Notes on Three “Acrostatic” Neuter S-Stems

Home , Caland system, Digamma, Narten present

Stefan Höfler Notes on three “acrostatic” neuter s-stems

Abstract: The aim of this paper is to investigate the three possible “acrostatic” neuter s-stems *mēd-es-,*sēd-es-, and *h₁ēd-es-, and their various continuants and derivatives within the IE daughter languages. I will argue that their lengthened root vowel is neither a result of a nominal ē : ĕ ablaut pattern nor due to systematic root-inherent characteristics, but can, in principle, be explained individually as the product of a secondary remodeling in analogy to verbal formations where a lengthened grade was regular.

Keywords: acrostatic, s-stems, Narten roots, μήδεα, γῆρας, síd, sætr, sēdēs, sersi, *h₁ed

Stefan Höfler: Universität Wien; [email protected]

1 Introduction

The neuter s-stem nouns represent one of the best-established word formation categories within the Indo-European languages. As a distinct stem class they are preserved in Indo-Iranian, in Anatolian, in Greek, in Latin, in Old Irish and in the Slavic languages; in some of them even to this day. Usually, they reflect full grade in the root throughout the paradigm, whereas the suffix shows ablaut from o-grade in the nom.-acc. sg. to e-grade in the oblique stem. Based on his interpretation of different relic forms, Schindler (1975c) argued that this paradigm (henceforth called the “standard paradigm”; e. g. *u̯éku̯-os :*u̯éku̯-es- ‘word, speech’) replaced an older proterokinetic pattern (e. g. *u̯éku̯-s :*uku̯-és-). Even though some aspects of his argumentation have been criticized until recently,1 Schindler’s assumptions remain the basis for today’s communis opinio concerning the ablaut pattern of neuter s-stem nouns. 1.1 At the end of his paper, Schindler mentioned that there could have been s-stems of the acrostatic type B (i. e. R(ḗ)-S(ø) : R(é)-S(ø)-)2 as well, considering forms that reflect lengthened grade in the root or pairs with root vowel alternations such as Skt. gas ‘sin’ (: Gk. ἄγος ‘guilt, curse’), Skt. vsas ‘garment’, Skt.

1 Cf. e. g. Litscher 2007; Hartmann 2012. 2 Cf. for this type Eichner 1973: 68f.

10.1515/if-2014-0016 Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 294 Stefan Höfler vhas ‘conveying’, Gk. γῆρας ‘old age’ (: γέρας ‘gift of honour’), Gk. ἦϑος ‘custom, habit’ (: ἔϑος ‘id.’), Gk. μήδεα and μέδεα pl. ‘male genitals’, Gk. ῥῆγος (also ῥέγος) ‘rug, blanket’, and OIr. síd (*sḗdos) ‘fairy mound’.3 These root vowel alternations can, in principle, be explained as a result of a paradigmatic split: The long vowel forms are believed to be the continuants of the strong stem, while the words reflecting a short vowel would then continue the root ablaut of the oblique stem. In terms of heuristics, this scenario is parallel to the root vowel alternation ĕ : ø that we find in a pair like Ved. váras- ‘width’ : Ved. úras- ‘breast’, whose existence can be explained via a paradigmatic split of the underlying proterokinetic s-stem *(h₁)u̯érH-os : *(h₁)u̯r̥H-és-.4 In the following years, some of these “acrostatic” s-stems have been more plausibly explained through various phonological or morphological developments, and/or different etymologies (cf. Höfler 2012 for an exhaustive overview). However, Schindler’s mere mentioning the possibility of their existence has consequently led to a broad acceptance of this type of neuter s-stems among many scholars to this day.5 One can, however, raise several objections against “acrostatic” neuter s-stems, or at least against the notion of “acrostatic”6 used for the group of s-stems that seemingly reflect an ē : e root ablaut. First, and maybe worst of all, is the basis on which this assumption was arrived at, namely the mere vowel alternation7 ē : e. It is true that this vowel alternation is crucial to the identification of words that we categorize as acrostatic nouns of the type B. Additionally however, these words usually reflect the expected zero grade of the suffix, as in *ḗku̯-r̥ :*éku̯-n̥-s ‘liver’, where the root ablaut as well as zero-grade suffix *-r̥ in the nom.-acc. sg. are

3 Examples cited after Schindler 1975c: 267. 4 Cf. Nowicki 1976: 27; Stüber 2002: 186. 5 Cf. for example, more or less unanimously, Schaffner 2001: 77f. and 587; Stüber 2002: 22 and passim; Widmer 2004: 50; Casaretto 2004: 553f.; Hartmann 2012: 79. 6 It is unwise and misleading to use the term “acrostatic” not only for the accent-ablaut classes acrostatic A (viz. R(ó)-S(ø):R(é)-S(ø)-) and acrostatic B (viz. R(ḗ)-S(ø):R(é)-S(ø)-) of the standard model (for this cf. Eichner 1973: 91 note 33; Schindler 1975c: 262–4; and also Meier-Brügger 2002: 203ff. for a historical overview, as well as Widmer 2004: 49ff.), but also for nouns with a columnal accent on the root syllable as a result of secondary remodeling, e. g. the s-stems following the standard paradigm (as done by Hartmann 2012: 52 et passim). 7 It would, in turn, be bold to assume the existence of s-stems of the acrostatic type A, that is with the root ablaut o : e, only because of the root vowel alternation of pairs like Gk. ὄχος ‘chariot’ and the Hesychius gloss ἔχεσφι · ἅρμασιν (dat. pl.), where the o-vocalism of the former is best explained through analogy after a thematic noun of the same meaning ὄχος m. (e. g. h.Cer. 19) or *ὄχᾱ f. (Myc. wo-ka; cf. Panagl 1982: 154–6). For similar analogical processes within Latin see also below note 28.

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 Notes on three “acrostatic” neuter s-stems 295

confirmed by Gk. ἧπαρ and Av. yākarə (< *ḗku̯-r̥), and Ved. yákṛ-t, Lat. iecur and NPers. ǰigar (< *éku̯-r̥). The zero-grade suffix of the oblique is reflected by gen. sg. Ved. yaknás, Gk. ἥπατος, Lat. *iecinis (< *éku̯-n-es) and the stem of the remodeled fem. pl. OLith. jẽknos.8 Our group of “acrostatic” s-stems, however, lacks this important feature.9 Instead, we find the plain o : e suffixal ablaut, that we know from the standard paradigm of neuter s-stems. In fact, there seems to be no difference at all between the descendants of the proterokinetic type s-stems and our group of “acrostatic” ones. One could, of course, argue that both inflectional types had already merged into the standard pattern in Proto-Indo-European times, but we would then hope for at least some evidence, other than the long vowel, to support the assumption of “acrostatic” s-stems in the first place. In fact, we do have s-stems other than the proterokinetic standard ones and the unjustly labeled “acrostatic” ones, but there is hardly any trace whatsoever of the fact that some of those coalesced into one category in PIE or in einzelsprachlich times. There are hysterokinetic s-stems, appearing predominantly as second

8 Reconstruction and forms after Eichner 1973: 68f.; Schindler 1975b: 5f.; Schaffner 2001: 77. Cf. NIL: 392–95 for a collection of various other suggestions (with lit.), and also for the peculiar occurrence of R(o), as in Lat. gen. sg. iocineris, iocinoris etc. De Vaan (2003: 68f.) argues that Av. yākarə might only be an error or occasional lengthening in the manuscript. 9 The only s-stem for which such an analysis is in principle acceptable is the pair Gk. γῆρας ‘old age’ (Il.+) and γέρας ‘gift of honour’ (Il.+), which could individually continue the strong and the weak stem of an acrostatic neuter *ǵḗrh₂-s :*ǵérh₂-s-. However, it is far more plausible that the form with the lengthened vowel is an inner-Greek innovation. Greek must have inherited regular γέρας < *ǵérh₂-s meaning ‘old age’ (for which cf. also γεραιός ‘old’ [Il.+], later analogically γηραιός [Hes.+], and also Ved. jarás- [m. or] f. ‘old age’ [RV+]), which was remodeled in analogy to the aorist ἐγήρᾱ (for which cf. Peters 1980: 314 note 259) to γῆρας. The original form γέρας was consequently restricted to the specialized meaning ‘gift of honour’ (thus also Stüber 2002: 83f.; Meissner 2006: 82). This instance of analogical interference can be compared to the pair βένϑος ‘depth’ (Il.) : βάϑος ‘id.’ (Ion.-Att.), where the former is the regular variant in the Iliad but later only found in poetry, while the latter is absent in Homer but later on appears very frequently (cf. Meissner 2006: 65–67 for the attestations) and owes its shape to the associated adjective βαϑύς ‘deep, high’ (Il.+) in which the zero grade is regular. An even closer example might be the pair πένϑος ‘grief, sorrow, mourning’ (Il.+) : πάϑος ‘suffering; incident, experience, impression’ (Aesch.+), semantically overlapping but not synonymous (cf. Meissner 2006: 67f. for the attestations and semantics), the latter of which seems to have been influenced by the aorist ἔπαϑον (thus Stüber 2002: 46). There are several more examples like these (cf. also Stüber 2002: 46–50; Meissner 2006: 65–72) that collectively suggest that a secondary remodeling of γέρας to γῆρας after ἐγήρᾱ etc. is fairly reasonable.

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 296 Stefan Höfler members of compounds,10 and amphikinetic s-stems, mostly as collectives,11 both of which seem systematically connected to the proterokinetic ones via internal derivation. There is no indication of a comparable derivational dependence concerning the “acrostatic” s-stems. One important question that has to be taken into account is: What would an original acrostatic s-stem look like? Let us consider, for instance, Ved. vsas ‘garment’. It clearly belongs to the root *√u̯es12 ‘[Kleidung] anhaben, bekleidet sein mit’. Due to the lengthened grade in the root, the noun has sometimes been labeled an “acrostatic” s-stem (cf. e. g. Nowicki 1976: 116; Stüber 2002: 171f.). How- ever, the PIE ancestor of this s-stem should then, at some time, have been none other than nom.-acc. sg. *u̯ḗs-s : gen. sg. *u̯és-s-s. Since geminated (and likewise also triple) *s was reduced to a single *s (cf. Mayrhofer 1986: 120f.), the outcome would have been *u̯ḗs :*u̯és. In view of language economy,13 I see no motivation in the heads of PIE speakers to generate such a form in the first place. But if they did, and conceding that they may have already introduced the gen. sg. ending *-os, therefore: *u̯ḗs :*u̯és-os, I see no reason why this word should not have been conceived as a root noun right away, because synchronically there was no intelli- gible suffix anymore, and thus leaving no trace whatsoever of the former s-stem inflection.14

10 For example *h₁su-men-és- in Ved. sumánas-, Av. humanah-, Gk. εὐμενής, and various formations in the daughter languages, but also uncompounded as possessive adjectives, cf. Ved. apás- ‘busy’ (: ápas- ‘work’) and Gk. ψευδής ‘mendacious’ (: ψεῦδος ‘lie’), for which cf. Schaffner 2001: 585f.; Stüber 2002: 27; Widmer 2004: 31f. and 65f., as well as the simplex Lat. Cerēs,-eris ‘Ceres; goddess of agriculture, etc.’, and some other forms. 11 For example Av. man <*mén-ōs as the synchronic nom.-acc. pl. of manah- ‘mind, thought’ < *mén-os, but also substantives without an evident collective meaning, such as *h₂éu̯s-ōs > Ved. uṣás-, Av. ušah-, Gk. (Ion.) ἠώς ‘dawn’ etc., for which cf. Stüber 2002: 22–6; Widmer 2004: 30 and 111ff.; Höfler 2013. 12 LIV²: 692f. 13 One significant difference between other athematic acrostatic nouns and the alleged “acrostatic” s-stems is that the suffix *-s- was unable to become vocalic and could not serve as a syllable nucleus. Therefore, “acrostatic” s-stems would hypothetically be monosyllabic, both in the nom.-acc. sg. and in the oblique stem, which would clearly separate them on the surface from the regular disyllabic acrostatic substantives (e. g. i-stems, u-stems, r/n-heteroclites, etc.). 14 The general assertion followed in this paper, viz. that lengthened grades in verbal formations had an analogical influence on short-vowel s-stems, and then gave rise to long-vowel equivalents, appears to be illicit in this case, since there seem to be no long-vowel verbal formations of this root in Indo-Iranian or elsewhere (except for irregular full grades in váste ‘wears’ etc. cf. Schindler 1994: 398). However, Malzahn (2010: 896f.) has suggested that, while Toch. B wäs- ‘don, wear (clothes)’ points to a non-Narten root aorist *us-to, the fem. pl. gerundive Toch. A waṣlaṃ must be derived from a thematic present with a pre-Proto-Tocharian long root vowel. This provides us

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 Notes on three “acrostatic” neuter s-stems 297

It is therefore fairly safe to project that the origin of our “acrostatic” s-stems must be relatively recent and date back to a point when the standard paradigm had already evolved, which means that at the same point the proterokinetic inflectional pattern had already become obsolete. It would seem only reasonable to accept that also the acrostatic pattern had, at that time, lost its systematic validity, reducing the possibility of newly formed “acrostatic” s-stems to absurdity. 1.2 Two decades later, Schindler (1994) offered an alternative explanation for long-vowel s-stems by suggesting the existence of “Narten”roots – roots that show a systematic ablaut ē : ĕ in verbal and nominal formations instead of the common ĕ : ø pattern.15 Following and elaborating his theory, the long-vowel s-stems would no longer hint at an original acrostatic paradigm, but present nothing else than the usual proterokinetic type. The languages that exhibit a lengthened root vowel would have generalized the strong stem of the noun (e. g. *sḗd-os), where the long *ē would have been the regular full grade of the “Narten” root, while the forms with a short *ĕ would continue the weak stem (e. g. *sĕd-és-), *ĕ being the reduced grade of the root. It is obvious that this explanation has certain advantages, as it gets along without the problematic assumption of two different accent-ablaut types that would have merged into one and the same paradigm. However, the assumption of two different types of roots is not at once compelling, and indeed, there are several examples to prove the opposite. We find, for example, zero grades of alleged “Narten” roots that are just as suspicious as the notorious lengthened grades. And, as Schindler himself admits: “Für das Material bestehen natürlich z. T. Alternativerklärungen.”16

with a possible source for the analogical influence and may also explain the long-vowel forms Toch. B yesti ‘garment’ (as if < *u̯ḗst-o; cf. Malzahn 2004: 217) and Welsh gwisg ‘id.’ (if, as per Klingenschmitt 2008: 196, from *u̯ēs-s-keh₂-, see also below note 170). Ved. váste and the like would then be analyzable as Narten present middles *u̯és-to etc. from a root *√u̯es with inherent punctual semantics ‘to put on (clothes)’ (contra ‘(Kleidung) anhaben, bekleidet sein mit’ as per LIV²: 692). The “stative” meaning ‘to wear’ of the characterized Narten present on the other hand could be interpreted in the same way as in pairs such as: standard root aorist *ku̯éu̯- ‘sich in Bewegung setzen’ (punctual) as in Gk. Hom. ἔσσυτο, σύτο ‘stürmte los’ etc. vs. Narten root present *ku̯ḗu̯- ‘sich bewegen, in Bewegung sein’ as in Ved. ptc. cyávāna- ‘in motion’ etc., whose semantics Kümmel (1998: 195ff.) (with additional examples) describes as “durative Folgehandlung”. We may, therefore, project: root aorist *u̯és- ‘to put on (clothes)’ (punctual) vs. Narten present *u̯ḗs- ‘to have on (clothes)’ (durative consequent action). 15 Similarly in recent years also Klingenschmitt 2008: 196ff. 16 Alternative explanations are, for example, proposed by de Vaan (2004) for the Iranian material. Cf. also Meissner (2006: 72–86 especially for the Greek material), whose proposed solutions are,

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 298 Stefan Höfler

It may, however, be noted that some of the alleged “Narten” roots show a very basic root structure *√CeC. There might have been a tendency to avoid zero grades of the shape *CC word-initially,17 and therefore to reintroduce the full vowel *e. Whether this encouraged the full grade *CeC to be upgraded under paradigmatic pressure by another *e to *CēC as a possible consequence18 is difficult to answer. In this paper, a thorough examination of three alleged “acrostatic” s-stems will be conducted, whose continuants show up in more than only one branch of the Indo-European languages. Incidentally, the roots featured in this paper, *√med,*√sed and *√h₁ed, share the same basic root structure. If it could be de- termined that they act similarly in their ablaut behavior and their derivational properties, this would indeed give a hint at the assumption that root structure had something to do with the occurrence of irregular lengthened grades. But as we will see after the discussion of the long-vowel s-stems and s-stem continuants in question, such an interpretation is not favored by the material. The theory of “Narten” roots might, however, stand a chance after all, yet only in a “light” version. It will be argued that there was no systematic correlation between verbal “Narten” paradigms and nominal formations, but that an analogical influence of verbal lengthened grades (of whatever origin) on nominal stems, and in particular on neuter s-stems, which predominantly functioned as verbal abstracts, is certainly to be expected.19 In fact, a comparable process did happen in historical times, as demonstrated by cases of similar remodeling within the individual languages (cf. πάϑος for πένϑος after ἔπαϑον, see note 9 above). 1.3 The working hypothesis of this paper will therefore be as follows: There are no “acrostatic” s-stems. There are no “Narten” roots sensu stricto. There is, however, a reasonable probability that lengthened grades in default verbal formations can be regarded as one possible source of analogical introduction of a long vowel into the root of a neuter s-stem. Sometimes, however, there might be other, more sensi- ble explanations for long vowels, such as regular sound developments or deriva-

however, not always convincing. A very promising analysis is presented by Peters (2002: 101) and Nussbaum (apud Peters 2002: 101 note 10), where it is asserted that “Narten” nouns only inflected acrostatically (in suffix and ending) when the suffix involved also exhibited acrostatic inflection elsewhere. 17 Word-internally, this does not seem to be the case, where we find *CC from the alleged “Narten” root *√sed ‘to sit down’ for example in *si-sd- (Ved. sdati, Gk. ἵζω, Lat. sīdō), and *ni-sd-o- (Lat. nīdus, Ved. nīḍá-, Germ. Nest). 18 Cf. Strunk’s (1985: 499) principle of “Sekundäraufstufung”. Of course, there is no palpable reason why the roots of the structure *√CeRC and the like should have evaded the zero grades *CR̥C etc., which otherwise occur completely regularly. 19 Similarly already Isebaert 1992: 203.

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 Notes on three “acrostatic” neuter s-stems 299 tional patterns where a lengthened grade is expected (viz. vṛddhi-derivatives). For every proposed solution, examples of similar developments will be given to con- firm that the approach in question is not an arbitrary assumption but can be paralleled by a comparable process.

