Trump was unfairly (for a change!) ridiculed for saying the latter in the traditional way (as I do). Beyond Latin, it explains the common proparoxytone (mis)pronunciations of such names as Esteban, Vladimir, Wallander, Martínez (as heard from British soccer announcers), Istanbul, Stalingrad; and Hanover is the standard English form of Hannover.
The second part of the rule explains the usual pronunciations of incognito and patina (not stamina, which in common enough use to have kept its traditional stress, as are such words as Africa and America), where, oddly enough, the stressed i sounds like “ee” (as though the words were Spanish or Italian) and not the usual “long i” sounds in Anglo-Latin words (such as vagina). People who told me about the movie Ex Machina usually pronounced the second word as “masheena”
It also explains the common English pronunciations of paprika, basmati and a great many Italian place and personal names such as Pesaro, Brindisi, Stefano, Cesare, Pecora — all proparoxytone in Italian — and Podestà (oxytone). In the movie Big Night, in which the main characters were supposedly Italian immigrants, the word timpano was pronounced timpano. The Irish surname Costello is invariably pronounced Costello in America (to be sure, the most famous Americans with that name — Lou and Frank — were not Irish and were not originally named Costello). Sofia, the name of the capital of Bulgaria (Sofiya in Bulgarian) is usually pronounced like the name Sophia. And then there’s karaoke, usually pronounced like “carry-okie.” Not to mention Bacardí, pronounced Bacardi (even in its own commercials) despite the prominent acute accent on its labels.