Kluge's Law in Italic? (was: Volcae and Volsci)

From: dgkilday57
Message: 68402
Date: 2012-01-23

--- In cybalist@yahoogroups.com, Miguel Carrasquer Vidal <miguelc@...> wrote:
>
> On Sat, 29 Mar 2008 01:04:50 -0000, "Anders R. Joergensen"
> <ollga_loudec@...> wrote:
>
> >--- In cybalist@yahoogroups.com, Miguel Carrasquer Vidal <miguelc@>
> >wrote:
> >>
> >> Henry Lewis and Holger Pedersen in "A Concise Comparative
> >> Celtic Grammar" (1937, 3rd. ed. 1974) posit Kluge's law for
> >> Celtic.
> >
> >Yes, and Lühr published a long article on the matter
> >(Sprachwissentschaft 10 (1985): 274-346). Getting the PIE accent
> >involved may of course help with the numerous counter-examples, but it
> >will be hard to control, given that we have no other trace of PIE
> >accent in Celtic.
>
> Except perhaps Dybo's pretonic long vowel shortening law.
> Another tricky one, but there must be some truth in it.
>
> The fact that the law (Kluge's, I mean) worked in exactly
> the same way in Germanic (also in pretonic position only)
> makes it all the more plausible that the same conditioning
> applied in Celtic. In fact, I find it very hard to avoid
> thinking that it is the _same_ law in both Germanic and
> Celtic.

This thread from 2008, in which evidence was presented for Kluge's assimilation in Celtic, got me thinking that the peculiar present stem of Latin <mitto:> 'I send', against <mi:si:> 'I sent' and <missus> 'sent', might be analyzed the same way. Old Latin <cosmittere> (Paul. Fest.) for <committere> 'to join, entrust to, commit' shows that the Proto-Indo-European root began with *sm-. It has been identified by some as *smeit- 'to throw', recognized in Avestan <mae:þ-> 'to throw', <hamista-> (*ham-[h]mista-) 'downcast, oppressed', etc. (Pokorny, IEW 968). The fricative in <mae:þ-> suggests that the Iranian words in fact reflect PIE *smeith2-, from which Lat. *mititus rather than <missus> would be expected, and the gemination in <mitto:> is not adequately explained.

De Vaan (Et. Dict. of Lat. and the Other Itc. Lgs. s.v., 2008) dismisses <cosmittere> as untrustworthy and argues instead for derivation from PIE *meith2- 'to exchange, remove'. This root is supported by Sanskrit <methete> 'he becomes hostile, quarrels' (i.e. 'exchanges blows'), Germanic *maidaz 'changed, abnormal' (Old English <gema:d> 'insane, mad'), Lat. <mu:to:> 'I (ex)change, remove', and other words. De Vaan places South Picene <meitims> nom. sg., <meitimúm> acc. sg. 'monument' here also. Again however the root-final laryngeal should give Lat. *mititus; there is no basis for dropping it to get *meit- going into Proto-Italic. De Vaan at least recognizes the difficulty of deriving <mittere> from *mi:tere by the so-called littera-rule (i.e. -V:C- replaced by -VCC-, dialectal as explained below) since there is no trace of *mi:tere.

Under my hypothesis, the pre-tonic PIE dental clusters *-tn-´, *-dn-´, and *-dHn-´ were all reflected as Proto-Italic *-tt-, and similar pre-tonic clusters with labial and velar articulations yielded *-pp- and *-kk-. From a PIE root *smeid-, a zero-grade /n/-suffixed present *smid-n-óh2 thus became Proto-Itc. *smitto:, Lat. <mitto:>. The sigmatic aorist stem *sme:id-s- continued into Proto-Itc., then underwent Osthoff's shortening (i.e. *-V:RC- > *-VRC-) and assimilation to yield the Proto-Latin 1sg. perf. *smeiss-ai, Lat. <mi:si:>, and the passive participle *smid-tó- became Proto-Itc. *smisso-, Lat. <missus>.

The original sense of PIE *smeid- is very likely retained in Modern English <smite>. (Pokorny did not recognize this as a distinct root, and Mallory & Adams cite it as 'smear', clearly a derived sense.) Reflexes include Gothic <ga-smeitan> 'to strike', <bi-smeitan> 'to besmear'; Old English <smi:tan> 'to strike, smite, injure', <be-smi:tan> 'to besmear'; Old High German <smi:zan> 'to strike, hit', <bi-smi:zan> 'to defile'; Modern HG <schmeißen> 'to throw, hurl; fling (a door) shut, slam'. Derivatives such as OE <smitte> 'spot, stain, smut', Middle HG <smitze> 'id.', OE <smittian> 'to blemish, infect' suggest an intermediate sense, 'to afflict (crops) with blight', between 'to strike' and 'to besmear' in the older Germanic languages. The Italic sense-development leading from *smid-n-óh2 to <mitto:> would be something like 'strike (with a missile)' > 'throw (a missile), shoot (an arrow)' > 'send out'. The sense of <schmeißen> renders such a development plausible.

An example of this type of present stem in Germanic is *bakko: beside the more usual *bako: 'I bake', reflected in Old High German as <backen> (Berth. v. Regensb.) beside <bahhan>, the latter corresponding to Old English <bacan>. Both stems reflect the zero-grade of PIE *bHeh3g- 'to bake, roast' found in Greek <phó:go:> 'I roast'. The more usual Gmc. stem evidently comes from *bHh3.gé/ó-, the less usual from *bHh3.gné/ó- with Kluge's assimilation. In the dialects which had the longer stem, the consonantism spread from the present to the past participle, then to the preterit. Luther used <backen> with the preterit <buch>, and several old dictionaries of MnHG have <backe>, <buch>, <gebacken> as the principal parts. When this matter was investigated by Paul (PBB 9:583-4, 1884), the dialect of Ruhla still had <buch> in use beside <buk>. (Later, of course, <buk> found itself in a struggle for existence with the weak preterit <backte>.)