2 PIE *mēd-es-

The first example of an alleged “acrostatic” s-stem is *mḗd-os. The lengthened grade is reflected by Gk. μήδεα pl. ‘counsels, plans’20 (Il.+) and Arm. mit ‘mind, thought’, whereas Umbr. meřs ‘ius, law’21 seems to go back to regular *méd-os with a short root vowel. 2.1 The supposed development PIE *méd-os > Umbr. meřs relies on two well- known phonological features of the Umbrian language. The change of intervocalic /d/ to /ř/ (cf. also zeřef, serse ‘sedēns, sitting’ < *sedens) and syncope in final syllables, which is also found in Oscan (cf. nom. sg. húrz ‘hortus, garden’ < *hortos). The chronology of these events, however, has been disputed, since it is usually accepted that syncope predates the development /d/ > /ř/, thus leading to an alleged nom.-acc. /meds/ and via assimilation /mets/. Equally problematic is the assumption that /ř/ was secondarily introduced from the oblique cases, since also there the suffix vowel should have been syncopated before /d/ became /ř/. Additional difficulties are induced by the apparent derivative /mersuva/ abl. sg. f. (confirmed by the spelling mersuva [III 11]), going back to *med-es-u̯a-, where the outcome /rs/seems rather unexpected, as well as by another alleged s-stem tuder ‘border, boundary’ (< *tud-es-?) which seems to contradict both syncope and /d/ > /ř/. Meiser has consequently outlined a framework of successive phonological and analogical developments that can positively account for the attested forms.

20 Schindler’s example Gk. μήδεα (Od., Androm. apud Gal., Call., Ant. Lib.) and μέδεα (Archil. 138) ‘male genitals’, also μέζεα (Hes. Op. 512, Lyc.), obviously belongs to a sphere where (tabuistic?) remodeling cannot be excluded. Since it is unclear whether the word was originally identical to the s-stem μήδεα ‘counsels, plans’ (as per Meissner 2006: 80 “[T]he semantic difficulties are not insurmountable”), or belongs to another PIE root *√med ‘swell’ (as per IEW: 706), or is of pre-Greek origin (as per Beekes 2010: 2, 941), it will be omitted from the discussion. 21 In the Umbrian alphabet (henceforth in bold letters) it is attested as meřs, in the Latin alphabet (henceforth in italics), where /ř/ is usually written as 〈rs〉, it appears as mers. The spelling 〈rs〉 in this case should be read as /řs/. For the different readings of the sequence 〈rs〉 cf. Buck 1904: 83 and also below 3.1.

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 300 Stefan Höfler

He assumes that, after the syncope of final syllables, *d first shifted to a voiced fricative *ð, resulting in a paradigm /mets/ : /meðez-/ (with intervocalic *s>*z) where *ð was analogically introduced in the nom.-acc. sg. /meðs/ and was not affected by the following -s. After the syncope of internal syllables, /meðez-/became /meðz-/ and was dissimilated to /mers-/, whereas adjacent to r the fricative *ð was dissimilated or backformed to d (hence tuder).22 In all other positions *ð became ř.23 In the Umbrian corpus the word only appears in the nom. sg., in all instances preceded by a relative pronoun or a conjunction and followed by the copula, in some cases even univerbated with it.24 There are also two derivatives of the s-stem, *med-es-u̯o-25 and *med-es-to-.26,27 The latter has a match in Latin modestus ‘mod-

22 However, this explanation implies that the suffixal vowel of tuder, as opposed to meřs, was not syncopated. Cf. Meiser 1986: 231–8 for a detailed and thorough discussion of the word, which is, unfortunately, best summarized by its last sentence: “Warum freilich die Entwicklung bei *tudes- anders verlief als beim strukturell ähnlichen *medos > meřs, bleibt ungeklärt.” Note however, that the phonologically expected outcome *tuřs,*turs appears as a morpheme in verbal forms, see next note. 23 Cf. Meiser 1986: 226–31 and etuřstamu (Ib 16), eheturstahamu (VIb 55), eturstahmu (VIb 53 [twice]) – as an imp. 3rd pl. of a denominative verb ‘exterminato, (they) should expel’ as if < *eχtudestāmōd (vel sim.) – which shows the proposed development of *d>řadjacent to s, which also explains /meřs/. 24 meřs (Ib 18; twice), mers (VIb 31, 55), mersest (VIb 55; univerbation mers + est, or simply without interpunct), mersei (VIa 28) and mersi (VIa 38, 48; univerbation mers + sei, si [pres. subj. 3rd sg.]), cf. Untermann 2000: 461. 25 mersus nom. sg. m. (III 6), mersuva abl. sg. f. (III 11) and mersuva acc. pl. n. (III 28), cf. Untermann 2000: 473f. 26 mersto acc. sg. m. (VIa 3, 4, 16, 17), mersta acc. sg. f. (VIa 3, 4, 16), meersta acc. sg. f. (VIa 17), merstu abl. sg. m. (VIa 1), merstaf acc. pl. f. (VIa 4) and mersta acc. pl. f. (VIa 3 [twice], 4, 18 [twice]), cf. Untermann 2000: 473. Following Meiser’s argumentation outlined above, the reading has to be /meřsto/, because /ð/ was not influenced by the voiceless /s/. 27 The Oscan magistrate title meddíss (also Marrucinian medix, Marsian medis, meddis, Paelignian medix, Volscian medix, cf. Untermann 2000: 456f.) seems to reflect a compound of *med-os, *med-es- and *dik- (similar to Lat. iūdex ‘judge’ from the s-stem iūs <*ou̯es-+*dik-), though the exact morphological analysis of the first compound member is unclear (*med-(e)s-diks,*medo-diks, *med-diks vel sim.). Nussbaum (1976: 242f. note 5, followed apparently by Tremblay 2010: 208) argues for the latter and supposes a neuter root noun *mḗd :*mĕd-ós, whose weak stem served as the first member of the compound, and whose strong stem yielded (in recharacterized form) Gk. μῆδος and Arm. mit. However attractive this interpretation may seem at first, it entails some major difficulties. The evidence of neuter root nouns in PIE other than names for body parts is scarce (cf. Schindler 1972a: 8; Balles 2006: 258 note 406), and also the existence of an ablaut ē : ĕ in root nouns is not absolutely clear from the material (the most prominent example being *h₃rḗǵ- ‘king’; cf. Schindler 1972b: 37; Schindler 1994: 399, and see note 62 below). At any rate,

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 Notes on three “acrostatic” neuter s-stems 301 erate, calm, restrained’, which differs from it only in the root vocalism. The derivative indicates that Latin must also quite certainly have had a neuter s-stem *medus at some point, which was not only influenced in its vocalism, but later also replaced by the (supposedly synonymous) masculine modus ‘measure, manner’.28 We have presupposed so far that meřs goes back to a short-vowel formation *mĕdos. In fact, an advocatus diaboli could object that vowel length was not con- sistently expressed in Umbrian spelling. There is one attestation of the derivative *med-es-to- occurring with plene spelling as meersta in VIa 17, perhaps hinting at an underlying *mēd-es-to-. This is, however, not very likely. To begin with, the word is written 14 times with a simple e in the very same text, once even in the same line, which makes a misspelling quite probable (cf. Meiser 1986: 140). Fur- thermore, the length indicated cannot easily go back to PIE *ē, as this was raised to /ẹ̄/ written as 〈e, i, eh〉 and 〈e, i, ei, eh, ehe〉 in the two alphabets.29 As one can easily see, there is virtually no guarantee that /mẹ̄řs/ < *mēdos would have been in any way graphically distinct from /meřs/ < *mĕdos, since both forms could by all means have been written as meřs and mers. We will only gain a certain amount of confidence if we happen to find a new inscription where the word appears as 〈miřs〉, 〈mehrs〉 or the like. For now, however, we should stick to the null hypothesis, viz. that meřs, mers does in fact stand for an accurately written /mĕřs/.30 2.2 Arm. mit ‘mind, thought’ (Bible+) is – unlike the other continuants of neuter s-stems in Armenian that were chiefly transferred to the o-stem declension (cf. Matzinger 2005: 37f.) – synchronically inflected as an a-stem.31

the equations Lat. iūs: Umbr. meřs, Lat. iūstus: Umbr. mersto-, Lat. iūdex : Osc. meddíss seem to indicate that Lat. *ou̯os and Osc.-Umbr. *medos were exact semantic matches at the time of their religious and juridical conceptualization within the individual languages of the Italic family, cf. Benveniste 1969: 123–32; Untermann 2000: 456–9. 28 A similar approach is also the most plausible explanation for the vocalism of the s-stem Lat. pondus,-eris ‘weight’, viz. for older *pendus, remodeled under the influence of *pondus m. (preserved only in pondō indecl. ‘in weight’ as a fossilized abl. sg.), cf. Meillet 1922: 96; Walde & Hofmann 1938–1956: 2, 278f. 29 Cf. Buck 1904: 34; Meiser 1986: 27 and 45. Examples include Umbr. fesnafe (IIb 16), which is compared to Osc. fíísnú <*fēsnā (cf. Lat. fēstus, fēriae), Umbr. sehmeniar (Ib 42), sehemeniar (VIIa 52), seme.nies (IIb 1), sehmenier (Vb 11, 16), if, as commonly accepted, they belong to Lat. sēmen (as if *sēmen-io- etc.), and Umbr. plener (VIIa 21, 34), plenasier (Va 2, 14), if akin to Lat. plēnus and plēnārius. 30 Bertocci (2012: 14ff.) argues for a general development *ē > Umbr. e (as far as I can see, limited to the second syllable of a word, which then resists syncope), yet rather on morphological than on phonological grounds. 31 Theoretically, the word need not continue a PIE neuter s-stem but could go back to *mēd-eh₂- (thus, e. g., Meillet 1922: 96) with a lengthened root vowel. There is some conspicuous evidence

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 302 Stefan Höfler

This peculiarity is best explained by the fact that the word is predominantly used in its plural form nom. mit-kʿ, gen.-dat. pl. mt-acʿ (cf. Martirosyan 2010: 470f.). If one supposes that this use was already common in pre-einzelsprachlich32 times (which is indeed suggested by the plurale tantum Gk. μήδεα ‘counsels, plans’, see below 2.3), one could assume that a putative nom.-acc. pl. *mēd-es-h₂ was inherited into Armenian (and into Greek, where it regularly produced μήδεα) and led via *miteʰa and *mita, after adding the common nom. pl. marker *-kʰ (cf. Matzinger 2005: 119ff.), to the attested nom. pl. mit-kʿ, which could then be interpreted as belonging to an a-stem substantive.33 Since this assumption makes perfect sense for both Armenian and Greek on phonological and morphological grounds, but cannot, however, account for Umbr. meřs, it consequently seems reasonable to assume that the formation dates from a common Proto-Graeco-Armenian period, and that also the lengthened grade might be a shared innovation. Within Armenian, mit is isolated, but the Greek material provides us with clues to a possible source of the alleged remodeling. 2.3 The Greek noun μήδεα, attested from the Iliad onwards, belongs to a group of different formations of the root *√med ‘messen, für Einhaltung sorgen, sich

of the existence of substantives with the structure R(ē)-eh₂- in PIE, cf. *bʰēr-eh₂- > Gmc. *bǣrō- f. ‘bier, litter’ in OHG bāra, Germ. Bahre, OEngl. bǣr etc., *h₁ēd-eh₂- in Lith. da f. ‘food’, Latv. ȩ̄da f. ‘bait’, Russ. едá (jedá) ‘food, meal’, OHG āza ‘id.’ etc. (see below 4.7), *sēd-eh₂- > Gmc. sǣtō- in ON sát, OEngl. sǣt ‘ambush’, MHG sāze ‘seat, residence, ambush’ (cf. Darms 1978: 91–102 for more examples and a thorough discussion, and also Isebaert 1992: 203, who proposes an influence of Narten presents). For most of these samples, there are, of course, other possible explanations. Interestingly enough, however, these formations belong to roots for which a “Narten” character has been proposed. In any case, for our Armenian word, this interpretation remains unattractive because of the formally possible and semantically attractive connection to the Greek word (see below). Be that as it may, the above-postulated *mēd-eh₂- seems to be directly reflected in another IE language, namely by OHG māza f. ‘measure, manner’, Germ. Maß f. ‘a mug of beer’. 32 It is unclear if neuter s-stems already formed a proper nom.-acc. pl. by adding *-(e)h₂ to the oblique stem in PIE times. This is, admittedly, suggested by equations like Gk. (Ion.) γένεα ~ Lat. genera (< *ǵenh₁-es-h₂) ~ OCS slovesa ‘words’ (< *ḱleu̯-es-eh₂) ~ OIr. tige ‘houses’ (< *(s)teg-es-(e)h₂), but in Avestan and Vedic the nom.-acc. pl. of neuter s-stems goes back to an amphikinetic collective formation (Av. man <*mén-ōs as the synchronic nom.-acc. pl. of manah- ‘mind, thought’; Ved. mánāṁsi is the result of an analogical transformation of an equally underlying *mén-ōs, viz. insertion of a nasal and addition of the neut. pl. marker -i), which is seemingly older than the forms with *-(e)h₂ that can easily have been formed in einzelsprachlich times (cf. Stüber 2002: 203). Note that in Hittite, where we would perhaps expect an archaic state of affairs, no nom.-acc. pl. is attested for the (commonly accepted) s-stems nēpiš ‘sky, heaven’ and aiš ‘mouth’. 33 Cf. for all this Clackson 1994: 147–9; Olsen 1999: 69; Stüber 2002: 125f.; Matzinger 2005: 17 and 47f.; Martirosyan 2010: 470f.

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 Notes on three “acrostatic” neuter s-stems 303 kümmern’ (LIV²: 423), including the thematic verb μήδομαι ‘to deliberate, contrive, decide’ (Il.+), the agent noun μήστωρ ‘advisor, counselor’ (Il.+), thematic μέδομαι ‘to care for, think of’ (Il.+) and μέδω ‘to rule’ (Emp., Soph.) with the participle μέδων ‘ruler’ (already Homeric), which, taken together, show a peculiar ē : ĕ alternation. There is an obvious semantic connection between the substantive μήδεα ‘counsels, plans’ and the verb μήδομαι ‘to deliberate, contrive, decide’, which justifies the assumption that during their prehistory one may have influenced the root vocalism of the other. As already mentioned in the premises of this paper, there is a better chance of explaining a remodeling of the substantive in analogy to the verb than the other way round; all the more since there are categories within the PIE verbal system where lengthened grades are more or less commonly accepted. If we can find a way of successfully explaining the origin of the lengthened grade in the verb*mḗd-e/o-, it will be only reasonable to accept that pre-einzelsprachlich *mdesa (vel sim.) was analogically remodeled to *mḗdesa, which then led to Arm. mit and Gk. μήδεα. One way of explaining the long vowel in μήδομαι is by assuming that *mēd- represents a contamination of two separate but semantically largely overlapping roots *√med and *√meh₁ ‘(ab)messen’ (LIV²: 424f.) in Proto-Graeco-Armenian times.34 This is, of course, not disprovable, but the coexistence of μήδομαι and μέδομαι would demand that the original root *√med had not entirely been given up in favor of the secondary root *mēd, which seems at best fairly unlikely. The peculiar pair μήδομαι : μέδομαι is far more easily understood if we consider them to be the result of an individual lexicalization of the two stem alternants *mḗd-: *méd- of some acrostatic verbal formation.35 It is clear that, being a medium tantum, the lengthened grade in μήδομαι must be of secondary origin, since we would expect a reduced grade in the middle. Beyond this, a conspicuous long vowel is also found in the Hesychius gloss μῆστο · βουλεύσατο. There are different ways of interpreting this form. Latte (1966: 663) emends it to (ἐ)μήσατο, the regular synchronic s-aorist of μήδομαι, attested since Homer. Chantraine (1968–1980: 693) suggests an original athematic

34 Thus Beekes 2010: 2, 941 (apparently discarding an older view, viz. *√meh₁d as in Beekes 1988: 30). Similar, but less convincing is the account of Meissner, who likes to derive the s-aorist (ἐ)μησάμην (Il.+) not from *md-, but from *meh₁-, from which then “for formal reasons and considering the close semantic relationship with μέδομαι, a present μήδομαι could have been created and μήδεα then may have been derived from it” (Meissner 2006: 81). For Gk. μέτρον (as *med-tro-; not from *√meh₁!) cf. Schindler apud Mayrhofer 1986: 111 and apud Peters 1999: 447 and note 2. 35 Cf. also Isebaert 1992: 195 note 14.

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 304 Stefan Höfler present *mēd-:*mĕd- for μήδομαι and therefore regards μῆστο as an athematic imperfect. However, a Narten present36 does not correspond to the alleged durative character of the root *√med ‘messen, für Einhaltung sorgen, sich kümmern’ that would call for a regular standard root present.37 It therefore seems conceivable that μῆστο reflects a characterized Narten root aorist formation *mēd-to.38 This approach would then also account for the preterite OIr. ro·mídar ‘judged’39 and maybe for the perfect (> preterite-present) Goth. ga-mōt ‘to find room, have permission’.40 Greek would then have generalized the aorist allomorph *mēd- in the (thematized) present stem,41 thus

36 Also proposed in LIV²: 423, as well as by Isebaert (1992: 201). 37 Cf. for this principle Meillet 1908: 84f.; Peters 1975: 41; Schindler 1975a: 62; Isebaert 1992: 194. 38 Cf. for this Peters 1980: 324. The lengthened grade would neither be original in a Narten imperfect middle nor in a Narten aorist middle. In any case, it is noteworthy that most of the attested verbal formations of the root go back to middle forms, so the middle may have, at an early stage, adopted the unmarked ablaut equivalents (in addition to μήδομαι and μέδομαι [active μέδω is only attested since Soph. and Emp.], cf. also OIr. midithir (see next note), OAv. masatā subj. med. 3rd sg. ‘will measure out’, YAv. vī-māδaiiaṇta opt. med. 3rd pl. ‘shall measure out’, Toch. B maistär ‘gages, estimates’ (cf. Malzahn 2010: 776–8) and Lat. medeor ‘to heal, relieve’; the latter differs significantly from Lat. madeō ‘to be full, drunk’ [from a different, homophonous root *√med ‘voll werden, satt werden’, LIV²: 423f.], for both of which LIV² assumes an essive formation

*med-h₁é-, whence medeor must have restored R(e) secondarily). 39 This preterite is quite peculiar anyway, since it behaves differently from all other *CeT-verb preterites. Seeing it as the continuant of a (Narten) root aorist would account for this curiosity. Other OIr. continuants of (standard) root aorists include ·cer ‘fell’, luid ‘went’, and ·lá ‘laid’ (cf. Schumacher 2004: 60f.). A different origin of ·mídar, viz. from the weak stem of an inherited perfect *me-md- that was (analogically) transformed to *mēd-, is proposed in Schumacher 2004: 74–76 and 481f. note (c), but the implied development seems rather ad hoc. The present midithir, ·midethar reflects *med-e/o- which developed apparently regularly from thematic *med-e/o- within (Proto)Irish; *med-e/o- is also required by Middle Welsh með- (not †meið-), cf. Schumacher 2004: 481 note (a). 40 LIV²: 423 projects *me-mōd- as a secondary perfect, analogically to the R(ē) of the Narten present. This account ignores, however, the fact that the verb is attested in Gothic as mitan ‘measure’ (< *med-e/o-) without any traces of a lengthened grade. Even if ga-mōt and mitan are no longer interpretable as belonging to the same root on a synchronic level and may therefore have developed independently from a relatively early stage, it seems more plausible to accept with Peters that the perfect formation in question was presumably derived from the aorist stem allomorph, rather than from the present, cf. Peters 1980: 97 and 324 (with further examples). 41 As Peters (1980: 28 sub a)) points out, this kind of leveling seems to have been more common – given the unmarked status of the Greek aorist – than a leveling in favor of the present stem allomorph. Cf. for example the pres. στόρνῡμι ‘I spread’ after aor. ἐστόρεσα (via metathesis from *stero[s]-; *√sterh₃).

