In Greek, the PIE verbal roots *dheh1 'put' or 'do', *Hieh1 'throw', and *deh3 'give' show up with an unexpected -k- in some aorist forms: ἔθηκα, ἧκα, ἔδωκα. In Latin, the reflexes of the first two roots have -k- throughout: facio, iacio (although the present forms are new analogical formations, so it's plausible that this -k- too was once confined to the perfect/aorist). LIV adds some data from other Italic languages (and tentatively from Phrygian) but says the Greek and Italic -k-'s are unrelated (citing two references which I can't access at the moment), which seems odd. What are the theories as to the origin of this -k- or these -k-'s?

Also, are there other IE languages in which these verbs may have once had -k- in the aorist but we can't tell because the resulting forms would be the same with or without it? (For example, if Sanskrit once had an aorist *adhākt it seems plausible that the regular reflex of this, **adhāk, would be analogically reformed as the attested adhāt.)

  • 3
    I've always wondered about this too. A connection with the perfect suffix -k-? The similarity may be coincidental, or indicating a common origin, or analogical. Note also that both the (pseudo-)sigmatic aorist and the perfect mostly have endings with -a-, and that aorist and perfect more or less fused together in Latin: perhaps they were intimately related in (late) Proto-Indo-European? And many Greek perfects don't have -k- but only -a-, so perhaps the -k- found in the Greek perfect is not that tightly connected to the perfect as such, but rather to some kind of "past" notion, I don't know.
    – Cerberus
    Commented Sep 16, 2013 at 22:09
  • 1
    As to facio, some analyse it as not analogical but owing to zero-grade ablaut (dʰh1k-yo) v. e-phase fec- (> dʰeh1k-): books.google.nl/…
    – Cerberus
    Commented Sep 16, 2013 at 22:15
  • Thanks, I'd been following Sihler who treats facio as an innovation, but the zero-grade etymology obviously works too.
    – TKR
    Commented Sep 17, 2013 at 2:07

4 Answers 4


Andreas Willi in his book Origins of the Greek Verb (2018) from the Cambridge University Press argue that the k-aorist functions synchronically in Greek as a transitive marker, but that in Proto-Indo-European the k-extended and unextended forms of the root appear to have both been used (both inflected as root aorists) with no clear distinction in meaning, with the k being generalised by analogy to the 1sg form and s-aorist.

Citing Untermann he identifies it as certainly of the same origin as the Italic k-extension in *dheh1 & *Hyeh1-, but then cites Orel 1997, Sowa 2007 & Rieken 2007 to argue that Phrygian and/or Luwian probably also show the same k-aorist.

Discussing possible origins, he cites Osthoff 1884, Hirt 1928, Markey 1980, Bammesberger 1984, Shields 2002, & Dunkel 2004 as thinking of it as a fossilised particle or deictic, Kortlandt 2001 as thinking of it as a "dative marker *ka" (quotation marks are his, and not mine), but then says a connection the k-presents seems more promising.

Discussing the origin of the k-presents, he first notes that, as with the k-aorists, they only occur in laryngeal-final roots. He then suggests that in the 1sg passive this would lead to a cluster of two laryngeals, in which a k may be inserted to avoid a geminate (assuming the laryngeal cluster would first assimilate), parallel to the way that s was inserted in clusters of dental stops to avoid a geminate. This k would then be either generalised or erased by analogy, resulting in the extended and unextended forms having no difference in semantics.

  • Why am I not surprised that Kortlandt sees an otherwise obscure dative marker in an aorist extension? Commented Feb 23, 2022 at 22:01

Untermann shows that the -k- in lat. fac- and iac- is part of the root and does not contribute "aoristic" semantics to the verbs. Nor does he find any relations in their functionality compared to the greek material. The -k- may however be of common italic origin as tentatively shown by suchs forms as ven. vhagsto (root *dheh1-k-) or osk./umbr. stakaz (root *steh2-(k?)). The presented material is hardly sufficient to determine the function or contexts of the k-extension.

Ref.: Untermann, J. Meiser, G. (Ed.) Gr. ἔθηκα = lat. feci, gr. ἧκα = lat. ieci? Indogermanica et Italica. Festschrift für Helmut Rix zum 65. Geburtstag., 1993, 461-468.

  • Thanks! Does he give any argument for why he regards the Latin -k- as unrelated to the Greek one, apart from its not contributing aoristic semantics? If -k- was taken over into the present from the perfect (i.e. if facio iacio are analogical innovations) then naturally you wouldn't expect perfective semantics in the present forms anyway.
    – TKR
    Commented Sep 23, 2013 at 16:42
  • Well, his conclusion are 'unsurmountable functional differences' between the Greek forms and the Italic forms and the fact that there can be found no specific commonalities between the verbal inflexion of Greek and Latin. Therefore he excludes them to be taken over at all.
    – user2498
    Commented Sep 24, 2013 at 7:32

The simple and surprising solution is that the k is part of the root. Kloekhorst investigated Lycian word material and concluded that the laryngeals *h2 and *h3 were uvular sounds *q: and *qṷ: in Proto-Anatolian and Early PIE as well. If the consonants really had been long (fortis), is a question of debate, but that laryngeals had developed from uvular stops is becoming more and more evident. See also Kortlandt 2018.