It is not immediately obvious why 'bake' should have had two PGmc present stems. Perhaps the nasal originally distinguished a progressive present from an ordinary present which could refer to habitual action, so that *bHh3.gnóh2 meant specifically 'I am baking (something) now', while *bHh3.góh2 could also mean 'I bake (something) regularly'. But the operation of Kluge's Law would have rendered the distinction between *bakko: and *bako: opaque, soon to be replaced by other means of distinguishing progressive action. By chance some OHG dialects retained the marked form as general.

Nasal suffixation of the zero-grade root is much less common than infixation in forming present stems. Greek <dákno:> 'I bite' (PIE *denk^-, *dn.k^-n-) and the Old Latin 3pl. <danunt> 'they give' (PIE *deh3-, *dh3.-n-) apparently have it. Both are regarded as recent analogical formations by Sihler (New Comp. Gr. of G&L §§466, 488A, 1995). However, the same author suggests that the nasal infix was originally a suffix followed by a root-extension (§453). Noting that PIE *jeu-, *jeu-h{x}-, and *jeu-g- all mean basically 'join, link', he explains the well-attested present stem *ju-né-g-/*jung- 'yoke, harness' as a zero-grade basic root followed by two suffixes. Once such a pattern was established in late PIE, it could be extended analogically to other suitable roots which did not carry extensions, and indeed we see a great efflorescence of nasal infixation in Indo-Iranian. Therefore, nothing forbids us from regarding <dákno:> and <danunt> as archaisms. Neither root was suitable for /n/-infixation, and by chance their /n/-suffixed presents survived into historical times.

When the PIE accent preceded the type of cluster in question, Latin shows assimilation of the stop to the nasal. Thus, <annus> 'year' reflects *h2ét-no- or *h4ét-no- (Gothic <aþnam> dat. pl. 'years'; cf. Skt. <átati> 'goes, wanders'), <somnus> 'sleep' reflects *swép-no- (OE <swefn>), and <dignus> 'worthy' reflects *dék^-no- (cf. <decet> 'it befits'; -gn- is written for [-Nn-], and raising of *-eN- to -iN- is regular). Most Italicists agree with Brugmann (IF 17:492, 1904/5) that Proto-Italic *atno- 'year' became *akno- in P-Italic, reflected as the Umbrian acc. pl. <acnu> and the Oscan loc. sg. <akeneí>, gen. pl. <akunum> (with regular anaptyxis). The principal exception was Buck (Osc. and Umb. Gr. §159.a, 1928), who doubted P-Italic *-tn- > *-kn- on the grounds of his assignment of Osc. <Patanaí> to Proto-Italic *Patna:- (discussed below) rather than *Patena:- (von Planta, Gr. der o.-u. Dial. §261.6, 1897). At any rate, the Latin data suffice to show that the assimilation hypothesized above for <mitto:> did not occur when the PIE accent preceded the stop-nasal cluster.

Under my hypothesis PIE *-tn- thus has either of two outcomes in Latin: -tt- if the PIE accent immediately followed, -nn- otherwise. Against this stands a much older and more popular hypothesis that PIE *-tn- can yield Lat. -nd- (Thurneysen, KZ 26:301-14, 1883; Sihler, NCG §222.2). As originally formulated by Thurneysen, *-tn- > -nd- was independent of original accent, the apparent exception <penna> 'feather' was referred to *pet-sna: rather than *pet-na:, and the gerundive was built on the present participle, e.g. <amandus> from *amant-nos. Subsequent scholars have rejected this explanation of the gerundive, and Go. <aþnam> renders *at-snos an unreasonable makeshift for <annus>, but some authors continue to support *-tn- > Lat. -nd-. Specifically, it has been applied for over a century to explain Lat. <pand-> in terms of earlier *patn-, which was allegedly retained in P-Italic.

By a remarkable feat of oversight, the foregoing scheme has been used to justify deriving the Lat. gerund and gerundive in -nd- from the oblique stem in *-tn- of a universally productive /r/n/-heteroclitic formation parallel to <iter> 'going, road, way' (Sihler, NCG §568.3). But the same gerundive appears in P-Italic with regular assimilation as -nn- (in Umbrian usually written -n-). Since this formation is clearly Common Italic, its derivation from *-tn- cannot be defended on the basis of a later sound-change allegedly affecting Latin but not P-Italic, as already pointed out by Buck (Stud. in Cl. Phil. 1:161, 1895). Apart from the phonetic anachronism, this explanation of the gerund(ive) suffers from other weaknesses. It presumes that heteroclites were productively formed in Proto-Italic as in Hittite, indeed one for every verb. But this declensional class is vestigial in Italic. Latin has <iter>, <jecur>, and <femur> in good use, and P-Italic attests the heteroclites for 'fire' and 'water' which Latin lacks. A universal Proto-Italic formation should have left more than a handful of odd relics. Moreover, each heteroclite originally had two oblique stems. With <iter> itself there is good evidence that the stem *h1itn- proper to the gen. sg. was replaced by *h1iten- proper to the loc. sg. (as explained below regarding Umbrian <une>) going into Proto-Italic. The lack of a satisfactory account of the Italic gerund(ive) does not justify abandoning logic and ignoring facts to construct an implausible makeshift.

With the gerund(ive) dismissed, it is certainly odd that the alleged Latin soundlaw *-tn- > -nd- has quotable examples only for a very specific phonetic shape. It has been claimed that Lat. <pando:> 'I stretch out, spread out, expand' comes from *pat-no:, a form corradical with <pateo:> 'I stretch out, extend, lie open'. To understand how these verbs are actually related, we first notice the connection between Grk. <lantháno:> 'I escape notice, elude' and Lat. <lateo:> 'I lie hidden, avoid detection'. The Grk. aorist <élathon> requires a zero-grade *lh2dH- or *lh4dH-, circum-nasalized in the present. That is, the basic root *leh2/4- has a /dH/-extension reflected in this Greek verb. Latin <lateo:> on the other hand is a stative built on the participle *lh2/4.-tó- 'hidden'. If <pando:> and <pateo:> are similarly related, the basic root is not *pet- but *peh{x}-, with the laryngeal to be determined.

