greek hippos and hippeus

The Journal of Indo-European Studies

The derivational history of Greek ·ppow and

flppeÊw

Michiel de Vaan Leiden University

[email protected]

The recent insight that the Proto-Anatolian word for ‘horse’ was *@ek-u- suggests that the non-Anatolian word *h1ekuo- ‘horse’ resulted from thematization. Its source may have been the genitive singular *h1kuós of the Early PIE u-stem for ‘horse’. In Greek, the vowel i in ·ppow may reflect a prop vowel which regularly arose in the cluster *h1ku-, showing the generalization of *h1kuó- in a prestage of Greek. The suffix of flppeÊw ‘horseman’ may have arisen from hypostasis of the locative singular *h1kèu ‘on the horse, on horseback’, yielding *h1kèus ‘horse-rider’; thence, the suffix spread to other occupational denominations.

1. In his recent etymological dictionary of Hittite, Alwin Kloekhorst (2008: 237-239) convincingly shows that the Anatolian words for ‘horse’ go back to a Proto-Anatolian u-stem *@ek-u- ‘horse’ from PIE *h1ék-u-. Compare the attestations: Hittite *ekku-(c.): ANSE.KUR.RA-us [nom.sg.], ANSE.KUR.RAHI.A-un [acc.sg.], ANSE.KUR.RA-as [gen.sg.], ANSE.KUR.RAMES-us [acc.pl.]; Cuneiform Luwian *ássu- or *azzu- (c.) (ANSE.KUR.RA-us [nom.sg.]), Hieroglyphic Luwian ásu- (c.) ‘horse’; Lycian esb- ‘horse’ (esbedi [abl.-ins.], esbehi [gen.adj. nom.sg.c.]). The Lycian word is mostly cited as esbe- (e.g. by Melchert 2004: 17), but, as Kloekhorst argues, “this is not necessarily correct as the -e- visible in abl.-instr. esbedi and gen.adj. esbehe/i- in both cases is inherent to the ending (-edi ~ CLuw. -áti, -ehe/i- ~ CLuw. -assa/i-).” Kloekhorst infers that the thematic stem *h1ekuo- ‘horse’ found in the other Indo-European languages must be the result of a thematization which was not shared by Anatolian. This, then, is one of the common innovations of the Indo-European dialects that remained a linguistic unity for some time after Proto-Anatolian split off, and one of the indications for the correctness of the Indo-Hittite hypothesis (Kloekhorst 2008: 7–11, Cowgill 1974,

The derivational history of Greek ·ppowand flppeÊw 199

Volume 37, Number 1 & 2, Spring/Summer 2009

and predecessors). In the remainder of this paper I will call Early Proto-Indo-European (EPIE) the prestage reconstructed for all attested IE languages, including Anatolian, and Late Proto-Indo-European (LPIE) the prestage which resulted after Proto-Anatolian split off. Since EPIE *h1éku- and LPIE *h1ékuo- both mean ‘horse’, a derivation of *h1éku-o- as ‘with speed, speedy’ = ‘horse’ from *h1éku- ‘speed’, as proposed by Schindler apud Balles (1997: 221, Anm. 8), cannot be defended anymore. The thematization attested outside Anatolian did not change the meaning ‘horse’, and is therefore best interpreted as the result of a formal reanalysis. Thematization of athematic nouns took place at a larger scale in the prehistory of many Indo-European nouns. Well-known examples include the agent noun suffix *-ter-/-tr- versus the instrument noun suffix *-tro-, and Hittite huuant- ‘wind’ < *h2uh1ent- vs. Sanskrit vàta-, Avestan váta-, Tokharian A wänt, B yente, Latin ventus, Welsh gwynt, Gothic winds ‘wind’ < *h2ueh1nto-. Thematizations from different ablaut grades of a single athematic noun attest to the productivity of this process in the prestages of the individual languages, as in the case of PIE *sue/op-r, *sup-n-os ‘sleep’ yielding OIc. svefn < *suep-no-, Skt. svápna-, Lat. somnus < *sue/op-no-, Lith. sãpnas < *suop-no-, OCS s∫n∫, Gr. Ïpnow < *sup-no-. The most attractive theory proposed to account for nominal thematization is the view that it originated from the hysterodynamic genitive/ablative ending *-os, which was reanalyzed as a nominative singular and led to the creation of the new category of o-stems (Beekes 1985: 167–207, 1995: 194, Kortlandt 2004: 166f.). This explanation is based on the hypothesis that at an earlier, Pre-Indo-European stage, the genitive/ablative could also function as an ergative case, indicating the agent of transitive verbs. This was proposed over a hundred years ago independently by Uhlenbeck 1901 and van Wijk 1902, on the one hand, and by Holger Pedersen 1907, on the other hand; a summary is provided by Kortlandt 1983a and 2008. When the nominative-accusative system of PIE arose, the genitive/ablative ending *-s was reinterpreted as a nominative ending with animate nouns. Some readers may be reluctant to accept that the o-stems arose from a reanalysis, since the attested languages often derive adjectives, collectives and compounds by adding *-o- to

