The Treatment(s) of * -u- after a Coronal in Oscan: Dialect Variation and Chronology

TheevidenceforthetreatmentinOscanof*- u -afteracoronalisexamined.Intheareas which use the Oscan alphabet (Campania and Samnium), this has become [i̯u]; in the areaswhichusetheGreekalphabet(LucaniaandBruttium)ithasbecome[y].Contrary to previous assumptions, there is evidence for a change to [y] in the Latin-alphabet Tabula Bantina from Lucania, since the ⟨i⟩ in the forms petiropert ‘four times’ and manim ‘hand’ismosteasilyexplainedascomingdirectlyfrom*- u -.Evidencefromboth relative and absolute chronology shows that this difference must be a dialectal rather than a chronological split between Campania-Samnium and Lucania-Bruttium, since the different reflexes of *- u - are already in place by the time of our earliest evidence, and are maintained throughout the history of Oscan.


Introduction
Compared to the question of the relationships between the Sabellic languages (on which see lately Clackson 2013, with copious references), rather less atten-Indo-European Linguistics 2 (2014) 112-125 tion has been devoted to the question of the identification of dialect variation within the individual languages; no doubt this is due to the relative paucity of our evidence for the languages, and because, where variation can be identified, this is often due to development over time (e.g. between the Umbrian of the Iguvine Tables i-v, and that of Tables v-vii). Nonetheless, Adiego Lajara (1992: 20-21, 78-79) and Weiss (1998: 708) argue for the existence of dialects in South Picene, and Rix (1996) has identified a range of possible dialect-based differences in Oscan. Oscan is a particularly fertile source for evidence of dialectal variation, because there is a relatively large amount of evidence spread over a wide geographical area (ancient Campania, Samnium, Lucania, Bruttium and Sicily).
In this article I examine the development of inherited *-u-after a coronal in Oscan, and in particular its treatment in the Latin-alphabet Tabula Bantina. I suggest that, contrary to previous claims, there is evidence for a change to *-uafter a coronal in the Tabula Bantina, which suggests that it had developed to [y] as in the Oscan spoken elsewhere in Lucania and Bruttium. Both spelling evidence from the Tabula Bantina, and evidence from the relative chronology of sound changes, suggest that there was a dialectal difference in the treatment of *-u-after a coronal; this became [y] in Lucania and Bruttium, while it developed to [i̯ u] in Campania and Samnium (perhaps via a stage [y]).

Evidence from the Oscan and Greek Alphabets
Inscriptions in both the Oscan and Greek alphabets suggest that a particular change affected *-u-after a coronal consonant1 (Buck 1928: 40;García Ramón 2011: 124 [y], at least in some places, but a mistake or an abbreviation seem more likely. In the form perisstu[leís] (Abella 3/Cm 3) 'of the peristyle' , borrowed from Gk. περίστυλον, the ⟨u⟩ is in fact written y, i.e. a Greek ⟨υ⟩.6 Presumably this suggests an attempt to render Attic-Ionic (koine) [y] in this late loan word, implying that [y] was not a part of the Campanian Oscan phonemic system at the date of this inscription (around 100bc, according to Crawford et al. 2011: 894).
The spelling with ⟨iu⟩ is also found in words borrowed from Greek: these are diumpaís (Teruentum 34 a.7, b.9/Sa 1) 'to the nymphs' ← Doric Greek νύμφᾱ and tiurrí (Pompeii 2/Po 34, Pompeii 3/Po 35) 'tower' ← Greek τύρσις or τύρρις. There are several possible explanations for this: In the case of νύμφᾱ, the vowel in the first syllable would have had the same value [u] as in Oscan (Buck 1910: 25-26), and been treated in the same way. If tiurrí was borrowed from an Ionic dialect, this vowel would have been [y]. If Oscan had no [y] at the time, this could have been borrowed as Oscan *-u-, which subsequently developed to [i̯ u] after a coronal along with inherited *-u-; or it could have been borrowed directly as [i̯ u], in attempt to reflect both the frontedness and roundedness of [y]. If *-u-became [y] after a coronal in the North as in the South, before further developing to [i̯ u] in the North (for which see the next paragraph and section 5), the words could have been borrowed at the stage when Oscan had [y], and then Greek and Oscan [y] were subsequently treated in exactly the same way.