De Vaan (EDL s.v.) notes the difficulty in connecting Lat. <patior> 'I undergo, experience, suffer' with Grk. <pêma> 'suffering' (the latter reflecting PIE *péh1-mn.). Since <patior> cannot be a denominative to PIE *ph1.-tó- or *ph1.-tí-, he suggests an alternative connection with *pet- 'to fly, fall' which is much more difficult semantically. I believe the connection with *peh1- is correct, but the *-t- does not come from a nominal formation. We find an explicit /t/-extension in a few verbs denoting customarily prolonged activity. The two most obvious are Lat. <pecto:> 'I comb, card (wool), Grk. <pektéo:> against <péko:>, <peíko:> (from *pék^-joh2), Lithuanian <pe.sù>; and Lat. <plecto:> 'I plait, braid', OE <fleohte> against Grk. <pléko:>. As a slightly less transparent example, watching over animals or children for long periods of time requires feeding them, so PIE *peh2-t- 'to protect for long' acquired the specialized sense 'to feed, graze, nourish' against *peh2- 'to guard, protect'. But obviously this prolongative /t/-extension was no longer productive during the recorded history of any IE language, and deverbatives of archaic formation incorporate it. Lat. <pecten> and Grk. <kteís> 'comb' reflect something like nom. *pék^tn., gen. *pk^tnós, loc. *pk^tén(i), variously remodelled, from *pek^-t- 'to comb for long'. And the root-noun *nókWts 'night' is based on *nekW-t- 'to stay dark' against the simplex *nekW- 'to get dark' reflected in Hitt. <ne-ku-zi> 'it becomes evening', whose alternant <ne-ku-uz-zi> appears to show *nekWt- losing its specific prolongative force and ousting *nekW-, just as *pek^t- and *plek^t- ousted the corresponding /t/-less verbs from Proto-Italic.

I propose that PIE *peh1- '(to be) open, exposed' with the prolongative extension became *peh1t- 'to stand open, stay exposed', effectively a stative root, typically used in the sense 'to be exposed to harm'. To this was built a /j/-present *ph1.t-jé/ó-, originally an inceptive 'to become exposed', later acquiring durative capacity and rendering the stative obsolete. A morphological parallel is the transitive stative *keh1p- 'to hold up' with its inceptive *kh1.p-jé/ó- 'to pick up', reflected in Latin as <ce:pi:> 'I took' and <capio:> 'I take'. (The English cognate is not 'have' but 'heave'; Lat. <capio:>, <capis>, <capit> correspond to OHG <heffu>, <hevis>, <hevit>, Streitberg, Urgerm. Gr. §206.B.a.I.2, 1896.) In this view, Lat. <patior> was originally the middle voice of the inceptive of *peh1t-, 'I expose myself (to harm)' > 'I suffer, undergo, experience'.

Now, regarding <pando:>, de Vaan (EDL s.v.) follows Schrijver in a modification of Thurneysen's theory. Taking the basic root implicitly as *pet-, he supposes that a circum-nasalized zero-grade *pn.t-n- was internally voiced to *pn.dn-, then acquired anaptyctic /a/ (according to a rule of Schrijver, *CCCC > *CaCCC), and the resulting *pandn- was reduced to *pand- (like Thurneysen's participial gerundive *-nt-n- > *-ndn- > *-nd-). The problem I have with this is that Greek actually has circum-nasalized zero-grade present stems (like <lantháno:> above), and the Sievers-Edgerton third resonant allophone is triggered, in this case *-n.n- > Grk. -an-. I would thus expect *pn.t-n- to become Proto-Italic *pentan-, Lat. *pentino:, and nothing of the sort is known. De Vaan regards *pat- as a secondary full grade developed in Italic from the zero-grade *pt-, but the details of this development are unclear, and it is odd that a full grade should appear in the stative <pateo:> and the participle <passus>, allegedly *pat-to-.

I take Lat. <pateo:> as parallel to <lateo:> above, a stative built on PIE *ph1.-tó- 'open, exposed'. And I see <pando:> as parallel to Grk. <lantháno:> above without the post-nasalization, representing the extended root *peh1-dH-. Its (weak) present stem is *ph1.ndH-, and its participle <passus> continues *ph1.dH-tó-, while <pansus> has introduced -n- from the present stem. The perfect <pandi:> (not in good use, but cited by Priscian) also has the present stem. On the basis of <di:-vi:si:> 'I separated' (*dis-we(:)id(H)-s-ai), I would expect *pe:si: as the inherited perfect of <pando:>, and it is not surprising that such a form would be dropped. In sum, <pando:> cannot be derived from <pateo:> or some imagined *patno:. These Latin verbs have the same root *peh1- with different formations, and their dental stops are not equivalent. But it remains to clarify the Oscan and Umbrian forms cited by some scholars in support of the false derivation, and to exclude from Common Italic the stem *patne/o- (which, if supported by evidence, would be very difficult for my hypothesis).

Development of Italic *-tn- to Latin -nd- was implicitly accepted by Buck (OUG §§81, 99.4, 213.2, 233), who rendered as Latin <Pandae> the Oscan dat. sg. <Patanaí>, a divine name which he derived from an Italic feminine stem *Pat-na:-. He also understood the Umbrian gen. sg. <Padellar> as reflecting an Italic derived fem. stem *Pat-no-la:-. Sihler gives *pat-na:- as the source of Lat. <pando:>, which makes no sense at all, since the infinitive is <pandere>. However, Buck's equation of the Roman and Oscan goddesses, which goes back to Mommsen (Unterital. Dial. 135-7, 1850), has some superficial plausibility. Regarding the obscure Panda, we have this from Nonius (63L.):

"Pandere: Varro existimat ea causa dici, quod qui ope indigerent et ad asylum Cereris confugissent panis daretur; pandere ergo quasi panem dare, et quod numquam fanum talibus clauderetur: de Vita Populi Romani lib. I: 'hanc deam Aelius putat esse Cererem; sed quod in asylum qui confugisset panis daretur, esse nomen fictum a pane dando, pandere, quod est aperire.'"

