A potential etymology for ἵππος

The horseposting will continue until morale improves

Timeo Danao et geminata ferentes

Meditating on the mysteries of the Proto-Indo-European *h₁éḱwos as a result of this abstruse technical monstrosity recently, it occurred to me that we still don’t have a good explanation for Greek ἵππος.

Today, I will attempt to provide one. You’re welcome.

ἵππος in brief

This reflex, híppos, is a regular development in Attic and some dialects of Ionic from an earlier Mycenaean word written 𐀂𐀦, i-qo, which preserves a labiovelar. The geminate -pp- is generally taken presuppose a geminate in Mycenaean and Proto-Greek, and thus the Proto-Greek word is reconstructed as *híkkʷos, with an evolution something like this:

1. *híkkʷos

2. *híkpos (labiovelars > labials in most positions in Attic and Western Ionic)

3. *híppos

(I presume that the geminate -pp- was considered as 2 separate sounds and one half assimilated to the other, because there’s no reason to assume that at this stage of the language it was considered a “lengthening” of kʷ.)

In addition, we also have some evidence of forms where a velar was retained: Eastern Ionic has ἴκκος ikkos (attested in the Byzantine Etymologicum Magnum, and a regular development in East Ionic by which kʷ is preserved as k before o).

The East Ionic reflex lacks initial h, but onomastic evidence suggests that this was a regular loss and h was preserved in other dialects: ῾Ικκότας and ῾Ικκότιμος are each attested once and clearly contain a morph of ῾Ικκο- , hikko- with initial h intact. These names are associated with Macedonians in the 4th and 3rd centuries BC, and we can presume that this form therefore reflects a Northwest Doric version of this name.

As an interesting comparison, the personal name Ἴκκος is famously associated with an Olympic victor from Taranto, a Spartan colony of Magna Grecia, indicating that “Doric proper” also retained this form in some capacity, but had lost its initial h, at least by the lifetime of Ἴκκος in the mid-5th century BC.

This is unacceptable

*híkkʷos looks superficially similar enough to *h₁éḱwos that a relation of some kind is beyond doubt. However, assuming a Proto-Greek *híkkʷos, which all available evidence indicates we must, is a problem, because there is no clear or satisfying explanation for:

1. The initial h

2. The vowel i

3. The gemination of kʷ

These discrepancies have been a bugbear in Greek for some time. In all other Indo-European families (except Anatolian), we find regular reflexes of a thematic noun *h₁éḱwos. In Greek, this should produce **épos or **ékos by the following pathway:

1. *h₁éḱwos

2. *h₁ékʷos (velar+consonantal w merge as a labiovelar in Proto-Greek)

3. *ékʷos (h₁ is lost without trace before e)

4. *ékos/*épos (kʷ > k before o in some dialects, and p in others)

Several explanations have been devised to account for each of the three discrepancies. Unfortunately, none of them are wholly satisfactory.

The change from e to i has been explained as a conditional raising in the neighbourhood of a labial in Mycenaean, but this doesn’t account for Aeolic and Doric reflexes which also show the same change, and the change is only marginally attested in a handful of words, all of which also have variants with e, which ἵππος never does.

The addition of h has been described as via analogy, although opinions differ as to the question “with what?”. An analogical origin would also be hard to explain as dialectical evidence suggest h was lost over time rather than added. Geminate -kkʷ- has been explained as compensatory lengthening, although it is again uncertain for what we are compensating.

How do we explain this?

Having dug myself into the weeds on the Anatolian question, I noted that the athematic Hittite, Luwian, and Lycian nouns ekkus, azzus, esb – which must be traceable to an original PIE athematic u-stem *h₁éḱus – are described by Kloekhorst as being relics of an originally hysterokinetic paradigm (or hysterodynamic after Beekes and Co). This means, for the uninitiated, that there would be an alteration in accent (and thus ablaut) between “strong” and “weak” stems of the noun.