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 Notes on three “acrostatic” neuter s-stems 305 resulting in μήδομαι, alongside regular μέδω and μέδομαι.42 Another welcome epiphenomenon of this theory is that it can explain why the latter is only attested in the present and imperfect, but never found in the aorist. 2.4 Of course, this explanation presupposes the existence of Narten root aorists, which is far from commonly accepted, despite some seemingly conclusive evidence.43 But even if one rejects a Narten aorist *mḗd-to (> μῆστο Hsch.) and plumps for a Narten present *mḗd-ti instead, it seems quite understandable how this led to a (Proto-Graeco-Armenian) verb *mḗd-e/o- that finally produced Gk. μήδομαι. It is also comprehensible that this verb caused an original verbal abstract *mĕd-es- (which independently developed to Umbr. meřs) to be remodeled to *mēd-es-, resulting in Arm. mit and Gk. μήδεα.

3 PIE *sēd-es-

The second s-stem of particular interest is PIE *sḗd-os. The short-vowel form *séd- os is the direct source of Ved. sádas- (RV+), Gk. ἕδος (Il.+)44 and ON setr, all of which have the meaning ‘seat, residence’, while OIr. síd ‘fairy mound; peace’ and ON sætr ‘a mountain pasture’ seem to go back to *sḗd-os. 3.1 Another possible continuant of the s-stem might lie in Umbr. sersi (VIa 5). The word appears in VIa 5 in the sequence sersi pirsi sesust, immediately before a relative clause introduced by the conjunction pirsi45 ‘when’, followed by the fut. perf. 3rd sg. sesust, probably ‘sederit’ (cf. Untermann 2000: 680f.), thus suggesting a meaning ‘in sede cum sederit, i. e. when he (the augur) has seated himself on the seat’ (Buck 1904: 263). According to the communis opinio,46 the word has

42 Of course, also this form is not regular. The expected stem allomorph of the root present middle *m̥d- must have been replaced by *med- from the singular active, maybe in order to prevent an odd allomorphy *med-:*m̥d->*med-:*ad- (?), or euphonically to avoid difficult-to-pronounce zero grades (?). 43 Cf. Tremblay 2005 for an overview (with literature). 44 The word might also be attested in Mycenaean Greek as o-pi-e-de-i if this is to be read as prep. *opi + dat. sg. *hedehi ‘at the seat, residence’, referring to the temple or sanctuary of a deity. Cf. DMic: 2, 39 with lit. 45 In the Umbrian alphabet found as peře (IIa 3). The various spellings in the Latin alphabet (persi, persei, perse, pirsi, pirse, all on VIa and VIb) partly seem to be the result of a rhyming connection to the preceding or the following word, cf. persi mersi (VIa 38), persei mersei (VIa 28), pirsi mersi (VIa 48), or the discussed sersi pirsi (VIa 5) itself, cf. Untermann 2000: 521f. For its various semantics and uses cf. also Weiss 2010: 61 note 113. 46 Cf. Untermann 2000: 658f. also for other, less convincing interpretations.

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 306 Stefan Höfler to be read as /seři/ and reflects the abl. or loc. sg. of an i-stem *sedi-. However, an i-stem of this kind from this root would be unique within the IE languages,47 all the more since the alleged comparandum Lat. sēdēs need not continue an i-stem formation (see below). In the Latin alphabet, the spelling 〈rs〉 is not exclusively used for designat- ing /ř/, but also for the sound sequences /řs/ and /rs/ proper. For our matters, this means that 〈rs〉 might also stand for two distinct sounds and not only one phoneme. An interpretation as 〈sersi〉 = /seřsi/48 or /sersi/49 permits the analysis as the expected outcome of a presupposed s-stem loc. sg. *sed-es-i, the obvious advantage of which being that Umbr. sersi, then, would no longer be an isolated formation, but would formally align with the well-attested group of Ved. sádas-, Gk. ἕδος and ON setr, all of which show a parallel meaning50 ‘seat, residence’.

47 The existence of the secondary s-stems YAv. hadiš- ‘Gottheit des Wohnsitzes’ and OPers. hadiš- ‘Wohnsitz, Palast’ (cf. Stüber 2002: 143) does not necessarily presuppose the erstwhile presence of an i-stem *séd-i-, but can be regarded as cognate to Ved. sádhiṣ- ‘Sitz, Stätte’ (< *sed-h₂-s-?; cf. EWAia: 2, 694). 48 This reading is not only suggested by the spelling mers (VIb 31, 55 [twice]), which appears as meřs (Ib 18 [twice]) in the Umbrian alphabet, but also by the forms mersei (VIa 28) and mersi (VIa 38, 48), which are best analyzed as juxtapositions of 〈mers〉 (viz. /meřs/) with the pres. subj. 3rd sg. si of the copula (viz. /meřs/+/si/ > /meřsi/). Incidentally, all the above-mentioned examples appear in the same tablet as sersi, and thus permit a reading /seřsi/. 49 There seems to be a derivative of the s-stem meřs that indicates a phonological development different from the one just assumed. The outcomes of an alleged form *medes-u̯o- (nom. sg. m. mersus (III 6), abl. sg. f. mersuva (III 11) and acc. pl. n. mersuva (III 28), all of which have 〈rs〉 for /rs/) suggest a dissimilation of ř +*z to rs (cf. Meiser 1986: 174f., 184f.; also Weiss 2010: 99f note 4). Unfortunately, there are no attestations of case forms of (regular) neuter s-stems in Umbrian other than the nom. sg. meřs (for tuder cf. immediately below; for Umbr. erus [secondary s-stem, only acc. sg.] cf. Weiss 2009b) that would be able to clarify whether this phonological development was indeed realized within the paradigm of neuter s-stems, thus resulting in a somewhat peculiar stem-alternating paradigm nom. sg. meřs : gen. sg. *merser, or if ř was generalized throughout the paradigm by analogical leveling (gen. sg. *meřser). In fact, the other attested s-stem tuder exhibits paradigmatic leveling in another direction (generalization of the oblique -er- also in the nom.-acc. sg., cf. Meiser 1986: 231–8 and above 2.1), which could, in theory, support the assumption that a leveling in either direction is possible, and may even be expected in Umbrian. This, then, would have led to a generalization of the stem variant of the nom.-acc. sg. meřs-, and similarly seřs-, thus again giving preference to the reading /seřsi/. 50 The concrete meaning ‘seat, chair, saddle etc.’ that is required by Umbr. sersi is also paralleled in Vedic and Greek:

RV 5.61.2 kvà vó ’śvāḥ kvā‘ bhśavaḥ katháṃ śeka kath yaya pṛṣṭhé sádo nasór yámaḥ ‘Wo sind eure Rosse, wo die Zügel? Wie habt ihr das vermocht, wie seid ihr gekommen? (Wo ist) der Sattel auf dem Rücken, der Zaum in den Nüstern (der Rosse)?’ (Stüber 2002: 143)

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 Notes on three “acrostatic” neuter s-stems 307

There is, however, a fundamental drawback to this analysis. The evidence of a locative (or ablative51) ending -i of consonantal stems in Umbrian is scarce.52 One would expect the ending ‑e < *-i,53 as in loc.-abl. sg. vapeře ‘stone (seat)’ (III 7) or kapiře/capirse ‘cup/bowl with handle used mainly for ritual purposes’54 (Ia 34, 41/VIb 24, 37).55 The ending -i (< *-īd), in turn, marks the regular ablative of Umbrian i-stems,56 which has led to the already mentioned analysis of sersi as the abl. sg. of an i-stem *sed-i-. In that case, the word could be identified with Lat. sēdēs, gen. sg. sēdis f. ‘seat, residence’ which shows a peculiar lengthened root vowel. Since the vowel /ẹ̄/ < PIE *ē is not always graphically distinguished from /e/ in Umbrian (see above 2.1), 〈sersi〉 could possibly stand for /sẹ̄ři/ as well.57 But the existence of an Italic i-stem *sēdi- is not conclusively imposed by the Latin word either. The three dissenting votes are the nom. sg. in -ēs,58 the gen. pl. sē-

Il. 9.193 ... ταφὼν δ’ ἀνόρουσεν Ἀχιλλεὺς αὐτῇ σὺν φόρμιγγι λιπὼν ἕδος ἔνϑα ϑάασσεν. ‘Erstaunt erhob sich Achilleus mitsamt der Leier und verliess den Sitz, wo er gesessen hatte’ (Stüber 2002: 144)

51 For the locative uses of the ablative in Umbrian cf. Buck 1904: 203f. The Umbrian abl. sg. of consonant stems seems to go back to the loc. sg. anyway (as opposed to Oscan, where we find the ending of o-stems), cf. Buck 1904: 125; Weiss 1993: 43. 52 There is one example of a consonant stem with a loc. sg. in -i: Umbr. scalsie ‘a kind of vessel’ (VIb 5, VIIa 37; loc. sg. scalsi+ enclitic *-en), where the original -i was presumably retained before the enclitic, cf. Buck 1904: 126. For the abl. sg. peři, persi see below in the text. 53 Cf. Meiser 1986: 113f., who casts some doubt on this sound law’s validity. 54 Cf. Weiss 2010: 342f. for an interpretation of its ritual purpose. 55 Cf. Untermann 2000: 825f. and 367f. 56 The locative of i-stems also has the ending -e, cf. loc. sg. ocre ‘mount, stronghold’ (VIa 26, 36, VIb 29), cf. Untermann 2000: 791f. 57 Cf. also Klingenschmitt 1992: 115. 58 Of course, this is the regular nom. sg. ending of hysterokinetic i-stems in Latin (cf. Klingen- schmitt 1992: 114; Schaffner 2001: 435; Weiss 2009a: 242–4), but as such one would expect a zero grade in the root (cf. Lat. fidēs ‘faith, trust’ < *bʰidʰ-ē ()[+s]:fīdō ‘I trust’ < *bʰedʰ-e/o-; Lat. clādēs ‘calamity’ < *kl̥h₂d-:per-cellō ‘I smite’ < *kelh₂d-), or at least a secondarily introduced full grade (cf. Lat. com-pāgēs ‘binding, framework’ < *peh₂ǵ-:pangō ‘I fix’; Lat. con-tāgēs ‘touch’ < *teh₂g-:tangō ‘I touch’), but not a lengthened grade. If one, therefore, supposes that sēdēs is not an original hysterokinetic formation, but was generated after a productive pattern as a feminine verbal abstract, one would then expect †sedēs (after sedeō, sedēre ‘to sit’) as an outcome, since these abstracts almost exclusively correspond in their root vocalism to the associated present stem (cf. Lat. caedēs ‘slaughter’ : caedō ‘I slaughter’; Lat. lābēs ‘disaster’ : lābor ‘I fall’ etc.). The assumption that the verbal abstract was derived from a secondary root variant *sēd- (as per Klingenschmitt 1992: 117; the evidence of which is limited to Celtic causative formations with *ō, viz. OIr. sáidid ‘thrusts, fixes’ and Middle Welsh gwahawd ‘to invite’) is hardly disprovable, yet

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 308 Stefan Höfler dum (Cic., Liv.),59 and of course the lengthened grade of the root. Because of these irregularities, it has been proposed that sēdēs should be regarded as a remodeled root noun.60 This seems to be an attractive solution, since it could explain the inflectional behavior61 and also the vowel length.62 The starting point would be a root noun *sēd-s, gen. sg. *sĕd-e/os, whence with leveled root ablaut in favor of the strong stem *sēd-s :*sēd-e/os, resulting in Latin †sēs(s)63 : sēdis. Why the nom. sg., then, was transformed to sēdēs, is an open question.64 But it may, in any case, be noted that such a remodeling is not exactly unique within Latin. It can be paralleled by the root nouns nūbs f. ‘cloud’ (Liv. Andron.) and saeps f. ‘hedge, fence’

unlikely. Another possibility is, however, that the verbal abstract was somehow built on the stem of the synchronic perfect sēdī (of whatever origin it may be), a suggestion that has also been made for above-mentioned com-pāgēs, con-tāgēs, and for rūpēs ‘cliff, crag’ (after *pāgī [only pēgī], *tāgī, rūpī?), and also for amb-āgēs ‘detour, meanderings’ (after *āgī [only ēgī]?; cf. for these examples Peters 1977: 68) for which the explanation given above (secondarily introduced full grade would have led to †amb-agēs) is not possible. But nevertheless, a secondary remodeling of †amb-agēs to amb-āgēs after com-pāgēs, con-tāgēs, pro-pāgēs ‘a stock’, etc. cannot be excluded, so sēdēs would remain the only significant example for this derivational process, which additionally also yields some semantic difficulties. 59 This gen. pl. appears beside the expected sēdium. As per Ernout 1965: 17; Benedetti 1988: 149 note 578; pace Klingenschmitt 1992: 116f., the former seems to be the older one. 60 Cf. Benedetti 1988: 149f.; Tremblay 2010: 204; and NIL: 593f. note 2 for a summary of the different other assumptions (with lit.). 61 Cf. for example the gen. pl. pĕdum of the root noun pēs ‘foot’. 62 One must, of course, concede that PIE had root nouns with an acrostatic R(ḗ):R(é) ablaut, for which the comparative evidence is not exactly overwhelming (cf. Schindler 1972b: 37; Schindler 1994: 399; Scarlata 1999: 759 with lit.; Tremblay 2010: passim with a collection of possible examples). Within Latin, the supporting evidence includes rēx, rēgis m. ‘king’ (cf. OIr. rí, rig, Ved. rj-), lēx, lēgis f. ‘law’ (*√leǵ ‘sammeln, auflesen’ [LIV²: 397]; cf. Marrucinian lixs [nom. sg.] and Oscan ligud [abl. sg.], for which cf. Untermann 2000: 434f.), maybe spēs, spēī f. ‘hope’ (if from *spḗh₂-s [Eichner’s law!] with *h₂ because of Ved. sphāyātai ‘soll fett werden’ etc. (pace LIV²: 584 *√spʰeh₁); cf. Weiss 1993: 25–7), and less convincing ēr, ēris m. ‘hedgehog’ (cf. Gk. χήρ Hsch.; if from *√ǵʰers ‘sich sträuben, erstarren’ [LIV²: 178], with ēr for *hēr as in ānser for *hānser), and finally rēnēs m. pl. ‘kidneys’ (if with Lith. strnos f. pl. ‘loins’ from *srḗn-; cf. Mastrelli 1979). Taken together, the assumption of an ē : e root noun *sḗd-s does at least not seem illusionary. 63 For *-sed- as a second compound member cf. Lat. dēses ‘idle’, praeses ‘guardian’, reses ‘listless, torpid’, subses ‘qui subtus sedet’ and obses ‘hostage’, cf. Benedetti 1988: 149–55; and OIr. arae, gen. arad ‘driver of a chariot’, if < *pr̥h₂ised-s,*pr̥h₂ised-os ‘sitting next (to the warrior)’; cf. Stifter 2006: 161. For the Vedic material cf. Scarlata 1999: 560ff. 64 Cf. e. g. also Untermann 1992: 146.

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 Notes on three “acrostatic” neuter s-stems 309

(Cic., Varro) which, in Classical times, occur as nūbēs and saepēs respectively, and maybe also by trabēs (Enn.) instead of the usual trabs ‘tree-trunk, beam’.65 Another possible continuant of a root noun *sēd-s is found in Lepontic. In the inscription of Prestino (CO·48) the form siteś appears as the apparent accusative object of the verb tetu ‘gave, dedicated’. It was taken as the acc. sg. of a neuter s-stem *sēd-es by Prosdocimi (1976: 214f.) but there are several serious objections to this assumption (cf. for these Uhlich 1999: 294f.). Therefore, it has been argued and is now widely accepted that siteś has the meaning ‘seats’ and reflects the acc. pl. of a root noun (viz. *sēd-n̥s).66 However it may be, an explanation based on an inner-Italic equation is, in principle, preferable to an attempt at interpreting the Umbrian word sersi as an s-stem with regard to outer-Italic parallels; all the more so since the latter option contains the pivotal problem that -i should not surface as the ending of an abl.-loc. sg. of a consonant stem, a difficulty that it shares with the analysis of sersi as a root noun, which, as has just been shown, is the most plausible origin of Lat. sēdēs and Lep. siteś. It is possible, yet unprovable, that the expected loc. sg. *serse was remodeled to sersi in order to avoid homophony with the participle serse (< *sedens), that itself appears in the same tablet three lines above and eleven lines below sersi, or due to rhyming purposes based on the following conjunction pirsi, which itself shows this particular tendency (see note 45 above), or simply by substituting the (too ambiguous?) ending -e by the more iconic desinence -i, which was used as the ablative ending of i- and u-stems. This is also a possible explanation for the abl. sg. peři (Ia 29, 32), persi (VIb 24, 37–39) ‘foot’,67 which should actually surface as †peře.68 Since this word continues a root noun as well, it seems fairly justified to assume that Umbr. sersi indeed reflects the abl. sg. of a root noun *sēd-s with matches in Lat. sēdēs and Lep. siteś. 3.2 The explanation as a root noun obviously does not make sense for OIr. síd ‘fairy mound’ and ON sætr ‘a mountain pasture’, which both seem to go back to a proper s-stem, as if < *sēd-os and *sēd-es- respectively.

65 A root noun *trēb-s might be suggested by Osc. trííbúm acc. sg. ‘house’ < *trēb-m̥ cf. Klingen- schmitt 1992: 117; de Vaan 2008: 626; ablehnend Weiss 1993: 75ff. 66 Initially Lejeune 1971a: 194f.; cf. also Uhlich 1999: 293–8 (with a full discussion of the form); Griffith 2005: 53f and 61–3 (for a plausible phonological development of *-n̥s to Lep. -eś). 67 Another explanation would be that there was an influence of the u-stem abl. sg. mani ‘hand’, cf. Klingenschmitt 1992: 111; Weiss 1993: 44. 68 Cf. Meiser 1986: 114 for another, less convincing explanation (viz. as an old instr. sg. *pedē).