Varro disagreed with Aelius's identification of Panda with Ceres, and represented them as distinct goddesses in his Menippean satire, the Sciamachia, from which Gellius (13:23.4) cited this invocation:

"Ted Anna Perenna, Panda Cela, te Pales
Nerienes et Minerva, Fortuna ac Ceres"

The reading of the first line is disputed. (Dumézil failed to recognize <Nerie:nes> as the genitive of the Sabine goddess <Nerio:>, whose morphology is parallel to the river-name <Anio:>, old acc. <Anie:nem>, and his placement of a comma at the end of the line is wrong. Pales Nerienes is one deity, 'Pales of Nerio'.) Mommsen (Ann. dell'Inst. 20:424, 1848) accepted the foregoing reading and understood Panda Cela as Proserpina. Varro (LL 5:42) also mentioned a Porta Pandana at Rome:

"Saturnia porta, quam Junius scribit, ibi, quam nunc vocant Pandanam."

Solinus (1:13) echoed this:

"Portam appellaverunt Saturniam, quae postmodum Pandana vocitata est."

Festus offered this etymology:

"Pandana porta dicta est Romae, quod semper pateret."

It is more plausible that the gate was named after a nearby temple or altar of Panda, like two other gates (Paul. Fest.):

"Salutaris porta appellata est ab aede Salutis, quae ei proxima fuit."

"Minutia porta Romae est dicta ab ara Minuti, quem deum putabant."

Arnobius (Adv. Nat. 4:3) passed along this folk-explanation:

"Et quod Tito Tatio, Capitolium ut capiat collem, viam pandere atque aperire permissum est, dea Panda est appellata vel Pantica."

Mommsen regarded this goddess as a local deity of the Capitoline Hill, distinct from the Panda characterized as Ceres or Proserpina, whom he also regarded as probably Sabine. This duality needlessly complicates matters. If Panda was a Sabine goddess, her temple or altar on the Capitoline, near the gate which took her name, would naturally be in a Sabine quarter of the city, which would explain the legendary connection with Tatius. Most important for our purposes is the alternate name Pantica. In Umbrian, the change of word-medial *-nt- to -nd- is regular (Buck, OUG §156), and it stands to reason that this also occurred in Sabine. Then <Pantica> can be understood as the native Latin name of this goddess, and <Panda> as the Latinized form of the syncopated Sabine *Pandca, since Latin did not tolerate such a cluster (but Umbrian did, with <totcor>, <todcome>, <todceir> from syncopated forms of *toutiko- 'urban').

For the etymology, the Labbaean gloss <Panda> 'eiré:ne:s theós' i.e. 'goddess of peace' suggests a connection with Lat. <pango:> 'I fasten, fix, agree upon (peace)'. This verb has three perfects in use, the (formally) sigmatic aorist <panxi:>, the (apparent) root-aorist <pe:gi:>, and the reduplicated <pepigi:>. The Greek synonym <pé:gnu:mi> has the passive aorist <epáge:n>, so Lat. <pe:gi:> (for expected *pa:gi:) is usually attributed to remodelling after <fre:gi:> 'I broke', and the PIE root is identified as *peh2/4g^-. Both <pa:ctus> and <pa:nctus> occur as past participles. Obviously the nasal in <pa:nctus> was introduced from <pango:>, like that in <sa:nctus> from <sancio:>. The regular phonetic development of *-Vnkt- was Lat. -V:kt- as shown by <qui:ntus> 'fifth' from *quinctos (thus the archaizing spelling Quinctius), from which <qui:nque> acquired a long vowel by analogy. With <ju:nctus> (for example), the dorsal stop was restored from <jungo:> to *ju:ntus (from earlier *junctos) on the template of <lego:>, <lectus>, and the like. Similarly with <sa:nctus>, getting -c- restored from <sancio:> was supported by <facio:>, <factus>. But the obscure divine name <Pa:ntica>, from *Panctica 'connected with making peace', once its origin had been forgotten, would have had no basis for analogical restoration of the stop from <pango:>. The Sabine *Pandca inferred above must represent an early borrowing from Old Latin *Pa:ntica:, perhaps when Attus Clausus and his 5000 clients joined the Romans ca. 500 BCE. Oscan and Umbrian, and presumably Sabine, developed their *-Vnkt- to -V:ht- as we see with Osc. <saahtúm> 'sanctum', Umb. <sahta>, <sahatam> 'sanctam'. Native Sabine *Pa:htca (or *Pa:(h)dca like Umb. <todcome>) would not have been Latinized as <Panda>.

Skeptics are free to dismiss my attempt at explaining Panda/Pantica, but if they do, they must also reject the use of Panda as evidence for *-tn- > -nd-, since no independent information supports equating Panda with Oscan Patana or Umbrian Padella. Regarding the latter, we have this from Arnobius (Adv. Nat. 4:7):

"Patellana numen est et Patella, ex quibus una est patefactis, patefaciendis rebus [v.l. frugibus] altera praestituta."

And we have this from Augustine (Civ. Dei 4:8):

"Praefecerunt ergo Proserpinam frumentis germinantibus, geniculis nodisque culmorum deum Nodutum, involumentis folliculorum deam Volutinam; cum folliculi patescunt, ut spica exeat, deam Patelanam."

