Sunday, September 25, 2011

The curious lives of French prepositions

When you study comparative Romance philology, it’s only a matter of time before you notice that French exhibits some prepositional anomalies. These have interested me for years. For heaven’s sake, you might wonder, how do we explain Spanish con, Italian con, Portuguese com, Romanian cu, all from Latin cum, but French avec? Where did parmi “among, between” come from? And how about dans “in”? What about chez “at the home of”, which is so useful it has made its way even into English? None of these are found in the other Romance languages — and there are plenty of other examples — but why?

I was asking myself questions like these twenty years ago, and though I have now long known the answers, it struck me that some of my readers might be interested as well. Since chez is my favorite example, I’m going to save it for last. Let’s start with the strange-looking avec.

This word is the modern reflex of Middle French avecques, in turn from Old French avoc, avuec, avoec. The latter is the spelling found in The Song of Roland, used prepositionally near the beginning of the poem (l. 186), but used adverbially near the end (l. 3626). The word is a contracted or elided form of Latin apud hŏc “with this (thing)”. The path would have been something like Latin apud hŏc > ap[ud] hŏc > Vulgar Latin *abhoc > Old French avoc. What is most interesting is that French kept a vestigial trace of the pronoun, hōc, the neuter form of hīc “this”. Quite separately, the Latin preposition apud eroded directly into an Old French preposition od, ot, o “with” — which occurs with greater frequency than avoc in Roland. There is also one occurrence of the construction o tot “with all” (l. 1357), with the same meaning as modern avec, and which looks something like the obsolete English withal.

Originally, Latin apud and cum had different connotations, the latter more often associated with coincidence of time than with people or things, but French took one path, all the other Romance languages the other. Why is a difficult question, one that would require a lot deeper investigation that we have time for here, but it was during the Carolingian/Merovingian dynasties that avoc began to outshine od, likely under the influence of Frankish (i.e., Germanic) constructions and preferences. (Let us remember too that Old English wið was originally “against”, preserved now only in withstand; whereas, it was mid that connoted the sense of our modern preposition “with”.)

Many of the other anomalous French prepositions evolved along similar lines from Latin collocations of either preposition + noun/pronoun, preposition + preposition, or preposition + adverb. By contrast, the other Romance languages (particularly Italian) usually derived their forms directly and solely from the original Latin prepositions. The “French model” explains parmi and dans, among many others.

Parmi, also attested in Roland, is formed from a preposition + noun, from L per mĕdium “in, through the midst of” > VL per mĕdiu > OF par mi, parmi. The preposition dans is similarly formed. The usual Romance preposition from Latin in “in(to)” became French en (cp. Sp en, It in, P em, Ro în), but dans came from OF denz, in turn from L de ǐntus “from within”. There is also an alternative (and redundant) form in dedans, from OF dedenz (< L de de ǐntus). Other “compound” prepositions of this sort include avant, dehors, dessous, dessus, delà, dépuis, avant, devant, envers, devers, etc. Some of these have direct cognates in the other Romance languages, but not all of them.

As promised, my favorite: chez. This wonderful preposition is unique among the Romance languages, and so valuable and concise that is has been borrowed from French. We all know what it means: “at the house of”, as in, “party this weekend chez Jason and Jennifer.” (That’s just an example; please do not knock on our door tonight unless you come bearing wassail! ;). This one is the real anomaly, because it is essentially just a noun repurposed into a preposition. This becomes pretty obvious when you consider that the only way to translate it requires the use of a noun, “house, home, etc.”.

So, as you may have guessed already, French chez goes back to Latin casa “house”. The c > ch sound change is among the most common in the language; cp. OF castel, chastel < VL castellu, OF cheval < VL caballu, OF chien < VL cane, OF chose < VL causa, and hundreds more.