Evidence from the Latin Alphabet
Since the evidence from both the Oscan-and Greek-alphabet inscriptions suggests a development of *-u-to either [i̯ u] or to [y], it would be strange if the same feature were not found in the Oscan of the Tabula Bantina, which is an inscription from Bantia in Northern Lucania dating to the early 1st century bc.7 As noted by Buck (1928: 40), the frequent word for 'money' in the Tabula Bantina is eituam, eituas (Bantia 1.9, .13, .18, .19, .27/Lu 1), not xeitiua-, as might be expected on the model of eitiuva-. However, if the Tabula Bantina underwent the development to [y] which we find elsewhere in Lucania, the absence of the spelling ⟨iu⟩ would not be particularly surprising. In fact, I would argue that there is some evidence for precisely this development, which consists of the two words manim (Bantia 1.24/Lu 1) 'hand' and petirupert, petiropert (Bantia 1.14, 15/Lu 1) 'four times' .8 As shown by Latin manus, -ūs, South Picene abl. pl. manus (Asculum Picenum 2/ap 2), Umbrian abl. sg. mani (it iia 32), mani (it vib 24), loc. sg. (with postposition *en) manuve (it iib 23) we would expect manim to be the accusative of a u-stem, and to go back to *manum.9 The usual explanation for unexpected -im in this word is due to levelling from the ablative, where *-ūd gave -id regularly, as demonstrated by castrid (Bantia 1.8/Lu 1) 'head (?)' (Buck 1928: 132, followed by e.g. Bottiglioni 1954: 117, Wallace 2007: 20-21 and Tikkanen 2011. Exactly what -id represents is not entirely certain: some scholars maintain that *-ū-became [ī] in Oscan in some contexts, but remained [ū] elsewhere; others recognise a development of *-ū-to [ȳ] everywhere,10 which could be written with ⟨u⟩ or with ⟨i⟩ in the Oscan alphabet, and which was written with ⟨i⟩ in castrid (Buck 1928: 41;Meiser 1986: 53;Seidl 1994: 349-351;and see Martzloff 2006: 116-117). Apart from castrid, the evidence is extremely slim, consisting only of fruktatíuf (Abella 1 a.21/Cm 1) 'produce' < *bhrūg-and tiium (Capua 34/Cp 37), tiú(m) (Saepinum 2/Sa 31) 'you' < *tū-om. If the change were to [ī], this would provide an additional reason for the generalisation of the vowel of -id to the accusative singular in *-um, since this would set up a closer parallel with the i-stems, which also had an ablative in -id and an accusative in -im (Poccetti 2002: 50-52, who sees this development as affecting both Oscan and Umbrian). Such a possibility cannot be denied, but if 7 On the date of the Tabula Bantina see Crawford (1996: 274-276). 8 The second vowel of petirupert, petiropert would normally be expected to have been lost by syncope; presumably it was restored by analogy with pettiur. 9 Umbrian also has acc. pl. manf (it iia 38), with a consonant stem ending (if not a mistake for manuf ;Untermann 2000: 450 the u-stem acc. sg. in -im were due to analogy with the i-stems, it is peculiar that this analogy did not affect the gen. sg., as shown by castrous (Bantia 1.3/Lu 1).
