How do you say "in excelsis"? You know, Latin for "in the highest," like in "gloria in excelsis Deo" or "hosanna in excelsis," quotations from angels in songs we seem to hear a lot in December.
Since it's not English, a common assumption is that the way it looks to English speakers is wrong. So even if you grew up singing it just like "excel sis," as an adult you've probably decided that must be incorrect. But what's the right way? Well, since the c in cælis and other Latin (and Italian) words with one of the "front" vowels i, e, œ, or æ following is said like "ch," it must be "ex chell cease," right?
Not really. In fact, you'd be better off singing "excel sis," though that's not necessarily the best way either. The reason for all this is a fascinating tale of sounds changing over time.
Let's start with the fact that in Classical Latin, c was always pronounced "k" everywhere. So if angels around the year 1 A.D. (or 4 B.C.) were singing Latin, they would probably have sung "ex kell cease." But they probably weren't singing Latin — the lingua franca of that time and place was Greek, but there's no reason to assume they would be singing that either. Their audience were shepherds, after all!
But their words made it into a Latin translation of the Bible. By the time Latin was the language of the Christian church, however, its pronunciation had shifted, and c before front vowels was being said like "ch." But! This didn't make "ex chell cease," because sc before front vowels was being said like "sh" — which is also the case in modern standard Italian (yes, prosciutto is "pro shoot toe," not "pros chew toe") — and xc was being said like "ksh"! So in church Latin, excelsis is like "ek shell cease."
The reason for all this is a thing called place assimilation. Compare two words, "coot" and "cute." Watch where your tongue is in them. In "coot" the back of the tongue touches at the back of the mouth. But in "cute," because the vowel sound starts at the front of the mouth, your tongue flexes forwards. It also moves farther forwards in "kit" than in "cat." It still sounds to us like "k," but over time that movement can push right up to the front of the mouth, so the tongue is pressing right against the ridge at the front… and then releasing more gradually because the vowel keeps the tongue close to the top. So, through a process called lenition (weakening), you can get what's called an affricate: a sound like "ch" that's a combination of a stop (such as "t") and a fricative (a hissy or buzzy sound produced by air friction, such as "sh" or "s").
But when there's the sound "s" before the "k," you can get a second kind of assimilation happening as the same time: manner assimilation. The "s" is already a fricative, and it's easier for the "k" to soften up as it moves forward into the "s" and become just a fricative, "sh." But you're not going to say "s-sh." The two sounds merge to "sh." And, just for added fun, in xc the "k" in the "ks" sound doesn't change, because it's not right before a vowel. So "ex kell cease" becomes "ek shell cease."
Latin (and its descendant Italian) isn't the only place this kind of change has happened. It happened in Old English, too, more than 1,500 years ago. Take the word ship. The old Germanic root started with "sk"; in Old English, the word was spelled scip — but pronounced "ship," because assimilation! But after we had established that, the English had ongoing contact with Norse speakers (because the Norse invaded), and in Old Norse the word was still skip. Which is where we borrowed skipper from. Ironically, since then Norse has also assimilated sk to "sh" in those places: Skip in modern Norwegian sounds like English "sheep." But by the time English borrowed skipper it had stopped running new words through that assimilation, so we keep the "k."
But all this doesn't mean that "ek shell cease" is the only way you can sing excelsis either. Latin changed further. By the Medieval era, it was no longer anyone's first language; it was being used fluently as a second language by scholars throughout western Europe, but each country had its own standard pronunciation. In Germany, for example, the c before front vowels moved forward not to "ch" but farther, to "ts." The languages that had descended from Latin — Italian, French, Spanish, Portuguese, Romanian — all conformed their Latin to how the sounds of Latin had changed when it became those languages. In France excelsis became like "ek sell cease" because the c before e and i had moved all the way forward and softened to "s" there, as you can see in any French word containing ce or ci.
And in English? We know what we do in English. Say the word excel. There you have it. Our pronunciation of Latin-derived c matched what the French did, due in no small part to French influence. (Meanwhile, many old English words that once had sc, like scip, got spelled with sh instead for clarity.) We also shifted the pronunciation of vowels to match how we said them in English, but that's a whole other ball of wax!
So anyway, at the time the songs that use in excelsis — "Angels We Have Heard on High" and "Ding Dong Merrily on High" — were written, the English pronunciation of Latin made it the same as you would say it as an English word. When were those songs written, by the way? Brace yourself: The English lyrics for both songs were written in the 1800s. But "Angels We Have Heard on High" was originally French. It's not entirely clear how much older the French version is, but evidence suggests that it, too, was written in the 1800s.
The Latin phrase, of course, was not made up in the 1800s; it's quoting from a Latin translation of the Bible done some 1,500 years earlier. If you want to go with the standard pronunciation for that, it's back to "ek shell cease." Since the angels wouldn't have been using Latin anyway, the choice is yours…
…but you don't really have a good excuse for singing "ex chell cease."