200 Michiel de Vaan


existing stems (Brugmann 1906: 156–165). A famous example of such a derivation is *Hrot-h2- ‘wheel’ (Latin rota) > *Hrot-h2-o- ‘with wheels, waggon’ (Skt. rátha-). Other instances are derived adjectives in *-dhh1o- (Balles 2003), *-iHo-, *-no-, -ro- (Nussbaum 1986: 243), etc., ordinal numbers such as *s(e)ptm-

ó- ‘seventh’, decasuative compounds such as Gr. §nãliow ‘in the sea’, and nouns which might continue adjectives, e.g. Skt. utsá- ‘spring’, presumably from *‘containing water’ (Balles 1997: 2208). Widmer (2004: 33) explicitly distinguishes two types of thematic vowel, denominated ‘o1’ and ‘ó2’: whereas the former does not change the meaning of the original stem (as in Hittite huuant- vs. Skt. vàta-‘wind’), the latter is regarded as a “possessive” suffix by Widmer. In defence of the reanalysis theory, I would like to make the following two remarks. Firstly, the distinction between ‘o1’ and ‘ó2’ is not always self-evident. For instance, Widmer (p. 33) interprets Gr. biÒw ‘bow’ as a derivative *gwiH-o- *‘with a bow-string’ to athematic Skt. jiyà- ‘bow-string’ < *gwieH-, but the meaning ‘bow’ could just as well be due to metonymical use of ‘bow-string’. Secondly and more importantly, the apparent bifold function of *-o- need not be old. An EPIE system in which gen.abl.sg. *-os and nom.sg. *-os existed side by side could easily lead to the reinterpretation of a genitive/ablative as an adjective: ‘(he is) a man of courage’ = ‘a courageous man’, ‘(it is) a path of stone’ = ‘a stony path’. Hence the productivity of o-stem adjectives. Such a reinterpretation of the genitive is actually found in Hittite syntax, cf. Friedrich (1974: 123), Yoshida (1987: 1-11): nominative taiazil ‘theft’ > genitive taiazilas ‘(he) of theft’ = ‘thief’; genitive arkammanas iianun ‘I made (these cities) to (such) of tribute’ ≈ nat-za arkammanallius [iianun] ‘I [made] them tribute-bearing’. Note that an origin from a hysterodynamic gen.sg. in *-ós may also account for the stressed character of “possessive” *-ó- (Balles 1997: 2208). Returning to the word for ‘horse’, a genitive singular in *-os points to an original hysterodynamic (in the terminology of Beekes 1995: 174–190) paradigm of earlier *h1ék-u-. The genitive singular must therefore have been *h1k-u-ós; the familiar full grade of the root in non-Anatolian *h1ékuo- must have been introduced from the old nom.sg. of the u-stem which survives in Anatolian. Thus, I reconstruct the following EPIE paradigm of *h1éku- ‘horse’:



nom.sg. *h1ék-u[-s] acc.sg. *h1k-éu-m gen.sg. *h1k-u-ós → nom.sg. *h1ku-ó-s >> *h1éku-o-s loc.sg. *h1k-éu-i / *h1k-èu

If the root of this noun is the same as in the adjectives Skt. áßú-, Gr. »kÊw ‘swift’, Latin ócior ‘faster’ < *h1ók- or *h1o-h1k-, the original meaning of EPIE *h1ék-u- must have been ‘the swift one’. 2. This hypothesis, which was designed to explain the co-occurrence of an Anatolian u-stem and a LPIE o-stem, has the additional advantage of yielding a plausible explanation for initial i- in Greek ·ppow, Mycenaean i-qo, which has always been a crux of Greek linguistics. It has been proposed that i- is due to raising of *e in the neighbourhood of a labial in the Mycenaean dialect (cf. Meier-Brügger 1992 I: 61), but the number of forms in support of this theory is very small (depa / dipa, temitija / timitija, iqo), and the first two forms show a vowel vacillation which is absent from iqo and ·ppow.

