Tocharian A and Tocharian B both have mildly interesting reflexes of the Proto-Indo-European word for “wheel”: kukäl and kokale, both meaning “wagon” or “chariot”.
They are mildly interesting because they can’t possibly have derived from the same version of the P.I.E. word for wheel. Let’s take a look at why.
Tocharian phonology: very briefly
Tocharian as preserved has undergone some extremely strange sound changes at some point in its deep past. Among other things, the classic P.I.E. catch-all vowel *e becomes *ä most of the time in in Proto-Tocharian, as does *i. Meanwhile, P.I.E. *o becomes *e, with occasional diversions into even more exotic forms. On the consonant side, PIE’s voiceless, voiced, and breathy voiced stops all reduce to a single unvoiced series, as do palato- and labiovelars.
I suspect we can blame these exoticisms on Tocharian’s prolonged contact with early forms of Samoyedic (Uralic languages are known for their dearth of consonants and their overabundance of vowels). Which raises interesting questions about who the speakers of Tocharian actually were and where and when they came from.
Conventional wisdom associates them with the Afansievo culture of southern Siberia. It’s easy to see why: they’re in roughly the right place to have had descendants who later ended up in the Tarim Basin, and we know they were Indo-European (a direct and contemporary offshoot of the Yamnaya culture). But they’re about 2000 years too early: Proto-Samoyedic likely dates to about 1000 BC, and the Afansievo, by then, had been long replaced in the region by their distant cousins, the Indo-Iranian speaking Andronovo culture.
But this is a subject for another time.
Getting back on track
Although words inherited from P.I.E. can still be easily identified, they look very different from in many other branches. (Although if you spend long enough looking, you’ll start to see Tocharian words as being a bit like Germanic if it hadn’t undergone Grimm’s law, Indo-Iranian if it hadn’t spent quite so much time palatalising every consonant it could get its hands on, or Balto-Slavic if it hadn’t become entirely deranged).
kukäl and koakle are both quite instructive of this process. Most lexica give them a P.Toch. predecessor *kuk(ä)le, also meaning “wagon” or “chariot” (the development from “wheel” to “wheeled vehicle” being seen in OCS but also more recently in English).
How do we get there? Well, as we can see, *kʷekʷlo- contains 2 consonant sounds and two vowel sounds. According to the normal processes of Tocharian historical phonology, they should undergo the following changes between P.I.E. and P.Toch.:
*kʷ > *k
*e > *jä via *je
*l > *l
o- > *e
Clearly however, something else is happening here because this would give us something like **čälke in P.Toch. (the palatal “on-glide” developed by *e serves to palatalise the preceding stop), from which we’d then expect to see very different reflexes in A and B. So there’s obviously a special case operating here. What could it be?
Most likely, it’s a product of rounded vowel umlaut, in which a non rounded vowel is rounded when followed by one that is. If P.I.E. *o persisted long enough before becoming *e, it would have affected a change from *e > *ọ in its neighbour before the change from *e > *je > *jä > *ä could be completed. (As for why it’s happening here, perhaps, see below.)
So we might have a process like:
*kʷekʷlo-
*keklo- (delabialization)
*kọklo- (rounded vowel umlaut)
*kọkle (P.I.E. *o > *e)
This is much more satisfying: *kọkle in P.Toch. can easily give kokale in B, with an epenthetic *ä emerging at some point and becoming a in the interim.
But then we still have to explain kukäl and its unusual u.
The assumption I’ve seen made for this is that both kukäl and kokale must descend from *k(ʷ)ukʷlo(s), the same zero-grade form of the noun that produces Greek κύκλος (the ordinarily-assumed form *kʷekʷlo- should give us a Greek **τεπλος, which is unattested).*
However, this isn’t without its problems: a hypothetical early *kukʷlo(s) would develop as follows in Tocharian:
*kukʷlo(s) (or *kʷukʷlos, which would probably resolve the same way)
*kuklos (delabialisation)
*kwäklos (P.I.E. *u > *wä following *k)
*kwäkle (P.I.E. *o > *e)
*kwäkäle (epenthesis)
This gets us neatly to A’s kukäl. A lost all final vowels early, and tends to delete P.Toch. *ä in open syllables which are immediately followed by another syllable with an open vowel (e.g. śäpm < *säpän “sleep”).
But we’re then still at a loss to explain kokale. Although Tocharian B does occasionally see P.Toch. *u change to *o, this appears to be an irregular development that occurs only when *u is directly adjacent to a resonant. This discounted, *kukʷlo(s) should yield Tocharian B **kukale. (This, by the way, is also the result we would expect to see if *kʷekʷlo- had become *kwäkle in P.Toch: both A and B the sequence -wä- usually becomes *-u-.)
Brief aside: Adams (2015) gives the preform as *käuk(ä)le from *kʷukʷlos, which is also problematic: *käukäle would give the “correct” result in A, but in B it would give us**kūk(a)le (the P.Toch. diphthongs *äw and *äy reliably reducing to ū and ī in B) which leaves us back in square one. Moreover, *kʷukʷlos would resolve in P.Toch. as *kwäk(ä)le, similar to above, rather than *käukäle. I’m tempted to assume this is a typographic error.
Why not both?
The simplest explanation for this contradiction is that both *kukʷlos and *kʷekʷlo- existed in Proto-Tocharian and were developed separately in A and B. I’m tempted to suggest the reason *kʷekʷlo- undergoes vowel rounding umlaut to eventually give us kokale in B is because the precise from we should reconstruct is *kʷekʷlóm, as also seen in reflexes preserved in Indo-Iranian and Germanic.
Which poses us with the question: were both of these forms “native” to the Tocharian branch, or was one of them borrowed from a neighbouring Indo-European group? It’s difficult to be sure, but perhaps worth noting that the earliest states of both Indo-Iranian and Tocharian reflexes of *kʷekʷlóm would have been identical.
*The Greek reflex is also sometimes derived from *kʷokʷlos, but this seems wrong: ordinary processes of change would give us **poplos from *kʷokʷlos or *koplos in some dialects. However, if we assume a u vowel in the original form, we can reach kúklos easily: *kʷukʷlos simply becomes *kuklos by application of the boukolos rule (a labiovelar delabialises when adjacent to u).