Since the i-stem gen. sg. in -eis was also shared with the o-and consonant stems, we might expect that if remodelling of the u-stems had occurred, it would have affected the gen. sg. as well as (if not before) the acc. sg. The alternative possibility would be to see the ⟨i⟩ of manim as reflecting the regular result of *-u-after a coronal elsewhere in Oscan, as proposed by von Planta (1892Planta ( -1897. Such an interpretation would also have the advantage of explaining the other attested accusative singular of this word, which is usually read μανομ̣ (Buxentum 1/Lu 62; thus Crawford et al. (2011) and all previous editions). However, having examined the inscription (Soprintendenza per i Beni Archeologici di Salerno Avellino Benevento e Caserta, Salerno, 11/04/2012), I conclude that the existence of the final ⟨μ⟩ is extremely doubtful, notwithstanding the comment "[l]eft-hand hasta only of final μ visible" (Crawford et al. 2011(Crawford et al. : 1330. The broken edge of the tablet occurs shortly after the ⟨ο⟩ of μανο[, and runs through a circular flaw in the bronze, of a sort that occurs elsewhere on the tablet (e.g. on side b.4, above the ⟨α⟩ of αυτ, and b.5, just before ]ις); this is visible on the photographs in Crawford et al. (2011) and earlier editions, and is not the "hasta" seen by Crawford et al. (2011). Just below the flaw is the end of a thin line running diagonally upwards from left to right, which is presumably the "hasta". However, this line is much more lightly scratched than the other letters, and only visible in certain lights, and in my view is just as likely to be a flaw or the result of the damage inflicted on the tablet when it was broken. As described above, in the Greek alphabet, [y] can be spelled the same as [u] (i.e. as ⟨ου⟩ and ⟨ο⟩), and in Buxentum 1/Lu 62, ⟨ου⟩ is the usual way of spelling [u]. Consequently, for this word I would read μανο [υμ], which would follow the usual spelling conventions of this tablet and the Greek alphabet inscriptions more generally. If one reads μανο[μ], it is necessary to posit two sources for the Oscan forms: either an i-stem for manim and a consonant-stem for μανο [μ] (which at least has the support of Umbrian manf),11 or separate analogical generalisations of the i-stem and o-stem accusative singular endings respectively, as assumed by Poccetti (2002: 51, and passim). The reading μανο[υμ] has the great advantage of allowing both it and manim to represent a single form [manym] from the u-stem which is the best-attested noun-class for this word in Italic.12

11
I am grateful to an anonymous reviewer for pointing out that μανο[μ] could be a consonant stem. 12 In fact, even if μανομ̣ were the correct reading, this would not necessarily rule out zair Indo-European Linguistics 2 (2014) 112-125 As for petiropert, petirupert 'four times' , these are clearly to be analysed as the neuter plural of pettiur 'four' (Aufidena 1/Sa 17) followed by -pert 'times' . One explanation for the spelling of the vowel of the second syllable with ⟨i⟩ is that petiropert reflects a preform *petrii̯ ā > *petirii̯ o > *petiri̯ o > *petirro (Thurneysen 1893: 565 fn. 3;Buck 1928: 52, 66, 139). But motivating such a form turns out to be rather an involved business. First of all, it has to be assumed that *peturā was remodelled to *petrii̯ ā after *trii̯ ā 'three' , which is not attested in Oscan (although reconstructed in τρι[οπερτ], Buxentum 1 a.3/Lu 62), but is found in Umbrian triia 'three' (it iv.2), triiuper (e.g. it ib.21) 'three times' . Then, after the usual Oscan anaptyxis to give *petirii̯ o, *-ii̯ o has to become *-i̯ o in a more or less ad hoc development,13 before *-i̯ -is lost after -r-(a development for which there is some evidence in Oscan; cf. víkturraí Teruentum 20/Sa 24 < Lat. Victoria). A far more straightforward approach is simply to accept that petiropert, petirupert is the regular result of *peturā-> [petyro-] (von Planta 1892-1897: 1.126).
