What became of "Sabine l"? An overlooked Proto-Italic sound law

What became of “Sabine l”? An Overlooked Proto-Italic Sound Law Blanca María Prósper Universidad de Salamanca indoling@usal.es Up to now, no more than four reliable examples of the change da- > la- have been found, and they have been explained away as due to “Sabine influence” or paretymology. I shall try to show that we are dealing with a sound law that operated as early as Proto-Italic and calls into question the received opinion on the IE stop system. 1 1. Introduction: The Sabines are not to blame As is well known, Latin provides many, not to say most of the habitually mentioned instances of the phonetic confusion of [d] and [l] in the languages of the world. Lateral and dental sounds, both stops and fricatives, share many traits and are consequently prone to occasional merger leading to both synchronic and diachronic variation. The shift [d] > [l], as far as it has been detected in Latin, both in real and fictitious cases, is usually labeled “Sabine l”, and attributed in a rather imprecise fashion to an Italic substrate or dialectal variation, but is always considered as a sporadic phenomenon, by no means a “neogrammarian” sound change. In sum, while all the handbooks feel the need to mention it in passing, no systematic account of its real conditions, nature and scope is ever found anywhere, and the dialect held to have exerted this influence on Latin phonetics is a ghost in spite of much scholarly adornment. Take, for instance, some recent monographs: Negri (1992: 232) calls the product of the contact of Latin with such an Italic dialect “la lingua di Numa”. Coleman (1990) confidently claims: 1 Abbreviations: PItal. = Proto-Italic, Lat. = Latin, O. = Oscan, U. = Umbrian, Ven. = Venetic, Pael. = Paelignian, SP. = South-Picene, part. = participle. I thank the editor of JIES as well as two anonymous reviewers for their comments. Volume 47, Number 3 & 4, Fall/Winter 2019 458 Blanca María Prósper “The only plausible explanation of all this is that original d forms in Latin have been partly replaced by forms from an Italic language which shared with Umbrian the loss of occlusion but had l as the output of the change. The most likely candidate is Sabine.” Finally, the non-committal stance of Meiser (1998: 100) is diagnostic of the case having come to a dead end: “Lautgesetzliche Bedingungen lassen sich für den Wechsel nicht angeben; Einfluß sabellischer Dialekte ist immerhin nicht auszuschließen (das Umbrische kennt in der Tat einen Wechsel 1 >  >
).” Varro (De Ling. Lat. 5, 123) is routinely believed to be the forerunner of the “Sabine l” hypothesis in ancient times, on the tenuous authority of his invention of a Greek form   to account for lepestae, quae etiam nunc in diebus sacris Sabinis vasa vinaria in mensa deorum sunt posita (in fact, although Varro may have been misled by   , the Latin form goes back to   ‘limpet-shaped vase’). This sentence has been taken to mean that he felt there was a correspondence between Greek d- and Sabine l- (see Negri 1992: 242-243). But, as observed by Weiss (2009: 475, fn. 59), the Sabine attribution is a modern myth, never explicitly found in the writings of the ancients. Undoubtedly, the alternation l/d in itself could not be overlooked by ancient authors and commentators, since it occurs in lexically related items, such as odor/ol
re, solium, consilium / sed
re, medulla / melilla and dus, lg; or was recorded as an alternative pronunciation, as in Novensides (Varro, De Ling. Lat. 174) vs. Novensiles (Livy, Ab Urb. Con. VIII, 9, 6, etc.), praesidium vs. praesilium, etc.; the ancients occasionally support the wrong form, or identify actually different forms, as in Paulus ex Festo 81, Lindsay: Delicata dicebant dis consecrata, quae nunc dedicata. Late authors have also been prone (or witness) to hypercorrection, possibly favored by unrelated forms, as in fidium pro filio (Paulus ex Festo 133, Lindsay). Marius Victorinus revealingly omits any reference to a particular dialect: communionem enim habuit <l> littera <cum d> apud antiquos, ut dinguam et linguam et dacrimis <et> lacrimis ... et communio cum Graecis, nos lacrimae, illi   (Gramm. VI 26, 1-5). In modern times, the “Sabine connection” apparently goes back to Conway (1893), who, undeterred by the fact that Varro The Journal of Indo-European Studies What became of “Sabine l”? 459 himself ascribed the variant Novensides (not Novensiles) and fedus (Lat. haedus) to the Sabines, enthusiastically pointed to Sabine as the Italic dialect that was responsible for the change, mainly on the grounds that it stood closer to Latin than Oscan or Umbrian. Some ancient glosses, since discredited, seemed to bear out this assumption. Conway’s work, however, is too messy, eclectic and reliant on disparate evidence, ranging from modern local hydronymy to the internal influence of Latin synonyms or equivalents, to be taken seriously. And, needless to say, we know virtually nothing about Sabine. On the other, hand, the process that should have led to “Sabine l” occurring in Latin forms is predictably tiptoed around: in terms of contact linguistics, one does not quite see why urban or rustic Latin, not directly transmitted by bilingual Sabines, should have adopted this alternative pronunciation for an (arbitrarily restricted) number of forms. What is more, it is quietly assumed that, because the Sabines hypothetically had generalized /l/ for IE /d/ (something intrinsically unlikely), they would turn any synchronic foreign /d/ into /l/, and not, say, into the nearest dental sound, e.g. /t/. If, alternatively, one sticks to the Sabine influence but assumes that we are dealing with loanwords, it is unclear what sort of cultural prestige the Sabines enjoyed that should have led the Romans to replace the equivalent Latin forms at all (and the forms at issue are not suspect of designating products of foreign provenance). Again, all this is unwarranted, and the logic under the expression “Sabine influence” is impressionistic to say the least. In sum, while the name of the hypothesis has caught on for no good reason, its contents remain as non-specific as ever, unrelated phenomena may have been inadvertently pooled together, and, as I shall argue at length below, too much credence has been attached to the generally unreliable opinions 2 of ancient grammarians. 2 See the deleterious reviews of the “Sabine l” hypothesis by Schrijnen (1914: 379), who stated: “Eine volkstümliche, populär-dialektische Umänderung auf eigenem Boden ließe sich vielleicht annehmen” and was duly sceptical regarding Volksetymologie; and much later Bottiglioni (1943) and Poucet (1966), who timidly suggest it could be an internally motivated change but again fail to provide further internal evidence for or against this idea. The question has been most recently addressed by Burman (2018: 51-58), who does Volume 47, Number 3 & 4, Fall/Winter 2019 460 Blanca María Prósper 2. Narrowing down the problem: the context of the sound change In spite of the usual accounts, there are two changes at issue, one taking place in initial and another in medial position, and in different phonetic contexts. Their essential difference has been blurred by the implicit idea that only forms synchronically or diachronically exhibiting both /d/ and /l/ can be sanctified as members of the group of roots showing the “Sabine” change /d/ > /l/. This is perhaps the only problem concerning Latin phonetics and etymology that has never been tackled for its own sake, but systematically weighed against what ancient scholarship had to contribute to it. The reason for this is that ancient grammarians treat the diachronically conditioned alternations as if they were comparatively late, that is to say, not preliterary. It is generally accepted that the first of these changes, by which #da- yields Latin #la-, is reflected in five forms: a) Latin larix ‘larch-tree’ has been related to OIr. dair ‘oak’, but this idea cannot be further substantiated and phytonyms are always suspected of being loanwords. b) Latin dingua (only in Marius Victorinus VI, 26, 2) gives lingua ‘tongue’, for which analogy with ling
and/or tabooistic effects are often invoked, and whose Oscan cognate fanguam (Cumae) calls for prudence. I shall consequently focus on the more reliable cases c), d) and e). c) Latin lacruma, lacrima ‘teardrop’ vs. dacrima, an early variant allegedly used by Livius Andronicus according to Paulus ex Festo 60, Lindsay, goes back to IE *daru- (whatever the previous stages, since this form has inspired a variety of imaginative reconstructions). Paulus reads dacrimas pro lacrimas Livius saepe posuit, nimirum quod Graeci appellant 
. This could be understood as overtly implying that Livius Andronicus was hypercorrecting, but Festus and his sources seemingly tended to label as Greek any forms for which they could find a plausible Greek cognate. Ennius, a slightly 1 later author than Livius, wrote lacrumis. The often quoted her best to disentangle the terminological mess incurred by most of the former scholarship but finally dismisses the problem as hopeless. The Journal of Indo-European Studies What became of “Sabine l”? 461 variant dacrumis does not exist as far as I know: Bergk (1859: 187) arbitrarily corrected Ennius’ text nemo me lacrumis decoret into nemo me dacrumis decoret in order to recreate an alliterative effect purportedly present in the original text, and at once corroborate the existence of an archaic form with d-. The variant form dacrimas in Livius does not exist either. It is an old conjecture by K. O. Müller that replaces the manuscript reading lacrimas in a verse recorded by Paulus ex Festo 182, Lindsay: simul ac lacrimas de ore noegeo detersit (in spite of the obvious alliteration lac-lac-!). It is transmitted by the Farnesianus, the only codex in which a part of Festus’ work has survived untouched by Paulus’ later compilation. The reading lacrimas is consequently favored in the most recent edition of Livius Andronicus (Blänsdorf et alii 2010: 26; Frag. 17, ex Od. , 88). Festus’ remark on dacrimas is by far the best evidence we have in favor of the idea that da- was evolving into la- at the outset of literary Latin, and we owe it to literary works that, crucially, are almost entirely lost. But, in spite of former accounts, the question does not revolve around whether dacrima is a loanword from Greek 
 μ, -  , which it is 3 definitely not, but, conversely, whether it was glossed over to look Greek (or, as we are going to see, simply to look authentic), and is consequently non-diagnostic. d) Latin l
vir/laevir goes back to *deh2i-u(e)r- ‘brother in law, husband’s brother’, attested in Skt. devár, Gk. , etc., and has been presumably modified under the influence of vir ‘man’. It is only mentioned by late grammarians, like Paulus ex Festo 102, Lindsay. Here, Nonius Marcellus (p. 894 Lindsay) comes to the aid of researchers with his often quoted quasi laevus vir, which has given some air to the possibility that the initial l- is paretymological after all and can be quietly swept under the rug. e) Latin lautia ‘entertainment for foreign guests’ is attested in Livy, Apuleius, or the Senatus Consultum de 2 Asclepiade (CIL I , 588, a text going back to 78 BC). It has a variant dautia, used by Livius Andronicus according to Paulus ex Festo 60, Lindsay: item dautia, quae lautia dicimus, et dantur legatis hospitii gratia. As for the etymology, Vine (2006: 238), 3 See Hamp (1972). Volume 47, Number 3 & 4, Fall/Winter 2019 462 Blanca María Prósper followed by EDLIL, s.u. dautia, proposes an agent noun *douHó‘bestowing’ from *deh3u- ‘give’ (Rix et alii 2001: 107) with laryngeal metathesis, which would have given *dauó- early on, and was enlarged by -et-, giving ‘bestowal’. In itself this is complex but unproblematic, since, as Vine himself has recognized, the accent-sensitive change /o/ > /a/ must have taken place prior to the split of the Italic languages, including 4 Venetic. On closer inspection, however, the argument according to which lautia is secondary and analogical is uncompelling: EDLIL’s account, s.u. dautia, leads one to believe that Paulus’ reference was to Livy, when in fact it records Livius Andronicus. While the reader is reminded of the fact that this form always occurs together with a case form of Latin locus and an alliterative effect is invoked as an explanation (as already suggested by Schrijnen 1914: 379, who, however, regards the etymological connection of lautia and lavare as conceivable), one does not quite see why this should be the case in a legal phrase, and, contrarywise, could expect the preservation of dautia given not only the inherent conservativeness of legal language, but also the convenience of preserving the lexical relatedness of dautia and dare. It is quite another thing to verify that lautia does appear in an alliterative context, as in Livy 30, 17, 14, aedes liberae, loca, lautia legatis decreta, which is immaterial to the final decision on its origins. It should be noted that it is slightly inconsistent to give credence to Livius as regards the change dacrima > lacrima and then look for an ad hoc explanation for dautia > lautia. 4 Alternatively, one could consider a potential adjective of the structure *dh3uetó- ‘to be given’ > *dauutV-. Some scholars would object to this possibility on phonotactic grounds and suggest that the reconstructed sequence should have given *duueto-, either because the more sonorous approximant -u- was more prone to become vocalized than the laryngeal (presupposing *dh3uueto-), or because the epenthetic vowel that resulted from laryngeal vocalization was assimilated to the following -u- (yielding C
u- > Cuu-, as defended for IndoIranian and Greek by Lipp 2009: 421-424). No certain examples of the enlarged *deh3u- occur outside Italic and isolated examples in Baltic, Skt. dúvas ‘reverence, oblation, favor’, and possibly OIr. dúas (- stem) ‘gift, recompense to poets’ < *deh3u-sth2- (but see an alternative account below). The Journal of Indo-European Studies What became of “Sabine l”? 463 What is more, dautia could become the touchstone that lays bare Livius’ tendency to invent pseudo-archaisms: As the 5 dictionaries often point out, lautia can also be a concrete departicipial neuter noun, derived from lautus ‘washed’ but also ‘refined, luxurious, sumptuous’, by means of *-(i)io- (which may have been the case with spatium, pal
tium, initium, etc.). In fact, it may have been the secondary association of lautia and dare which, again, have led Livius to recreate dautia and disregard the “corrupt” form lautia. But, this time, he simply did not hit the target as with dacrimas, since he had no Greek form to hand and possibly relied on indtiae ‘armistice’ (a moment in which gifts are officially exchanged), which was still possible in view of the increasingly opaque morphophonemic alternation between -au- (in the first syllable), as in lautus, vs. - (in any non-first syllable), as in illtus. On the other hand, a relatedness of lautia with Celtic *la/outo- in OIr. folud ‘property, substance’, MW. golud ‘riches’ cannot be ruled out. While the Welsh form has been traced back to *-louto-, Isaac (2007) has argued for neutralization of tautosyllabic /ou
