Indo-European wheel words

by Edward Pegler on 25 May, 2011

What exactly is the evidence that Proto-Indo-European’s had wheels and wagons? And what is the significance of *kwekwlo-?

A partially reconstructed, wheeled toy from the Cucuteni Tripolye B2 culture

Wheels, it appears, are not that old. They first turn up, in the form of moulded clay wheels on toys, in Ukraine’s Tripolye B2 culture (dated around 3800BC). After this date there is an explosion of evidence for wheels across Europe and down into the Middle East.

So (just going back 500 years to be safe) wheels probably weren’t invented much before, say, 4300 BC then. Why does this matter? Frankly it doesn’t.

But for people who study or are interested in the origins of Indo-European (IE) languages then this date can get them quite heated. This is because there are currently two main stories for the origin of IE languages. These are that the original “Proto-Indo-Europeans” (or PIEs)

1)    lived in the Ukrainian steppe around 4000BC and spread their language through conquest using wheeled vehicles and horses. This model is currently championed by Jim Mallory in Belfast and David Anthony in New York.

2)   lived in Anatolia (modern Turkey) around 7000BC and spread their language through the spread of farming. This model is championed by Colin Renfrew in Cambridge and, increasingly, by Russell Gray in Auckland.

There are, of course, other stories, but even these depend on whether the PIEs had wheels or not.

The “wheel” related word list

Indo-European "wheel-related" words, together with the distribution of IE language branches. Based on David Anthony's "The Horse, the Wheel & Language".

Indo-European "wheel-related" words, together with the current or most recent distribution of IE language branches (extinct language branches are in italics). Based on the delightful Fig. 4.2 of Anthony (2007), but with additions and changes discussed in the text.

What’s important to the opposing groups is a simple question: Did the PIEs have words for “wheel” or “wagon” or bits of wagons? If they did then the PIEs can be no older than about 4300BC. If they didn’t then they can.

Most linguists in fact argue that the PIEs did have words for wheel. The candidates put forward nine reconstructed PIE word forms, selected by the them as evidence for the PIEs having wheels. These are:

  • *hurki , argued to mean “wheel”
  • *roteh2,  argued to mean “wheel”
  • *kwekwlo-, argued to mean “wheel”
  • *kwelh1-, argued to mean “turn” perhaps in the sense of a turning wheel.
  • *h2eks-, argued to mean “axle”
  • *h2ih3s-, argued to mean “thill” or “wagon shaft”
  • *wéĝh-, argued to mean “convey in a vehicle”
  • *h3nebh-, argued to mean “nave” or “wheel hub”
  • *iugó-, argued to mean “yoke”

The aim of this post is to cast a critical eye over the linguistics of each of these forms using evidence available on the internet. I am not a linguist. Feel free to shoot me down in flames. I also have no vested interest in either story. However, I do feel frustrated when another wesbite trots out the same old cant.

The end of the PIE

One example of an Indo-European family tree, based on Atkinson & Gray (2006)

I should mention the IE branches which are important to this discussion.

Currently it’s generally accepted by all sides that the Anatolian language branch was the earliest to become separated from the rest of the PIE group. This branch, which included Hittite, is now extinct but it was once present across much of modern day Turkey.

Because the Anatolian languages are quite different from other IE languages they are sometimes excluded from PIE. Instead an earlier grouping is defined, called Proto-Indo-Hittite. For the purposes of this post I will include all branches in PIE.

Although there is some discussion, the next language branch generally thought to have split from the main IE group is the Tocharian branch. This is also extinct, but was present in the Taklamakan desert, north of the Himalayas.

After that no-one can really agree which language branch or branches were the next to separate off. Candidates include Celtic, Germanic, Greek and Armenian.

*kwelh1-, “to turn”

This root is reconstructed regularly from Latin colus, meaning “spun thread”, Old Indic cárati, Old Irish cul meaning “vehicle”, Old Norse hvel, meaning “wheel”, Old Prussian kelan, meaning “mill wheel”, Ukrainian коло (kolo), meaning “circle” (corrected 19/6/11), Bulgarian кола (cola), meaning “cart” and Greek πολος (polos), meaning “axis”, and Albanian sjell, meaning “to turn” (Armenian sjel, supposedly meaning “turn around”, may or may not exist). The form seems to have resulted in a variety of different words although many, but not all of them relate to wheels and transport.