If manim and petiropert, petirupert are correctly analysed as containing [y], which seems much simpler than the alternative proposals, how are we to understand the spelling of eitua-with a ⟨u⟩? There are two possible explanations. The first is that both ⟨u⟩ and ⟨i⟩ are being used to represent [y] as the reflex of *-u-after a coronal. This would be because, in the absence of a letter ⟨y⟩ in the Latin alphabet (not yet borrowed from the Greek alphabet), the writer of the Tabula Bantina did not have a separate letter to represent this sound, and chose whichever of ⟨i⟩ = [i] or ⟨u⟩ = [u] sounded to him closer to [manym]. Firstly, the scribe could have simply missed out ⟨υ⟩ (there are several other mistakes in this tablet). Secondly, as already noted, ⟨ο⟩ is also attested as a spelling for [y] in other inscriptions, and consistency in spelling is not a characteristic of this tablet (the Oscan vowel [e] is spelled ⟨ε⟩, ⟨ει⟩ and ⟨ι⟩). Although there are a large number of instances of ⟨ου⟩ for [u] in what remains of Buxentum 1/Lu 62, this does not rule out the possibility that ⟨ο⟩ was also used: it may simply be a coincidence that the fragment that we have happens to present a large number of uses of ⟨ου⟩. Indeed, there may be an example of the use of ⟨ο⟩ for [u] in ρεκος (Buxentum 1 b.5/Lu 62), whose context suggests comparison with ⟨ f ⟩acus of the Tabula Bantina (Bantia 1.30/Lu 1) (Poccetti 2002: 59-63). Poccetti tries to explain the form ρεκος as reflecting [-os] by means of a highly implausible series of analogical remodellings, but if ρεκος is rightly compared with facus, it is far more likely that this is simply to be seen as an example of the use of ⟨ο⟩ for [u] in this inscription. 13 The sequence *-ii̯ V-was reduced to *-i̯ V-in Oscan, but at a far earlier stage (the development is shared with Umbrian, and took place prior to final syncope, which is also a shared feature with Umbrian; Buck 1928: 66-67;Meiser 1986: 63). Hence Buck's description of the proposed development of *petirii̯ o > *petiri̯ o as "a local change of vocalic i" (Buck 1928: 67); it is not clear that the same change can be seen in zicolom 'day' < *di̯ ēkelo-< *dii̯ ē-kelo-, as implied by Buck: see Untermann (2000: 868-869) and Rix (2004: 496-497

Relative and Absolute Chronology
Relative and absolute chronology suggest that the change of *-u-after a coronal was fairly early in the pre-history of Oscan. In Oscan, *-ō-and *-u-fell together in quality to give [u], and at some point it seems likely that distinctions of vowel length were lost outside initial syllables in Oscan, which had taken place at least by about 300bc (Lejeune 1975: 244-245).17 However, the spelling ⟨iu⟩ is never found after a coronal followed by original *-ō-, even in a non-initial syllable, e.g. regatureí 'ruler' (Teruentum 34 a.12, b.15/Sa 1) < *regatōr-, líkítud (Abella 1 b.10-11/Cm 1) 'it is permitted' < *likētōd. Consequently, we can deduce that *-uhad already begun to change before loss of vowel length in non-initial syllables.
14 If *-ū-gave [ȳ], this is paralleled by the use of both ⟨i⟩ and ⟨u⟩ in the Oscan alphabet. One could also compare the spelling with ⟨u⟩ and ⟨i⟩ of vowels in non-initial syllables before labials in Latin in forms like optumus/optimus, although in this instance these vowels had probably become [ǝ] (Weiss 2009: 118). 15 I owe this suggestion to an anonymous reviewer. 16 This ingenious suggestion is also owed to an anonymous reviewer. 17 The evidence for this claim consists of the fact that we almost never find double writing of vowels to mark length outside initial syllables. There are three exceptions: trístaamentud (Pompeii 24/Po 3) is an adapted loan-word from Latin; in αfααματεδ (Potentia 9/Lu 6), αfααμα[τεδ] (Potentia 10/Lu 7), the long vowel is presumably analogical on the simplex faamated (e.g. Pompeii 2/Po 34). The shortening may have taken place much earlier than 300bc, but it can only be identified once the custom of using double letters to represent long vowels begins. Our earliest instance of double letters for long vowels is perhaps fiis[ias] (Capua 7/Cp 16), which is dated to 325-300 bc. zair Indo-European Linguistics 2 (2014) 112-125 Our inscriptional evidence backs up this claim to an early change, and allows us to be more precise about the date. The earliest usage of ⟨υ⟩ in Oscan written