/ and /au
/ in Welsh. Both forms would then go back to IE *léh2u-to-, which may be translated as ‘the things that are enjoyed, riches’. What is more, this casts some light on the allegedly secondary meaning of the Latin past part. lautus as ‘luxurious’: it could in fact be the primary meaning, and in that case it would also go back to *leh2u-to-. This can be neatly explained as a substantivization of *lh2u-to-, the past part. of *leh2u- ‘to enjoy; benefit’, in Lat. lucrum < *lh2u-tlo- ‘benefit’, Germ. launa- < *le/oh2u-no- ‘reward’, Gk.  ‘enjoy’ (not included in Rix et alii 2001). Somewhere down the line, it became associated with lavere, and this was made easier because the partly identical lav
re had a past part. lav
tus. In that case, the reconstructed past part. *luh3-to- > *lto- would have survived only in compounds, and the preexistent *lauto-, which automatically incorporated the meaning ‘washed’, took its place as the past part. of the simplex lavere because it had the shape of most forms of lavere and lav
re (whatever the 5 Cf. DLL, s.u. lautia; EDLIL takes an altogether different stance by implying that the past part. lautus could have favored the irregular change d- > l- in lautia, and thus conveniently rules out this form, too. Volume 47, Number 3 & 4, Fall/Winter 2019 464 Blanca María Prósper origin of -au- in these forms) and shared the scheme of plaud
, 6 plausus; cave
, cautus, etc. All this would ipso facto eliminate the need to find an etymology for dautia, and suggests that it is not lautia that is the product of assimilation to the context (because it was surrounded by words beginning by l- like loca, legatis), but, conversely, it is dautia that owes its existence to unconscious hypercorrection, suggesting that Livius or the tradition after him were more interested in “purifying” some aspects of the lexicon than in achieving or maintaining pleasant auditory effects. In addition, this would prove fatal for the idea that Livius is our only, but trustworthy witness for the ongoing and unpredictable change d- > l-, since it would automatically discredit dacrimas as a real Latin form. To recap, I contend that the traditionally adduced cases of preserved da-, if they have existed at all, are artificial and deceptive. If taken seriously, they suggest that the change is sporadic, late and even analogical (and then non-existent). But, as we have seen, the only unassailable cases of diachronic variation are dacrima and dautia; regarding these, the Latin author in question, in both cases Livius, could arguably start from the contemporary forms lacrima and lautia, venture a guess and actually get at the right, original form. It should be clear by now that he hit the target in the first case and succeeded in fooling ancient and modern scholarship in the second. Unfortunately, Paulus’ testimony, in spite of being fragile, th late, indirect (in the 8 C. AD he epitomizes Festus, a nd grammarian thought to have lived in the 2 C. AD, who in turn draws upon the earlier Verrius Flaccus, who may have been 7 indebted to Varro and Aelius Gallus), poorly transmitted and containing a number of unwarranted grammatical opinions which threaten objectivity and may have subtly contaminated a 6 Vine (2006: 239) reconstructs a verbal adjective, as if from *louh3-etó-. Garnier (2010: 369-370) posits *laui-to-, and acknowledges that the expected past part. must have been *luh3-to-, disguised among compounded forms like colltus, and mentions Arm. luac’ ‘he has washed’, from a frequentative *l-tse-, but finally reconstructs *kon-lau(i)-to-. 7 Cf. Glinister et alii (2007: 1-9). The Journal of Indo-European Studies What became of “Sabine l”? 465 number of entries in unrecoverable ways, has never been subject to an adequate Quellenkritik, as modern scholarship has been seduced by any opportunity of regaining shreds of Old Latin. I assume Livius and his contemporaries are unlikely to have actually heard the phonetic sequence da- in those particular words, which they (or Livius at least) used for precisely that reason: it did not belong to everyday Latin usage. And it is not immaterial to this argument that Livius, about whom we know very little, was defined as semigraecus by Suetonius, possibly stemmed from Magna Graecia, and upon his arrival in Rome became a protégé of a senatorial family and a teacher of the élites. If this is something to go by, it could even be the case that his dacrima, dautia were simply due to slips of the tongue, viz. to interference of his native language with Latin, and then not even the product of a conscious experiment, or he may have sought to cause an aesthetic impression on his partly bilingual readers or spectators, and then his non-Latin d- forms were generated for the sake of word-play, intended to emphasize the connection of Greek and Latin (the expression saepe posuit leaves us wondering whether he often, but not systematically selected this variant for alliterative or other reasons; and our only manuscript reading for this word in Livius is lacrimas). But we cannot pursue this thread further, and his command of both languages, his taste for archaic forms, and his ability to romanize expressions and situations speak in favor of a planned linguistic effort to produce the first Roman epic poem. Therefore, the limited but universally accepted evolution from Old Latin da- to Classical Latin la-, whose input is claimed to be miraculously attested in Livius, never took place, because only the variants lacrima, lautia and laevir (or *laever) were actually available by that time. While we can hardly speculate about his linguistic conscience, Livius may have felt, in view of Latin lacrima, that initial l- could be a corrupt, barbarian version of pristine d-, and he may have seen the word medial alternation -d/l- as a confirmation of this suspicion (he in fact chose the Latinate Ulixes, but the form with -l- was already Volume 47, Number 3 & 4, Fall/Winter 2019 466 Blanca María Prósper 8 known in Greek as Olys(s)eus besides Odysseus). It would probably be off the mark to presume that Livius was put on this track because he reckoned that too many forms had initial la-. But I shall seek to demonstrate that there was a reason why Livius could deduce there was a correspondence between Latin la- and Greek - (and not, for instance, li-: di- or l
-: d
-) potentially justifying this facet of language-polishing (for instance the existence of laevir vs. , attested in the Iliad, even if a putative †daever has not reached us), and 9 consequently rectified the actual Latin forms. We must keep in mind that even the cultivated Roman élites and poets could not be expected to distinguish the boundaries between Latin borrowing from Greek, Latin somehow being genetically an offshoot of Greek, and fashionable manipulation of some lexical items so as to make them sound closer to Greek. Whether Livius or his secondary compilers set about to “reconstruct” the uncontaminated, viz. Greek appearance of other Latin forms beginning with l- we will probably never know, since what has come down to us is the tip of the iceberg of ancient encyclopedic literature. What is likely to have happened is that, following a typical abductive procedure, viz. by hypothesizing the existence of the d- forms, Livius created them, and paradoxically with disastrous consequences not for his own language, where they patently never prospered, but for modern historical linguistics. In order to obtain a thitherto non-existent Latin dacrima, Livius simply had to remodel lacri/uma. As a consequence, what had been a close cognate of the Greek form 
μ was by a sleight of 8 Kretschmer (1896: 280-281) held the view that the frequent λυσσε
ς was the original form vis à vis δυσσε
ς, which in turn belonged to the language of the Epos and was redone in analogy to δ
σσεσθαι ‘be wrathful’. By contrast, Solmsen (1909: 211) reckons with dental-to-dental dissimilation. 9 Note that this is a far cry from Marotta’s approach (1993: 69), which revives the Sabine hypothesis in a more sophisticated way. She brings to bear a diaphasic variation motivated by diglossia between Greek and Latin in the Latium Vetus. There, a Latin speaker who was also competent in Greek could assimilate the Aeolic existing variation d/l to the point of borrowing it into Latin, whereupon it received an extra prestige linked to the Sabines as antiquitas. (Bear in mind, however, that it is the secondary /l/ that is usually termed Sabine, not the old-fashioned or ancient-looking /d/.) The Journal of Indo-European Studies What became of “Sabine l”? 467 the hand granted an ancestor that automatically became legitimized as its full match. That Varro had gone so far as to create 
 out of the blue only to account for lepestae should have promoted awareness of our limited knowledge both of predocumentary Latin and the linguistic agenda of Latin authors and scholars. 3. The two origins of Latin #laWe have seen that at least two forms, lacrima and laevir, can be traced back to IE forms beginning by *d-, which does not mean that the actual attestation of Old Latin d- forms is reliable (as shown in the case of dautia). Occasionally, individual steps have been taken to add further materials to the accepted list, but no global attempt has been made to find out how many forms beginning with l- and lacking a good etymology could be given one simply by reconstructing d- in its stead, regardless of the ancient grammarians’ silence about them. In other words, while the substantial number of forms exhibiting initial la- and lacking an etymology, however uncompelling, should lead one to find out if they may go back to da-, try to formulate a rule, and test it inductively, we have resigned ourselves to the idea that the forms showing la- for da- are the haphazard product of comparatively recent, dialectal, analogical or simply arbitrary 10 changes. By doing this, we have been reassured that the forms in la- are otherwise true to their original look. In other words, with few exceptions, we have made a point of subtracting as many examples of the rule #da- > #la- as possible, hoping that the problem itself might vanish, instead of adding more examples, and confirming there is a problem. 10 See the same complaint expressed by Pulju (1998). But, again, he dares not walk off the beaten path and takes into account the same problematic evidence as the rest of the scholars, only to conclude that the problematic cases originally contained a cluster /dl/ simplified over time in one direction or another, which, needless to say, is ad hoc, and already suggested as the input of a regular sound change by Schrijnen (1914: 390). Since, however, such clusters are generally dispreferred in languages because their acoustic parameters do not differ enough to allow phonemic contrast (see Marotta 1993: 62), the mechanic reconstruction of such a cluster for a comparatively high number of cases violates the uniformitarian principle. Volume 47, Number 3 & 4, Fall/Winter 2019 468 Blanca María Prósper What happens if we change our perspective in accordance with the above ideas? The stubborn fact is that, for a sequence #lC- to be possible at all in any Italic language, the root must be in most cases *lH- (*slH-, *tlH- and *ulH- would have yielded #l
-), and this paradox remains unresolved: the only compatible verbal roots recognized by Rix et alii (2001: 399-401) are *leh1- ‘to let, allow’, *leh2- ‘to bellow’, *leh2- ‘to pour’, *leh2- ‘to hide’ (late), and the enlarged *leh2k- ‘to make noise’, *leh2p- ‘to illuminate’, and none of them is very promising when one has to come to terms with the sizable number of forms lacking an explanation thus far. The rest of the certain cases of initial #lC- must be traced back to the doubtful phoneme /a/ (*las- in lascivus), to   schwa secundum (*l d- in lassus, *sl g- in laxus) or to the outcome of Thurneysen-Havet-Vine’s rule (*lou- > lau-, as in lav). I shall argue below that, if we simply ignore Livius, we are left with a sizable number of cases of initial #laC- going back to *#daC- (which can be extended to #l
C- > *#d
C-), at least when no morphophonemic alternations interfered. What usually stops us from venturing an etymology with a /d/ for forms showing a /l/ is the idea that the change can only be confirmed if both variants are attested. And yet, as observed above, this may be largely illusory, and it is the extant doublets that should be held suspect. The age and nature of this shift has been obscured by the monoglottic, tolerated but sporadic change -d> -l- in medial position, which produced such doublets as odor, olre (late olor is probably analogical), solium, sedre, etc. All these forms have one thing in common: /d/ yields /l/ only when 11 a front vowel or yod follows. This shift is paralleled for instance in the Romance languages, where we find Valencian Catalan rélit < reditu, 11 This is why Eichner’s account on HAVELOD (Forum cippus) as a direct h cognate of haud ‘not’ and ultimately from * auedo- ‘falsch’ (Eichner 1995, fn. 8) remains problematic as a case of “Sabine l”. If the underlying formations are identical, which is not completely certain (one could reckon with, e.g., h * a/ou
-lo-), and assuming that the ablative form had become stereotyped in legal formulas, medial -l- could be explained as a case of sporadic or even nonce case of dissimilation of dentals. I thank Heiner Eichner (Vienna) for sending to me this interesting work. The Journal of Indo-European Studies What became of “Sabine l”? 469 melecina < medicina, Leonese portalgo < portaticu. But the opposite change is also well attested, as in the Calabrian regular change of initial (postvocalic) and intervocalic -l- > -
- in the neighborhood of any vowel, e.g. in [a '
ana] ‘the wool’ (< lana) or ['pa
a] ‘spade’ (< pala), for which cf. Rohlfs (1966: 217, 309). A shift -
a > -la in the Romance languages mostly affects final syllables: cicada, cauda > Sp. cigala, cola, and comes up at a stage in which the intervocalic voiced stop [d] had already fricativized into [
]. The phonetic processes behind this two12 way change have not been completely clarified. Elsewhere I have tried to show that a regular dissimilation has taken place in Very Old (Preliterary) Latin, by which a sequence of two dental sounds -t/
V-
V- has evolved into -t/