Now, in ridiculing the plethora of divinities in traditional Roman religion (most of which they collected from Varro, whose works were still extant), these two churchmen had no reason to cite Oscan or Umbrian deities. Patella and Patel(l)ana are minor Roman goddesses whose names are transparently connected with <pateo:>, and very likely with Umb. Padella as well. Those who accept Lat. <pando:> from *patno: are inclined to derive <Padella> from Italic *Patnola:, which would make the Latin counterpart *Pandola under their soundlaw. (Going into Umbrian, *Patnola: > *Patn.la: > *Padn.la: > *Padenla: > *Padella: > *Padello:, gen. -a:r, cf. Poultney, Bronze Tables of Iguvium 314, 1959.) I find it more plausible to equate Lat. Patella directly with Umb. Padella on the basis of an Italic adjective *patros 'standing open, patent, manifest', Lat. *pater, the source of <patra:re> 'to make manifest, accomplish' (whose alleged derivation from 'father', though popular, is absurd). Reflecting PIE *ph1.t-ró-, this Itc. *patros is morphologically parallel to OL <sakros> 'sacred', Lat. <sacer>, whence <sacra:re> 'to make sacred, sanctify'. The Italic goddess of opening flowers or whatever would then be *Patrola:, going into Umbrian *Patr.la: > *Padr.la: > *Paderla: > *Padella: > *Padello:, going into Latin *Patr.la: > *Paterla: > *Patella: > Patella.

The phrase in the Iguvine Tables (VIa:14) mentioning Padella is <pertome Padellar>, which Poultney (BTI 236-8) rendered 'to the gate of Padella'. Since he related Padella to Panda, he drew a parallel with the Porta Pandana at Rome. However, Iguvium had only three city-gates, named after the cities to which their roads led. Poultney regarded Umb. *perto- as transferred from the fourth declension to the second, thus continuing PIE *pértu- (whence Gmc. *ferþu- 'inlet, estuary, firth', ON <fjörðr>), and I can find no fault with this etymology. However, the meaning is much more likely to be 'narrow passage, alley' than 'gate', which is the sense of the obliquely related Lat. <porta>. The place mentioned was probably an alley leading to, or passing by, the temple or altar of Padella at Iguvium.

Twice on the Tablet of Agnone we find the dative phrase <Patanaí Piístíaí>, which Buck renders 'Pandae Fidiae'. He capably explains the epithet (OUG §21) as a borrowing from Greek <Pístios>, noting that <Zeùs Pístios> renders 'Juppiter Fidius' (Dion. Hal. 4:58), and suggests that the Oscan -ií- is due to contamination with the word for 'pious', which occurs on the Tablet in the masc. dat. sg. as <Piíhiúí>. The Tablet has good examples of regressive anaptyxis following light syllables, e.g. <Patereí> 'to the Father', so it is reasonable that the immediate antecedent of <Patanaí> was *Patnai. The crucial question now is whether this continues a Proto-Italic *Patna:-, as Buck favored, or *Patena:-, as von Planta advocated (and Buck admitted as possible, OUG §159.a). The latter requires syncope to precede the Oscan anaptyxis, which is not a problem. Post-tonic */e/ was regularly syncopated in this environment, as we see from Osc. <úpsannam> which corresponds exactly to Lat. <operandam> from Italic *opesanda:m. Buck was aware that refusing to accept *-tn- > *-kn- in P-Italic means rejecting the sense of *akno- as 'year', which fits the texts very well, and his alternative derivation of *akno- from *ak-, supposedly extracted from deverbatives of *ag- 'to drive' (OUG ib.) is not satisfying.

Von Planta supported the morphology *Patena:- with the Sabine gloss <terenum> 'molle' (cognate with, or borrowed from, the neuter of Greek <térenos> 'soft, smooth, delicate', more commonly <tére:n>), the Umbrian ethnonyms <Tesenakes>, later <Tesenocir> abl. pl. 'Tessenacis' (name of a city-gate, pl. tant.) and <Talenate> dat. sg. 'Talenati' (name of a decuvia), and some Latinized place-names possibly containing *-en- as a formant (Sarsina, Fulginia, Fu:cinus, Frusino:). This is admittedly a slim foundation for postulating *-eno- as a productive Italic deverbative. I would add Lat. <sarcina> 'bundle, pack', usually connected with <sarcio:> 'I mend, patch, repair', and possibly identical with <Sarsina>. This town was in Umbria, and the Umb. adverb <sarsite> 'wholly, publicly'(?) has been tentatively referred to Itc. *sarki:te:d 'sewn together' > 'collectively' > 'publicly' (Poultney, BTI 322). Lat. <patina> 'dish, platter' is usually taken as borrowed from Grk. <patáne:>, but <patera> 'shallow bowl' is considered native and referred to <pateo:>. Since <sarcina> is native Latin, <patina> could equally well be native, reflecting Itc. *patena:-, and formally identical to the Oscan goddess above. I identify Italic *-ena:- as the feminine of an admittedly rare PIE agential suffix *-enó- taking zero-grade of the root and reflected in Skt. <kr.p-aná-> 'lamenting, miserable'. Then PIE *ph1.t-enéh2-, Itc. *patena:- could mean 'the female who opens' or 'the object (with fem. gender) which spreads, broad dish, platter'. In Latin and Umbrian, perhaps the wide use of the second sense made renaming the goddess necessary to avoid irreverence, but not in Oscan.

(The corresponding barytone suffix *-eno- takes full grade, e.g. Lat. <pa:gina> 'that to which vines are fixed, trellis', Skt. <vá:hana-> 'conveyance, vehicle'.)