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 310 Stefan Höfler

Wagner (1969: 246 note 107) suggested that the long-vowel forms OIr. síd and ON sætr69 must be explained as a vṛddhi-derivative *sēdos (sic!) of the s-stem *sedos ‘seat’, the original meaning of which should have been ‘belonging to, being near a (human) settlement (*sedos)’. This interpretation is, at first glance, quite promising, as it offers a comprehensible explanation for the semantics: In Irish folk belief, as Wagner points out, the dwellers of these fairy mounds, the síde (nom. pl.), were believed to reside in the immediate vicinity of human settlements, on higher ground, in elf-mounds and ancient tumuli or burying places. He adds that the meaning of ON sætr is likewise understandable, since mountain pastures usually belonged to the whole village community, the parallelism in form and meaning between síd and sætr therefore being obvious. However, Darms (1978: 67–74), in his book on vṛddhi-derivation in Germanic, raises some justified objections against Wagner’s supposition, especially in view of OIr. síd, for which such an analysis is morphologically impossible, since vṛddhi- derivatives inflect thematically (see below 3.3). After a thorough discussion of the material, Darms tries to explain ON setr and sætr as the result of a paradigmatic split of an ablauting *sēd-os,*sĕd-es-, with reference to Schindler 1975c. He finds support for this theory in Swiss German sess n. (< *setez- or *seta-), also signifying ‘a mountain pasture, alp’, which, to him, proves that this meaning can also have developed in primary formations of the root without the detour of a vṛddhi-derivative. Despite this verdict, however, we may be inclined to believe that the interpretation of sætr as a vṛddhi-derivative is the far more plausible solution after all, since not only formally, but also semantically, as Darms indeed has to admit, it makes perfectly sense. For the base form setr, the meaning ‘seat, settlement, farmyard’ is well-attested. The alleged meaning of the derivative ‘belonging to, being near the seat, settlement, farmyard’ fits into the picture well, since for sætr, Darms determines the meaning ‘a mountain pasture, summer pasture, alp; chalet’, which implies a viable semantic development.70 On the formal side, it is noteworthy that basically all inherited s-stems were thematized in North Germanic and are synchronically inflected as neuter a-stems (e. g. nom.-acc. sg. setr, gen. sg. setrs).71 In this light, ON setr regularly goes back

69 He also included Swiss German Sāss, which is found in many names of alpine pastures, but cf. Darms 1978: 71f. 70 A possible equivalent may be found in Upper German Maiensäß n. (only marginally) ‘unterste Stufe einer Alm’, to which the cattle are driven in May, and Swiss German Säss n., which are both put in reference to ON sætr in Kluge & Seebold 2002: 24, 591, where a vṛddhi-derivative is the preferred explanation as well. 71 Cf. Casaretto 2004: 555 and note 1813.

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 Notes on three “acrostatic” neuter s-stems 311

via *set-iR-a-<*set-iz-a-(vel sim.) to a thematized *sĕd-es-o-, and likewise, an alleged vṛddhi-derivative *sēd-es-o- leads via *sēt-iz-a->*sāt-iR-a- with umlaut lautgesetzlich72 to ON sætr. Beyond this, it is, in my opinion, improbable that an ablauting paradigm would have survived long enough to produce some sort of paradigmatic split whose individual continuants happen to have survived as a pair exclusively in Old Norse. Additionally, there are parallel cases of vṛddhi-derivatives being used in the field of topographical terms in Germanic,73 which makes this analysis all the more preferable. And finally, another vṛddhi-derivative of an s-stem base might be found in Old Norse, supporting the formal analysis outlined above. The neuter fær ‘lamb, sheep’ is traditionally connected with Gk. πόκος m. ‘fleece’ and is thought to go back to PGmc. *fahaz (thus IEW: 797). But neither the gender nor the semantics ad- vise such an interpretation. On the other hand, a connection to a homophonous s-stem *fahaz has been proposed74 to account for ON fax n. ‘mane’ (as if75 < *fahsa-), ignoring, however, that such an s-stem (as if *póḱ-os) is very unlikely to have ever existed. Considering Gk. πέκος n. ‘fleece’ (only marginally) and Lat.

72 Note that the raising of *e to *i in non-first syllables and the development *ē >*ā predate the i-umlaut. This process, then, affects /a, ā, ō, u, ū/ and u-diphthongs, but not e (cf. Krahe & Meid 1967–1969: 1, 59; pace Darms 1978: 72 (ON hatr ‘hate’ without umlaut might have retained its root vowel analogically after the verb hata), who is, however, right when he admits that “Die Umlautsbedingungen im An. sind aber nicht so klar, daß sie ein i oder j der Folgesilbe auch dann erzwingen können, wenn dieses sonst nicht begründet werden kann.”). 73 Cf. PGmc. *mari-:*mōra- (in OHG mari, meri ‘sea’, OEngl. mere ‘sea, lake’ etc. : OEngl. mōr ‘moor, marsh’, Germ. Moor ‘id.’ etc., cf. Darms 1978: 158–66), PGmc. *dala-:*dōli- (in OEngl. dæl ‘valley’, OIcl. dalr ‘id.’, Germ. Tal ‘id.’ etc. : OIcl. dœll ‘valley dweller’ < ‘belonging to the valley’, cf. Darms 1978: 208–18). 74 Thus de Vries 1961: 149 and 114; Magnússon 1989: 221 and 167. 75 Admittedly, the new etymology of fær outlined here cannot account for fax either. The word appears also in OHG (fahs ‘shock of hair’) and OEngl. (feax ‘id.’). IEW: 797 invokes < *-po ḱ-s-o- with dubious o-grade. It is wise to separate fax from fær, at least from a synchronic, inner-Germanic point of view. It might be somehow connected to the stem of Ved. pákṣ-man- n. ‘eyelashes’, YAv. pašna- ‘id.’ (of whatever origin; cf. EWAia: 2, 62f.). Alternatively, one could hypothesize a PIE derivative *poḱ-s-o- with a peculiar structure R(o)-S(ø)-o- that would be to *peḱ-es-as*h₂omǵʰ-s-o- (Toch. A eṃts,Bentse m. ‘Gier, Neid’) is to *h₂emǵʰ-es- (Ved. áṁhas- n. ‘Bedrängnis, Not’, YAv. ązah- n. ‘Bedrängung, Enge’, ON angr n. (m.) ‘Verdruss, Betrübnis’), or as *tomH-s-eh₂- (Lith. tamsà ‘darkness’) is to *temH-es- (Ved. támas- ‘id.’ etc.), but for now this remains speculation (cf. Peters apud Adams 1985: 12 note 21; Hilmarsson 1987: 72).

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 312 Stefan Höfler pecus,-oris n. ‘sheep, livestock’,76 and in view of the ordinary development of neuter s-stems in Germanic,77 the Proto-Germanic equivalent should have been *feh-iz-.78 An alleged vṛddhi-derivative of this word would then have led to *fēh- iz-a-79 > *fāh-iR-a-(vel sim.) > fær, parallel to *sēt-iz-a->*sāt-iR-a-(vel sim.) > sætr. On the semantic side, presupposing a meaning ‘sheep’ for the base *feh-iz-, the semantics of *fēh-iz-a- would have been ‘belonging to the sheep (= ewe)’ > ‘lamb’ or ‘belonging to the sheep (= flock of sheep)’ > ‘(one single) sheep’. Coin- cidentally, there are various similar examples of vṛddhi-derivatives in the field of (domestic) animal names in Germanic,80 which adds to the likelihood of this new etymology.81 3.3 This interpretation, however, does not solve the problem of OIr. síd ‘fairy mound’, which, as Darms points out, cannot continue a vṛddhi-derivative *sēdos (as suggested by Wagner). Vṛddhi-derivatives appear almost exclusively as thematic stems or, to a far lesser extent, as i-stems, but never as s-stems. A vṛddhi- derivative to an s-stem *sĕd-os should have yielded *sēd-es-o-82 (or perhaps *sēd- s-o-), which would then have led to OIr. †síde.83 But for all that, síd is inflected as an s-stem in Old Irish. Unless one admits that the word was secondarily trans-

76 Even if the original semantics of the s-stem might have been a verbal noun ‘Rupfung’ (hence Greek ‘fleece’; cf. LIV²: 467 *√peḱ ‘[Wolle oder Haare] rupfen, zausen’), it is fairly safe to project a meaning ‘sheep, livestock’ (< ‘what is being plucked’) for PIE *peḱ-os (thus also Stüber 2002: 135). 77 Cf. *(h₁)régu̯-os > PGmc. *rekʷ-iz- thematized as Goth. riqis ‘darkness’, ON røk(k)r ‘id.’ (with labial umlaut of *e before *kʷ). 78 The regular outcome of *feh-iz-(a-) in Old Norse would probably have been †fér. One might suggest that the word itself was replaced by the synonymous u-stem ON fé n. ‘cattle, sheep’ (from *peḱ-u-; cf. Goth. faihu, OHG fihu, Lat. pecū, Ved. páśu-, etc. ‘cattle, livestock’) and the alleged vṛddhi-derivative fær ‘lamb, sheep’ respectively. 79 A long-vowel s-stem *fēh-iz was already proposed by Schmidt (1889: 148f.), but of course he did not envisage a vṛddhi-derivative. Needless to say that the same objections can be made against the originality of an s-stem *fēh-iz as outlined above in the introduction 1.1. 80 Cf. PGmc. *han-en- ‘rooster’ : *hōn-n-a- n. ‘chicken’ (in Germ. Hahn : Germ. Huhn etc.; cf. Darms 1978: 122–33) and others (cf. Darms 1978: 134–42). 81 There is, however, a major blemish in this analysis: OSwed. fār n. ‘sheep’, Swed. får n. ‘id.’ etc. do not show any sign of i-umlaut, suggesting again a pre-form *fahaz- and implying that ON fær reflects affection of R-umlaut. Since the cognates of ON sætr regularly appear with i-umlaut (ModIcel. sætrur ‘summer grazing’, Norw. sæter, Swed. säter; cf. de Vries 1961: 576), one would have to assume that the intervocalic *h somehow had an umlaut-inhibiting effect on the preceding vowel before its loss and subsequent contraction to defend the proposed etymology. Since the phonological processes involved are not at all clear to me, this has to remain an open question. 82 Cf. Debrunner 1954: 142f. 83 Cf. gen. sg. nime ‘of the sky, heaven’ < *nem-es-os.

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 Notes on three “acrostatic” neuter s-stems 313 ferred to this stem class (for which there are only a few parallels),84 the interpretation as a vṛddhi-derivative is problematic, both on phonological and morphological grounds. OIr. síd therefore seems to be the regular continuant of a long-vowel formation *sēdos. Semantically, the problem is aggravated by the formally identical word OIr. síd ‘peace’. Most probably, the word belongs to the same root because of its Welsh counterpart hedd ‘id.’, which allegedly goes back to the short-vowel form *sĕ- dos.85 Darms, therefore, suggests an ablauting paradigm *sēd-os,*sĕd-es-, with reference to Schindler 1975c, and asserts that Irish and Welsh would individually have generalized the strong and the weak stem. In Irish, the meaning would have specialized from ‘seat, residence’ to ‘seat, residence of fairies’. The development to the second meaning of ‘peace’, shared by both languages, is left open.86 Stüber (2002: 144f.) objects to the existence of an ablauting paradigm *sēd-os, *sĕd-es- within Insular Celtic,87 since this would be a unique case of preserved root ablaut of a suffixal stem. She therefore favors a secondary origin of the Welsh vocalism (but see note 85), while she regards OIr. síd as the regular continuant of an acrostatic s-stem *sḗd-os. Following the premises of this paper, one would, however, rather assume the Welsh hedd to be the regular continuant of the short-vowel s-stem *sedos, and OIr. síd to be the remodeled form, probably in analogy to associated verbal forms. This is the strategy deployed by Meissner (2006: 75), who suggests an analogical influence of the verb saidid ‘sits’ and its suppletive preterite síasair, from which the stem síad- would have been abstracted, which could then easily have influ-

84 Cf. de Bernardo Stempel 1999: 149–51 for a small number of examples. 85 It is unclear whether Welsh sedd ‘seat’ also goes back to *sedos and was secondarily separated from hedd on a formal level by generalizing the different anlaut variants s- and h-, or if it continues a different formation, cf. Stüber 2002: 144. She also takes into consideration a remodeling in analogy to verbal forms like eisteddaf ‘I sit’, which is, however, problematic, since this, as Schumacher (2000: 218) has shown, goes back to a compound verbal noun *eχs-sodiā (> eistedd), whereas *sed-e/o- is not attested in Welsh; cf. also Schumacher 2004: 562 (d). 86 Stüber (2002: 144) proposes a development ‘worüber man (zu Rate) sitzt’ → ‘Friede(nsabkom- men)’ and compares Engl. settlement meaning ‘colony, village’ and ‘resolution, agreement’. 87 It has yet to be clarified whether the Gaulish toponyms Mello-sedum and Viro-sidum (cf. Matasović 2009: 326 with lit.) can possibly serve as evidence for the co-existence of the two stem variants *sed- and *sīd-. It is, in any case, clear that °-sedum and °-sidum would not have to be in immediate relation to an s-stem, but could just as well point to a thematic stem or a root noun (for which see below), even though original s-stems apparently do come up as thematic second compound members in Gaulish place names, cf. °-dunum and °-δουνον besides s-stem OIr. dún ‘fort, rampart’ (cf. Dottin 1985: 115).

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 314 Stefan Höfler enced the noun. There are several necessary objections88 to this theory, the first one being that the connection between the meanings ‘fairy mound’ and ‘peace’, and ‘to sit’ is not obvious enough to encourage an analogical remodeling of this sort. Since the word is isolated within Old Irish both semantically and formally, I see no reasonable chance how it could have obtained its long vowel as the result of an analogical remodeling. But if one assumes some sort of analogy, this alleged remodeling would have had to have taken place at a time when, at a synchronical stage, there were still long-vowel verbal forms, e. g. from a Narten present, representing one of the expected characterized present stem formations to the punctual root *√sed ‘to sit down’. This Narten present is, however, only doubtfully attested by the not un- ambiguous present OLith. sdmi and the Vedic participle sādád- (as if < *sēd-n̥t-), a hapax in the compound sādád-yoni- (RV 5.43.12).89 And finally, the comparison with an entirely different s-stem *sīd-os,90 which is reconstructed for Lat. sīdus,-eris, may seem possible on phonological grounds but is not convincing on the semantic side, since the meanings ‘fairy mound, peace’ on the one hand and ‘constellation, star’91 on the other are rather difficult to reconcile. The word, therefore, seems to persistently hint at either an “acrostatic” s-stem or an s-stem of a Narten root. But both of these options should rather be dismissed, the former one due to the objections already made above,92 and the latter one because there are good counter-examples to this assumption, e. g. the zero grades in the old reduplicated present Ved. sdati, Gk. ἵζω, Lat. sīdō, and derivatives like PIE *ni-sd-o- in Lat. nīdus, Ved. nīḍá-, Germ. Nest, OIr. net etc.93 The remaining option, therefore, is to compare OIr. síd with Lat. sēdēs, Umbr. sersi and Lep. siteś, and somehow trace it back to a root noun. Admittedly, this is

88 Cf. also Stüber 2007: 40, who additionally remarks that under these conditions, the s-stem would have had to be remodeled to †síad, not síd. 89 The compound can be regarded as a nonce-formation and perhaps owes its long vowel to the preceding word sādayadhvam, cf. Lubotsky apud Pronk 2012: 240. Nikolaev (2008: 554 note 31) is also skeptical about its originality. 90 Proposed by Thurneysen 1887: 153f. 91 For Lat. sīdus, whose prehistory is somewhat opaque, cf. Stüber 2002: 181f. 92 A paradigm like nom.-acc. sg. *sḗd-s : gen. sg. *séd-s-s is very unlikely to have ever existed, but if it did, it seems quite plausible that it would have been conceived as a root noun and consequently merged with the alleged feminine *sḗd-s :*séd-os. 93 Cf. most recently Pronk 2012: 240f. As far as long-vocalic formations such as *sōd-o- (Engl. soot) etc. are concerned, I am afraid to admit that I have as yet no satisfactory explanation for these.

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 Notes on three “acrostatic” neuter s-stems 315 not the most elegant solution, but in view of the alleged inner-Celtic parallel, its likelihood might increase a little. The regular outcome of an already leveled root noun *sḗd-s : gen. sg. *sḗd-o s might have been †sí :†síd (parallel to rí : ríg m. ‘king’ < *(h₃)rḗg-s : *(h₃)rḗg-os), while the regular standard s-stem *séd-os :*séd-es-os would have led to †sed :†side. It now appears feasible to assume that these two words merged into one paradigm at some point within Proto-Irish as some instance of étymologie croi- sée.94 One could hypothesize that the possible Scharnierform was the dat. sg. in phrases such as ‘in (the) seat’ and ‘in peace’, which would have produced †í síd for the root noun and †í sid for the s-stem in (classical) Old Irish.95 Since the two forms differed only in vowel length, it probably would not have been too unreasonable to confound them and eventually fuse them into one lexeme. This bold assumption would, then, also be able to explain the two very different meanings ‘fairy mound’ and ‘peace’: One could suppose that the root noun carried the semantics ‘seat, residence’ (thus still Lep. siteś?) > ‘seat, residence of fairies’ > ‘fairy mound’, whereas the s-stem had allegedly developed the specialized meaning ‘peace’ already in common (insular) Celtic times, whence also Welsh hedd ‘id.’ < *sĕd-os. This account may seem quite arbitrary at first, but after a thorough look through the attested Old Irish s-stems one will note that as a category, they are a rather heterogeneous group.96 Beside a few inherited words with parallels in other IE languages, there are a number of s-stems that can be traced back to PIE roots, but without s-stem parallels elsewhere, and also quite a few neuters without any etymological links at all, suggesting that the two latter groups re- ceived their s-stem inflection only in Celtic or Irish times. But more interestingly, there might be one or two97 instances of étymologies croisées within the squad of

94 Similarly Schrijver 1991: 376. 95 Their Proto-Irish pre-forms might have been something like *sīδi and *seδih (cf. McCone 1996: 100; Stifter 2006: 177 and 148) whence probably *sīδə and *siδə and finally †síd and †sid. 96 Cf. de Bernardo Stempel 1999: 140–54; Höfler 2012: 84–96. 97 A third possible example might be OIr. tír ‘land, earth’ (Welsh, Corn., Bret. tir ‘id.’) from alleged PCelt. *tīros <*tēros, seemingly another long-vowel s-stem. It is usually etymologically linked to the root *√ters ‘vertrocknen; durstig werden’ (LIV²: 637f.), so the expected s-stem should have been *ters-os. Etymological and semantic parallels can be found in Lat. terra f. ‘land, earth’ (*ters-eh₂-) and Osc. terúm n. ‘area (of a temple)’ (*ters-o-), and traces of the s-stem might be present in Lat. terrēnus ‘earthly’ (as if < *ters-es-no-) and terrestris ‘terrestrial’. Accordingly, one possible way to account for the long vowel in tír is to assume a cross between an original s-stem *ters-os >†terr and a root noun *ters(-s) (which might have led to *tēr via regular sound development already in PIE if *ph₂tḗr is correctly analyzed as **ph₂tér-s etc.) > OIr. †tír. This, however, remains pure speculation, since such a root noun is nowhere attested.