VlV-, and that the inherited, mostly denominal adjectives in -i-losimply do not exist and are allomorphs of the common type in idus (see Prósper 2016, 35-50). In all likelihood, both changes, affecting on the one hand intervocalic /d/ preceding short front vowels, and, on the other, initial /d/ or perhaps a more complex segment preceding a short open front/central vowel [a], are 13 probably unconnected. What they have in common is that they unexceptionally occur in syllable onset. While there is a 12 Marotta (1993) has underlined what she calls “the special status of dental stops throughout the history of Latin”, which involves a vast array of cases of deletion, substitution and assimilation. The coronal class is by definition, because of its unmarked nature, the most prone to change, and dentals are the weakest stops. In her view, the substitution of /l/ for /d/ means that a more marked segment is introduced, indicating that only coronal stops are unmarked and typically undergo these changes; in onset, this is in her view suggestive of a strengthening process. 13 The showcase examples are Dakota vs. Lakota and Myc. da-pu2-ri-to-jo vs. Labyrinthos (in which the direction of the change might be the same in spite of the usual explanation that compares Gk.  ), or Gk.   vs.   ‘laurel’. Interestingly, a similar phenomenon involving language contact is attested in Eastern Iranian languages and Anatolian. The Anatolian languages have no initial d-, which has been devoiced, possibly earlier in Hittite. As a consequence, l- is substituted for d- in loanwords. Interestingly, most examples contain the sequence da-: thus, according to Melchert (2003: 181 and fn. 13), Luv. *dabarna- ‘the strong one’ > Hitt. labarna-; Gk. μ  > Lydian Lamtru, Akkadian personal name Ddbnu > Luv. La-tà-pa-nu. I owe this reference to my colleague J. Virgilio García Trabazo (Santiago de Compostela). Note that this is not the case in such Greek loanwords in Oscan as dam<u>sennias, since the initial contrast d-/t- was still available. Volume 47, Number 3 & 4, Fall/Winter 2019 470 Blanca María Prósper close natural connection between coronals and front vowels, this falls short of explaining their context-bound merger with 14 laterals at two different, probably very distant stages. The comparative regularity of the change #da- > #larequires further phonetic grounding. We can speculate with the possibility that the articulatory effort was minimized when [d] preceded [a] because this is the inherently longest short vowel. However, there is a chance that we are dealing with two different kinds of /d/. In the change that gives rise to the wordinternal variants in l/d, the targeted phoneme is certain to be /d/. In addition, if the only reliable examples of medial l/d go h back to /d/, not /d /, it may be the case that this change has to be situated in a time in which Latin still distinguished medial [d] from [
], namely prior to fortition and merger of plain voiced stops and voiced aspirates, which took place later than assumed (see Prósper 2016, 35-50). But in Proto-Italic #d-, many Indo-Europeanists would recognize a glottalized sound, probably an implosive sound []. Greenberg (1970: 129) noted that it is a typical feature of the languages possessing an implosive series that [] is retracted, becoming an alveolar, and this is often accompanied by retroflexion. In turn, this “often leads in a further stage of development to a preglottalized sonant r or l
”. Additionally, Greenberg observes (ib., 131) that implosives are often found in initial, but not in medial position, where they are neutralized with their plain counterparts. 14 Conversely, an exclusively Sabellic change l > d seems to be favored by high front vowels. Oscan has DIVMPAIS from *lump. See Prósper (2015, 42-47) on the Lusitanian divine name LVMBIS or LVMPIS (dat. pl.), occurring once in an indigenous onomastic context: this form is most unlikely to be reflecting the epigraphically very uncommon divine name LYMPHIS, which is restricted to half a dozen cases in Italy and a recently uncovered one in Baetica, and is a cultivated form. Moreover, this is unlikely for phonetic reasons (if the rendition was careless or defective we would expect to find LINFIS). SouthPicene kduíú ‘I am called’ corresponds to Lat. clue ‘to be known’, etc. Rix (1994) has made a case for the South-Picene shift -l- > -d- being triggered by a following palatal vowel, and consequently for /u:/ having evolved into /y:/ in this dialect. We could obtain a unified picture of this tendency in Sabellic by assuming that l- > d- in O. DIVMPAIS is actually caused by the epenthetic dorso-palatal -i-, which in turn responds to context-bound backing of coronals. In fact, Matisoff (2013:88) has described deltacism as “partially epenthetic in nature”, by which -lj- > -ldj- > -dj- > -d-. The Journal of Indo-European Studies What became of “Sabine l”? 471 It could consequently be the case that Italic [
] survived only in initial position, and even there it tended to merge with [d]. But when it preceded [a] or its immediate ancestor, it shifted to [l], whereupon it eventually merged with the only IE inherited lateral sound [l]. That an early dissimilation of glottalic features is involved that led to the divergent outcome (as we are going to see, virtually all the sequences would go
back to # H- or #aH-, where H = any of the three 15 reconstructed laryngeals) remains an intriguing possibility. But Greenberg observes that when [
] becomes retracted or retroflex, the glottal feature becomes redundant and tends to be 16 eliminated anyway. In some languages like Bedik (NigerCongo), initial [
] has become [l]. This alternative reconstruction may prove of some heuristic value when it comes to the problem of the high number of Latin words beginning with *la- for which there is no plausible connection 17 whatsoever. 15 After writing this work, I have come by an unpublished master thesis by Jakob (2017: 13, 24-25) who, for the first time, makes a point of \finding a phonetic explanation for the usual suspects, and an ingenious one: in his view, Livius was hypercorrecting, and no external influence can be demonstrated for lacrima, lvir and lautia. Consequently, there must be a reason for the change that is common to these three forms: a word initial sequence *dHV-. To my mind, however, the reconstructed preforms are patently ad hoc: *dh2eru- > lacrima is a possibility like any other. Since it is often treated as a compound which at some stage became *draru-, it is unclear how and when this preform would come into being. Jakob’s *dh2eiue r > lvir cannot have originated from any allomorph of his own paradigm *deh2i-ur, acc. sing. *dih2-uer-m, gen. sing. *diuh2r-es. Finally, lautia is taken from *deuh3 -t- (which cannot give dau- anyway) or, to account for /l/, from *dh3eu-et-, which amounts to wishful thinking. In other words, all three forms probably contain a laryngeal, but not necessarily in that position. And, as usual, other potential cases of this shift remain unconsidered. 16 Note that a similar dissimilation is likely to lurk behind the loss of /d/ in the h Baltic forms ìlgas ‘long’, etc., from *dlh1g h1ó-, on which see below. 17 The general idea that glottalization is somehow behind the deviant behavior of voiced plosives, including the case at issue, is not new. Cf. Lehmann (19861: 487), who favors an original ejective /t/, however: “Change of initial glottalic dental is most prominent in Latin, where l < d has the label ‘Sabine l’ after a proposed explanation no longer widely accepted though the label has remained”. There is an explanation for the tendency of /d/ and /l/ to merge partly based on the (traditionally neglected) phonetics of /l/ in Leonard (1980): in his view, there are some reasons (such as the disparate Romance outcomes Volume 47, Number 3 & 4, Fall/Winter 2019 472 Blanca María Prósper 4. Neglected instances of #da- > #laIn this section, I shall try to identify a number of cases of initial la- which find a more compelling explanation if they are traced back to da-. In doing this, positive advantages can be obtained: first, a registered root can now be identified; second, the word formation and Ablaut become transparent; third, for the first time, the semantic side of the equation is satisfactory; and last, but not least, plausible, identical cognates can be found in other Indo-European languages for most of these forms, notably in Greek and Indo-Iranian, and they occasionally help illuminate or provide parallels for other Latin forms. *deh2i-t- > laetus A root *deh2- ‘to give out, distribute, divide’ (Rix et alii 2001: 103-104) is well attested, especially in Greek, Vedic and Albanian, both in verbs and nouns. By contrast, it is apparently absent from Latin except for the family of daps ‘banquet’. In the following lines, I shall argue that virtually all known forms and extensions of this root are actually present in Italic. To begin with, the adjective laetus ‘happy’ may ultimately go back to *deh2-(i)- (Rix et alii 2001: 103-104 *deh2i-). It can be analyzed as a possessive derivative *déh2i-t-o- ‘partaking’ derived from a noun *deh2i-t-, attested in Greek ,   ‘banquet, feast, food’. Then, it would be identical to the Greek personal name
  (a son of Kephalus in Pausanias, 1, 37, 6), which additionally shows the expected accent position. In that case, the original meaning is less general: either ‘wealthy, partaking of riches’ or simply ‘having good fortune, receiving one’s lot’. This is consistent with what DEL, s.u. laetus considers its original meaning, ‘gras’, that is to say ‘fatty’, as in glande sues laeti redeunt (Verg., Georg. 2, 520), and the evolution is not very different from that of Lat. f
lix ‘(suckling) > well fed > satisfied > happy’ to Sp. feliz, It. felice. Finally, it is fully consistent with the reconstruction of an inherited (because impossible to explain as an internal creation of Greek) of /l/, and the confusion of /d/ and /l/ in Latin and Umbrian in some contexts) to believe that, in preliterary times, /l/ had a retracted, obstruentized articulation that made it similar to /d/; this realization was retained in lower sociolects or non-urban varieties. The Journal of Indo-European Studies What became of “Sabine l”? 473 acrostatic noun *doh2i-t-/*deh2i-t-, as per Nussbaum (2004), cf. Vij
nas (2009: 76-77). *deh2i-tor- > Laetrius The existence of a gens Laetria leaves little room for doubt: the underlying personal name *Laitr is identical to the Greek personal name  (a Trojan hero killed in combat in Il. , 275; the same name or a variant form *daitros may also occur in Mycenaean Greek as da-to-ro in Knossos) and to   ‘conviva’ (Aeschylus). This form goes back to an agent noun *deh2i-tor- ‘dispensing, providing for’. A *daitr is also attested in the anthroponymy of the “Illyrian” complex, judging from DASSIVS DA/ETORIS FIL(IVS) (on the epitaph of a Dalmatian soldier uncovered in Wiesbaden, Germania Superior, CIL XIII, 7581). Needless to say, an enlarged root *(s/u)leh2i- is unattested. As is well known, the “Italic rule” prescribes that Italic agent nouns in -tr be modeled on the past participle. In this case, the PIE full grade of the root is preserved because this is an inherited formation, unrelated to a synchronic verb and confined to the realm of onomastics very early. Ironically, the fact that the centurion Marcus Laetorius is recorded as the first plebeian magistrate, who took office as early as 495 BC, identifies this neglected gentilic as the earliest attestation of the change da- > la-, and decries the much later forms dacrimas and dautia ascribed to Livius Andronicus as the result of hellenization or hypercorrection. *dh2-t
r > lat
r The -r-stem later, -eris ‘brick, block’ (diminutive laterculus) is in all likelihood an unproductive agentive formation in -t
r. Its original meaning is probably ‘part, division, section’ (cf. EDLIL, s.u. later) and the only root one can reasonably have recourse to in order to explain it is *dh2- ‘to divide, segment’. Its nearest cognate is Gk.   ‘divider, distributer’. It corresponds to the hysterokinetic type described in Tichy (1995), which has not a general but a “potential” function, and in the Latin case shows a generalized full grade of the suffix. There are two ways in which we can bridge the gap between ‘divider’ and ‘divided object, segment, resulting part’. To begin with, later may originally have designated the mold that was Volume 47, Number 3 & 4, Fall/Winter 2019 474 Blanca María Prósper used for making the actual clay bricks. Secondly, some agent nouns develop a passive meaning, as in Pahlavi griftr ‘prisoner’, frftr ‘easily deceived’, etc., and this is especially understandable if the noun has no synchronic link with a verb. As an anonymous reviewer kindly points to me, however, Lazzeroni (1992) has contended that nouns in -ter could be used as instrument nouns, which disposes of the problem straightforwardly. It should additionally be noted that the old agentive meaning ‘divider’ vel sim. has been preserved in the gentilic Laterius, which is probably derived from this noun but is unlikely to be related to bricklaying. h *deh2i-d h1-e/o- > laed Lat. laed ‘to damage’ remains equally devoid of external (h) (h) cognates. Since we lack a root *(s)leh2id - or *(s)lh2eid -, and the connection with Germ. *slte/a- ‘to tear, destroy’ cannot be defended for phonetic reasons, laed can no longer be taken at face value. In fact, it looks like a neo-root going back to a collocation and if the structure proposed above has something to recommend itself, it would originally have meant ‘provide one’s lot’ from which a negative sense ‘to impart a blow’ would have developed (note, by contrast, the original warlike meaning uh h in fend < *g en-d h1- ‘to place blows, strike’); the evolution in malam partem is, mutatis mutandis, comparable to that of damnum ‘expense > loss > damage’, on which see below (cf. Hackstein 2002 for the origins of the process). Interestingly, this warlike meaning of the extended root is probably attested in the fossilized Greek dative
 ‘in battle’ (attested from the Iliad onwards). Martin Peters (apud Nikolaev 2010: 168, fn. 52) has analyzed it as the only remnant of a root-noun *deh2i- in the dative case, which in his view had evolved as follows: *deh2i-ei > *di-ei >> *diii [sic]. The sense of the neo-root preserved in Lat. *lai
- as ‘to wage war, inflict damage’ becomes more plausible from this comparative perspective. This could help clarify the origins of another unexplained word: laus ‘praise’ (from Livius Andronicus). As usual, it has no near cognates, and the connection with Goth. awiliu ‘song of praise’, G. Lied, etc. (< *leut-) fails to convince in view of the incompatible Ablaut. OIr. lúaidid ‘to move, set in motion, incite; utter, proclaim’ is unrelated. Vine (2006, 238) has disposed of The Journal of Indo-European Studies What became of “Sabine l”? 475 the phonetic inconvenience by reconstructing an agent noun *louó-, in turn enlarged by -ed-, giving *laued-. But the formation still looks obscure to me, and the unenlarged root *leu- does not exist (in fact, the ultimate analysis or appurtenance of the Germanic forms remains unknown). Alternatively, if we reconstruct a secondary rooth compound *deh2u-d (h1)-s ‘composition-making’, we get a form very close in semantics to Gk. 