In other words, nominative *h₁éḱus would have a genitive form *h₁ḱuós. This is backed up by the presence of Anatolian genitives ending in -aš for these nouns, and presents an interesting possibility as to how or why many athematic roots in PIE are later thematicised: genitives from an older, obsolete paradigm are reinterpreted as nominatives.

Great minds

Of course, I’m not the first person to think of this, as I discovered when I set out to see if we might use this as the basis for a new explanation of ἵππος. Michiel de Vaan wrote an article on exactly this subject far back in prehistory (2009).

He also advances the idea that Proto- or Pre-Proto-Greek’s reinterpretation of the original genitive *h₁ḱuós might also be paralleled with a reinterpretation of the bare-locative *h₁ḱēu “on horseback” as an agentive noun, giving rise to the associative/occupational suffix -εύς in historically attested Greek.

I won’t go into his explanation of ἱππεύς here, except to say that I find it much more convincing than his explanation of ἵππος.

Backing the wrong horse

De Vaan’s hypothesis, in brief, is that the re-analysed nominative *h₁ḱuós in Proto-Greek was augmented in accordance to a pattern observed by Mayrhofer (1987) and Vine (1999) in which an expected sequence *(s)CCRV, *(s)CRV appears as *(s)CiCRV, *(s)CiRV, in Greek. Vine describes i in these sequences as a “non-phonemic prop vowel”.

Observant readers will note that this explanation, while ingenious, still isn’t without its problems.

It cannot account for initial h. De Vaan recruits one particular example of the “by analogy” line but does not consider the issue any further except in his note that the h appears to be absent in the name Λεύκιππος and its variants (although we find in 5th century Laconia an Αἱρήἱππος and in Oytilos in 196BC a Κρατήἱππος, the forms of which appears to preserve h through the compound).

Indeed, the proposed pathway of development HCC > ʔiCC > iCC presumes the loss of the initial consonant, forcing us to assume an innovation where there is no need to do so.

Further, this model doesn’t explain the gemination of *-kkʷ-, and makes no real attempt to do so. The only mention of it at all in the article in question is in relation to its derivative, ἱππεύς, which is stated to have acquired the geminate -pp- by analogy to ἵππος.

I think these weaknesses can be ascribed to attempting to date all of *híkkʷos’s oddities to the Pre-Proto-Greek period. Even though the author notes that in his schema:

1. The thematicisation of *h₁ḱwos could easily see e restored either by analogy to most other PIE nouns (and especially to the original strong nominal stem *h₁éḱus)

2. The Greek addition of i would ipso facto pre-date such a restoration among all other branches

It may be that because he began his line of thinking in the frame of the Indo-Anatolian hypothesis, he did not puzzle this out fully. Happily or unhappily for him, I am here to do the heavy lifting.

Towards a unified horse theory

I believe we can, with small modifications, take de Vaan’s hypothesis and create a complete explanation for the origin of ἵππος as an entirely and uniquely Greek development, separating it from any “post-Anatolian” innovation.

We’ll start with the supposition that Proto-Greek *híkkʷos is a direct continuation of the weak-stem genitive *h₁ḱuós. We can presume that this change happened in all non-Anatolian branches by trivial analogy to the old strong stem, but I would contend our unique Greek reflex demands us to assume that Greek’s circumstances were unique.

I believe this begins with the early merger of the sequence -*kw- as *kʷ- in Proto-Greek. In other branches this change is either absent or occurs at a far later date.

This shift would have left a strong stem *h₁ékus alongside a weak stem genitive *h₁kʷós in a highly unstable and irregular arrangement that requires some kind of regularisation.

h₁okʷos pokʷos

I believe that we can explain all of *híkkʷos’s irregular developments by presuming that the Proto-Greeks attempted to resolve the discrepancy in the *h₁ékus~*h₁kʷós paradigm with reference to the word’s closest linguistic relative, the parent of the adjective ὠκύς, “swift”.

In comparing *h₁ékus~*h₁kʷós to the reduplicated preform *h₁oh₁ḱús, it is possible that the early Greeks reanalysed the original unreduplicated stem **h₁eku- as an irregular development, and hypercorrected the genitive*h₁kʷós to *h₁h₁kʷós as a presumed zero-grade of a root *h₁eh₁kʷ-o-.