The other Oscan word we need to examine is the third-plural imperfect subjunctive <patensíns> from the Cippus Abellanus, which Buck renders as 'aperirent', alternatively 'panderent'. He gives the Italic verb as *patno: or *pateno: (as if Lat. *patino:), and explains the formation as the /e:/-subjunctive of an /s/-aorist (an Italic innovation, since this aorist originally took a short-vowel subjunctive, but there are other examples). Sihler wants to equate <patensíns> directly with Lat. <panderent> through Italic *patnese:nt, citing also Epic Greek <pítne:mi> 'I spread out' and its Attic counterpart <petánnu:mi> (remodelled after the aorist <epétasa>). In discussing Greek verbs with this morphology (NCG §§472-3), Sihler has no explanation for the /i/ in <pítne:mi>, since according to his analysis the stem was *pt-ne-h2-. As deduced above, however, the prolongative verb-root was not *pet- but *peh1-t-, so the (strong) present stem of <pítne:mi> was actually *ph1.t-né-h4- (the second laryngeal will be covered below), later with the Greek accent retracted to the reflex of *h1 in the strong forms. This ties in very nicely with another problem which Sihler finds intractable (§496.2), the vocalism of the Greek imperative <ísthi> 'be!'. "The question of how attested <ísthi> arose has no answer," says he, "but even so that is an easier question than WHY *ésthi would have been vulnerable to replacement at all." Both questions, as well as the difficulty with <pítne:mi>, vanish if we assume that sonant *h1 regularly became Greek /i/, not /e/, when it was originally pre-tonic but later took the accent by retraction. There is no problem with taking the original form of the imperative 'be!' as *h1.s-dHí, or with taking the original form of <pítne:mi> as *ph1.t-néh4-mi. And the aorist <epétasa> most likely had its original accent on the augment, as in Sanskrit aorists of any length, so that sonant *h1 was post-tonic in *h1é-ph1.th4.-sm. and took the Greek accent by protraction, not retraction. Generalization of the new strong vocalism to the (originally) weak present forms of <pítne:mi> later in Greek is trivial.

(Greek <thésis> 'setting, placement' is not fatal to the rule proposed above, since it was probably inherited from *dHh1.´tis. The Sanskrit cognate <dhitis.> does not have a quotable accent; it is apparently a late analogical formation, replacing *hitis. (cf. <hitás> from *dHh1.tós, Grk. <thetós> 'set, placed'). However, <thésis> is morphologically parallel to <stásis> 'standing, position', whose Skt. cognate is <sthítis.>. Most likely the paroxytonic Greek accent of both nouns is inherited.)

The reason I reconstruct *h4 above, not *h2, is that we have clear evidence of an /a/-coloring laryngeal extension in the PIE verbal root 'fly, fall', both *pet- and *peth{a}-, but the Skt. set.-forms show no aspiration, <patitás>, <pátitum>, etc. In words like <sthítis.> above, PIE *-th2.- regularly yields Skt. -thi-, with the aspiration of /t/ generalized to all forms derived from the same root. Since this does not happen with <patitás>, we must be dealing with *peth4-. Now, I cannot prove that *h2 did not also function as a root-extension, but without positive evidence to the contrary, I will provisionally assume that it did not, and that *h4 is the extension found above in *peh1-t-h4-, nasally infixed as *ph1.t-né-h4-, in <pítne:mi>. The sense added by *h4 probably referred to the completion of physical action, something like this: *pet- 'fall', *pet-h4- 'fall down', *peh1-t- '(slowly) spread', *peh1-t-h4- '(slowly) spread out'. Of course, like other root-extensions, it became obsolete well before the breakup of PIE.

Returning now to Osc. <patensíns>, we should notice that the Cippus Abellanus was executed carefully, with consistent attention to orthographic detail, and this word occurs twice. One spelling error for *patnesíns would be quite remarkable, two such errors unthinkable. Therefore, Sihler's daydream of deriving the word from Itc. *patnese:nt (in order to equate it with Lat. <panderent>) must be excluded. The segment -en- either continues original *-en- or syllabic *-n.-. De Vaan (EDL s.v. <pando:>) opts for the latter, postulating Itc. *pt-n.-s- (where *-s- is the aorist marker, but this gives enough consonants to apply Schrijver's rule, *CCCC > *CaCCC, which is marginally better than an ad-hoc shwa secundum). Since I work with full-grade *peh1-t- rather than *pet-, in my view the Itc. zero-grade was already *pat-, and no shwa-secundal shenanigans are necessary. Among the Oscan examples of post-tonic syncope of */a/ (Buck, OUG §89.2) is the neut. pl. <prúftú> 'placed (boundaries)', from *pro-fato-, identical to Greek <pró-theto-> (§244.1), with Italic */a/ from a sonant laryngeal. I propose that in contrast to the Proto-Greek /n/-infixed present stem *ph1.t-né-h4-, Proto-Italic used an /n/-suffixed stem (as in Gmc. *bakko:) to form the present from the zero-grade of *peh1-t-h4-, then added the /s/-marker to this nasal present to form the aorist stem (cf. Lat. <panxi:>). The 3pl. aorist (later "imperfect") subjunctive would then formally continue PIE *ph1.th4.-n-s-eh1-nt, Proto-Italic *patanse:nt. In Proto-P-Italic, syncope (as in Osc. <prúftú>) would yield a secondary syllabic nasal in *patn.se:nt, and this would regularly yield Osc. <patensíns>.

Latin <damnum> 'loss, damage, penalty' and <scamnum> 'bench, prop, support' pose an apparent problem for the idea sketched in this post. They are traditionally reconstructed as *dapnom and *skabHnom, as though based on the zero-grades of PIE roots *deh2/4p- and *skeh2/4bH-, which would suggest an accented suffix *-nóm. The diminutive <scabellum> shows that *scabnolom already had short /a/ before *-bn- > -mn- in <scamnum>, so Osthoff's shortening cannot be invoked. However, nominal formations in *-no- not equivalent to substantivized passive adjectives can have the accent on a zero-grade root. Thus, Greek <erípne:> 'steep slope' is clearly connected with <ereípo:> 'I dash down', <é:ripon> 'I fell down', the root being *h1reip-, and the PIE noun *h1ríp-neh2. And though it differs in meaning, <damnum> appears formally identical to Old Norse <tafn> 'sacrificial animal or meal'. Had the latter reflected *-nóm, Kluge's Law would have given PGmc *-pp- instead of *-fn-. The sense of PIE *dh2/4.´nom was perhaps something like 'mandatory distribution, recompense' which in Germanic came to denote what man owes to the gods, but in Italic meant originally what one man owes to another whom he has wronged, then came to mean the legal basis for the penalty, the actual damage done. (Old Irish <dúan> 'poem', if it represents Proto-Celtic *dafna: from the same root, does not show Kluge's assimilation and points to *dh2/4.´neh2 under the deductions about Celtic elsewhere in this thread. Perhaps the PCelt word originally referred to a poem of thanks to the gods.) At any rate, these Lat. nouns in -amnum do not disprove Kluge's assimilation for Italic.

