David J. Peterson (dedalvs) wrote,
David J. Peterson

O Circumfix, O Circumfix...

So way, way back on August 20th, Twitter user bestilheiro tweeted the following:

@Dedalvs I've been reading about your work in GoT (congrats!). You said there are no circumfixes in English. What about Dylan's a-changing?
Good question! As I told him then, though, it's probably not answerable in 140 characters, so that's why I've come here.

First, the circumfixes I know of are never quite so simple as they appear. For example, in German, you often see what might be described as a "circumfix" in the past participle. If you have a verb like wohnen, "to live", the past participle is gewohnt. Based on those two samples alone, one might say that -en is the infinitive ending and ge- -t is the past participle ending.

Unfortunately, it's not that sample. Most of the time, the -en suffix is associated with the infinitive, but there are times when it isn't (e.g. sein, "to be")—and, more to the point, you see that -en ending in several other places in the paradigm, for example:

  • wir wohnen "we live"

  • sie wohnen "they live"

  • wir wohnten "they lived"
Etc. The suffix is everywhere. And the same is true of the other suffixes commonly associated with the verbal paradigm: -e, -st and -t.

Looking at the past participle, most (not all, but most) past participles begin with ge-. That prefix is pretty much only associated with past (or perhaps "perfect" is a better term) participles. Calling ge- -t a circumfix proper posits a particular ordering—i.e. the construct ge- -t is added to the stem wohn. Is that right? I'm not so sure.

In fact, taking a glance at a number of German verbs, it looks to me like the perfect participle is built off the preterite stem if it's regular (i.e. ends in some kind of t). If it isn't, the perfect participle is built off the infinitive (with the possibility of a vowel change). Here are some examples (infinitive ~ preterite stem ~ perfect participle):

  • wohnen ~ wohnt- ~ gewohnt "live"

  • arbeiten ~ arbeitet- ~ gearbeitet "work"

  • laufen ~ lief- ~ gelaufen "run"

  • nehmen ~ nahm- ~ genommen "take"

  • wissen ~ wusst- ~ gewusst "know"

  • schützen ~ schützt- ~ geschützt "protect"
So, in a phonological sense, you might call it a circumfix (i.e. you couldn't have *gearbeit), but doing so seems to be missing a generalization, in which case calling it a circumfix might not be so useful.

The other languages I know of with "circumfixes" are rather similar. Take Arabic, for example. You have a root k-t-b which has to do with writing (e.g. kitaab "book", kaatib "writer", etc.). With that root, you then have a word like maktaba, which means "library". That looks like you added a ma- -a circumfix to a root (with some other phonological stuff going on in the center). But you also have a word maktab, which means "office", and a word kataba which means "he wrote" (the suffix isn't quite the same on kataba, but there are plenty of words that use that suffix by itself), so calling it a circumfix doesn't seem quite right.

Even Georgian (whose derivational strategies ultimately inspired Dothraki) has some question marks regarding its circumfixes. Many of them involve shared endings, for example:

  • titi "finger" ~ satite "thimble" (sa- -e)

  • lamazi "beautiful" ~ silamaze "beauty" (si- -e)

  • kartveli "Georgian" ~ sakartvelo "Georgia" (sa- -o)

  • marili "salt" ~ umarilo "without salt" (u- -o)
In fact, there is no part of any circumfix in Georgian that's unique (even the -et used in the sa- -et place name circumfix enjoys use as a suffix on its own to derive some place names).

In Dothraki, the circumfixes are a little more stable, but even there there's some repetition. For example:

  • ido "wooden" ~ savidosalat "to be protected"

  • qoralat "to take hold of" ~ eqorasalat "to let go of"
I think you'd really have to delve into the histories of these languages to figure out just where these pieces came from and what status they've had (and continue to have) in the language to figure out just how robust any given circumfix is.

Now let's go to English and the question at hand. In Dylan's "The Times They Are a-Changin'", we see something that might be a circumfix. Of course, he didn't invent it; that "a- -ing" thing's been in English a long time. And its etymology isn't hard to piece together. See, the "-ing" form of the verb is a participle, and not finite (i.e. can't serve as the main verb of a sentence). It can also be used as a noun (e.g. "Running is tiring"). Way back when, you used the nominal form with the preposition "at" to indicate the progressive (e.g. "I am at going"). Over time, that got whittled down, so you got "I'm a'goin'". That form still may be current in some dialects; I'm not sure. Now, though, we have it as simply a "folksy" version of the progressive, which we form simply by using the present participle along with "to be" (the preposition no longer necessary).

So, does that count as a circumfix? Personally, I don't think so. Certainly the "a'" no longer has any real tie to "at", but we use the "-ing" form by itself plenty. There is no grammatical difference between "-ing" and "a- -ing", just a kind of sociological difference (in most dialects). While there's no question that phonologically the "a-" part and the "-ing" part are attached to a single stem, I'm not sure that's enough to rescue the circumfix analysis (note also that "un- -s" is also attached to "cool" in the same way in "History of the uncool's rise to coolness"). But I'm only, like, 85% sure.

Perhaps a more convincing circumfix in English is the "e- -ate" in "elongate". True, we have "e-" elsewhere and "-ate" elsewhere, but there's no *elong or *longate. And while there's a historical explanation (we simply derived it from the Latin ēlongāre), the presence of this mini-paradigm (along with "embolden") has contributed to the pressure we feel to have a causative circumfix. Consider the nonce (but sensical) coinage from The Simpsons "embiggen". Certainly "biggen" should've been cromulent, but somehow, "embiggen" makes sense. Go ahead and google up some nonce verbs like "enlargen" and see how many hits you get.

So, to answer the question, I think the "at x'ing" construction was on its way to becoming a circumfix, but given the fact that it dropped off and we have "x'ing" which means the exact same thing, I don't know if I'd consider it a true circumfix (or at least a true morphological process distinct from the "-ing" form). But you can decide for yourself. That's what the internet is for. :)
Tags: conlang, dothraki, linguistics

Comments for this post were locked by the author