μ ‘composition > poem’. Consequently, I believe Martin Kümmel’s suggestion (Rix et alii 2001: s.u. *deuh2-) that IE *deuh2- ‘zusammenfügen’ be corrected as *deh2u- because of Goth. taujan ‘to do, make’ (< *deh2u-ie/o-) to be right (it could further be argued that -i- and -u- are different enlargements of a single root *deh2-). And from here it is only one step to ‘praise, laudatio’, the poem par excellence in public life. It should be noted that OIr. dúan (fem.) ‘poem’, traditionally traced to *dh2p-no- and identified with Lat. damnum ‘expense, damage’, OIc. tafn ‘sacrificial animal/meal’, OIc. tapa ‘to lose’ (see EDPG, s.u. *tapp/bon-), Arm. tawn ‘feast’ (Watkins 1976), finds an alternative explanation if we reconstruct *deh2u-no-. This is more plausible, in spite of Watkins’ efforts to integrate dúan in a system of gift exchange and reciprocity, since the semantic leading thread that unites the terms going back to *dh2p-no- is ‘consumption, destruction, expenditure, sacrifice’ > ‘(sacrificial) meal, feast’, and is unrelated to the notion of ‘giving’. What is more, this paves the way for a new etymology of OIr. dúas ‘gift, recompense to poets’ (-), usually taken from *deh3u-sth2- (from the enlarged root ‘to give’ mostly found in Italic; see fn. 4 above), but perhaps to be traced back to *deh2u-sth2-. Latin words enlarged by a suffix -(e)d- are by no means usual (I deem the explanation of the adjectives in -idus as -(e)do- to be untenable for most forms), and it is astonishing that such an unproductive formation has no cognates anywhere. h (h) Accordingly, Lat. fraus ‘deceit’ (<< *d reu-d -), whose vocalism 18 equally presents problems, may have been remodeled by analogy with laus on account of its antonymic meaning. 18 U. frosetom ‘fraudatum’, probably from a periphrasis containing a nom. sing. *frVu
-s and a past part. of *h1ei-, is inherently ambiguous. The often raised objection that we would expect Lat. -b- by Lex RUBL is not Volume 47, Number 3 & 4, Fall/Winter 2019 476 Blanca María Prósper *deh2-es- > lr Lr, laris ‘tutelary god’ (the long vowel in the nom. sing. is assured by Priscian) remains unexplained. If (Very) Old Latin *las-, transmitted in the nom. pl. LASES in the Carmen Arvale (CIL VI, 32388) and in the abl. pl. ab lasibus lares (Varro, De Ling. Lat. 6, 2), and lasibus ... pro... laribus (Paulus ex Festo 323, Lindsay), is not artificial, it could be explained as decompositional to a possessive adjective like, e. g., *h1uesudeh2-es- > *uesu-ds-, and would originally have meant ‘dispenser of goods/riches’. For similar formations, cf. the Avestan possessive adjective hu-dh- beneficent from *h1su(h) d eh1/3-és- or its antonym du-dh- wrongdoer (that is to say, not giving bad things or giving bad gifts) and yau-dhmaking healthy’, as opposed to Skt. su-ds- generous. The slight semantic hesitation suggests that two Indo-Iranian h preforms *daHas- and *d aHas- have fallen together in Iranian. Additionally, Armenian dik ‘gods’ probably goes back to a h plural form *d s-es, which in my view is decompositional to an h ancient possessive adjective *X-d eh1-es- having/receiving a (...) ritual act. Cf. on this stem and the etymology of Lat. fstus, friae, Gk. , etc. Prósper (2018a). The lares would then belong to the Indo-European tradition of fortune dispensing divinities, exemplified by OCS. bog, OPers. baga- ‘god’, such divine names as Skt. Bhaga, Lith. th Datanus (glossed as donator bonorum seu largitor, XVI C.); Gk. μ , a class of divinities allotting fate, etc. The IE divine names are often accompanied by epithets alluding to this very quality, cf. Indra vasu-d- ‘giver of wealth’, Homeric   , etc. Finally, Hesiod, Op. 122, 126 claims that the men of the golden generation eventually became μ  and calls 19 them    . The only formational problem lies in the fact that we would expect a gen. sing. †lris, not laris, and a nom. pl. †lres, not lares, unless the short vowel is analogical to unassailable, since there is no single case of -au- from -au
-. This may point to a realization [a
] which made the sequence immune from labialization and would equally well explain claud, caud, plaud without recourse to a (in itself conceivable) root enlargement *-d(h3)-. 19 See West (2007: 132). I owe the last reference to my colleague and friend Marco Antonio Santamaría Álvarez (Salamanca). The Journal of Indo-European Studies What became of “Sabine l”? 477 m
s, maris or simply, like probably this latter form, the other cases retain the original hysterokinetic inflection attributed to these adjectives, namely acc. *-dh2-es-m > lrem, gen. *-dh2-s-és > lris, which smoothly accounts for the attested form (cf. Schindler 1975: 263). When I was writing these lines I discovered an interesting work by Matasovi
(2014), who has come to a very similar conclusion. However, he reconstructs *deh2s-s / *dh2s-os ‘share’, which is unclear to me in formational terms. At all events, iubar, iubris poses the same problem (and cf. Caesar, -ris). Given the anomalous apophony (we would expect †iuberis, †Caeseris), this set of nouns seems to have undergone leveling early on. Dunkel (1997) is probably right in taking the short vowel from the abbreviated nom.-acc. Accordingly, -
s, -
ris would have become -ar, -aris, and lris, lres could be modeled on this type, like the obscure m
s, mris ‘male’. Interestingly, the connection of l
r and l
rua ‘monster, evil spirit’ may cast some light on the matter. It looks like the fossilized feminine form of a denominative adjective in -uus. As such, it is expected to preserve the original vocalism of the oblique stem of its base, thus presupposing an inherited structure *deh2-és-, gen. *deh2-es-ós. Note that this would point to the original non-rhotacized form actually preserving the long vowel of the stem; accordingly, the first two words of the verse ENOS LASES IVVATE in the Carmen Arvale would have the scansion U - - -, as in the immediately following formula ENOS MARMAR IVVATO, and is equally compatible with the admitted structure of the Saturnian verse, whether this is conceived of as of quantitative or accentual nature (and if this text is conducted 20 in Saturnians at all, which is debated). Gk.   and its derivative    are epithets of the Erinyes, Carybdis and Hecate. In a recent study, Kölligan Macedo (2015) have cogently argued that   is a 20 It follows that the association of the divine name Acca Lrentia with the Lares is probably paretymological. Lrentia could go back, for instance, to the present participle of a desiderative formation *d(e)h2/1s-nt-, which would pave the way for its identification with the personal names DASAS, DASANTIS (Dacia, Dalmatia, Germania) and DASES, DASENTIS (Pannonia), or Skt. (RV) abhidsant- ‘persecuting’, from *d(e)h2/1- (Rix et alii 2001: 103 ‘aufspüren, antreffen’). Volume 47, Number 3 & 4, Fall/Winter 2019 478 Blanca María Prósper compound of the type attested in Lat. locu-pl s (see Nussbaum 2004), and reconstruct *dns-pleh1-t- ‘who fills/is filled with magical power’. The first member is PIE * dns-, the zero grade of a root noun *dens- or of the s-stem *dens-es- ‘magic power, cunning, plan, craft’. Alternatively, I think the first member could be the compositional zero grade of the stem *deh2-os underlying the possessive compound *X-deh2-es, and consequently we might as well reconstruct *dh2s-pleh1-t- ‘who fills with gifts’, or more broadly, since compounds of this root usually govern a DO, ‘fulfilling one’s destiny’. Note the h ‘divine, god-spoken’, etc. parallelism with *d h1s- in Since the Erinyes are known to have been invoked as protectors of fertility, this quality may have contributed to their - with the first member epithet. Whether the similarity of (Hesychius) is indicative of the gloss of cognacy at all is debatable. *deh2-mn > l mina Lat. l mina ‘thin sheet (especially of metal)’, which thus far resists all etymological attempts, may be a secondary feminine singular (of the type opera, -ae, arma, -ae, ra, -ae, which gained currency in the Romance languages and presupposes a collective sense of the neuter plural form), and then a reinterpretation of the plural neuter *d mina. In turn, it goes back to an action noun *deh2-mn ‘division > slab, slice’, and is directly paralleled by such terms as st men ‘thread, warp’ or ab-d men ‘belly’, built from the *CeH- roots. As Sergio μ ‘division > Neri (Munich) points out to me, Gk. population’ could be its thematic derivative *deh2-m(n)-ó-. If the old connection of l mina with l tus ‘broad’ were accepted, it would mean ‘broadening’. But in that case, we would expect an evolution *stel(h2)-mn > stol(u)men (with a full grade of the root, as is usually the case with the proterokinetic primary formations inherited by Latin, like agmen, germen, cr men, columen, termen, men, etc.), and *stel- is not even certain to terminate in a laryngeal (see Rix et alii 2001: 594). A middle participle in -meno-, that would connect l mina with f mina, looks less likely, but cannot be rejected out of hand. The Journal of Indo-European Studies What became of “Sabine l”? 479 *dh2p- > lapi One of Pacuvius’ fragments (of his play Periboea, 276) reads lapit cor cura, aerumna cor conficit. EDLIL, s.u. lapit, while quoting in passing some possibilities considered in the past, like *lep- ‘to peel’ is noncommittal as to its etymology. But EDLIL, s.u. leps ‘charm, grace’, goes so far as to reconstruct a present *lop-i- ‘to peel’ with labial dissimilation, and relates it to Gk.  ‘to peel’. LEW, s.u. lapit, reconstructs an implausible  Ablaut variant *l p- of the same root. While the obvious meaning is stated by Paulus ex Festo 105, Lindsay, who glosses lapit as afficit, an alternative interpretation has found its way that goes back to Nonius Marcellus, who states: Lapit significat obdure facit, et lapidem facit (p. 34, Lindsay). Although Nonius’ translation is mere etymological play, recent etymologies indirectly rely on its superficial attraction. For instance Weiss, in spite of giving a correct assessment of its meaning (2010: 157-158) resorts to the hypothetical connection with lapis to underpin his interpretation of U. vaputu as a possessive deinstrumental *lapto-: “a verbal root could have produced a thematic noun *lapos ‘cutting’ from which a *lapto‘endowed with cutting’ > ‘knife’”. I find this argument unsubstantiated, however, since such a verbal root does not exist (*lapid- is probably a non-IE Wanderwort), and I doubt if a knife can be described as being endowed with an action of which it is the instrument. Instead, the obvious reconstruction is *dh2p-ie/o- ‘to 21 lacerate, tear to rags’, the full match of Gk. 
. The A identification with Toch. verb t
p(p)- ‘to eat’, first proposed by Fränkel (1932: 7) is still occasionally invoked as a cognate, but it is not likely to belong here under acceptance of Winter’s rule (whereby IE /d/ > /ts/; see Winter 1962). At all events, if the proposed change da- > la- is of Proto-Italic age, vaputu can be equally explained as a thematicized (agent?) noun *dh2p-t‘cutter’ (with /o:/ spread from the nominative), admittedly an 22 infrequent formation. 21 As suggested in passing by Conway (1893: 157) in his treatment of “Sabine l”, and rejected without as much as an argument by LEW. 22 Garnier (2015) denies the existence of a laryngeal and favors a (otherwise poorly attested) root *dep- ‘presser vivement’, of which *dap- would be a neo- Volume 47, Number 3 & 4, Fall/Winter 2019 480 Blanca María Prósper *dh2-tu > U. vatuva, Lat. latus If the above conclusions are right, all the Umbrian forms beginning with va- must be reconsidered, since in principle they can go back to Italic *da-. It could be the case, for instance, that the much debated form vatuva, vatuvu, a neuter plural that performs the role of the direct object in the sentence vatuva 23 ferine feitu, belongs here. It is often translated as ‘sacrifice the victims on a tray’. But, in view of some expressions preceding it that refer to portions of meat (proseseto), Weiss (2010: 230-234), has argued that the instruction contained in this sentence cannot refer to the slaughtering, which has already been performed, and that vatuva is more likely to designate some animal parts. One may consequently be tempted to reconstruct an adjective *dh2-teu-o- > Proto-Sabellic *latouo- > *atuo- (after syncope) > *uatuuo- (with anaptyxis) and translate ‘place the parts to be distributed on a tray’, that is to say, the animal parts that are to be offered or rather given out to mortals. Alternatively one could postulate the athematic neuter plural noun *dh2-tu-eh2. There is an obvious parallel in the ritual associations of Gk.   ‘meal’ (with remodeling of the suffix as a feminine in -t- and root extension favored by the synchronic association with  μ ‘offer a meal’, 
μ ‘ate a meal’), and Skt. d
tu (neuter) ‘part’, Av. v-
tu- ‘division, destruction’. An overall comparison suggests that the extended root *deh2i- is original in the agent nouns   and possibly also μ  ‘minor divinity, daemon’, but analogical in   and   , whose cognates show the basic form *deh2- in IndoIranian and Italic. Since neuter forms in -tu are thin on the ground, this constitutes a remarkable archaism. zero grade. Nonetheless, Rieken (2017) draws attention to an overlooked Anatolian instance of this root which confirms the presence of a laryngeal sound and provides a near cognate of vaputu: Hitt. tap(p)aštai‘slaughtering block’, in her view from a possessive adjective *dh2p-s-tó-. An isolated gentilic LAPSCIDIVS (CIL X, 4200, Latium et Campania; IX, 3406, Samnium), identified with LAFCIDIVS (Rome, CIL VI, 35647) by Bücheler (1891: 235) is a trivial derivative in -idius from a form *dh2p-s-(i)ko- which might contain the same -s-stem, and, as he prudently observes, cannot be related to either lapis or scind. 23 See Meiser (1986: 155) for the velarization of initial l- in Umbrian. The Journal of Indo-European Studies What became of “Sabine l”? 481 Interestingly, this might go some steps towards a final solution of U. vatra, a form attested only once in the formula sakre vatra ferine feitu (Tab. Ig. III, 30-31) and occasionally explained away as a misspelling for the quasi-synonymous vatuva, attested twelve times. As pointedly stated by Weiss (2010: 225, 231), “the similarity of vatra and vatuva, whatever its ultimate explanation, is hardly fortuitous” and “it would be hard to imagine that vatra were anything but a part of the 24 animal”. In short, if both forms have compatible origins, because they share the same root, their respective etymologies will support each other. In my view, nothing stands in the way of identifying vatra, mutatis mutandis, with Skt. d
trám ‘part, awarded part’ and Gk.   ‘share, part’ (once again showing the root extension -i-). In itself, this does not rule out the possibility that the scribe has misperceived or misspelled the expected vatuva, which differs from it by one single stroke (if we further resign ourselves to a “corrected” form vatua for vatuva, which, as remarked by Weiss 2010: 226, is uncompelling). But, even if he has, his error may be suggestive of the existence of a form that was readily confusable with the correct one, both for formal and semantic reasons. Sakre(s) vatra can be taken to mean ‘parts of the piglet’. Vatuva, on the other hand, may be specifically associated with distribution and not only with physical segmentation, and the functional specialization of these nouns is the expected thing to happen, as in Greek or Sanskrit. We may add to the above list Lat. latus ‘side’ (< ‘part’), which would come from a secondary stem *dat-os with the -textension attested in Gk. 
μ, and probably in Goth. un24 The connection of both Umbrian forms and the proposed meaning are not new, however: To my knowledge, the connection of vatuva with Lat. latera was first put forward by Bottiglioni (1954: 260, fn. 5), who translated ‘le parti delle vittime tagliate’ and equated vatuva: vatra to Lat. pecu: pecora without further ado. Soon after, Poultney (1955: 77-79) independently arrived at the same conclusion, translating ‘side portions, ribs’. The word formation remains unclear in both cases, and it is unwarranted to identify Lat. *latesa > *latera and U. vatra, since Umbrian syncope preceded rhotacism. Weiss’s own solution (227-228) starts from a heteroclitic noun *uat-ur, *uat-uen-. The Sabellic neuter pl. of a secondarily leveled paradigm would be *u
a tur, which would give rise to U. vatra, while vatuva would have been mechanically created after the singular form became *uatu by (unparalleled) deletion of -r-. Volume 47, Number 3 & 4, Fall/Winter 2019 482 Blanca María Prósper ga-tassans, acc. pl. masc. ‘undisciplined, idle’, etc. (see Lehmann 1986-2: 377). On a slightly different, more attractive possibility, however, latus could ultimately reflect the full match of vatuva, namely a neuter *dh2-tu. Since it lacked a synchronic relationship with a verb, it was refashioned in the same direction as gls, glris ‘sister in law’. Perhaps one could suppose, in the wake of Nussbaum (1986, 145-146), that a number of adjectives in -u- were enlarged by -s- to form endocentric derivatives whose neuter was reinterpreted as a noun (conjecturally in *uetu- ‘old’ *uetu-s ‘id.’). An isolated neuter form *latu could simply have been dragged along by this tendency and redone as latus. Other proposed etymologies of latus, for instance the one which identifies it with OIr. sliss ‘side’ or leth ‘side’ (cf. DLL, s.u. latus), or the one that relates it to l
tus ‘broad’ (cf. LEW, s.u. latus), are phonetically untenable. To be sure, given the above evidence one could argue that we are dealing with a change that affected exclusively the root *deh2- and its enlargements. But most of the reviewed examples have probably been disconnected from each other at a stage previous to Proto-Italic. Instead, I reckon with the possibility that the unitary color of the vowel in all the possible instances of this root automatically turned it into l-. This would seem to leave the apparent preservation of dain daps and damnum (< *dap-no-) unexplained. But, beside the fact that some forms may have fallen under the influence of d, dre early on, such derivatives as dapslis and daps
lis are, in spite of EDLIL’s silence on the matter, nothing but early loanwords from Greek   ,   ‘abundant’; and the apparent synchronic derivative dapn, -
re is by all accounts a borrowing from Gk. 