Initial i- in ‘horse’ may now be understood as a lone survivor from word-initial *h1k- before the introduction of the full vowel *-é- into the first syllable. Greek exhibits a number of forms with unexpected -ι- instead of zero within a consonant cluster (cf. Mayrhofer 1986: 176), as in p¤tnhmi ‘to

spread out’ < *ptnámi, sk¤dnhmi ‘to scatter’ < *skdnámi, Hom.

p¤surew ‘four’ < *kwtwr-, fipnÒw ‘oven’ < *sp-no- (etymology proposed by Vine 1999). According to Vine (who builds on earlier proposals), the i in these forms is an inner-Greek syllabification of a non-phonemic prop vowel in clusters of the structure *(s)TTRV- and *sTRV-. A slightly different structure is found in =¤za ‘root’, Myc. wiriza, which Vine explains from a secondary zero grade *wrdj- beside regular full grade *wrád- < PIE *urh2d- (Lat. rádíx).

We can now see that LPIE ‘horse’, when it was syllabified as *h1k-u-o-, would also contain three initial consonants, and Greek i- could reflect vocalization in the environment *h1CC-. Possibly, the development was *h1CC- > *@CC- > *@iCC- > iCC-. Diachronically, the vocalization to i would have to precede the vocalization of *h1CV- to *eCV- which is otherwise found in Greek. The same environment of *h1- plus two consonants

202 Michiel de Vaan


may be seen in the cluster laryngeal + s + stop in Greek be!’ (OAv. zdí) < *h1s-d

hi, thus supporting the solution for i- in ‘horse’. Finally, a sequence of glottal stop plus two consonants may have arisen in Gr. fixyËw ‘fish’ < PIE *dghuH- (Arm. jukn, Lith. zuvìs, Latv. zuvs), if we assume that *d was preglottalized: *@dgh- > *@idgh- > *@ikth- > ikhth- (Haye van den Oever, pers.com.).

A different solution is proposed by Ruijgh (1995: 353–355), who compares to Gr. fixyËw ‘fish’, fixyuw ‘strength’ < PIE *sgh-uH- and ‘kite’ < *tkiH-ino- (?; to Arm. c‘in, Skt. ßyená- < *(t)kieH-in-, cf. Beekes 2003: 200). Since initial clusters xy, sx, kt are otherwise unproblematic in Greek, Ruijgh surmises that the forms in i- were borrowed from a cognate and neighbouring “proto-grécoide” language or dialect, in which these initial clusters were not allowed. A similar dialect would be responsible for the development *h1kwó- > *kwó- > *ikwó-. While this scenario is possible, it requires the assumption of an extra, unknown dialect from which Greek borrowed a couple of words. It is striking that in fixyËw, fixyËw and fikt›now, the initial cluster of two stops is followed by a glide which would serve as a consonant when the suffix had the full grade: *dgh-ueH-, *sgh-ueH-, *tk-ieH-. Hence, initial i- could have arisen as an alternative vocalization *iTTRV- as opposed to the type *TiTRV- seen in etc. Although not all the details in the prehistory of these words are clear, I would reckon with the possibility that we are dealing with regular reflexes in both cases, rather than with dialect borrowings. Note that we find a vocalization to *eCV- in *dkMtom > Gr. •katÒn ‘100’ and *duidkMti > Gr. ‘20’, where *d is followed by only one stop plus a vocalic resonant or glide (cf. Kortlandt 1983b).

The introduction of -e- in the stem of ‘horse’ was a trivial development found in all non-Anatolian languages except Greek, unless the personal name ÉEpeiÒw (Homer, Pindar), the mythical constructor of the Trojan Horse (Odyssee 11.523), and/or the ethnonym ÉEpeio¤ (Iliad, Pindar) from Elis contain *ekwo- ‘horse’ and thus show that Proto-Greek contained both formations. The survival of a preform *ékus or *ékwos until a relatively recent date may also explain the barytone accent of

(Ruijgh 1995: 355), viz. as the result of analogy. The only element still defying an explanation is the initial h-,

* Ä



which is not found in the PN LeÊk-ippow. Ruijgh (1995: 355)

explains it from analogy with ërmata ‘chariot’. This interpretation of the Greek data implies that the

image of a uniform post-Anatolian preform *h1ékuo- must be abandoned. That seems a high price to pay, but note that both thematization (of *h1éku-) and paradigmatic ablaut levelling (in *h1kuó-) frequently occur in all branches of Indo-European. 3. The identification of a EPIE u-stem *h1ék-u- ‘horse’ also allows for a new analysis of Greek flppeÊw ‘horseman’. This noun represents a category of derived agent nouns which “kennzeichnen einen Mann nach seinem Beruf oder Stand” (Risch 1974: 157). The suffix synchronically alternates between /eu/ in front of consonants and /éw/ in front of vowels; it could thus reflect a uniform long-vowel form *-éu- (with shortening in front of tautosyllabic consonants by Osthoff’s Law), or go back to an earlier ablaut *eu versus *éu. There are some dialectal variants with a short vowel in the oblique cases, in particular gen.sg. -eow in Arcadian, dat.sg. -e› in Attic. Haug (2002: 111–115) argues that these may continue PGr. *-ew-, but I agree with Perpillou (1973: 63–72) and Widmer (2008: 626, fn. 24) that these can be analogical to the paradigm of the s-stems.