The heteroclitic noun 'water' presents another apparent problem. Umbrian has the acc. sg. <utur> (in the old Etruscan-based alphabet for /udo(:)r/; assibilation of -d- to -r^- was inhibited by the following -r) and abl. sg. <une> (for /unne/, assimilated from *udne). This last form, if referred to a loc. sg. *udní, would indicate that Kluge's assimilation did not affect P-Italic and could not have occurred in Proto-Italic. Now, while the PIE gen. sg. of *wódr. 'water' was evidently *udnés (Skt. <udnás>, Hitt. <ú-i-te-na-as^>), the loc. sg. was *udén(i ) (Skt. <udán>, <udáni>). In Gothic the nom./acc. sg. <wato:> represents PGmc *wato:r, a conflation of the expected *watur with *uto:r, this from the PIE collective *udó:r (Skt. <udá:> nom./acc. pl., Hitt. <ú-i-da-a-ar>, Greek <húdo:r> serving as nom./acc. sg.). The Go. dat. sg. <watin> represents PGmc *watini, conflating *watur with *utini from the PIE loc. sg. *udéni. And the Go. gen. sg. <watins>, PGmc *watenez, evidently conflates *watur with *utenez, which cannot directly reflect PIE *udnés, for this would have yielded PGmc *uttes. At some stage of Paleo-Germanic (by which I mean the language before the Grimm-Verner-Kluge sound-shifts and associated phenomena defining Proto-Germanic), the inherited gen. sg. *udnés must have been remodelled to *udénes, acquiring the stem found in the loc. sg. *udéni.

Something similar must have happened in Proto-Baltic (or Paleo-Baltic), with the Lithuanian gen. sg. <vandeñs> against the nom. sg. <vandu°~>. (Nasal-insertion is presumably by analogy with the /n/-infixed verb *unéd-, *und- 'to make wet, soak, moisten', Skt. 3sg. <unátti>, 3pl. <undáti>. Analogical nasalization likewise explains Lat. <unda> 'wave' without recourse to ad-hoc *-dn- > -nd- implied by taking Grk. <halos-údne:> 'sea-wave, billow' as a direct cognate.)

Latin does not attest this word for 'water', but the heteroclite <femur> 'thigh' has the gen. sg. <feminis>, not *femnis. (The nom./acc. sg. <femen> and gen. sg. <femoris> are late analogical creations.) The heteroclites <iter> 'way' and <jecur> 'liver' have introduced the final syllables from their nom./acc. sg. to the other cases, but the new gen. sg. <itineris> and <jocinoris> must be based on older *itinis, *jocinis, again having the internal gradation of the PIE loc. sg., which must have been extended to the gen. sg. at an early date. (The /je-/jo-/-alternation in the 'liver' paradigm appears to be ancient. Avestan <ya:kar@> has evidently generalized /a:/ resulting from Brugmann's lengthening in the loc. sg. *jokWén(i).) Thus there is no problem in assuming that Proto-Italic, like Paleo-Germanic, generalized the internal /e/-grade of the loc. sg. to the other oblique cases of /r/n/-heteroclitic nouns. Nothing suggests that the reverse process, generalization of the internal zero-grade of the gen. sg. to the loc. sg., might have occurred in Italic. The Lat. abl. sg. <femine> is formally a loc. sg., and so presumably is Umb. <une>, representing Proto-Italic *udeni, syncopated in P-Italic to *udni and lowered to *udne in prehistoric Umbrian, then assimilated to /unne/. Post-tonic syncope of */e/ in P-Italic polysyllables is regular; cf. Oscan <úpsannam> 'faciendam' (directly cognate with Lat. <operandam>); Umb. <osatu> 'facito' from *opesa:to:d, <mersto> 'justum' from *medestom (Buck, OUG §88.3). This eliminates any objection to Kluge's assimilation posed by Umb. <une>. Now that the most obvious objections have been addressed, other possible Italic examples can be cited.

Lat. <glittus> 'sticky' is usually considered related to Grk. <gloiós> 'sticky, clammy'; Gmc. *klaijo:, OE <clæ:g> 'clay', but the exact relation is a sticky problem. I suspect we have to do with a PIE verb *gleidH- 'to make sticky' yielding *glidH-nós, Lat. <glittus> 'made sticky'. The root-extension *-dH- (nasalized in <pando:> and <lantháno:> above) apparently makes factitives from adjectival and stative roots. From the adjective *wei- 'apart, separate' a verbal root *weidH- 'to put apart, separate, divide' is derived. From the stative *steh2- 'to be standing' a factitive *steh2dH- 'to place standing' is implied by *sth2dH-mós, Grk. <stathmós> 'standing-place, stable, station'. The Gmc. preterit stem *sto:d- 'stood' (OE <sto:d>, <sto:don>) can be understood as the root-aorist of this extended verb without further ado.

Lat. <lacca> 'swelling on the shin-bone' (Chiron, Vegetius) is perhaps from *lh2/4k-néh2 'injured part', from *leh2/4k- 'to injure, tear, lacerate' (Grk. <lakís>, <lákisma> 'rent, shred, tatter', <lakízo:> 'I rend, tear') also found in *lh2/4k-rós, Lat. <lacer> 'injured, torn, lacerated'. The fact that <lacca> is attested only late does not deny its antiquity, for it is a technical veterinary term.