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 316 Stefan Höfler s-stem nouns that could perhaps support our audacious assumption of *sḗd-s × *séd-es-→*sḗd-es- (OIr. síd). The first example is the s-stem ond (gen. sg. uinde, uinne) ‘stone’ which might owe its peculiar o-vocalism to an analogical influence of, or a merger with a thematic noun that regularly had an o-grade in the root, just as it is proposed for Lat. pondus n. ‘weight’ after *pondus m. (see above note 28), which might be etymologically identical with it (as if from *pend-os ‘heaviness’).98 We could therefore project a cross between *pénd-es-×*pónd-o-→ *pónd-es- (OIr. ond). The second example is an even more obvious candidate, namely OIr. nem ‘sky, heaven’. It is recognizably connected to the more or less synonymous group of Hitt. nepiš, Ved. nábhas-, Av. nabah-, Gk. νέφος, OCS nebo etc. ‘cloud, sky’. These continuants can be traced back to PIE *nébʰ-os, the regular outcome of which, however, should have been OIr. †neb. The preferable explanation for the actual attested nem is to regard it as an étymologie croisée of two individual s-stems, *nébʰ-es- and *ném-es- (as in Lat. nemus ‘(sacred) grove, glade’, Gk. νέμος ‘id.’, Ved. námas- ‘worship, adoration’, Av. nəmah- ‘id.’99) of the root *√nem100 ‘to as- sign’, whose ritual connotation (cf. also Gaul. νεμετον and OIr. neimed ‘holy place, sanctuary’101) must have played a vital role in this process. 3.4 As we may now conclude, there seems to be no need to project a long-vowel s-stem *sḗd-os for PIE. ON sætr is morphologically and semantically best analyzable as an inner-Germanic vṛddhi-derivative *sēd-es-o-, whereas OIr. síd most likely represents a cross between the regular s-stem *séd-os, as in Ved. sádas-, Gk. ἕδος, ON setr and Welsh hedd, and the root noun *sḗd-s, continued most probably by Lat. sēdēs, Umbr. sersi and Lep. siteś.

4 PIE *h₁ēd-es-

The third “acrostatic” s-stem in this paper is *h₁ḗd-os, whose existence in PIE is not as evident. There are no immediate descendants of the s-stem noun in any Indo-European language. We shall, however, see that its existence in PIE times is suggested by different derivatives or remodelings and therefore very probable.

98 Cf. Matasović 2009: 137. 99 Schrijver (1995: 35) actually thinks that OIr. nem is the direct continuant of *ném-os, which is semantically unattractive without conceding an influence of *nébʰ-os. 100 *√nem ‘zuteilen’ LIV²: 453. 101 Stüber (2002: 131) proposes an interplay of assimilatory processes (lenited b lenited m) and the influence of OIr. neimed for OIr. nem. ∼

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 Notes on three “acrostatic” neuter s-stems 317

4.1 The first pair of words in this respect is Lith. desis/ėdesỹs (LDW: 1, 5163) ‘fodder, feed’ and Latv. ēdesis (LVV: 1, 573) ‘pig feed’, both of which are often analyzed as deverbal abstracts.102 However, it can easily be demonstrated that these are better explained as denominal derivatives and thus presuppose the existence of a neuter s-stem *h₁d-es- in Proto-Baltic. From a synchronic point of view, the suffix Lith. -esis (-esỹs)103 is used for de- riving abstract nouns (nomina actionis) from verbs.104 As the examples suggest, the suffix has become quite productive105 in Lithuanian, especially for verbs ex- pressing all different kinds of sounds and noises; but taken as a whole, derivatives of verbs from a great variety of different semantic fields can be found. On these grounds, Lith. desis/ėdesỹs can be interpreted as deverbal from Lith. sti, du (LDW: 1, 532) ‘eat, devour’, as it also denotes the process of ‘eating’ as a nomen actionis (cf. Bammesberger 1973: 82), from which the concrete meaning ‘fodder, feed’ might easily have developed.106 In Latvian, the parallel suffix -esis is far less common, but still found in a handful of words that can be analyzed as deverbal substantives, appearing as concrete nomina rei actae (see below for the examples). In this light, Latv. ēdesis ‘pig feed’ regularly corresponds to the verb ēst, ȩ̄du ‘eat’ as ‘what is eaten’ with subsequent semantic narrowing.107 From a diachronic perspective, it is generally accepted that the origin of the suffix should be sought in an *-io-derivative of an s-stem base (viz. *-es-io-).108 The few inherited PIE neuter s-stems in the Baltic languages109 show a simi-

102 Irslinger (2009: 217) however mentions Lith. desis as an example for inherited s-stems that were transferred to vocalic stem classes in Baltic and reconstructs an underlying PIE *h₁ēd-es-. Similarly also Casaretto 2004: 570 note 1887 and NIL: 210. 103 For the form reflecting métatonie douce cf. Derksen 1996: 149 and 158. The Latvian word does not exhibit metatony. 104 Beside these examples, only a few nouns without a verbal base are found, e. g. trobesỹs ‘building, house’ (: trobà ‘id.’), debesìs,-iẽs and debesỹs, dẽbesio ‘cloud’ (: PIE *nebʰ-os, cf. below) and nuogesỹs ‘nudity’ (: nuõgas ‘nude, bare’), cf. Bammesberger 1973: 84f. 105 Leskien 1891: 592–94 lists approx. 20 examples; Bammesberger 1973: 82–86 has over 50. 106 For this development cf. also Germ. das Essen, Fr. le manger. 107 LVV: 1, 577. Note that in Old Prussian, there are no traces of such a suffix. 108 Cf. Ambrazas 1994: 288. 109 For some other s-stems a conversion to the masculine stems in -as has been proposed, motivated by the homophonous nom. sg. in *-os (cf. Bammesberger 1973: 43f.). While I do not think that two of the proposed words can by any chance be reliable examples for this process (namely Lith. mẽlas ‘lie’ and mẽtas ‘year’), I do believe that Lith. mẽnas ‘art, skill’, and Lith. véidas ‘face; appearance’, Latv. veĩds ‘form, appearance’ could at least possibly continue the PIE s-stems *mén-os (cf. Ved. mánas- ‘mind, sense, understanding’ [RV+], Av. mánah- ‘id.’, OPers. manah-

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 318 Stefan Höfler lar development110: PIE *nébʰ-os111 is continued as an i-stem in Lith. debesìs112 ‘cloud’ and Latv. debess113 ‘sky, heaven’,114 PIE *h₂éu̯s-os115 as an i-stem in Lith. ausìs,-iẽs, f. ‘ear’, Latv. àuss f. ‘id.’ and OPruss. acc. pl. āusins ‘id.’,116 and PIE

‘thinking power’, Gk. μένος ‘mind, courage, anger’ [Il.+], cf. Stüber 2002: 128f.) and *u̯éd-os (cf. Ved. védas- ‘knowledge; property’ [RV+], YAv. vaēδah- ‘id. (?)’, Gk. εἶδος ‘form, shape, appearance, look’ [Il.+], cf. Stüber 2002: 166–9) respectively (thus also Petit 2010: 170). Indeed, I believe that one word can be added to these examples, namely Lith. pẽnas ‘food’ (PIE *pén-os; cf. Lat. penus, -oris ‘provisions’, and maybe Skt. panasá- m. ‘breadfruit tree’, if < *pen-es-ó-, but ablehnend EWAia: 3, 303f.), for which the analysis as an inherited s-stem, to my knowledge, has not yet been proposed. 110 This quasi derivational process did not implicate any semantic modification of the base (similarly also Lith. jentė, gen. sg. jenter̃s ‘husband’s brother’s wife’ < *Hénh₂ter- as opposed to Latv. ìetere ‘id.’ < *Hénh₂ter-eh₂-; cf. NIL: 204). The development is surely motivated by the gradual decline of both the genus neutrum and the consonant stem inflection. Apparently, many continuants of PIE consonant stems (i. e. athematic stems and root nouns) survived into the Baltic languages as (masculine or feminine) i- and io-stems. To name only a few parallel examples, regardless of their exact PIE reconstruction, one may consider Lith. obuolỹs and Latv. âbuolis ‘apple’ (as masculine io-stems), Lith. naktìs and Latv. nakts ‘night’ (as feminine i-stems), Lith. širdìs and Latv. siȓds ‘heart’ (as feminine i-stems), Latv. sā‘ ls ‘salt’ (as a feminine or masculine i-stem), Lith. sẽnis ‘old man’ (as a masculine io-stem), cf. Fraenkel 1936: 176f.; Stang 1966: 223. The question of whether they were really extended by the addition of an *-i- or *-io-suffix or simply merged into these paradigms due to mis- or reinterpretation of different case forms as possible Scharnierforms, need not concern us here. Therefore, I will continue to speak of it as a derivational process, even if this may not be unmitigatedly accurate. 111 Cf. Hitt. nepiš-, CLuw. tappaš- and HLuw. tipas- ‘sky’, Ved. nábhas- ‘mist; cloud; sky’, Av. nabah- ‘cloud’, Gr. νέφος ‘id.’, OCS nebo ‘sky, heaven’, air. nem ‘id.’ – The occurrence of anlauting d- instead of n- is not entirely clear. It could be due to a contamination with a semantically associated word. Pokorny thinks of Lith. dangùs ‘sky, heaven’, Fraenkel considers a noun related to Gk. δνόφος ‘Dunkelheit, Finsternis, dunkles Gewölk’ that otherwise left no traces in Baltic (cf. IEW: 315; LEW: 1, 85). Petit (2010: 29) compares debesìs for †nebesìs to Lith. devynì ‘nine’ (instead of †nevynì). For Hitt. nepiš- cf. also Höfler 2013. 112 Gen.-iẽs, m. (and dialectal f.), also: debesỹs, gen. dẽbesio m. (*-io-stem), LDW: 1, 421. For the geographical distribution of these and some other variants cf. ABL: 66–8 and 140f. 113 Gen. debess, f.; used predominantly in its plural form debesis, LVV: 1, 449f. 114 Both nouns still have a non-palatalized gen. pl. (Lith. debesų̃, Latv. dȩbȩsu) from the consonantal stem inflection. 115 Cf. OIr. áu, ó, OCS ucho (and Alb. vesh?) ‘ear’ – reconstructed according to Schindler 1975b: 264. However, the word has been subject to many discussions with regard to its stem formation, its inflectional type and the quality of the anlauting laryngeal. For a comprehensive overview of the different opinions cf. NIL: 339–43. 116 The Baltic forms (and independently Lat. auris) are most probably back-formations from the dual *h₂éu̯s-iH (with leveled root ablaut instead of *h₂us(-s)-iH), cf. Nussbaum 1986: 211 note 31.

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 Notes on three “acrostatic” neuter s-stems 319

*púH-os117 as an *-io-stem in Lith. puvsis118 ‘rotten stuff’ and Latv. puvesis119 ‘purulence, rot’. It is, therefore, only reasonable to assume that the abstract nouns in -esis must continue PIE neuter abstracts in *-os/-es- in some way or other. But, as Bammesberger (1973: 86) points out, the above mentioned inherited s-stems are obviously not abstract nouns. The origin of the suffix must therefore lie in a PIE verbal abstract that was inherited into the Baltic languages and was then able to serve as the starting point for the productive suffix -esis.120 Despite the reasonably manageable amount of data that comes into consideration, this starting point has not yet been found. Let us therefore reconsider the Latvian evidence, where the suffix is no longer productive. Leskien (1891: 594) lists a handful of Latvian words in -esis, all of which denote concrete nouns and can synchronically be associated with corre- sponding verbs, although in some cases the semantic relation seems somewhat far-fetched. Two nouns, the already mentioned Latv. puvesis ‘purulence, rot’ (cf. pūt ‘to rot’) and Latv. gŗuveši [pl.] ‘ruins’ (cf. grūt ‘to collapse’), have counter- parts in Lithuanian (Lith. puvsis ‘rotten stuff’ and Lith. griuvsiai (pl.) ‘ruins’), the other ones being limited to Latvian: Latv. kruvesis ‘frozen mud’ (cf. kŗaũt ‘to heap’?), Latv. tupesis ‘haystack’ (cf. tupēt ‘to cower’?) and Latv. dzeresis ‘a sour drink’ (cf. dzert ‘to drink’). For some reason, Leskien does not mention Latv. ēdesis, which has an equivalent in Lith. desis/ėdesỹs. Yet, it is exactly this word that must have been the source for the spreading of the suffix -esis in Lithuanian, and to a lesser extent in Latvian. It seems very probable that Proto-Baltic inherited a PIE s-stem *h₁d-es-

117 Cf. Ved. puvas- (Lubotsky apud de Vaan 2005: 62), Gk. πύος, Lat. pūs ‘purulence’, and perhaps Arm. how ‘purulent blood’. All the words reflect zero grade of the root, which can be interpreted as a grundsprachlich generalization of the weak stem *puH-és-. However, I do not believe that the strong stem *péu̯H-os ever existed in the first place. It is an observable phenomenon that roots in **-eu̯H show a tendency to occur in what looks like a zero grade where one would expect a normal full grade, thus appearing almost exclusively as *-uH (cf. Nussbaum 1986: 66 note 53 for this phenomenon in root nouns). The same principle can furthermore explain the zero-grade s-stem PIE *sríHg-os > Gk. ῥῖγος, Lat. frīgus ‘cold, frost, chill’, cf. Höfler 2012: 157f. 118 Gen. -io, m. or f., also: puvėsỹs, pùvėsio m., LDW: 3, 2046. The long vowel of the suffix is clearly secondary (cf. Ambrazas 1993: 86f.). 119 Predominantly used in the pl. puveši (m.); cf. LVV: 3, 443. 120 “Wir müßten somit Ausschau halten nach einem indogermanischen Verbalabstrakt, das ins Baltische ererbt wurde und der Ansatzpunkt für das produktive Suffix -esis/-esỹs sein konnte. Ein eindeutiges Vorbild habe ich jedoch nicht finden können” (Bammesberger 1973: 86).

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 320 Stefan Höfler with the twofold121 meaning ‘eating’ and ‘what is eaten’ (> ‘food, fodder’). In a next step, it was remodeled to *d-es-io- in some sort of mechanical process that did not induce any change in semantics, just as is shown by some of the other122 inherited s-stems. Because, synchronically in Lithuanian, desis was interpretable as an abstract to the verb sti, du ‘eat, devour’ via the suffix -esis/-esỹs, this suffix could then be used to form verbal abstracts from all different kinds of verbs. In Latvian, however, where the meaning of an action noun ‘eating’ was supposedly given up in favour of a specialized nomen rei actae ‘what is eaten (by animals)’, it served as a model for only a small group of concrete nomina rei actae, the most obvious and semantically close example being ‘what is drunk’ as Latv. dzeresis ‘a sour drink’. There is one more indication of positive evidence of the erstwhile existence of a Proto-Baltic neuter *d-es-: Apparently, some inherited s-stems survived into einzelsprachlich times not only extended by *-i- and *-io-, but occasionally also by *-ti(o)-. This seems to be the case with the hapax Lith. augestis (LDW: 1, 243/2) ‘growth’ (as if < *h₂eu̯g-es-ti(o)-; cf. *h₂eu̯g-es- in Ved. ójas- ‘strength, vigor, power’ [RV+], Av. aojah- ‘strength’) and is most certainly the source of the marginal Lith. ėdestis (LKŽ: 2, 1043/1) ‘fodder’.

121 As Stüber (2002: 243 et passim) points out, most PIE s-stems from transitive verbal roots show the semantics of nomina rei actae (e. g. ‘what is eaten’). Originally, however, they also served as nomina actionis (e. g. ‘eating’), which explains their being remodeled and grammaticalized as infinitives in many languages. 122 In fact, the pair Lith. puvsis ‘rotten stuff’ and Latv. puvesis ‘purulence, rot’ exhibits almost exactly the same development. Since it is very probable that the two words are inherited from PIE but, at the same time, stand in a synchronic relation to the verbs Lith. púti, pų̄vù ‘rot, decay’ (LDW: 3, 2044) and Latv. pũt, puvu ‘rot’ (LVV: 3, 452), one could, of course, argue that the productivity of the suffix -esis originates from this substantive. I am inclined to accept that Latv. puvesis could have served as a model for the semantically not too remote Latv. kruvesis ‘frozen mud’ (unless one wants to see in this word the Latvian equivalent of the Greek neuter s-stem κρύος ‘icy cold, frost’, which is formally possible and semantically at least not impossible. In that case, both forms would go back to a stem like *kruH-os :*kruH-es-, whose phonological and morphological development in the two languages would have been exactly as in *puH-os :*puH-es- > Gk. πύος, Latv. puvesis. As to the root in question, one would easily accept that Latv. kruvesis and kŗaũt belong to *√kreu̯H ‘aufhäufen, bedecken’ (LIV²: 371) and that the verbal noun underwent a semantic specialization – cf. a (dung) heap, ein Haufen (Mist) etc. –, but it seems quite hard to account for Gk. κρύος ‘icy cold, frost’ under these premises. For (other) possible etymological connections which do not, however, fully satisfy on morphological and semantic levels, cf. Chantraine 1968–1980: 588f.; Frisk 1960–1972: 2, 28f.; Beekes 2010: 1, 786), but I rather doubt that a word of such specialized semantics could be a better starting point for the spreading of the suffix than the everyday word ‘to eat’.

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 Notes on three “acrostatic” neuter s-stems 321

As for the vocalism of the s-stem in question, however, the Baltic words are of little explanatory power. It is true that both forms seem to point towards a long- vowel derivative *ēd-es-io-, but the vowel length can, of course, be of secondary origin. All nominal derivatives of the root123 in Baltic reflect a long ē and may have generalized this vocalism analogically to the verb. As for the verbum, there are two possible explanations for the long vowel: It may be the result of Winter’s law,124 or go back to a Narten present *h₁ḗd-/h₁éd-.125 Even if the Baltic languages inherited an s-stem *h₁ḗd-os, as I have attempted to demonstrate, the long root vowel cannot serve as proof for a PIE lengthened grade. 4.2 Evidence for a PIE *h₁ḗd-os126 is also found in Latin. At a first glance however, the infinitive ēsse ‘to eat’ (Naev.+)127 seems inconclusive for our purposes, because, even though Latin infinitives are believed to go back to locatives of neuter s-stems that served as verbal abstracts,128 one would expect the outcome †ēdere or ĕdere129 (from *h₁ēd-es-i or *h₁ĕd-es-i). Yet, some supposedly archaic infinitive formations in Latin do also reflect a zero-grade suffix plus the assumed loc. sg. ending (cf. esse ‘to be’, uelle ‘to want’, ferre ‘to bring’, with -se as if < *-s-i130).