Nouns with this suffix are well attested in Mycenaean and Homer (cf. Schindler 1976: 350, Barto™ek 2003: 279–290; the lists given below are modelled on Schindler’s account). In Mycenaean, the phonological interpretation, the etymology and the status as appellative or personal name are often unclear. I therefore exclude from the enumeration the approximately 150 personal names, in which -eus became very productive. The appellatives mostly indicate male agents (professions), as well as names for vessels (for an explanation of this type see Leukart 1983) and sheep terms. Among the formations with a relatively certain meaning and etymology, we find:

a. Derived from o-stems: apiporeu ‘amphora’ (to forÒw‘bearing’), ijereu ‘priest’ (to flerÒw ‘holy’), kakeu ‘smith’ (to xalkÒw ‘copper, bronze’), kanapeu ‘fuller’ (to knãfow‘carding-comb’), kerameu ‘potter’ (to k°ramow ‘potter’s earth’), wirineu ‘tanner’ (to =inÒw ‘hide’).

204 Michiel de Vaan


b. From other stems: amotewo ‘cartwright/assembler’ (to ërmat- ‘chariot’), meritewo ‘bee-keeper’ (to m°lit- - ‘honey’), kotonewe ‘who possesses a property’ (to kto¤na‘piece of land’), opiteukeewe ‘inspector of arms?’ (to teËxow ‘armor’), perekeu ‘who plaits?’ (to pl°kv ‘to plait’ or a derivative thereof), zeukeusi ‘driver of a yoke of oxen?’ (to zeËgow‘pair of oxen’). c. Unknown basis (possibly a noun or adj. in *-lo-): qasireu ‘local officer’.

In Homer, by contrast, the o-stems provide the large majority of the derivational bases for the eu-stems:

a1. From concrete o-stem nouns: flppeÊw ‘charioteer, horseman’, kerameÊw ‘potter’, ofikeÊw ‘inmate of one’s house’, oÈreÊw ‘mule’ (to oÔrow ‘boundary’?), porymeÊw‘ferryman’ (to porymÒw ‘ferry; strait’), xalkeÊw ‘copper-smith’. a2. From abstract o-stem nouns: nomeÊw ‘herdsman’ (to nomÒw ‘place of pasturage’), tokeÊw ‘parent’ (to tÒkow‘offspring; parturition’), foneÊw ‘murderer’ (to fÒnow‘murder’). To a1 or a2 belong: ÙxeÊw ‘band, strap’ (to ˆxow ‘holder, stead’), pompeÊw ‘guide’ (to pompÒw ‘escort, guide’), foreÊw‘bearer’, émfiforeÊw‘amphora’. a3. From o-stem adjectives: èlieÊw‘fisher; sailor’ (to ëliow‘of the sea’), éristeÊw‘chief, leader’, flerÊwpriest’. b. From other stems: boeÊw ‘rope of ox-hide’ (to boËwor bÒeow), donakeÊw ‘thicket of reed’, trapezeÊw ‘at a table’, épereÊw‘thwarter’, ±peropeÊw‘deceiver’. c. basileÊw‘king’.

Thus, although Homeric Greek is more recent than Mycenaean, it shows a more restricted productivity of the eu-stems. Schindler (1976: 351) resolves this paradox by linguistic reasoning: since a secondary restriction to o-stem bases as found in Homer cannot be motivated within Greek, the reverse must be true: the Homeric pattern is more ancient than the one found in Mycenaean. I agree with this conclusion. In its favor we may also regard the fact that four out of the six Mycenaean nouns derived from o-stems are also attested in Homer: émfiforeÊw, flereÊw, xalkeÊw, and kerameÊw. If

we add knafeÊw, attested in an Old Attic inscription and



Herodotus+, and basileÊw ‘king’ (if derived from a stem in *-lo-), the overlap between Linear B and the oldest alphabetical Greek becomes even more impressive. 4. There are three central questions regarding the suffix -eu-/-éw-: What was its function? Why does it predominantly have the form -éw-? Why were these stems originally restricted to o-stem base nouns?