Sore eyes are frequently rubbed. Possibly Lat. <lippus> 'sore-eyed' continues PIE *h2/4libH-nó- 'rubbed, smeared, anointed' from *h2/4leibH- (Greek <aleípho:> 'I anoint'). Likewise <lippe:s> 'sore eyes' could reflect a noun *h2/4libH-ní- '(frequently) rubbed object' vel sim.

Lat. <siccus> 'dry' can be regarded as the regular reflex of PIE *sikW-nós 'poured out, emptied of fluid, drained', from *seikW- 'to pour out' (Skt. <sécate>, <sincáti>, OHG <si:han>, etc.), without any of the usual etymological makeshifts: "expressive gemination" and the resulting problem of explaining why dryness requires more "expression" than wetness, arbitrary syncope and assimilation of an invented *siticus 'thirsty', ad-hoc fortition of *sisko-, or morphologically unusual *sikW-ko-.

Lat. <bucca> 'cheek', later 'mouth', rather than being a borrowing from an Illyrian language (as I used to think), can be understood as continuing *bukW-néh2 'inflated part, puffy part', from PIE *beukW- 'to blow hard, inflate, puff', reflected in Greek <búkte:s> 'blustering', and in my view OE <pyffan>, Gmc. *pufjan, from the Gaulish zero-grade *bup- borrowed before the consonant-shift, probably in a noun referring to local windy weather. (Possibly the obscure Northumbrian <peuf>, <peufa> continue a borrowing from the Gaul. full grade, with OE <pohha> 'bag, swelling' continuing a direct Gmc. inheritance from PIE *búkWon-. If Nhmb. (Lind.) <pocca> is not merely an error for (Rushw.) <pohha>, it may continue the related pre-shifted borrowing from Gaul. *bucca < *bukW-néh2.) Chiron's veterinary term <buccula> 'swelling on a horse's body' retains the generic sense of the Latin word. Gallo-Latin personal names <Buccus>, <Bucco:>, <Buccio:>, and the like led Ernout & Meillet to consider <bucca> a possible borrowing from Celtic. However, Varro attests <bucco:> 'chatterer, blockhead' as a form of address, and his background was Sabine, not Celtic. We also have <bucco:> used by Plautus, a native of Umbria.

Not all Latin tenues geminatae can be attributed to nasal assimilation. <Juppiter> is obviously a nominalized vocative (cf. Greek <Zeû páter>) in which *-u:p- has become -upp-, which is not a regular feature of Roman Latin and must come from the dialect used by the Sabine priests. That is, we have Sabino-Latin <Juppiter>, <Jovis>, etc. against Roman Latin <Die:spiter>, <Dia:lis>. It is likely that the Sabino-Latin dialect (which must be rigorously distinguished from the Sabine language itself) was also used by many plebeians resident in Rome and many rustics dwelling in Latium, so not all Sabino-Latinisms need represent priestly terms. Other words showing dialectal -VC:- in accented position for etymological -V:C- are <vitta> 'band, ribbon, fillet' for expected *vi:ta (from *wei- 'to turn, twist') and <pannus> 'cloth, rag' for expected *pa:nus (cf. Grk. <pênai> 'aggregate of threads, web'). With <cuppa> 'drinking vessel, cup' against <cu:pa> 'cask, tub' the Sabino-Latin term has developed a sense-distinction. Etymological -u:p- is shown by Sanskrit <kú:pas> 'hole, pit, well' which must have an undetermined laryngeal in the zero-grade root, *kuh{x}p-, so these words should not be ascribed to a root *keup-.

On the other hand <mu:cus> 'mucus' can be understood as a Roman Latin hypercorrection for <muccus>, the latter being the etymologically correct form in both dialects, and derived from *mug-nós, whose nasalized zero-grade appears in <(e:)mungo:> 'I blow my nose'. Otherwise, one must conflate the distinct roots *meuk- and *meug-, leading to a snotty mess in one's etymological dictionary. In my view, the gentilicium <Mu:cius>, with its long and distinguished history at Rome, thus has nothing to do with <mu:cus> and did not originally mean 'Snotty', as some scholars have suggested. Pompeian Latin has <exmucca:vit> for Roman <e:munxit>, and this verb survives in Italian <smoccare>, <smoccolare> 'to snuff' (a candle-wick). Apparently the pool of wax around the wick was reminiscent of snot.

The situation is trickier with <su:cus>, <succus> 'juice, sap, thick fluid'. The sense is compatible with 'that which is sucked', but <su:gere> 'to suck' along with OE <su:can>, Old Irish <su:gim>, and Latvian <su:zu> imply a zero-grade *suh{x}g^-, while OE <su:gan>, OHG <su:gan>, and Old Norse <su:ga> imply *suh{x}g^H-, unless borrowing can be invoked for these Gmc. forms with -g-. Either way, *suh4g^(H)-nós 'that which is sucked, juice, sap' would presumably yield Proto-Italic *su:ccos after assimilation. Evidently <su:cus> has undergone regular Roman Latin degemination after a long vowel and <succus> is the corresponding Sabino-Latin form, like <cuppa> to <cu:pa>. At any rate, Kluge's assimilation allows us to discard the ad-hoc variant *su:k- to Lat. *su:g-.

I have no compelling evidence for Kluge's assimilation in Venetic. The personal name <Bukka> (whether or not it means 'Cheeky', 'Chatterer', or the like) could well be from Illyrian substrate. Such a substrate in Padanian Venetic is shown by the river-name <Plavis> (now <Piave>). This would be *Plovis if it were native Venetic, since inscriptions show PIE */o/ retained as /o/, not converted to */a/ as in Illyrian. In Macedonian, the personal name <Perdíkkas> might be attributed to Kluge's assimilation, if the meaning is 'Outstanding' (literally 'thoroughly shown, shown out'; cf. German <ausgezeichnet>) from *per-dik^-nós, but this is purely speculative. Nevertheless, it seems possible that Kluge's assimilation could have occurred independently several times in the prehistory of the IE languages.

Douglas G. Kilday