The use of chez as a preposition comes along after Roland. In Old French, chez was not a preposition, but rather a noun meaning “house”. The prepositional use today has pushed this noun out of the language. Instead, the common French word for a house is maison, of which the English cognate form is mansion (< VL mansiōne < L manēre “to stay, remain”). The other Romance languages retain the derivatives of L casa in common use, cp. Sp casa, It casa, P casa, Ro acasă — but the Latin noun still survives in modern French as the specialized noun case “a small house or hut; or a square on a chess-board”. In the 11th century, the usual construction would have been je vais à chez Gautier (translating Latin vado ad casam Walterii), but à chez contracted rapidly to chez alone.

And the rest, as they say, is histoire. :)


  1. Actually Swedish "hos" means exactly "chez" - and its etymology also includes a word for house, this time the Gmc one.

  2. How do you do breves and longae signs over vowels? Your computer or sth anyone can do by html?

  3. Jason

    Brilliant, this excellent post made me look over at my pile of Old French books and say come on take the plunge!! But Old Irish first!!!

    Cheers Andy

  4. Excellent article! I studied French from kindergarten to college, without much exposure to its Romance cousins, so I never thought anything of its uniqueness in this respect. My pronunciation went out the window when I started learning Japanese a few years ago... mais j'adore toujours le français.

  5. @Hans-Georg: Thank you for letting me know that something exactly analogous occurred in Swedish. Very interesting! As for the diacriticials, anyone can produce them, so long as the font in question contains the glyph. The Unicode standard defines a whole welter of consonants and vowels with diacriticals and other decorations. Exactly how to produce them will depend on the kind of computer and software you are using, but here is one useful page (just as a starting point).

    @Andy: Glad you enjoyed it. I’ve never done much with Old Irish beyond a word here or there, but I can highly recommend Old French as a rich and rewarding field of study.

    @Geoff: Thanks! Kindergarten to college — seventeen years is a long time to study a language, even one’s own cradle tongue! I have formally studied French for less than half as long (sixth grade through twelfth), though I began before that on my own and have persisted on and off ever since: ça me plaît beaucoup aussi. :)

  6. And of course the reason French is strange, as distinct from the equally interesting details of how it is strange, are that it got a heavy Germanic overlay from those East Germanic speakers the Franks, even becoming temporarily a verb-second language. Indeed, it is almost fair to say that Old French is essentially Middle English relexified with Romance vocabulary.

  7. @ John Cowan:
    I like your essentialist definitions, however:

    - Franks are West Germanic (and Goths are East Germanic, never mind that Visigoths got west of France into Spain and Franks got France from the East, earlier on Visigoths had been a bit East of at least North as seen from Franks)

    - Germanic overlay is true about accent for a period, also about some lexical items

    - as for Middle English relexified with Romance vocabulary, the problem is Middle English itself is Anglofrench partially relexified with Anglo-Saxon vocabulary. But in some writers due to family heritage, no doubt it may be Frankish (a k a Dutch) relexified with Romance vocabulary of surroundings. No doubt there were talkers who differed from that, though.

    However, les bonnes gens vs les gens normaux seems to hanker back to sth like die gute Frau vs die Frau ist gut, at least if other nouns with similar adjective gender shift on adjectives before and after noun follow same pattern, since Fr masculine is close to German undeclined forms.

  8. @ Jason Fisher: thank you for the page I abridged as!

    It is not the only trace of Gallic influence on Swedish, presumably during Middle Ages.

    Older o under some conditions (notably=ō) becomes pronounced as oo in both Swedish and provençal, and Sw. nation, tradition is very close to monegasque pronunciation of same words.

  9. Wait, I will try lower case o with macron again, using html: ō

  10. phew, it worked, when I tried alt button I got higher case M!

  11. Interestingly, Catlan, which IMHO is closer to Old French than any other living language that I know, uses both mas and casa for house, with mas being more a country house any hence akin to the English meaning of mansion. It also uses amb for with rather than the more usual romance cum/con stem. I guess amb is somehow related to avec.