123 The only counter-example is Lith. dantìs m. ‘tooth’, OPr. dantis ‘id.’ (*h₁d-ont-), which was, however, presumably already lexicalized in PIE and therefore no longer linked to the verbal root. 124 Proposed by Winter 1978: 438f. 125 Proposed by Narten 1968: 15 note 44; with further implications cf. Schindler 1975a: 62; Isebaert 1992: 194; Kümmel 1998: 203f. 126 Very doubtful is the account by Festus that Lat. ador n. ‘a kind of coarse grain’ had an early form edor that implies a connection with the verb ‘to eat’ (“ador farris genus, edor quondam appellatum ab edendo (…)”; Paul. Fest. p. 3M). The desinence -or (instead of expected †edus) would then be reminiscent of other neuter s-stems with a leveled nom.-acc. sg. like aequor,-oris ‘sea’, rōbur,-oris ‘oak tree, hard timber’ and fulgur,-uris ‘thunderbolt’. But a change from *edor to ador is completely ad hoc. The “modern” etymology of ador, however, is also not unproblematic: It might be related to the s-stem OIr. ad ‘a kind of grain’ that it glosses (cf. Stokes 1887: 293) and belong to the root *√h₂ed ‘vertrocknen’ (LIV²: 255). As for the semantics cf. Festus’ folk-etymological explanation “(…) uel quod aduratur, ut fiat tostum (…)”. 127 The spelling 〈ss〉 is secondary. The length of the vowel is vouched for by the demand of Nisus, a grammarian of the 1st century A.D., for a spelling comese, since the vowel in the second syllable was long, and by a Latin defixio in the Greek alphabet that spells ησσε, cf. Weiss 2009a: 431 note 27. 128 Of the type *ǵenh₁-os, loc. sg. *ǵenh₁-es-i > genus, genere, that could then be referred to a thematic present of the same root (here: OLat. genunt ‘they beget’), cf. Meiser 1998: 225. 129 This form is, in fact, the analogically created infinitive and in common use since the Roman imperial period, cf. Meiser 1998: 223. 130 Certainly, these forms can also be analyzed as consisting of the athematic stem plus *-si, which had at some stage been reinterpreted as an infinitive suffix, all the more so because it is doubtful whether the s-stems *h₁es-os,*u̯el (h₁)-os and *bʰer-os ever existed in the first place,

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 322 Stefan Höfler

If one, as per Peters 2002: 123, accepts that the origin of infinitives of the type Lat. dīxe (synchronically a perfect infinitive)131 and Gk. δεῖξαι (synchronically a sigmatic aorist infinitive) lies in a directive/allative in *-a of an s-stem (viz. *deḱ- s-a132),133 implying that the all. sg. of proterokinetic stems (as much as the instr. sg.)134 followed the hysterokinetic pattern, then Lat. ēsse might also be analyzed in this respect as an archaic formation *h₁d-s-a (vel sim.) with leveled root ablaut. But even if this interpretation were correct, the vowel length could be explained, for example, via Lachmann’s law135 and need not be original. 4.3 The Vedic compound ríśdas- (RV+) is used as an epithet for various gods. There are two main interpretations of the underlying stems.136 The first option would be ‘Sorge um den Fremdling tragend’ with ri° for arí- in composition (*Hri°; cf. also Peters 1986: 370 note 18) and the s-stem *śādas- (cf. Gk. κῆδος ‘care, mourning’, Goth. hatis137 ‘hate’),138 the other one being ‘Speise rupfend’ (= ‘fastidious, picky’) with *riśa° from √riś ‘pluck, rip’ (cf. VIA: 228) and *adas- from *h₁ed-es-. Even if the latter analysis is the correct one, it is of little help for

despite Ved. bháras- ‘care, maintenance’ (AV), Gk. προ-φερής ‘excellent’ (Il.: προφερέστερος +), for both of which Stüber (2002: 64) considers an einzelsprachlich origin, plus arm. ber(k‘) ‘harvest, fruit’ which need not continue an s-stem, pace Matzinger 2005: 41f. Therefore, ēsse may also be analyzed as an analogical formation of the athematic stem ed- plus -se. 131 Unless it stands for dīxisse by haplology, cf. Sommer 1914: 589f. The form appears e. g. in Plaut. Poen. 961. 132 Of course, Latin must have replaced the ending *-a analogically by *-i or *-e(), or one assumes an original directive ending *-a which would perhaps have ended up as -e (as per Weiss 2009a: 446). 133 Ved. jiṣé (RV 1.111.4; 1.112.12), which also perhaps belongs here, has been identified by Stüber as an infinitive of the root √ji (VIA: 187) ‘to conquer’ (PIE *√gu̯e ‘to prevail, win’, LIV²: 206), viz. from a dat. sg. *gu̯i-s-é, cf. Stüber 2000: 152. Of course, she assumes that the underlying substantive was non-neuter because of the structural correspondence to the amphikinetic s-stems bhiyás- m. or f. ‘fear’ (instr. sg. bhīṣ <*bʰih₂-s-éh₁) and uṣás- f. ‘dawn’ (gen. abl. sg. uṣás < *h₂us-s-és). In the light of the aforementioned proposal, the form could, however, reflect the perfectly shaped all. sg. *gu̯i-s-á of a neuter s-stem *gu̯e-os. 134 Cf. Stifter 1997: 219, with reference to Schindler, Nussbaum, and Peters. 135 Cf. Weiss 2009a: 175 and also pres. ind. 2nd sg. ēs (< *h₁ed-s), 3rd sg. ēst (from *h₁ed-t >†ēs plus analogically restored -t), unless one ascribes the length to the Narten present (cf. Isebaert 1992: 195f.; Weiss 2009a: 431), which might be furthermore suggested by the subj. (← opt.) edī- (cf. Kümmel 1998: 203 and note 49). 136 Cf. EWAia: 2, 451. 137 The Germanic continuants (cf. also ON hatr, OE hete) could reflect the zero-grade root ablaut of the proterokinetic weak stem of this word (*ḱeh₂d-os :*ḱh₂d-és-), or the short vowel was analogically introduced from the verb (Goth. hatan ‘to hate’ etc.; cf. Casaretto 2004: 561). 138 Cf. Pinault 2000: 441ff. for this interpretation and a thorough discussion of the compound.

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 Notes on three “acrostatic” neuter s-stems 323 our purposes, since it could, of course, also reflect *riśa-ādas- with a long-vocalic *h₁ēd-es- as the second member of the compound. 4.4 Some severe problems also lie behind Umbr. ezariaf139 (IV 27), if the interpretation as an acc. pl. of a derivative *h₁ed-es-āso- is correct and the meaning is something like ‘food (as an oblation)?’. We would then, however, expect an unrhotacized outcome of the suffix *-āso-, as suggested by plenasier urnasier (Va 2)140 etc. Besides, *d should be reflected as ř, or at least, adjacent to *z (from intervocalic *s), dissimilated to rs.141 Meiser, therefore, suggests a series of con- ditioned sound changes142 to account for the peculiar spelling. Yet it is far from certain that the word belongs here, so it should better be left out. 4.5 In Greek we find some words that, at a first glance, seem to reflect derivatives of a stem ἐδεσ-. To this small group belong ἐδεστής ‘eater’ (Hdt., Antiph.), ἔδεσμα n. ‘food’ (Att.), ἐδεστέον ‘one must eat’ (Plat.) and ἐδεστός ‘eatable, good for food’ (Att.). However, these formations are usually regarded as deverbal. Frisk, for example, explains ἐδεστής, ἔδεσμα and ἐδεστός as built, in some way or other, on the stems of ἠδέσϑην (aor. pass.) and ἐδήδε(σ)μαι (perf. med.), which themselves are Greek innovations, probably after ἐτελέσϑην, τετέλεσμαι, ᾔδέσϑην, ἀλήλε(σ)μαι and the like.143 This account, however, seems somewhat arbitrary. Benveniste showed144 that ἐδεστής is better analyzed as a remodeling of a simplex agent noun *ἐστής (< *ἐδ-τής; for ἐδ- cf. also εἶδαρ ‘food’ [Il.+] < *ἐδ-ϝαρ) – that was at a synchronic level semantically opaque145 – by re-adding ἐδ- in order to restore the relationship with ἔδω, ἔδομαι etc. From then on, the newly created stem ἐδεσ- (actually containing double ἐδ- from two different chronological lay-

139 It is unclear which phoneme was expressed by 〈z〉, but possibly /dz/ or /ts/, cf. Meiser 1986: 240. 140 Both forms are in the abl. pl. as if < *pln-āsos *orden-āsos (?), cf. Untermann 2000: 563f. and 806f. 141 Of course, there is only one example for this development, see note 49 above. 142 He assumes that, before the operating of the regular rhotacism, in a sequence of three frica- tives (as in *eðezāziā-or*eðezāsā-) the third one was dissimilated to r, and that consequently, in syncopated *eðzārā- the *ð was dissimilated in vicinity of r to *d again, leading to *edzāra- or *etsāra-, written as 〈ezaria-〉, cf. Meiser 1986: 239f. 143 Cf. Frisk 1960–1972: 1, 444f. Similarly Chantraine 1968–1980: 312f. and more recently Beekes 2010: 1, 375. 144 Cf. Benveniste 1964: 28–30, but similarly already Chantraine 1933: 317. 145 The simplex survived in compounds such as ὠμηστής ‘eater of raw flesh’ > ‘ferocious’ (with -η- from compositional lengthening; cf. also Ved. āmd- ‘Rohes essend’ (RV 10.87.7d), cf. Scarlata 1999: 34), where the semantic connection to the verb had (gradually) been lost, cf. Benveniste 1964: 29.

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 324 Stefan Höfler ers) was able to serve as the basis for formations like ἔδεσμα and ἐδεστός.146 The ungainly detour via the passive aorist may therefore easily be bypassed. What remains conspicuous, however, is the obvious but hitherto neglected connection of these forms with other derivatives of s-stem bases. For instance, from τέλος n. ‘end, goal, fulfillment; executive function, office; tax, expense; mil- itary unit etc.’ (Hom.+) we find τελεστής ‘an official, priest, initiator’ (Cleanth.) and Hsch. βουτελέστην · ϑύτην ‘sacrificer’, τέλεσμα ‘money paid or to be paid, payment’ (GDI 3749.55 etc., Diod. S.), τελεστός ‘fulfilled’ (IG II² 4548) and ἀ- τελεστός ‘without end, unaccomplished’ (Hom.+). It seems evident that these, to some extent rather late and marginal formations, are derived from the denominative verb τελέω, τελείω (as if < *telés-e/o-147) ‘to finish, complete, initiate; to discharge, pay’ (Il.+).148 But it is difficult on a semantic level,149 and nearly impossible on a formal one,150 to decide whether the derivational base was the nominal or the verbal stem. In principle, the same can be said about ἄκος n. ‘cure, remedy’ (Il.+) and ἀκέομαι ‘to cure; repair’ (Il.+): We find ἀκεστής ‘patcher, tailor’151 (Xen.+), ἀκέσματα n. pl. (Il. +), ἄκεσμα (Aesch.+) ‘remedy, medecine’ and ἀκεστός ‘curable’ (Il. 13.115, Hp., Antiphon.).152

146 Benveniste even shows that these two formations (plus ἐδεστέον) may have been created in immediate analogy to the derivatives of their semantic counterpart πίνω ‘to drink’, viz. πόμα (Pind.), πῶμα (Aesch.), ποτός (Hom.+) and ποτέον. 147 But cf. in detail Peters 1984: 99. 148 Yet, Chantraine 1968–1980: 1102 and Frisk 1960–1972: 2, 871f. regard ἀ-τελεστός as denominal, as well as dial. τελεστα ‘some kind of official’ (from Elis, cf. Bechtel 1923: 848 and also Chantraine 1933: 313), which must, in my opinion, be identical with the (perhaps only coincidentally) late attested τελεστής, and also with Myc. te-re-ta ‘id.?’ (cf. DMic: 2, 338f.). 149 The clear deverbative meaning of ἐδεστός ‘eaten’ (Soph. Ant. 206) is attested at the same time as ‘eatable, good for food’ and ἐδεστά pl. ‘meats’ (Eur. Fr. 472.19), for which the semantic analysis as deverbative ‘(what is) eaten’ > ‘eatable(s)’ is also acceptable. Cf. also ποτός ‘for drinking’ and ποτόν ‘a drink’. A denominative interpretation would require a development ‘provided with eating, having food’ (cf. the type Lat. barbātus, Lith. barzdótas ‘having a beard’) > ‘eatable(s)’, which might seem less convincing. 150 The deverbative use of -μα is well-attested while there is only marginal evidence for denominal formations (cf. Schwyzer 1939: 522–4; Risch 1974: 49f.). For -τής and -τός both formation patterns are well documented (cf. Schwyzer 1939: 499–501 and 501–03; Risch 1974: 33–5 and 19–21). 151 In this case, the meaning clearly indicates that the form is deverbal, since only the verb ἀκέομαι also has the specialized meaning ‘to repair’, which is needed to account for ‘patcher, tailor’. 152 For the latter, Frisk 1960–1972: 1, 56 for some reason accepts a denominal origin.

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 Notes on three “acrostatic” neuter s-stems 325

But this pattern cannot account for κηδεστής ‘relative by marriage’153 (Eur.+) and perhaps ἀ-κήδεστος ‘careless, unburied’154 (Il.+), since there is no denominative verb †κηδέω from κῆδος n. ‘care, mourning, funeral rites, connection by marriage, affinity’ (Il.+) that could have served as a verbal base. The same applies, except for the accent, to κεράστης ‘horned (being)’ (Soph., Eur., of ἔλαφος, Πάν, etc.) formed directly from the stem of the neuter κέρας ‘horn’. If derivatives in -τής (and maybe also -μα and -τός) could thus be directly derived from s-stem bases, one could argue the same origin for ἐδεστής (and also ἔδεσμα and ἐδεστός), and therefore claim the existence at some time of a simplex *ἔδος : *ἔδεσ-, that for some reason was not preserved in any written account of the Greek language. As absurd as this assumption may sound at first, it is exactly this strategy that is commonly accepted for explaining ἀργεστής, an epithet for the south wind (νότος, Il.) and the west wind (Zέφυρος, Hes.), also Ἀργέστης (Arist. etc.) as the name of this wind,155 where a verbal base does not exist. It is hardly possible to disprove either of the two outlined attempts to explain ἐδεστής, ἔδεσμα and ἐδεστός, since both approaches provide lucid arguments.156 But it is, after all, suspicious that there is no undoubted trace of a simplex *ἔδος in Greek.157 Benveniste’s hypothesis might therefore be preferred.

153 Both Frisk 1960–1972: 1, 836f.; and Chantraine 1933: 313; Chantraine 1968–1980: 523 designate it as denominal. 154 Frisk identifies it with the synonymous ἀ-κηδής (Il.+) and implies a denominal origin, whereas Chantraine (1968–1980: 523) interprets it as deverbal from ἀκηδέω (Hom., Aesch., S.), which itself would be denominative from ἀ-κηδής. For the coexistence of internally derived possessive compounds (ἀ-κηδής) and compounds with a possessive suffix (ἀ-κήδεστος) cf. also Widmer 2013: 190. 155 With contrastive accent; cf. Frisk 1960–1972: 1, 132; Chantraine 1968–1980: 104. The s-stem *ἄργος is furthermore implied by the compound ἐναργής ‘clearly visible, prominent, splendid’ (Hom.) and the adjective ἀργεννός ‘white’ (Il.) < *ἀργεσ-νός. Benveniste (1964: 29) also cites ἀργεστής and κηδεστής as examples of denominal derivatives of the type that served as a model for reinterpreting ἐδεστής (< ἐδ- + *ἐδ-τής) as ἐδεσ-τής. 156 Another point for Benveniste’s proposition comes from a seemingly parallel formation ἐδωδή ‘food, meal’ (Il.+), which he analyzes as ἐδ- + *ὠδή, cf. Benveniste 1964: 32; and more recently Vine 1998: 695–7 as ἐδ- + *ὤδη. 157 Maybe, however, it is not *ἔδος that we should be looking for, but *ἦδος. There is a neuter of this appearance in Homer ἦδος ‘delight, pleasure’ (Il.+) and later ‘vinegar (as a flavoring)’ (Att., semantics based on the denominative ἡδῡνώ ‘make pleasant, season (a dish)’, cf. Chantraine 1968–1980: 406), that, as opposed to ἥδομαι ‘to rejoice’ and ἡδύς ‘sweet, tasteful, pleasant, pleasing’ (Il.+), exhibits an absence of aspiration and only doubtful traces of the digamma (but cf. Dor. ἇδος (in both senses) and Hsch. γᾶδος · γάλα, ἄλλοι ὄξος), for which there is no satisfactory explanation (cf. Chantraine 1958: 184 and 151). Is it possible that *ἦδος ‘food’ and *ἧδος ‘pleasure’ merged into one word? The merging point might have been the possessive compound ἀηδής

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 326 Stefan Höfler

4.6 Another piece of evidence of PIE *h₁ed-es- can be found in Finn. ateria158 ‘meal’, which was identified by Schindler (1963: 205f.) as a borrowing from Proto-Norse *āteRja < PGerm. *ētez(i)jan, ultimately reflecting a derivative *h₁ḗd- es-(i)o-m159 ‘food’. If this etymology is to be taken seriously, we also have to find a reasonable explanation for the long root vowel that, in the case of Germanic, cannot be traced back to a short ĕ by any expected regular sound development.160 4.7 There is another derivative from a base *h₁ēd-(e)s- in the Germanic languages, namely *h₁ēd-s-o- n. (in OEngl. ǣs ‘food, carrion’, OFris. ēs, MHG âs ‘cadaver’, etc.161), which seems to have an equivalent in Tocharian (B yetse,Ayats m. ‘(outer) skin’; cf. Adams 1999: 507f.) and in Russ. ясá (jasá) f. ‘meal, course of a meal’162 (< *h₁ēd-s-eh₂-). Morphologically, these forms either reflect thematic derivatives of a long-vocalic base *h₁ēd-(e)s- or vṛddhi-derivatives of a short-vocalic stem

‘unpleasant’ (Sappho, Hdt., Pl.) with a specialized meaning ‘distasteful, nauseous (of food, drugs etc.)’ (Hippocr., Pl.), which could have been interpreted as both ἀ + ἡδής ‘having/providing no delight’ and ἀ(ν) + ἠδής ‘having/providing no eating/food’ (with *ἀνηδής for *νηδής, as in ἀνωφελής ‘useless’ for *νωφελής (cf. Myc. no-pe-re-a₂ /nōpʰeleʰa/; DMic: 1, 477f.) from ὄφελος ‘promotion, use, advantage, gain’; there seems to have been a certain amount of oscillation between ἀ- and ἀν- before a vowel, h- and former digamma, leading to “irregular” combinations of ἀ- plus an original vowel (cf. Lejeune 1971b: 37ff.), that would have encouraged the proposed confusion of ἀ + ἡδής and ἀ + ἠδής (← ἀν + ἠδής)). Once the overlapping semantics ‘unpleasant (to eat)’ and ‘unpleasant (in general)’ coalesced, the second compound member could no longer be distinguished. Thus, the reanalyzed simplex might have inherited qua étymologie croisée the meaning of the one and the form of the other. But since the compound is attested relatively late, this idea remains idle speculation. 158 Cf. also the dialectal variants atria, aria, arja and Vepsian ateŕǵ, ateŕ ‘time between meals’. 159 The formal similarity to the proposed PBalt. *ēd-es-io- might only be of coincidental origin, as the remodeling of PIE s-stems to i- and io-stems was identified above as an inner-Baltic development, whereas in Germanic most of the neuter s-stems were directly thematized (cf. Schaffner 2001: 588; Casaretto 2004: 555). However, the parallelism is conspicuous. 160 A vṛddhi-derivative (see above 3.2) from a base *h₁ed-esis likewise both morphologically improbable (expected **h₁ḗd-(e)s-o- [see below 4.7] would have had to be remodeled to *h₁ḗd-es-o-, unless one concedes the marginal existence of vṛddhi-derivatives with *-o-, as per Darms 1978: 31, for bases in *-eh₂-) and semantically doubtful (‘belonging to food’ ≥ ‘food, meal’?), but see below 4.7. 161 Cf. EWAhd: 1, 407. Differently Seebold (1970: 180) who assumes *ēd-to->ǣs-a-; cf. also Lat. ēsus ‘eaten’ (< *h₁ed-to- with Lachmann’s law?), OPr. īstai (dat. sg. n.) ‘food’ and OCS jasto n. ‘meal, portion’ (both with Winter lengthening?), cf. NIL: 211; thus also Ringe 2006: 88. 162 Cf. REW: 3, 495. The word is reminiscent of similar *-eh₂-formations with a lengthened root vowel in Balto-Slavic, that are usually regarded as feminine vṛddhi-derivatives, cf. Nussbaum 1986: 8; Villanueva Svensson 2011: 30–2; differently Pronk 2012: 211–3. The long vowel can, however, be analogical, cf. also Russ. едá (jedá) ‘meal, food’ (as if < *h₁ēd-eh₂-) and Russ. ежá (ježá) ‘meal, food’ (as if < *h₁ēd-(i)eh₂-), REW: 1, 391f.