These words can be divided into two categories. Those such as hvel, kelan, cárati and sjell are all regularly derived from the PIE root *kwelh1-. Those such as colus, cul, коло and polos are regularly derived seemingly from the PIE with ablaut *kwolh1-. Whilst this is a regular phenomenon it does, arguably, suggest subtly separate origins for “wheel” between, say, Old Norse and Russian.

Interestingly, Luwian and Hittite include the word kaluti, which seems to mean “a turn” or “a circle”. However, linguists seem to think it is not derived from the same root (presumably because it would have become something like kual-). Tocharian A lutk & B klutk, meaning “to turn”, may be also from *kwelh1– but are irregular if so.

*roteh2-, “wheel” (n)

This is reconstructed regularly from Latin rota, Irish rath, Welsh rhōd, Lithuanian rãtas, Latvian rats and German rad (amongst others), all of them meaning “wheel” in the sense of a wagon, as well as Albanian rreth, meaning “circle”, and Old Indic rátha, meaning “chariot”. The etymology is pretty clear and unambiguous and it would be difficult to put forward an argument that the speakers of the root language of all of these did not have wheeled vehicles of some kind.

It seems implausible that Tocharian A ratäk & B retke, meaning “army”, is derived from *roteh2 as tentatively suggested by Douglas Q Adams (PIE *roteh2– should become Tocharian *racä- as far as I can tell). Generally, it’s thought to be a Proto-Iranian loanword from *rataka, thought to mean “order” or “series”, although this itself has problems.

*ak’s– or *h2eks-, “axle” (n)

This is reconstructed regularly from Old Indian ákṣa-, Latin axis, Irish ais, Welsh echel, German achse, Lithuanian ašìs, Russian ось (osi) and Old Greek άξονας (axonas) (among other languages). All mean either “wheel axle” or “axis on which something turns”. Therefore some relationship with a wheel seems reasonable.

*h2ih3s-, “thill” or “shaft” (n)

This is reconstructed from Old Indic īṣá, Hittite hissa and Russian vojë (among other languages) meaning “shaft”, as well as Old Greek óiαξ (óiaks), meaning “tiller”, and English oar. I admit to finding this PIE stem’s derivation tricky to follow. Regardless, its necessary connection with a wheeled wagon seems tenuous.

*wéĝh-, “convey in a vehicle” (v)

This root is reconstructed regularly from German weg, English weigh, Latin vehō, Bulgarian веза (vesа) and Lithuanian vèžti . Meanings are generally on the lines of “carry” or “convey”, sometimes by means of a wagon. In Old Indic its cognate occurs in váhati, meaning “transport”. It’s also recognised in Tocharian A wkäṃ and B yakne, meaning “habit” or “manner”.

Words meaning “wagon” have been derived from many of these words. However, there is nothing to indicate that the original PIE root particularly signified carrying something in any form of wheeled transport. In the case of Tocharian, the link with wheeled transport, or in fact any form of transport, is non-existent.

*h1wŗgis, “wheel” or “having circular form” (n)

Based on the PIE root *h1werg-, thought to mean “to turn around”. This is argued to have descendants in both Hittite hurki, and Tocharian A yerkwanto and B wärkänt, all of which mean “wheel”.

However, to quote David Anthony, “Tocharian specialist Don Ringe sees serious difficulties in deriving either Tocharian term from the same root that yielded Anatolian hurki-, suggesting that the Tocharian and Anatolian terms were unrelated and therefore do not require a Proto-Indo-European root.”

*h3nebh-, “nave” or “hub” (n)

This is reconstructed from Old Indic nábhya, meaning “wheel hub”, and many other languages including Greek omphalós, Latin umbilīcus, Old Irish imbliu, English navel, where it always means “navel”. Although it has come to mean “hub” in Old Indic, trying to assign any other meaning to the root but “navel” seems unreasonable.

*iugó-, “yoke” (v)

This is reconstructed from Old Indic yoga, Greek zdügo, Latin iugum, Welsh iau, English yoke, Russian иго (igo), all meaning “a yoke” or “yoking”. Other words thought to be related are Tocharian A & B yuk, meaning “conquer” and Hittite juga-, of unknown meaning, but possibly “yoke”. This word appears to have meant “yoke” as in to tie animals to something such as a plough or wagon.

