Indo-European Words for “Wheel”: Evidence for Transition from Agriculture to Pastoralism

The impetus for this post comes from a recent review by David W. Anthony and Don Ringe “The Indo-European Homeland from Linguistic and Archaeological Perspectives” (Annual Review of Linguistics 1 (2015): 199-219) that summarizes the best-in-class linguistic and archaeological arguments in favor of the Bronze Age Pontic Steppe theory of Indo-European homeland (contra the Neolithic Anatolian model). The Proto-Indo-European (PIE) “wheel” terminology and the earliest archaeological attestation of wheeled transport is an absolutely essential evidential nexus that can prove or disprove the Recent Pontic Steppe theory beyond reasonable doubt. From an archaeological standpoint,

“The invention of the wheel-and-axle principle, which first made wagons and carts possible, is solidly dated by radiocarbon after 4000-3500 BCE, a very well studied external fact (Bakker et al. 1999; Fansa & Burmeister 2004, Anthony 2007). This external fact ties late PIE to a real-world date after wheeled vehicles were invented, that is, after 4000–3500 BCE” (Anthony & Ringe 2015, 201-2).

Indo-European linguists supporting the Recent Pontic Steppe hypothesis adduce as proof not one word for ‘wheel’ but a whole lexico-semantic set related to wheeled vehicles. Anthony & Ringe (2015) illustrate the geographical distribution of reflexes of this set on a map:

KinshipStudies-IndoEuropeanWheelMap copy 2

The distribution of the reflexes of the PIE wheeled transport vocabulary by subfamily is pooled in the table below:

KinshipStudies-IEWheelWords copy

An important observation that can be made from these distributions is that the most divergent Anatolian and Tocharian branches have clear gaps in the wheeled-transport vocabulary compared to the rest of the IE subfamilies. These gaps are different: Anatolian has a reflex of PIE *H2/3eyHos ‘thill’, while Tocharian has reflexes of *kwekwlo– ‘wheel’ (with a meaning shift to ‘chariot’) and *wegheti ‘s/he transports it in a vehicle’. Outside of Anatolian and Tocharian, most branches display transparent reflexes of the reduplicated form*kwekwlo– with the same meaning ‘wheel’ derived from PIE *kwel– ‘to turn, to move around, to cultivate’ (Gk tellomai ‘turn around in circles’, pelomai ‘am in motion, go’, poleoo ‘go around, range, haunt (intrans.); turn up the soil (trans.)’, Lat coloo ‘I cultivate, I inhabit’, Skrt cárati ‘he roams’). Anatolian and Tocharian are also united in having reflexes of PIE *H2werg– with the same meaning ‘turn’: Hitt hurki-, Toch A warkant (< *H2wergwnt-, Adams, Douglas Q. A Dictionary of Tocharian B, 1999, 506), Toch B yerkwanto all mean ‘wheel’. Outside of Anatolian and Tocharian this root is poorly attested and, unlike PIE *kwel-, does not show a plethora of reflexes. The morphology of Hitt hurki and Toch *H2wergwnt- is different, which is usually taken as an indication of a possibility of independent formation from the same underlying root.

Anthony & Ringe spend a long time demonstrating the uniqueness of the *kwekwlo-/*kwekwlos formation shared by most IE languages. It’s the only verb-derived reduplicated formation shared by a large number of IE branches. Plus, it’s a zero-grade root with a thematic vowel and a nominative singular ending. They make a strong case that *kwekwlo– could not have been invented multiple times in already well-differentiated IE branches – the interpretation that the proponents of the Neolithic Anatolian model advocate for.