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 Notes on three “acrostatic” neuter s-stems 327

*h₁ĕd-(e)s-. Semantically, both interpretations do not really produce smooth re- sults: The first option would imply a possessive derivative ‘having or providing food’,which, in the case of Germanic, would have been synonymous to ‘food’ and then develop, via a process of semantic narrowing and degeneration,163 to ‘carrion, cadaver’. A vṛddhi-derivative ‘belonging to or consisting of food’ would make just as much sense theoretically, at least for Germanic. Another possibility164 is that *h₁d-es- already developed a meaning ‘meat, flesh’ in early times and that the derivative thus denoted ‘belonging to the meat, flesh’ as the edible parts of a killed animal, in contrast to the bones etc. In Tocharian, the meaning would have been specialized from ‘belonging to the flesh’ to ‘(outer) skin’. But even if we ascribe the length in *h₁ēd-s-o- and its continuants to the deus ex machina-process of vṛddhi-derivation, we still find additional long-vowel forms in the thematic neuters ON át ‘food, meal’, OEngl. ǣt, OHG āz ‘meal’ (as if <*h₁ēd-o-) and feminine OHG āza ‘meal’ (as if < *h₁ēd-eh₂-, cf. Lith. da, Russ. едá [jedá])165 etc. that seem to indicate that also in *h₁ēd-s-o- the length had better be regarded as original. As opposed to the aforementioned examples in Balto-Slavic and Latin, attempted explanations such as Winter’s and Lachmann’s law or analogical influence from the verb are virtually impossible in the case of Germanic. Neither is there any (accepted) sound law that would account for long ē before certain consonant clusters, nor does the verb in Germanic show a lengthened grade in the present stem166 that could have been transferred analogically to other formations

163 Cf. for these notions in general Bloomfield 1935: 426f., and in particular the similar example OEngl. mete ‘food’ > meat ‘edible flesh’. 164 Melanie Malzahn (p. c.). 165 Cf. EWAhd: 1, 406f. As mentioned above (see note 31) a long vowel is found frequently in *-eh₂-formations, at least in Germanic, and might therefore not come as a great surprise (cf. also Villanueva Svensson 2011: 7). 166 Cf. Goth. itan, ON eta, OEngl. etan, OHG ezzan, etc. as thematized continuants (*h₁ed-e/o-> PGmc. *iti-:*eta-) of the weak stem of a Narten present *h₁ḗd-:*h₁éd-?, cf. Ringe 2006: 174.The lengthened grade is found in the preterite (cf. Goth. et, ON át, OEngl. ǣt, OHG āz, etc.; from *h₁e-h₁d-?, cf. LIV²: 231 note 13 and Ringe 2006: 185) but I see no chance that the preterite stem could have influenced the nominal forms.

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 328 Stefan Höfler of this root.167 It seems, therefore, as if they reflected derivatives of an s-stem with a genuine lengthened grade.168 4.8 The remaining derivatives of an s-stem base with a suffixal zero grade known to me are predominantly from Balto-Slavic. In most of the cases, it is impossible to tell whether the form is a secondary derivative of an s-stem or a derivative via a complex suffix with an anlauting s, whose origin may or may not be based on a reinterpretation of some of these secondary derivatives. The vowel length can in all places be basically explained by Winter’s and/or Lachmann’s law, but there is no reason at all to assume that it cannot just as well be original. The forms of certain particular interest are169 *h₁ēd-s-l(i)o- (Latv. ēslis ‘some- one who constantly masticates, glutton’ and OCS jasli, Russ. я́сли (jásli) etc. pl. ‘crib, manger’, which match formally but obviously not semantically), *h₁ēd-s- meh₂- (Latv. ȩ̂sma f. ‘bait for wolves’), *h₁ēd-s-neh₂- (Lith. snos f. pl. ‘gingiva,

167 Suspiciously enough, there are also substantives that clearly reflect a short root vowel (cf. OHG ezza ‘indulgence’, as if < *h₁ed-eh₂-; OHG ezzo ‘glutton’, as if < *h₁ed-on-; OEngl. etol, ettul, OHG filu-ezzal ‘gluttonous’, as if < *h₁ed-ulo-). However, these could be more recent, deverbal formations, as opposed to the above-mentioned, long-vocalic forms, that, considering their occurrence in North- and West- and, with Goth. uz-eta* ‘crib’ and af-etja ‘glutton’ (GothED: 208) also in East-Germanic, would date back to an earlier stage. 168 Another conspicuous feature of the discussed Germanic derivatives is that they reflect two different derivational bases *h₁ēd-es- and *h₁ēd-s-. It is hard to determine the formal or semantic difference between the two stems. More generally speaking, it is far from clear what secondary derivatives of s-stems are supposed to look like and what conclusions can be drawn from the various formations showing different root and suffixal ablaut. In Vedic, for instance, we find in total four different types of thematic derivatives of s-stems (cf. for this Debrunner 1954: 126f. and 136f.). There are regular vṛddhi-derivatives of the structure R(ā)-as-á- (cf. āyasá- ‘made of iron’ from áyas- ‘iron’), possessive derivatives without vṛddhi as R(a)-as-á- (cf. rabhasá- ‘wild, violent’ from rábhas- ‘violence’), some forms with a zero-grade suffix R(a)-s-á-: Cf. vatsá- ‘calf’ < ‘yearling’ from *u̯et-es- ‘year’, as in Gk. ἔτος ‘id.’ and also OIr. feis, Middle Welsh gwys, OBret. guis, OCorn. guis ‘sow’ < *u̯et-s-ih₂- (cf. Stüber 2002: 188)), and one instance of a derivative with zero grade in the root and the suffix R(ø)-s-á-(útsa- ‘spring, well’ < ‘having water’ (cf. EWAia: 1, 215, also for equivalents in the Iranian languages) from *u̯ed-es- ‘water’ as in Arm. get,-oy ‘river’ (cf. Olsen 1999: 45f.; Matzinger 2005: 43) and Gk. ὕδος n. (Call. Fr. 475), dat. sg. ὕδει (Hes.Op.61) ‘water’ (cf. Chantraine 1968–1980: 1153; according to Nussbaum 1986: 203 note 16, the latter can also belong to a root noun) with a zero grade by analogy to the far more frequent synonymous heteroclite ὕδωρ, which itself, according to Schindler (1975b: 3f.), generalized the root vocalism of the weak stem. The lexicalized semantics of the latter two seem to indicate that their formation is the more ancient one. Yet it remains unclear how the Germanic forms fit into this picture. 169 Cf. for this IEW: 288f.; NIL: 210.

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 Notes on three “acrostatic” neuter s-stems 329 gums’) and *h₁ēd-s-keh₂- (Lat. ēsca170 f. ‘food, bait’, Lith. ėskà f. ‘food, fodder, ap- petite’, Latv. ēšķa and ēšķis ‘glutton, scolding person’). 4.9 In conclusion and consideration of the abundance of derivatives, it seems fairly safe to assume that an s-stem *h₁d-os must have existed well into einzelsprachlich times, even though there is no trace of the simplex itself. This peculiarity might be due to the variety of other synonymous and therefore competing derivatives of the root *√h₁ed and simultaneously a consequence of the tendency to recharacterize apparently sub-characterized derivatives (like a simple s-stem) by extending them via additional stem formants.171 It is also clear, however, that the discussed derivatives have, in principle, only little significance when we try to determine the root ablaut of the underlying simplex. As we have seen, it is possible to explain the vowel length in some cases as resulting secondarily from certain presupposed sound laws or from the analogical influence of associated verbal forms. Yet in some cases, as in OEngl. ǣs ‘food, carrion’ etc. and Toch. B yetse ‘(outer) skin’, such an approach is doubtful. Taken together, therefore, the evidence speaks in favor of a long-vowel formation *h₁ēd-os :*h₁ēd-es-, which then accounts for all the discussed continuants and derivatives. And assuming a long- vowel pre-form makes perfect sense in the light of the working hypothesis of this paper if we assert that the s-stem ‘food’ was connected to the secondary seman- tics172 of the Narten present *h₁ḗd-ti ‘eats’ rather than to the root *√h₁ed ‘to bite’,173 and that that it consequently owes its long vowel to the present ‘to eat’.

5 Conclusion

After going through all these examples, we are now able to draw a conclusion. As we have seen, it is perfectly possible to explain long-vowel s-stems like Gk. μήδεα, Arm. mit, and the various derivatives and/or continuants of *h₁ēd-(e)s- as being

170 Cf. for this pattern also Gk. λέσχη ‘resting place’ < *legʰ-s-keh₂- from *légʰ-os ‘bed’ (λέχος ‘id.’), cf. Isebaert 1992: 203, and Welsh gwisg ‘garment’ < *u̯ēs-s-keh₂-, for which see above note 14. 171 Cf. for similar processes Matzinger 2008: 25ff. 172 Cf. Schindler 1975a: 62; Peters 1980: 24 note 24; Isebaert 1992: 194; Kümmel 1998: 203f. 173 Therefore, a hypothetical s-stem *h₁ĕd-es- should have had the semantics of ‘a bite, a bit, a chunk’. It may be exactly this word that served as the basis for Hitt. ezza-, izza- n. ‘chaff’ (HEG: 1, 119: “Stroh; Spreu”; HED: 321: “chaff”, also symbolic of or idiomatic for “(stored) holdings, (material) goods”; HW²: 2, 141: “Spreu, Häcksel”), whose etymology has until now been obscure (cf. HEG; HED; HW²: loc. cit.). On the phonological side, Hitt. ezza- would be the perfectly expected outcome of a thematic derivative *h₁ĕd-s-o- whose morphology, on the other hand, can be neatly compared to *u̯ĕt-s-o- > Ved. vatsá- ‘calf’ from *u̯et-es- ‘year’. As for the semantic relation in ques-

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 330 Stefan Höfler built on, or remodeled after verbal Narten formations, rather than to assume the PIE acrostatic substantives *mḗd-s :*méd-s-, *h₁ḗd-s :*h₁éd-s- and the like. On the other hand, identifying ON sætr as a vṛddhi-derivative and proposing an étymolo- gie croisée for OIr. síd probably also solves the riddle of the mirage *sḗd-s : séd-s-. I am inclined to believe that similar solutions might also apply to the remaining “acrostatic” s-stems, for some of which a proposal has indeed been uttered in the course of this paper. The concept of “acrostatic” s-stems as a whole might therefore have to be abandoned. Perhaps the same holds true for Narten roots, at least sensu stricto. But considering the many different possible interpretations of long- vowel nominal formations, one has to conclude that, yet again, all has not been said and done.

Acknowledgement: This paper is based on my master’s thesis Untersuchungen zum Ablaut der neutralen s-Stämme des Indogermanischen (Höfler 2012), elaborating the assumptions and improving the views presented there. I am much obliged to Prof. Melanie Malzahn, Prof. Martin Peters, and my colleague Oliver Plötz for sharing their thoughts with me, for answering my questions, and for helping me with some detailed problems. I am also indebted to the anonymous reviewers of the IF for the very helpful comments. Of course, however, I alone am responsible for all conclusions drawn here and for any errors of fact or judgment. The work on this paper was made possible by a DOC scholarship of the Austrian Academy of Sciences, for which I am also very grateful.

tion, one would have to start from an already metaphorically used base ‘a bit’, the possessive derivative of which would have developed from ‘having, consisting of bits’ to ‘mass of small bits’, whence ‘chaff’. A similar semantic development is not only vouched for by English ‘to bite’ and ‘a bit’, ‘bits’, but also for example by Ved. mardáyati ‘grinds, crushes, squelches’, Lat. mordeō ‘I bite’ and East Fris. murt ‘bröckelige Masse, Staub’, Swiss Germ. murz, morz ‘small pieces’ (IEW: 736f.) or Lith. kándu, ką́sti ‘to bite’ and Latv. kņadas ‘Nachbleibsel beim Getreidereinigen’ (LVV: 2, 252), and by the English word chaff itself (OE ceaf, Dutch kaf, German Kaff n. etc.), which (as insinuated e. g. by Holthausen 1934: 44; Onions 1966: 160) possibly belongs to a root *√ǵebʰ ‘essen, kauen’ (LIV²: 161; cf. OLith. žėbmi ‘esse langsam, kaue’, OCS i-zobati ‘verzehrt’ etc.) and its derivatives German Kiefer ‘jaw’, Käfer ‘beetle’, English chafer, MHG kiven, kiffen ‘to gnaw’ plus their various cognates within the Germanic languages (for which cf. IEW: 382). But of course, this etymology remains a mere construct since it will be hard to either verify or falsify it.

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 Notes on three “acrostatic” neuter s-stems 331

Abbreviations

ABL Anna Stafecka et al. (2009). Atlas of the Baltic Languages. A Prospect. Rīga & Vilnius: Latvijas Universitātes Latviešu valodas institūts & Lietuvių kalbos institutas. DMic Francisco Aura Jorro & Francisco Rodríguez Adrados (1985–1993). Diccionario micénico. 2 vols. Madrid: Consejo Superior de Investigaciones Cientifícas, Insti- tuto de Filología. EWAhd Albert L. Lloyd, Otto Springer & Rosemarie Lühr, eds. (1988–). Etymologis- ches Wörterbuch des Althochdeutschen. Göttingen & Zürich: Vandenhoeck & Ruprecht. EWAia Manfred Mayrhofer (1986–2001). Etymologisches Wörterbuch des Altindoari- schen. 3 vols. Heidelberg: Winter. GothED Winfried P. Lehmann (1986). A Gothic Etymological Dictionary. Leiden: Brill. HED Jaan Puhvel (1984–). Hittite Etymological Dictionary. 9 vols. Berlin & New York: Mouton. HEG Johann Tischler (1974–). Hethitisches etymologisches Glossar. Mit Beiträgen von Günter Neumann und Erich Neu. 4 vols. Innsbruck: Institut für Sprachwis- senschaft. HW² Johannes Freidrich & Annelies Kammenhuber (1975–). Hethitisches Wörterbuch. 2nd ed. 3 vols. Heidelberg: Winter. IEW Julius Pokorny (1989). Indogermanisches etymologisches Wörterbuch. 2nd ed. Vol. 1. Bern & Stuttgart: Franke. LDW Alexander T. Kurschat & Wilhelm Wissmann (1968–73). Litauisch-deutsches Wörterbuch. Thesaurus linguae Lituanicae. 4 vols. Göttingen: Vandenhoeck & Ruprecht. LEW Ernst Fraenkel (1962–1965). Litauisches etymologisches Wörterbuch. 2 vols. Göttingen & Heidelberg: Vandenhoeck & Ruprecht, Winter. LIV² Helmut Rix (2001). Lexikon der indogermanischen Verben. Die Wurzeln und ihre Primärstammbildungen. Unter Leitung von Helmut Rix bearbeitet von Martin J. Kümmel, Thomas Zehnder, Reiner Lipp, Brigitte Schirmer. 2nd ed. Wiesbaden: Reichert. LKŽ Lietuvių kalbos žodynas (1941–2002). 20 vols. Vilnius: Mintis & Mokslas. LVV Jānis Endzelīns, ed. (1923–1932). K. Mǖlenbacha Latviešu valodas vārdnīca. 4 vols. Rīga: Izdevusi izglītības ministrija. NIL Dagmar S. Wodtko, Britta Irslinger & Carolin Schneider (2008). Nomina im indogermanischen Lexikon. Heidelberg: Winter. REW Max Vasmer (1953–1958). Russisches etymologisches Wörterbuch. 3 vols. Hei- delberg: Winter. VIA Chlodwig H. Werba (1997). Verba Indoarica. Die primären und sekundären Wurzeln der Sanskrit-Sprache. Wien: Akademie der Wissenschaften.

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 332 Stefan Höfler

Bibliography

Adams, Douglas Q. (1985). “Sanskrit púmān, Latin pūbēs, and related words”. In: Die Sprache 31.1, 1–16. — (1999). A Dictionary of Tocharian B. Amsterdam & Atlanta: Rodopi. Ambrazas, Saulius (1993). Daiktavardžių Darybos Raida. Lietuvių Kalbos Veiksmažodiniai Vediniai. Vilnius: Mokslo ir Enciklopedijų Leidykla. — (1994). “Zur Entwicklung der Abstraktbildungen im Baltischen”. In: Indogermanische Forschungen 99, 277–300. Balles, Irene (2006). Die altindische Cvi-Konstruktion. Form, Funktion, Ursprung. Bremen: Hempen. Bammesberger, Alfred (1973). Abstraktbildungen in den baltischen Sprachen. Göttingen: Vanden- hoeck & Ruprecht. Bechtel, Friedrich (1923). Die griechischen Dialekte. Berlin: Weidmannsche Buchhandlung. Beekes, Robert S. P. (1988). “PIE. RHC- in Greek and other languages”. In: Indogermanische Forschungen 93, 22–45. — (2010). Etymological Dictionary of Greek. With the assistance of Lucien van Beek. 2 vols. Leiden & Boston: Brill. Benedetti, Marina (1988). I composti radicali latini. Esame storico e comparativo. Pisa: Giardini Ed. Benveniste, Émile (1964). “Renouvellement lexical et dérivation en grec ancien”. In: Bulletin de la Société de Linguistique de Paris 59, 24–39. — (1969). Le vocabulaire des institutions indo-européennes. 2 vols. Paris: Les Éditions de Minuit. De Bernardo Stempel, Patrizia (1999). Nominale Wortbildung des älteren Irischen. Stammbildung und Derivation. Tübingen: Niemeyer. Bertocci, Davide (2012). “Survivings of the *-eh₁-stative morphology in Umbrian and Latin”. In: The Sounds of Indo-European. Papers on Indo-European Phonetics, Phonemics and Morphophonemics. Ed. by Roman Sukač & Ondřej Šefčík. Vol. 2. München: Lincom Europa, 14–28. Bloomfield, Leonard (1935). Language. London: Allen & Unwin. Buck, Carl Darling (1904). A Grammar of Oscan and Umbrian. Boston: Ginn & Company. Casaretto, Antje (2004). Nominale Wortbildung der gotischen Sprache. Die Derivation der Sub- stantive. Heidelberg: Winter. Chantraine, Pierre (1933). La formation des noms en grec ancien. Paris: Klincksieck. — (1958). Grammaire homérique. Vol. 1: Phonétique et morphologie. 2nd ed. Paris: Klincksieck. — (1968–1980). Dictionnaire étymologique de la langue grecque. Histoire des mots. 2 vols. Paris: Klincksieck. Clackson, James (1994). The Linguistic Relationship between Armenian and Greek. Oxford: Black- well. Darms, Georges (1978). Schwäher und Schwager, Hahn und Huhn. Die Vddhi-Ableitung im Germanischen. München: Kitzinger. Debrunner, Albert (1954). Altindische Grammatik. Vol. 2.2: Die Nominalsuffixe. Göttingen: Van- denhoeck & Ruprecht. Derksen, Rick (1996). Metatony in Baltic. Amsterdam & Atlanta: Rodopi. Dottin, Georges (1985). La langue gauloise. Grammaire, textes et glossaire. Paris: Klincksieck.