The solution adhered to in a majority of articles and handbooks is that the suffix *-éw- somehow developed out of a regular hysterodynamic u-stem paradigm (cf. Meier-Brügger 1992 II: 26, Rix 1992: 147, Beekes 1994: 10, Haug 2002: 114). The exact way in which this happened has never been clarified, however. Also, the link with o-stems remains unexplained in this scenario.

Schindler himself in 1976 argued that *-éw- arose by secondary suffixation of *-u- to thematic stems. He arrives at his conclusion by the following chain of arguments:

1. The large majority of Homeric and Mycenaean eu-stems belong to an o-stem basis. 2. Therefore, they cannot be primary u-stems, but must have been built on o-stems (p. 351). 3. “In IE secondary formations, the thematic vowel of the o-stems can be replaced only by *-i-; otherwise it is preserved.” (p. 351) 4. “The eu-stems must contain the thematic vowel throughout their paradigm” (p. 351).

Schindler therefore concludes that “The thematic vowel *-e- is followed by the proterokinetic allomorphs of the suffix *-ew-: nom. *ekue-u-s, gen. *ekue-eu-os” (p. 351–352).

As pointed out by Beekes (1994: 10), the reasoning in points 2 to 4 is far from compelling. Firstly, Schindler does not consider an alternative reason why the eu-stems are closely linked to o-stems: the possibility that they developed from a small nucleus of forms in which an o-stem and an eu-stem were derived from the same root. Such a development is well-known in historical morphology. Consider the example of the Latin adjectives in -idus which are formed especially often to presents in -ére (cf. Nussbaum 1999). The reason is not that -idus contains (an element of) the vowel -é-, but that both formations ultimately go back to different PIE derivations from

206 Michiel de Vaan


identical root forms. For example, lúcidus ‘bright’ and lúcére ‘to shine’ are both derived from *le/ouk- ‘light’. This objection removes the basis for step 4 in Schindler’s scheme, which posits the presence of the thematic vowel with the suffix *-ew-.

Secondly, Schindler proposes to derive the lengthened grade suffix *-éw- from a PIE sequence *ekue-eu-, by the addition of a vowel-initial suffix to a stem already ending in a thematic vowel. Such a morphological procedure seems highly unlikely to me. A word-internal sequence of two full vowels is at variance with the (generally acknowledged) principle that there can only be one full vowel (*e, *o, *é, *ó) per syllable in PIE. Indeed, it is unlikely that a sequence *-e-e- was allowed phonotactically in PIE at all. There is no certain example of a derivational sequence *-e-e- (or, for that matter, *-e-o-, *-o-e-, or *-o-o-) in nominal derivation. Of course, we do find -o-ei and *-o-es in nominal endings of the o-stems, but these are clearly very recent formations. In verbal derivation, Greek and Indo-Iranian show thematic subjunctive endings in *-e-e- (23sg, 2pl.) and *-o-o- (1sg, 13pl.). Additional support for a sequence *-e-e- is often seen in the Italic subjunctives in *-(s)é-, such as those of the first conjugation (Latin amet ‘love’, Oscan deiuaid ‘swear’) and of s-presents (Latin amaret < *-s-é-, Oscan fusíd ‘be’); thus e.g. Meiser (1998: 200), Tichy (2000: 99), Meier-Brügger (2002: 167), Fortson (2004: 95). An alternative explanation for *é in these forms is the PIE optative suffix *-ieh1- (Jasanoff 1991: 87, 96), and there is strong distributional evidence that this accounts for all Italic subjunctives in -é- much better than *-e-e- (Schrijver 2006: 58f.). Hence, subjunctive *-e-e- and *-o-o- in Greek and Indo-Iranian are best regarded as post-PIE (cf. Beekes 1995: 245).

Thirdly, although he has a theory about the ‘how’, Schindler does not explain the ‘why’ of the problem. Why was the suffix *-(e)u- used for deriving nouns indicating professions and occupations; in other words, what was the semantic motivation? And: why was it specifically with o-stems that the suffix *-eu- came to be productive in Greek? 5. The conclusions reached in section 1, that the PIE word for horse was a u-stem, and in section 2, that Greek ·ppowmay directly continue a form of that stem, suggest the following solution to the formal and semantic problems. The loc.sg. of the EPIE word *h1éku- ‘horse’ that we have posited