It’s noteworthy, however, that in Baltic – a subfamily with no reduplicated formations for ‘wheel’ – the same reduplicated root is found with the meaning ‘neck’ (Lith kaklas, Latv kakls). It’s an example of a reduplicated verb-derived noun that, unlike *kwekwlo– ‘wheel’, is found in just one branch of IE. But the very fact that *kwekwlo– can have a meaning other than ‘wheel’ in an IE subfamily suggests that at the time of the formation of the IE wheel-transport vocabulary and long thereafter the root *kwel– ‘to turn’ was productive and, as such, generated nouns in already isolated early IE subfamilies. The Baltic forms for ‘neck’ could not have been derived from *kwekwlo– with the existing meaning ‘wheel’ because the latter are not attested in the Baltic languages (OPruss kelan ‘wheel’, Latv du-celis ‘two-wheeled vehicle’). Slavic languages (that form a sister subfamily to Baltic, by many accounts), too, have only unreduplicated reflexes of IE *kwel– ‘to turn’ with the meaning ‘wheel’ (e.g., Rus koleso) suggesting that Lith kaklas and Latv kakls ‘neck’ are independent reduplicated formations from the underlying verb *kwel– ‘to turn’ that occurred in an early IE geographic area devoid of *kwekwlo– ‘wheel’. Besides Baltic and Slavic, other IE branches have unreduplicated derivatives of *kwel- in their wheeled-transport vocabulary (e.g.., Gk polos ‘axle’).

This also suggests that Fortson’s idea (2010, p. 130) that kwekwlo– was “an expressive neologism for a new gadget,” which Anthony & Ringe (2015, 205) consider a “reasonable speculation” is not reasonable at all, as neck is not a “new gadget.” A more reasonable speculation would be that reduplication iconically represented repetitive circular movement and a product or an agent thereof. This understanding may be helpful to semantically differentiate between two forms for ‘wheel’ reconstructed by Anthony & Ringe (2015) for PIE: *kwekwlo– and *Hroto-: if *kwekwlo– emphasized repetitive, circular but ultimately static movement, *Hroto– interpreted wheel as an enabler of linear, forward movement (from PIE *Hret– ‘to run, to ride’: OIr rethim ‘I run’, Lith ritu ‘I roll’). The two concepts may bleed into each other in daughter languages (comp. Lat rota ‘wheel’ but also ‘circle’ and ‘potter’s wheel’), but the possibility that in PIE wheel was interpreted in two very different ways has implications for the question of PIE homeland and the timing of IE migrations and language divergence.

Earlier, I attempted to show that IE labiovelars (kw in PIE *kwel- and *kwekwlo-) have a more complex system of reflexes in daughter languages. The triple split of PIE *kw, *gw and *gwh into k/p/t, g/b/d and gh/bh/dh (in different phonetic environments, namely before u, back vowels and front vowels, respectively) described for Ancient Greek is in fact a PIE phenomenon. One of the illustrative examples of this overlooked phonetic law has a bearing on the origin of the IE term for ‘wheel’. It’s known that IE root *pol-/*pel- yielded Lat pollen ‘finely milled flour’, pulvis ‘dust’, Gk palee ‘finely milled flour, dust’, Skrt palalam ‘ground seeds’, Lith pelenai ‘ashes’, pelene ‘hearth’, Slav *poleeti ‘burn’, *polmen ‘flame’ and *pepelu/*popelu ‘ashes’. If the etymological material that I advanced in my previous post is correct and labiovelars indeed gave labial reflexes before back vowels already in PIE times, then PIE *kwel– and *pol– are cognates. The meaning ‘ashes, burn, flame, hearth’ found in Balto-Slavic suggests that burning was perceived by early Indo-Europeans as a repetitive, circular movement and ashes as the final product of a cycle. (Comp. in this context tellomai ‘turn around in circles’, teleoo ‘finish’, telos ‘end’.) The robustness of this hypothesis is indicated by the fact that it provides a verbal source for the pol-nouns, which is missing if the two cognate sets are treated separately. Another indication that we are on the right track here is Slav *pepelu/*popelu ‘ashes’ that has the same rare reduplicated morphology as IE *kwekwlo– ‘wheel’ and Lith kaklas/Latv kakls ‘neck’. Importantly, there is a strong semantic link between the two cognate sets: Lat pollen ‘finely milled flour’, Gk palee ‘finely milled flour’, Skrt palalam ‘ground seeds’ make perfect sense in the light of OPruss maluna-kelan ‘mill wheel’ and Lat coloo ‘I cultivate’. Importantly, the semantic link between the two branches of PIE *kwel– – the large *kwel-group and the small *pol-group – exists not on the level of wheeled-transport vocabulary but on the deeper level of agricultural vocabulary. It appears that early Indo-Europeans may have developed the concept of a wheel first on the basis of the idea of repetitive, circular but static movement as manifested in such agricultural tools as a quern, a millstone or a mill wheel. The concept of linear, progressive, forward movement, which is very appropriate for a population that migrates to new lands) and is best represented by PIE *Hroto-, now completely devoid of static agricultural connotations, emerged as a second phase in the evolution of IE wheeled transport.