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 Notes on three “acrostatic” neuter s-stems 333

Eichner, Heiner (1973). “Die Etymologie von heth. mehur”. In: Münchener Studien zur Sprachwis- senschaft 31, 53–107. Ernout, Alfred (1965). Philologica III. Paris: Klincksieck. Fraenkel, Ernst (1936). “Die indogermanischen -l-Stämme”. In: Zeitschrift für vergleichende Sprachforschung 63, 168–201. Frisk, Hjalmar (1960–1972). Griechisches etymologisches Wörterbuch. 3 vols. Heidelberg: Winter. Griffith, Aaron (2005). “*-n(C)s in Celtic”. In: Die Sprache 45, 44–67. Hartmann, Markus (2012). “Zum Ablaut der neutralen s-Stämme des Indogermanischen”. In: International Journal of Diachronic Linguistics and Linguistic Reconstruction 9, 51–85. Hilmarsson, Jörundur (1987). “Reflexes of I.-E. *suH₂n̥to-/-ōn ‘sunny’ in Germanic and Tocharian’”. In: Die Sprache 33.1–2, 56–78. Höfler, Stefan (2012). Untersuchungen zum Ablaut der neutralen s-Stämme des Indogermanischen. Diplomarbeit. Wien. — (2013). Ist der Wettergott ein Himmelsgott? Indogermanische Götternamen und ihr Beitrag zur internen Derivation. Paper presented at the 1. Grazer Kolloquium zur indogermanischen Altertumskunde, 14.–15. November 2013, Karl-Franzens-Universität Graz. Holthausen, Ferdinand (1934). Altenglisches etymologisches Wörterbuch. Heidelberg: Winter. Irslinger, Britta (2009). “Indogermanische Abstraktsuffixe – synchron und diachron”. In: Protolan- guage and Prehistory. Akten der XII. Fachtagung der Indogermanischen Gesellschaft, Krakau, 11. bis 15. Oktober 2004. Ed. by Rosemarie Lühr & Sabine Ziegler. Wiesbaden: Reichert, 215– 235. Isebaert, Lambert (1992). “Spuren akrostatischer Präsensflexion im Lateinischen”. In: Latein und Indogermanisch. Akten des Kolloqiums der Indogermanischen Gesellschaft, Salzburg, 23.–26. September 1986. Ed. by Oswald Panagl & Thomas Krisch. Innsbruck: Institut für Sprachwissenschaft, 193–205. Klingenschmitt, Gert (1992). “Die lateinische Nominalflexion”. In: Latein und Indogermanisch. Akten des Kolloqiums der Indogermanischen Gesellschaft, Salzburg, 23.–26. September 1986. Ed. by Oswald Panagl & Thomas Krisch. Innsbruck: Institut für Sprachwissenschaft, 89–135. — (2008). “Erbe und Neuerung bei Akzent und Ablaut in der litauischen Morphologie”. In: Kalbos istorijos ir dialektologijos problemos 2. Ed. by Danguolė Mikulėnienė & Saulius Ambrazas. Vilnius: Lietuvių kalbos institutas, 180–215. Kluge, Friedrich & Elmar Seebold (2002). Etymologisches Wörterbuch der deutschen Sprache. 24th ed. Berlin & New York: de Gruyter. Krahe, Hans & Wolfgang Meid (1967–1969). Germanische Sprachwissenschaft. 7th ed. 3 vols. Berlin & New York: de Gruyter. Kümmel, Martin J. (1998). “Wurzelpräsens neben Wurzelaorist im Indogermanischen”. In: His- torische Sprachforschung 111, 191–208. Latte, Kurt, ed. (1966). Hesychii Alexandrini Lexicon. Vol. 2: E–O. Hauniae: Munksgaard. Lejeune, Michel (1971a). Lepontica. Paris: Les Belles Lettres. — (1971b). “Observations sur les composés privatifs”. In: Mémoires de philologie mycénienne. Ed. by Michel Lejeune. Vol. 2. Roma: Ateneo, 35–45. Leskien, August (1891). “Die Bildung der Nomina im Litauischen”. In: Abhandlungen der philologisch-historischen Classe der königlich sächsischen Gesellschaft der Wissenschaften 12, 151–618.

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 334 Stefan Höfler

Litscher, Roland (2007). “κρέας, kravíḥ and the original nom.-acc. sg. of the IE s-stem neuters”. In: Greek and Latin from an Indo-European perspective. Ed. by Coulter H. George et al. Cambridge: Philological Society, 107–120. Magnússon, Ásgeir B. (1989). Íslensk orðsifjabók. Reykjavík: Orðabók Háskólans. Malzahn, Melanie (2004). “Toch. B yesti nāskoy und der Narten-Charakter der idg. Wurzel *wes ‘(Kleidung) anhaben’”. In: Die Sprache 43.2, 212–220. — (2010). The Tocharian Verbal System. Leiden & Boston: Brill. Martirosyan, Hrach K. (2010). Etymological Dictionary of the Armenian Inherited Lexicon. Leiden & Boston: Brill. Mastrelli, Carlo A. (1979). “Una nota su lat. rēnēs e gr. ῥάχις”. In: Incontri linguistici 5, 37–42. Matasović, Ranko (2009). Etymological Dictionary of Proto-Celtic. Leiden & Boston: Brill. Matzinger, Joachim (2005). Untersuchungen zum altarmenischen Nomen. Die Flexion des Sub- stantivs. Dettelbach: Röll. — (2008). “Multiple suffixes in Indo-European languages”. In: The Discourse Potential of Un- derspecified Structures. Ed. by Anita Steube. Berlin & New York: de Gruyter, 21–44. Mayrhofer, Manfred (1986). Indogermanische Grammatik. Vol. 1.2: Lautlehre [Segmentale Phonologie des Indogermanischen]. Heidelberg: Winter. McCone, Kim R. (1996). Towards a Relative Chronology of Ancient and Medieval Celtic Sound Change. Maynooth: Department of Old Irish. Meier-Brügger, Michael (2002). Indogermanische Sprachwissenschaft. 8th ed. Berlin & New York: de Gruyter. Meillet, Antoine (1908). “Sur l’aoriste sigmatique”. In: Mélanges de linguistique offerts à M. Ferdinand de Saussure. Ed. by Honoré Champion. Paris: Société de Linguistique, 79–106. — (1922). “La forme du génitif pluriel en ombrien”. In: Mémoires de la Société de Linguistique de Paris 22, 258–259. Meiser, Gerhard (1986). Lautgeschichte der umbrischen Sprache. Innsbruck: Institut für Sprach- wissenschaft. — (1998). Historische Laut- und Formenlehre der lateinischen Sprache. Darmstadt: Wis- senschaftliche Buchgesellschaft. Meissner, Torsten (2006). S-stem Nouns and Adjectives in Greek and Proto-Indo-European. A Diachronic Study in Word Formation. Oxford: Oxford University Press. Narten, Johanna (1968). “Zum ‘proterodynamischen’ Wurzelpräsens”. In: Pratidānam. Indian, Iranian and Indo-European studies presented to Franciscus Bernardus Jacobus Kuiper on his sixtieth birthday. Ed. by Jan C. Heesterman, Godard H. Shokker & Vadasery I. Subramoniam. The Hague: Mouton, 9–19. Nikolaev, Alexander S. (2008). “Gedanken über ein neues Buch (Review of: Paul Widmer, Das Korn des weiten Feldes)”. In: Acta Linguistica Petropolitana 4.1, 541–570. Nowicki, Helmut (1976). “Die neutralen s-Stämme im indo-iranischen Zweig des Indogermanis- chen”. PhD thesis. Universität Würzburg. Nussbaum, Alan J. (1976). “Caland’s ‘Law’ and the Caland System”. PhD thesis. Harvard University Press. — (1986). Head and Horn in Indo-European. Berlin & New York: de Gruyter. Olsen, Birgit A. (1999). The Noun in Biblical Armenian. Origin and Word-Formation. Berlin & New York: de Gruyter. Onions, Charles T. (1966). The Oxford Dictionary of English Etymology. Oxford: Clarendon.

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 Notes on three “acrostatic” neuter s-stems 335

Panagl, Oswald (1982). “Homerisch ὄχεα: ein verkappter lautlicher ‘Mykenismus’”. In: Serta Indogermanica. Festschrift für Günter Neumann zum 60. Geburtstag. Ed. by Johann Tischler. Innsbruck: Institut für Sprachwissenschaft der Universität Innsbruck, 251–257. Peters, Martin (1975). “Altpersisch ašiyava”. In: Die Sprache 21, 37–42. — (1977). “Review of: Klaus Strunk, Lachmanns Regel für das Lateinische”. In: Die Sprache 23, 67. — (1980). Untersuchungen zur Vertretung der indogermanischen Laryngale im Griechischen. Wien: Akademie. — (1984). “Indogermanische Chronik 30a – VII. Altgriechisch”. In: Die Sprache 30.1, 67*–105*. — (1986). “Probleme mit anlautenden Laryngalen”. In: Die Sprache 32, 365–85. — (1999). “Ein tiefes Problem”. In: Compositiones Indogermanicae in memoriam Jochem Schindler. Ed. by Hans Chr. Luschützky & Heiner Eichner. Praha: Enigma, 447–456. — (2002). “Indogermanische Chronik 35”. In: Die Sprache 39.3 (Sonderheft) = Chronicalia Indoeuropaea 39 (1997), 94–129. Petit, Daniel (2010). Untersuchungen zu den baltischen Sprachen. Leiden & Boston: Brill. Pinault, Georges-Jean (2000). “Le nom primitif de la rétribution rituelle en védique ancien”. In: Bulletin d’études indiennes 17–18, 427–476. Pronk, Tijmen (2012). “Proto-Indo-European long vowels and Balto-Slavic accentuation”. In: Bal- tistica 47, 205–247. Prosdocimi, Aldo L. (1976). “L’iscrizione di Prestino”. In: Studi etruschi 35, 199–222. Ringe, Donald A. (2006). “A sociolinguistically informed solution to an old problem: the Gothic genitive plural”. In: Transactions of the Philological Society 104.2, 167–206. Risch, Ernst (1974). Wortbildung der homerischen Sprache. 2., stark veränderte Auflage. Berlin & New York: de Gruyter. Scarlata, Salvatore (1999). Die Wurzelkomposita im Ṛgveda. Wiesbaden: Reichert. Schaffner, Stefan (2001). Das Vernersche Gesetz und der innerparadigmatische grammatische Wechsel des Urgermanischen im Nominalbereich. Innsbruck: Institut für Sprachen und Literaturen. Schindler, Jochem (1963). “Zu einigen Lehnwörtern im Finnischen”. In: Die Sprache 9, 203–206. — (1972a). “Das Wurzelnomen im Arischen und Griechischen”. PhD thesis. Universität Würzburg. — (1972b). “L’apophonie des noms-racines indo-européens”. In: Bulletin de la Société de Linguistique de Paris 67.1, 31–8. — (1975a). “Armenisch erkn, griechisch ὀδύνη, irisch idu”. In: Zeitschrift für vergleichende Sprachforschung 89, 53–65. — (1975b). “L’apophonie des thèmes indo-européens en -r/n”. In: Bulletin de la Société de Linguistique de Paris 70.1, 1–10. — (1975c). “Zum Ablaut der neutralen s-Stämme des Indogermanischen”. In: Flexion und Wort- bildung. Akten der V. Fachtagung der Indogermanischen Gesellschaft. Ed. by Helmut Rix. Wiesbaden: Reichert, 259–267. — (1994). “Alte und neue Fragen zum indogermanischen Nomen”. In: In memoriam Holger Pedersen. Kolloquium der Indogermanischen Gesellschaft vom 26. bis 28. März 1993 in Kopenhagen. Ed. by Jens E. Rasmussen. Wiesbaden: Reichert, 397–400. Schmidt, Johannes (1889). Die Pluralbildungen der indogermanischen Neutra. Weimar: Böhlau. Schrijver, Peter (1991). The Reflexes of the Proto-Indo-European Laryngeals in Latin. Amsterdam & Atlanta: Rodopi. — (1995). Studies in British Celtic Historical Phonology. Amsterdam & Atlanta: Rodopi.

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 336 Stefan Höfler

Schumacher, Stefan (2000). The Historical Morphology of the Welsh Verbal Noun. Maynooth: Dept. of Old Irish. — (2004). Die keltischen Primärverben. Ein vergleichendes, etymologisches und morphologis- ches Lexikon. Unter Mitarbeit von Britta Schulze-Thulin und Caroline aan de Wiel. Innsbruck: Institut für Sprachen und Literaturen. Schwyzer, Eduard (1939). Griechische Grammatik. Vol. 1: Allgemeiner Teil: Lautlehre – Wortbildung – Flexion. München: Beck. Seebold, Elmar (1970). Vergleichendes und etymologisches Wörterbuch der germanischen starken Verben. The Hague: Mouton. Sommer, Ferdinand (1914). Handbuch der lateinischen Laut- und Formenlehre. 2. & 3. Heidelberg: Winter. Stang, Christian S. (1966). Vergleichende Grammatik der baltischen Sprachen. Oslo, Bergen & Tromsø: Universitetsforlaget. Stifter, David (1997). “Celtiberian -unei, Luguei”. In: Die Sprache 39.2, 213–223. — (2006). Sengoídelc. Old Irish for Beginners. Syracuse, NY: University Press. Stokes, Whitley (1887). “Irish feminine stems in î and u, and neuter stems in s”. In: Zeitschrift für vergleichende Sprachforschung 28, 289–294. Strunk, Klaus (1985). “Flexionskategorien mit akrostatischem Ablaut und die sigmatischen Aoriste”. In: Grammatische Kategorien. Funktion und Geschichte. Akten der VII. Fachta- gung der Indogermanischen Gesellschaft, Berlin, 20.–25. Februar 1983. Ed. by Bernfried Schlerath. Wiesbaden: Reichert, 490–515. Stüber, Karin (2000). “Zur Herkunft der altindischen Infinitive auf -sáni”. In: Münchener Studien zur Sprachwissenschaft 60, 135–167. — (2002). Die primären s-Stämme des Indogermanischen. Wiesbaden: Reichert. — (2007). “Review of: Torsten Meissner, S-stem Nouns and Adjectives in Greek and Proto-Indo- European. A Diachronic Study in Word Formation”. In: Kratylos 52, 38–42. Thurneysen, Rudolf (1887). “IV. Irisch síd”. In: Zeitschrift für vergleichende Sprachforschung 28, 153–154. Tremblay, Xavier (2005). “Zum Narten-Aorist”. Apophonica IV. In: Indogermanica. Festschrift Gert Klingenschmitt. Indische, iranische und indogermanische Studien dem verehrten Jubilar dargebracht zu seinem fünfundsechzigsten Geburtstag. Ed. by Günter Schweiger. Taimering: VTW, 637–64. — (2010). “Jenseits von Schindler? Die Bedeutung der drei Wurzelnomina-Ablauttypen”. In: Münchener Studien zur Sprachwissenschaft 64, 181–221. Uhlich, Jürgen (1999). “Zur sprachlichen Einordnung des Lepontischen”. In: Akten des Zweiten Deutschen Keltologen-Symposiums (Bonn, 2.–4. April 1997). Ed. by Stefan Zimmer, Rolf Ködderitzsch & Arndt Wigger. Tübingen: Niemeyer, 277–304. Untermann, Jürgen (1992). “Wurzelnomina im Lateinischen”. In: Latein und Indogermanisch. Akten des Kolloqiums der Indogermanischen Gesellschaft, Salzburg, 23.–26. September 1986. Ed. by Oswald Panagl & Thomas Krisch. Innsbruck: Institut für Sprachwissenschaft, 137–153. — (2000). Wörterbuch des Oskisch-Umbrischen. Heidelberg: Winter. De Vaan, Michiel A. C. (2003). The Avestan Vowels. Amsterdam & New York: Rodopi. — (2004). “‘Narten’ roots from the Avestan point of view”. In: Per aspera ad asteriscos. Studia Indogermanica in honorem Jens Elmegård Rasmussen sexagenarii Idibus Martiis anno MMIV. Ed. by Adam Hyllested, Anders R. Jørgensen & Jenny H. Larsson. Innsbruck: Institut für Sprachen und Literaturen, 591–599.

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 Notes on three “acrostatic” neuter s-stems 337

— (2008). Etymological Dictionary of Latin and the Other Italic Languages. Leiden & Boston: Brill. — (2005). “Review of: Karin Stüber, Die primären s-Stämme des Indogermanischen”. In: Kraty- los 50, 60–65. Villanueva Svensson, Miguel (2011). “Indo-European long vowels in Balto-Slavic”. In: Baltistica 46.1, 5–38. Vine, Brent (1998). “The etymology of Greek κώμη and related problems”. In: Mír Curad. Studies in honor of Calvert Watkins. Ed. by Jay H. Jasanoff, H. Craig Melchert & Lisi Oliver. Innsbruck: Institut für Sprachwissenschaft, 685–702. De Vries, Jan (1961). Altnordisches Etymologisches Wörterbuch. Leiden: Brill. Wagner, Heinrich (1969). “The origin of the Celts in the light of linguistic geography”. In: Transac- tions of the Philological Society 68, 203–250. Walde, Anton & Johann B. Hofmann (1938–1956). Lateinisches Etymologisches Wörterbuch. 2 vols. Heidelberg: Winter. Weiss, Michael L. (1993). “Studies in Italic Nominal Morphology”. PhD thesis. Ithaca: Cornell University Press. — (2009a). Outline of the Historical and Comparative Grammar of Latin. Ann Arbor: Beech Stave. — (2009b). “Umbrian erus”. In: East and West. Papers in Indo-European Studies. Ed. by Kazuhiko Yoshida & Brent Vine. Bremen: Hempen, 241–264. — (2010). Language and Ritual in Sabellic Italy. The Ritual Complex of the Third and the Fourth Tabulae Iguvinae. Leiden & Boston: Brill. Widmer, Paul (2004). Das Korn des weiten Feldes. Interne Derivation, Derivationskette und Flexions- klassenhierarchie. Aspekte der nominalen Wortbildung im Urindogermanischen. Innsbruck: Institut für Sprachen und Literaturen. — (2013). “Akzent und Ablaut, externe und interne Derivation in der Nominalkomposition”. In: Indo-European Accent and Ablaut. Ed. by Götz Keydana, Paul Widmer & Thomas Olander. Kopenhagen: Museum Tusculanum, 187–195. Winter, Werner (1978). “The distribution of short and long vowels in stems of the type Lith. sti : vèsti : mèsti and OCS jasti : vesti : mesti in Baltic and Slavic languages”. In: Recent Developments in Historical Phonology. Ed. by Jacek Fisiak. The Hague: Mouton, 431–446.

Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33 Bereitgestellt von | Vienna University Library / University of Vienna Angemeldet Heruntergeladen am | 17.12.14 13:33