above would either have had the form *h1kéui, or, as an endingless locative, *h1kèu ‘on the horse, on horseback’. A hypostasis of the latter form (that is, a reinterpretation as a strong case form) with subsequent sigmatization in the nom.sg. would have yielded a new noun *h1kèu-s ‘the one on horseback, horse-rider’. A similar process of reinterpreting the locative as the nominative has been claimed for Skt. sákhay-, Av. haxaii- ‘companion’ < *sokw-h2-oi- to a loc.sg. *sokw-h2-ei ‘in the retinue/loyalty’ (Widmer 2008: 621), Skt. obl. ßír§án- ‘head’ < loc.sg. *k rh2-s-én, Av. xsapan- ‘night’ < loc.sg. *kwsp-én ‘at night’ to xsap- ‘night’ (Nussbaum 1986). In order to explain the resulting geminate -pp- in flppeÊw, we must assume that the stem form *h1ku- was reintroduced into the original loc.sg., yielding Pre-Greek *h1kuèus. This form could lead directly to flppeÊw if one adopts the explanation of Greek -i- as given above; even if one is not prepared to believe this, one could assume that i- was adopted in Greek from the word for ‘horse’.

This hypothesis surmises that the entire Greek type of nouns in –euwhad as its starting point the noun flppeÊw; there could have been other loc.sg. forms of u-stem nouns where the same process took place, but none are attested. This would seem a small basis, especially since the word for ‘horseman’ itself is not attested in Mycenaean. However, there are two weighty arguments in favor:

• Of all the Homeric eu-nouns derived from o-stems, flppeÊw is the only one that can also be linked to a PIE

u-stem. Hence the connection between o-stems and ευ-stems that is so striking.

• The meaning of flppeÊw can be understood as ‘horserider’ > ‘one who is occupied with horses, drives horses, charioteer’. This is not so, for instance, with kerameÊw ‘potter’, xalkeÊw ‘copper-smith’ or flereÊw ‘priest’, which cannot be explained as ‘someone who is in/at X’ but rather denote ‘someone concerned with X’. Thus, the meaning of ‘occupation’ which is the productive meaning in Greek, can be understood from an original locatival meaning only in the case of flppeÊw. The model ‘horse’ : ‘horseman’ was then extended to other occupational denominations.

208 Michiel de Vaan


6. The discussion of flppeÊw is often combined with that of

the Greek óu-stems dm≈w ‘slave’, pãtrvw ‘father’s brother’ and

mÆtrvw ‘mother’s brother’. Their nom.sg. in -vw may hide an earlier asigmatic ending *-óu, the interpretation of which is disputed. In the following, I will not put forward any new analysis of my own, but it may be useful to summarize what seems to me the most likely explanation. For the semantics we may follow Pinault, who on several occasions (1997: 227, 2000: 89) has argued in favor of a possessive or collective interpretation of the suffix *-u-. Thus, I assume that *dm-u-, *ph2tr-u- meant ‘of/belonging to the house’, ‘of/belonging to the father’. We have corroborating evidence for these u-stems in other IE languages: PIE *dom-u- in Proto-Slavic *domû ‘house’, *domovû [adj.] ‘house-’ and Skt. dámúnas- (Pinault 2000), and PIE *ph2tr-u- ‘kinsman on the father’s side’ in Lat. patruus, Skt. pit®vya-, Av. túiriia-; see Rau (2004) for a discussion of the latter formations.

A similar explanation has been proposed for Gr. ¥rvw‘warrior; forefather’ by Peters 2002, who connects it with the root *ser- ‘to rob, seize’. He regards Hom. ∏ra [acc.] ‘tribute’ as a reflex of a PIE root noun *sér, *srós ‘robbing, loot’. Hittite sáru- ‘booty, plunder’ < *sóru and Welsh herw ‘raid’, Old Irish serb ‘theft’ < Proto-Celtic *serwá- point to the existence of a PIE u-stem to the same root. The noun ¥rvw, Peters argues, may then be explained on the basis of an original nom.sg. *sér-óu ‘who captures, robber’ with long -é- presumably taken from the root noun. The Myc. theonym tiriseroe /triséró(h)ei/ [dat.sg.] may show generalization of a suffix form *-óh- on the basis of the nom. and acc.sg., thus Peters (2002: 363).

As for the morphology, the Greek nouns show the structure nom.sg. *CC-óR, acc.sg. *CC-óR-m, which was one of the derivational types derived from an EPIE hysterodynamic type *CéC-R, *CC-éR-m by Beekes (1985: 161). In the root, the zero grade *CC- was generalized from the accusative, whereas in the suffix, the full grade in *-óR arose in the nominative and then spread to the accusative. A trace of the asigmatic ending *-óu is found in the Hittite au-stems, in particular harnáu(s), harnu- ‘birthing chair’ (< *h3ér-nou-), where the nom.sg. is attested both as commune har-na-a-us and as neuter har-na-a-ú. As argued by Kloekhorst (2008: 310), this points to a PIE nom.sg. *h3ér-nóu which was either sigmatized to mark its



common gender, or was reinterpreted as a neuter (a phenomenon described by Weitenberg 1995).