, and a very primitive one if it th regularly underwent vowel reduction in the 6 -5th C. BC (or earlier). Daps would be from the start more naturally associated with dapn, etc., than with lapit. Since this root noun remains otherwise unattested, one could even entertain the idea that it was back-formed to the abovementioned verb and adjectives in Proto-Latin. This would match the meaning well: note that a root-noun *dh2p- would be more likely to mean ‘cutter’ or ‘cutting, carving, etc.’ than its result ‘(sacrificial) feast, banquet’. The Journal of Indo-European Studies What became of “Sabine l”? ? 483 h *dh2b -s > labr/labs The -s-stem labr/labs ‘work, labor’ (since Naevius) has no visible connections either. EDLIL prudently rejects the cognacy with lab ‘to waver’ (which in turn seemingly goes back to *slh2b- and has cognates in Balto-Slavic and Germanic only). By contrast, Lith. dobti ‘to strike, beat, kill, (dial.) torment, exhaust’ would provide a formally and semantically unimpeachable parallel. Germ. *tabjan ‘to hinder’ could possibly be added. The metaphoric usage of words originally related to suffering and torture to designate ‘work’ is familiar from tripalium > Sp. trabajo, PSl. *mka > Rom. munc, etc. (h) h EDBIL, s.u. dobti, traces this form to *d eh2b - and cites the 25 opinion of W. Smoczy
ski, who suggests a direct comparison with *deh2p-. Accordingly, we may be dealing with an alternative enlargement of the same root *deh2-. The Germanic 2 evidence is conflicting, and EDPG, s.u. *dapp/bn- pools MEng. dabben, etc. ‘to beat’ with the Baltic forms above under a root h h form *d eh2b -, which would destroy the connection with the Latin form unless an unrestricted version of Grassmann’s Law is accepted for Latin, which will never be the case. But it cannot be ruled out that there is one single verb *dapp/bn- from h h *d ob -nh2- embracing both the transitive meaning ‘to diminish, oppress’ and the intransitive ‘to deteriorate’, since both are only present in Nw. dabba. h *dlh1g h1ó- > largus The adjective l
rgus ‘abundant, generous’ remains, again, 26 unexplained. In principle we could be dealing with a 25 Slownik etymologiczny jzyka litewskiego, Vilnius 2007: 117 (non vidi). The h h h h tentative reconstruction *d b -ie/o- in Rix et alii (2001: s.u. *d eb ‘vermindern’) is most unlikely. 26 The long vowel is secured by the apex in CIL VI, 32521, but grammars and monographs usually omit the fact that it is only used in fragment B, col. 2, which reads LAR´[-] with a displaced apex, while the rest of the inscription mentions the consul as LARGO without apex. Incidentally, this would mean that the Latin version of Osthoff’s Law took place in Latin, if at all, before (this layer of) syncope. An often advocated rule a > /_rC disposes of the problem directly. Volume 47, Number 3 & 4, Fall/Winter 2019 484 Blanca María Prósper 27 compound. On a more intriguing possibility, l
rgus could originally mean ‘long’ (the default meaning inherited, e.g., by h Spanish) and then become the direct match of *dlh1g h1ó- in Skt. drghá-, Lith. ìlgas, OCS. dlg ‘long’, cf. Gr.   , 28   , Goth. tulgus, Hitt. daluki-. For this to be possible, the expected outcome *dl
o- should have undergone metathesis early on to avoid the sequence dl-, giving PItal. *d
lo-, thereupon, ex hypothesi, *l
lo-, and eventually by regular dissimilation the attested Latin form (the metathesis putatively h preceding the whole process is impossible for *dloh1ng ó- > 29 longus, in which the simplification of the cluster may be an 30 earlier, more general western-IE phenomenon). Of course, under acceptance of the “palma-effect”, the predicted outcome 31 could alternatively be *dalao-, which would directly lead to 32 the attested largus by dissimilation and syncope. 27 From *d
l-go- ‘part-bearer’ (and belonging to the compositional subtype of prdigus, with the same syncope as the compounded verbs in -age/o- like purg, iurg), through the stages *ll-ago- > *lr-(i)go- (by dissimilation and vowel reduction/syncope), and then a direct cognate of IE *deh2-lo- in OIr. dál ‘part’. 28 The word is originally a collocation with different variants, whose second h member (the governing verb) is IE *g eh1- ‘to reach’ and the first is *delh1‘length’. Cf. Neri (2007: 53-54, fn. 149). Pinault (2017) has defended an h etymology *dlh1-b u(h2)-ó- ‘being/becoming distant’ for Toch. tsälp- ‘to pass away, be released’ (I owe this reference to Sergio Neri, München). Its relatedness with the causative or stative verb form indulge ‘be generous, concede’ is not certain (Rix et alii 2001: 113 and EDLIL do not even consider it). But Havet (1889: 232) had already traced back both largus and indulge to h (the contemporary version of) *dlh1g h1ó-; his now unjustly forgotten idea is however quoted with due appreciation by DLL, 609. 29 h For Pinault (2017: 653) *dlh1g (h1)-ó- became the basis of a present with h nasal infix ‘to lengthen’, which gave rise to a thematic action noun *dlónh1g h o- > *dlong o- by the “Saussure effect”, which later became adjectivized. 30 Cf. Marotta (1999) on the different strategies employed by Latin to bypass the “coronal syndrome”, a name for the inability of coronal stops to govern liquids in the slope of a syllable. Latin has neither (-)dl-, nor (-)dr-, nor (-)tl-. 31 Cf. Stuart-Smith (2004: 47-48) on intervocalic [
]. 32 With syncope favored by medial -r-. In another work (Prósper 2017) I have advocated an explanation of the “palma-effect” based on context rather than on accent (note that the present case would require accent retraction, which is unwarranted). In my view, it is the essential phonetic incompatibility of the laryngeal in coda position with the following labial consonant (especially [u]) The Journal of Indo-European Studies What became of “Sabine l”? 485 *dHk-/*d- > lak-  laqueus Laqueus ‘loop, rope’ equally lacks an etymology. The often vindicated connection of laqueus with the root of lacit, lax (see below) on the grounds that the latter forms end up in a labiovelar (Manu Leumann et alii 1977: 148, LEW, s.u. laci) is unwarranted from the formal point of view and has no semantic support whatsoever (see EDLIL, s.u. laqueus). DLL, s.u. laqueus, in describing it as a “terme technique qui est sans doute emprunté, comme beaucoup de mots en -eus. Etrusque?” simply avoids the question, poses another concerning the u
“borrowed” sequence [ku
] or [k ], and, faute de mieux, predictably resorts to the obscura per obscuriora fallacy, which in Latin linguistics typically consists of ascribing to Etruscan everything that is not immediately understandable (not to mention that the adduced loanwords that could justify this escape route are Greek and their sources are perfectly identified, as in coccineus, purpureus, pniceus). Instead, it can be more persuasively related to a number of IE forms meaning ‘hair, thread, wick’ which according to the traditional account presuppose a root *de- (in Skt. dá
‘fringe’, OIr. dúal ‘lock’, Goth. tagl, SCr. dlàka ‘a single hair’), but in view of some forms are more convincingly traced to *deh1k- by EDPG, s.u. tahjan. This work does not take the nonGermanic cognates into account and sets up this root for Goth. tahjan ‘to tear, lacerate’ < *dh1k-, MHG. z
che ‘wick’ (< *deh1kon-), Far. tág ‘sinew, fibre’ (< *deh1k-eh2), Goth. tagl ‘hair’ (*dh1k-ló-). Accordingly, I would reconstruct for the base of  laqueus a verbal adjective *dh1k-uó- (or alternatively *d -uó-) which equally came to mean ‘hair'. Its substantivated collective A form would then be preserved in Toch. 
ku ‘head hair’, if 33 from *deh1k-u-eh2. Latin adjectives in -eus typically refer to the materials an object is made of. And, albeit metaphorical that causes a schwa to be inserted between them, giving rise to a trisyllabic structure. In our case, a fricative velar or postvelar sound would be the factor behind the resyllabification. 33 Cf. Adams (1982-83), who sticks to the root-form *de- and reconstructs A/B *d-u-eh2, whose singular form *duom/*du would occur as yok ‘body hair’. The laryngeal formulation looks superior in that it reveals previously unnoticed lexical associations and spares us from reconstructing two homonymous roots *de- and an unmotivated lengthened grade. Volume 47, Number 3 & 4, Fall/Winter 2019 486 Blanca María Prósper extensions are of course usual, giving rise, for instance, to the designation of colors, and other nuances have been attached to it in the course of time, this particular derivative has no noun beside it, and is consequently more likely to preserve the typical structure of a Stoffadjektiv, like v
mineus, argenteus or aureus (see Leumann et alii 1977: 286). In this vein, we would expect laqueus to be built from a noun meaning precisely ‘thread, hair, wick’, after which it was substantivated to 34 designate specifically an item made of plaited strands. *dh3-k- > Lat. laci, lax; U. VAS, antervakaz, vaçetum A change #da- > #la- neatly explains the Venetic preterite la.g.sto (if the restitution is right at all) in the sequence dona.s.to ke la.g.[s/to] ‘donated and presented’ (Lejeune 1974: 12A, Este, translated in 12B as DONOM DEDIT), which is habitually related to Lat. laci, perfect -lexit, translated as ‘entice’. In fact it is only attested as a simplex lacit in a late text: decipiendo inducit and lacit: inducit in fraudem (Paulus ex Festo 103, 104, Lindsay), and the meaning may consequently be suspected of being abstracted from compounds. In my view, Ven. la.g.sto, Lat. lacit, -lexit are the remainders of an Italic verb *dak-ie/o- which was created early on by means of a suffix -k-, and then following the same procedure as with *fak-ie/o-. It conveyed some specialized meaning with regard to Lat. dedit and Ven. doto, dona.s.to ‘gave, offered’ which, as we are going to see, was probably close to ‘to 34 Under this assumption, Lat. lacer ‘mutilated’, as well as its derivative lacerare ‘to hurt, tear’, whose meaning is identical to that of Goth. tahjan, could be traced to a noun *dh1k-ló- ‘rag’ which can be matched with Goth. tagl, possibly SCr. dlàka ‘a single hair’ and goes back to a verbal derivative in -lo- with passive meaning. This would only be possible as long as OIr. dúal ‘plait, lock, tress’, which is compatible with a variety of preforms, is given up as a member of this set of forms; both dúal and Skt. da may go back to *de‘perceive’ in the secondary sense of ‘adornment’, *den- ‘bite, press together’ and refer to the seams stitching the fringe of a tissue or to its usual appearance (cf. Fr. dentelle ‘old lace’). Were this true, the nasal stem reconstructed by EDPG, s.u. tahjan, for MHG. zache ‘wick’ would have the holokinetic structure nom.-acc. stem *deh1k-on- (for the Germanic forms), locative stem *dh1k-en- (possibly for Lat. lacinia ‘edge of a garment’). The traditional connection with Gk. λακ
ζω ‘to tear’ is equally conceivable, however, for which see EDLIL, s.u. lacer. The Journal of Indo-European Studies What became of “Sabine l”? 487 exhibit, display, place on show, within reach’ > ‘show something off as an attractive asset’. The compound verbs share some traits with the compounds of dc and have a personal DO suggesting a meaning ‘(mis)lead’ which may be favored by the derivatives of the compounded root nouns: like artifex gives rise to artificium, once lax had arrived at the meaning ‘decoy’ > ‘fraud, transgression’ (see below on Umbrian vas), and a compound in -lax like in-lex means ‘seducer, inductor’, its derivative inlicium is the ‘action of seducing, carrying away’ but also ‘summoning’. In a well-known work, Untermann (1993) called into question the idea that there is a close prehistoric relatedness 35 behind the equation of Gk.  and Lat. ic,  and fc, and argued in favor of a root enlargement -k-, by means of 36 which the new present stem was derived. Be it as it may, the Latin present faci, the Sabellic subjunctive O. fakiiad, U. façia 37 and Lat. iaci are early Italic innovations and in my view it strains imagination to separate them from the Latino-Faliscan rd inherited aorist form 3 sing. *fced, pl. *fakond that gave rise to Latin sing. FECED, pl. FECRONT vs. Faliscan faced, facet by 38 opposed processes of paradigm leveling. Consequently, Italic *dak-ie/o- becomes the missing piece, even if the preterite form, which in prehistoric times must have been an aorist *d(e)h3-k- like Gk. , has been redone into a sigmatic preterite *lak-s- both in Latin and in Venetic. There is A Venetic preterite vha.g.s.to, but neither the inherited †fk-e-t nor †fak-to. The apparently monoglottic replacement of †fk-et by Venetic *f
k-s-to may have been triggered by the combined influence of *l
k-s- and the future/desiderative va.g.son.t. in the 35 This form in turn must be considered a match of the Phrygian prefixed rd preterite αδδακετ and, I would add, Celtiberian 3 person singular tekez rd (Luzaga bronze tablet, from Celtic *dket), 3 person plural tako (Iniesta lead letter, from Celtic *dakont, cf. Prósper 2007: 85-86). 36 See Har