*kwékwlo-, “wheel” (n)

This was originally reconstructed from Old Indic cakrá, meaning “circle” or “wheel”, Avestan caxrem, meaning “wheel” and Old English hweogol, hweowul or hwēol, meaning “wheel” or “circular band”.

Words also thought to derive from this root are Greek kuklos meaning “circle” or “wheel”, Tocharian A kukäl and B kokale, meaning “wagon”, and Lithuanian kãklas, meaning “neck”. Hittite kugullas could also derive easily from *kwékwlo-. However, its meaning, something along the lines of “lump” or “measure” or even “bread roll”, although unclear, is unlikely to be related to wheels.

The reconstructed form *kwékwlo- is almost ideal PIE. It appears to derive from the PIE root *kwélh1– (seen above) meaning something along the lines of “to turn”, by a process known as reduplication.

Reduplication in IE is where a verb is expanded, often to give a past tense. This might have produced something like *kwékwelh1– or *kwékwolh1-, meaning “to have turned” (excuse my poor PIE). From this would come a noun, something like *kwékwlos, meaning “thing that turned”, hence “wheel”.

The chance of reduplicating a verb then turning it into a noun is reasonably high and may have been done many times in history. However, if it had been done by one of the descendants of IE it should have produced a distinct form, recognisable to an IE linguist.

Indeed, in the opinion of one such linguist, Andrew Garrett,  “… such an account is hardly possible for PNIE *kwekwlos ‘wheel’ (in Germanic, Greek, Indo-Iranian, Tocharian…): though derived from *kwélh1– ‘turn’, a reduplicated C1e-C1C2-o- noun is so unusual morphologically that parallel independent formation is excluded.”

This makes *kwékwlo- the flagship word for the wheelies. It would suggest that any IE language that includes a word derived by regular rules of IE sound change from *kwékwlo- must have been part of a larger grouping somewhere in the second half of the fifth millennium BC or later. Following this argument, as both Tocharian and Greek have words supposedly derived from *kwékwlo- they must have split from the main IE group after wheels were invented.

However, I seem to see weak points with this argument. Therefore it’s perhaps worth looking at the individual derived words in more detail.

Old Indic cakra, meaning “wheel’

This, word, closely related to Avestan čaxra, meaning “circle” or “wheel”, can be derived by regular means from a PIE form *kwékwlo-, via Proto-Indo-Iranian *kekro-.

It is notably comparable to some Finno-Ugric words, such as Finnish kekri, which have meanings varying from perhaps “yearly cycle” to “circular”. Although this raises the possibility that the Indo-Iranian word is derived from a Finno-Ugric language this is unlikely. Few, perhaps no, words have gone into Indo-Iranian from Finno-Ugric.

English wheel, meaning “wheel”

This is thought to come from Old English hweogul, hweowul or hwēol. PIE *kwekwlo- would develop into a form like *hwigwl. With the softening of gw to give *hwiwl this seems close enough to be reasonable.

Greek kuklos, meaning “circle”

χύχlος (kuklos, with a plural kukla, meaning “wheels” – NB the ancient Greek for “wheel” is τροχός (trochos)) is suggested to derive from a Proto-Greek form *kwukwlos, although without explanation for the change of vowel from PIE *kwékwlo-. No standard IE linguistic rule can do this.

Sihler (1995) has suggested that “strongly labial environments” may change a PIE “e” to PIE “o” (a result of ablaut), which would naturally change to “u” in Proto-Greek. One of just two good examples of this is, of course, *kwékwlo- (the other being PIE *gwenH2 becoming Greek γυνή (guní)). So his preferred PIE form is *kwokwlo-.

Postulating the Proto-Greek form *kwukwlos seems to be an attempt by linguists to explain why the recorded ancient Greek form was not *téklos or *péklos. For comparison, ancient Greek téssares or péttares (depending on dialect), both meaning “four”, are thought to be from PIE *kwetwares. This makes the derivation of kuklos from *kwekwlos strange, although not perhaps entirely impossible.

Lithuanian kãklas, meaning “neck”

This word can be from PIE *kwekwlo- via Balto-Slavic *k:ikla-. However, it would be preferable to derive kãklas from *kwokwlo-, as with Greek. If the form originally meant “wheel” it has changed this meaning considerably. This could possibly be achieved with a corruption of meaning to do with, say, wheeling your head around, although this seems a stretch.

