Horse sense
Semitic, Hurrian, and Sumerian words for "horse" are probably derived from Luwian
When we first meet horses in the historical record, it’s in Akkadian, Hurrian, and Sumerian texts from late 3rd millennium BC Mesopotamia. All three languages refer to them with terms that seem self-evidently derived from the same source.
In Akkadian, the best understood of the trio, the common word for horse is sīssûm, with Old Assyrian variant sisium. This most probably reflects a pronunciation t͡sit͡sium.
To the north, the Hurrian word for horse was variously written as eš-še-, iš-ši-y(v)-, and eš-še-e. Our understanding of Hurrian isn’t as complete as it might be, but these variants all appear to point to an underlying lemma *essy- or *essiy- (or possibly *essə[y]).
To the south, the Sumerians use anšesi2-si2. This most likely reflects * t͡sit͡si, in line with Akkadian.
If you’re wondering where all the confusion around s, š, t͡s is coming from, it’s a vagary of the writing system: the series of signs traditionally transcribed ša, še, etc in cuneiform probably reflect sibilants of some kind, while those transcribed sa, se etc probably reflect alveolar affricates.
These three terms are obviously close etymological relatives of each other, as are other terms for horse in neighbouring Semitic languages (Ugaritic 𐎒𐎒𐎆 ssw, Hebrew sûs, Classical Syriac ܣܘܣܝܐ sūsyā, and dialectical Arabic sīsiyy “pony”, and even Egyptian ssm), but there’s been considerable debate over the years as to their ultimate origin.
The usual suspects
The first port of call has traditionally been an Indo-European language of the satem subgroup. Questions as to the validity of satem languages as a subgroup aside, this seems like a neat fix. Whatever sound we presume our s-phones above to ultimately represent, we can easily trace them to a hypothetical early satem language loanword. Perhaps something like *ečwos.
This is especially tempting given the local presence of Mitanni, whose Hurrian dialect absolutely does contain a very similar looking word aš-šu-wa-, probably reflecting an early instance of Western Iranian asva- (compare Old Persian 𐎠𐎿 asa or 𐎠𐎿𐎱 asp[a]).
However, something doesn’t add up, and that something is time. Although we can make a case that Indo-Iranian speakers (with emphasis on the Iranian) were present in Mesopotamia as early as around 1800 BC, perhaps slightly earlier, a late 3rd Millennium date would put us in the vicinity of Proto-Indo-Iranian proper.
Furthermore, we have non-linguistic evidence of horses in Hurrian territory (datable to around 2450BC per Hauser) from an even earlier period which would push the date of a possible loan back even further – before even Proto-Indo-Iranian was so much as a twinkle in *Dyeus Ph2ter’s eye.
Fortunately we don’t have to look very far for an alternative, because there is a palatalising Indo-European dialect that we can plausibly place in the general vicinity of the Hurrians (and various Semitic groups) in the mid-3rd Millennium.
Enter the Luwians
The Luwic sub-branch of the Anatolian family (including Luwian, Lycian, Carian, Milyan, Sidetic, Pisidian, and possibly, more distantly, Lydian) is an interesting case: it appears to be neither centum nor satem in its affiliation, preserving some evidence of a distinct palato-velar series of stop consonants under specific conditions.
One such condition arises in the Luwian inherited reflex of *h₁éḱus:
Cuneiform 𒊍𒍪𒉿𒀸 az-zu-wa-as
Hieroglyphic a2-su3 (nom. /asus/ or /ʔasus/)
These renderings presumably reflect a pronunciation *at͡s:us via the following path from PIE:
*h₁éḱus (archaic athematic of *h₁éḱwos, preserved only in Anatolian)
*ʔéḱ:us (P. Anat. *h₁ > *ʔ, voiceless stop becomes fortis in this position)
*(ʔ)éč:us (P. Luwic ḱ: palatalises to *č: before *u)
*(ʔ)áč:us (merger of e and a in Proto-Luwian)
*(ʔ)át͡s:us (fronting of č: to t͡s:)
It’s worth noting that the precise phonological value of the sound transliterated as zz or s in the Luwian material is unknown, but can be analysed from an etymological standpoint (per Hawkins, Kloekhorst) as a parallel development of the sound that yields Hittite -kk- in ekkus.
A borrowing around stage 3 here (in which č can be taken to represent an alveolar or palatal stop/affricate consonant of uncertain value) takes us a long way to explaining our Hurrian lemma *essiy-, which can be taken to reflect our proposed *éč:us at a few centuries remove (alongside a regular process of vowel harmonising in which an early Luwic *u is raised to *i.)
I would hazard a guess that this loan diffused quickly through proximal Semitic languages. The reduplication of the affricate/sibilant/palatal/whatever it is in the initial position might be explained as an attempt by Semitic speakers to “familiarise” an alien sounding Hurrian borrowing by analogy to their common triradical pattern, taking *eč:u- or *et͡s:u- or *et͡s:iy- or whatever form they originally encountered and reinterpreting it with reference to a new Semitic root S-S-W or S-S-Y.
Different chains of borrowing, borrowing from separate Hurrian dialects, or borrowings at different times can be recruited to explain the alternation between y and w seen in the semivowel here, although it’s beyond my patience or purpose here to explain this fully. The Akkadian evidence, though inconclusive, points to an original variant in i/y (w and y regularly simplify to ʔ and then disappear in Akkadian, although note the Old Assyrian reflex), as does the Syriac, while Amorite/Ugaritic and Hebrew both point to a variant in w.
And?
I like little forensic excursions like this as we can use them (sometimes) to put tentative boundaries on other historical events. In this case, if the case is proven, we can place early forms of Luwian (or “Luwic”) more broadly in proximity to the north of Syria and to Hurrian territory circa 2500 BC or slightly earlier. This isn’t surprising, as we know Luwians were roughly in this area in the 2nd Millennium BC as well. We also know that the Hurrians were a dominant party in horse trading and culture – for a number of reasons – from a very early date, and that horses were most likely introduced to the area by Indo-Europeans of one kind or another, and this is a more plausible scenario entirely than assuming time-travelling Indo-Iranians.
And based on this, we can also anchor ourselves with regard to multiple other assumptions about the dates of earlier stages of the protolanguage and where it may have been spoken.