arson (1993: 149-150) for an overview. 37 See Wallace (1988) and Coleman (1990: 10-11) on the dubious status of dialectal Latin PROIECITAD. 38 Cf. Wallace (2005). In another work I shall try to show that Latin never inherited an IE perfect form VHEVHAKED. Volume 47, Number 3 & 4, Fall/Winter 2019 488 Blanca María Prósper 39 Tavola d’Este. It is consequently not necessary to posit a prehistoric aorist *fax for Latin. The enlarged variants of these roots have also given rise to the agent nouns Lat. -fex ‘maker’ as in ponti-fex, arti-fex (< *fak-), lax etenim fraus est (Paulus ex Festo, 103, Lindsay) and inlex ‘seductive’ (< *dak-), sub-ics ‘underlying parts’ (< *iak-), and cf. the parallelism of facess ‘to carry out, perform’ and 40 lacess ‘to challenge’. Additionally, this seems to confirm that the vocalism of iaci is secondary, and contains a morphological zero grade which has been modeled on faci and laci (**ci would be the phonetically expected form). In sum, all the Greek CeH-verbs having a -k- aorist are likely to have a cognate of the structure C-k-ie/o- in Italic. We would in principle not expect a formation *st-k-ie/o-, since the corresponding secondary, transitive Greek aorist is 
‘placed’, beside intransitive , Skt. ásth
t ‘stood’, which is suggestive of -k- being restricted to originally exclusively transitive roots. In fact, we find U. stakaz (nom. sing., past part.) which is usually traced back to *st/
k-
-to- ‘established, fixed (vel sim.)’ (cf. WOU, AB 700). But cf. also Toch. t
k
-, found in subjunctive, preterite, and imperative stems for ‘to be, become’, ultimately from PIE *(s)teh2-k- (see Malzahn 2010: 157 for formational details). More recently, Lazzarini - Poccetti (2001: 168-180) have put forward a number of interesting hypotheses on the segmentation and reading of Palaeo-Italic  
[.] q  (cippus of Tortora, C3). After the pronoun   ‘nobody’ (< u *ne-k is), they identify an imperative form expressing a prohibition, which is in turn amenable to a number of interpretations, hampered by the small lacuna. One of the 41 restorations they propose is ( )
[] q , which they take from *st
-k-ie/o- + -se/o- (I would slightly correct this pre-form as *st-k-ie/o- in line with the above arguments). We can only 39 See Prósper (2018c: 466 and fn. 20) for a discussion of this form, in which <v> in all likelihood stands for /f/. 40 Benedetti (1988: 121) ventures a relationship with Skt. words meaning ‘rope’ (rami-, etc.), and reconstructs a root *la- which is merely ad hoc. 41 The correctness of the reading <κ> has been corroborated by Crawford et alii (2011 III: 1337, on autopsy). The Journal of Indo-European Studies What became of “Sabine l”? 489 speculate about how this structure came about, but the stem of rd the plural forms, for instance a 3 person plural *fakond (directly attested in Celtiberian tako, see fn. 34), may have constituted the derivational base for a present in -i- or -ie/o-. Thereupon, most of the resulting stems give rise to secondary st 1 conjugation stems, especially in compounds, namely Lat. fic
re, -lic
re and U. *stk
-. Interestingly, Umbrian might preserve still another continuant of this stem, if the past part. antervakaz ‘interrupted’ (vel sim.) is traced back to *anter- + dk-
-to- (Lat. *inter-lic
re, cf., mutatis mutandis, Lat. interfici ‘to put to death’) and the connection with Lat. vac
re ‘to be empty’, which is as formally impeccable as it is semantically unconvincing, is consequently abandoned. It is unclear whether other Umbrian forms traditionally linked with antervakaz should be brought to bear here, but it is at least very tempting ? to identify Lat. lax ‘fraud’ with U. VAS ‘ mistake, fault, transgression’ (Tab. Ig. VIa 28, 38, 48; VIb 30), a connection which has not been suggested to my knowledge. This form is unfortunately compatible with several reconstructions, but, given its plausible relation with vaçetum/VASETOM, an ancient periphrasis probably meaning ‘(gone) wrong, faulty’, it could go back to a root noun *dk-s ‘fraud, deception’ (< ‘decoy, lure’ < ‘something offered to sight’), only extant in the formula TVER PESCLER VIRSETO AVIRSETO VAS EST ‘if there is any transgression in your sacrifice, whether visible or invisible’. Since it is likely to be a feminine given the appearance of its epithets, a masculine past participle in -tos and a neuter in -os can be ruled out. None of the root etymologies offered by WOU has any power of persuasion. As regards the Oscan evidence, Mancini (2006) has interpreted the debated verb form aflukad in the longest preserved Oscan defixio (ST Cp 37, Crawford et alii 2011 I: 443445; l. 3) as the subjunctive of a stem *af-lk- which contains the root we are discussing with lowering/rounding caused by the following velar segment. He additionally treats the problem of the future perfect aflakus in the same inscription (l. 11) which, in view of the different vocalism, he ascribes to a different origin: while aflukad is ‘take out, cause to come out, elicere’, in the ritual context ‘to summon her will (of the Volume 47, Number 3 & 4, Fall/Winter 2019 490 Blanca María Prósper goddess Ceres), the legions of the demons’, aflakus is taken to mean ‘to sacrifice, give as a sacrificial victim (Pacius Clovatius, the victim of the curse)’. This is not indisputable, however, since the same hesitation is attested in FACVS, PRAEFVCVS, FEFACIT, FEFACVST (Bantia, ST Lu 1, Crawford et alii 2011 III: 1438) and it is strains imagination that such similar forms would have completely different origins; unfaithful rendition of a non-phonemic, back unrounded vowel is preferable as an explanation. Accordingly, I deem a meaning ‘to draw forth’ plausible, which one may freely translate as ‘unleash’ in the first case, and in the second, sakrim: svai: puh aflakus huntrus teras huntrus a[pas, as ‘if you have dragged him under ground, under water, as a victim?’ (context and syntax are unclear, however). Accordingly, starting from the present form *dakie/o- > *lak-ie/o-, Oscan has created a preterite based on *dk-, reflected in a root subjunctive stem *lk-
- and a future perfect *lk-us-. This leaves us with two innovative and divergent perfect formations, *lk-s- in Venetic and Latin vis à vis *lk- in rd Sabellic. The form
f (3 pers. sing., ST Lu 13), which Mancini and WOU 59 tentatively relate to aflukad, may reflect the asigmatic perfect indicative. In sum, a present form *dk-ie/o- > *lk-ie/o- constitutes a formal and semantic match of the rest of the enlarged verb forms *CeH-  *CH-k-. As in the case of faci vs. con-d, the relatedness of enlarged vis à vis -k-less forms was irremediably lost for the speakers of Latin. Since compounded forms of h *deh3- and *d eh1- would have tended to merge formally, it was the enlarged form *dk-ie/o-, and not the simplex, which specialized in the creation of compounds in the first place (alternatively, the basic forms actually merged, and the enlarged forms, which carried different meanings, progressively found their place in compounds). While Untermann is right in that -k- serves a specific function as an enlargement preventing phonetic confusion of the shorter root form f/dV- and probably also provides further semantic specialization, the fact that we now find -k- in all the roots for which there is a -k- aorist in Greek can hardly be overlooked. The Journal of Indo-European Studies What became of “Sabine l”? 491 *dm
h2- > lam-  O. lamatir All the above arguments could go some steps towards rd clarifying the obscure etymology of the Oscan verb lamatir (3 person singular middle) in the sequence suai neip dadid lamatir (Capua, ST Cp 37, Crawford et alii 2011 I: 444). In my view it is likely to reflect a root *dam-, which is both phonetically and semantically plausible: It means something like ‘if he does not give it back, let him be punished’ and given its context it must have been very prone to be misperceived or unconsciously dissimilated. But the same form is attested in Bantia (ST Lu 1, Crawford et alii 2011 III: 1438) as SVEIPIS NEIP CEBNVST ... COMENEI LAMATIR ‘if he does not come to the census [...] let him be punished in public’ and then points to an early change. Predictably, lamatir has been taken from a poorly attested root *lem(H)- (Rix et alii 2001: 412, corrected into *h3lem(H)- in Kümmel 2019) which is said to have survived in the Venetic 42 personal names Lemetorei, Lemetorna (Este) and in Lat. lanius ‘butcher’ (from Plautus), a form whose appurtenance is in turn unclear, and which is uncertain to belong here in view of its root vocalism. EDLIL, s.u. lanius, reconstructs an agent noun *h3lomH-io- assuming the root vowel has undergone a “labial dissimilation”, in turn posited ad hoc by Schrijver (1991: 475) to account for the hypothetical evolution *lo- > *la- in *loku- > lacus ‘lake’. However, in view of recent research lacus probably contains an inherited sequence *la- after all, and most scholars would probably agree that the paradigm had the form *loku-, gen. *laku-s, whether for phonetic or analogical reasons, in 43 Late-PIE. The allegedly related MIr. laime ‘ax’, only attested 44 once glossing bíail and on the whole unreliable, contains a zero grade, possibly *lm
(H)ió-, and can by no means be equated to lanius under a common ancestor *(H)lomió- (pace EDPG, s.u. 42 See Prósper (2018c) on the possibility that it goes back to *nemh1-, as first proposed by Lejeune (1974: 52). 43 See Bichlmeier (2009) for a number of possible reconstructions of this stem as well as several Celtic hydronymic instances of *lakuV- > -lape. Since, on the other hand, a greater degree of vowel-to-consonant coarticulation is expected if the consonant is tautosyllabic, one does not quite see why we would get a dissimilatory reaction to *lo- in lanius but preservation in longus, where it was followed by a velarized nasal [
]. 44 This filiation was suggested in passing by Stokes (1904: 258). Volume 47, Number 3 & 4, Fall/Winter 2019 492 Blanca María Prósper *lamjan). If the Latin form, in turn, came from *lm(H)ió-, we u would expect †lenius (otherwise, IE *g m-ie/o- would never have given Lat. veni). Alternatively, we could start from *lm(H)-iió-. But in that case, this form should never have given lanius with context-bound delabialization of -m-, but either †lamius or †lomius, like eximius or praemium (see below on the expected root vocalism). Needless to say, the association of lanius with the family of Lat. lac- in lanius, qui disci<n>dendo lacerat pectora (Paulus ex Festo, 103, Lindsay) is phonetically impossible: *lak-ni- would have given †lagnius, and a nasalized *lank-ni- (as in lancin
re) would have resulted in †l
nius. In short, this outcome cannot be neatly accounted for, and, again, lanius can be traced back to *dh2-nó-/-ni-, a derivative of *deh2- ‘to divide’ (cf. Skt. diná- ‘divided’, Gk. 
 ‘debt, loan’, or the gloss 
 ˙ μ ‘portion’ in Hesychius). If lanius were a regular derivative of *dh2-ni-, either *dh2-ni-ó- or *dh2ni-ió-, it could be plausibly equated with Venetic dane.i., if it is the dat. sing. of a noun *d(e)h2-ni- ‘division’ occurring in the Tavola d’Este (see Prósper 2018c, 468), in which the preservation of *da- has to be explained via crossing with the outcome of *deh3-, as in Lat. d, dare, or directly belongs to a PItal. proterokinetic noun *deh3-ni-s, *dh3-nei-s, like OCS. dan, Lith. duõnis ‘tribute’ and directly related to the Latin -t-stem ds, dtis ‘dowry’. The butcher (in fact a specialized office) is simply the person who has the skills required to cut up the animal (Sp. des-piezar, G. zer-legen, etc.), not the one who, much 45 more broadly speaking, breaks, hurts or destroys. The gentilic Lanius is likely to preserve the original basic meaning ‘distributer, divider’, like Laterius and Laetorius. On the semantic side, the root *lem(H)- is taken to mean ‘to break’, and not ‘beat’ or even ‘put to death’, as habitually assumed in order to come to terms with the gap between a legal 45 For all DLL’s efforts to consider the verb lani ‘dechirer, mettre en pieces’ as basic, and the form lanius and its meaning ‘butcher’ as derived, this possibility does not withstand scrutiny. Lani must be denominative to the early attested lanius (in turn a primary derivative from the root, by no means a productive formation in Latin) and the meaning ‘tear to shreds’ (with teeth, nails) must be secondary: cf. the figurative meaning of Sp. carnicería, Eng. butchery as ‘carnage, slaughter’. Only the alternative etymology *d(e)h2enables us to follow the semantic path correctly. The Journal of Indo-European Studies 493 What became of “Sabine l”? 46 document and a defixio. And one can be reasonably sceptical as to whether such a horror was destined for offenders respectively guilty of robbery, as in the curse of Vibia (which is quite conceivable as an exaggeration in the general context of a malediction) and of not coming to the census for the correct valuation of his property, as in the Tabula Bantina (which teeters on the brink of the ludicrous). In addition, this meaning is mostly attested for the causative stem with an /o/ grade (as in OCS. lomiti, Germ. *lamjan, Lith. lamìnti). A number of indeterminacies surround the meaning of the basic form with /e/ or zero grade of the root, like Celtic *lam-ie/o- (deponent) ‘to dare’, Lith. lémti ‘to decide, determine’, and perhaps Gk. μ , adv. ‘unceasingly, without pause, restlessly’. Malzahn A (2010: 843-845) adds Toch. läm- ‘to sit’ and proposes an original meaning ‘to rest’ for this root. As a consequence, the proposed attribution of lamatir and lanius (and not †lomatir or †lonius) founders not only on formal but possibly also on semantic grounds. From the lexical point of view, lamatir may go back to the root *demh2- ‘to subdue, tame’, which has a nasal present Gk. 