The old asigmatic nominative is confirmed by Greek. Since the ending of flppeÊw must go back to *-éus with

Osthoff’s shortening, we would expect an outcome *-ouw for an original ending *-óus if it were equally old. The fact that we find nom.sg. -vw suggests that these forms remained asigmatic at least until after Osthoff’s shortening had ceased to work, with an ending *-≈ that was identical to that of the f. ói-stems

of the type peiy≈‘persuasiveness’. For the development of the attested paradigm of the Gr. ou-stems see Peters 2002: 362f. In summary: *dom-, *dem- ‘house’ *dm-u- ‘belonging to the house’: nom.sg. *dmòu ‘belonging to the house(hold), slave’ > Pre-Proto-Greek *dmò >> Proto-Greek nom.sg. *dmòs *ph2ter- ‘father’ *ph2tr-u- ‘belonging to father’: nom.sg. *ph2tr-òu ‘relative on father’s side’ > PPGr. *patrò >> PGr. *patròs This explanation removes the semantic objections put forward by Widmer (2008: 624, fn. 20), who argues that it is difficult to get from *dem-u- ‘house’ to *dmòu- ‘household’ by means of internal derivation. I start from different premises: the word for ‘house’ was a root noun, while *de/om-u- must have had a derivational meaning. Widmer’s own solution of a locative singular *dom-eu ‘in house, in the household’, whence with hypostasis *dm-óu- ‘slave’, does not explain the origin of the locatival suffix. References

Balles, Irene Griechisch êf(e)now “Reichtum”. Historische Sprachforschung 110:

215–232.

210 Michiel de Vaan


2003 Die lateinischen idus-Adjektive und das Calandsystem. In: Tichy, Eva, Dagmar Wodtko and Britta Irslinger (eds.), Indogermanisches Nomen. Derivation, Flexion und Ablaut. Akten der Arbeitstagung der Indogermanischen Gesellschaft, Freiburg, 19.-22. Sept. 2001, 9–29. Bremen.

Barto™ek, Antonín 2003 Handbuch des mykenischen Griechisch. Heidelberg. Beekes, Robert 1985 The origins of the Indo-European nominal inflection. Innsbruck. 1994 The neuter plural of thematic nouns. Derivatives from a stem in -e-

from thematic nouns. In: Dunkel, George e.a. (eds.) Früh-, Mittel- und Spätindogermanisch, 1–15. Wiesbaden.

1995 Comparative Indo-European Linguistics. Amsterdam / Philadelphia. 2003 Historical phonology of Classical Armenian. In: Kortlandt,

Frederik: Armeniaca. Comparative Notes. With an appendix on the historical phonology of Classical Armenian by Robert S.P. Beekes, 133–225. Ann Arbor.

Brugmann, Karl 1906 Grundriss der vergleichenden Grammatik der indogermanischen

Sprachen. II. Band: Lehre von den Wortformen und ihrem Gebrauch. 1. Teil: Allgemeines. Zusammensetzung (Komposita). Nominalstämme. 2. Bearbeitung. Straßburg.

Cowgill, Warren 1974 More evidence for Indo-Hittite: the tense-aspect systems. In:

Heilmann, Luigi (ed.) Proceedings of the Eleventh International Congress of Linguists, 557–570. Bologna.

Fortson, Benjamin 2004 Indo-European Language and Culture. An Introduction. Oxford Friedrich, Johannes 1974 Hethitisches Elementarbuch. Erster Teil: Kurzgefasste Grammatik. Dritte,

unveränderte Auflage. Heidelberg. Haug, Dag 2002 Les phases de l’évolution de la langue épique. Göttingen. Jasanoff, Jay 1991 The origin of the Italic imperfect subjunctive. Historische

Sprachforschung 104: 84–105. Kloekhorst, Alwin 2008 Etymological Dictionary of the Hittite Inherited Lexicon. Leiden. Kortlandt, Frederik 1983a PIE verbal syntax. Journal of Indo-European Studies 11: 307–324.



1983b Greek numerals and PIE glottalic consonants. Münchener Studien zur Sprachwissenschaft 42: 97–104.

2004 Indo-Uralic consonant gradation. In: Hyvärinen, Irma, Petri Kallio and Jarmo Korhonen (eds.) Etymologie, Entlehnungen und Entwicklungen. Festschrift für Jorma Koivulehto zum 70. Geburtstag, 163–170. Helsinki.

2008 C. C. Uhlenbeck on Indo-European, Uralic and Caucasian. www.kortlandt.nl

Leukart, Alex 1983 Götter, Feste und Gefäße. Mykenisch -eus und -éwios: Strukturen

eines Wortfeldes und sein Weiterleben im späteren Griechisch. In: Heubeck, Alfred and Günter Neumann (eds.) Res Mycenaeae, 234–252. Göttingen.