in the Greek alphabet is συπ (Metapontum 1/ Lu 37), which belongs to c. 400-375bc, while ⟨iu⟩ is already in use in niumediis (Bouianum 116/Sa 27, c. 325 bc) and the abbreviation ni is already found in the iúvila inscriptions which use the unreformed alphabet and hence belong to before c. 300 bc (Capua 7/Cp 16, Capua 14/Cp 26). On the other hand, we do not find ⟨iu⟩ in supruis, supṛ[us (Capua 34.7, .10/Cp 37) 'above' < *supro-(cf. Latin super 'above'),18 which is to be dated early since it uses the unreformed Oscan alphabet; its consistency in avoiding double letters for geminate consonants and long vowels also suggests an early date (Buck 1928: 100).19 It seems possible that the adoption of ⟨iu⟩ to spell [i̯ u] after a coronal may have taken place towards the end of the fourth century bc (or a little before, given the establishment of ni as the standard abbreviation for niumsis by this time). The absence of ⟨iu⟩ in Capua 34/Cp 37 would then be an 'old-fashioned' feature of the orthography of this inscription, along with the absence of double letters.20 18 Possibly also in turumiiad, whose etymology and meaning are, however, very uncertain (Untermann 2000: 778). 19 This inscription is dated by Crawford et al. (2011) to 200-150bc, on the basis of Bücheler (1878: 74-76). According to Bücheler the archaeological evidence does not allow a dating prior to the third century bc. If this dating is reliable, we should posit a date as early as possible in the third century, i.e. around or just after 300. Bücheler's attribution of the inscription to the first half of the second century rests on very dubious assumptions about its linguistic idiosyncracies, and cannot be relied on. 20 The use of ⟨u⟩ for ⟨iu⟩ may have continued as a possibility, on the basis of sup (Teanum Sidicinum 34/Si 1b, 150-100 bc). This may come from *sup (cf. συπ, Lat. sub), but this inscription consists of several broken blocks of badly damaged tufa, with numerous natural holes. Consequently, the identification of sup as 'below' cannot be confirmed by context, and it may instead be part of a longer word supi[. A preform *sōp-therefore cannot be ruled out. On the basis of the photographs in Crawford et al. (2011: 572) it also seems at least possible that ⟨u⟩ should be read ⟨ú⟩ (there is an off-centre hole above the ⟨u⟩, which looks much the same as that which forms part of the ⟨ú⟩ in the preceding tríí]ḅúm).

Evidence for Dialect Variation
The identification of [y] < *-u-after a coronal in the Tabula Bantina suggests a dialectal connection with the Oscan spoken elsewhere in Lucania and Bruttium, which had [y] already in the early fourth century; while in Samnium and Campania we only have good evidence for [i̯ u]. One of the explanations for the use of ⟨u⟩ in eitua-discussed in Section 3 implied that [y] in Lucania and Bruttium was a later development of the [i̯ u] attested in Campania and Samnium, and this is the development assumed by Lejeune (1970: 296-299 -u-to [y] in Lucania by c. 400-375bc (which was still in existence in the Tabula Bantina in the early 1st century bc); by comparison, the old-fashioned spelling with ⟨u⟩ was still possible in Campania around the beginning of the third century, which suggests that the development to [i̯ u] had taken place not too long before (the spelling ⟨iu⟩ is found in Samnium around 325). If the change were *-u-> [i̯ u] everywhere, it would have to have taken place early in the 4th century, and the further development to [y] in Lucania would have to have taken place very quickly indeed.

Conclusion
Original *-u-after a coronal in Oscan gave [y] in Lucania and Bruttium, written with ⟨υ⟩ as well as ⟨ου⟩ and ⟨ο⟩ in the Greek alphabet, and with ⟨i⟩ and ⟨u⟩ in the Latin alphabet of the Tabula Bantina. In Campania and Samnium *-udeveloped to [i̯ u], written ⟨iu⟩, perhaps also via a stage [y]. The treatment of *-u-after a coronal thus represents a dialect boundary between Northern and Southern varieties of Oscan. Relative chronology shows that the change to [y] took place early in the history of Oscan; the terminus ante quem for this change is our earliest example of ⟨υ⟩, from 400-375bc. The development to [i̯ u] in the North had taken place by the end of the 4th century bc, but perhaps not long before, since one early 3rd century inscription still uses ⟨u⟩ rather than ⟨iu⟩.