That means that the k-forms are an old relic of the times when uvular sounds were in use. A form like Gr. ἔϑηκα is indeed Early PIE *téQm̥ and Middle PIE *dhéh1m̥. I do not know the quality of the uvular stop corresponding to h1, so I am denoting it with a Q. And the endings for the passive root aorist in -θην are nothing else then later forms of *dhéh1m̥.

Lat. faciō, iaciō developed from *dhǝh1i̯ō and *(h)i̯ǝh1i̯ō.

The same explanation is valid for many perfect forms where Old Indic has a -u.

Gr. πέπωκε, OI papáu, PIE *(pe)póh3e

Gr. ἔγνωκε, OI jajñáu, PIE *(g̑e)g̑nóh3e

An example for an odd behaviour of Lycian is the word Lyc. qla- ['kwla] versus Hitt. Éhēla- ['χe:la], court. Lyc. q is a velar sound whereas Hitt. h is a uvular fricative. As Lyc. q cannot come from a velar *kṷ, only an uvular *qṷ is a reasonable solution to it.

Kloekhorst 2018: Anatolian evidence suggests that the Indo-European laryngeals *h2 and *h3 were uvular stops, Indo-European Linguistics 6, 69-94. Kortlandt 2018: The Indo-European k-aorist. Farnah [Fs. Lubotsky] (Ann Arbor: Beech Stave Press), 137-142.

  • 1
    Welcome to the site and thanks for this answer! I think what's needed to make this account plausible is some explanation of why the proposed uvulars went to κ/c in just those forms, and some evidence that *h1 was also uvular. Also, can you explain how the Indic forms are relevant?
    – TKR
    Commented Feb 11 at 17:49
  • 1
    In particular, how is *dʰeh₁m̥ seemingly responsible for both -θηκα in ἔθηκα and -θην in aorist passives? If the same input – indeed the exact same form of the same verb – yields different outputs, there needs to be a very strong reason for that. Commented Feb 11 at 20:30
  • the issue is we know the regular reflexes of the laryngeals. Without giving a plausible conditioning environment for the k > q shift I think this is only have an answer. I do remember a paper, I think it was from Kloekhorst suggesting such a conditioning environment though, I think it was before a tautosyllabic *t, with analogy spreading it across the rest of the paradigm, but I don't have the citation to hand
    – Tristan
    Commented Feb 12 at 10:09

That uvulars can change to velars, is simply conceivable. For the question, why some uvulars did not change to laryngeals or glottal stops, more research is needed. I also cannot explain the simultaneous vowel lengthening. It seems that the uvular is like an operator that causes an additional vowel lengthening. Kortlandt is assuming an interim state *Ɂk for it.

In regard to Hitt. hastai, bone, the question, why Bosnoserbocroatian favours kost instead of **ost, cannot be answered at the moment.

The Indic forms are not relevant. But they can give a hint why at least one of the laryngeals could have had a labial part. Some authors assume for *h1, *h2 and *h3 the glottal stops *Ɂ, *ʕ and *ʕṷ (see the discussion at Kümmel).

If you take the given glottal stops for granted, you just have to work out which uvular sounds could have preceded the glottal stops. I dare to make a guess: *q > *h1, *q: > *h2, *qṷ(:) > *h3. But this is by no means sure. And it raises more questions. Had there been perhaps more uvulars (3 “normal” uvulars (preglottalised, short and long) and 3 labio-uvulars) that merged into 3 laryngeals? I am sorry that I really cannot say more to this subject.

Kümmel 2022: On new reconstructions of PIE “laryngeals”, especially as uvular stops, Acta Linguistica Petropolitana. 2022. Vol. 18.1. P. 199–215, footnote 3.

  • 3
    (This answer should be an addition to or comment on your previous one, as it's not an independent answer to the question.) Interesting as this proposal is, it doesn't seem tenable for several reasons: (1) AFAIK there's no evidence *h₁ was ever uvular; (2) if *h₃ was labialized then the result of fortition should be *kʷ, which wouldn't yield κ; and most crucially (3) as long as no conditioning environment for the fortition can be given the theory is ad hoc and unfalsifiable.
    – TKR
    Commented Feb 11 at 20:25
  • 2
    The main problem with taking *h₁ back to **q and h₂ back to *qː is that this would mean Pre-PIE allowed geminate uvulars in syllable- and word-initial position, even next to other consonants. Things like *h₂ner- ‘man’ and th₂eu̯s- ‘be silent’ would go back to **qːner- and **tqːeu̯s-, structures which are extremely rare typologically. Moreover, unless we also posit other initial geminates (which apparently merged completely with their non-geminate counterparts), Pre-PIE would be, I think, the only language ever posited to allow initial /qː/, but not other initial geminates. Commented Feb 11 at 20:37
  • @JanusBahsJacquet iirc Kortlandt & Kloekhorst generally hold that the "voiceless" stops were geminates in PIE, so this wouldn't be unique to the laryngeals
    – Tristan
    Commented Feb 12 at 10:11
  • This does not provide an answer to the question. Once you have sufficient reputation you will be able to comment on any post; instead, provide answers that don't require clarification from the asker. - From Review
    – cmw
    Commented Feb 18 at 15:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.