We can then explain the insertion of i between *h₁h₁ in light of Mayrhofer and Vine’s rule, in which a sequence CCRV resolves to CiCRV (with labiovelar kʷ acting as the “glide” or resonant in this case).

This would leave us with another strange form: *h₁ih₁kʷós, or perhaps *h₁íh₁kʷos, with accent retraction by analogy. Ordinarily, a sequence of *Cih₁C would yield:

1. *Cih₁C-

2. *Cyeh₁C- (palatalisation of first consonant in proximity to y, vowel relaxed to e)

3. *CyēC- (loss of laryngeal + compensatory lengthening of adjoining vowel)

Without vowel insertion, it may also lead directly from *Cih₁C > *CīC.

We do not observe this pattern of resolution in *híkkʷos, but I believe this is because there are two other processes at work.

1. In historically attested Greek, a combination of a laryngeal + y can resolve as h. This process is not frequent enough to merit a name, but we might call it “laryngeal palatalisation”.

2. A hypercorrect pronunciation of *h₁íh₁kʷos could easily prompt a change h₁kʷ > kkʷ (possibly via ʔkʷ) if h₁ especially in an environment where h₁ was already on the way to being lost as a distinct phoneme.

In this scenario, we can also see how the normal process of palatalisation in a *CiC- sequence was interrupted, leading to perhaps an allophonic form *h₁yíʔkʷos, which was resolved as *híkkʷos after the final loss of the laryngeals.

We can explain the regular development of *h₁oh₁ḱús, > ὠκύς (in which *CoHC- yields *CōC- as expected) because it is a phonetically quite normal word for late-PIE/dialectical Proto-Greek, and would also have been quite a regularly formed word in the protolanguage.

By contrast, the sequence *h₁íh₁- appears would be highly irregular, occurring perhaps only in one or two other contexts in PIE, usually in optatives, and never word initially. Its sheer unusualness might therefore explain why this word has such a singular resolution in historical Greek.

That phonetic evolution in full

We can now attempt to reconstruct the diachronic changes that lead from a Proto-Indo-European *h₁éḱus to late-Proto-Greek *híkkʷos.

1. *h₁éḱus~*h₁ḱuós (“hysterodynamic” athematic u-stem with weak stem genitive)

2. *h₁ékus~*h₁kuós (“centum”-merger of ḱ and k)

3. *h₁ékus~*h₁kʷós (early Proto-Greek merges velar k with consonantal w as labiovelar kʷ, creating an unstable nominal paradigm)

4. *h₁kʷós is re-analysed as a thematic nominative with *h₁ékus seen as an irregular development by reference to *h₁oh₁kús

5. *h₁h₁kʷós (hypercorrection of *h₁kʷós)

6. *h₁ih₁kʷós (conditional anaptyxis)

7. *h₁íh₁kʷos (accent retraction to first syllable, as part of hypercorrection?)

8. *h₁yíʔkʷos (palatalisation of h₁ to h₁y before i allophonic retraction/assimilation of h₁ to kʷ)

9. *híkkʷos (palatalised h₁y > h, ʔkʷ fully assimilates to kkʷ)

Well we got there in the end

I realise that we’ve got to do quite a bit of special pleading to get to this outcome. That said, this is perhaps one of those instances where we can’t rely on Occam’s Razor to do the job for us. This is a relatively simple and extremely well-worn word from perhaps the best-studied ancient language in the world, which has defied all attempts to etymologise it for decades. Perhaps we shouldn’t even be looking for a simple solution.

It’s worth noting as a point for further discussion that this development would have had to have taken place separately from those which yield other thematic reflexes of *h₁éḱus. With some other marginal changes, this points to the idea that if Anatolian was the earliest family to leave PIE, Greek must have been hard upon its heels.

As usual, I mean no harm to any of the people whose work I make reference to here. It would not be possible for me to indulge myself with these little flights of fancy if not for their work, and they are free to accept, modify, or reject it as they find it.