Mayrhofer, Manfred 1986 Lautlehre. In: Indogermanische Grammatik. Band I, 2. Halbband, 87–

181. Heidelberg. Meier-Brügger, Michael 1992 Griechische Sprachwissenschaft. Two volumes. Berlin / New York. 2002 Indogermanische Sprachwissenschaft. 8., überarbeitete und ergänzte

Auflage. Berlin / New York Meiser, Gerhard 1998 Historische Laut- und Formenlehre der lateinischen Sprache. Darmstadt. Melchert, H. Craig 2004 A Dictionary of the Lycian Language. Ann Arbor / New York. Nussbaum, Alan 1986 Head and Horn in Indo-European. Berlin [etc.]. Pedersen, Holger 1907 Neues und nachträgliches. Zeitschrift für vergleichende

Sprachforschung 40: 129–217. Perpillou, Jean-Louis 1973 Les substantifs grecs en –eÊw. Paris. Peters, Martin 2002 Aus der Vergangenheit von Heroen und Ehegöttinnen. In: Fritz,

Matthias and Susanne Zeilfelder (eds.) Novalis Indogermanica. Festschrift für Günther Neumann zum 80. Geburtstag, 357–380. Graz.

Pinault, Georges-Jean 1997 Remarques sur le pluriel tokh. B akrúna, A ákrunt. In: Lubotsky,

Alexander (ed.) Sound Law and Analogy. FS Beekes, 219–233. Amsterdam / Atlanta.

2000 Védique damúnas-, latin dominus et l’origine du suffixe de Hoffmann. Bulletin de la Société de Linguistique de Paris 95: 61–118.

212 Michiel de Vaan


Rau, Jeremy 2004 PIE *ph2trou-/*ph2t®u- and its derivation. Handout, XII. Fachtagung

der Inodgermanischen Gesellschaft. Cracow, October 2004. Risch, Ernst 1974 Wortbildung der homerischen Sprache. 2., völlig überarbeitete

Auflage. Berlin. Rix, Helmut 1992 Historische Grammatik des Griechischen. Darmstadt. Ruijgh, Cornelis 1995 Observations sur les voyelles d’appui en proto-indo-européen et

en grec ancien. In: Smoczy!ski, Wojciech (ed.) Kurylowicz Memorial Volume, part I, 345–356. Cracow.

Schindler, Jochem 1976 On the Greek type flppeÊw. In: Morpurgo Davies, Anna and

Wolfgang Meid (eds.) Studies in Greek, Italic and Indo-European Linguistics (FS Palmer), 349–352. Innsbruck.

Schrijver, Peter 2006 Review of: Gerhard Meiser, Veni Vidi Vici. Die Vorgeschichte des

lateinischen Perfektsystems (München: Beck, 2003), Kratylos 51, 46–64.

Tichy, Eva 2000 Indogermanistisches Grundwissen für Studierende

sprachwissenschaftlicher Disziplinen. Bremen. Uhlenbeck, Christiaan Cornelis 1901 Agens und Patiens im Kasussystem der indogermanischen Sprache.

Indogermanische Forschungen 12: 170–172. Vine, Brent 1999 Greek =¤za ‘root’ and “Schwa Secundum”. In: Ivanov, Vyacheslav

and Brent Vine (eds.) UCLA Indo-European Studies, Volume I, 5–30. Weitenberg, Jos 1995 Sigmatization and thematization in Hittite. In: van den Hout, Theo

and Johan de Roos (eds.) Studio Historiae Ardens, Ancient Near Eastern studies presented to Philo H.J. Houwink ten Cate on the occasion of his 65th birthday, 333–344. Istanbul.

Widmer, Paul 2004 Das Korn des weiten Feldes. Interne Derivation, Derivationskette und

Flexionsklassenhierarchie: Aspekte der nominalen Wortbildung im Urindogermanischen. Innsbruck.

2008 Drei griechische *-óu-Stämme. In: Huber, Brigitte, Marianne Volkart and Paul Widmer (eds.) Chomolangma, Demawend und



Kasbek. Festschrift für Roland Bielmeier zu seinem 65. Geburtstag, 615–630. Halle an der Saale.

van Wijk, Nicolaas 1902 Der nominale Genitiv singular im Indogermanischen in seinem Verhältnis

zum Nominativ. Zwolle. Yoshida, Daisuke 1987 Die Syntax des althethitischen substantivischen Genitivs. Heidelberg.

greek hippos and hippeus

Documents