μ μ, OIr. damnaid, Skt. dam
ya-, and a root aorist (as in the Greek participle 
μ -), etc., which has active sense and directly reflects its prevocalic zero grade *dmh2-. This would point to a Proto-Italic change *-m/nHV- > -amV-/-anV- in the first syllable (as opposed to -mo- in medial position in ordinal numbers and superlatives). While this outcome is far from uncontroversial, the alternative etymology *(H)lm(H)- equally depends on its acceptance or on complex analogical processes. And, if it is categorically refuted (the current opinion is presently divided between those who favor an outcome -emVand those who prefer -omV-), we face the usual problem: lamatir would have to be considered a denominative formation 47 from *lH-mV- which, again, begs the question. 46 See Murano (2010: 61-63) for the state of the art. The evidence in favor of #Cmm- > Italic #Com- is rather paltry: it is not demonstrated that sum, sumus directly continue *(e)smm- and not *(e)som- (cf. EDLIL, s.u. sum, esse), and hom is a conflictive enough form to base anything upon it and possibly reflects an older *hem as traditionally assumed (additionally, the preceding consonant favors an obscure vowel). Humus 47 Volume 47, Number 3 & 4, Fall/Winter 2019 494 Blanca María Prósper By contrast, Latin preserves the iterative /o/-grade stem dom
re, and the alternation l-/d- may testify to early paradigm split. Our hypothetical stem †dam
- possibly continues an aoristic subjunctive stem *dmh2-eh2- directly built from the root and then equatable to Lat. tag
-, fuat, -tulat, U. habas. Therefore, it may be formationally identical to but syntactically different from the “preventive” reconstructed for this particular formation by Rix (1998), which is logically preceded by a sentence negation: in our case it apparently functions as its positive version, namely as an imperative. Whether this is due to a secondary functional development like those of the Latin extra-paradigmatic -
-subjunctives remains debatable. *deh2-nu- ‘river’  L
nuvium I think there can be little doubt that we shall find further examples of this rule in the comparatively unexplored domain of onomastics. L
nuvium is a city in Latium situated to the SW of Rome, whose construction probably dates back at least to the beginnings of the first millennium BC. It goes back to *d
neu-, a derivative of *d
nu- ‘river’ in Skt., Av. d
nu- ‘stream, river’, also attested as a RV. thematic derivative d
navá- ‘name of a class of demons’ (*d
neu-ó-), an Av. athematic ethnonym nom. ? pl. d
nauu ‘ inhabitants of the river area’ (*d
neu-es), as well as probably the Greek ethnonym
. The original noun may have been a proterokinetic *deh2-nu-s, gen. *dh2-neu-s. Accordingly, Greek
 (if it is not a loanword, as routinely assumed) could go back to *dh2-neu-ó- (which makes some difficulties with the vocalism of the suffix, however), but the root vocalism /a:/ would have been generalized early on in Indo-Iranian, where the alternation *d
nu-, *d(i)nau- was thus avoided. A further enlarged *d
neu-(i)io- has given rise to a number of (Pre-)Celtic river names, like D
nuvius ‘Danube’, the 48 Donwy in Wales, etc. probably contains an /o/ grade, also reflected in U. hunte, O. huntras ‘who is below’, etc. We are possibly dealing with late instances of the “Lindeman effect”, and then not quite the same context. At any rate, the few words that could contain this initial sequence are amenable to different explanations. Some other Latin forms like am or fams might go back to #CmHV-. 48 Cf. on the etymological ambiguities Mayrhofer 1992: s.u. dnu-, Mallory and Adams (1997: 486). Whether this root is homonymous with or identical to The Journal of Indo-European Studies What became of “Sabine l”? 495 What is more, there is a Roman gentilic Lanuvius, Lanivius and an Oscan name in the gen. sing.  [la:nu
iei
s] (300 BC, Campania, this is the new reading in Crawford et alii 2011 II: 927, Picentia 3) with syncope of the medial vowel. We may conclude that the Italic peoples inherited a river name or appellative, or, on a more cavalier assumption, an ethnonym *l
neuo- > *l
nouo-, possibly conveying long lost religious connotations, which regularly provided the base of the Latin and Oscan forms, including the famous place name. A beautiful, probably Celtic match of these personal names can be found in the filiative formula POLLIVS DANOVI / F(ILIVS) (CIL III, 4544, st Gottlesbrunn/Carnuntum, Pannonia Superior, 1 C. AD), besides D(IS) M(ANIBVS) DANVVI QVINTI (CIL II, 6301, Monte Cildá, Palencia, Cantabri Orgenomesci) or C(AIVS) RETONIVS / nd DANVVIVS (CIL III, 3581, Kalocsa, Pannonia Inferior, 2 -3rd C. AD). Consequently, L
nuvium clarifies a number of points: the western continuants of *deh2-nu- are not borrowed from Iranian, as occasionally assumed; Italic and Celtic have inherited an adjective *d
neuo- or *d
neu-i(i)o- from a conceivably common ancestor; the Italic forms unmistakably show a voiced segment (traditionally reconstructed as /d/) and h consequently the alternative etymology *d e/on- (as in Lat. fns ‘fountain’) can now be safely ruled out. Finally, Lat. l
ma ‘marshy place, bog’ (from Ennius) has been traced back to *leh2-meh2 on purely descriptive grounds, and compared with a number of Baltic and Slavic forms of possibly related, but not identical meanings: dialectal Lith. lãmas ‘piece, lump, plot, nest’, Latv. lams ‘piece-wages’, OCS. lom ‘marshy spot’, ORuss. lom ‘breaking, marsh, pool, woods ravaged by a storm, etc.’ and can be traced to *lom(H)o- (cf. EDBIL, s.u. lamas). There is also a form that is fully compatible with a reconstruction *leh2-meh2, namely Lith. lomà, lõmas ‘hollow, valley’, PSl. *lam ‘hollow’. It should be noted, however, that EDBIL, s.u. loma, relates this to the former group *deh2- ‘divide, distribute’, and there has been a considerably primitive semantic shift from ‘abundant, rich’ to ‘flowing’, is impossible to decide. The divine name *Danu-, extracted from the ethnonym danann, has a short root vowel /a/, which in principle should be taken from the oblique stem. Volume 47, Number 3 & 4, Fall/Winter 2019 496 Blanca María Prósper and posits a lengthened grade form *lm(H)-, explicitly jettisoning the cognacy of these forms with Latin l
ma. But if the Balto-Latin connection has to be dismissed as fortuitous, what is *leh2-? According to Rix et alii (2001: 401), Lat. l
ma may belong together with Hitt. l
hui, lahuanzi, the only continuant of a root *leh2- ‘to pour’, but Kümmel (2019) alternatively reconstructs *leh3u-, and the former entry has consequently disappeared. Again, a more persuasive hypothesis would link l
ma both to L
nuvium, etc., and, as a collective formation, to a number of forms traced back to *deh2-mo- and sharing a common meaning ‘fluid’: Gk. μ  ‘fat’, Arm. tamuk 49 ‘damp’. 5. Conclusions By way of conclusion, I suggest we consider the following scenario: At a very early stage, (what we usually reconstruct as) 50 #da- regularly gave #la-. The validity of the hypothesis that this change is of Proto-Italic age (a language stage which can be defined as the oldest common ancestor of Sabellic, Latin and Venetic) is inescapable if an implosive sound [
] was involved. In turn, this remains contingent upon our judgment of O. lamatir, U. vatuva, vatra, O.   (= Lat. L
nuvius) and panItalic *lk-. No evidence from the Sabellic languages decisively opposes these tenets, except perhaps the prefix and preposition *d
(-d), which in view of the fact that it looks like the ablative feminine of a pronoun whose Proto-Italic stem must have been *do- (cf. Lat. dum, d-nec), inferable at least from a nucleus of adverbials, and its plausible cognacy and synchronic synonymy with the preposition *d, does not qualify as counterevidence. 49 This form is especially relevant for the dialectal filiation of Lusitanian, which I have classed as an Italic or para-Italic language (the terminological debate would take us too far afield). In fact, the Lusitanian indigenous inscription of Lamas de Moledo contains a phrase in the acc. sing. ANGOM LAMATICOM, which designates an animal to be sacrificed to the local gods. It must be translated as ‘pasture lamb’ or agnus pascualis, meaning that it had reached a certain age, a well-known ritual condition for sacrifice. LAMATICOM is an obvious derivative of a form identical to Lat. lma. See Prósper (2002: 64), where I still stuck to the older interpretation of ANGOM as ‘valley’. 50 See Prósper (2018b) for the close genetic relatedness between Latin and Venetic. The Journal of Indo-European Studies What became of “Sabine l”? 497 Taken as a whole, all this casts many doubts on the traditional reconstruction of the IE stop system, confirms that there was a Proto-Italic language stage and that the Italic languages do not owe their common traits to secondary contact within the Italian Peninsula, and is suggestive of Lusitanian being an Italic language (see fn. 48 above and Prósper forthc. on the Lusitanian divine name LAEPO). As far as Latin is concerned, only the paradigms of dare and Lat. daps count as exceptions, probably because the change was reverted early on. In the first case, only some forms of the paradigm actually contained da- in Italic, specifically the past part. *da-to- in Lat. datus, O. datas and the Lat. agent noun dator ‘giver’ (an Italic formation replacing IE *d-tr, which has st been remodeled on the past part.). By contrast, cf. Latin 1 person singular d, subj. dem, duim, perf. ded, and related nouns like ds, dnum. This paradigm is comparatively well attested in Sabellic, e.g. in the reduplicated present *di-de-ti in Vestinian DIDET, O. fut. DIDEST, the perfect U. dede, O. deded. If the Latin present forms d
s, dat, damus, etc. do not directly continue a root aorist with some modifications in the 51 vocalism, but ultimately go back to de-reduplicated forms (as often assumed on the strength of the Sabellic counterparts and of compounds like reddere) and if, additionally, the infinitive 52 dare is also a secondary form, then only the past participle *da-to- would fail to undergo the change proposed here, or simply, as contended above, is the product of analogy. In the case of Latin daps, damnum, the Anlaut da-, as implied above, may be due to the influence of the early Greek loanwords dapn, dapslis on a form *lap-, *lap-no- regularly 51 As we have seen, Italic may rather have inherited an active aorist *d(e)h3-kwhich gave rise to the present *dak-i-e/o-, whereupon it was redone as a sigmatic form (at least in Latin and Venetic); and the reduplicated perfect was assigned as a preterite to a reduplicated present. This would mean that Ven. doto has been remodeled starting from a middle aorist *dh3-to. Meiser (1998: 188) refutes the possibility that the Latin present ultimately goes back to a reduplicated present on the grounds that the replacement of *di-da- by *di-deis Proto-Italic, like in all polysyllabic forms, but this is only a petitio principii. 52 See Prósper (2018a) for the idea that dare cannot continue an inherited locative case *dh3s-i, since there is no inflection which could possibly have such a locative form. Volume 47, Number 3 & 4, Fall/Winter 2019 498 Blanca María Prósper inherited from Proto-Italic (whose expected phonetic structure is in fact preserved in lapit). However, the fact that there is no Sabellic counterpart for daps in spite of the rich ritual terminology preserved in Oscan and especially in Umbrian leads me to suspect that daps could be a back-formation, and 53 that only *lap-no- is inherited. No credence can be attached to the very indirectly transmitted testimony of Livius Andronicus, and dautia, dacrimas are in all likelihood nonce forms, created for a specific 54 occasion or poem, which never became received forms. It may even be the case that they were only recorded as curiosities by the first grammarians, and later interpreted as archaic in a more general way by successive compilers, who eventually treated them as tokens of the treasured but enigmatic Old Latin language. Consequently, they contribute nothing in modern terms to the chronology and scope of the sound change. The rest of the forms failing to undergo the change are loanwords, and this typically embraces technical terms, like danus ‘fenerator’ (< Greek  ), onomastic items, like the divine name Damia (< Greek
μ), and plant and animal names, like damma ‘roe’ (from Gaulish). Finally, as far as lexical evidence can be trusted to establish genetic relatedness, the considerable amount of Italic forms going back to *deh2- and their evident connections with Greek and Indo-Iranian, as opposed to the virtual absence of this root in Celtic, tip the scales against the existence of an Italo-Celtic stage of Indo-European dialectalization. 53 One could additionally argue that damnum is a Proto-Latin borrowing from Gk.   that gave *dap-na by syncope, was understood as a neuter plural and eventually gave rise to a singular form *dapnom, but this is mere speculation not supported by other evidence 54 And Livius’ prestige was so great in his day that these, even if they were already old-fashioned by his time, would hardly be expected to be the few forms with an initial d- to evolve into the classic forms with l- by a sporadic and unexplained change that, if lacrimis in Ennius can be taken at face value, had been completed by the next generation. The possible objection that he was using archaisms long dead is not defensible, since he could not be expected to handle written sources that old, at least not for a non-specialized, every-day word like dacrimas. As we have seen, the only fragments of Livius and Ennius preserving this form have l-. The Journal of Indo-European Studies What became of “Sabine l”? 499 References Adams, Douglas Q. 1983 Tocharian A 