Tocharian kukäl (A) & kokale (B), meaning “chariot” or “wagon”

(NB The Tocharian for “wheel” is, as stated above, A wärkänt, B yerkwanta)

Kukäl and kokale are thought to derive from a Proto-Tocharian form, perhaps *kukäle. As with Greek there are problems deriving this form from PIE *kwekwlo-. The form necessary for giving the correct Tocharian should be, as Tocharian expert Douglas Adams noted, “closely related to, but phonologically distinct from … *kwekwlo- .” He suggests PIE *kwukwlo-.

Alternatively, Don Ringe (2009) suggests the following, slightly tortuous derivation process (references removed):

*kwékwlos > *kwékwlë > *kwyékwlë > *kwyә́kwlë → Proto-Tocharian *kwә́kwlë ‘chariot, wagon’ (with adjustment of palatalization in a reduplicated form; or is this just straightforward assimilation?); > *kŭkl ~ *kŭkla- > *kukäl ~ kukla- → Tocharian A kukäl ~ kukla-; > *kwәkwә́lë > Tocharian B kokale.)” (added 29/07/11)

This is because PIE *kwekwlo- would give a Tocharian form of *käkla- or, following the sound change laws to their extremes, *käśla- or *śäśla- or *śla- (c.f. PIE *kwetares becomes Proto-Tocharian *ś(ä)twer, becoming Tocharian A śtwar and B śtwer).

Looking at it another way, *kukäle could be derived from PIE *kwukwelo-, *gwugwelo-, *kukelo- or some other similar forms, but not obviously from *kwekwlo-.

On the other hand, the –käl– element of both kukäl and kokale could simply be derived from PIE *kwelh1 (the form thought to mean “turn” mentioned earlier) without any reduplication. There are also other PIE roots which could give rise to the same element. For example PIE *kwele-, meaning to “move around” or “drive”, is thought to have produced the Tocharian stem käl-, to “lead” or “bring”. This would satisfy the Tocharian form in kukäl and kokale.

Speculating wildly and unsensibly, the first element of kukäl and kokale could, conceivably, be related to the Tocharian word for “cow”, which is ko in Tocharian  A and keǔ in Toch B (probably from another PIE root *gwow- through Proto-Tocharian *kew). Using this logic the words for “lead” and “cow” could be brought together in the Proto-Tocharian word *kukäle, meaning “cow bring thing”…

Well, perhaps not. Either way, while the latter is probably an unlikely derivation of kukäl and kokale it seems just as plausible, or implausible, as their derivation from *kwekwlo-.

The significance of *kwekwlo-

Looking at the words supposedly derived from *kwekwlo- it seems that all apart from cakrá and caxrem and hwēol require some manipulation of the PIE form. There is at least the need for a second, possibly ablauted, form, *kwokwlo- and possibly for a third (say *kwukwelo-).

At the moment I can’t help having doubts about whether such a word as *kwekwlo-, meaning “wheel”, ever existed in the original PIE vocabulary. It’s descendants, unlike those of *roteh2, are rather scarce among the different branches. More preferable is that different reduplicated forms occurred, with different meanings perhaps related to turning, at different times. So, to contradict Andrew Garrett, parallel independent formation from derivatives of *kwelh1 or *kwolh1 seems just as easy as anything.

Discussion – who did have wheels?

Certain forms, such as *roteh2 and *h 2eks-, appear to mean “wheel” and “axle” respectively. They seem to reflect a genuine, wheel-related origin and one or both appear in all the language families, excluding Tocharian and Anatolian. This suggests that perhaps wheels appeared after the separation of Anatolian and Tocharian. Greek and Armenian do not include a form related to *roteh2. More debatable is whether wheels appeared after the separation of Greek and Armenian.

*kwelh1– seems to be related to wheels in some, but not all languages. The languages where that relationship is clear do not include Hittite or Tocharian but may include Greek. This is similar to the pattern discussed for *roteh2 and *h2eks-.

Some words, such as *h2ih3s-,*wéĝh-,*iugó- and *h3nebh-, have meanings which do not necessarily require the use of wheels or wagons. Therefore their distribution in the language groups is not enough evidence for the use of wheels by PIEs. *hurki can be also excluded from the discussion as this form may well not have existed.

The case for *kwekwlo- seems more difficult. Regardless of whether there was such a PIE word, the argument that a derivative is clearly present in Tocharian is weak and even its presence as a derivative in Greek is not entirely convincing.