IE *H2werg– ‘to turn’ manifested in Hitt hurki-, Toch A warkant, Toch B yerkwanto (< *H2wergwnt-) ‘wheel’ is a puzzle because it’s clearly old but it does not have strong cognates (and absolutely no cognates in cultural vocabulary) outside of the two most divergent branches of IE languages. But there’s one IE cognate set that needs to be re-examined in the light of a possibility that early Indo-Europeans derived their first terms for wheeled transport drawn forward by horses from pre-existing static, agricultural applications of the wheel. It’s IE *gwreH2won/*gwreH2nu– ‘millstone, quern’ represented by Toch B karwene ‘stone’, Skrt graavan ‘stone for pressing the soma’, Arm erkan ‘quern’, OIr brau (Gen. broon) ‘millstone, quern, hand-grinder’, Goth asilu-qairnus ‘donkey-mill’, OHG quirn, OEng cweorn, ONorse kvern ‘quern’, OPruss girnoywis, Lith girna ‘millstone’, girnos (pl.) ‘quern’, Latv dzirnus (pl.) ‘quern’, OCS zruny ‘quern’. IE *H2werg– ‘to turn’ and IE *gwreH2won/*gwreH2nu– ‘millstone, quern’ have a lot in common phonetically and morphologically: both have a second laryngeal, a resonant and a labiovelar (Toch *H2wergwnt also shares with *gwreH2won the suffixal enlargement -n-) but the laryngeal and the labiovelar are placed differently in the two roots. Their placement suggests that the two roots are related through a metathesis, so that the more fundamental PIE *gwerH2w– ‘quern’ metathesized into *H2wergw– that yielded Hitt hurki-, Toch A warkant, Toch B yerkwanto ‘wheel’. Consequently, it’s unlikely that *gwreH2won/*gwreH2nu– is derived from *gwerH2– ‘heavy’ (EIEC, 474). Some key reflexes of IE *gwerH2– ‘heavy’ (Gk barus, Lat gravis) do not map well onto the reflexes of *gwreH2won/*gwreH2nu– (missing from Greek and Latin). Instead, there must have been a verbal root *gwerH2– ‘to turn’ that survived in Toch as wark- (Adams, A Dictionary of Tocharian B, 547-8).

A special mention should be made of the fact that Arm erkan ‘quern’ shows remarkable similarity to Toch B yerkwanto ‘wheel’. It’s currently assumed that the Armenian form underwent a typical Armenian metathesis from the CrV-type root to the rC-type root + prothetic vowel e-. But now it appears that, at least in this case, metathesis affected the Hittite and Tocharian forms as well.

Pending the ultimate phonetic validity of these interpretations, this analysis paints a picture of early Indo-Europeans developing their wheeled-vehicle vocabulary on the basis of pre-existing agricultural vocabulary. Rotation as a repetitive, circular, static movement employed by agriculturalists was reinterpreted as linear, progressive movement forward. Querns and millstones, which are archaeologically attested from the 9th millennium BC (first in Neolithic West Asia), seem to be the direct antecedents of wheels used for transportation, warfare and migration. This analysis also suggests that, by the time wheeled-vehicle vocabulary began developing among early Europeans, the agricultural vocabulary had long been in place, which is consistent with the Bronze Age Pontic Steppe theory. IE *gwreH2w-/*H2wregw- should be included into the IE wheeled-transport vocabulary. Although as a form for ‘wheel’ it’s an Anatolian-Tocharian isogloss, it belongs with a wider IE cognate set having a clearly older meaning – ‘quern’. It means that, although early Anatolian and early Tocharian may have been geographically isolated from the rest of IE languages, they don’t appear to be divergent (at least not in this cognate set) from other IE subfamilies in a phylogenetic sense. The gaps in the wheeled-transport vocabulary observed between Anatolian, Tocharian, on the one hand, and the rest of IE languages, on the other, should be interpreted as lexical loss in Anatolian and Tocharian, not as the preservation in Anatolian and Tocharian of an earlier “stage” in the IE lexico-semantic evolution.


Leave a Reply