ku ‘headhair’ and AB yok ‘(body)hair’. Zeitschrift für vergleichende Sprachforschung 96: 167-169. Benedetti, Marina 1988 I composti radicali latini. Esame storico e comparativo. Pisa: Giardini. Bergk, Thomas 1859 Philologische Thesen. Philologus 14: 180–187. Bichlmeier, Harald 2009 Arelape, Tergolape, Interlaken. Ein Beitrag zur keltischen Ortsnamenschicht im Alpenraum. Historische Sprachforschung 122: 254–269. Blänsdorf, Jürgen, Carl Buchner and Willy Morel 2010 Fragmenta Poetarum Latinorum Epicorum et Lyricorum. Berlin New York: Walter de Gruyter. Bottiglioni, Giulio 1943 Di alcuni presunti sabinismi del lessico latino. Studi Etruschi 17: 315-326. 1954 Manuale dei dialetti italici. Osco, umbro e dialetti minori. Grammatica, testi. Bologna: S.T.E.B. Bücheler, Ernst 1891 Altes Latein (Fortsetzung). Rheinisches Museum für Philologie, Neue Folge 46: 233-243. Burman, Annie C. 2018 De Lingua Sabina: A Reappraisal of the Sabine Glosses, doctoral thesis, on line: https://doi.org/10.17863/CAM.18502. CIL = Theodor Mommsen et alii 1862- Corpus Inscriptionum Latinarum. Berlin: De Gruyter. Coleman, Robert 1990 Dialectal variation in Republican Latin, with special reference to Praenestine. The Cambridge Classical Journal 36: 1-25. Conway, Robert S. 1893 On the change of d to l in Italic. Indogermanische Forschungen 2: 157-167. Volume 47, Number 3 & 4, Fall/Winter 2019 500 Blanca María Prósper Crawford, Michael H., Broadhead, William M., Clackson, James P. T., Santangelo, Federico, Thompson, S. and Watmough, Margaret T. (eds.) 2011 Imagines Italicae. A Corpus of Italic Inscriptions. London: Institute of Classical Studies, School of Advanced Study, University of London. DLL = Alfred Ernout and Antoine Meillet 1951 Dictionnaire étymologique de la langue latine. Histoire de mots. Paris: Klincksieck. Dunkel, Georg E. 1997 Latin iubar and f
s: sound-law and analogy. In: Alexander Lubotsky (ed.), Sound Law and Analogy: Papers in Honor of Robert S. P. Beekes on the Occasion of his 60th Birthday, 27-36. Amsterdam and Atlanta. EDBIL = Rick Derksen 2015 Etymological Dictionary of the Baltic Inherited Lexicon. Leiden: Brill. EDG = Robert S. P. Beekes 2010 Etymological Dictionary of Greek. Leiden: Brill. EDLIL = Michiel De Vaan 2008 Etymological Dictionary of Latin and the other Italic Languages. Leiden: Brill. EDPG = Guus Kroonen 2013 Etymological Dictionary of Proto-Germanic. Leiden: Brill. Eichner, Heiner 2 Zu frühlateinischen Wortformen auf dem Forumscippus CIL I , 1 1995 (1. HAUELOD, 2. LOIUQUIOD, 3. KAPIA(D): DOTAUE[RE], 4. Eventuelles Gesamtszenario. In: Michaela Ofitsch and Chr. Zinko (eds.), Studia Onomastica et Indogermanica. Festschrift für Fritz Lottner von Hüttenbach zum 65 Geburtstag, 65-73. Graz. Fränkel, Ernst 1932 Zur tocharischen Grammatik. Indogermanische Forschungen 50: 120. Garnier, Romain 2010 Sur le vocalisme du verbe latin. Étude synchronique et diachronique, Innsbruck, IBS. 2015 Daps, epulum et sollemnis: une famille méconnue en latin. In: Elbieta Maczak-Wohlfeld -
Barbara Podolak (eds.), Words and Dictionaries. A Festschrift for Professor Stanisaw Stachowski on the th Occasion of his 85 Birthday, 127-137. Cracow. The Journal of Indo-European Studies 501 What became of “Sabine l”? Glinister, Fay, Clare Woods, John A. North and Michael H. Crawford (eds.) 2007 Verrius, Festus & Paul. Lexicography, Scholarship and Society. Bulletin of the Institute of Classical Studies, Suppl. 93. London: Institute of Classical Studies at the University of London. Greenberg, Joseph 1970 Some generalizations concerning glottalic consonants, especially implosives. International Journal of American Linguistics 36: 123145. Hackstein, Olav 2002 Uridg. *CH.CC > *C.CC. Historische Sprachforschung 115: 1-22. Hamp, Eric P. 1972 Latin dacrima, lacruma and Indo-European ‘tear’. Glotta 50: 291299. Har
arson, Jon A. 1993 Studien zum urindogermanischen Wurzelaorist und dessen Vertretung im Indoiranischen und Griechischen. Innsbruck: IBS. Havet, Louis 1889 Varia. Memoires de la Societé de Linguistique 6: 230-245. Isaac, Graham R. 2007 The reflexes of the British diphthong *au. Journal of Celtic Linguistics 11: 23-47. Jakob, Anthony 2017 The Metathesis of *-Hu- and *-Hi- in PIE, Master thesis. Leiden, on line: http://hdl.handle.net/1887/52572. Kölligan, Daniel and Macedo, José M. 2015     . Glotta 91: 129-151. Kretschmer, Paul 1896 Einleitung in die Geschichte der griechischen Sprache. Göttingen: Vandenhoeck und Ruprecht. Kümmel, Martin 2019 Addenda und Corrigenda zu LIV
, http://www.martinkuemmel.de/liv2add.html. on line: Lazzarini, M. Letizia and Poccetti, Paolo 2001 Il mondo enotrio tra VI e V secolo a. C. Atti dei seminari napoletani (1996-1998), L’iscrizione paleoitalica da Tortora. Naples: Loffredo. Volume 47, Number 3 & 4, Fall/Winter 2019 502 Blanca María Prósper Lazzeroni, Romano 1992 L’espressione dell’agente come categoria linguistica. I nomi indoeuropei in *-ter/-tor. Studi e saggi linguistici 32: 233-245. Lehmann, Winfred P. 1986-1 Reflexes of PIE d < t’. In: Dieter Kastovsky and Aleksander Szwedek (eds.), Linguistics across Historical and Geographical Boundaries. In Honour of Jacek Fisiak on the Occasion of his Fiftieth Birthday. Volume 1. Linguistic History and Historical Linguistics, 483-489. Berlin - New York. 1986-2 A Gothic Etymological Dictionary. Leiden: Brill. Lejeune, Michel 1974 Manuel de la langue venète Heidelberg: Carl Winter Verlag. Leonard, Clifford S. Jr. 1980 The role of /l/ in rhotacism. In: Herbert J. Izzo (ed.), Italic and Romance. Linguistic Studies in Honor of Ernst Pulgram, 43-53. Amsterdam. Leumann, Manu, Hofmann, Johann B. and Szantyr, Anton 1977 Lateinische Grammatik. I. Laut- und Formenlehre. Munich: Beck. LEW = Alois Walde and Johann B. Hoffmann 1965 Lateinisches Etymologisches Wörterbuch. Heidelberg: Carl Winter Verlag. Lindsay, Wallace M. (ed.) 1903 Nonii Marcelli De Conpendiosa Doctrina I-III. Leipzig: Teubner. 1913 Sexti Pompei Festi De verborum significatu quae supersunt cum Pauli epitome. Leipzig: Teubner. Lipp, Reiner 2009 Die indogermanischen und einzelsprachlichen Palatale im Indoiranischen. II. Thorn-Problem, Indo-Iranische Laryngalvokalisation. Heidelberg: Carl Winter Verlag. Mallory, James P. and Adams, Douglas Q. 1997 Encyclopedia of the Indo-European Culture. Chicago - London: Fitzroy Dearborn. Malzahn, Melanie 2010 The Tocharian Verbal System. Leiden: Brill. Mancini, Marco 2006 Osco aflukad nella defixio Vetter 6. In: Domenico Caiazza (ed.), Samnitice loqui. Scritti in onore di A. L. Prosdocimi, 73-90. Piedimonte Matese. The Journal of Indo-European Studies 503 What became of “Sabine l”? Marotta, Giovanna 1993 Dental stops in Latin: A special class. Rivista di Linguistica 5: 55101. 1999 The Latin syllable. In: Harry Van der Hulst et alii (eds.), The Syllable, 285-311. Berlin - New York. Matasovi
, Ranko 2014 The etymology of Lares. In: Mislav Jei
, Ivan Andrijani
and Kreimir Krnic (eds.), Vita litterarum studiis sacra. Zbornik u ast Radoslavu Katii
u, 247-250. Zagreb. Matisoff, James A. 2013 The dinguist’s dilemma: Regular and sporadic l/d interchange in Sino-Tibetan and elsewhere. In: Tim Thornes et alii (eds.), Functional-Historical Approaches to Explanation: In Honor of Scott DeLancey, 83-104. Amsterdam. Mayrhofer, Manfred 1992 Etymologisches Wörterbuch des Altindoarischen I. Heidelberg: Carl Winter Verlag. Meiser, Gerhard 1986 Lautgeschichte der umbrischen Sprache. Innsbruck: IBS. 1998 Historische Laut- und Formenlehre der lateinischen Sprache. Darmstadt: Wissenschaftliche Buchgesellschaft. Melchert, Craig H. 2003 The Luwians. Leiden: Brill. Murano, Francesca 2010 Verbi e formule di defissione nelle laminette di maledizione osche. Quaderni del Dipartimento di Linguistica - Università di Firenze 20: 51-76. Negri, Mario 1992 La lingua di Numa. In: Bela Brogyányi and Reiner Lipp (eds.), Historical Philology. Greek, Latin and Romance. Papers in Honour of Oswald Szemerényi II, 229-265. Amsterdam-Philadelphia. Neri, Sergio 2007 Cadere e abbattere in lndo-Europeo. Sull’etimologia di tedesco fallen, latino aboleo e greco μ. Innsbruck: IBS. Nikolaev, Alexander S. 2010 Issledovanija po praindoevropejskoj imennoj Petersburg: Russian Academy of Sciences. morfologii. St. Nussbaum, Alan J. 1986 Head and Horn in Indo-European. Berlin - New York: De Gruyter. Volume 47, Number 3 & 4, Fall/Winter 2019 504 2004 Blanca María Prósper A -t- party: Various IE nominal stems in *-(o/e)t-. Talk held at the th 16 Annual UCLA Indo-European Conference, november 2004: 111. Pinault, Georges.-J. 2017 Tocharian tsälp- in Indo-European perspective. In: Bjarne S. Hansen et alii (eds.), Usque ad Radices. Indo-European Studies in Honour of Birgit Anette Olsen, 643-658. Copenhagen. Poucet, Jacques 1966 L’origine sabine de la “commutatio” du -d- en -l-, un mythe linguistique? L’Antiquité Classique 35: 140-148. Poultney, James W. 1955 Two problems in the Iguvine tables. The American Journal of Philology 76: 77-82. Prósper, Blanca M. 2002 Lenguas y religiones prerromanas del occidente de la Península Ibérica. Salamanca: Ediciones Universidad de Salamanca. 2007 Estudio lingüístico del plomo celtibérico de Iniesta. Salamanca: Ediciones Universidad de Salamanca. 2015 Celtic and non-Celtic divinities from ancient Hispania: power, daylight, fertility, water spirits and what they can tell us about Indo-European morphology. Journal of Indo-European Studies 43, 1/2: 1-56. 2016 The Indo-European Names of Central Hispania. A Study in Continental Celtic and Latin Word Formation. Innsbruck: IBS. 2017 Proto-Italic laryngeals in the context ClHC- and new Italic and Celtic etymological connections. Rivista Italiana di Linguistica e Dialettologia 19: 1-24. 2018a The Venetic names of Roman Siscia. Voprosy Onomastiki 15/3: 105-124. 2018b The Venetic inscription from Monte Manicola and three termini publici from Padua: A reappraisal. The Journal of Indo-European Studies 46: 1-61. 2018c The Venetic agent nouns in -t
r- revisited. In: José M. Vallejo, Iván Igartua and Carlos García Castillero (eds.), Studia Philologica et Diachronica in Honorem Joaquín Gorrochategui, Veleia Minor 35: 453-471. forthc. Celtic and Venetic in Contact: The Dialectal Attribution of the Personal Names in the Venetic record. Pulju, Tim 1998 Indo-European *d, *1, and *dl. In: John Charles Smith, Richard M. Hogg, Delia Bentley and Linda van Bergen (eds.), Historical Linguistics 1995. Selected Papers from the 12th International Conference on Historical Linguistics, Manchester, August 1995, Volume 1, 311-326. Amsterdam-Philadelphia. The Journal of Indo-European Studies 505 What became of “Sabine l”? Rieken, Elisabeth NA4 tap(p)a
tai- u.ä. ‘Schlachtblock’ und die uridg. Wurzel 2017 Heth. *deh2p- ‘schlachten, zerfleischen’. In: Ivo Hajnal, Daniel Kölligan and Katharina Zipser (eds.), Miscellanea Indogermanica. Festschrift für José Luis García Ramón zum 65. Geburtstag, 699-704. Innsbruck. Rix, Helmut 1994 Südpikenisch kduíú. Historische Sprachforschung 107: 105-122. 1998 Eine neue frühsabellische Inschrift und der altitalische Präventiv. Historische Sprachforschung 111: 247-269. Rix, Helmut et alii 2 Lexicon der indogermanischen Verben. Die Wurzeln und ihre 2001 Primärstammbildungen. Wiesbaden: Reichert. Rohlfs, Gerhard 1966 Grammatica storica della lingua italiana e dei suoi dialetti, I. Fonetica. Torino: Einaudi. Schindler, Jochem 1975 Zum Ablaut der neutralen s-Stämme des Indogermanischen. In: Helmut Rix (ed.), Flexion und Wortbildung. Akten der V. Fachtagung der Indogermanischen Gesellschaft, Regensburg, 259267. Wiesbaden. Schrijnen, Jos 1914 Das “sabinische l” im Lateinischen. Zeitschrift für vergleichende Sprachforschung 46: 376-380. Schrijver, Peter 1991 The Reflexes of the Proto–Indo–European Laryngeals in Latin. Amsterdam: Rodopi. Solmsen, Felix 1909 Odysseus und Penelope. Sprachforschung 42: 207-233. Zeitschrift für vergleichende ST = Helmut Rix 2002 Sabellische Texte. Die Texte des Oskischen, Umbrischen und Südpikenischen. Heidelberg: Carl Winter Verlag. Stokes, Whitley 1904 Hibernica (Fortsetzung). Sprachforschung 37: 250-361. Zeitschrift für vergleichende Stuart-Smith, Jane 2004 Phonetics and Philology. Sound Change in Italic. Oxford: Oxford University Press. Volume 47, Number 3 & 4, Fall/Winter 2019 506 Blanca María Prósper Tichy, Eva 1995 Die Nomina agentis auf -tar- im Vedischen. Heidelberg: Carl Winter Verlag. Untermann, Jürgen 1993 Gr.  = Lat. feci, Gr.  = Lat. ieci. In: Gerhard Meiser (ed.), Indogermanica et Italica. Festschrift für Helmut Rix, 461-468. Innsbruck Vij
nas, Aurelijus 2009 The Indo-European Primary t-Stems. Innsbruck: IBS. Vine, Brent 2006 On Thurneysen-Havet’s Law in Latin and Italic. Historische Sprachforschung 119: 211-249. Wallace, Rex 1988 Dialectal Latin fundatid, proiecitad, parentatid. Glotta 66: 211-220. 2005 A Faliscan inscription in the Michael and Judy Steinhardt collection. Zeitschrift für Papyrologie und Epigraphik 153: 175-182. Watkins, Calvert 1976 The etymology of Irish dúan. Celtica 11: 270-277. Weiss, Michael 2009 An Outline of the Historical and Comparative Grammar of Latin. Ann Arbor - New York: Beech Stave Press. 2010 Language and Ritual in Sabellic Italy: The Ritual Complex of the Third and Fourth Tabulae Iguvinae. Leiden: Brill. West, Martin L. 2007 Indo-European Poetry and Myth. Oxford: Oxford University Press. Winter, Werner 1962 Die Vertretung der indogermanischen Dentale im Tocharischen. Indogermanische Forschungen 67: 16-35. The Journal of Indo-European Studies