So overall I would say that a good case can be made for the presence of wheel-related vocabulary in much of the IE family, including the Italic, Celtic, Slavic, Indo-Aryan, ?Albanian and Germanic branches, which suggests that these branches had not separated by the time that wheels were invented.

The case for Greek is more marginal but certainly indicates perhaps some close connection of Proto-Greek with the other language families during the time that the wheel was invented. However, the case seems weak for the Tocharian and Anatolian branches separating from the other language branches after wheels were invented.

(I offer no opinion about Armenian. This language, like Albanian, has suffered from much loss of original vocabulary due to the influence of Iranian. However, for all IE language branches it highlights a problem – the lack of evidence for wheel-related words derived from PIE doesn’t prove that the ancestor language didn’t once have them. Nor can archaeology prove that ancient peoples didn’t have wheels.)

If what I’ve observed is true then the spread of IE branches from the Ukraine (or from elsewhere) after the invention of wheels is still possible. However, this would need to be without Anatolian and Tocharian (which has implications for the Afanasevo culture). It also allows for the possibility that the initial spread of farming into Europe from Anatolia could have included PIE speakers.

All this, of course, is only an opinion by someone who’s not a linguist and based on the limited information I could access. There are other lines of evidence, such as the IE word for wool (*h2/3wlh1-), also used to argue a late date for PIE.

However, from the information I’ve seen it seems that some archaeologists are happy to find the most speculative of evidence as support for the existence of wheel-related words derived from PIE, based on highly speculative discussions by linguists.

On the other hand linguists themselves seem a little complacent about selling their ideas to us, assuming that the illingual should trust them without question. This is not healthy science. I think they should sell their ideas better and explain to us why we should believe them.


Adams, D.Q. 1999 A Dictionary of Tocharian B, Rodopi, pp864 (see also Indo-European etymological dictionary (dead link?))

Anthony, D.W. 2007 The Horse, the Wheel and Language. Princeton, pp553.

Atkinson, Q.D. & Gray, R.D. 2006 How old is the Indo-European language family? Illumination or more moths to the flame?, In J. Clackson, P. Forster and C. Renfrew (eds) ‘Phylogenetic Methods and the Prehistory of Languages’, MacDonald Institute: Cambridge, 91-109.

Brugmann, K. 1891 (translation) A Comparative Grammar of the Indo-Germanic Languages, Vol. II Morphology, Westermann, p96.

Erkut, S. 2006 The Hittite Word kugulla-, Tarih Araştırmaları Dergisi 25, p107-111.

Fortson, B.W. 2010 Indo-European Language and Culture: An Introduction, Wiley-Blackwell, pp568.

Garrett, A. 2006 Convergence in the formation of Indo-European subgroups: Phylogeny and chronology.  In Phylogenetic methods and the prehistory of languages, Forster, P. & Renfrew, C. (eds.) Cambridge, p139-151.

Parpola, A. 2005 The Nasatyas, the Chariot and Proto–Aryan Religion. Journal of Indological Studies, 16-17, p1-63.

Parpola, A. 2007 Proto-Indo-European speakers of the Late Tripolye culture as the inventors of wheeled vehicles: Linguistic and archaeological considerations (unpublished?)

Quiles, C. & Lopez-Mechero, F. 2009 A Grammar of Modern Indo-European, Second Edition: Language and Culture

Wikipedia – Indo-European Sound Laws (online resource)

Indo-European etymological dictionary, Leiden (online resource) (dead link?)

Starostin, S. Tower of Babel (online resource)

Early Indo-European Online, University of Texas

Supplementary References

Murphy, M. 2009 Olszanica 6: A Sledge Shed? Theoretical Structural Archaeology post

Ringe, D. 2009 Inheritance vs. lexical borrowing: some Indo-European cases, (online, accessed 28/6/11)

(An important reference detailing the suggested derivation of words from  *kwekwlo-)


Figure 4.2 of David Anthony’s book indicates that the Tocharian languages include words derived from *h2ih3s– and *ak’s-. However, he does not provide evidence for what these words are and I can’t find evidence of them recorded anywhere else. Furthermore, an earlier version of the diagram from 1995 does not indicate the existence of such words so I’m not sure what their status is. It’s possible that, like *roteh2, these are not generally accepted derivations.

{ 21 comments… read them below or add one }

Leave a Comment